Sensitivity Optimization for NV-Diamond Magnetometry

John F. Barry,1, 2, 3, 4, ∗ ,† Jennifer M. Schloss,1, 2, 4, 5, ∗ ,‡ Erik Bauch,3, ∗ Matthew J. Turner,3, 4 Connor A. Hart,3 Linh
M. Pham,1, 2 and Ronald L. Walsworth2, 3, 4, 6, 7, 8, §
1
Lincoln Laboratory,
Massachusetts Institute of Technology,
Lexington, Massachusetts 02421,
USA
2
Harvard-Smithsonian Center for Astrophysics,
Cambridge, Massachusetts 02138,
USA
3
Department of Physics,
Harvard University, Cambridge,
Massachusetts 02138,
USA
arXiv:1903.08176v2 [quant-ph] 28 May 2020

4
Center for Brain Science,
Harvard University, Cambridge,
Massachusetts 02138,
USA
5
Department of Physics,
Massachusetts Institute of Technology,
Cambridge, Massachusetts 02139,
USA
6
Quantum Technology Center,
University of Maryland,
College Park, Maryland 20742,
USA
7
Department of Electrical and Computer Engineering,
University of Maryland,
College Park, Maryland 20742,
USA
8
Department of Physics,
University of Maryland,
College Park, Maryland 20742,
USA
(Dated: May 29, 2020)

Solid-state spin systems including nitrogen-vacancy (NV) centers in diamond consti-

tute an increasingly favored quantum sensing platform. However, present NV ensemble
devices exhibit sensitivities orders of magnitude away from theoretical limits. The sen-
sitivity shortfall both handicaps existing implementations and curtails the envisioned
application space. This review analyzes present and proposed approaches to enhance
the sensitivity of broadband ensemble-NV-diamond magnetometers. Improvements to
the spin dephasing time, the readout fidelity, and the host diamond material properties
are identified as the most promising avenues and are investigated extensively. Our anal-
ysis of sensitivity optimization establishes a foundation to stimulate development of new
techniques for enhancing solid-state sensor performance.

CONTENTS B. Alternatives to Ramsey magnetometry 9

1. CW-ODMR 9
I. Introduction and background 2 2. Pulsed ODMR 10
A. NV-diamond magnetometry overview 2 C. Parameters limiting sensitivity 14
B. Magnetometry introduction 4
C. The NV- ground state spin 4 III. Limits to relaxation times T2∗ and T2 15
D. Spin-based measurements on NV- 5 A. Motivation to extend T2∗ 15
E. Spin dephasing and decoherence 6 B. Ensemble and single-spin T2∗ 16
F. DC and AC sensing 7 C. Dephasing mechanisms 17
D. Nitrogen limit to T2∗ 18
II. Measurement sensitivity considerations 7 E. Nitrogen limit to T2 19
A. Magnetic field sensitivity 7 F. 13 C limit to T2∗ 20
G. NV- limit to T2∗ 21

IV. Methods to extend T2∗ and T2 22

A. Dynamical decoupling for AC magnetometry 22
∗ These authors contributed equally B. Double-quantum coherence magnetometry 24
† [email protected] C. Spin bath driving 26
‡ [email protected] D. Transverse strain and electric field mitigation 27
§ [email protected]
 2

V. Methods to increase readout fidelity 28 Maze et al., 2008) and NV ensembles (Acosta et al.,
A. Spin-to-charge conversion readout 28 2009) circa 2008. In the decade following, both single-
B. Photoelectric readout 30
C. Ancilla-assisted repetitive readout 31
and ensemble-NV-diamond magnetometers (Doherty et al.,
D. Level-anticrossing-assisted readout 32 2013; Rondin et al., 2014) have found use for applications
E. Improved photon collection methods 33 in condensed matter physics (Casola et al., 2018), neuro-
F. Near-infrared absorption readout 34 science and living systems biology (Schirhagl et al., 2014;
G. Green absorption readout 36
H. Laser threshold magnetometry 36 Wu et al., 2016), nuclear magnetic resonance (NMR) (Wu
et al., 2016), Earth and planetary science (Glenn et al.,
VI. Diamond material engineering 37 2017), and industrial vector magnetometry (Grosz et al.,
A. Conversion efficiency 37
2017).
B. NV charge state efficiency ❈❤❛♣t❡r ✶✿
37 ◆❱ ❇❛❝❦❣r♦✉♥❞
1. Non-optical effects on NV charge state efficiency 38
2. Optical effects on NV charge state efficiency 38 a b 6&& ' ,2&.45
8:
C. Diamond synthesis and high pressure high temperature . .

treatment 39
;7
D. Electron irradiation 41 . ;

E. Low pressure high temperature annealing 42 23*.45

F. Other common impurities in synthetic or treated single
crystal diamond 43 V !" # $%
G. Preferential orientation 45
;:
. . !" # '%
N
VII. Miscellaneous sensing techniques 45
A. Rotary echo magnetometry 45 87
. 9 ( ( ) *+,-./01
B. Geometric phase magnetometry 45
!" # &
C. Ancilla-assisted upconversion magnetometry 45 ¯¯
NV || [111]
D. Techniques for the strong NV- -NV- interaction regime 46
FIG. 1 Overview of the nitrogen-vacancy (NV) center quantum
VIII. Conclusion and outlook 48
system. a) Diagram of diamond lattice containing an NV cen-
Acknowledgments 49
ter, which consists of a substitutional nitrogen adjacent to a
A. 49 lattice vacancy. The green arrow marks the NV symmetry axis,
1. Derivations 49 oriented along the [11̄1̄] diamond crystallographic axis for the
a. Ramsey DC magnetic field measurement 49 particular NV center shown here. From Ref. (Pham, 2013). b)
b. Spin-projection-noise-limited sensitivity 51 Energy level diagram for the negatively charged NV- center in
c. Photon-shot-noise-limited sensitivity 52 diamond, with zero-field splitting D between the ground-state
d. Overhead time 53 electronic spin levels ms = 0 and ms = ±1. The ms = ±1 energy
2. Optimal precession time 53
levels experience a Zeeman shift in the presence of a magnetic
3. Considerations for increasing sensor number 53 ~ which forms the basis for NV- magnetometry. Adapted
❋✐❣✉r❡ ✶✳✶✿ ❋♦✉r ♦r✐❡♥t❛t✐♦♥s ♦❢ t❤❡ ◆❱ ❝❡♥t❡r ✐♥ ❞✐❛♠♦♥❞✳ ❈❛r❜♦♥ ❛t♦♠s ❛r❡
4. Choosing nitrogen concentration in diamond samples 54
field B,
from Ref. (Schloss et al., 2018).
❞❡♣✐❝t❡❞ ✐♥ ❜❧❛❝❦✱ ♥✐tr♦❣❡♥ ✭◆✮ ❛t♦♠s ✐♥ ❜❧✉❡✱ ❛♥❞ ✈❛❝❛♥❝✐❡s ✭❱✮ ✐♥ ✇❤✐t❡✳
5. Spin resonance linewidth and T2∗ 55
6. Estimating T2∗ from spin resonance linewidths of N0S 55
❚❤❡ ◆❱ ❡❧❡❝tr♦♥✐❝ s♣✐♥ ✐s ✐♥❞✐❝❛t❡❞ ❜② ❣r❡❡♥
7. Stretched exponential parameter 56
❛rr♦✇s✳defects
Solid-state ❋♦✉rsuch❛❞❞✐t✐♦♥❛❧
as NV centers♦r✐✲ exhibit quantum
❡♥t❛t✐♦♥s ❛r❡ ♣♦ss✐❜❧❡ ❜② ✢✐♣♣✐♥❣ t❤❡ ♥✐tr♦❣❡♥
8. Isotopic purity confusion in the literature 57
properties ❛t♦♠s
similar❛♥❞ ✈❛❝❛♥❝✐❡s
to traditional ✐♥ ❡❛❝❤
atomic systems yet confer
❝♦♥✜❣✉r❛t✐♦♥ ❛❜♦✈❡❀ ❤♦✇❡✈❡r✱ ♦r✐❡♥t❛t✐♦♥s ✇✐t❤point
❡q✉✐✈❛❧❡♥t s②♠♠❡tr② ❛①❡s
9. Linear Stark and Zeeman regimes 57
10. Example annealing calculations 58 technical and logistical advantages for sensing applications.
❛r❡ s♣❡❝tr❛❧❧② ✐♥❞✐st✐♥❣✉✐s❤❛❜❧❡ ❛♥❞ ♠❛②gent❤❡r❡❢♦r❡
11. The diamond type classification system 59 NVs are ❜❡ ❝♦♥s✐❞❡r❡❞ ✐♥ t❤❡ s❛♠❡
defects composed of a substitutional nitro-
References ◆❱ ♦r✐❡♥t❛t✐♦♥ ❝❧❛ss✳ 64
fixed adjacent to a vacancy within the rigid carbon lat-
tice (see Fig. 1a). Each NV center’s symmetry axis is con-

▼♦r❡ r❡❝❡♥t❧②✱ ❝❤❡♠✐❝❛❧ ✈❛♣♦r ❞❡♣♦s✐t✐♦♥ ✭❈❱❉✮ While ❤❛s ❡♠❡r❣❡❞

NVs are❛sobserved
❛♥♦t❤❡rto✈✐❛❜❧❡
strained to lie along one of the four [111] crystallographic
directions. exist in three charge
♠❡t❤♦❞ ❢♦r ❞✐❛♠♦♥❞ ♣r♦❞✉❝t✐♦♥✳ ■♥ ❈❱❉ s②♥t❤❡s✐s✱
states (NVs♦✉r❝❡ , NV ❣❛ss❡s ❝♦♥t❛✐♥✐♥❣ ❝❛r❜♦♥
I. INTRODUCTION AND BACKGROUND
- 0 + -
and NV ), the negatively charged NV
❛r❡ ✐♥tr♦❞✉❝❡❞ ✐♥t♦ ❛ ❣r♦✇t❤ ❝❤❛♠❜❡r ✇❤❡r❡
A. NV-diamond magnetometry overview centert❤❡②applications
is ❛r❡ ❡♥❡r❣✐③❡❞
favored for t♦ ❢♦r♠
quantum ❛ ♣❧❛s♠❛
sensing and quantum infor-
t❤❛t ❞❡♣♦s✐ts ❝❛r❜♦♥ ❛t♦♠s ♦♥t♦ ❛ s✉❜str❛t❡✳ mation❈♦♠♣❛r❡❞ t♦ ❍P❍❚ s②♥t❤❡s✐s✱ ❈❱❉
(Doherty et al., 2013). The NV -
defect
❣r♦✇t❤ ♦❝❝✉rs ❛t ❧♦✇❡r ♣r❡ss✉r❡s
Quantum sensors encompass a diverse class of devices
✭ ❜❛r✮ ❛♥❞ t❡♠♣❡r❛t✉r❡s ✭ ❈✮✱ ✇❤❡r❡
exhibits a spin-1 triplet electronic ground state with long

signals. As theirt❤❡ t❤❡r♠♦❞②♥❛♠✐❝❛❧❧② st❛❜❧❡constants,

❛❧❧♦tr♦♣❡ ♦❢ times
❝❛r❜♦♥T ✐s≈❣r❛♣❤✐t❡ r❛t❤❡r t❤❛♥ ❞✐❛♠♦♥❞✳
that exploit quantum coherence to detect weak or nanoscale spin lifetimes at room temperature; longitudinal relaxation
❍♦✇❡✈❡r✱
behavior
❜② ✐♥tr♦❞✉❝✐♥❣ ❤②❞r♦❣❡♥
is tied to physical
❣❛s ✭❍ ✮ ✐♥ ❛❞❞✐t✐♦♥
1
t♦ ❝❛r❜♦♥✲♣r♦✈✐❞✐♥❣ ♠❡t❤❛♥❡
6 ms (Jarmola et al., 2012; Rosskopf et al.,

❣❛s ✭❈❍ ✮✱ ❛♥② ❣r❛♣❤✐t❡ ❞❡♣♦s✐ts

quantum devices can achieve accuracy, repeatability, and
❛r❡ r❛♣✐❞❧② are achievable (Balasubramanian et ✐♥
❛♥❞ s❡❧❡❝t✐✈❡❧② ❡t❝❤❡❞✱ r❡s✉❧t✐♥❣
2014) are typical, and coherence times
al.,❛ 2009).
♥❡t
T up to a few ms
2

malis, 2007). As a❣r♦✇t❤ ♦❢ ❞✐❛♠♦♥❞ ❡✈❡♥shown

✉♥❞❡rutility
t❤❡ ❞❡s❝r✐❜❡❞ spin♠❡t❛st❛❜❧❡ ❝♦♥❞✐t✐♦♥s✳ ❋✉rt❤❡r♠♦r❡✱
precision approaching fundamental limits (Budker and Ro- The defect’s

❈❱❉ s②♥t❤❡s✐s ❤❛sboth

❛ ♥✉♠❜❡r ♦❢ applied
❛❞✈❛♥t❛❣❡sfields,♦✈❡r ❍P❍❚ s②♥t❤❡s✐s❀ ❞✐❛♠♦♥❞s ❝❛♥ ❜❡(Doherty et al.,
result, these sensors have in energy levels are sensitive to magnetic fields, electric
a wide range of applications spanning pure and
science (Degen et❣r♦✇♥ ♦✈❡r A❧❛r❣❡ ❛r❡❛semerging
❛♥❞ ♦♥ ❞✐✛❡r❡♥t
quantums✉❜str❛t❡2013), ♠❛t❡r✐❛❧s✳ ❆❧s♦✱tot❤❡operate
✐♥tr♦❞✉❝t✐♦♥ ♦❢
strain, and temperature variations
al., 2017). rapidly -
allowing NV as a multi-modal sensor.
sensing platform ❝❤❡♠✐❝❛❧ ✐♠♣✉r✐t✐❡s
employs atomic-scale ✐♥t♦ t❤❡
defects ❣r♦✇t❤ ❝❤❛♠❜❡r
in crystals. ❝❛♥
Coherent ❜❡ ✜♥❡❧② ❝♦♥tr♦❧❧❡❞
spin control ❛♥❞
is achieved s♦ t❤❡r❡❢♦r❡
by application of resonant
❝❛♥ t❤❡ has
In particular, magnetometry ♣r♦♣❡rt✐❡s ♦❢increasing
t❤❡ ❞✐❛♠♦♥❞
using nitrogen ♣r♦❞✉❝❡❞✳microwaves
vacancy (NV) ❚❤❡s❡ ❛❞✈❛♥t❛❣❡s
(MWs) near ♠❛❦❡2.87❈❱❉
GHz.t❤❡ Upon ♠❛✐♥optical excitation,
s②♥t❤❡s✐s t❡❝❤♥✐q✉❡ ❢♦r ❞✐❛♠♦♥❞s ✉s❡❞ ✐♥ r❡s❡❛r❝❤
color centers in diamond garnered interest. ❛♣♣❧✐❝❛t✐♦♥s✳
nonradiative decay through a spin-state-dependent inter-

❖♥❡Taylor
♦❢ t❤❡et ♠♦st ❝♦♠♠♦♥ ✐♠♣✉r✐t✐❡s spin-state-dependent
❢♦✉♥❞ ✐♥ ❞✐❛♠♦♥❞✖❜♦t❤ s②♥t❤❡t✐❝ ❛♥❞and optical spin
The use of NV centers as magnetic field sensors was first system crossing (Goldman et al., 2015a,b) produces both
proposed (Degen, 2008; al., 2008) and demon- fluorescence contrast
strated with single NVs (Balasubramanian et al., 2008; -
initialization into the NV center’s m = 0 ground state
✸
s
 3

(see Fig. 1b). tion (Davis et al., 2018); nanoscale thermometry (Kucsko
Relative to alternative technologies (Grosz et al., 2017), et al., 2013; Neumann et al., 2013); single protein detec-
sensors employing NV- centers excel in technical simplic- tion (Lovchinsky et al., 2016; Shi et al., 2015); nanoscale
ity and spatial resolution (Arai et al., 2015; Grinolds and micron-scale NMR (Bucher et al., 2018; DeVience et al.,
et al., 2014; Jaskula et al., 2017). Such devices may op- 2015; Glenn et al., 2018; Kehayias et al., 2017; Loretz et al.,
erate as broadband sensors, with bandwidths up to ∼ 2014; Rugar et al., 2015; Staudacher et al., 2013; Sushkov
100 kHz (Acosta et al., 2010b; Barry et al., 2016; Schloss et al., 2014); and studies of meteorite composition (Fu et al.,
et al., 2018), or as high frequency detectors for signals up 2014) and paleomagnetism (Farchi et al., 2017; Glenn et al.,
to ∼ GHz (Aslam et al., 2017; Boss et al., 2016, 2017; Cai 2017).
et al., 2013; Casola et al., 2018; Hall et al., 2016; Horsley Despite demonstrated utility in a number of applica-
et al., 2018; Loretz et al., 2013; Lovchinsky et al., 2016; Pel- tions, the present performance of ensemble-NV- sensors re-
liccione et al., 2014; Pham et al., 2016; Schmitt et al., 2017; mains far from theoretical limits. Even the most sensi-
Shao et al., 2016; Shin et al., 2012; Steinert et al., 2013; tive ensemble-based devices demonstrated to date exhibit
Tetienne et al., 2013; Wood et al., 2016). Importantly, ef- readout fidelities F ∼ 0.01, limiting sensitivity to at best
fective optical initialization and readout of NV- spins does ∼ 100× worse than the spin projection limit. Addition-
not require narrow-linewidth lasers; rather, a single free- ally, reported dephasing times T2∗ in NV-rich diamonds re-
running 532 nm solid-state laser is sufficient. NV-diamond main 100 to 1000× shorter than the theoretical maximum
sensors operate at ambient temperatures, pressures, and of 2T1 (Bauch et al., 2018; Bauch et al., 2019; Jarmola
magnetic fields, and thus require no cryogenics, vacuum et al., 2012). As a result, whereas present√state-of-the-art
systems, or tesla-scale applied bias fields. Furthermore, di- ensemble-NV- magnetometers exhibit pT/ Hz-level sensi-
amond is chemically inert, making NV- devices biocompat- tivities, competing technologies such as superconducting
ible. These properties allow sensors to be placed within quantum interference devices (SQUIDs) and spin-exchange
∼ 1 nm of field sources (Pham et al., 2016), which enables relaxation-free
√ (SERF) magnetometers exhibit sensitivities
magnetic field imaging with nanometer-scale spatial resolu- at the fT/ Hz-level and below (Kitching, 2018). This
tion (Arai et al., 2015; Grinolds et al., 2014; Jaskula et al., ∼ 1000× sensitivity discrepancy corresponds to a ∼ 106 ×
2017). NV-diamond sensors are also operationally robust increase in required averaging time, which precludes many
and may function at pressures up to 60 GPa (Doherty et al., envisioned applications. In particular, the sensing times re-
2014; Hsieh et al., 2018; Ivády et al., 2014) and tempera- quired to detect weak static signals with an NV-diamond
tures from cryogenic to 600 K (Plakhotnik et al., 2014; Toyli sensor may be unacceptably long; e.g., biological systems
et al., 2013, 2012). may have only a short period of viability. In addition, many
Although single NV- centers find numerous applications applications, such as spontaneous event detection and time-
in ultra-high-resolution sensing due to their angstrom-scale resolved sensing of dynamic processes (Marblestone et al.,
size (Balasubramanian et al., 2008; Casola et al., 2018; 2013; Shao et al., 2016), are incompatible with signal av-
Maze et al., 2008), sensors employing ensembles of NV- eraging. Realizing NV-diamond magnetometers with im-
centers provide improved signal-to-noise ratio (SNR) at the proved sensitivity could enable a new class of scientific and
cost of spatial resolution by virtue of statistical averaging industrial applications poorly matched to bulkier SQUID
over multiple spins (Acosta et al., 2009; Taylor et al., 2008). and vapor-cell technologies. Examples include noninva-
Diamonds may be engineered to contain concentrations of sive, real-time magnetic imaging of neuronal circuit dy-
NV- centers as high as 1019 cm-3 (Choi et al., 2017a), namics (Barry et al., 2016), high throughput nanoscale
which facilitates high-sensitivity measurements from single- and micron-scale NMR spectroscopy (Bucher et al., 2018;
channel bulk detectors as well as wide-field parallel mag- Glenn et al., 2018; Smits et al., 2019), nuclear quadrupole
netic imaging (Davis et al., 2018; Fescenko et al., 2018; resonance (NQR) (Lovchinsky et al., 2017), human mag-
Glenn et al., 2015; Le Sage et al., 2013; Pham et al., 2011; netoencephalography (Hämäläinen et al., 1993), subcellu-
Steinert et al., 2010, 2013; Taylor et al., 2008). These en- lar magnetic resonance imaging (MRI) of dynamic pro-
gineered diamonds typically contain NV- centers with sym- cesses (Davis et al., 2018), precision metrology, tests of fun-
metry axes distributed along all four crystallographic orien- damental physics (Rajendran et al., 2017), and simulation
tations, each primarily sensitive to the magnetic field pro- of exotic particles (Kirschner et al., 2018).
jection along its axis; thus, ensemble-NV- devices provide This review accordingly focuses on understanding present
full vector magnetic field sensing without heading errors or sensitivity limitations for ensemble-NV- magnetometers to
dead zones (Le Sage et al., 2013; Maertz et al., 2010; Pham guide future research efforts. We survey and analyze meth-
et al., 2011; Schloss et al., 2018; Steinert et al., 2010). NV- ods for optimizing magnetic field sensitivity, which we di-
centers have also been employed for high-sensitivity imag- vide into three broad categories: (i) improving spin dephas-
ing of temperature (Kucsko et al., 2013), strain, and electric ing and coherence times; (ii) improving readout fidelity;
fields (Barson et al., 2017; Dolde et al., 2011). Recent exam- and (iii) improving quality and consistency of host diamond
ples of ensemble-NV- sensing applications include magnetic material properties. Given the square-root improvement
detection of single-neuron action potentials (Barry et al., of sensitivity with number of interrogated spins, we pri-
2016); magnetic imaging of living cells (Le Sage et al., marily concentrate on ensemble-based devices with & 104
2013; Steinert et al., 2013), malarial hemozoin (Fescenko NV- centers (Acosta et al., 2009; Barry et al., 2016; Cleven-
et al., 2018), and biological tissue with subcellular resolu- son et al., 2015; Le Sage et al., 2012; Wolf et al., 2015b).
 4

However, we also examine single-NV- magnetometry tech- the external magnetic field. For example, the voltage in-
niques in order to determine their applicability to ensem- duced across a pickup coil varies with applied AC magnetic
bles. Moreover, while this work primarily treats broadband, field, as does the resistance of a giant magnetoresistance
time-domain magnetometry from DC up to ∼ 100 kHz, sensor. In atomic systems such as gaseous alkali atoms,
narrowband AC sensing techniques are also analyzed when the Zeeman interaction causes the electronic-ground-state
considered relevant to future DC and broadband magne- energy levels to shift with magnetic field. Certain color
tometry advances. Alternative phase-insensitive AC mag- centers including NV- in diamond also exhibit magneti-
netometry techniques, such as T1 relaxometry (Ariyaratne cally sensitive energy levels. For both NV- centers and
et al., 2018; Casola et al., 2018; Hall et al., 2016; Pelliccione gaseous alkali atoms, magnetometry reduces to measur-
et al., 2014; Romach et al., 2015; Shao et al., 2016; Tetienne ing transition frequencies between energy levels that dis-
et al., 2013; van der Sar et al., 2015), are not discussed. play a difference in response to magnetic fields. Various
This document is organized as follows: the remainder approaches allow direct determination of a transition fre-
of Sec. I provides introductory material on NV- magne- quency; for example, frequency-tunable electromagnetic ra-
tometry, with Sec. I.B introducing magnetic field sens- diation may be applied to the system, and the transition
ing, Sec. I.C presenting the NV- spin Hamiltonian and frequency localized from recorded absorption, dispersion,
its magnetic-field-dependent transitions, Sec. I.D describ- or fluorescence features. Transition frequencies may also
ing quantum measurements using the NV- spin, Sec. I.E be measured via interferometric techniques, which record
outlining how spin dephasing and decoherence limit mag- a transition-frequency-dependent phase (Rabi, 1937; Ram-
netometry, and Sec. I.F summarizing differences between sey, 1950).
DC and AC sensing approaches while focusing subsequent
discussion on DC sensing. Section II concentrates on mag-
netic field sensitivity, with Sec. II.A introducing the math- C. The NV- ground state spin
ematical formalism governing sensitivity of Ramsey-based
ensemble-NV- magnetometers, Sec. II.B reviewing common The NV- center’s electronic ground state Hamiltonian
alternatives to Ramsey protocols for DC magnetometry, can be expressed as
and Sec. II.C overviewing key parameters that determine
H = H0 + Hnuclear + Helec|str , (1)
magnetic field sensitivity. Section III examines the NV-
spin ensemble dephasing time, T2∗ , and coherence time, where H0 encompasses the NV- electron spin interaction
T2 . In particular, Sec. III.A motivates efforts to extend with external magnetic field B ~ and zero-field-splitting pa-
T2∗ , Sec. III.B highlights relevant definitional differences of rameter D ≈ 2.87 GHz, which results from an electronic
T2∗ for ensembles and single spins, Sec. III.C characterizes spin-spin interaction within the NV- ; Hnuclear character-
various mechanisms contributing to NV- ensemble T2∗ , and izes interactions arising from the nitrogen’s nuclear spin;
Secs. III.D-III.G investigate limits to T2∗ and T2 from dipo- and Helec|str describes the electron spin interaction with
lar interactions with specific paramagnetic species within electric fields and crystal strain. Defining z to be along the
the diamond. Section IV analyzes methods to extend the NV- internuclear axis, H0 may be expressed as
NV- ensemble dephasing and coherence times using DC and
radiofrequency (RF) magnetic fields. Section V analyzes a ge µB  ~ ~ 
H0 /h = DSz2 + B·S , (2)
variety of techniques demonstrated to improve the NV- en- h
semble readout fidelity. Section VI reviews progress in en-
where ge ≈ 2.003 is the NV electronic g-factor, µB is
gineering diamond samples for high-sensitivity magnetome- ~ =
the Bohr magneton, h is Planck’s constant, and S
try, primarily focusing on increasing the NV- concentration
(Sx , Sy , Sz ) is the dimensionless electronic spin-1 operator.
while maintaining long T2∗ times and good readout fidelity.
H0 is the simplest Hamiltonian sufficient to model basic
Section VII analyzes several additional NV-diamond mag-
NV- spin behavior in the presence of a magnetic field.
netometry techniques not covered in previous sections. Sec-
The NV- center’s nitrogen nuclear spin (I = 1 for 14 N and
tion VIII provides concluding remarks and an outlook on
I = 1/2 for 15 N) creates additional coupling terms charac-
areas where further study is needed. We note that this
terized by
document aims to comprehensively cover relevant results
reported through mid-2017 and provides limited coverage Hnuclear /h = Ak Sz Iz + A⊥ (Sx Ix + Sy Iy )
of results published thereafter.
+ P Iz2 − I(I + 1)/3


(3)
gI µN  ~ ~
− B·I ,
B. Magnetometry introduction h

Magnetometry is the measurement of a magnetic field’s where Ak and A⊥ are (respectively) the axial and transverse
magnitude, direction, or projection onto a particular axis. magnetic hyperfine coupling coefficients, P is the nuclear
A simple magnetically-sensitive device is a compass nee- electric quadrupole parameter, gI is the nuclear g-factor
dle, which aligns along the planar projection of the ambi- for the relevant nitrogen isotope, µN is the nuclear magne-
ent magnetic field. Regardless of sophistication, all magne- ton, and I~ = (Ix , Iy , Iz ) is the dimensionless nuclear spin
tometers exhibit one or more parameters dependent upon operator. Experimental values of Ak , A⊥ , and P are re-
ported in Table 16. Note that the term proportional to P
 5

most magnetic sensing applications, H0 can be taken as the

Hamiltonian describing the NV- ground state spin for each
of the hyperfine states.
In the presence of a magnetic field oriented along the NV
~ = (0, 0, Bz ), H0 is given in matrix form
internuclear axis B
by

ge µB
 
D+ h Bz 0 0
(z)
H0 /h =  0 0 0 . (5)
 
ge µB
0 0 D − h Bz

with eigenstates |ms = 0i, |ms = −1i, and |ms = +1i and
magnetic-field-dependent transition frequencies
FIG. 2 Energy level diagram for the NV- ground state spin in
the presence of an axial magnetic field Bz and ignoring nuclear ge µB
ν± = D ± Bz , (6)
spin, as described by Eqn. 5. Population in the |ms = 0i state h
results in higher fluorescence under optical illumination than
population in the |ms = ±1i states. In this diagram, resonant which are depicted in Fig. 2. For the general case of a
MWs (gray oval) address the |ms = 0i → |ms = +1i transition. ~ with both axial and transverse components
magnetic field B
Eqn. 8 describes the pseudo-spin-1/2 subspace occupied by these Bz and B⊥ , the
transition frequencies are given to third
two levels. order in gehµB D
B
by
"    2
g e µB B 3 ge µB B
vanishes for I = 1/2 15 -
in NV , as no quadrupolar moment ν± = D 1 ± cos θB + sin2 θB
h D 2 h D
exists for spins I < 1. # (7)
~
The NV- electron spin also interacts with electric fields E

ge µB B
3 
1 3 1 2
± sin θB tan θB − sin θB cos θB ,
and crystal stress (with associated strain) (Kehayias et al., h D 8 2
2019). In terms of the axial dipole moment dk , trans-
verse dipole moments d⊥ and d0⊥ , and spin-strain coupling where tan θB = B⊥ /Bz .
parameters {Mz , Mx , My , Nx , Ny }, the interaction is Magnetic sensing experiments utilizing NV- centers of-
presently best approximated by (Barfuss et al., 2018; Do- ten interrogate one of these two transitions, allowing the
herty et al., 2012; Udvarhelyi et al., 2018; Van Oort and unaddressed state to be neglected. For example, choosing
Glasbeek, 1990) the |0i and |+1i states and subtracting a common energy
(z)
offset allows H0 from Eqn. 5 to be reduced to the spin-1/2
Helec|str /h = dk Ez + Mz Sz2


Hamiltonian H given by
+ (d⊥ Ex + Mx ) Sy2 − Sx2


!
D
+ 21 gehµB Bz 0
+ (d⊥ Ey + My ) (Sx Sy + Sy Sx ) (4) H/h = 2
1 ge µB
. (8)
0 −D
2 − 2 h Bz
+ (d0⊥ Ex + Nx ) (Sx Sz + Sz Sx )
+ (d0⊥ Ey + Ny ) (Sy Sz + Sz Sy ) . This simplification is appropriate when off-resonant excita-
tion of the |ms = −1i state can be ignored and operations
Experimental values of d⊥ and dk are given in Table 16. on the spin system are short compared to T1 . From this
In magnetometry measurements, the terms proportional to simple picture, the full machinery typically employed for
(d0⊥ Ei +Ni ) for i = x, y are typically ignored, as they are two-level systems can be leveraged.
off-diagonal in the Sz basis, and the energy level shifts they
produce are thus suppressed by D (Kehayias et al., 2019).
Furthermore, many magnetometry implementations oper- D. Spin-based measurements on NV-
ate with an applied bias field B~0 satisfying d⊥ Ei +Mi 
ge µB
h B0  D for i = x, y in order to operate in the lin-
We now outline Norman Ramsey’s Method of separated
ear Zeeman regime, where the energy levels are maximally oscillatory fields when adapted for magnetic field measure-
sensitive to magnetic field changes (see Appendix A.9). In ment using one or more NV- centers in a two-level sub-
the linear Zeeman regime, the terms in Helec|str propor- space, e.g., {|0i,|+1i}. After initialization of the spin state
tional to (d⊥ Ei +Mi ) can also be ignored. The sole re- to |0i, a periodically varying magnetic field B1 (t), with po-
maining term in Helec|str acts on the NV- spin in the same larization in the x-y plane and frequency ν+ resonant with
way as the temperature-dependent D and is often combined the |0i ↔ |+1i transition, causes spin population to oscil-
into the parameter D for a given NV- orientation (Glenn late between the |0i and |+1i states at angular frequency
et al., 2017). Except for extreme cases such as sensing in ΩR ∝ B1 , called the Rabi frequency. The resonant field
highly strained diamonds or in the presence of large electric B1 (t) is applied for a particular finite duration π/(2ΩR )
fields, the values of all the electric field and strain param- known as a π/2-pulse, which transforms the initial state
eters in Helec|str are ∼ 1 MHz or lower. Consequently, for |0i into an equal superposition of |0i and |+1i. This state
 6

Initial
/2-pulse Free precession inverval  Final
/2-pulse Spin state readout

2 2 2 2

2 2 2 2

FIG. 3 Bloch sphere depiction of Ramsey sequence. After initialization to the spin state |↑i, a sinusoidally-varying magnetic field
rotates the state vector by π/2, thus preparing a superposition of |↑i and |↓i spin states. Next, the Bloch vector undergoes free
precession for duration τ , accumulating a phase φ proportional to the static magnetic field being sensed. After time τ , a second
π/2-pulse maps the accumulated phase onto a population difference between the |↑i and |↓i states. Here, a φ = π phase accumulation
is shown, which maps back to the state |↑i. Finally, a projective spin state measurement detects the population difference, allowing
determination of the static magnetic field sensed by the spin.

is then left to precess unperturbed for duration τ , during time of a free induction decay (FID) measurement, wherein
which a magnetic-field-dependent phase φ accumulates be- a series of Ramsey sequences are performed with varying
tween the two states. Next, a second π/2-pulse is applied, free precession interval τ , and an exponential envelope de-
mapping the phase φ onto a population difference between cay is observed (see Fig. 7a). Inhomogeneous fields limit
|0i and |+1i. Figure 3 provides a Bloch sphere depiction T2∗ by causing spins within an ensemble to undergo Lar-
of the Ramsey sequence, where the states |0i and |+1i are mor precession at different rates. As depicted in the second
denoted by |↑i and |↓i respectively. See Appendix A.1.a for Bloch sphere in Fig. 5, the spins dephase from one another
a full mathematical description of a Ramsey magnetic field after free precession intervals τ ∼ T2∗ .
measurement.
The subsequent spin readout process is fundamentally
limited by quantum mechanical uncertainty. If a measure-
ment of the final state’s spin projection Sz is performed
in the {|0i, |+1i} basis, only two measurement outcomes
are possible: 0 and 1. The loss of information associated
with this projective measurement is commonly referred to
as spin projection noise (Itano et al., 1993). A projection-
noise-limited sensor is characterized by a spin readout fi-
delity F = 1. Other considerations, such as the photon shot
noise, may lead to reductions in the fidelity F, which de- FIG. 4 Diamond containing spin impurities. NV- centers [thick
grade magnetic field sensitivity. The sensitivities of S = 1/2 red arrows (→→)] experience magnetic fields caused by other
magnetometers at the spin-projection and shot-noise limits spin defects in the diamond, including substitutional nitrogen
are discussed in Sec. II.A and treated in detail in Appen- [thin green arrows (→)], 13 C nuclei [small black arrows (→)],
dices A.1.b and A.1.c. and other paramagnetic impurities [blue (→) and purple
(→) arrows]. The inhomogeneous and time-varying dipo-
lar magnetic fields generated by these spins dephase and
decohere the NV- spin ensemble.
E. Spin dephasing and decoherence

Pulsed magnetometry measurements benefit from long Dephasing from fields that are static over the measure-
sensing intervals τ , as the accumulated magnetic-field- ment duration can be reversed by application of a π-pulse
dependent phase φ typically increases with τ . For example, halfway through the free precession interval. In this pro-
in a Ramsey measurement, φ = γe Bτ with γe = ge µB /~; tocol (Hahn, 1950), the π-pulse alters the direction of spin
maximal sensitivity of the observable φ to changes in B precession, such that the phase accumulated due to static
is therefore achieved when dB dφ
= γe τ is maximized. At fields during the second half of the sequence cancels the
the same time, contrast degrades with increasing τ due to phase from the first half. Thus, spins in nonuniform fields
dephasing, decoherence, and spin-lattice interactions, with rephase, producing a recovered signal termed a "spin echo"
associated respective relaxation times T2∗ , T2 , and T1 . The (Fig. 5). The decay of this echo signal, due to fields that
optimal interrogation time must therefore balance these two fluctuate over the course of the measurement sequence, is
competing concerns. characterized by the coherence time T2 , also called the
The parameter T2∗ characterizes dephasing associated transverse or spin-spin relaxation time. In NV- ensemble
with static or slowly varying inhomogeneities in a spin sys- systems, T2 can exceed T2∗ by orders of magnitude (Bauch
tem, e.g., dipolar fields from other spin impurities in the et al., 2018; Bauch et al., 2019; de Lange et al., 2010). As
diamond, as depicted in Fig. 4. T2∗ is the characteristic 1/e the T2 -limited spin echo sequence is intrinsically insensitive
 7

/2
/2

FIG. 5 Recovery of spin phase coherence with central π-pulse. Bloch sphere depiction of spin dephasing due to static field inho-
mogeneities (characterized by T2∗ ) followed by application of a π pulse at time τ /2 and then spin rephasing at time τ . The π-pulse
cancels T2∗ dephasing as well as sensitivity to static signal fields.

to DC magnetic fields, it is frequently employed for de- II. MEASUREMENT SENSITIVITY CONSIDERATIONS
tecting AC signals. Meanwhile, the T2∗ -limited Ramsey se-
quence is commonly employed for DC sensing experiments. A. Magnetic field sensitivity

The spin-projection-limited sensitivity of an ensemble

F. DC and AC sensing magnetometer consisting of N non-interacting spins is ap-
proximately given by (Budker and Romalis, 2007; Taylor
Quantum sensing approaches may be divided into two et al., 2008)
broad categories based on the spectral characteristics of the
fields to be detected, summarized in Table 1. In particular, ensemble ~ 1
ηsp ≈ √ , (9)
DC sensing protocols are sensitive to static, slowly-varying, ∆ms ge µB N τ
or broadband near-DC signals, whereas AC sensing proto-
cols typically detect narrowband, time-varying signals at where ge ≈ 2.003 is the NV- center’s electronic g-factor (Do-
frequencies up to ∼ 10 MHz (Boss et al., 2016, 2017; Cai herty et al., 2013), µB is the Bohr magneton, ~ is the re-
et al., 2013; Loretz et al., 2013; Pham et al., 2016; Schmitt duced Planck constant, τ is the free precession (i.e., in-
et al., 2017; Shao et al., 2016; Shin et al., 2012; Steinert terrogation) time per measurement, and ∆ms is the dif-
et al., 2013; Wood et al., 2016), although AC sensing exper- ference in spin quantum number between the two interfer-
iments of ∼ 100 MHz signals have also been demonstrated ometry states (e.g., ∆ms = 1 for a spin S = 1/2 system,
for niche applications (Aslam et al., 2017). Both DC and and ∆ms = 2 for a S = 1 system employing ms = +1 and
AC sensors employing NV- ensembles exhibit sensitivities ms = −1 states). Certain pulsed magnetometry schemes
limited, in part, by the relevant NV- spin relaxation times. such as Ramsey-based protocols can allow sensitivities ap-
DC sensitivity is limited by the ensemble’s inhomogeneous proaching the spin-projection limit, in part by ensuring that
dephasing time T2∗ , which is of order 1 µs in most present the spin state readout does not interfere with the mag-
implementations. AC sensitivity is limited by the coher- netic field interrogation (Ramsey, 1950). However, even
ence time T2 , which, as mentioned above, is typically one when employing Ramsey protocols, NV- ensemble magne-
to two orders of magnitude longer than T2∗ (Bauch et al., tometers suffer from at least three major experimental non-
2019; de Lange et al., 2010), and which can be extended idealities, which deteriorate the achievable magnetic field
through use of dynamical decoupling protocols to approach sensitivity.
the longitudinal spin relaxation time T1 (see Sec. IV.A). Ad- First, for NV- ensemble magnetometers, the spin-state
ditionally, alternative forms of T1 -limited AC sensing such initialization time tI and readout time tR may be significant
as T1 relaxometry allow phase-insensitive detection of sig- compared to the interrogation time τ . By decreasing the
nals at frequencies in the ∼ GHz regime (Casola et al., fraction of time devoted to spin precession, the finite values
2018; Hall et al., 2016; Pelliccione et al., 2014; Shao et al., of tI and tR deteriorate the sensitivity by the factor
2016; Tetienne et al., 2013). In general, the enhanced field r
tI + τ + tR
sensitivities afforded by longer AC sensor coherence times . (10)
coincide with reduced sensing bandwidth as well as insensi- τ
tivity to static fields, restricting the application space of of Second, the conventional NV- optical readout tech-
sensors employing these techniques (see Table 1). This re- nique (Doherty et al., 2013), which detects the spin-state-
view concentrates primarily on DC sensing protocols with dependent fluorescence (also commonly referred to as pho-
particular focus on sensors designed to detect broadband toluminescence or PL) in the 600 - 850 nm band, does
time-varying magnetic fields from DC to ∼ 100 kHz. not allow single-shot determination of the NV- spin state
to the spin projection limit (i.e., the standard quantum
limit) (Itano et al., 1993). An NV- center in the elec-
tronically excited spin-triplet state will decay either di-
rectly to the spin-triplet ground state or indirectly though
 8

Broadband DC sensing AC sensing

Common techniques Ramsey (Sec. II.A), CW-ODMR (Sec. II.B.1), Hahn echo, dynamical decoupling (Sec. IV.A)
pulsed ODMR (Sec. II.B.2)
Relevant relaxation Inhomogeneous spin dephasing (T2∗ ) Homogeneous spin decoherence (T2 ) and longitudi-
nal relaxation (T1 )
Frequency/bandwidth 0 to ∼ 100 kHz (pulsed), 0 to ∼ 10 kHz (CW) Center frequency: ∼ 1 kHz to ∼ 10 MHz; band-
width: . 100 kHz
Example magnetic Biocurrent detection, magnetic particle tracking, Single biomolecule and protein detection, nanoscale
sensing applications magnetic imaging of rocks and meteorites, imag- nuclear magnetic resonance, nanoscale electron spin
ing of magnetic nanoparticles in biological systems, resonance, magnetic resonant phenomena in mate-
magnetic imaging of electrical current flow in ma- rials, noise spectroscopy
terials, magnetic anomaly detection, navigation

TABLE 1 Operational regimes and selected applications of broadband DC and AC sensing protocols employing NV- ensembles in
diamond. T1 relaxometry methods are not considered.

a cascade of spin-singlet states (Rogers et al., 2008) via

!"#$%&'%('%)$*+%),*$-.)"(/+&0
an inter-system crossing (Goldman et al., 2015a,b; Thier-
ing and Gali, 2018). Conventional NV- optical readout ex-
ploits the ms = ±1 states’ higher likelihood to enter the
singlet-state cascade more often than the ms = 0 state (see
Table 13). An NV- center that enters the singlet state cas-
cade does not fluoresce in the 600 - 850 nm band, whereas
an NV- center decaying directly to the spin-triplet ground
state can continue cycling between the ground and excited
triplet states, producing fluorescence in the 600 - 850 nm
band. The ms = ±1 states therefore produce on average
less PL in the 600 - 850 nm band, as shown in Fig. 6. Un-
fortunately the ∼ 140 − 200 ns (Acosta et al., 2010b; Gupta
et al., 2016; Robledo et al., 2011) spin-singlet cascade life-
time and limited differences in ms = ±1 and ms = 0 decay FIG. 6 Fluorescence of the NV- spin states. NV- centers pre-
behavior allows for only probabilistic determination of the pared in the ms = 0 state emit photons at a higher rate than
centers prepared in the ms = ±1 states. This spin-dependent
NV- initial spin state. Following Ref. (Shields et al., 2015), fluorescence forms the basis of conventional NV- readout. Data
we quantify the added noise from imperfect readout with courtesy of Brendan Shields.
the parameter σR ≥ 1, such that σR = 1 corresponds to
readout at the spin projection limit. This parameter is the
inverse of the measurement fidelity: F ≡ 1/σR . For im- dephasing occurs with characteristic time T2∗ so that η is
perfect readout schemes, the value of σR can be calculated additionally deteriorated by the factor
as (Shields et al., 2015; Taylor et al., 2008)
1
∗ p , (13)
s
2(a + b) e−(τ /T2 )
σR = 1 + (11)
(a − b)2 where the value of the stretched exponential parameter p
depends on the origin of the dephasing (see Appendix A.7).
s
1
= 1+ 2 , (12) NV- spin resonance lineshapes with exactly Lorentzian pro-
C navg
files correspond to dephasing with p = 1, and spin reso-
where a and b respectively denote the average numbers of nance lineshapes with Gaussian profiles correspond to p = 2
photons detected from the ms = 0 and ms = ±1 states of (see Appendix A.5).
Combining Eqns. 9, 10, 11, and 13 gives the sensitivity
a single NV- center during a single readout. In Eqn. 12
for a Ramsey-type NV- broadband ensemble magnetome-
we identify C = a−b
a+b as the measurement contrast (i.e., the ter (Popa et al., 2004) as
interference fringe visibility) and navg = a+b
2 as the aver- ensemble
age number of photons collected per NV- center per mea- ηRamsey ≈ (14)
surement. Although sub-optimal initialization and readout
s r
~ 1 1 1 tI +τ +tR
times tI and tR can degrade the value of C, it is henceforth √ ∗ 1+ 2
∆ms ge µB N τ e−(τ /T2 )p C navg τ
assumed that tI and tR are chosen optimally. | {z } | {z } | {z } | {z }
Spin projection limit Spin dephasing Overhead time
Third, the sensitivity η is degraded for increased values Readout

of τ due to spin dephasing during precession. For Ramsey- where N is the number of NV- centers in the ensemble and
type pulsed magnetometry (i.e., with no spin echo), the ∆ms = 1 for the effective S = 1/2 subspace employed for NV-
 9

magnetometry using the ms = 0 and ms = ±1 basis. How- a b !"%

)+(2*3 #$%&'()
ever, in the limit of measurement contrast C  1 and when
!"$

45&-6'78+27
the number of photons collected per NV- center per optical #$%&'(
!1% !1%
readout is much less than one, the readout fidelity is lim- !" " !"!

ited by photon shot noise and can be approximated using !!"$

)*+,
1/F = σR ≈ C √1navg . Defining N = N navg to be the aver- ./0
!!"%
age number of photons detected per measurement from the !"! !"# $"! $"# %"!
98**&:8*;*22<6'&7<(*&"
%"# &"!

ensemble of N NV- centers yields the following shot-noise- c d #%$$

limited sensitivity equation for a Ramsey scheme (Pham, -"./0!)
#56 *+

@68(+><A*,&45
2013): #$%&'( $%('

√ !" $%($
ensemble,shot ~ 1 tI + τ + tR
ηRamsey ≈ √ .
N
∗ )p )*+, $%&'
∆ms ge µB Ce −(τ /T2 τ
(15) $%&$
!! !" !# $ # " !
Hereafter, we assume broadening mechanisms produce 98*?=*';3

Lorentzian lineshapes, so that p = 1. For negligible tI and

e f #%$$
4=>2*,&/0!)
-
tR , the optimal measurement time is τ = T2∗ /2, whereas

@68(+><A*,&45
#123'(4 *+ ,
#$%&'( $%('
!./0
for tI + tR  T2∗ , the optimal τ approaches T2∗ (see Ap-
!" "7 $%($
pendix A.2). Equation 15 illustrates the benefits attained
by increasing the dephasing time T2∗ , the measurement con- )*+, $%&'

trast C, the number of NV- spin sensors N , and the average

$%&$
number of photons detected per NV- per measurement navg . !! !" !# $
98*?=*';3 # " !

Table 2 lists values of σR and N achieved using conventional

optical readout in pulsed and CW magnetometry measure- FIG. 7 Overview of Ramsey, CW-ODMR, and pulsed ODMR
ments, with both single NV- centers and ensembles. At magnetometry protocols. a) Schematic of Ramsey magnetom-
present, conventional optical readout is insufficient to reach etry protocol. b) Representation of free induction decay asso-
the spin projection limit for both single- and ensemble-NV- ciated with a Ramsey protocol versus free precession time τ .
sensors. Appendix A.1 derives the sensitivity for a Ramsey- Fringes exhibit contrast CRamsey and decay exponentially with
dephasing time T2∗ . c) Schematic of CW-ODMR sensing proto-
type magnetometer in both the spin projection and shot
col. d) Representation of CW-ODMR spectrum with contrast
noise limits. CCW and linewidth ∆ν. e) Schematic of pulsed ODMR sensing
In addition to Ramsey-type methods, other protocols al- protocol with MW π-pulse time τπ ∼ T2∗ . f) Representation
low measurement of DC magnetic fields. These alternative of pulsed ODMR spectrum with contrast Cpulsed and linewidth
methods, including continuous-wave and pulsed optically ∆ν ∼ 1/(πT2∗ ).
detected magnetic resonance, offer reduced sensitivity com-
pared to Ramsey-type sequences (for a fixed number of NV-
centers addressed), as discussed in the following sections. Alternatively, the MW frequency may be tuned to a spe-
cific resonance feature’s maximal slope, so that incremental
changes in magnetic field result in maximal changes in PL.
B. Alternatives to Ramsey magnetometry The sensitivity of this latter approach can be further im-
proved by modulating the MW frequency to combat noise
1. CW-ODMR or by exciting multiple hyperfine transitions simultaneously
to improve contrast (Barry et al., 2016; El-Ella et al., 2017;
Continuous-wave optically detected magnetic resonance Schloss et al., 2018).
(CW-ODMR) is a simple, widely employed magnetometry CW-ODMR does not require pulsed optical excitation,
method (Acosta et al., 2009; Barry et al., 2016; Dréau et al., MW phase control, fast photodetectors, multichannel tim-
2011; Fuchs et al., 2008; Schloss et al., 2018; Schoenfeld ing generators, or switches; the technique is therefore
and Harneit, 2011; Tetienne et al., 2012) wherein the MW technically easier to implement than pulsed measurement
driving and the optical polarization and readout occur si- schemes. Additionally, CW-ODMR is more tolerant of MW
multaneously (see Fig. 7c). Laser excitation continuously inhomogeneities than pulsed schemes and, when properly
polarizes NV- centers into the more fluorescent ms = 0 implemented, may yield similar sensitivities to pulsed mag-
ground state, while MWs tuned near resonance with one of netometry protocols when a larger number of sensors are
the ms = 0 ↔ ms = ±1 transitions drive NV- population interrogated with the same optical excitation power (Barry
into the less fluorescent ms = ±1 state (reducing the emit- et al., 2016).
ted light). A change in the local magnetic field shifts the The shot-noise-limited sensitivity of an NV- magnetome-
resonance feature with respect to the MW drive frequency, ter employing CW-ODMR is given by (Barry et al., 2016;
causing a change in the detected fluorescence, as illustrated Dréau et al., 2011)
in Fig. 7d.
In the simplest CW-ODMR implementation, the MW
frequency is swept across the entire NV- resonance spec- 4 h ∆ν
ηCW = √ √ , (16)
trum, allowing all resonance line centers to be determined. 3 3 ge µB CCW R
 10

Reference Readout method Single NV- /ensemble σR N [counts/measurement]

(Shields et al., 2015) conventional single 10.6 9.45 × 105 cps × tR

(Shields et al., 2015) spin-to-charge conversion single 2.76 -
(Lovchinsky et al., 2016) conventional single 35 ∼ 105 cps × tR
(Lovchinsky et al., 2016) ancilla-assisted single 5 -
(Fang et al., 2013) conventional single 80 0.01
(Hopper et al., 2016) conventional single 48 0.04
(Hopper et al., 2016) spin-to-charge conversion single 3 -
(Jaskula et al., 2017) conventional single 54 0.022
(Jaskula et al., 2017) spin-to-charge conversion single 5 -
(Neumann et al., 2010a) ancilla-assisted single 1.1 -
(Le Sage et al., 2012) conventional ensemble 67 2×108
(Wolf et al., 2015b) conventional ensemble ∼ 1000 1012
(Chatzidrosos et al., 2017) NIR absorption† ensemble 65 -
(Barry et al., 2016) conventional† ensemble ∼ 5000 -
(Schloss et al., 2018) conventional† ensemble ∼ 5000 -

TABLE 2 Example literature values for readout schemes employing conventional optical readout or alternative techniques. The
parameter σR characterizes the factor above the spin projection limit and N is the average number of photons collected per
measurement. Conventional NV- readout is unable to reach the spin projection limit (σR = 1), whereas alternative schemes can
allow readout to approach this limit. The best demonstrated pulsed readout methods with ensembles are presently ∼ 100× away
from the spin projection limit. The symbol † denotes non-pulsed schemes for comparison, and dashed lines (-) indicate values not
reported (or not applicable to non-pulsed schemes).

with photon detection rate R, linewidth √ ∆ν and CW- existing CW-ODMR devices (Barry et al., 2016; Schloss
ODMR contrast CCW . The prefactor 4/(3 3) originates et al., 2018). Moreover, the poor readout fidelity accom-
from the steepest slope of the resonance lineshape when as- panying the low required initialization intensity is particu-
suming a Lorentzian resonance profile and is achieved for larly deleterious to applications where volume-normalized
a detuning of 2∆ν √ from the linecenter (Vanier and Au-
3
sensitivity (i.e., the sensitivity within a unit interrogation
doin, 1989). Operation of a CW-ODMR magnetometer volume) is important.
can be modeled using the rate equation approach from
Refs. (Dréau et al., 2011; Jensen et al., 2013).
2. Pulsed ODMR
However, CW-ODMR is not envisioned for many high-
sensitivity applications for multiple reasons. First, CW-
Pulsed ODMR is an alternative magnetometry method
ODMR precludes use of pulsed methods to improve sen-
first demonstrated for NV- centers by Dréau et al. in
sitivity, such as double-quantum coherence magnetometry
Ref. (Dréau et al., 2011). Similar to Ramsey and in con-
(see Sec. IV.B), and many readout-fidelity enhancement
trast to CW-ODMR, this technique avoids optical and MW
techniques. In particular, the readout fidelity is quite poor
power broadening of the spin resonances, enabling nearly
compared to conventional pulsed readout schemes, as shown
T2∗ -limited measurements. In contrast to Ramsey magne-
by the last two entries in Table 2. Second, CW-ODMR
tometry, however, pulsed ODMR is linearly sensitive to spa-
methods suffer from MW and optical power broadening,
tial and temporal variations in MW Rabi frequency. When
degrading both ∆ν and CCW compared to optimized Ram-
such variations are minimal, pulsed ODMR sensitivity may
sey sequences. Optimal CW-ODMR sensitivity is achieved
approach that of Ramsey magnetometry without requir-
approximately when optical excitation, MW drive, and
ing high Rabi frequency (Dréau et al., 2011), making the
T2∗ dephasing contribute roughly equally to the resonance
method attractive when high MW field strengths are not
linewidth (Dréau et al., 2011). In this low-optical-intensity
available.
regime, the detected fluorescence rate per interrogated NV-
In the pulsed ODMR protocol, depicted schematically in
center is significantly lower than for an optimized Ramsey
Fig. 7e, the NV- spin state is first optically initialized to
scheme, which results in readout fidelities ∼ 103 below the
ms = 0. Then, during the interrogation time τ , a near-
spin projection limit (Barry et al., 2016). This low opti-
resonant MW π-pulse is applied with duration equal to
cal intensity requirement becomes more stringent as T2∗ in-
the interrogation time, τπ = τ , where the Rabi frequency
creases, meaning that CW-ODMR sensitivity largely does
ΩR = π/τπ . Finally, the population is read out optically.
not benefit from techniques to extend T2∗ .
A change in the magnetic field detunes the spin resonance
Overall, the combination of poor readout fidelity (and no
with respect to the MW frequency, resulting in an incom-
proposed path toward improvement) combined with an in-
plete π-pulse and a change in the population transferred to
ability to benefit from extended T2∗ suggests that prospects
the ms = ±1 state prior to optical readout.
are poor for further sensitivity enhancement over the best
For a Lorentzian resonance lineshape (see Appen-
 11

dices A.5 and A.6), the expected shot-noise-limited sen- interrogated NV- centers and same mean photon collec-
sitivity may be calculated starting from the shot-noise- tion rate R) because pulsed ODMR enables use of high
limited CW-ODMR sensitivity given by Eqn. 16. For pulsed optical intensities that would degrade CCW (Dréau et al.,
ODMR, the resonance profile is given by a convolution of 2011). Although Cpulsed may approach the Ramsey con-
the T2∗ -limited line profile and additional broadening from trast CRamsey (see Fig. 7a,b), Cpulsed < CRamsey is ex-
the NV- spin’s response to a fixed-duration, detuned MW π- pected in practice for several reasons: first, because the
pulse, as shown in Fig. 8. When the interrogation time τπ is technique requires Rabi frequencies to be of the same or-
set to ≈ T2∗ , these two broadening mechanisms contribute der as the NV- linewidth set by T2∗ , the MW drive may be
approximately equally to the resonance linewidth (Dréau too weak to effectively address the entire inhomogeneously-
et al., 2011). Assuming τπ ≈ T2∗ , we write the pulsed broadened NV- ensemble. Second, while the high Rabi fre-
ODMR linewidth ∆ν as ∆ν ≈ Γ = 1/(πT2∗ ) (see Fig. 7f), quencies ∼ 2π ×10 MHz commonly employed in Ramsey se-
while noting that this approximation likely underestimates quences effectively drive all hyperfine-split NV- transitions
the linewidth by . 2×. of 14 NV- or 15 NV- (Acosta et al., 2009), the weaker π-pulses
required for pulsed ODMR cannot effectively drive all hy-
perfine transitions with a single tone. Pulsed ODMR op-
Tπ
=
450
ns
1.00 eration at the excited-state level anticrossing (Dréau et al.,
0.95
2011) or utilizing multi-tone MW pulses (Barry et al., 2016;
El-Ella et al., 2017; Vandersypen and Chuang, 2005) could
0.90
allow more effective driving of the entire NV- population
Tπ
=
900
ns and higher values of Cpulsed . However, when multi-tone
1.00
pulses are employed, care should be taken to avoid degra-
0.95
dation of Cpulsed due to off-resonant MW cross-excitation,
0.90 which may be especially pernicious when the T2∗ -limited
Tπ
=
1800
ns linewidth (and thus MW Rabi frequency) is similar to the
Normalized PL

1.00 hyperfine splitting.

0.95 Although pulsed ODMR may sometimes be preferable to
0.90 Ramsey, the former technique ultimately √ provides inferior
sensitivity. Several factors of order 2 (which arise from
Tπ
=
3600
ns
1.00
a lineshape-dependent numerical prefactor (Dréau et al.,
2011), MW Fourier broadening, nonuniform ensemble driv-
0.95
ing, and hyperfine driving inefficiencies) combine to degrade
0.90 the pulsed ODMR sensitivity with respect to that of Ram-
Tπ
=
14400
ns sey. Furthermore, unlike double-quantum Ramsey magne-
1.00
tometry (see Sec. IV.B), pulsed ODMR has not been exper-
0.95 imentally demonstrated to mitigate line broadening from
0.90
temperature fluctuations or other dephasing mechanisms
2.892 2.896 2.900 2.904 common-mode to |ms = −1i and |ms = +1i. Hypotheti-
Frequency [GHz] cal double-quantum analogs to pulsed ODMR (Fang et al.,
2013; Taylor et al., 2008) might likely require, in addition
FIG. 8 Pulsed ODMR spectra for various π-pulse durations τπ . to the sensing π-pulse, high-Rabi-frequency MW pulses to
When τπ  T2∗ , the resonance lineshape is Fourier-broadened
initialize the | ± 1i superposition states, similar to those
beyond the T2∗ -limited linewidth. When τπ  T2∗ the photo-
luminescence (PL) contrast is diminished due to spin dephas- employed for double-quantum Ramsey, which would under-
ing. Choice of τπ ∼ T2∗ (≈ 3 µs here) allows nearly-T2∗ -limited mine pulsed ODMR’s attractive low MW Rabi frequency
linewidths while preserving PL contrast. From Ref. (Dréau requirements.
et al., 2011). A generalization of pulsed ODMR is Rabi beat sens-
ing (Fedder et al., 2011; Rabi, 1937), wherein the spins
Choosing initialization and readout times tI and tR and are driven through multiple Rabi oscillations during the
interrogation time τπ = T2∗ reduces the time-averaged pho- interrogation time. Under optimal conditions, Rabi beat
ton collection rate R by the readout duty cycle tR /(tI + magnetometry, like the specific case of pulsed ODMR, may
T2∗ + tR ). Then, defining N = RtR to be the mean number exhibit sensitivity approaching that of Ramsey magnetom-
of photons collected per optical readout cycle and replac- etry. For the regime where the Rabi frequency ΩR is large
ing CCW with the pulsed-ODMR contrast Cpulsed yields the compared to the resonance linewidth (∼ 1/T2∗ ), sensitivity
pulsed-ODMR sensitivity is optimized when the detuning is chosen to be similar to the
Rabi frequency (∆ ∼ ΩR ), when the interrogation time is
tI + T2∗ + tR
p
8 ~ 1 similar to the dephasing time (τ ∼ T2∗ , see Appendix A.2),
ηpulsed ≈ √ √ . (17)
3 3 ge µB Cpulsed N T2∗ and when τ is chosen to ensure operation at a point of
maximum slope of the Rabi magnetometry curve. How-
The value of Cpulsed under optimized conditions is ex- ever, Rabi beat magnetometry is sensitive to spatial and
pected to be higher than CCW (for the same number of temporal variations in the MW Rabi frequency ΩR (Ram-
 12

Sensitivity optimization
Parameter
Method Method description and evaluation
optimized
Double-quantum Doubles effective gyromagnetic ratio. Removes dephasing from mechanisms inducing shifts
coherence magnetometry common to the |ms = ±1i states, such as longitudinal strain and temperature. Minor
(Sec. IV.B) additional MW hardware usually required. Generally recommended.
Bias magnetic field Suppresses dephasing from transverse electric fields and strain at bias magnetic fields of
(Sec. IV.D) several gauss or higher. Generally recommended.
Spin bath driving Mitigates or eliminates dephasing from paramagnetic impurities in diamond. Each impurity’s
(Sec. IV.C) spin resonance must be addressed, often with an individual RF frequency. Additional RF
hardware is required. Recommended for many applications.
Dynamical decoupling Refocuses spin dephasing using one or more MW π-pulses, extending the relevant relaxation
Dephasing (Sec. IV.A) time from T2∗ to T2 , with fundamental limit set by 2T1 . Recommended for narrowband AC
∗
time T2 sensing; generally precludes DC or broadband magnetic sensing.
Rotary echo Extends measurement time using a MW pulse scheme but offers reduced sensitivity relative
magnetometry to Ramsey. Not recommended outside niche applications.
(Sec. VII.A)
Geometric phase Offers increased dynamic range, using a MW spin manipulation method, at the cost of
magnetometry reduced sensitivity relative to Ramsey. Not recommended outside niche applications.
(Sec. VII.B)
Ancilla-assisted Employs NV- hyperfine interaction to convert DC magnetic fields to AC fields to be sensed
upconversion using dynamical decoupling. Operates near ground-state level anticrossing (103 gauss) and
magnetometry offers similar or reduced sensitivity relative to Ramsey. Not generally recommended.
(Sec. VII.C)
Spin-to-charge Maps spin state to charge state of NV, increasing number of photons collected per mea-
conversion readout surement. Allows σR ≈ 3 for single NV centers, and initial results show improvement over
(Sec. V.A) conventional readout for ensembles. Substantially increased readout time likely precludes
application when T2∗ . 3 µs. Requires increased laser complexity. Technique is considered
promising; hence, further investigation is warranted.
Ancilla-assisted Maps NV- electronic spin state to nuclear spin state, enabling repetitive readout and in-
repetitive readout creased photon collection. Allows σR to approach 1 for single NVs; no fundamental barriers
(Sec. V.C) to ensemble application. Substantially increased readout time likely precludes application
when T2∗ . 3 µs. Requires high magnetic field strength and homogeneity. Technique is
considered promising, although further investigation is warranted.
Improved photon Improves σR by reducing fractional shot noise contribution, subject to unity collection and
collection (Sec. V.E) projection noise limits. Near-100% collection efficiency is possible in principle, making this
mainly an engineering endeavor. While many schemes are incompatible with wide-field
imaging, the method is generally recommended for optical-based readout of single-channel
bulk sensors.
NIR absorption readout Probabilistically reads out initial spin populations using optical absorption on the 1 E ↔1 A1
(Sec. V.F) singlet transition. Demonstrated σR values are on par with conventional ensemble readout,
Readout and prospects for further improvement are unknown. Technique is best used with dense
fidelity ensembles and an optical cavity but is hindered by non-NV- absorption and non-radiative
F = 1/σR NV- singlet decay. Further investigation is warranted.
Photoelectric readout Detects spin-dependent photoionization current. Best for small 2D ensembles; has not yet
(Sec. V.B) demonstrated sensitivity improvement with respect to optimized conventional readout.
Level-anticrossing- Increases the number of spin-dependent photons collected per readout by √ operation at the
assisted readout excited-state level anticrossing. Universally applicable, but at best offers a 3 improvement
(Sec. V.D) in σR . Not recommended outside niche applications.
Green absorption Probabilistically reads out initial spin populations using optical absorption on the 3 A2 ↔3 E
readout (Sec. V.G) triplet transition. Performs best with order unity optical depth. Demonstrations exhibit
contrast below that of conventional readout by 3× or more. Prospects are not considered
promising.
Laser threshold Probes magnetic field by measuring lasing threshold, which depends on NV- singlet state
magnetometry population. Moderately improved collection efficiency and contrast are predicted compared
(Sec. V.H) to conventional readout. Challenges include non-NV- absorption and system instability near
lasing threshold. Prospects are not considered promising.
Entanglement-assisted Harnesses strong NV- dipolar interactions to improve readout fidelity beyond the standard
magnetometry quantum limit. Existing proposals require 2D ensembles, impose long overhead times, and
(Sec. VII.D) exhibit unfavorable coherence time scaling with number of entangled spins. While existing
protocols are not considered promising, further investigation toward developing improved
protocols is warranted.

TABLE 3 Summary analysis of approaches to optimize ensemble-NV-diamond magnetic sensitivity

 13

Diamond material optimization

Parameter
Method Method description and evaluation
optimized
N-to-NV CVD Common synthesis method that can produce high-quality ensemble-NV diamonds. Rel-
conversion synthesis atively easy to control dimensions and concentrations of electronic and nuclear spins.
efficiency χ (Sec. VI.C) May introduce strain and unwanted impurities, which can limit achievable ζ, χ, and T2∗ .
(Sec. VI.A) Effective for producing NV- -rich-layer diamonds.
NV-to-NV- HPHT Common synthesis method that can produce high-quality ensemble-NV diamonds with
charge state synthesis lower strain and fewer lattice defects than CVD. Control over doping and impurity
efficiency ζ (Sec. VI.C) concentration may be more difficult than in CVD. Not intrinsically amenable to creating
(Sec. VI.A) NV- -rich-layer diamonds. Ferromagnetic metals may incorporate into diamond.
Paramagnetic Irradiation Diamond treatment method that, combined with subsequent annealing, converts substi-
impurities (Sec. VI.D) tutional nitrogen to NV centers. Electrons are preferred irradiation particle. Dose should
(Sec. VI.F) be optimized for diamond’s nitrogen concentration to create high ζ without degrading
χ. Generally recommended with annealing for producing NV- -rich diamonds.
Strain LPHT Low-pressure annealing that, combined with prior irradiation, converts substitutional
(Sec. III.C) annealing nitrogen to NV centers. Heals some diamond lattice damage. NV- centers are created
(Sec. VI.E) effectively at ∼ 800 ◦ C; additional treatment at ∼ 1200 ◦ C may eliminate some unwanted
impurities. Generally recommended with irradiation for producing NV- -rich diamonds.
Nuclear spins HPHT High-pressure annealing may reduce strain and eliminate some unwanted impurities.
(Sec. III.F) treatment May enable increases in ζ and χ. Recommended for diamonds with balanced aspect
(Sec. VI.C) ratios.
Isotopic Diamond synthesis with isotopically enriched source (gas for CVD and typically solid
enrichment for HPHT) allows reduction of unwanted nuclear spin concentration (e.g., 13 C) and
(Sec. III.F) selection of nitrogen isotope (14 N or 15 N) incorporated into NV- . CVD diamonds with
[13 C] ≈ 20 ppm have been synthesized. Recommended for achieving long T2∗ .
Surface Surface termination with favorable atomic elements can stabilize the desired NV charge
treatment state near the surface and extend relaxation times. Generally recommended.
(Sec. VI.A)
Preferential CVD synthesis of diamond with NV centers preferentially oriented along a single axis.
orientation At present, preferential orientation is only maintained in unirradiated diamonds, largely
(Sec. VI.G) hindering its capability to produce NV- -rich diamonds. Not generally recommended.

TABLE 4 Summary analysis of diamond engineering parameters and methods for high-sensitivity ensemble-NV- magnetometry.
Colored lines indicate methods that may be employed to optimize each parameter.
 14

sey, 1950). For high values of ΩR , MW field variations may (iii) Sensor Number, Density, or Interrogation
limit the Rabi measurement’s effective T2∗ . Hence, prac- Volume | In theory, the number of sensors N can be
tical implementations of Rabi beat magnetometry on NV- increased without limit. However, practical consid-
ensembles likely perform best when ΩR ∼ 1/T2∗ , i.e., when erations may prevent this approach. First, a larger
the scheme reduces to pulsed ODMR. value of N (and an associated larger number of pho-
tons N) can increase some types of technical noise
that scale as N , e.g., noise from timing jitter in de-
C. Parameters limiting sensitivity vice electronics or from excitation-laser intensity fluc-
tuations.
√ As photon shot noise scales more slowly
Examination of Eqn. 14 reveals the relevant parameters as N , achieving a shot-noise-limited sensitivity be-
ensemble
limiting magnetic field sensitivity ηRamsey : (i) the dephas- comes more difficult with increasing N . Second, large
ing time T2∗ ; (ii) the readout fidelity F = 1/σR ; (iii) the values of N can require impractically high laser pow-
sensor density [NV- ] and the interrogated diamond vol- ers, since the number of photons needed for NV- spin
ume V , which together set the total number of sensors initialization scales linearly with N . While larger N
N = [NV- ] × V ; (iv) the measurement overhead time can be achieved either by increasing the NV- den-
tO = tI +tR ; and (v) the relative precession rates of the two sity or increasing the interrogation volume, both ap-
states comprising the interferometry measurement. Sensi- proaches result in distinct technical or fundamental
tivity enhancement requires improving one or more of these difficulties. Increasing N by increasing the interroga-
parameters. As we will discuss, parameters (i) and (ii) are tion volume with fixed [NV- ] may increase the dia-
particularly far from physical limits and therefore warrant mond cost and creates more stringent uniformity re-
special focus. quirements for both the bias magnetic field (to avoid
(i) Dephasing Time T2∗ | In current realiza- degrading the dephasing time T2∗ ) and the MW field
tions, dephasing times in application-focused broad- (to ensure uniform spin manipulation over the sens-
band NV- ensemble magnetometers (Barry et al., ing volume). Furthermore, increasing interrogation
2016; Chatzidrosos et al., 2017; Clevenson et al., 2015; volume is incompatible with high-spatial-resolution
Kucsko et al., 2013) are typically T2∗ . 1 µs. Con- sensing and imaging modalities (Fu et al., 2014; Glenn
sidering the physical limit T2∗ ≤ 2T1 (Alsid et al., et al., 2017, 2015; Le Sage et al., 2013; Pham et al.,
2019; Bauch et al., 2019; Jarmola et al., 2012; Levitt, 2011; Simpson et al., 2016; Steinert et al., 2010; Teti-
2008), with longitudinal relaxation time T1 ≈ 6 ms for enne et al., 2017). On the other hand, increasing
NV- ensembles (Jarmola et al., 2012), a maximum NV- density will increase dephasing from dipolar cou-
T2∗ ≈ 12 ms is theoretically achievable, corresponding pling and decrease T2∗ unless such effects are miti-
to a sensitivity enhancement of ≈ 100×. Although gated (see, e.g.,√Sec. IV.C). Finally, because sensitiv-
the feasibility of realizing T2∗ values approaching 2T1 ity scales as 1/ N , we expect increasing N to allow
remains unknown, we consider improvement of T2∗ to only modest enhancements (e.g., . 5×) over standard
be an effective approach to enhancing sensitivity (see methods. To date no demonstrated high sensitivity
Sec. III.A). While the stretched exponential parame- bulk NV-diamond magnetometer (Barry et al., 2016;
ter p can provide information regarding the dephasing Chatzidrosos et al., 2017; Clevenson et al., 2015; Wolf
source limiting T2∗ , its value (typically between 1 and et al., 2015b) has utilized more than a few percent of
2 for ensembles) does not strongly affect achievable the available NV- in the diamond, suggesting limited
sensitivity (Bauch et al., 2018). utility for increasing sensor number N in current de-
vices. See Appendix A.3 for additional analysis.
(ii) Readout Fidelity | Increasing readout fidelity
F = 1/σR is another effective method to enhance (iv) Overhead Time | Although measurement over-
sensitivity, as fractional fidelity improvements re- head time can likely be decreased to ∼ 1 µs, max-
sult in equal fractional improvements in sensitivity. imum sensitivity enhancement (in the regime where
With conventional 532 nm fluorescence readout, cur- T2∗ ∼ tI + tR ) is expected to be limited to order unity,
rent NV- ensemble readout fidelities F are a fac- (. 3×). See Sec. III.A for a more detailed discussion.
tor & 67× removed from the spin projection limit (v) Precession Rate | Use of the NV- center’s full
σR = 1 (Le Sage et al., 2012), indicating large im- S = 1 spin can allow ∆ms = 2 in Eqns. 14 and 15,
provements might be possible. For comparison, mul- i.e., a 2× increase in the relative precession rate of
tiple readout methods employing single NV- centers the states employed compared to use of the standard
achieve F within 5× of the spin projection limit, i.e., S = 1/2-equivalent subspace (see Sec. IV.B) (Bauch
σR < 5 (Ariyaratne et al., 2018; Hopper et al., 2016, et al., 2018; Fang et al., 2013). However, further im-
2018b; Jaskula et al., 2017; Lovchinsky et al., 2016; provement is unlikely, as the NV- spin dynamics are
Shields et al., 2015) with Ref. (Neumann et al., 2010a) fixed.
achieving σR = 1.1.
We note that the derivation of Eqn. 14 makes cer-
In contrast, we believe prospects are modest for improv- tain assumptions (in particular, independence of the N
ing sensitivity by engineering parameters (iii), (iv), and (v). sensors) that do not apply to some exotic approaches,
 15

such as exploiting strong NV- -NV- interactions via Flo- T2∗ values may relax certain technical requirements by al-
quet techniques and harnessing entanglement for sensing lowing lower duty cycles for specific experimental protocol
(see Sec. VII.D) (Choi et al., 2017). steps. In a standard Ramsey-type experiment, the opti-
Table 3 summarizes our analysis of present and proposed cal initialization and optical readout each occur once per
techniques to optimize ensemble-NV- magnetic field sensi- measurement sequence. Assuming a fixed mean number of
tivity. Table 4 summarizes our review of engineering meth- photons are required for spin polarization and and for read
ods for producing optimized diamond samples for high- out of the NV- ensemble, the time-averaged optical power
sensitivity ensemble-NV- magnetometry. and resulting heat load are expected to scale as 1/T2∗ . Re-
ducing heat loads is prudent for minimizing temperature
variation of the diamond, which shifts the energy splitting
III. LIMITS TO RELAXATION TIMES T2∗ AND T2 between |ms = 0i and |ms = ±1i and may require correc-
tion (see Sec. IV.B). Minimizing heat load is also important
A. Motivation to extend T2∗
for many NV-diamond sensing applications, particularly in
the life sciences. Assuming a fixed overhead time tO , the
A promising approach to enhance DC sensitivity focuses
realization of higher values of T2∗ , and thus τ , necessitates
on extending the dephasing time T2∗ (Bauch et al., 2018).
processing fewer photons per unit time, which may relax
The effectiveness of this approach may be illustrated by
design requirements for the photodetector front end and
close examination of Eqns. 14, 15. First, optimal sensitiv-
associated electronics (Hobbs, 2011).
ity is obtained when the precession time τ is similar to the
Extended T2∗ times can provide similar benefits to the
dephasing time T2∗ (see Appendix A.2), so that the approx-
MW-related aspects of the measurement. A standard
imation τ ∼ T2∗ is valid for an optimized system. There-
Ramsey-type measurement protocol employs a MW π/2-
fore, for the simple arguments presented in this section, we
pulse before and after every free precession interval. If the
assume that T2∗ extensions translate to proportional exten-
length of each π/2-pulse is held fixed, the time-averaged
sions of the optimal τ . When the dephasing time T2∗ is
MW power and resulting heat load will scale as 1/T2∗ . Addi-
similar to or shorter than the measurement overhead time
tionally, higher T2∗ values can allow for more sophisticated,
(T2∗ . tO ≡ tI + tR ), which may be typical for Ramsey
longer-duration MW pulse sequences, in place of simple
magnetometers employing ensembles of NV- centers in dia-
π/2-pulses, to mitigate the effects of Rabi frequency in-
monds with total nitrogen concentration [NT ] = 1-20 ppm,
homogeneities (Angerer et al., 2015; Nöbauer et al., 2015;
the sensitivity enhancement may then be nearly linear in
Vandersypen and Chuang, 2005) or allow for other spin-
T2∗ , as shown in Fig. 9.
manipulation protocols. Finally, higher T2∗ values could
The above outlined sensitivity scaling can be intuitively
make exotic readout schemes that tend to have fixed time
understood as follows: when the free precession time is
penalties attractive, such as spin-to-charge conversion read-
small relative to the overhead time, i.e., τ ∼ T2∗  tO ,
out (Shields et al., 2015) (see Sec. V.A) and ancilla-assisted
doubling T2∗ (thus doubling τ ) results in twice the phase
repetitive readout (Jiang et al., 2009; Lovchinsky et al.,
accumulation per measurement sequence and only a slight
2016) (see Sec. V.C).
increase in the total sequence duration; in this limit, mag-
netometer sensitivity is enhanced by nearly 2×. This fa-
vorable sensitivity scaling positions T2∗ as an important pa- Reference No. NV- probed tI tR
rameter to optimize when T2∗ . tO . (Shields et al., 2015) single 150 ns -
Typical NV- ensemble T2∗ values are ∼ 500 ns in (de Lange et al., 2012) single 600 ns 600 ns
[NT ] ≈ 20 ppm chemical-vapor-deposition-grown diamonds
(Hopper et al., 2016) single 1 µs 200 ns
from Element Six, a popular supplier of scientific diamonds.
(Fang et al., 2013) single 2 µs 300 ns
Even when employing extraordinarily optimistic values of
(Maze et al., 2008) single 2 µs 324 ns
tI = 1 µs and tR = 300 ns in Ramsey sequences performed
(Neumann et al., 2009) single 3 µs -
on such ensembles, only roughly one quarter of the total
measurement time is allocated to free precession. In this (Le Sage et al., 2012) ensemble 600 ns 300 ns
regime, as discussed above, the sensitivity scales as ∼ 1/T2∗ . (Bauch et al., 2018) ensemble 20 µs -
Although values of tI and tR vary in the literature (see (Wolf et al., 2015b) ensemble 100 µs 10 µs
Table 5), the use of longer tI and tR may be desired to (Mrózek et al., 2015) ensemble 1 ms -
achieve better spin polarization and higher readout fidelity. (Jarmola et al., 2012) ensemble 1 ms -
Notably, initialization times are typically longer for NV- en-
sembles than for single NV- defects, as higher optical excita- TABLE 5 Initialization and readout times in the literature used
tion power is required to achieve the NV- saturation inten- for conventional optical readout of NV- defects. In general, NV-
ensembles require longer initialization times than single NV- de-
sity over spatially-extended ensembles, and, furthermore, fects, in part due to the often non-uniform optical excitation
non-uniformity in optical intensity (e.g., from a Gaussian intensity applied to the ensemble (Wolf et al., 2015b). Dashed
illumination profile) can be compensated for by increasing lines (-) indicate values not reported.
the initialization time (Wolf et al., 2015b).
Longer dephasing times T2∗ offer additional benefits be-
yond direct sensitivity improvement. For example, higher
 16

with an electronic or nuclear spin bath. The strength of

these fluctuations can vary spatially throughout a sample
Relative sensitivity enhancement (arb. units)

due to the microscopically nonuniform distribution of bath

100
spins. As a result, different NV- centers in the same sample
50 display different T2∗ {single} values (Dobrovitski et al., 2008;
Hanson et al., 2008, 2006; Ishikawa et al., 2012). For ex-
ample, an NV- spin in close proximity to several bath spins
10 will experience faster dephasing than an NV- spin many
lattice sites away from the nearest bath spin.
5
Although ensemble T2∗ values are also influenced by spin-
bath fluctuations, as discussed in Secs. III.D and III.F, an
ensemble T2∗ value is not equal to the most common value of
1 T2∗ {single} within the ensemble. For one, the ensemble value
0.5 1 5 10 50 is limited by sources of zero-frequency noise that do not
T2* ( s) contribute to T2∗ {single} , such as spatially inhomogeneous
magnetic fields, electric fields, strain, or g-factors (de Sousa,
2009). These inhomogeneities cause a spatially-dependent
FIG. 9 Sensitivity enhancement scaling with dephasing time T2∗ distribution of the single-NV- resonance line centers, which
for a Ramsey-type magnetometer normalized to the same device broadens the ensemble resonance line and thus degrades
with T2∗ = 500 ns. The different curves assume overhead times T2∗ . Figure 10 depicts broadening contributions to T2∗ from
(tO = tI + tR ) of 1 µs ( ), 10 µs ( ), and 100 µs ( ). The
both varying single-NV- line centers and varying single-NV-
sensitivity enhancement is bounded from above by the fractional
T2∗ improvement ( ) and from below by the square root ( ) linewidths (∝ 1/T2∗ {single} ). The relative contribution to an
of the fractional T2∗ improvement. For simplicity the precession ensemble’s T2∗ value from these two types of broadening is
time τ is set to T2∗ . See Appendix A.2 for details on determining expected to be sample-dependent. Although measurements
the optimal precession time. in Ref. (Ishikawa et al., 2012) on a collection of single NV-
centers in a sparse sample found the distribution of single-
NV- line centers to be narrower than the median single-NV-
B. Ensemble and single-spin T2∗ linewidth, such findings are not expected to hold generally
(e.g., due to strain).
As discussed above, the dephasing time T2∗ is a critical Even in the absence of static field inhomogeneities, the
parameter for broadband DC magnetometry. Importantly, spin-bath-noise-limited T2∗ value of an ensemble is expected
T2∗ is defined differently for a single spin than for a spin to be shorter than the most likely T2∗ {single} value, as the
ensemble. While an ensemble’s T2∗ characterizes relative ensemble value is strongly influenced by the small minority
dephasing of the constituent spins, a single spin’s T2∗ char- of NV- centers with bath spins on nearby lattice sites (Do-
acterizes dephasing of the spin with itself, i.e., the distribu- brovitski et al., 2008). In fact, theoretical calculations in
tion of phase accumulation from repeated measurements on Refs. (Dobrovitski et al., 2008; Hall et al., 2014) reveal
the spin over time (Ishikawa et al., 2012; de Sousa, 2009). that single spins and ensembles interacting with surround-
Since this work focuses on ensemble-based sensing, single- ing spin baths each exhibit free-induction-decay (FID) en-
spin dephasing times are herein denoted T2∗ {single} , while velopes with different functional forms (see Appendices A.5
the term T2∗ is reserved for ensemble dephasing times. and A.7), a result borne out by experiments (Bauch et al.,
2018; Bauch et al., 2019; Maze et al., 2012). In general, the
ensemble T2∗ value cannot be predicted from T2∗ {single} of
any constituent spin (Dobrovitski et al., 2008), and appli-
!"#$%&'(&)(&

" " cation of single-spin measurements or theory to ensembles,

! !
#$% #$% or vice versa, should be done with great care.

!%&*#&)(+ !%&*#&)(+

FIG. 10 Contributions of individual spin resonances to the over-

all spin ensemble lineshape. The ensemble resonance lineshape
( ) is broadened both by the distribution of line centers (left)
and the distribution of linewidths (right) of the constituent spins
( , , ).

Values of T2∗ {single} are affected by slow magnetic, elec-

tric, strain, and temperature fluctuations. Variations in the
magnetic environment may arise from dipolar interactions
 17

C. Dephasing mechanisms remaining defects in the electronic spin bath, such as neg-
atively charged single vacancies (Baranov et al., 2017), va-
The various contributions to an NV- ensemble’s spin de- cancy clusters (Iakoubovskii and Stesmans, 2002; Twitchen
phasing time T2∗ can be expressed schematically as et al., 1999b) and hydrogen-containing defects (Edmonds
et al., 2012).
1 1 1 The quantity T2∗ {nuclear spin bath} in Eqn. 18 describes
≈ ∗ +
T2∗ T2 {electronic spin bath} T2∗ {nuclear spin bath} NV- ensemble dephasing from nuclear spins in the dia-
1 1 mond lattice. In samples with natural isotopic abundance
+ ∗ + of carbon, the dominant contributor to nuclear spin bath
T2 {strain gradients} T2∗ {electric field noise}
1 1 dephasing is the 13 C isotope (I = 1/2), with concentra-
+ ∗ + ∗ tion [13 C] = 10700 ± 800 ppm (Wieser et al., 2013), so
T2 {magnetic field gradients} T2 {temperature variation}
that T2∗ {nuclear spin bath} ≈ T2∗ {13 C} (Balasubramanian
1 1
+ ∗ + , (18) et al., 2009; Dréau et al., 2012; Hall et al., 2014; Zhao et al.,
T2 {unknown} 2T1 2012). Other nuclear spin impurities exist at much lower
concentrations and thus have a negligible effect on dephas-
where the symbol notation T2∗ {X} denotes the hypotheti-
ing. The T2∗ {13 C} scaling with concentration [13 C] is dis-
cal limit to T2∗ solely due to mechanism X (absent all other
cussed in Sec. III.F and can be minimized through isotope
interactions or mechanisms). Equation 18 assumes that all
engineering (Balasubramanian et al., 2009; Teraji et al.,
mechanisms are independent and that associated dephas-
2013).
ing rates add linearly. The second assumption is strictly
Another major source of NV- ensemble dephasing is non-
only valid when all dephasing mechanisms lead to single-
uniform strain across the diamond lattice. Because strain
exponential free-induction-decay envelopes (i.e., Lorentzian
shifts the NV- spin resonances (Dolde et al., 2011; Jamon-
lineshapes); see Appendices A.5, A.6, and A.7. Here we
neau et al., 2016; Trusheim and Englund, 2016), gradients
briefly discuss each of these contributions to NV- ensemble
and other inhomogeneities in strain may dephase the en-
dephasing, and in later sections we examine their scalings,
semble, limiting T2∗ . Strain may vary by more than an
and how each mechanism may be mitigated.
order of magnitude within a diamond sample (Bauch et al.,
The electronic spin bath consists of paramagnetic impu-
2018), and can depend on myriad diamond synthesis pa-
rity defects in the diamond lattice, which couple to NV-
rameters (Gaukroger et al., 2008; Hoa et al., 2014). For a
spins via magnetic dipolar interactions. The inhomoge-
given NV- orientation along any of the [111] diamond crys-
neous spatial distribution and random instantaneous ori-
tal axes, strain couples to the NV- Hamiltonian approxi-
entation of these bath spins cause dephasing of the NV-
mately in the same way as an electric field (though with a
spin ensemble (Bauch et al., 2018; Bauch et al., 2019; Do-
different coupling strength) (Barson et al., 2017; Doherty
brovitski et al., 2008; Hanson et al., 2008). Electronic spin
et al., 2012; Dolde et al., 2011) (see Appendix A.9 for fur-
bath dephasing can be broken down into contributions from
ther discussion). Thus, the quantity T2∗ {strain gradients}
individual constituent defect populations,
may be separated into into terms accounting for strain cou-
1 1 1 pling along (k) and transverse to (⊥) the NV- symmetry
= ∗ 0 + ∗ - axis,
T2∗ {electronic spin bath} T2 {NS } T 2 {NV }
1 1 1 1
+ ∗ + . (19) = ∗
T2 {NV0 } T2∗ {other electronic spins} T2∗ {strain gradients} T2 {straink gradients}
1
Here T2∗ {N0S } denotes the T2∗ limit from dephasing by para- + ∗ . (20)
magnetic substitutional nitrogen defects N0S (S = 1/2), also T2 {strain⊥ gradients}
called P1 centers, with concentration [N0S ] (Cook and Whif- Application of a sufficiently strong bias magnetic field mit-
fen, 1966; Loubser and van Wyk, 1978; Smith et al., 1959). igates the transverse strain contribution to dephasing (Ja-
As substitutional nitrogen is a necessary ingredient for cre- monneau et al., 2016), (see Sec. IV.D), while the longitu-
ation of NV- centers, N0S defects typically persist at con- dinal contribution may be mitigated by employing double-
centrations similar to or exceeding NV- (and NV0 ) con- quantum coherence magnetometry (see Sec. IV.B).
centrations and may account for the majority of electronic Inhomogeneous electric fields also cause NV- ensemble
spin bath dephasing (Bauch et al., 2018). Sec. III.D ex- dephasing (Jamonneau et al., 2016), with associated limit
amines T2∗ {N0S } scaling with [N0S ]. For NV-rich diamonds, T2∗ {electric field noise}. This dephasing source may also
dipolar interactions among NV- spins may also cause de- be broken down into components longitudinal and trans-
phasing of the ensemble, with associated limit T2∗ {NV- }. verse to the NV- symmetry axis, and the contributions can
Sec. III.G examines the T2∗ {NV- } scaling with [NV- ] and be suppressed by the same methods as for strain-related
other experimental parameters. In NV-rich diamonds, the dephasing.
neutral charge state NV0 (S = 1/2) is also present at con- In addition, external magnetic field gradients may cause
centrations similar to [NV- ] (Hartland, 2014) and may also NV- spin dephasing by introducing spatially-varying shifts
contribute to dephasing, with limit T2∗ {NV0 }. The quantity in the NV- energy levels across an ensemble volume, with
T2∗ {other electronic spins} encompasses dephasing from the associated limit T2∗ {magnetic field gradients}. Design of
 18

uniform bias magnetic fields minimizes this contribution to action strength between NV- spins and N0S spins. The in-
NV- ensemble dephasing, and is largely an engineering chal- verse linear scaling of T2∗ {N0S } is supported by both the-
lenge given that modern NMR magnets can exhibit sub-ppb ory (Abragam, 1983a; Bauch et al., 2019; Taylor et al.,
uniformities over their cm-scale sample volumes (Vander- 2008; Wang and Takahashi, 2013; Zhao et al., 2012) and
sypen and Chuang, 2005). experiment (Bauch et al., 2018; Bauch et al., 2019; van
Even though T2∗ is considered the inhomogeneous dephas- Wyk et al., 1997). However, reported values of the scal-
ing time, homogeneous time-varying electric and magnetic ing factor AN0S from theoretical spin-bath simulations vary
fields may appear as dephasing mechanisms if these fields widely; for example, Ref. (Zhao et al., 2012) predicts
fluctuate over the course of multiple interrogation/readout AN0S = 56 ms-1 ppm-1 , whereas Ref. (Wang and Taka-
sequences. Such a scenario could result in the unfortu- hashi, 2013) predicts AN0S = 560 ms-1 ppm-1 , a 10× dis-
nate situation where the measured value of T2∗ depends crepancy. The authors of Ref. (Bauch et al., 2018; Bauch
on the total measurement duration (see Sec. A.2). By et al., 2019) measure T2∗ {N0S } on five samples in the range
the same argument, temperature fluctuations and spatial [N0S ] = 0.75 − 60 ppm (see Fig. 11) and determine AN0S =
gradients can also appear as dephasing mechanisms and
101 ± 12 ms-1 ppm-1 , such that for a sample with [N0S ] =
can limit the measured T2∗ . Temperature variations cause
1 ppm, T2∗ {N0S } = 9.9 ± 1.2 µs. The experimental value of
expansion and contraction of the diamond crystal lattice,
AN0S is consistent with numerical simulations in the same
altering the NV- center’s zero-field splitting parameter D
work (Bauch et al., 2019). The authors calculate the sec-
(dD/dT = −74 kHz/K (Acosta et al., 2010a)) and, depend-
ond moment of the dipolar-broadened single NV- ODMR
ing on experimental design, may also shift the bias magnetic
linewidth (Abragam, 1983a,b) for 104 random spin bath
field. Finally, we include a term in Eqn. 18 for as-of-yet un-
configurations and, by computing the ensemble average over
known mechanisms limiting T2∗ , and we note that T2∗ is lim-
the distribution of single-NV- linewidths (Dobrovitski et al.,
ited to a theoretical maximum value of 2T1 (Levitt, 2008;
2008), find good agreement with the experimental value
Myers et al., 2017).
AN0S = 101 ms-1 ppm-1 .
Importantly, Eqn. 18 shows that the value of T2∗ is pri-
marily set by the dominant dephasing mechanism. There- Electron paramagnetic resonance (EPR) measurements
fore, when seeking to extend T2∗ , one should focus on reduc- of nitrogen N0S defects in diamond (van Wyk et al., 1997)
ing whichever mechanism is dominant until another mech- from 63 samples also confirm the scaling 1/T2∗ ∝ [N0S ] (see
anism becomes limiting. Reference (O’Keeffe et al., 2019) Appendix A.6 and Fig. 35) and the approximate scaling
aptly expresses the proper strategy as a “shoot the alliga- constant AN0S . With the likely assumption that the dephas-
tor closest to the boat” approach. For example, even if the ing time for ensembles of substitutional nitrogen spins in a
dephasing due to substitutional nitrogen is substantially nitrogen spin bath can approximate T2∗ {N0S } for NV- ensem-
decreased in a particular experiment, the improvement in bles (Dale, 2015) (see Appendix A.6), the measurements in
T2∗ may be much smaller if, say, strain inhomogeneity then Ref. (van Wyk et al., 1997) suggest AN0S ≈ 130 ms-1 ppm-1 ,
becomes a limiting factor; at that point it becomes more which is in good agreement with the measured AN0S =
fruitful to shift focus towards reducing strain-induced de- 101 ± 12 ms-1 ppm-1 from Ref. (Bauch et al., 2019) (see
phasing. Appendices A.5 and A.6).
In addition, the data in Ref. (van Wyk et al., 1997) sug-
gest that dipolar dephasing contributions from 13 C at nat-
D. Nitrogen limit to T2∗ ural isotopic abundance [10700 ppm (Wieser et al., 2013)]
and from substitutional nitrogen are equal for [N0S ] =
In nitrogen-rich diamonds, the majority of electronic 10.8 ppm. The measured values of AN0S (Bauch et al., 2018)
spins contributing to the spin bath originate from substitu- and A13 C (see Sec. III.F) for NV- ensembles predict the two
tional nitrogen defects, since N0S may donate its unpaired contributions to be equal at N0S = 10.3 ppm, which is con-
electron to another defect X and become spinless N+ S , via sistent to within experimental uncertainty.
the process (Khan et al., 2009), In Appendix A.4, we present a simple toy model (Klein-
sasser et al., 2016) for the case when nitrogen-related de-
N0S + X0 ↔ N+ -
S +X . (21) fects dominate T2∗ . In this regime, under the assumption
that the conversion efficiency of total nitrogen to NV- ,
In these samples, the electronic spin concentration is closely
NV0 , and N+ is independent of the total nitrogen concen-
tied to the total concentration of substitutional nitrogen
∗ tration [NT ], the dephasing time T2∗ scales inverse-linearly
donors [NT S ], and thus T2 {electronic spin bath} is primar-
T with [NT ], while the number of collected photons N scales
ily set by [NS ]. In unirradiated nitrogen-rich diamonds,
linearly with [NT ]. These p scalings result in a shot-noise-
however, N0S serves as the primary contributor to the elec-
limited sensitivity η ∝ 1/ N · T2∗ , which is independent of
tronic spin bath (Bauch et al., 2018). The N0S contribution
to dephasing obeys [NT ]. However, as discussed in Sec. II.C and Appendix A.3,
technical considerations favor lower nitrogen concentrations
1 [NT ], which result in lower photon numbers N and longer
= AN0S [N0S ] (22)
T2∗ {N0S } dephasing times T2∗ (Kleinsasser et al., 2016).

where [N0S ] is the concentration of neutral substitutional

nitrogen, and AN0S characterizes the magnetic dipole inter-
 19

E. Nitrogen limit to T2 the second interval (after the π-pulse). Consequently, the
characteristic decay time of the NV- spin state measured
through Hahn echo, denoted by T2 (the coherence time),
is substantially longer than the inhomogeneous dephasing
a time T2∗ , typically exceeding the latter by one to two orders
103 of magnitude (Bauch et al., 2019; de Lange et al., 2010).
By design the Hahn echo sequence and its numerous exten-
sions (Gullion et al., 1990; Meiboom and Gill, 1958; Wang
100 et al., 2012) restrict sensing to AC signals, typically within
a narrow bandwidth, preventing their application in DC
T2* ( s)

sensing experiments. Nonetheless, the Hahn echo T2 plays

10 a crucial role in diamond sample characterization and for
9.
9
AC sensing protocols (Sec. IV.A) and merits brief discus-
±
1.
2
sion here.
1
s·p
nat. ab. 13 C samples Like T2∗ , T2 depends on the nitrogen concentration [N0S ],
pm
99.95 %+ 12 C samples which sets both the average dipolar-coupling strength be-
fit + 95% C.I.
tween NV- and nitrogen bath spins (i.e., AN0s [N0s ] from
0.1 Eqn. 22), as well as the average coupling strength between
10-3 0.01 0.1 1 10 100 103
nitrogen bath spins (Bar-Gill et al., 2012; de Sousa, 2009).
Nitrogen concentration (ppm) Furthermore, it can be shown that when nitrogen is the
b dominant decoherence mechanism, T2 {N0S } depends inverse
103
linearly on the nitrogen concentration [N0S ] (Bauch et al.,
2019), revealing a close relationship to T2∗ {N0S }. The depen-
dence of T2 {N0S } on [N0S ] was recently determined experi-
mentally through NV- ensemble measurements on 25 dia-
16

100
0

mond samples (see Fig. 11b), yielding (Bauch et al., 2019)

±
12
T2 ( s)

1
s·

= BN0S [N0S ].
pp

10 (23)
T2 {N0S }
m

Here, BN0S = 6.25 ± 0.47 ms-1 ppm-1 , such that an NV- en-
1 nat. ab. 13 C samples semble in a 1-ppm-nitrogen sample is expected to exhibit
99.95 %+ 12 C samples
fit + 95% C.I. T2 ' 160±12 µs. The scaling in Eqn. 23 should also be com-
pared to that of T2∗ {N0S } (Eqn. 22), with T2 {N0S }/T2∗ {N0S } =
0.1
10-3 0.01 0.1 1 10 100 103 BN0S /AN0S ≈ 17. A straightforward application of these
Nitrogen concentration (ppm)
results is the calibration of the total nitrogen spin con-
centration in diamond samples through T2∗ measurements,
FIG. 11 Substitutional nitrogen spin bath contribution to
T2 measurements, or both, provided that nitrogen remains
ensemble-NV- dephasing time T2∗ and coherence time T2 . a) the primary source of dephasing and decoherence in such
Measured spin-bath contribution to T2∗ vs. nitrogen concentra- samples. Here, T2 measurements are advantageous over
tion measured by secondary ion mass spectrometry (SIMS) for T2∗ (or linewidth) measurement schemes, as the latter are
five diamond samples. Fit yields 1/T2∗ {N0S } = AN0 [N0S ] with more likely to be limited by non-nitrogen dephasing mech-
S
AN0 = 101±12 ms-1 ppm-1 . b) Measured Hahn echo T2 vs. nitro- anisms (Bauch et al., 2018).
S
gen concentration for 25 diamond samples. The linear contribu- Lastly, we note that the inverse linear scaling of T2∗ {N0S }
tion to the fit is attributed to substitutional nitrogen and yields and T2 {N0S } with [N0S ], as well as the hierarchy T2  T2∗ , are
1/T2 {N0S } = BN0 [N0S ] with BN0 [N0S ] = 6.25 ± 0.47 ms-1 ppm-1 . consistent with earlier EPR studies of N0S nitrogen defects
S S
The nitrogen-independent contribution to the fit is given by in nitrogen-rich diamonds (Stepanov and Takahashi, 2016;
T2 {other} = 694 ± 82 µs. Adapted from Ref. (Bauch et al.,
van Wyk et al., 1997) and other comparable spin systems
2019).
in silicon (Abe et al., 2010; Witzel et al., 2010). Predicting
the values of BN0S and AN0S for NV- based on equivalent
Contributions to the NV- spin dephasing time T2∗ from EPR scaling parameters measured with P1 centers in dia-
static and slowly-varying inhomogeneities are largely miti- mond (van Wyk et al., 1997) is expected to be crudely ef-
gated by employing a π/2 − π − π/2 Hahn echo pulse se- fective. However, accuracy at the 10% level or better likely
quence (see Sec. IV.A). In contrast to a π/2 − π/2 Ramsey requires accounting for various experimental specifics [e.g.
sequence (see Appendix A.1.a), the added π-pulse reverses the magnetic field value (Hall et al., 2014)].
the precession direction of the sensor spins halfway through
the free precession interval. As a result, any net phase accu-
mulated by the NV- spin state due to a static magnetic field
vanishes, as the accumulated phase during the first interval
(before the π-pulse) cancels the accumulated phase during
 20

F. 13
C limit to T2∗ a 0.5 Natural Abundance Sample

Contrast (%)
- 13
Dipolar coupling between NV electronic and C nuclear
spins can also limit T2∗ (Balasubramanian et al., 2009; Do-
0.0
brovitski et al., 2008; Hall et al., 2014; Zhao et al., 2012). T2,* DQ = 0. 445(30) s
Reducing the 13 C content below the natural abundance con- 0.5
centration [13 C] = 10700 ± 800 ppm ≈ 1.1% (Wieser et al., 0.0 0.3 0.6 0.9 1.2 1.5
Free Precession Time ( s)
2013) through isotope engineering is the most direct way to b
1.0

FFT (arb. units)

mitigate this effect (Balasubramanian et al., 2009; Bauch
et al., 2018; Itoh and Watanabe, 2014). In the “dilute”
spin limit where [13 C]/[12 C]  0.01 (Kittel and Abrahams, 0.5
1953), the 13 C dephasing contribution is well-approximated
by
0.0
1 0 3 6 9 12 15
= A13 C [13 C], (24) Frequency (MHz)
T2∗ {13 C}

where A13 C is a constant characterizing the magnetic dipole FIG. 12 T2∗ measurement of a low-nitrogen-content diamond
interaction strength between NV- spins and 13 C nuclear with natural abundance [13 C] = 10700 ppm to assess the 13 C
contribution to dephasing. a) Double-quantum Ramsey free in-
spins, in accordance with theoretical predictions (Abragam,
duction decay (FID) (•) and associated fit ( ) suggest T2∗ is
1983c; Dobrovitski et al., 2008; Hall et al., 2014; Kittel and 445 ns in the double-quantum basis. This data sets a bound
Abrahams, 1953) (see Appendix A.8). Although experi- A13 C < 0.105 ms-1 ppm-1 . Correcting for the test diamond’s
mental measurements relating T2∗ to [13 C] are only avail- approximately known [N0S ] ≈ 0.5 ppm content allows further re-
able for single NV- centers (Balasubramanian et al., 2009; finement to A13 C ≈ 0.100 ms-1 ppm-1 . b) Fourier transform of
Mizuochi et al., 2009) and not for NV- ensembles, the scal- the FID shown in the top panel. The three peaks arise from
ing in Eqn. 24 is consistent with experimental findings hyperfine interactions associated with the NV- center’s 14 N nu-
clear spin I = 1 and exhibit intra-peak spacing double that of
in a similar ensemble spin system: EPR linewidth mea-
an equivalent single-quantum Ramsey measurement. The unbal-
surements on substitutional phosphorus spin ensembles in anced peak heights are attributed to nuclear spin polarization
a 28 Si crystal exhibit the same scaling for various dilute induced by the 150 gauss bias magnetic field.
concentrations of 29 Si (Abe et al., 2010; Morishita et al.,
2011). Figure 4b in Ref. (Abe et al., 2010) suggests that
∗{single,mp}
Eqn. 24 is approximately valid for [29 Si]/[18 Si] . 0.05, so it to be T2 = 1.8 ± 0.6 µs [measured in a 20 G
is plausible that A13 C can be inferred from measurements bias field, Fig. 4a in Ref. (Maze et al., 2012)]. From re-
on diamonds with natural 13 C isotopic abundance where lations in Ref. (Hall et al., 2014) we estimate A13 C in terms
[13 C]/[12 C] ≈ 0.0107. We make this assumption henceforth. of the coupling constant Asingle
13 C for a single NV- , A13 C ≈
While the value of A13 C is not known precisely for NV- single,mp
2.2 A13 C , which yields A13 C = 0.11 ± 0.04 ms-1 ppm-1 .
ensembles, T2∗ measurements in diamond with natural 13 C Our measured value A13 C ≈ 0.100 ms-1 ppm-1 is also in
abundance set an approximate upper bound on A13 C , since reasonable agreement with first-principles theoretical cal-
necessarily 1/T2∗ > 1/T2∗ {13 C}. Figure 12 shows a Ram- culations by Ref. (Hall et al., 2014), suggesting A13 C ≈
sey FID for a diamond with natural 13 C abundance and 0.057 ms-1 ppm-1 for NV- ensembles in natural isotopic dia-
low nitrogen concentration; these data suggest A13 C ≈ mond in tens-of-gauss bias fields. Note that the experimen-
0.100 ms-1 ppm-1 . With this value for A13 C , the expected tal determination of A13 C outlined in this section represents
limit for a 99.999% 12 C isotopically-enriched diamond is an upper bound on the true value of A13 C in the dilute
T2∗ {13 C} ≈ 1 ms, at which point dipolar interaction with (dipolar-broadened) limit; if substantial broadening arises
13
C nuclear spins is unlikely to be the leading-order de- from Fermi-contact contributions in addition to dipolar in-
phasing mechanism (see Eqn. 18). Comparing A13 C with teractions in natural abundance 13 C samples, or if [13 C] =
the measured AN0S = 101 ms-1 ppm-1 for dephasing of NV- 10700 ppm does not qualify as the dilute limit (Abragam,
ensembles by substitutional nitrogen (see section III.D), de- 1983c; Kittel and Abrahams, 1953), the value of A13 C given
phasing from natural abundance [13 C] = 10700 ppm and here will be overestimated.
substitutional nitrogen with concentration [N0S ] = 10.6 ppm Engineering diamonds for low 13 C content may be chal-
should be equivalent, in good agreement with Ref. (van lenging (Dwyer et al., 2013; Markham et al., 2011; Teraji
Wyk et al., 1997), which observes equivalence for [N0S ] et al., 2013). The isotopic purity of a diamond grown by
≈ 10.8 ppm. Conveniently, it is easy to remember that plasma-enhanced chemical vapor deposition (PE-CVD) is
T2∗ {13 C} is 1 µs for natural abundance 13 C diamond to expected to be limited by the purity of the carbon source
better than 10%. gas, which is most commonly methane (CH4 ). However,
The bound on A13 C derived above can be crudely con- diamonds grown with isotopically-enriched methane may
firmed using a mix of theoretical predictions from Ref. (Hall exhibit higher fractional 13 C content than the source gas
et al., 2014) and data from Ref. (Maze et al., 2012). The due to extraneous carbon sources in the CVD chamber
authors of Ref. (Maze et al., 2012) find the most proba- (Dwyer et al., 2013). Nonetheless, Teraji et al. achieve
ble T2∗ for a single NV- center in natural isotopic diamond
 21

[12 C] = 99.998% as measured by secondary ion mass spec- initialization of NV- centers (Doherty et al., 2013). For
trometry (SIMS) when using isotopically-enriched methane example, off-resonant NV- populations polarized into the
with 99.999% 12 C (i.e., [13 C] ≤ 10 ppm) (Teraji et al., 2013, spinless ms = 0 state during initialization should not con-
2015). Although such isotopically-enriched methane is cur- tribute to dephasing of the NV- centers used for sensing,
rently 103 - 104 times more expensive than natural abun- giving ς∦ ' 0.
dance CH4 , order unity conversion of the methane’s carbon Flip-flop interactions between NV- spins in different
content into diamond is attainable (Teraji et al., 2013). groups are off-resonant and are thus suppressed, whereas
flip-flop interactions can occur resonantly between spins in
the same group. The extra resonant interaction terms in the
G. NV- limit to T2∗ dipole-dipole Hamiltonian for spins in the same group re-
sult in a slightly increased dephasing rate (Abragam, 1983a;
Kucsko et al., 2018). Following Ref. (Abragam, 1983a), it
is expected that ANVk = 3/2 ANV∦ .
1.00
The lack of published data at present for T2∗ {NV- } in
Normalized PL signal (arb)

samples with varying NV- concentration prevents definitive

determination of ANVk and ANV∦ . However, both terms can
0.98 be estimated from the experimentally determined value of
AN0S = 101 ± 12 ms-1 ppm-1 (Bauch et al., 2019), which de-
scribes the scaling of NV- ensemble T2∗ with substitutional
0.96
nitrogen concentration (see Sec. III.D). Assuming an NV-
electronic spin bath couples to NV- sensor spins with ap-
0.94
proximately the same strength as a substitutional nitrogen
spin bath (Hanson et al., 2008), and accounting for the
2.80 2.85 2.90 2.95
higher spin multiplicity of NV- centers (SNV- = 1) com-
Frequency (GHz) pared to substitutional nitrogen spins (SN0S = 1/2), we cal-
culate (Abragam, 1983a)
FIG. 13 ODMR spectra from the same NV- ensemble in differ-
ent applied bias magnetic fields. A bias field with a different s
projection on each of the four NV- crystallographic orientations SNV- (SNV- + 1)
separates the ms = 0 ↔ ms = ±1 spin resonances into dis- ANV-∦ ' A 0
tinct groups ( ). A bias field that projects equally onto all four
SN0S (SN0S + 1) NS
orientations overlaps the spin resonances ( , offset for clarity).
p (26)
' 8/3AN0S
' 165 ms-1 ppm-1
Dipolar interactions among negatively-charged NV- cen-
ters may also limit the dephasing time T2∗ . Dephasing from
and find ANV-k = 3/2 ANV-∦ ' 247 ms-1 ppm-1 . From this ar-
NV- -NV- interactions arises from NV- spins in both the
same and different groups as the NV- centers used for sens- gument, although the precise value of T2∗ {NV- } depends on
ing, where groups are defined as follows: NV- centers with experimental conditions including optical initialization frac-
approximately the same spin resonance frequency are con- tion (determining ςk and ς∦ ) and bias magnetic field orienta-
sidered to be in the same group, whereas spins with differ- tion (setting the ratio [NV-k ]/[NV-∦ ]), the value of T2∗ {NV- }
ent resonance frequencies are in different groups (Kucsko for a given NV- concentration is well approximated (up to
et al., 2018). Depending on the strength and angle of the a factor of order unity) by the dephasing time T2∗ {N0S } for
applied bias field, the spin resonances of the four NV- orien- an identical concentration of N0S spins.
tations may be spectrally separated, or two or more may be Magnetometer operation in the NV- -NV- interaction
overlapped, changing the fraction of NV- spins in the same limit may occur as the N-to-NV- conversion efficiency Econv
group (see Fig. 13). The NV- -NV- dipolar contribution to approaches its theoretical limit of 50% (Felton et al., 2009;
T2∗ is then given by Pham et al., 2011) (see Sec. VI.A). Under these circum-
stances, and when other sources of dephasing can be ne-
glected [e.g., magnetic, electric, and strain gradients as well
1 1 1 as 13 C nuclear spins (see Sec. III.F) and other paramag-
= ∗ +
T2∗ {NV- } T2 {NV- } k T2∗ {NV- } ∦ (25) netic defects (see Sec. VI.F)], the interaction among NV-
= ςk A NV-k [NV-k ] + ς∦ A NV-∦ [NV-∦ ]. spins becomes the dominant source of dephasing. However,
maximal N-to-NV- conversion efficiency is not necessarily
Here [NV-k ] is the concentration of NV- spins in the group required to operate in the NV- -NV- interaction limit. When
being used for sensing and [NV-∦ ] is the concentration of Econv < 50%, dephasing due to other paramagnetic impu-
NV- spins in other groups, with [NV-k ] + [NV-∦ ] = [NV-total ]; rities may be reduced through spin bath driving techniques
the constants ANV-k and ANV-∦ characterize the dipolar in- described in Sec. IV.C. Spin bath driving can also decouple
teraction strength for pairs of NV- spins in the same group the NV- centers in different groups from the NV- centers in
and different groups respectively; and ςk and ς∦ are dimen- the group used for sensing (Bauch et al., 2018), suppressing
sionless factors of order unity accounting for (imperfect) the second term in Eqn. 25 (1/T2∗ {NV- }∦ ) and leaving only
 22

the first term (1/T2∗ {NV- }k ) as a fundamental limit to NV- / /

ensemble T2∗ . / x
x x
pump read-out
While this section has focused on the negatively charged Hahn echo
NV- center, NV centers are also present in the neutral
/ /
charge state NV0 (S = 1/2) (see Sec. VI.A). NV0 has not x
/ y / y x
been observed in its ground state in EPR, a phenomenon CPMG-2
tentatively attributed to resonance line broadening from
dynamic Jahn-Teller distortion (Felton et al., 2008). Sim- / / /
/ y / y / y / y
ilarly, magnetic noise created by NV0 may be reduced by x … x

this motional-narrowing-type effect to less than otherwise CPMG-n …

expected for a S = 1/2 defect (see Sec. IV.C). Consequently,

the contribution of NV0 spins to dephasing of NV- spins
may be smaller than expected. More recent work, however, FIG. 14 Select pulse sequences for AC magnetometry. The
suggests an alternative hypothesis: the NV0 ground state Hahn echo sequence includes a refocusing π-pulse midway
is not be visible in EPR due to large strain effects (Barson through the interrogation time, allowing phase-sensitive lock-in-
et al., 2019). What little, if any, NV+ is present in the type measurements of AC magnetic fields (top). Hahn echo is
maximally sensitive to AC fields with nodes coincident with the
sample is expected to be spinless (see Table 9) and should three MW pulses. Detection of AC fields with the quadrature
not contribute substantially to dephasing. phase can be achieved using the Carr-Purcell-Meiboom-Gill-2
Recently, several protocols have been proposed to mit- (CPMG-2) sequence (middle). Employing additional π-pulses
igate strong NV- -NV- dipolar interactions and extend (CPMG-n) achieves more efficient decoupling of the NV- from
T2∗ (O’Keeffe et al., 2019) or T2 (Choi et al., 2017b) while substitutional nitrogen and other paramagnetic defects in the
retaining magnetic field sensitivity. In addition, it has been diamond and provides sensitivity to higher-frequency AC mag-
proposed that under certain circumstances the NV- -NV- netic fields (bottom). From Ref. (Pham et al., 2012a).
dipolar interaction could enhance magnetometry sensitiv-
ity through enabling entanglement of multiple NV- cen-
the pulse sequence. In the ideal case where the three MW
ters (Choi et al., 2017) (see Sec. VII.D). Harnessing entan- pulses are commensurate with consecutive nodes of the AC
glement could enable superior scaling of measurement SNR magnetic field, the sensitivity of a Hahn-echo-based mea-
with number of spins addressed
√ N , exceeding the standard surement limited by shot noise and spin projection noise is
quantum limit SNR ∝ N and approaching the Heisen- given by
berg limit, SNR ∝ N . Controlled coupling of NV- spin ensemble
ηecho ≈ (27)
pairs (Bernien et al., 2013; Dolde et al., 2013; Hensen et al., s r
2015; Jakobi et al., 2016; Neumann et al., 2010b; Yamamoto π ~ 1 1 1 tI +τ +tR
√ −(τ /T2 )p
1+
et al., 2013a) has been demonstrated; however, applying 2 ∆ms ge µB N τ e
| {z } C 2 navg τ
entanglement-enhanced techniques to larger ensembles is } |
| {z } | {z {z }
Spin projection limit Spin decoherence Overhead time
Readout
expected to be challenging.
where ∆ms is the spin projection quantum number differ-
ence between the two states in the interferometry measure-
IV. METHODS TO EXTEND T2∗ AND T2 ment, N is the number of interrogated spins, τ is the full
field interrogation time, p is a stretched exponential param-
A. Dynamical decoupling for AC magnetometry
eter set by ensemble averaging of the local NV- spin environ-
ments (see Appendix A.7), C is the measurement contrast
While this review primarily addresses the broadband DC
prior to precession (see Sec. II.A), navg is the average num-
sensing modality of ensemble-NV- magnetometers, many
ber of photons detected per NV- center per readout, and
of the sensitivity-improvement techniques described herein
tI and tR are the optical initialization and readout times,
can also be applied to detecting narrowband AC magnetic
respectively. For more realistic measurements in which the
fields. Here we provide a brief overview of standard AC
pulse sequence cannot be phase-locked to the √ AC magnetic
sensing schemes; we discuss several approaches to improv-
field, the magnetic sensitivity is degraded by 2.
ing AC magnetic field sensitivity; and we highlight chal-
The shot-noise-limited sensitivity given by Eqn. 27 has
lenges unique to the AC sensing modality.
several key differences to that of a Ramsey-based DC sens-
The Hahn echo (alternatively referred to as the spin echo)
protocol, shown in Fig. 14, builds upon the Ramsey proto- ing protocol (Eqn. 15). First, since typically T2  T2∗ ,
col with an additional central MW π-pulse, which refocuses AC sensing schemes can achieve better sensitivity than DC
dephasing of the NV- spin ensemble (Hahn, 1950). The de- sensing schemes. Second, choice of spin interrogation time
cay of spin coherence measured with this pulse sequence is τ is more straightforward for Ramsey schemes than for
characterized by T2 , which is typically one to two orders echo-based schemes. For Ramsey-based sensing of DC or
of magnitude longer than T2∗ in NV- ensemble measure- quasi-static fields, τ ∼ T2∗ is optimal (Appendix A.2). In
ments (see Secs. III.C, III.D, and III.E). Furthermore,
contrast, while τ ∼ T2 is optimal for Hahn-echo-based pro-
while the refocusing pulse decouples the NV- spin from DC
magnetic fields, its presence makes Hahn echo measure- tocols, τ should also be matched to the period TB of the
ments particularly sensitive to oscillating magnetic fields AC magnetic field to be measured. As a result the scheme
with period TB matching the spin interrogation time τ of is maximally sensitive to fields of period TB ∼ T2 , with a
 23

detection bandwidth set by the relevant filter function (Cy- pulses k (Pham et al., 2012a), where s is set by the noise
wiński et al., 2008). Finally, coherent interactions between spectrum of the decohering spin bath and is typically sub-
the NV- spin and other spin impurities in the diamond can linear. For example, a bath of electronic spins, such as N0S
modulate the Hahn-echo coherence envelope. At best, these defects in diamond, exhibits a Lorentzian noise spectrum
and results in a power-law scaling of the coherence time
effects introduce collapses and revivals that do not affect T2 with s = 2/3, assuming the electronic spin bath is the domi-
and merely complicate the NV- magnetometer’s ability to nant decoherence source (de Sousa, 2009). The multi-pulse
measure AC magnetic fields of arbitrary frequency. When AC magnetic field sensitivity limited by shot noise and spin
the bias magnetic field is aligned to the NV- internuclear projection noise is given by
axis in diamond samples containing a natural abundance of ensemble
13 ηmulti ≈ (28)
C, collapse-and-revival oscillations occur with frequency s
set by the 13 C Larmor precession. At worst, misalignment
r
π ~ 1 1 1 tI +τ +tR
√ 1+ 2
between the bias magnetic field and the NV- internuclear 2 ∆ms ge µB N τ |e−[τ /(k
{z
s T )]p
2
} C n avg τ
axis results in anisotropic hyperfine interactions, which en- | {z } | {z } | {z }
Spin decoherence
Spin projection limit Overhead time
Readout
hance the nuclear-spin Larmor precession rate for 13 C (and
15
N in 15 NV-diamonds) as a function of separation between with an optimal number of pulses
the nuclear spins and NV- centers (Childress et al., 2006; 1
  p  p(1−s)
Maurer et al., 2010). These effects ensemble average to an 1 2T2
effectively shorter coherence time T2 (Stanwix et al., 2010), kopt = , (29)
2p(1 − s) TB
which degrades AC sensitivity.
Despite these differences, the Ramsey and spin-echo mea- for an AC magnetic field with period TB , assuming full in-
surement schemes share many of the same components; terrogation time τ = k2 TB and π-pulses commensurate with
consequently, many techniques for improving spin readout the nodes of the oscillating√ magnetic field. As before, the
fidelity (analyzed in Sec. V) apply to both DC and AC sensitivity is degraded by 2 when measuring AC magnetic
sensing modalities. For example, ancilla-assisted repeti- fields with unknown phase.
tive readout (Sec. V.C), level-anticrossing-assisted readout Equations 28 and 29 illustrate that multi-pulse measure-
(Sec. V.D), and improved fluorescence collection methods ment schemes improve sensitivity to magnetic fields with
(Sec. V.E) increase the number of detected photons per periods TB < T2 and enable sensing of higher frequencies
measurement N; preferential NV- orientation (Sec. VI.G) than can be accessed with Hahn-echo-based measurements.
enhances the measurement contrast C; and spin-to-charge- For example, the authors of Ref. (Pham et al., 2012a)
conversion (SCC) readout (Sec. V.A) and NV- charge state demonstrate a 10× improvement (compared to Hahn echo)
optimization (Sec. VI.B) increase both C and N. We in ensemble AC sensitivity at 220 kHz by using a multi-
note that because typically T2  T2∗ , advanced read- pulse sequence. However, the increased number of control
out techniques such as repetitive readout and SCC read- pulses, which are typically imperfect due to NV- hyper-
out presently offer greater sensitivity improvement for AC fine structure and inhomogeneities in the system, can re-
schemes than for DC schemes, as their long-readout-time sult in cumulative pulse error and thus degraded AC sensi-
requirements introduce smaller fractional overhead in AC tivity (Wang et al., 2012). Compensating pulse sequences,
measurements with longer interrogation times. including schemes in the XY, concatenated, and BB-n fam-
Additionally, techniques to extend T2∗ for DC and broad- ilies, may be employed to restore AC field sensitivity in
band magnetometry may also improve AC magnetic field the presence of pulse errors (Farfurnik et al., 2015; Gullion
sensitivity. For example, double-quantum (DQ) coherence et al., 1990; Low et al., 2014; Rong et al., 2015; Wang et al.,
magnetometry (Sec. IV.B) is expected to improve AC sen- 2012).
sitivity both by introducing a 2× increase in the NV- spin AC sensing techniques are also pertinent to noise spec-
precession rate (Fang et al., 2013; Mamin et al., 2014) troscopy. By mapping out a diamond’s spin bath spectral
and, in certain cases, by extending the NV- coherence time noise profile, tailored sensing protocols can be designed to
T2 (Angerer et al., 2015). Similarly, spin bath driving more efficiently extract target signals. To this end, dy-
(Sec. IV.C) and operation at a sufficiently strong bias mag- namical decoupling sequences, such as those in the CPMG
netic field (Sec. IV.D) may extend T2 by suppressing mag- and XY families, are employed for noise mapping (Bar-Gill
netic and electric/strain noise, respectively. et al., 2012, 2013; Bauch et al., 2019; Chrostoski et al.,
Another technique for enhancing NV- magnetic sensitiv- 2018; Romach et al., 2015). By varying both the total pre-
ity, unique to the AC sensing modality, is the application of cession time and the number of refocusing pulses, noise at
multi-pulse sequences, whose timing is based on the Carr- a variable target frequency can be isolated, ultimately al-
Purcell-Meiboom-Gill (CPMG) family of pulse sequences lowing measurement of the entire spin bath spectral noise
well-known in NMR (Cywiński et al., 2008; Pham, 2013)
(see Fig. 14). By applying additional MW π-pulses at a profile. However, such measurements are often complicated
rate of 2T1B , these multi-pulse sequences (i) extend the NV- by non-idealities in certain sequences’ filter functions, such
coherence time T2 by more effectively decoupling the NV- as sensitivity to harmonics or the presence of sidelobes (Cy-
spins from magnetic noise and (ii) increase the time dur- wiński et al., 2008). Recently, AC magnetometry protocols
ing which the NV- spins interrogate the AC magnetic field. with enhanced spectral resolution have been demonstrated,
The coherence time has been found to scale with a power such as the dynamic sensitivity control (DYSCO) sequence
(k)
law s (T2 → T2 = T2 k s ) as a function of the number of and its variants (Lazariev et al., 2017; Romach et al., 2019),
which provide simpler, single-peaked filter functions at the
 24

cost of reduced sensitivity. Additional dynamical decou- Use of the full spin-1 nature of the NV- center and the
pling sequences with increased spectral resolution or other double-quantum basis {|−1i, |+1i} allows for several sens-
advantages have been employed (Boss et al., 2017; Glenn ing advantages. First, at fixed magnetic field, an NV-
et al., 2018; Hernández-Gómez et al., 2018; Schmitt et al., spin prepared in a superposition of the | + 1i and | − 1i
2017) or proposed (Cywiński et al., 2008; Poggiali et al., states precesses at twice the rate as in the standard SQ
2018; Zhao et al., 2014). subspace of {|0i, | − 1i} or {|0i, | + 1i}, enabling enhanced
A final consideration in the application of multi-pulse magnetometer sensitivity. Moreover, measurements in the
sequences for enhancing AC magnetometry with NV- cen- DQ basis are differential, in that noise sources perturbing
ters is that extension of the T2 coherence time (and thus the |0i ↔ | + 1i and |0i ↔ | − 1i transitions in common-
enhancement of AC magnetic field sensitivity) is eventu- mode are effectively rejected. Sources of common-mode
ally limited by the T1 spin-lattice relaxation time, beyond noise may include temperature fluctuations, which enter
which increasing the number of π-pulses is ineffective. This the NV- Hamiltonian via the zero-field splitting parameter
limitation can be overcome by reducing the magnetome- D ( ∂D
∂T ≈ −74 kHz/K) (Acosta et al., 2010a; Kucsko et al.,
ter operating temperature, thereby suppressing the two- 2013; Toyli et al., 2013); axial strain gradients; axial elec-
phonon Raman process that dominates NV- spin-lattice re- tric fields; and transverse magnetic fields. For a detailed
laxation near room temperature and extending T1 (Jarmola discussion see Ref. (Bauch et al., 2018).
et al., 2012). Multi-pulse sequences performed at 77 K have If the spin bath environment is dominated by magnetic
demonstrated > 100× extensions in T2 compared to room noise, as is common for high-nitrogen and natural 13 C abun-
temperature measurements (Bar-Gill et al., 2013), and cor- dance diamond samples, measurements in the DQ basis ex-
responding improvements to AC magnetic field sensitivity hibit an increased linewidth and shortened associated de-
are expected. phasing time, as the 2× enhanced sensitivity to magnetic
Although the limit T2 , T2∗ ≤ 2T1 is well established fields causes the spin ensemble to dephase twice as quickly
∗ ∗
theoretically (Slichter, 1990; Yafet, 1963) and observed in as in the SQ basis, i.e., T2,DQ ≈ T2,SQ /2. This increased
other spin systems (Bylander et al., 2011), the maximum T2 dephasing and decoherence is confirmed experimentally for
values achieved in NV-diamond through dynamical decou- single NV- centers by the authors of Ref. (Fang et al., 2013),
pling protocols have historically never exceeded measured who observe a 2× decrease in T2∗ , and by the authors of
values of T1 , with Ref. (Bar-Gill et al., 2013) achieving Ref. (Mamin et al., 2014), who observe an ≈ 2× decrease
T2 = 0.53(2)T1 for NVs in bulk diamond and Refs. (Myers in the Hahn-echo coherence time T2 . Similar results are re-
et al., 2014; Romach et al., 2015) observing T2 . 0.1T1 for ported for NV- ensembles (Bauch et al., 2018; Kucsko et al.,
shallow NVs. While this discrepancy is not fully resolved, it 2018).
is partially accounted for by the observation that the typical In the SQ basis, non-magnetic noise sources such as tem-
measurement protocol for T1 [e.g., that described in (Pham, perature fluctuations, electric field noise, and inhomoge-
2013)] yields a T1 value that does not encompass all pos- neous strain may also contribute to spin dephasing (see
sible decays of the full spin-1 system but rather only the Sec. III.C). However, values of T2∗ in the DQ basis are in-
(0)
decay T1 in the pseudo-spin-1/2 subspace of |0i and either sensitive to these common-mode noise sources. When such
|+1i or |−1i (Myers et al., 2017). The value of T1 for the noise dominates dephasing in the SQ basis, the DQ dephas-
(0)
full S = 1 system is typically shorter than T1 thanks to ing time T2,DQ may exceed T2,SQ , allowing for additional
non-negligible decay from |+1i to |−1i and vice versa. This sensitivity improvement. For example, DQ measurements
spin-1 T1 , which can be measured using methods described reported in Ref. (Bauch et al., 2018) on NV- ensembles
in Ref. (Myers et al., 2017), is the relevant relaxation time demonstrate a ∼ 6× increase in T2∗ (narrowing of linewidth)
limiting T2 and T2∗ . in an isotopically purified, low-nitrogen diamond, leading to
an effective 13× enhancement in phase accumulation per
measurement when considering the twice faster precession
B. Double-quantum coherence magnetometry rate in the DQ basis. In Ref. (Bauch et al., 2018), the stan-
dard SQ basis T2∗ is found to be limited by strain inhomo-
Standard NV- magnetometry techniques, such as CW- geneities, whereas the T2∗ value measured in the DQ basis
ODMR (Sec. II.B.1), pulsed ODMR (Sec. II.B.2), and is likely primarily limited by interactions with residual 13 C
pulsed Ramsey or echo-type schemes (Sec. IV.A), are typi- nuclear spins (∼ 100 ppm). This T2∗ limitation emphasizes
cally performed in the pseudo-spin-1/2 single-quantum (SQ) the importance of isotopic purification when low-nitrogen
subspace of the NV- ground state, with the |ms = 0i and samples are employed (see Sec. III.F).
either the |ms = +1i or the |ms = −1i spin state (∆ms = 1) For AC magnetometry, dephasing due to strain inhomo-
employed for sensing. In contrast, double-quantum (DQ) geneities and temperature fluctuations can be largely al-
coherence magnetometry (∆ms = 2) works as follows for a leviated by using Hahn echo or similar dynamical decou-
Ramsey-type implementation (see Fig. 15). First, an equal pling sequences (see Sec. IV.A) (Pham, 2013). Neverthe-
superposition of the |+1i and |−1i states is prepared (e.g., less, double-quantum coherence magnetometry should still
|+DQ i = √12 (| + 1i + | − 1i)). Then, after a free precession yield benefits. First, ensemble
√ AC magnetometry bene-
interval, the final population in |+DQ i is mapped back to fits from the expected 2× sensitivity gain due to twice
|0i, allowing for a magnetic-field-dependent population dif- faster precession (Fang et al., 2013). Second, sensitivity
ference between |0i and |−DQ i to be read out optically. may be further enhanced if T2,DQ exceeds T2,SQ /2. For ex-
 a b SQ, spin basis 25
Green !" !"
(532 nm) #" #"


SQ 2


SQ 2
$)
&'(
. 0
 MW SQ
+
/'( +
,&'(
1 1
DQ, spin basis



SQ 2 SQ 2 !" !"
#" #"
MW SQ
- $%

$)
&'(
&'( .
DQ DQ
1 1


MW DQ
±
DQ, bright/dark basis
(DQ = 2 SQ)
!*( #*( !*( #*(

&*( + .0
Readout -
,&'( -
/*( + + ,/'(
2&*(
1 1

FIG. 15 a) Pulse sequence for Ramsey-type double-quantum coherence magnetometry, as implemented in Ref. (Bauch et al., 2018).
The single-quantum (SQ) |0i ↔ | + 1i and |0i ↔ | − 1i transitions are driven by MWs at or near respective resonance frequencies
ν+ and ν− . For simultaneously applied resonant MWs with Rabi frequencies Ω+ = Ω− = ΩSQ , double-quantum (DQ) √ transitions
occur between |0i and an equal superposition of | + 1i and | − 1i with corresponding
√ DQ Rabi frequency Ω DQ = 2ΩSQ . A DQ
Ramsey sequence requires MW pulses of duration τDQ = π/ΩDQ = π/(ΩSQ 2) to prepare the state used for sensing. b) Comparison
between a conventional SQ π/2-pulse on a single transition (top panel) and a DQ π-pulse used to prepare the superposition state
|+DQ i = √12 (|+1i+|−1i) for the sequence in (a) (lower two panels). Middle panel shows the DQ state preparation pulse in the
bare spin basis, while the bottom panel shows the same pulse in the basis of |0i and the bright and dark states |+DQ i and |−DQ i,
i.e., the orthogonal superposition states respectively coupled to and blind to the MW drive. During a DQ Ramsey free-precession
interval, spin population oscillates between |+DQ i and |−DQ i, (as these states are not energy eigenstates), at a rate proportional to
the magnetic field. For additional detail see Section 5.2 in Ref. (Schloss, 2019). Adapted from Ref. (Schloss, 2019).

a b c
%&'()*+(,-.+/"#$+*0-1-.2 67+(,-.+/"#$+*0-1-.2 899:
! !

!"#$ !"#$ !"#$

!"# !"# !"# !"# !"# ! !"#

345 345 345

FIG. 16 Selected pulse sequences for concurrent manipulation of NV- spins and the surrounding paramagnetic spin bath. a) Pulsed
spin bath driving protocol combining a Ramsey sequence on the NV- center(s) with a central RF π-pulse on the spin bath. b)
Continuous spin bath driving protocol combining a Ramsey sequence with continuous resonant RF spin bath drive. c) Hahn echo-
based double electron-electron resonance (DEER) protocol consisting of a Hahn echo sequence performed on the NV- center(s)
combined with a resonant RF π-pulse performed on the spin bath. Recreated from Ref. (Bauch et al., 2018).

ample, the authors of Ref. (Angerer et al., 2015) observe magnetic fields, these two transitions must be addressed
T2,SQ = 1.66 ± 0.16 ms and T2,DQ = 2.36 ± 0.09 ms for with separate and phase-locked MW frequencies (Angerer
single near-surface NV- center with T2,SQ likely limited by et al., 2015; Mamin et al., 2014). While equal Rabi fre-
electric field noise. In addition to magnetic sensing, mea- quencies on the two transitions are desirable, the MW pulse
surements employing the full spin-1 basis can enhance sensi- durations may be adjusted to compensate for unequal Rabi
tivity for temperature sensing (Toyli et al., 2013) and noise frequencies. MW pulses for each spin transition may be ap-
spectroscopy applications (Kim et al., 2015; Myers et al., plied simultaneously (Bauch et al., 2018; Fang et al., 2013;
2017). Mamin et al., 2014), as depicted in Fig. 15a, or sequen-
As shown schematically in Fig. 15, implementation of tially (Toyli et al., 2013). At low magnetic field, electric
double-quantum coherence magnetometry is a straightfor- field, and strain, a single MW frequency is adequate (Fang
ward extension of standard pulsed magnetometry. The DQ et al., 2013). In either case, care must be taken to en-
technique requires applying MW pulses to drive both the sure that both the upper and lower spin transitions are
|0i ↔ |−1i and |0i ↔ |+1i transitions. For sufficiently large addressed with adequate MW pulses to achieve an equal
 26

superposition of |+1i and |−1i (Bauch et al., 2018; Mamin To effectively suppress NV- dephasing, all nitrogen spin
et al., 2014). Due to the minimal increase in experimen- transitions must typically be driven. Elemental nitrogen
tal complexity, the ability to suppress common-mode noise occurs in two stable isotopes, 14 N with 99.6% natural iso-
sources, and the increased spin precession rate, we expect topic abundance, and 15 N with 0.4% natural isotopic abun-
DQ coherence magnetometry to become standard for high- dance. Diamonds may contain predominantly 14 N, where
performance pulsed-measurement DC magnetometers em- the 99.6% natural abundance purity is typically deemed
ploying NV- ensembles. sufficient, or 15 N, which requires isotopic purification. 14 N
exhibits nuclear spin I = 1 while 15 N exhibits nuclear spin
I = 1/2, resulting in 3 and 2 magnetic-dipole-allowed tran-
C. Spin bath driving sitions for each isotope, respectively (Cook and Whiffen,
1966; Loubser and van Wyk, 1978; Smith et al., 1959). Like
Residual paramagnetic impurity spins in diamond con- NV- centers, substitutional nitrogen defects possess a trig-
tribute to NV- dephasing, thereby reducing T2∗ . This effect onal symmetry as a result of a Jahn-Teller distortion (Am-
can be mitigated by directly driving the impurity spins, merlaan and Burgemeister, 1981; Davies, 1979, 1981). The
which is particularly useful when dynamical decoupling (see Jahn-Teller distortion defines a symmetry axis along any
Sec. IV.A) of the NV- sensor spins is not applicable, such of the 4 crystallographic [111] axes, leading to 4 groups of
as in DC sensing protocols. This technique, termed spin N0S spins. For an axial bias magnetic field B0 satisfying
bath driving, has been successfully demonstrated with sub- ge µB /~B0  AHF where AHF ∼ 100 MHz is the substitu-
stitutional nitrogen spins N0S (S = 1/2) (Bauch et al., 2018; tional nitrogen hyperfine interaction, ms and mI are good
Knowles et al., 2013; de Lange et al., 2012). Due to the high quantum numbers, and the 14 N spectrum consequently ex-
typical concentrations of N0S spins in NV-rich diamonds, we hibits up to 12 distinct resonances, each of which needs to
focus our discussion on this implementation. be driven (Belthangady et al., 2013; de Lange et al., 2012).
In pulsed spin bath driving (see Fig. 16a), a resonant π- If B0 is aligned with any of the diamond [111] axes, the
pulse is applied to the N0S spins halfway through the NV- 12 resonances reduce to 6 partially-degenerate groups with
Ramsey sequence, decoupling the N0S spins from the NV- multiplicity 1:3:1:3:3:1 (see Fig. 19a). Similarly, the 15 N
spins in analogy with a refocusing π-pulse in a Hahn echo spectrum shows up to 8 distinct resonances, which reduce
sequence (see Fig. 14) (Bauch et al., 2018; de Lange et al., to 4 partially-degenerate groups with multiplicity 1:3:3:1
2012). Alternatively, the spin bath can be driven contin- for B0 aligned to an NV internuclear axis (see Fig. 19b).
uously (see Fig. 16b) (Bauch et al., 2018; Knowles et al., Spin-bath driving is expected to be easiest to execute
2013; de Lange et al., 2012). In the latter case, the driving when the bias magnetic field B0 and hyperfine coupling AHF
Rabi frequency ΩN must significantly exceed the NV- -N0S are not of the same order. When gµhB B0 ∼ AHF , additional
coupling rate γN (i.e., satisfy ΩN /γN  1) to achieve ef- nuclear-spin-non-conserving transitions arise, resulting in
fective decoupling. [Note that γN ∼ 2π × (0.01 − 10) MHz reduced oscillator strength for the nuclear-spin-conserving
for nitrogen concentrations in the 1 − 1000 ppm range (see transitions. Thus, given fixed RF power, the drive efficiency
Sec. III.D).] Under this condition, the nitrogen spins un- for each addressed transition decreases. Although spin bath
dergo many Rabi oscillations during the characteristic dipo- driving has to date only been demonstrated in the regime
lar interaction time 1/γN . As a result, the NV- ensemble is gµB
h B0 & AHF (Bauch et al., 2018; Knowles et al., 2013;
decoupled from the nitrogen spin bath and the NV- dephas- de Lange et al., 2012), we expect driving in the gµhB B0 
ing time is enhanced. This phenomenon is similar to mo- AHF regime to also be effective.
tional narrowing observed in many NMR and ESR systems, The N0S electron spin resonance spectra for 14 N and 15 N
such as rotation- and diffusion-induced time-averaging of are readily observed in EPR experiments [see for example
magnetic field imhomogeneities (Abragam, 1983c; Slichter, Ref. (Smith et al., 1959) and (Drake et al., 2016)]. Al-
1990). ternatively, the nitrogen resonance spectra in a diamond
The authors of Ref. (de Lange et al., 2012) perform pulsed can be characterized with NV- centers using a Hahn-echo-
spin bath driving in a diamond with [NT ] . 200 ppm and based double electron-electron resonance (DEER) tech-
increase T2∗ for a single NV- 1.6×, from 278 ns to 450 ns. nique (Bauch et al., 2018; de Lange et al., 2012). In this
Similarly, in Ref. (Knowles et al., 2013), T2∗ for an individ- case, the NV- electronic spin is made sensitive to deco-
ual NV- is extended from 0.44 µs to 1.27 µs, a 2.9× improve- herence from N0S target impurity spins via application of
ment, using continuous spin bath driving in nanodiamonds frequency-selective π-pulses at the targeted spins’ resonance
with [N] . 36 ppm. An NV- ensemble study in Ref. (Bauch frequency. A schematic of the DEER pulse sequence is
et al., 2018) finds that if another mechanism, such as lattice shown in Fig. 16c, and the resulting DEER spectra for
strain or magnetic field gradients, is the dominant source both nitrogen isotopes are compared in Fig. 19. Extra
of dephasing, spin bath driving becomes less effective, as resonance features associated with substitutional-nitrogen-
shown in Fig. 17 (see also Sec. III.C). Nonetheless, at high related dipole-forbidden transitions and additional param-
nitrogen concentrations ([NT -
S ] & 1 ppm), NV ensemble de- agnetic spins are also commonly observed and may reveal
phasing due to dipolar interaction with nitrogen spins can additional sources of dephasing.
be greatly reduced by spin bath driving (Bauch et al., 2018), The experimental requirements for effective spin bath
as also demonstrated in single-NV- experiments (Knowles driving depend on the substitutional nitrogen concentra-
et al., 2013; de Lange et al., 2012). tion. At lower impurity concentrations, reduced spin
 27

bath drive strength (i.e., RF power) is needed to mitigate

nitrogen-induced dephasing. However, dephasing mecha-
nisms unrelated to nitrogen may exhibit larger relative con-
tributions to T2∗ in this regime, limiting the achievable T2∗
increase from nitrogen spin bath driving. In particular, in
samples with nitrogen content [N0S ] . 1 ppm, lattice-strain
gradients may dominate the ensemble dephasing time, as
is found in Ref. (Bauch et al., 2018). In this instance,
strain-insensitive measurement techniques, such as double-
quantum coherence magnetometry (see Sec. IV.B) must be
employed for spin bath driving to extend T2∗ . In the inter-
mediate regime ([N0S ] ∼ 1 ppm), where strain gradients and
NV- dipolar interactions with the nitrogen spin bath are of
similar magnitude, neither spin bath driving nor DQ coher-
ence magnetometry alone can achieve significant enhance-
ment of the dephasing time. However, Ref. (Bauch et al.,
2018) demonstrates a ∼ 16× improvement in T2∗ (effectively
a ∼ 32× improvement when considering the twice faster
precession rate in the DQ basis) for a [N0S ] ' 0.75 ppm
diamond when both techniques are combined, as shown in
FIG. 17 Ensemble free induction decay envelopes as measured
Fig. 18. In contrast, employing spin bath driving alone im-
using SQ and DQ Ramsey magnetometry, with and without con-
proves the dephasing time only by ∼ 1.1× (see Fig. 17), as tinuous spin bath driving. Measurements are shown for the fol-
strain-induced dephasing is left unmitigated. lowing: in the SQ basis without spin bath driving ( , first from
In nitrogen-rich diamonds ([N0S ] & 1 ppm), achieving the top); in the SQ basis with spin bath driving ( , second from
motional narrowing condition ΩN /γN  1 may be techni- top); in the DQ basis without spin bath driving ( , third from
cally difficult; increases in [N0S ] necessitate linear propor- top); in the DQ basis with spin bath driving ( , fourth from
tional increases in ΩN , which correspond to quadratic in- top). Measurements in the DQ basis mitigate strain-induced
dephasing, while spin bath driving mitigates dipolar dephasing
creases in the RF power required (Bauch et al., 2018).
from the paramagnetic substitutional nitrogen in the diamond.
We expect both pulsed and continuous spin bath driving in The data illustrate the synergistic effect of combining DQ co-
nitrogen-rich samples to be ultimately limited by parasitic herence magnetometry and spin bath driving; the aggregate ap-
effects. These effects include induced AC Zeeman shifts, proach vastly outperforms either technique employed indepen-
strain gradients, and sample heating due to the strong ap- dently. Even with twice faster precession, T2∗ is extended from
plied RF fields (Knowles et al., 2013). 1.8 µs to 29 µs. The DQ protocol with spin-bath driving is
We note that spin bath driving should be applicable to depicted in the inset. From Ref. (Bauch et al., 2018).
any paramagnetic spin species in diamond, such as dark
electron spins (i.e., non-optically-active paramagnetic de-
D. Transverse strain and electric field mitigation
fects), NV- centers, or even nuclear spins. The effectiveness
of the driving for dilute bath spins (fractional concentration
Spatial and temporal variations in electric fields or in
 0.01) is expected to depend on the target spin’s concen-
diamond crystal strain can degrade T2∗ , as described in
tration but not its gyromagnetic ratio, as both ΩN and γN
Secs. III.C and IV.B. Measurements performed in the NV-
vary linearly with the target spin’s gyromagnetic ratio. In
spin’s double-quantum basis are insensitive to variations in
other words, species with small gyromagnetic ratios are dif-
the axial components of the electric field Ez and spin-strain
ficult to drive but also do not contribute much to dephas-
coupling Mz , as these terms cause common-mode shifts in
ing for a given concentration (Bauch et al., 2018). Given
the energies of the |+1i and |−1i states (Barson et al., 2017;
its relatively high concentration, spin bath driving of the
13 Glenn et al., 2017). In contrast, broadening due to trans-
C in a natural abundance diamond ([13 C]= 10700 ppm)
verse electric fields Ex , Ey and transverse spin-strain cou-
is expected to be quite challenging (Bauch et al., 2019).
plings Mx and My may remain in DQ measurements (Bar-
Lastly, as the nitrogen spin bath contributes to T2 decoher-
son et al., 2017; Udvarhelyi et al., 2018). However, by oper-
ence (Bar-Gill et al., 2012; Pham et al., 2012a), nitrogen
ating at a sufficiently strong axial bias magnetic field B0,z ,
spin bath driving would be expected to extend the Hahn
the resonance line broadening from inhomogeneities in Ex ,
echo T2 and, to a lesser extent, coherence times achieved
Ey , Mx , and My can be mitigated (Jamonneau et al., 2016;
with dynamical decoupling sequences (Bauch et al., 2018),
Schloss et al., 2018), as illustrated in Fig. 20 and discussed
although neither application has been demonstrated at
further in Appendix A.9.
present.
The frequency shifts of the NV- ground state spin reso-
nances due to transverse strain and electric fields at zero
magnetic field are given by
q
2 2
± ξ⊥ = ± (d⊥ Ex + Mx ) + (d⊥ Ey + My ) , (30)
 28

and NV- wavelength-dependent excitation and associated

fluorescence (Aslam et al., 2013; Waldherr et al., 2011). Key
advantages of SCC readout over conventional spin-state-
dependent fluorescence readout are: (i), a slightly increased
spin contrast (Jaskula et al., 2017); and (ii), the ability to
read out the charge state for extended durations and thus
collect more photons per readout, leading to high-fidelity
charge-state determination. Larger photon-numbers per
readout reduce the relative contribution of shot noise to
the measurement, allowing for readout fidelities within or-
der unity of the spin-projection limit σR = 1 (see definition
in Sec. II.A).
Successful spin-to-charge conversion requires control of
the NV charge state. Characterization of charge dynam-
ics under optical excitation (Aslam et al., 2013; Beha et al.,
2012; Hacquebard and Childress, 2018; Manson et al., 2018)
indicate power- and wavelength-selective photo-ionization
FIG. 18 Ramsey interference fringes versus an applied test mag- processes, which allow for controlled switching between NV-
netic field, measured in the SQ basis ( , top), DQ basis ( , and NV0 . For example, green ∼ 532 nm light transfers sin-
middle), and DQ basis with N0S spin bath driving ( , bottom).
The longer dephasing times achieved when combining DQ coher-
gle NV centers preferentially to NV- with 70 − 75% proba-
ence magnetometry and spin bath driving allow for denser Ram- bility (Aslam et al., 2013; Beha et al., 2012; Waldherr et al.,
sey fringes and enhanced sensitivity. The decreased contrast 2011); strong yellow ∼ 589 nm (Hopper et al., 2018a) or red
for magnetic fields > 0.05 mT in the bottom plot results from ∼ 637 nm (Jaskula et al., 2017; Shields et al., 2015) light can
magnetic-field-induced detuning of the nitrogen spin resonances selectively ionize NV- to NV0 via absorption of two photons
with respect to the RF drive frequencies. From Ref. (Bauch by an electron in the triplet ground state; and near-infrared
et al., 2018). ∼ 900 -1000 nm (Hopper et al., 2016) can similarly ionize
NV- via absorption of two photons by an electron in the
singlet metastable state. Readout of the NV- charge state
where d⊥ = 0.17 Hz/(V/m) (Dolde et al., 2011; Michl
is commonly performed by applying weak yellow laser light
et al., 2019; Van Oort and Glasbeek, 1990) is the trans-
at ∼ 594 nm. At intensities well below the NV- saturation
verse electric dipole moment of the NV- ground state spin.
intensity Isat ∼ 1 - 3 mW/µm2 (Wee et al., 2007), yellow
Application of an external axial magnetic field B0,z intro-
light efficiently excites the NV- electronic spin transition
duces additional magnetic-field-dependent shifts and sup-
with zero phonon line (ZPL) at 637 nm without inducing
presses the effect of ±ξ⊥ on the spin resonances. When
ionization, while hardly exciting the NV0 transition (with
βz ≡ (ge µB /h)B0,z  ξ⊥ , contributions to T2∗ from tem-
ZPL at 575 nm) (Aslam et al., 2013; Beha et al., 2012;
poral fluctuations or spatial variations in ξ⊥ (Fang et al.,
Waldherr et al., 2011). Through introduction of a photon-
2013; Jamonneau et al., 2016) are diminished (see Ap-
detection threshold combined with appropriate spectral fil-
pendix. A.9). For the nanodiamonds characterized in
tering, NV- (which fluoresces under the yellow excitation)
Ref. (Jamonneau et al., 2016), with ξ⊥ = 7 MHz, B0,z ∼
may thus be distinguished from NV0 (which produces lit-
30 G is required to suppress the contribution to T2∗ from
tle if any fluorescence). Figure 21a displays a photon-count
transverse electric fields and strain. For the lower-strain
histogram characteristic of single-NV charge readout repro-
bulk diamonds used in Refs. (Fang et al., 2013; Jamonneau
duced from Ref. (Bluvstein et al., 2019). The clear sepa-
et al., 2016), with ξ⊥ ∼ 10 kHz, B0,z . 100 mG is sufficient.
ration of photon distributions from NV0 and NV- at low
excitation powers allows charge-state determination with
V. METHODS TO INCREASE READOUT FIDELITY fidelity > 99% (Hopper et al., 2018b).
The original work by Ref. (Shields et al., 2015) demon-
A. Spin-to-charge conversion readout strates SCC readout on a single NV center in Type IIa di-
amond nanobeams (see Appendix A.11 for overview of dia-
Spin-to-charge conversion (SCC) readout is an alterna- mond types). First, utilizing green laser light (see Fig. 21b)
tive to conventional fluorescence-based readout of the NV- and appropriate MWs, the NV center is prepared in the
spin state. The technique has been demonstrated for single ms = 0 or one of the ms = ±1 spin states of the NV-
NVs (Ariyaratne et al., 2018; Hopper et al., 2016; Jaskula triplet electronic ground state. A moderate power, 594 nm
et al., 2017; Shields et al., 2015) and for small ensem- yellow “shelving” pulse (145 µW, ∼ 0.9 mW/µm2 ) then ex-
bles in nanodiamonds (Hopper et al., 2018a) and bulk dia- cites the spin population to the triplet excited state. Due
mond (Jayakumar et al., 2018). In SCC readout, the NV- to the spin-dependent intersystem crossing from the triplet
center’s spin state is mapped optically onto the NV’s neu- excited state, the ms = 0 population is more likely to de-
tral and negative charge states (NV0 and NV- ). The charge cay back to the ground state, whereas the ms = ±1 pop-
state, and thus the original NV- spin information, can then ulation is more likely to be shelved into the metastable
be accurately read out by exploiting differences in the NV0 singlet states. The spin-to-charge conversion is then re-
 29

!"
#
!"#$#%&
!"#$#%&)*
!"#$#(

!"#$#'&
!
"#$!
.22:>-<# 9:8;3<<-/
48./=343:/= 48./=343:/=

!"#$#'&
!"#$#'&)*
!"#$#(
!"#$#%&
,--!./#0123443/5
+#$#( 671-8
/-#0123443/5
!
!+#

!$
#
!"#$#'&)*
!=#$#%&)*

!"#$#%&)*
!
"#$!
.22:>-<# 9:8;3<<-/
48./=343:/= 48./=343:/=

!"#$#%&)*
!=#$#'&)*

!"#$#'&)*
,--!./#0123443/5
+#$#( 671-8
/-#0123443/5
!
!+#

FIG. 19 Substitutional nitrogen N0S spin energy levels (left panes) and associated double electron-electron resonance (DEER) spectra
(right panes), for 14 N (top) and 15 N (bottom). Simulated spectra depict allowed-transition resonances (∆mI = 0) of the primary
nitrogen isotope ( ), forbidden-transition resonances (∆mI 6= 0) of the primary nitrogen isotope ( ), and spurious features
associated with allowed transitions of impurity isotopes ( ). The simulated data resonance linewidths and amplitudes are chosen to
approximately match the experimental data ( ). Spectra are simulated for and experimentally measured in an external magnetic
field aligned along the diamond crystallographic [111] axis. Adapted from Ref. (Bauch et al., 2018).

alized with a ∼ 10 ns high intensity resonant 637-nm pulse relaxometry and AC field sensing), where the penalty due
(22.5 mW, ∼ 140 mW/µm2 ), which ionizes (i.e., converts to additional readout overhead is less severe. To date, the
NV- to NV0 ) the triplet ground state population, corre- best SCC readout demonstrations improve field sensitiv-
sponding to ms = 0, but leaves the shelved population cor- ity only when interrogation times exceed ∼ 10 µs (Hopper
responding to ms = ±1 unaffected. Last, the NV charge et al., 2018a; Shields et al., 2015), which further motivates
state is read out by applying weak ∼ 594 nm light. The improvement of spin ensemble properties to achieve suffi-
∼ 594 nm light with lower energy than the NV0 ZPL at ciently long dephasing times (see Sec. III.A).
575 nm, ensures that only NV- is excited while the weak Given the clear success of SCC readout with single NVs,
intensity (∼ 1−10 µW, ∼ 6−60 µW/µm2 , Fig. 21a) ensures application to NV- -rich ensembles is a logical progres-
that NV- is not ionized during readout. sion, especially given the low conventional readout fideli-
The single-NV SCC result by Ref. (Shields et al., 2015) ties achieved for NV- -rich ensembles (F . 0.015, see Ta-
achieves a factor over spin projection noise σR = 2.76 ble 2). However, the prospect for SCC readout to substan-
(F = 1/σR = 0.36, see comparison in Table 2). As the fi- tially improve F in NV- ensembles likely hinges on whether
delity of the charge readout process itself approaches unity the additional complex charge dynamics present in NV-rich
(F CR = 0.975), the dominant inefficiency is attributed diamonds can be mitigated (Hopper et al., 2018a). Promis-
to the imperfect spin-to-charge conversion step (F SCC = ing SCC readout results on small NV ensembles in Type Ib
0.37). Several alternative SCC readout variants have been nanodiamonds demonstrate σR = 20, compared to σR = 70
demonstrated, providing similar sensitivity gains while of- with conventional readout in the same setup, allowing the
fering reduced experimental complexity (Hopper et al., authors to observe improved sensing performance for in-
2018a), or utilizing the singlet state for ionization (Hop- terrogations times > 6 µs (Hopper et al., 2018a). How-
per et al., 2016). For all SCC readout implementations, ever, this and other studies (Choi et al., 2017a; Manson
however, the improved values of σR come at the cost of sub- et al., 2018) report intricate NV- and NV0 charge dynam-
stantially prolonged spin readout times tR , which increase ics absent in single NV experiments. The effectiveness of
the sequence’s overhead time and diminish the overall sen- SCC readout in the complex charge environment inherent
sitivity improvement (see Sec. II.A). For example, the best to NV-rich ensembles (e.g., due to ionization and charge
reported readout fidelity (F = 0.36) (Shields et al., 2015) is dynamics of substitutional nitrogen and other impurity de-
achieved for readout times tR = 700 µs, which exceed con- fects) warrants further investigation (see Sec. VI.B). Nev-
ventional fluorescence-based readout times (tR ∼ 300 ns) by ertheless, SCC readout overcomes one sensing disadvantage
∼ 1000×. SCC readout is therefore most advantageous for specific to ensembles, namely that NV- orientations not be-
measurement modalities with long sensing intervals (e.g., T1 ing used for sensing can be preferentially transferred to
 30

-150 -100 -50 0 50 100 150

120
110 a (c)
100
NV 0.15
90 NV -
NV1GIJKGIJKGIJ
80 - = 0.78(1) 0.10

P(n)
70
60
0.05
T2* (µs)

50

40 0
0 10 20 30
(d)NV 0 0.15GIJKGI
30 b n photon counts1V
NV
Green (532 nm)
20 Initialize

Microwave (3 GHz)
-150 -100 -50 0 50 100 150
Spin
Bz (mG) Manipulation
Red (637 nm)
Ionize

FIG. 20 Suppression of dephasing from transverse electric fields NV- Fluorescence

Gate (> 690 nm)
and strain. For the single NV- center measured in Ref. (Jamon-
Yellow (594 nm)
neau et al., 2016), the dephasing time T2∗ {single} at zero magnetic Shelve Readout

field is limited by electric field fluctuations transverse to the NV-

symmetry axis, Ex and Ey . An applied axial magnetic field sup- FIG. 21 a) Probability histogram depicting photon emission
presses this source of dephasing by decoupling the NV- center from NV0 and NV- under weak yellow excitation. The striking
from transverse electric fields and strain. For magnetic fields difference in photon emission rate between NV0 and NV- al-
larger than ∼ 100 mG, the value of T2∗ {single} is limited by mag- lows the NV charge state to be determined with fidelity & 99%.
netic noise, reaching ∼ 100 µs in this isotopically enriched, [13 C] Adapted from Ref. (Bluvstein et al., 2019). b) Schematic of
= 20 ppm sample. For NV- ensembles at zero magnetic field, the SCC readout protocol used by Shields et al. (Shields et al.,
in addition to temporal fluctuations in Ex and Ey that limit 2015). Adapted from Ref. (Jaskula et al., 2017).
T2∗ {single} , spatial variations in Ex and Ey and in the transverse a
spin-strain coupling terms Mx and My may also limit T2∗ for 1.40s105 3.25
ensembles. Recreated from Ref. (Jamonneau et al., 2016).
Photoluminescence intensity

1.33s105 ODMR 3.20

NV0 during the ionization step. This results in reduced

(photons s–1)

background fluorescence and potentially allows for an ad- 1.26s105 3.15

ditional ∼ 2× sensitivity improvement relative to conven-
tional NV- readout. Overall, beyond the long overhead 3.10
times already discussed, SCC readout’s demanding power 1.19s105 PDMR
requirements are expected to further hamper ensemble-
based implementation. In particular, high required optical 3.05
1.12s105
intensities (& 150 mW/µm2 ) (Hopper et al., 2016, 2018a;
2,850 2,860 2,870 2,880 2,890
Shields et al., 2015) suggest scaling of SCC readout to larger
Microwave frequency (MHz)
bulk sample sizes (& 100 × 100 µm2 ) will be challenging.
FIG. 22 Photoelectrically detected magnetic resonance
(PDMR) of NV- centers. Spectra are simultaneously measured
B. Photoelectric readout
by ODMR ( ) and PDMR ( ) in the absence of an external
magnetic field. From Ref. (Bourgeois et al., 2015).
Another method to interrogate the NV- spin state is
photoelectric (PE) readout, which relies on measuring
a current of charge carriers resulting from NV- photo- duces NV- photo-ionization, and the ejected electrons gen-
ionization (Bourgeois et al., 2015, 2017; Gulka et al., 2017; erate a current, which is collected at the electrodes. NV-
Hrubesch et al., 2017; Siyushev et al., 2019). Since NV- photo-ionization may occur via single- or two-photon exci-
photo-ionization is spin-state dependent, (see Secs. V.A and tation. Single-photon ionization of the NV- 3 A2 electronic
VI.B) (Shields et al., 2015), the spin state can be inferred ground state requires photon energies of 2.7 ± 0.1 eV or
from the photocurrent signal in analogy to fluorescence- higher (wavelength . 460 nm) (Aslam et al., 2013; Bour-
based readout. Figure 22 shows a photoelectrically de- geois et al., 2017; Deák et al., 2014; Londero et al., 2018).
tected magnetic resonance (PDMR) spectrum measured si- PE readout implementations employing lower photon en-
multaneously with an ODMR spectrum, from Ref. (Bour- ergies, such as from 532 nm (2.33 eV) light common for
geois et al., 2015). One promised benefit of PE readout spin-state initialization, ionize the NV- centers via a two-
is that the photoelectron collection efficiency can approach photon process, namely 3 A2 → 3 E → conduction band (see
unity (Bourgeois et al., 2015). Fig. 30) (Bourgeois et al., 2015, 2017; Gulka et al., 2017;
In PE readout, a bias voltage is applied across electrodes Heremans et al., 2009). Whereas the rate of single-photon
fabricated on the diamond surface. An excitation laser in- ionization scales linearly with optical intensity (Hacque-
 31
a b
bard and Childress, 2018), two-photon ionization depends

threshold
500
quadratically on intensity (Aslam et al., 2013). Repetitive

Number of events
Optically illuminating the diamond for PE readout may readout

Readout fidelity
400
also induce background photocurrent from ionization of
other defects present in the sample. Most unfortunately, 300
532 nm green light ionizes substitutional nitrogen N0S de-
200
fects in a single-photon process (Heremans et al., 2009). Single
The background N0S photocurrent may exceed the signal conventional 100
NV- photocurrent, resulting in poor NV- measurement con- readout
trast. This problem is exacerbated for excitation intensities 0
well below the NV- saturation intensity, where two-photon Repetitive readout cycles Number photons / 5 ms
NV- ionization may be weak compared to single-photon c
ionization of N0S , and at elevated nitrogen concentrations
[N0S ]  [NV- ] (Bourgeois et al., 2017; Londero et al., 2018).
Multiple approaches can partially mitigate the unwanted
photocurrent associated with N0S ionization. For example,
lock-in techniques can remove the DC background from the
nitrogen photocurrent (Gulka et al., 2017). Additionally, a
shorter-wavelength laser can be employed to induce single-
photon ionization from the NV- 3 A2 state, thereby improv-
ing the NV- ionization rate relative to that of N0S . However,
the authors of Ref. (Bourgeois et al., 2017) observe that un-
der optimized experimental conditions, single-photon ion-
FIG. 23 Overview of ancilla-assisted repetitive readout. a)
ization using 450 nm light provides no contrast improve-
Readout fidelity F is improved with the number of repetitive
ment compared to two-photon ionization with 532 nm light. readout cycles. Fidelity for repetitive readout (red) is plotted
A variety of challenges accompany implementation of relative to a single conventional readout (blue, dashed). From
PE readout, not only for single NV- centers and small Ref. (Lovchinsky et al., 2016). b) The clear difference in to-
NV- ensembles (Bourgeois et al., 2015, 2017; Gulka et al., tal number of collected photons associated with the initial ms
2017; Hrubesch et al., 2017; Siyushev et al., 2019) but states allows determination of ms with fidelity approaching 1 in
also for envisioned extensions to larger detection volumes some implementations. Here F ≈ 0.92 in Ref. (Neumann et al.,
2010a). From Ref. (Neumann et al., 2010a). c) Quantum cir-
& (100 µm)3 using NV-rich diamonds. In addition to back-
cuit diagram and magnetometry pulse sequence with detection
ground photocurrent from ionization of nitrogen and other via ancilla-assisted repetitive readout. Application of an RF π-
defects, another expected obstacle to PE readout is electri- pulse between two weak MW π-pulses maps the NV- electronic
cal cross-talk between MW-delivery electrodes (used to ma- spin superposition onto the ancilla nuclear spin. Subsequently
nipulate the NV- spin states) and photocurrent-detection the superposition state may be repeatedly mapped back onto the
electrodes (Gulka et al., 2017; Siyushev et al., 2019). Fluc- electronic spin via a weak MW π-pulse and optically read out
tuations in the applied electric field could also add addi- without destroying the ancilla spin’s quantum state. Adapted
from Ref. (Lovchinsky et al., 2016).
tional measurement noise by coupling to fluctuations in
photoelectric collection efficiency.
Scaling PE readout implementations to larger NV- en-
minished in NV-rich diamonds due to charge traps, non-
sembles may introduce additional challenges. Because the
uniform electric fields, and space-charge limitations (Bour-
electrodes reside on the diamond surface, collecting pho-
geois et al., 2015; Bube, 1960; Rose, 1963). The applicabil-
tocurrent from NV- centers located & 100 µm deep may
ity of photoelectric gain to improving PE readout fidelity in
prove difficult (Bourgeois et al., 2015). Achieving the
ensemble-based extensions remains to be shown. Although
necessary bias electric field strength and uniformity over
PE readout shows promise for nanoscale sensing and inte-
& (100 µm)3 volumes may also be challenging; bias electric
grated quantum devices (Morishita et al., 2018), and may
field gradients across large detection volumes could reduce
prove beneficial when combined with PIN structures (Kato
NV- ensemble T2∗ values. Moreover, the presence of charge
et al., 2013), this technique’s utility for ensemble magne-
traps in NV-rich diamonds might hinder photoelectric col-
tometry in NV-rich diamonds remains unknown.
lection efficiency (see Sec. VI.F), especially from deeper NV-
centers. In addition, Johnson noise in the readout elec-
trodes may induce magnetic field fluctuations that could C. Ancilla-assisted repetitive readout
limit the achievable sensitivity (Kolkowitz et al., 2015).
In certain PE readout implementations, the detected sig- In conventional readout, the fast ∼ 500 ns repolarization
nal amplitude may be increased by photoelectric gain, an of the NV- electronic spin limits the number of photons an
intrinsic charge-carrier amplification arising from the dia- NV- emits before all initial spin state information is lost
mond’s charge dynamics and the electrode boundary con- (see Fig. 6). Even when implementing conventional read-
ditions (Bourgeois et al., 2015; Hrubesch et al., 2017; Rose, out with the best present collection efficiencies, the average
1963). However, photoelectric gain is expected to be di- number of collected photons per NV- center navg is less than
 32

1, and for many implementations navg  1, making photon NV- orientation (Schloss et al., 2018). Even slight angu-
shot noise the dominant contributor to the parameter σR lar misalignments introduce measurement back action on
(see Eqn. 12, Table 2, Sec. V.E). An alternative method the nuclear spin Iz , which spoils T1,n (Neumann et al.,
to increase the readout fidelity F = 1/σR circumvents this 2010a). The reduction in T1,n limits the available read-
problem by instead first mapping the initial NV- electronic out duration. Ensemble implementations would therefore
spin state information onto an ancilla nuclear spin. In the require highly uniform bias magnetic fields over ensemble
second step, the ancilla nuclear spin state is mapped back sensing volumes, conceivably on the ∼ (100 µm)3 scale. En-
onto the electron spin, which is then detected using conven- gineering such fields is within current technical capability
tional fluorescence-based readout. This second step may be but difficult nevertheless (see Sec. III.C and Ref. (Vander-
repeated many times with each marginal readout improv- sypen and Chuang, 2005)). Additionally, the MW and RF
ing the aggregate readout fidelity, as shown in Fig. 23a,b. control pulses would ideally manipulate the entire ensem-
While first demonstrated with a nearby 13 C nuclear spin as ble uniformly; spatial inhomogeneities of the control pulses
the ancilla (Jiang et al., 2009), the technique was later re- are likely to result in reduced readout fidelity unless mit-
alized using the NV- center’s 14 N (Neumann et al., 2010a) igated (Vandersypen and Chuang, 2005). Assuming suffi-
and 15 N nuclear spin (Lovchinsky et al., 2016). In the 13 C ciently strong and homogeneous B0 fields and MW driving
realization (Jiang et al., 2009; Maurer et al., 2012), the cou- can be realized, and that the additional overhead time is ac-
pling to the ancilla spin depends on the distance between ceptable, repetitive readout appears to be a promising but
the NV- defect and the nearby 13 C atom, making the tech- technically demanding method to improve F for ensembles.
nique difficult to implement for NV- ensembles where this
distance varies. This discussion instead focuses on the more
scalable realization using the NV- nitrogen nuclear spin as D. Level-anticrossing-assisted readout
the ancilla, which ensures the electron spin to ancilla spin
coupling remains fixed over the NV- ensemble. In conventional readout (Doherty et al., 2013), the read-
Figure 23c shows a quantum circuit diagram from out fidelity F = 1/σR depends on the number of photons
Ref. (Lovchinsky et al., 2016) depicting the repetitive read- navg collected per measurement sequence (see Eqn. 12).
out scheme. After the final MW pulse in an NV- sensing The value of navg is limited by the time the spin popu-
protocol, the NV- electronic spin state (denoted by sub- lation originally in the ms = ±1 states spends shelved in
script e) is mapped onto the nitrogen nuclear spin state the singlet state before decaying to the triplet ms = 0 state.
(subscript n). In Ref. (Lovchinsky et al., 2016), this map- Steiner et al. engineer the NV- spin to pass through the sin-
ping is achieved using a SWAP gate (CNOTe|n −CNOTn|e − glet state multiple times before repolarization, extending
CNOTe|n ), where CNOT denotes a controlled NOT gate. the readout duration per sequence to increase navg (Steiner
The SWAP gate consists of a MW π-pulse, then an RF et al., 2010), as depicted in Fig. 24. Using NV- centers with
14
π-pulse, then another MW π-pulse, where the MW pulses N, which has nuclear spin I = 1, three cycles through the
flip the electronic spin and the RF pulse flips the nuclear √ readout, yielding a ∼ 3× increase
singlet state occur during
spin. This procedure swaps the electronic and nuclear spin in navg and thus a ∼ 3× improvement in the fidelity F.
states, importantly, storing the electronic spin state infor- For NV- centers with 15 N with I = 1/2, the spin only passes
mation in the ancilla nuclear spin. Then an optical pulse twice through√the singlet state before repolarization, yield-
re-polarizes the electronic spin to ms = 0. Next, a set ing only a ∼ 2× improvement in F.
of repetitive readouts is performed. In each readout, the The technique is implemented as follows: the bias field B0
nuclear spin state is copied back onto the electronic spin is tuned to the excited-state level anticrossing at BLAC ≈
with a MW pulse, (a CNOTe|n gate), and then the elec- 500 G (Fuchs et al., 2008; Neumann et al., 2010a) to al-
tronic spin is optically read out without affecting the nu- low resonant flip-flops between the NV- center’s electronic
clear spin state. This process can be repeated many times spin and its 14 N nuclear spin (I = 1). Operation at
(& 102 ), and is limited in principle by the nuclear spin life- the level anticrossing polarizes the nuclear spin into the
time T1,n . In Ref. (Lovchinsky et al., 2016), while the initial state |mI = +1i (Jacques et al., 2009). At completion
RF pulse used in the SWAP gate requires ∼ 50-60 µs, each of a sensing sequence, immediately prior to readout, the
readout cycle requires only ∼ 1 µs. The large number of NV- electronic spin occupies a superposition of the states
readouts allow the aggregate readout fidelity F = 1/σR to |ms = 0, mI = +1i and |ms = −1, mI = +1i. Before the NV-
approach 1; notably, Ref. (Neumann et al., 2010a) achieves electronic spin state is read out using a conventional green
F = 0.92 (σR = 1.1) as depicted in Fig. 23b. laser pulse, two sequential RF π-pulses flip the nuclear spin
Extending ancilla-assisted repetitive readout to ensem- into the mI = −1 state, conditional on the electronic spin
bles is expected to be fruitful but necessitates overcom- occupying the ms = −1 state. This CNOT gate relies on
ing further challenges. The scheme requires a large mag- the RF drive being resonant with the nuclear transitions
netic field to minimize coupling between the NV- nuclear between the mI states for population in the ms = −1 state
and electronic spins, with Refs. (Lovchinsky et al., 2016; and off-resonant for population in the ms = 0 state. Dur-
Neumann et al., 2010a) employing fields of 2500 gauss and ing readout, the population in |ms = −1, mI = −1i cycles
6500 gauss respectively. Further, the bias magnetic field through the long-lived singlet state three times before the
must be precisely aligned along a single NV- symmetry axis, information stored in the NV- electronic spin is lost, al-
presently precluding its use for sensing from more than one lowing more signal photons to be collected. After the first
 33
a b 0 laser pulse duration [arb. units]
|-1,0〉
rf 3
|-1,+1〉 (trigger)
|-1,-1〉

# photons [1000/ns]
n|0,+1〉
mw n|-1,+1〉
(operating 2
transition)
|0,0〉 n|-1,-1〉
|0,+1〉 |0,-1〉
1
n|0,+1〉-n|-1,-1〉 = 3 • ( n|0,+1〉-n|-1,+1〉 )
IPL [arb. units] |0,+1〉
bright

|0,-1〉 |0,0〉 0 n|0,+1〉-n|-1,+1〉

|-1,-1〉 |-1,0〉 |-1,+1〉
dark 0 1 2
0 laser pulse duration [arb. units] Laser pulse duration [μs]

FIG. 24 Level-anticrossing-assisted readout as demonstrated in Ref. (Steiner et al., 2010). At the excited-state level anticrossing
near B = 500 G, green optical excitation polarizes NV- into the spin state |ms = 0, mI = +1i. a) Upon completion of a sensing
sequence, two RF pulses transfer population in the electronic spin state |ms = −1i from the nuclear spin state |mI = +1i to
|mI = −1i without affecting the |ms = 0i state. During optical readout, this population passes three times through the singlet
states before being repolarized to |ms = 0, mI = +1i, increasing the time over which the state-dependent fluorescence contrast
persists. b) Time-resolved photon detection comparing conventional readout (gray) and LAC-assisted readout (blue). The optimal
readout duration is extended by 3× and the difference in detected photon number between the two spin states is increased by 3×.
From Ref. (Steiner et al., 2010).

and second pass through the singlet to the ms = 0 state, tively for photons emitted directly through the {100} dia-
an electron-nuclear spin flip-flop returns the electronic spin mond surface (Le Sage et al., 2012), as depicted in Fig. 25.
state to ms = −1, as shown in Fig. 3a of Ref. (Steiner Although anti-reflection coatings can allow for higher col-
et al., 2010), enabling another cycle through the singlet lection efficiencies, present implementations demonstrate
state. The third pass repolarizes the NV- spin into the only modest improvement (Yeung et al., 2012). While
stable |ms = 0, mI = +1i state. great effort has resulted in high values of ηgeo for single
This technique’s
√ utility for magnetic sensing depends on NV- centers through use of various nano-fabrication ap-
whether the ≤ 3× increase in fidelity F outweighs the proaches (Babinec et al., 2010; Choy et al., 2013, 2011;
cost of additional overhead time (see Eqn. 10) introduced Häberle et al., 2017; Hadden et al., 2010; Li et al., 2015;
by the RF pulses. Although the authors of Ref. (Steiner Marseglia et al., 2011; Momenzadeh et al., 2015; Neu et al.,
et al., 2010) assert that microsecond-scale RF nuclear spin 2014; Riedel et al., 2014; Schröder et al., 2011; Shields et al.,
π-pulses are attainable, achieving such nuclear Rabi fre- 2015; Yeung et al., 2012), such methods do not easily trans-
quencies over large ensemble volumes ∼ (100 µm)3 may late to large ensembles.
prove very difficult, making this method impractical for Successful approaches for bulk ensemble magnetome-
NV- ensembles with T2∗ . 1 µs. Additional challenges ters have so far focused on collecting NV- fluorescence
for implementation with NV- ensembles include realizing that has undergone total internal reflection in the dia-
the requisite uniformity in the MW/RF fields and in the mond (Le Sage et al., 2012; Wolf et al., 2015b). While ab-
500 G bias magnetic field over ensemble volumes. Finally, sorption of NV- fluorescence by various defects may limit
the scheme presently precludes sensing from more than one ηgeo (Khan et al., 2013, 2009) for some diamonds, nitro-
NV- orientation (Schloss et al., 2018). gen (Weerdt and Collins, 2008) and NV- centers (Fraczek
et al., 2017) are expected to hardly absorb in the NV- PL
band ∼ 600-850 nm. A collection efficiency of 39% is demon-
E. Improved photon collection methods strated in Ref. (Le Sage et al., 2012) by detecting fluores-
cence from the four sides of a rectangular diamond chip sur-
In the limit of low contrast, the readout fidelity F rounded by four photodiodes (see Fig. 26). However, the in-
is proportional to the square root of the average num- creased experimental complexity associated with employing
ber ofpphotons collected per NV- per measurement, i.e., four detectors in contact with the diamond may be problem-
√
F ∝ N/N = navg (see Eqn. 12). Under these condi- atic for certain applications. In another approach, Wolf et
tions, sensitivity can be enhanced by increasing the geomet- al. employ a trapezoidally-cut diamond chip and a parabolic
ric collection efficiency ηgeo , defined as N/Nmax , where N concentrator to improve collection efficiency (Wolf et al.,
and Nmax are the number of photons collected and emitted 2015b). Although the authors calculate ηgeo to be between
respectively by the NV- ensemble per measurement. 60% and 65%, this result is not confirmed experimentally.
Efficient photon collection in diamond is hindered by Ma et al. (Ma et al., 2018) demonstrate a collection effi-
total-internal-reflection confinement resulting from dia- ciency of 40% by eliminating all air interfaces between the
mond’s high refractive index of approximately 2.41. For diamond and detector, in conjunction with coupling prisms
example, air and oil immersion objectives, with numeri- which direct light exiting the diamond’s four side faces to
cal apertures of 0.95 and 1.49 respectively, provide calcu- the detector.
lated collection efficiencies of only 3.7% and 10.4% respec- In the future, collection efficiency in bulk NV-diamond
 34

&ollection efficiency (%)

magnetometers is expected to improve to near 100%, lim-
10
ited only by losses due to absorption. For example, light oil-immersion objective
lost from the top and sides of the diamond in Ref. (Wolf 8
air objective
et al., 2015b) could be redirected to the detector by coat- 6
ing these sides of the diamond with a metallic (Choy et al., 4
2011) or dielectric reflector (Boesch et al., 2016). The
authors of Ref. (Wolf et al., 2015b) might also see an 2
improvement in collection efficiency by designing an op- 0
timized parabolic concentrator rather than using a com- 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
mercially available part. Hypothetical geometries for light 1umerical aperture
collection using parabolic or ellipsoidal reflectors are dis-
FIG. 25 Calculated collection efficiencies of NV- fluorescence
cussed in Ref. (Boesch et al., 2016). Whereas multiple- by oil-immersion or air microscope objectives through the {100}
reflection methods are suitable for bulk magnetometers, in- surface of a diamond chip, as a function of numerical aperture.
creasing ηgeo by collecting light undergoing multiple reflec- Figure and caption from Ref. (Le Sage et al., 2012).
tions in the diamond may substantially complicate accu-
rate image reconstruction for NV- magnetic imaging mi- a(b) c 10
croscopes (Le Sage et al., 2013; Pham et al., 2011; Steiner
laser light in
(d) 10
side-collection signal
et al., 2010). 109

detected photons
diamond
Is is also natural to consider exploiting the Purcell ef-
fect to improve readout fidelity. By engineering physical fluorescence 108 x100
out
structures around a chosen NV- center or ensemble, sev- b(d) 107
eral works have increased the triplet excited state’s ra- objective signal
106
diative decay rate (Choy et al., 2011; Kaupp et al., 2016;
Riedel et al., 2017). The increased radiative decay allows 105
for more PL photons to be collected from population orig- 0 1 2 3 4 5
4 mm laser power (W)
inally in |ms = 0i while population originally in |ms = ±1i
is shelved in the singlet states. Although theoretical in-
vestigations suggest that Purcell enhancement might im- FIG. 26 Fluorescence side-collection method (Le Sage et al.,
prove readout fidelity (Wolf et al., 2015a) or leave fidelity 2012). a) Green optical excitation is applied normal to the large
face of the diamond chip, and red fluorescence is collected from
unchanged (Babinec et al., 2012), the only reported exper- the sides. b) Red fluorescence from actual diamond chip. c) The
imental demonstration of Purcell-enhanced NV- spin read- depicted implementation results in a 100× increase in detected
out to date finds reduced fidelity (Bogdanov et al., 2017). photons relative to a 0.4 numerical aperture air objective. From
The authors surmise that charge effects related to dense Ref. (Le Sage et al., 2012).
NV- ensembles may contribute to the observed fidelity de-
crease. Achieving high Purcell factors for NV- ensemble
applications may also impose undesirable geometric con- proach to be effective, the lower singlet state 1 E lifetime
straints. For a clear and detailed discussion of radiative ∼ 140 − 220 ns at room temperature (Acosta et al., 2010b;
lifetime engineering for NV- spin readout, we recommend Gupta et al., 2016; Robledo et al., 2011) makes measuring
Ref. (Hopper et al., 2018b). Along similar lines, a re- the 1 E population via absorption on the 1 E ↔1 A1 transi-
cent proposal suggests that NIR fluorescence-based read- tion at 1042 nm viable.
out could exhibit improved fidelity over conventional read- Near-infrared (NIR) absorption has attractive benefits
out when combined with Purcell enhancement (Meirzada for certain applications: a) Contrast (and therefore sen-
et al., 2019). While this scheme requires high IR excitation sitivity) is not reduced by background fluorescence from
intensities likely incompatible with large NV- ensembles, non-NV- defects (such as NV0 ). b) The directional nature
it may show utility for small ensembles within a confocal of the 1042 nm probe light allows maximal collection ef-
volume. ficiency (ignoring absorptive losses) to be obtained more
easily than in a fluorescence-based measurement; for exam-
ple, this benefit was exploited in the first demonstration
F. Near-infrared absorption readout of microwave-free magnetometry with NV- centers (Wick-
enbrock et al., 2016). c) Owing to the upper singlet 1 A1
The sensitivity of conventional fluorescence-based read- lifetime of . 1 ns (Acosta et al., 2010b; Ulbricht and Loh,
out is limited by shot noise on the collected photons due 2018), the saturation intensity of the 1 E ↔1 A1 transition
to low fluorescence contrast C (see Eqn. 11). As an alter- sat
is unusually large (I1042 ∼ 50 megawatts/cm2 (Dumeige
native to fluorescence-based readout, population in one or et al., 2013)), allowing high intensity 1042 nm probe radi-
both NV- singlet states may be directly probed via absorp- ation to be used, so that fractional shot noise is reduced to
tion, giving a probabilistic measure of the initial ms spin well below that of an equivalent fluorescence-based mea-
state prior to readout. While the upper singlet state 1 A1 surement (Acosta, 2011). d) NIR absorption readout is
lifetime . 1 ns at room temperature (Acosta et al., 2010b; nondestructive, allowing for a single NV- center in the 1 E
Ulbricht and Loh, 2018) is likely too short for such an ap- singlet state to absorb multiple 1042 nm photons before
 35

degree of spin initialization into the ms = 0 state. Ulti-

a +1 b mately the device in √ Ref. (Jensen et al., 2014) achieves a
-1
3
E Diamond sensitivity√of 2.5 nT/ Hz, well above the shot-noise limit
mounting Photodiode
0 of 70 pT/ Hz.
Dichroic
1 Cavity
A notable recent implementation of NIR absorp-
A1 tion (Chatzidrosos et al., 2017) is depicted in Fig. 27b.
Pirror
532 One diamond face forms a reflector while the addition of
nm
532 1042 a dual-wavelength concave mirror results in an optical cav-
nm nm ity with a finesse of 160 and a cylindrical sensing volume
Filter
of ∼ 76 µm diameter and ∼ 390 µm length (Chatzidrosos
1042 et al., 2017). With 500 mW of 532 nm radiation and 80 mW
1 nm Piezo
E of 1042
√ nm radiation, a DC magnetic field sensitivity of 28
+1 pT/ Hz is achieved with this compact setup, with a band-
Diamond
3
A2 -1 width of about
√ 530 Hz. The shot-noise-limited sensitivity
is 22 pT/ Hz and √ the spin-projection-noise-limited sensi-
ms=0
tivity is 0.43 pT/ Hz.
FIG. 27 Near-infrared (NIR) absorption readout. a) Level dia- The NIR absorption approach is hindered, however, by
gram for the NV- center. Population accumulating in the 1 E several non-idealities, which so far limit readout fidelity to
singlet state is probed via absorption of 1042 nm radiation values far from the spin projection noise limit, similar to
resonantly addressing the 1 E ↔1 A1 transition. b) Miniature conventional optical readout (i.e., σR = 65 for the NIR ab-
NIR cavity-enhanced diamond magnetometer as described in
Ref. (Chatzidrosos et al., 2017). Dashed black lines depict sur-
sorption approach in Ref. (Chatzidrosos et al., 2017) ver-
faces forming the dual-wavelength optical cavity. Components sus σR ≈ 67 for conventional readout in Ref. (Le Sage
pertaining to MW delivery are omitted for clarity. et al., 2012)). First, the predominantly non-radiative
decay of the 1 A1 singlet greatly reduces the absorption
cross section σ1042 of the 1 E ↔1 A1 transition compared
eventual decay to the 3 A2 triplet. In principle such ab- to a radiative-decay-only transition (Acosta et al., 2010b;
sorption by a single NV- center can allow readout fidelity Rogers et al., 2008). Estimates suggest σ1042 = 3+3 −1 ×
near the spin-projection limit, even in the presence of non- 10−18 cm2 (Dumeige et al., 2013; Jensen et al., 2014; Ke-
negligible optical losses. hayias et al., 2013), whereas the purely radiative 3 A2 ↔3 E
NIR absorption readout has been successfully imple- transition is measured to have a much larger absorption
mented in several proof-of-principle magnetometers. In the cross section σ532 ≈ 3 × 10−17 cm2 for 532 nm excita-
first demonstration (Acosta et al., 2010), a diamond con- tion, as shown in Table 17. Realizing the full potential
taining [NV- ] ∼ 16 ppm is continuously illuminated with of this method requires 1042 nm laser intensities of order
532 nm radiation (driving the 3 A2 ↔3 E transition to op- sat
I1042 ∼ 50 MW/cm2 (Jensen et al., 2014). This satura-
tically polarize the NV- spin state) and 1042 nm NIR ra- tion intensity appears to limit interrogation cross sections
diation (resonantly addressing the 1 E ↔1 A1 transition), to .100 µm2 for ∼100 mW-scale 1042 nm radiation pow-
as shown in Fig. 27a. MW radiation transfers population ers, assuming a cavity finesse of ∼160. Laser intensities of
between the ground state Zeeman sublevels. In this first this magnitude may lead to undesirable ionization dynam-
demonstration (Acosta et al., 2010), a single pass of the ics (see Sec. VI.B). We note that many absorption cross
1042 nm radiation through the diamond sample resulted in section measurements neglect ionization/recombination dy-
a peak-to-peak contrast of ∼ 0.003 at room temperature. namics, which may skew reported values (Meirzada et al.,
The contrast can be enhanced by increasing the num- 2018, 2019). Second, as described in Ref. (Dumeige et al.,
ber of NV- defects interacting with each 1042 nm photon, 2013), non-resonant losses for 1042 nm radiation compro-
such as by employing a higher NV- density or a larger mise sensitivity by reducing the effective achievable collec-
diamond. Alternatively, for a fixed number of NV- cen- tion efficiency. For example, in Ref. (Chatzidrosos et al.,
ters, the 1042 nm radiation can be recirculated through 2017), 80 mW of 1042 nm radiation input to the dual-
the diamond. An NIR absorption magnetometer employ- wavelength cavity results in 4.2 mW transmitted to the
ing an optical cavity to increase the 1042 nm interaction detector. Third, the NIR absorption has only been demon-
length is analyzed in Ref. (Dumeige et al., 2013) and ex- strated for dense ensembles with [NVT ] ∼ 10 ppm to ensure
perimentally realizes a peak-to-peak contrast of 7.1% in appreciable 1042 nm absorption; the performance of this
Ref. (Jensen et al., 2014) when the bias magnetic field is method for diamonds with more dilute NV- concentrations
directed along the [100] crystallographic direction (mak- and longer T2∗ values remains unknown, and will likely de-
ing the magnetic resonances of all four NV- orientations pend on the scaling of cavity finesse with [NT ] or [NVT ]
degenerate). For this demonstration, the diamond is anti- density.
reflection coated (Yeung et al., 2012) and placed in the cen- While NIR absorption readout is effective and may find
ter of a dual-wavelength optical cavity, which allows recir- preference for certain applications (Chatzidrosos et al.,
culation of both 1042 nm and 532 nm radiation. The more 2017; Wickenbrock et al., 2016), without further ad-
efficient use of the 532 nm light enabled by the cavity allows vances enabling readout fidelity enhancement, (e.g., re-
both a larger NV- ensemble to be addressed and a higher duced 1042 nm non-resonant absorption or reduced non-
 36

radiative 1 A1 singlet decay rate), this method will remain

approximately on par with conventional fluorescence read-

Normalized Intensity (arb. units)

Infrared
out while requiring the non-trivial overhead of an NIR sin- Fluorescence
gle frequency laser and an optical cavity.
514 nm
Absorption
G. Green absorption readout 3
E Infrared
1
A Absorption
Red

Alternatively, NV- readout may be achieved by monitor- 514

nm
Fluor.
1042 Infrared
Fluorescence
nm

ing absorption of green probe laser radiation, which off- 1

E
Red
Fluorescence
resonantly drives the triplet 3 A2 ↔ 3 E transition (Bauch,
3
A2
2010; Walsworth, 2017). When resonant MWs drive the
ms = 0 ↔ ms = −1 or ms = 0 ↔ ms = +1 ground state Microwave Frequency (MHz)
spin transitions and facilitate population transfer to the
NV- singlet states, it is expected that the 3 A2 state will be FIG. 28 Simultaneous measurement of absorption and fluores-
depleted, resulting in increased green probe transmission cence on both the triplet and singlet NV- electronic transitions
and decreased red fluorescence. For absorption measure- (see inset lower left). For both transitions, the absorption and
fluorescence features have opposite signs and mirror one another
ments (both NIR and green), the change in transmitted up to a scaling factor. Adapted from Ref. (Bauch, 2010).
light upon resonant MW drive is expected to mirror the
change in fluorescence light up to a scaling constant, since
transmission is minimal when fluorescence is maximal and der resonant MW application (Bauch, 2010; Le Sage and
vice versa (Bauch, 2010). Data consistent with this under- Arai, 2011), leading to an anomalous inversion of the green
standing is shown in Fig. 28 for NV- centers illuminated absorption signal. This deviation from expected behav-
with 514 nm light. ior has been independently observed in multiple research
The absorption contrast, denoted Cabsorb , may differ sub- groups (Bauch, 2010; Le Sage and Arai, 2011). The anoma-
stantially in magnitude from the fluorescence contrast Cfluor lous Cabsorb reveals a strong wavelength and power depen-
(see Fig. 28). Because absorption measurements monitor dence (Bauch, 2010), which suggests that green absorption
transmitted light, the detected signal (and thus Cabsorb ) readout is hindered by an unknown effect competing with
depends on the optical depth of the absorbing material. and sometimes dominating otherwise expected behavior.
For example, even for the idealized case where Cfluor = 1, The wavelength and power dependence of this effect sug-
if only a small fraction of the incident light is absorbed in gests NV0 /NV- charge dynamics could play a role. Fur-
the absence of MWs, the change in transmission upon ap- ther investigation of this behavior might reveal presently
plication of resonant MWs will necessarily also be small, unknown NV dynamics. Overall, given the low absorption
yielding a low absorption contrast Cabsorb . Additionally, contrast Cabsorb , and yet unknown mechanism of anoma-
the absorption contrast may be further decreased due to lous absorption behavior, the utility of green absorption
the presence of non-radiative decay pathways. readout remains questionable.
Observed magnitudes of Cabsorb in the literature are lower
than Cfluor by ∼ 3× (Bauch, 2010; Le Sage and Arai, 2011)
or more. For example, the authors of Refs. (Ahmadi et al., H. Laser threshold magnetometry
2017, 2018a,b) use a CW-ODMR-based magnetometer em-
ploying a resonant optical cavity to recycle the green ex- Another approach for bulk NV- magnetometry is the cre-
citation light through the diamond multiple times, and ation of a NV-diamond-based laser threshold magnetome-
observe Cfluor ∼ 0.01, (which is typical), while measuring ter, as suggested by Ref. (Jeske et al., 2016). Lasing is
Cabsorb ∼ 10−6 . In Ref. (Ahmadi et al., 2018a) the same induced on the NV- 3 E (v = 0) ↔3 A2 (v 0 ≥ 1) transition;
experimental setup performs magnetometry simultaneously then, when a magnetic-field-dependent population accumu-
using both green absorption and red fluorescence, √ as shown lates in the singlet state, the lasing threshold increases, and
in Fig. 29. The green absorption yields ∼100 nT/ Hz sen- the laser’s output power is reduced. As theoretically out-
sitivity while the conventional
√ readout based on red fluo- lined (Jeske et al., 2017; Savitski, 2017), the laser thresh-
rescence reaches ∼ 400pT/ Hz, about 250× better. As old approach has a number of benefits relative to generic
with NIR absorption readout (see Sec. V.F), recycling the CW-ODMR methods (Sec. II.B.1): (i), effective contrast
green excitation light via a resonant optical cavity can in- is enhanced near the lasing threshold due to competition
crease the absorption signal by (i) addressing a larger NV- between stimulated and spontaneous emission, (ii) collec-
population, (ii) improving initialization into the ms = 0 tion efficiency is substantially improved by virtue of the
state, or (iii) enhancing Cabsorb . Although effectively im- lasing process. Although the emission cross sections for
plemented absorption readout may achieve higher optical NV- and NV0 have been measured (Fraczek et al., 2017),
collection efficiency than fluorescence detection, the low re- and stimulated emission from NV- was recently demon-
alized absorption contrasts are a current major drawback. strated (Jeske et al., 2017), substantial work remains to
Furthermore, absorption behavior for 532 nm probe ra- address potential problems. For example, absorption by
diation can result in increased probe laser transmission un- substitutional nitrogen or other defects may obstruct the
 37

Green
in the neutral and ionized charge states, NV centers in all
Modulation absorption three charge states, and other nitrogen-containing defects
signal
in the diamond, such as NVH (see Sec. VI.F). We define
MW source Lock-in the fraction of NV centers residing in the negative charge
Modulated
state as the charge state efficiency ζ,
microwaves Cavity
Cavity
532 nm
polarized Mirror mirror [NV- ] [NV- ]
Piezo ζ= = , (33)
light
[NVT ] [NV- ] + [NV0 ] + [NV+ ]
Diamond

Red
Photodiode so that Econv = χ · ζ. Although Refs. (Hauf et al., 2014;
fluroescence
Cavity lock
Pfender et al., 2017) show evidence for NV+ , this state has
photodiode
Photodiode
Photodiode so far required application of external voltages for observa-
tion. The rest of this section therefore assumes [NV+ ] is
Cavity lock
negligible and can be ignored.
To red
As the N-to-NVT conversion efficiency χ is determined
lock-in by the physical location of nitrogen and vacancies in the
diamond lattice, the value of χ is expected to be invariant
FIG. 29 Cavity-enhanced magnetometry based on green absorp-
tion as demonstrated in Refs. (Ahmadi et al., 2018a,b). A power under ambient conditions. Modification of χ requires condi-
build-up cavity allows green excitation light to pass through the tions severe enough to rearrange atoms within the diamond
diamond sample multiple times, increasing the effective path lattice, such as irradiation, implantation, high temperature,
length. The red fluorescence is measured simultaneously along or high pressure. With suitable electron irradiation and
with the green absorption. Adapted from Ref. (Ahmadi et al., subsequent annealing, N-to-NVT conversion efficiencies ap-
2018b). proaching 1 can be achieved, although such high values are
not necessarily desirable (see Secs. VI.D and VI.E).
In contrast, the charge state efficiency ζ depends on local
lasing process (Dodson et al., 2011), and it will need to
conditions in the diamond and can be affected by external
be shown that other sources of noise affecting the lasing
fields and optical illumination. Increasing ζ benefits sensi-
threshold or output power can be either controlled or nor-
tivity in two ways: first, by increasing the NV- concentra-
malized out (Jeske et al., 2016). More concerning, however,
tion and thus the number of collected photons N from the
is that both theory (DeGiorgio and Scully, 1970) and ex-
NV- ensemble; and second, by decreasing the concentration
periment (Lim et al., 2002) find large laser field fluctuations
of NV0 and the associated background fluorescence, which
in the vicinity of the lasing threshold.
improves measurement contrast. In the following section
we discuss factors contributing to the charge state efficiency
VI. DIAMOND MATERIAL ENGINEERING and methods to optimize it for sensing.

A. Conversion efficiency
B. NV charge state efficiency
In an idealized case in which all other parameters are held
constant, increasing the NV- density in a fixed detection The charge state efficiency ζ from Eqn. 33, depends on
volume will result in enhanced sensitivity. Since the NV- many factors both internal and external to the diamond.
density is limited by the density of nitrogen introduced into For both native NVs (Iakoubovskii et al., 2000) and NVs
the diamond, the N-to-NV- conversion efficiency created by irradiation and annealing of nitrogen-rich dia-
monds (Mita, 1996), the NV- and NV0 charge states can
[NV- ] coexist in a single sample. In general, for a given sample
Econv ≡ (31) and experimental procedure, the steady-state charge state
[NT ]
efficiency is difficult to predict. Contributing factors include
must be increased in order to achieve a high density of NV- the concentration of substitutional nitrogen and other de-
spins while minimizing the concentration of residual para- fects serving as charge donors or acceptors (Groot-Berning
magnetic substitutional nitrogen. Converting a substitu- et al., 2014) and their microscopic distributions (Collins,
tional nitrogen NS into a NV- defect requires both introduc- 2002; Doi et al., 2016); the wavelength, intensity, and duty
ing a vacancy to a lattice site adjacent to a substitutional cycle of optical illumination (Aslam et al., 2013; Doi et al.,
nitrogen (to create the NV), and capturing an electron (to 2016; Ji and Dutt, 2016; Manson and Harrison, 2005);
change the NV center’s charge state to NV- ). We denote the application of a bias voltage (Doi et al., 2014; Grotz
the efficiency with which nitrogen atoms in the diamond et al., 2012; Kato et al., 2013; Schreyvogel et al., 2014);
are converted to NVs as and, for near-surface NVs, the diamond surface termina-
tion (Chu et al., 2014; Cui and Hu, 2013; Fu et al., 2010;
[NVT ] Groot-Berning et al., 2014; Hauf et al., 2011; Kageura
χ= T
, (32)
[N ] et al., 2017; Newell et al., 2016; Osterkamp et al., 2015;
Rondin et al., 2010; Santori et al., 2009; Yamano et al.,
where [NT ] = [N0S ] + [N+ - 0 +
S ] + [NV ] + [NV ] + [NV ] + [N
other
] 2017). The charge state efficiency is likely affected by
accounts for the concentration of substitutional nitrogen NS
 38

the conditions of diamond growth, as well as the irradia- Moreover, irradiation and annealing to create NV centers
tion dose (Mita, 1996) (see Sec. VI.D), the annealing dura- may further convert desirable substitutional phosphorus
tion and temperature, and possibly the operation temper- into PVs (Jones et al., 1996). PVs in diamond will compete
ature (Manson and Harrison, 2005). Moreover, the value with NVs for electrons, undermining the benefit of phosphor
of the charge state efficiency ζ during an NV- sensing pro- doping to the charge state efficiency. Additionally, PL emis-
cedure can be difficult to measure. NVs may be reversibly sion at wavelengths overlapping the NV- PL spectrum was
converted between NV- and NV0 by various optical and observed in phosphorus-doped diamond (Cao et al., 1995),
non-optical processes (Aslam et al., 2013; Bourgeois et al., further complicating the use of phosphorus in NV-diamond
2017; Khan et al., 2009). Because ζ is strongly affected sensing. For additional discussion see Ref. (Doherty et al.,
by the illumination laser intensity and wavelength (Aslam 2016).
et al., 2013; Bourgeois et al., 2017), characterization of ζ The irradiation and annealing procedures applied to in-
by methods such as Fourier-transform infrared spectroscopy crease the N-to-NVT conversion efficiency χ can also af-
(FTIR), ultraviolet-visible (UV-Vis) spectroscopy, and elec- fect the charge state efficiency ζ. In Type Ib diamonds
tron paramagnetic resonance (EPR) may be misrepresen- grown by high-pressure-high-temperature (HPHT) synthe-
tative of NV charge state behavior under the optical illu- sis (see Sec. VI.C), with [NS ] & 50 ppm, ζ approaching 1 is
mination employed in most NV-diamond sensing devices. seen after low- and moderate-dose irradiation and anneal-
ing (Manson and Harrison, 2005; Mita, 1996). As discussed
in Sec. VI.D, at higher irradiation doses, the NV0 concen-
1. Non-optical effects on NV charge state efficiency tration is seen to abruptly increase (Mita, 1996), which can
be attributed to the combination of an insufficient concen-
Here we discuss the charge state efficiency ζ in nitrogen- tration of nitrogen defects NS available to donate electrons
rich diamond in the absence of optical illumination. For to the increasing overall NV population, and an increase
shallow NVs, the charge state is strongly affected by the in deep acceptor states such as multi-vacancy defects (Pu
surface chemical termination (Cui and Hu, 2013; Fu et al., et al., 2001; Twitchen et al., 1999b).
2010; Groot-Berning et al., 2014; Hauf et al., 2011; Rondin
et al., 2010). Based on the work in Ref. (Groot-Berning
et al., 2014), surface termination should provide enhanced 2. Optical effects on NV charge state efficiency
charge state stability to a depth of at least 60 nm and pos-
sibly farther (Malinauskas et al., 2008; Santori et al., 2009). Optical illumination of diamond may also change the NV
The charge state efficiency ζ can also be controlled electri- charge state efficiency ζ through ionization of NV- to NV0
cally (Doi et al., 2014; Forneris et al., 2017; Grotz et al., and also recombination of NV0 back to NV- (Aslam et al.,
2012; Hauf et al., 2014; Karaveli et al., 2016; Kato et al., 2013; Hopper et al., 2016, 2018a; Manson and Harrison,
2013; Murai et al., 2018; Schreyvogel et al., 2015, 2014). 2005; Waldherr et al., 2011). The steady-state value of ζ
Because diamond is an approximately 5.47 eV wide band is seen to depend on the illumination intensity and wave-
gap insulator (Wort and Balmer, 2008), Ref. (Collins, 2002) length, although most of the reported measurements have
contends that an NV center’s charge state depends less on been taken on single NV centers (Aslam et al., 2013; Hop-
the position of the Fermi level and more on the distance to per et al., 2016; Waldherr et al., 2011). For example, an
the nearest charge donor. In nitrogen-rich diamonds, these excitation wavelength band from 510 nm to 540 nm was
donors are mainly substitutional nitrogen defects NS , and found to produce the most favorable single-NV charge state
the charge state efficiency ζ is seen to increase with the con- efficiency in steady state compared to longer and shorter
centration [NS ] (Collins, 2002; Manson and Harrison, 2005). wavelengths (Aslam et al., 2013). In particular, when sin-
Other defects in the diamond lattice can alter ζ as well; for gle NVs were illuminated by 532 nm light at intensities
example, in Ref. (Groot-Berning et al., 2014), the NVs in typical for pulsed sensing protocols (Hopper et al., 2016;
separate implanted regions containing phosphorus (an elec- Waldherr et al., 2011) or similar wavelength light at lower
tron donor) and boron (an electron acceptor), were seen to intensities (Aslam et al., 2013), a charge state efficiency
have increased, and respectively decreased, NV charge state ζ ∼ 70 - 75% was observed. However, the value of ζ un-
efficiencies. der these conditions is likely to differ for dense NV ensem-
Introduction of electron donors other than nitrogen into bles (Manson and Harrison, 2005; Meirzada et al., 2018).
diamond might appear to be a promising avenue for in- For example, measurements in Ref. (Manson and Harrison,
creasing the NV charge state efficiency. For example, phos- 2005) on an NV ensemble in a diamond with [NT S ] ∼ 70 ppm
phorus (Doi et al., 2016; Groot-Berning et al., 2014; Murai and [NVT ] ∼ 1 ppm show the charge state efficiency drop-
et al., 2018), with donor level 0.6 eV below the conduction ping to ∼ 50% as the 532 nm power approaches the satu-
band (Katagiri et al., 2004), is a shallower donor than nitro- ration power of the NV- optical transition. More study is
gen, which lies 1.7 eV below the conduction band (Farrer, required to determine the relative contributions to the NV
1969; Wort and Balmer, 2008). However, creating n-doped ensemble ζ of optical charge-state switching, the presence
diamond through introduction of phosphorus has proven of nearby charge donors/acceptors, and other effects.
difficult (Kalish, 1999), as phosphorus doping is correlated Recently, several studies on single NV centers have shown
with the introduction of a deep acceptor tentatively iden- improved optical initialization to NV- by applying near-
tified as the phosphorus vacancy (PV) (Jones et al., 1996). infrared radiation (NIR) in combination with the 532 nm
 39

green excitation light (Chen et al., 2017; Hopper et al.,

2016; Ji and Dutt, 2016). This enhanced charge-state ini-
tialization has been demonstrated with 780 nm CW radi- Conduction band
ation (Chen et al., 2017), 1064 nm CW radiation (Ji and 3E Ion
Dutt, 2016), and 900 nm - 1000 nm pulsed radiation, achiev- 1A iza
tion

Electron energy
1
ing in the third case ζ > 90% (Hopper et al., 2016). The NV-

Hole energy
effect is theoretically explained as follows: after absorption 1E
3A 2A
of a green photon to enter the electronically excited state, 2
an NV0 absorbs an NIR photon, which promotes a hole to
Re NV0
the valence band and forms NV- (Hopper et al., 2016; Ji com
bin
and Dutt, 2016). The mechanism, visualized schematically atio 2E
n
in Fig. 30, is the same as the two-photon ionization and
recombination of NV- and NV0 by 532 nm radiation, but Valance band
with the second absorbed photon being an NIR photon.
In Ref. (Hopper et al., 2016) the NV0 -to-NV- recombina-
tion process is found to occur with up to a ∼ 7× higher
likelihood than the analogous ionization process converting FIG. 30 Energy level diagrams for NV- and NV0 , represent-
NV- to NV0 , wherein the excited-state NV- absorbs an NIR ing optical ionization and recombination processes through ei-
photon, promoting an electron to the conduction band. ther absorption of two 532 nm photons [green arrows (→)] or a
532 nm photon and an NIR photon [brown arrows (→)]
NIR-enhancement of charge state efficiency is expected
to be compatible with pulsed initialization and readout.
However, when employing 532 nm intensities I ≈ Isat ≈ C. Diamond synthesis and high pressure high temperature
2.7 mW/µm2 (Wee et al., 2007) typical for pulsed exper- treatment
iments, Ref. (Hopper et al., 2016) finds enhancements in
ζ to be lessened compared to operation at lower green in- Fabricated bulk diamonds are commonly synthesized us-
tensity. Furthermore, if the charge switching rate under ing one of two methods. In high pressure high temper-
green-plus-NIR illumination approaches or exceeds the op- ature (HPHT) synthesis, a process mimicking natural di-
tical spin polarization rate, spin readout fidelity could be amond formation, a carbon source material is mechani-
degraded by the increased photoionization during the read- cally compressed (pressure > 5 GPa) and heated (tem-
out pulse. Refs. (Ji and Dutt, 2016) and (Chen et al., 2017) peratures & 1250 ◦ C) to create conditions where diamond
report charge switching rates near ∼ 1 µs-1 , approaching is the thermodynamically favored carbon allotrope. Dis-
the singlet state decay rate of 4 µs-1 (Acosta et al., 2010b). solving the carbon source (typically graphite) in a metal
Nonetheless, Hopper et al. achieve enhanced charge state "solvent-catalyst" can increase the growth rate, decrease
initialization with much lower charge switching rates of the required temperature and pressure, and allow for bet-
∼ 10 ms-1 . ter composition control. Consequently, solvent-catalysts
Further work is required to determine if the NIR-plus- are nearly always employed. A small seed diamond facili-
green illumination technique can be extended to increase tates the growth; the dissolved carbon precipitates out of
the charge state efficiency ζ in NV ensembles. While the the metal catalyst solution and crystallizes onto the seed
technique has shown success for near-surface NV centers, diamond, growing the size. Nitrogen easily incorporates
Ref. (Meirzada et al., 2018) observes no enhancement in into the diamond lattice, and is historically the primary
the NV- PL from NIR-plus-green illumination compared impurity element in HPHT diamonds. However, nitrogen
to green-only excitation for NV centers in bulk diamond. content in HPHT-synthesized diamonds can be reduced by
Moreover, even if NIR-plus-green illumination can enhance varying the atomic composition of the metal solvent cata-
the ensemble value of ζ, the power requirements may limit lyst to “getter” the nitrogen, and recent advances in getter
the technique’s application to large ensembles. Although technology have allowed direct creation of electronic grade
the required NIR power for confocal setups addressing sin- HPHT diamond with [N0S ] . 5 ppb (D’Haenens-Johansson
gle NV- centers or small ensembles is modest (∼ mW), the et al., 2015; Tallaire et al., 2017b). References (Dobrinets
NIR intensity is & 10× higher than the typical 532 nm in- et al., 2013; Kanda, 2000; Palyanov et al., 2015) discuss
532 nm
tensities used for NV- spin initialization (INIR ≈ 23 Isat HPHT synthesis in detail.
in Ref. (Hopper et al., 2016)). Thus, when applying the Plasma-enhanced chemical vapor deposition (PE-CVD)
technique to macroscopic ensemble volumes, the maximum diamond synthesis (Angus et al., 1968) is a popular alterna-
addressable ensemble size will quickly be limited by the tive to HPHT synthesis, and can leverage established semi-
available laser power. For example, a (50 µm)2 spot would conductor fabrication techniques. In the most widespread
require & 100 W of NIR (Wee et al., 2007). At present, variant of this method, employing homo-epitaxial growth, a
NIR-enhancement of charge state efficiency appears un- diamond seed is exposed to a hydrocarbon plasma consist-
likely to yield substantial improvements to ensemble-NV- ing of approximately 95 - 99% hydrogen, with the balance
magnetometer sensitivities. composed of carbon and possibly other species such as oxy-
gen or argon. Methane is the most popular carbon source.
Radicalized carbon atoms bond with the growth surface,
 40

forming a mixture of sp2 and sp3 bonded orbitals. Although ing synthesis (Hounsome et al., 2006; Khan et al., 2013).
hydrogen etches both sp2 and sp3 bonded carbon, the etch As vacancy clusters, chains, and rings are typically para-
rate for sp2 bonded carbon is much greater (Schwander and magnetic (Baker, 2007; Iakoubovskii and Stesmans, 2002;
Partes, 2011) and, if the hydrogen etching and carbon de- Lomer and Wild, 1973; Yamamoto et al., 2013b), these
position rates are carefully tuned, diamond synthesis can clusters can increase NV- ensemble dephasing, reducing
be achieved (Angus et al., 1968). Unlike HPHT synthe- T2∗ . Additionally, since such vacancy chains and clusters
sis, PE-CVD (alternatively simply called CVD) synthesis are deeper electron acceptors than NV- (Edmonds et al.,
can easily allow the production of thin or delta doped lay- 2012; Khan et al., 2009), their presence may decrease mea-
ers from nanometer to micron scale (McLellan et al., 2016; surement contrast (Tallaire et al., 2017a). Naturally oc-
Ohno et al., 2012), masked synthesis of diamond struc- curring diamond that has not undergone irradiation rarely
tures (Zhang and Hu, 2016), or layered epitaxial growth contains vacancies (Mainwood, 1999), suggesting that va-
required for PIN (Kato et al., 2013) or NIN structures (Mu- cancies and vacancy clusters should be uncommon in well-
rai et al., 2018). synthesized HPHT diamond. As point defects, disloca-
In the past 15 years, the majority of NV-diamond liter- tions, and other extended defects are believed to be the
ature has employed diamonds grown by PE-CVD. First, dominant sources of strain in Type IIa diamonds (Fisher
much early work focused on single NV- centers; and et al., 2006), HPHT-synthesized diamonds may also exhibit
most HPHT-synthesized diamonds were not available at lower strain than their CVD-grown counterparts. While
that time with the requisite low nitrogen concentration dislocation densities of ≈ 104 - 106 cm-2 are typical in
(. 100 ppb), as HPHT impurity control can be chal- CVD-grown diamonds (Achard et al., 2014), certain HPHT-
lenging (Gaukroger et al., 2008; Martineau et al., 2009). synthesized diamonds can demonstrate dislocation densi-
Second, the layered deposition inherent to CVD allows ties of ≈ 100 - 1000 cm-2 (Martineau et al., 2009; Tallaire
straightforward growth of epitaxial layers (as would be re- et al., 2017b) and substantially lower strain (D’Haenens-
quired for magnetic imaging devices) and the application of Johansson et al., 2015, 2014).
semi-conductor techniques to control diamond composition. Although more research is needed, it is observed that
Third, the PE-CVD method was historically more popular the high quantity of hydrogen present during CVD growth
with commercial collaborators (such as Element Six and can result in hydrogen incorporation into the diamond lat-
Apollo Diamond) responsible for producing the majority of tice (Charles et al., 2004; Goss et al., 2014), (see Sec. VI.F).
scientific diamonds containing NV- centers. In contrast, diamonds synthesized directly by HPHT are
In addition, several challenges accompany direct HPHT unlikely to have hydrogen defects, as only one hydrogen-
synthesis of high-quality NV-diamonds. For one, solvent- related defect has been found to incorporate into HPHT-
catalyst incorporation into the diamond lattice may result synthesized diamond (Hartland, 2014).
in metal inclusions with size visible to the naked eye. Such Alternatively, mixed-synthesis approaches can combine
inclusions could be particularly problematic for magnetic the strengths of CVD and HPHT. One popular method
sensing applications, since the common materials employed is HPHT treatment, where an existing CVD diamond is
in the solvent catalyst alloys are the ferromagnetic ele- heated and subjected to high pressure, resulting in atomic-
ments Fe, Co, and Ni (Palyanov et al., 2015). The pu- scale reconfigurations of atoms in the lattice while leaving
rity of HPHT-synthesized diamonds may be limited by the the macro-scale diamond largely unchanged (Dobrinets
solid precursor materials, which may not be available with et al., 2013). HPHT treatment effectively removes sin-
as high chemical or isotopic purity as the gas-phase pre- gle vacancies (Collins et al., 2000; Dobrinets et al., 2013)
cursor elements employed for CVD synthesis. Finally, the and causes vacancy clusters to dissociate (Collins et al.,
HPHT process is not intrinsically amenable to fabrication of 2000; Dobrinets et al., 2013) or aggregate (Bangert et al.,
NV- -rich layers, as are needed for imaging applications. In 2009). Thus, this method is effective to treat CVD-grown
spite of these challenges, HPHT-fabricated diamonds with diamonds, which can exhibit vacancies and vacancy clus-
good characteristics for ensemble-NV- DC magnetometry ters (Charles et al., 2004; Hartland, 2014; Khan et al.,
- including long T2∗ (& 2 µs), high Econv (∼ 30%), and 2013). The approach of applying HPHT treatment to CVD
[NT ] ∼ 1-4 ppm - have been recently reported in the liter- diamonds was proposed in Ref. (Twitchen et al., 2010) and
ature (Grezes et al., 2015; Stürner et al., 2019; Wolf et al., realized by the author of Ref. (Hartland, 2014), wherein
2015b) (see Table 6). a CVD-grown diamond was HPHT treated after synthe-
While the exact motivation for HPHT diamond synthesis sis but prior to irradiation and subsequent annealing (see
is not always explicitly stated (Teraji et al., 2013), HPHT Secs. VI.D and VI.E). The diamond produced in Ref. (Hart-
synthesis may circumvent undesired characteristics inherent land, 2014) exhibits a notably high 30% conversion effi-
to CVD-synthesized diamonds (Charles et al., 2004; Hart- ciency Econv = [NV- ]/[NT ] as shown in Table 10. A simi-
land, 2014). A serious disadvantage of CVD synthesis is lar process pioneered by Lucent Diamonds employs HPHT
the incorporation of unwanted impurities and charge traps treatment of diamonds prior to irradiation and anneal-
into the lattice (see Sec. VI.F). In addition, CVD-grown ing (Vins, 2007). This process results in a final material
diamonds may display undesirable strain non-uniformities with an intense red hue and photoluminescence dominated
or contain a high dislocation density. For example, CVD- by NV- emission (Dobrinets et al., 2013; Wang et al., 2005),
grown diamonds sometimes exhibit a brown coloration, suggesting that HPHT treatment can be effective to in-
which is attributed to vacancy cluster incorporation dur- crease the charge state efficiency ζ, likely by eliminating
 41

Reference T2∗ T2 Econv [NV- ] [NV0 ] [NT ] [13 C] Synthesis

(Grezes et al., 2015) ∼ 2.6 µs 84 µs 29% 0.4 ppm 0.4 ppm 1.4 ppm 300 ppm HPHT
(Wolf et al., 2015b) - ∼ 50 µs 30% 0.9 ppm - 3 ppm - HPHT
(Zheng et al., 2019) ≥ 1.4 µs - - ∼ 0.9 ppm - > 2.9 ppm 300 ppm HPHT
(Hartland, 2014) - - 28% 1.2 ppm 0.7 ppm 4.1 ppm 10700 ppm CVD+HPHT
(Schloss, 2019) 1.55 µs 15.7 µs ∼ 30% ∼3 ppm - ∼ 10 ppm 100 ppm CVD
(Barry et al., 2016) 580 ns 5.1 µs 6.3% ∼ 1.7 ppm - 27 ppm 10 ppm CVD
(Schloss et al., 2018) 450 ns 7 µs ∼ 14% 3.8 ppm 2.0 ppm ∼ 28 ppm 10700 ppm CVD

TABLE 6 Partial literature survey of diamonds with properties well-suited to ensemble-NV- magnetometry. Diamonds with long
T2∗ , high N-to-NV- conversion efficiency Econv , and [NV- ] & 1 ppm, are expected to be particularly favorable for high sensitivity
magnetometry applications. Dashed lines (-) indicate values not reported or unknown.

charge traps. radiating particles knock carbon atoms out of the diamond
However, HPHT treatment cannot address all diamond lattice, creating both interstitial carbon atoms and mono-
deficiencies, CVD-related or otherwise. For example, vacancies (Newton et al., 2002; Twitchen et al., 1999a). Al-
should a CVD-synthesized diamond incorporate high con- though theoretical calculations have not yet completely con-
centrations of hydrogen or other elemental impurities into verged with experimental observations (Deák et al., 2014;
the diamond lattice during growth, HPHT treatment is in- Zaitsev et al., 2017), the widely accepted model posits that
effective to remove these impurities (Charles et al., 2004). upon subsequent annealing (discussed in Sec. VI.E), dif-
Such treatment is also limited to diamonds with balanced fusing vacancies are captured by substitutional nitrogen
aspect ratios, as thin plates or rods will likely crack under atoms, forming NV centers (Acosta et al., 2009). Primary
the high applied pressure. considerations in the irradiation process are the particle
In addition to HPHT treatment of existing diamonds, type, energy, and dose.
other mixed-synthesis approaches have also been pursued. The irradiation of diamond has been performed using
For example, utilizing type IIa HPHT seeds for CVD a variety of particles: protons, ionized deuterium atoms,
growth rather than CVD-grown seeds can yield material neutrons, and electrons (Ashbaugh III, 1988). Gamma ray
with lower strain and reduced densities of dislocations and irradiation from 60 Co has also been used (Ashbaugh III,
other unwanted defects (Gaukroger et al., 2008; Hoa et al., 1988; Campbell and Mainwood, 2000). Many of these par-
2014; Martineau et al., 2009; Tallaire et al., 2017b). An- ticles are suboptimal for NV creation, however, where only
other mixed-synthesis method exploits the fine composi- single monovacancies V0 are desired, and other created
tion control and high chemical purities available with CVD defects are likely deleterious. A particular problem for
synthesis to create the carbon precursor for HPHT syn- certain irradiation methods is the production of “knock-
thesis (Teraji et al., 2013). The diamond composition can on-atoms” (Campbell and Mainwood, 2000; Davies et al.,
thus be carefully controlled, and HPHT synthesis can take 2001), where the irradiating particle has sufficient energy
advantage of high-purity or isotopically enriched gaseous not only to displace an initial carbon atom from the lattice,
sources (e.g., methane or 15 N2 ). but to impart enough kinetic energy to that carbon that it
Given the prominent role lattice defects and elemental displaces additional carbon atoms, resulting in localized lat-
impurities play in determining the charge state efficiency tice damage (Buchan et al., 2015). Although annealing (see
and coherence times for NV- , additional research focused Sec. VI.E) can partially alleviate such damage, the lattice
on synthesizing sensing-optimized diamonds is warranted. damage can never be completely repaired (Balmer et al.,
2009; Fávaro de Oliveira et al., 2017, 2016; Lobaev et al.,
2017; Twitchen et al., 2010) and may result in unwanted
D. Electron irradiation paramagnetic defects or charge traps. For irradiation with
protons, neutrons, or ionized deuterium atoms, damage
For unmodified as-grown CVD diamond, realized con- from such knock-on-atoms can be severe. Similar lattice
version efficiency values Econv can be far less than unity, damage occurs from ion implantation of various species
as shown in Table 7, where the majority of substitutional such as nitrogen (Fávaro de Oliveira et al., 2017; Naydenov
nitrogen is not converted to NV- (Edmonds et al., 2012; et al., 2010; Yamamoto et al., 2013b), carbon (Naydenov
Hartland, 2014). In fact, for some CVD diamonds (see et al., 2010; Schwartz et al., 2011), and helium nuclei (Him-
Table 8) the concentration of grown-in monovacancies is ics et al., 2014; Kleinsasser et al., 2016; McCloskey et al.,
insufficient to achieve good Econv for total nitrogen con- 2014; Schwartz et al., 2011; Waldermann et al., 2007). Elec-
centration [NT ] & 1 ppm regardless of location; even if trons, with their lower mass, transfer less kinetic energy to
every monovacancy were adjacent to a substitutional nitro- the carbon atoms and are therefore better suited to creat-
gen, the conversion efficiency Econv would still be low (Deák ing isolated monovacancies. Electron irradiation is favored
et al., 2014; Mainwood, 1999). However, the monovacancy over gamma ray irradiation because the former can be ac-
concentration can be augmented after growth by irradiating complished in hours whereas the latter, when implemented
the diamond with energetic particles. The high-energy ir-
 42

Econv [NT ] Growth location Reference

0.0007 − 0.005 0.3 − 30 ppm Element Six (Edmonds et al., 2012)

0.0006 − 0.03 0.35 − 2.4 ppm Apollo Diamond Inc. (Edmonds et al., 2012)
0.02 − 0.03 4 ppm Warwick University (Hartland, 2014)

TABLE 7 Native N-to-NV- conversion efficiencies Econv and total nitrogen concentrations [NT ] in unmodified bulk CVD diamond

[V0 + V- ] Reference The irradiation dose should also be approximately

matched to the diamond’s total nitrogen concentration [NT ]
as suggested in Sec. VI.A; if too many vacancies are created,
60 ppb (Rutledge and Gleason, 1998)
then > 50% of Ns will be converted to NV0 , and the num-
. 20 ppb (Twitchen et al., 2010) ber of electrons donated by the remaining Ns will be insuffi-
. 0.03 ppb (Mainwood, 1999) cient to convert every NV0 to NV- . Figure 2 in Ref. (Mita,
1996) illustrates the importance of matching the irradia-
TABLE 8 Native monovacancy concentrations in unmodified tion dose to [NT ] to achieve maximal Econv . When de-
bulk CVD diamond termining irradiation dose, in-situ recombination between
a vacancy and an interstitial carbon should be accounted
using 60 Co, can take weeks (Collins, 2007). In summary, for (Campbell and Mainwood, 2000; Davies et al., 2001).
electron irradiation is preferred to create NV- ensembles Current estimates suggest approximately 30% (Campbell
optimized for sensing applications (Campbell et al., 2002; and Mainwood, 2000) to 50% (Davies et al., 2001) of ini-
Twitchen et al., 2010; Uedono et al., 1999), as this method tially created vacancies are immediately lost to spontaneous
allows for evenly distributed monovacancies to be created recombination. For example, using 1 MeV electrons (gen-
throughout the diamond in a timely manner, with less lat- erating ∼ 2 × 10−4 vacancies/electron/µm according to
tice damage than alternative methods. Ref. (Campbell and Mainwood, 2000), and assuming 40% of
Theoretical calculations predict monovacancy creation vacancies recombine immediately and two nitrogens are re-
requires electron energies & 165 keV (Campbell et al., quired to make a single NV- center, we expect a sample with
2002), roughly consistent with experiments observing va- [NT ] ∼ 1 ppm to require a dose of 7.3×1016 cm-2 . However,
cancy creation down to 145 keV (Eichhorn et al., 2019; fine-tuning of the irradiation dose is often done empirically,
McLellan et al., 2016). Older, less reliable experiments find suggesting either the presence of dynamics more compli-
vacancy creation for electron irradiation along the [100] di- cated than those included in the simple model presented
rection at 180 keV but not 170 keV (Koike et al., 1992). here (i.e., the presence of other vacancy traps, the formu-
Crude estimates suggest electron irradiation energies lower lation of divacancies, loss at surfaces, etc.) or errors in the
than ∼ 0.8 MeV will create mainly single vacancies (Loub- measured electron flux or substrate temperature (Camp-
ser and van Wyk, 1978; Mitchell, 1965) and avoid produc- bell and Mainwood, 2000). For example, while the produc-
ing multi-vacancy complexes. While this estimate is con- tion rate of neutral monovacancies from irradiation with
sistent with Ref. (Dannefaer et al., 1992) where divacan- 2 MeV electrons is found to be temperature-independent
cies are detected after irradiation with 3.5 MeV electrons, from room temperature to ∼ 300 ◦ C, the rate decreases
Ref. (Twitchen et al., 2010), however, finds no evidence of notably for higher temperatures (Newton et al., 2002).
vacancy pairs after irradiation with 4.6 MeV electrons, sug-
gesting that several-MeV irradiation energies may be safe. E. Low pressure high temperature annealing
The optimal irradiation energy may also depend on sample
geometries; thicker diamonds should require higher ener- For the successful creation of NV- centers, substitutional
gies to ensure vacancies are created uniformly through the nitrogen and monovacancies must be relocated to occupy
entire thickness (Campbell and Mainwood, 2000; Twitchen adjacent sites in the diamond lattice. This process can be
et al., 2010). For small ensembles close to the diamond accomplished via diffusion at elevated temperature, i.e., an-
surface, an electron microscope can provide the needed ir- nealing. Since monovacancies migrate in the neutral charge
radiation (Farfurnik et al., 2017; Kim et al., 2012; McLel- state V0 (Breuer and Briddon, 1995) with an activation en-
lan et al., 2016; Schwartz et al., 2012). More study is re- ergy of Ea = 2.3 ± 0.3 eV (Davies et al., 1992; Mainwood,
quired to resolve remaining discrepancies between experi- 1999), compared to measured values of Ea = 4.8−6.2 eV for
mental data and detailed simulations of the electron irradi- substitutional nitrogen (Deák et al., 2014; Dobrinets et al.,
ation process (Campbell et al., 2002; Campbell and Main- 2013; Jones et al., 2015), neutral monovacancies diffuse
wood, 2000). For example, recent measurements of mono- throughout the lattice during annealing until they reach the
vacancy density profiles versus depth, as judged by GR1 more immobile nitrogens. The negatively charged mono-
intensities in 1 MeV electron irradiated diamonds (Zaitsev vacancy’s higher activation energy (Breuer and Briddon,
et al., 2017), are inconsistent with Monte Carlo simulations 1995) ensures that monovacancy diffusion occurs predom-
in Refs. (Campbell et al., 2002; Campbell and Mainwood, inantly in the neutral charge state (Breuer and Briddon,
2000).
 43

1995), although a negative monovacancy can convert to a consisting primarily of substitutional nitrogen (and, as the
neutral monovacancy in a reversible charge transfer pro- process progresses, NVs), thereby reducing divacancy for-
cess (Davies et al., 1992). The diffusion constant D of the mation. Although preliminary work in Ref. (Nöbauer et al.,
neutral monovacancy is (Hu et al., 2002; Orwa et al., 2012) 2013) finds electron irradiation with in-situ annealing in-
creases T2∗ , no increase in Econv is observed, and further
D = D0 e−Ea /kB T , (34) investigation is warranted.
Following NV formation, further LPHT annealing above
where kB is the Boltzmann constant, T is the tempera- 800 ◦ C may reduce strain or paramagnetic impurities re-
ture, and D0 is a diffusion prefactor (see Appendix A.10). sulting from lattice damage. For example, divacancies can
Measurements of the diffusion constant D have yielded combine into other defects at ∼ 900 ◦ C (Twitchen et al.,
∼ 1.1 nm2 /s at 750 ◦ C (Martin et al., 1999) and 1.8 nm2 /s 1999b). Reduction of a given defect species may be effected
at 850 ◦ C (Alsid et al., 2019), suggesting values for D0 in by consolidation into other larger defect species, which may
agreement with an independently measured upper bound be paramagnetic (Baker, 2007; Hartland, 2014; Lomer and
(Acosta et al., 2009), and in moderate agreement with first Wild, 1973; Yamamoto et al., 2013b). Annealing to tem-
principles theoretical calculations (Fletcher and Brown, peratures of 1000 ◦ C to 1200 ◦ C is shown to extend the
1953) (see Appendix A.10). Other sources, however find T2 of both single NV- centers (Naydenov et al., 2010; Ya-
or employ different values for D0 or Ea (Hu et al., 2002; mamoto et al., 2013b) and ensembles (Tetienne et al., 2018)
Onoda et al., 2017; Orwa et al., 2012), suggesting that fur- created by ion implantation. As this increase is attributed
ther measurements are warranted. Once an NV center is to a reduction in paramagnetic multi-vacancy defects (Teti-
formed, the deeper binding energy of the nitrogen-vacancy enne et al., 2018; Yamamoto et al., 2013b), improvement in
bond relative to the neutral vacancy ensures that the bound T2∗ is expected as well, though this expectation has not
vacancy does not diffuse away (Goss et al., 2005; Hartland, been systematically confirmed in experiment. Practically,
2014). this additional LPHT treatment is limited by the temper-
The procedure described here is commonly termed low ature at which NVs anneal out, which is typically around
pressure high temperature (LPHT) annealing to distinguish 1400 ◦ C to 1500 ◦ C (Hartland, 2014; Pinto et al., 2012;
it from high pressure high temperature (HPHT) anneal- Zaitsev, 2001) and can vary depending on the presence of
ing (discussed in Sec. VI.C). Given the role of diffusion in other defect species within the diamond (Zaitsev, 2001).
LPHT treatment, the annealing temperature and anneal- While a systematic study of annealing temperatures and
ing duration are important control parameters. A tem- durations is warranted for engineering optimal samples for
perature of ∼ 800 ◦ C is usually employed (Botsoa et al., ensemble-NV- sensing, a standard recipe for samples is at
2011), given that monovacancies become mobile around least several hours at ∼ 800 ◦ C followed by several more
600 ◦ C (Davies and Hamer, 1976; Davies et al., 1992; Ki- hours at ∼1200 ◦ C (Breeze et al., 2018; Chu et al., 2014;
flawi et al., 2007; Uedono et al., 1999), and annealing times Fraczek et al., 2017). Some example calculations for an-
of several hours are typical, e.g., 2 hours in Ref. (Acosta nealing are detailed in Appendix A.10.
et al., 2009), 4 hours in Ref. (Lawson et al., 1998), 8 hours
in Ref. (Twitchen et al., 2010), 12 hours in Ref. (Barry
et al., 2016), and 16 hours in Ref. (Fraczek et al., 2017). F. Other common impurities in synthetic or treated single
Diamonds with lower values of [NT ] are expected to require crystal diamond
longer annealing times due to the greater initial distances
between vacancies and substitutional nitrogens. A study Unwanted species in the diamond lattice can degrade
by Element Six found no observable deleterious changes in magnetometer performance by decreasing the NV charge
diamond properties between samples that were annealed at state efficiency ζ = [NV- ]/[NVT ], creating local magnetic
∼ 800 ◦ C for ∼ 8 hours and samples that were annealed at noise, or reducing the fraction of substitutional nitrogen
the same temperature for longer periods (Twitchen et al., NS converted to NV- . This section restricts detailed discus-
2010). This ∼ 800 ◦ C annealing step is typically performed sion to multivacancy clusters and NVH (Khan et al., 2013),
under vacuum or in a non-oxidizing, inert gaseous environ- species present in diamond at sufficient concentrations to
ment to avoid graphitization (Dobrinets et al., 2013). Un- likely affect NV spin and charge dynamics. Extended dis-
der vacuum, present understanding is that diamond graphi- cussion of other defects can be found in Refs. (Deák et al.,
tization begins roughly around 1500 ◦ C (Davies and Evans, 2014; Newton, 2007); see also Table 9 for relevant defects
1972). commonly found in diamond.
Although the 800 ◦ C LPHT treatment is effective to cre- Multivacancy clusters are common in some diamonds
ate NVs, unwanted defects may form as well. For exam- grown by chemical vapor deposition (CVD) (Hounsome
ple, diffusing monovacancies can combine to form divacan- et al., 2006; Pu et al., 2000), and are believed to cause
cies (Twitchen et al., 1999b), which are immobile at 800 ◦ C. the brown coloration in CVD-grown diamond (Fujita et al.,
As deeper electron acceptors than NVs (Deák et al., 2014; 2009; Hounsome et al., 2006). During CVD synthesis, the
Miyazaki et al., 2014), the presence of divacancies reduces diamond surface can become rough and stepped. When
Econv . To mitigate divacancy formation, electron irradi- these steps are rapidly covered with additional deposited
ation with in-situ (i.e., simultaneous) annealing has been material, small voids, i.e., clusters of vacancies, can be left
proposed (Nöbauer et al., 2013). Under such conditions, in the diamond (Hounsome et al., 2006; Khan et al., 2013).
single vacancies are continuously created in an environment
 44

Diamond defect Ground state spin Another common impurity in diamond is hydrogen,
which gives rise to many defects (Dobrinets et al., 2013;
N0S S = 1/2 Goss et al., 2014; Zaitsev, 2001). For typical CVD dia-
N+S S =0 mond growth, the plasma is composed predominantly of
NV- S =1 hydrogen (& 95%) (Tokuda, 2015), which can incorporate
NV0 S = 1/2 into single crystal diamond at concentrations as high as
NV+ S =0 1000 ppm (Sakaguchi et al., 1999). The hydrogen incor-
NVH- S = 1/2 poration rate into the lattice is partially dependent upon
NVH0 S =0 the diamond growth recipe (Tang et al., 2004), and fur-
N02 S = 0 (Tucker et al., 1994) ther investigation into the hydrogen quantity incorporated
N+2 S = 1/2 and methods to mitigate hydrogen incorporation is war-
N2 V- S = 1/2 ranted. Hydrogen-related defects may influence the NV
N2 V0 S =0 charge state (Hauf et al., 2011; Lyons and de Walle, 2016).
N3 V0 S = 1/2 Additionally, at high enough concentrations the nuclear
N2 VH0 S = 1/2 spin of hydrogen may result in non-negligible dephasing or
VH- S =1 decoherence. At present we are unaware of any published
VH0 S = 1/2 method to effectively remove hydrogen from the bulk dia-
mond lattice (Charles et al., 2004; Hartland, 2014).
Vn H- S =1
The presence of hydrogen in the diamond lattice can en-
V- S = 3/2 (Baranov et al., 2017)
able formation of the NVH defect (Glover et al., 2003),
V0 S = 0 (Baranov et al., 2017)
wherein the hydrogen occupies the vacancy of an NV.
V+ S = 1/2 (Baranov et al., 2017)
In as-grown nitrogen-enriched CVD diamond, the ratio of
VV- S = 3/2 (Kirui et al., 2013)
([N+ 0 - -
S ]+[NS ]):[NVH ]:[NV ] was found to be approximately
VV0 S = 1 (Twitchen et al., 1999b) 300:30:1 in Ref. (Edmonds et al., 2012) and 52:7:1 in
Ref. (Hartland, 2014). The NVH species is undesirable be-
cause: (i) it lowers the conversion efficiency of incorporated
TABLE 9 Common defects in diamond and their ground state
electronic spin nitrogen to NV centers; (ii) it reduces the concentration of
substitutional nitrogen NS available to donate electrons to
turn NV0 defects into NV- ; (iii) NVH competes with NV as
Multivacancy cluster incorporation has been observed to an electron acceptor; (iv) NVH- is paramagnetic, causing
increase at high growth rates (Hounsome et al., 2006), and magnetic noise; and (v) the hydrogen in NVH may rapidly
may be correlated with nitrogen content (Pu et al., 2000). tunnel among the three adjacent carbon atoms at GHz fre-
Using positron annihilation, the authors of Ref. (Dannefaer quencies, resulting in high-frequency magnetic or electric
et al., 1993) found the density of multivacancy clusters was noise (Edmonds, 2008).
found to be roughly 1017 −1018 cm-3 for their growth condi- No known treatment can transform existing NVH defects
tions. Such vacancy clusters can trap electrons (Campbell into NV defects. The NVH complex is stable against an-
et al., 2002; Deák et al., 2014; Edmonds et al., 2012; Jones nealing up to approximately 1600 ◦ C but anneals out com-
et al., 2007), reducing the ratio of NV- to NV0 and also pletely by 1800 ◦ C (Hartland, 2014; Khan et al., 2013).
generating magnetic noise resulting from their trapped un- However, removal of NVH via annealing is not associated
polarized electron spins. The neutral divacancy V02 (Deák with increased NV concentration; rather, further isochronal
et al., 2014; Lea-wilsonf et al., 1995; Slepetz and Kertesz, annealing to 2000 ◦ C and 2200 ◦ C is accompanied by in-
2014; Twitchen et al., 1999b) and neutral multivacancy creases in N2 VH0 and N3 VH0 species (Hartland, 2014), sug-
chains (V0n , n ≥ 3) are paramagnetic (Baker, 2007; Iak- gesting the NVH concentration is reduced via aggregation
oubovskii and Stesmans, 2002; Lomer and Wild, 1973), and of NVH with one or more nitrogen atoms. NVH0 exhibits
increase environmental magnetic noise. Irradiation or im- absorption at 3123 cm-1 (Cann, 2009) but is otherwise not
plantation followed by annealing can also produce such de- known to be optically active.
fects (Naydenov et al., 2010; Yamamoto et al., 2013b). Diamonds subject to temperatures at which substitu-
Low pressure high temperature annealing is effective to re- tional nitrogen or interstitial nitrogen become mobile may
move certain multivacancy clusters. However, as the re- exhibit defects consisting of aggregated nitrogen, such
moval of multivacancy clusters is effected by aggregating as N2 (Boyd et al., 1994; Davies, 1976; Tucker et al.,
these species together or combining them with other de- 1994), N2 V (Green et al., 2015), N2 VH (Hartland, 2014),
fects, the reduction of smaller multivacancy defects may be N3 V (Green et al., 2017), N3 VH (Hartland, 2014; Liggins,
accompanied by an increase in larger multivacancy clusters 2010), N4 V (Bursill and Glaisher, 1985), or other aggre-
or other defects. High pressure high temperature (HPHT) gated nitrogen defects (Goss et al., 2004). The presence of
treatment effectively removes single vacancies (Dobrinets aggregated nitrogen defects reduces the quantity of nitro-
et al., 2013) and causes some vacancy clusters to dissoci- gen available to form NV centers or donate electrons to NV0
ate (Dobrinets et al., 2013), which may aggregate to form to form NV- , and can cause additional paramagnetic noise.
different multivacancy clusters (Bangert et al., 2009). See Other defects such as VH (Glover, 2003; Glover et al., 2004),
Sec. VI.C. V2 H (Cruddace, 2007; Shaw et al., 2005), and OV (Cann,
 45

2009; Hartland, 2014) have been identified in synthetic dia- VII. MISCELLANEOUS SENSING TECHNIQUES
mond and may act as charge acceptors or create additional
paramagnetic noise. However most defects discussed in this A. Rotary echo magnetometry
paragraph are observed at concentrations low enough to
be neglected for diamonds fabricated for NV- magnetome- Broadband magnetometry can also be performed us-
try, as shown in Table 10, reproduced from Ref. (Hartland, ing a MW pulse scheme called rotary echo (Aiello et al.,
2014). Additional defect species are inferred to exist from 2013; Mkhitaryan and Dobrovitski, 2014; Mkhitaryan et al.,
charge conservation arguments but have not been directly 2015). In this technique pioneered by Aiello et al. (Aiello
observed (Khan et al., 2009). More research is needed to et al., 2013), rotary echoes are produced by periodic rever-
better understand defects in synthetic diamond grown for sals of the driving field. The simplest protocol inverts the
magnetometry applications. phase of the driving field to reverse the sign of the Rabi
oscillations. The rotary echo technique may have utility for
certain niche applications such as event detection (Aiello
G. Preferential orientation et al., 2013), but the method so far yields worse sensitivity
than a Ramsey protocol. Like other dynamical-decoupling-
In naturally occurring and many fabricated diamonds, type methods, rotary echo can be tailored to reject noise
NV- centers are distributed evenly among all four crys- at certain frequencies and also has applications for certain
tallographic orientations. However, under certain cir- narrowband AC sensing, such as detection of individual nu-
cumstances, CVD-grown diamond can exhibit preferential clear spins (Mkhitaryan et al., 2015).
orientation of NV- centers along certain crystallographic
axes (Edmonds et al., 2012; Pham et al., 2012b). Sev-
eral research groups have achieved almost perfect alignment B. Geometric phase magnetometry
of all NV- centers along the a single [111] axis. Michl et
al. demonstrated 94% alignment (Michl et al., 2014), Lesik In the presence of particular DC and RF magnetic
et al. demonstrated 97% alignment (Lesik et al., 2014), and fields, an NV- spin may accumulate a measurable geo-
Fukui et al. demonstrated 99% alignment (Fukui et al., metric phase (Berry, 1984) in addition to a dynamical
2014). The mechanism for preferential orientation is ex- phase. Following demonstrations of control and readout
plained in Ref. (Miyazaki et al., 2014). of an NV- center’s geometric phase (Arroyo-Camejo et al.,
An ensemble-NV- magnetometer utilizing a single NV- 2014; Maclaurin et al., 2012; Yale et al., 2016; Zhou et al.,
orientation in a diamond with no preferential orientation 2017; Zu et al., 2014), the authors of Ref. (Arai et al., 2018)
suffers from reduced measurement contrast due to un- implemented geometric phase measurements for DC mag-
wanted PL from NV - centers of other orientations. A netometry. In their protocol, depicted in Fig. 31, the phase
diamond with 100% preferential orientation may allow a of a MW Rabi drive is swept adiabatically around a closed
4× increase in contrast. In practice, though, the enhance- phase-space loop during two intervals separated by a central
ment is typically somewhat less than 4×, since polarized π-pulse. Whereas the π-pulse cancels the dynamic phase
excitation light can already be used to selectively address accumulated during the sequences, the acquired geometric
particular NV- orientations (Lesik et al., 2014), and high phase depends on the strength of the DC magnetic field.
bias fields can suppress fluorescence from off-axis NV- cen- While this technique enables wide-dynamic-range field sens-
ters (Epstein et al., 2005; Tetienne et al., 2012). ing by avoiding a 2π phase ambiguity inherent to Ramsey
Diamonds grown with preferential orientation have at magnetometry, it is unlikely to enhance sensitivity with re-
least two main drawbacks. First, NV- concentrations for spect to optimized Ramsey.
preferentially grown diamonds in the literature are cur-
rently relatively low (Fukui et al., 2014; Lesik et al., 2014;
C. Ancilla-assisted upconversion magnetometry
Michl et al., 2014), typically around 1012 cm-3 although
concentrations up to 1015 cm-3 have been achieved (Tahara
A clever and novel magnetometry scheme pioneered by
et al., 2015). Second, it appears that the N-to-NV- con-
Ajoy et al. in Ref. (Liu et al., 2019) utilizes frequency up-
version efficiency cannot be increased through irradiation
conversion via an ancilla nuclear spin to make broadband
and subsequent annealing without destroying the preferen-
measurements of an external magnetic field. The method
tial alignment, although conflicting evidence on this topic
works as follows: A large magnetic field is aligned along the
has been reported (Fukui et al., 2014). Since electron ir-
NV- internuclear axis and tuned to near the ground state
radiation followed by annealing can increase the N-to-NV
level anti-crossing (GSLAC) at ≈ 1024 gauss, allowing the
conversion efficiency by ∼ 10× to 100×, preferential ori-
relative strengths of the Zeeman term and the hyperfine
entation is not currently believed to be a viable method to
coupling of the NV- electronic spin to the ancilla nuclear
achieve better ensemble magnetometry sensitivity. However
spin to be precisely tuned. In this regime, the NV- elec-
it is possible that future technical advances or treatment
tronic spin is first-order insensitive to magnetic fields per-
could alter this understanding. Additionally, preferential
pendicular to the NV- symmetry axis. However, an applied
orientation precludes the implementation of vector magne-
transverse magnetic field B⊥ modulates the strength of the
tometry (Schloss et al., 2018).
hyperfine interaction, resulting in amplitude modulation of
 46

Defect As-grown 1500 ◦ C anneal Irradiation 800 ◦ C anneal

[N0S ] (ppb) 1620 (160) 1100 (100) 200 (20) 120 (15)
[N+S ] (ppb) 1500 (150) 2200 (250) 3000 (300) 1000 (100)
[NV0 ] (ppb) ≤ 10 ≤ 10 ≤ 10 695 (70)
[NV- ] (ppb) 60 (5) 40 (5) 35 (5) 1160 (120)
[NVH0 ] (ppb) 500 (50) 310 (30) 380 (40) 290 (30)
[NVH- ] (ppb) 405 (40) 200 (20) obscured 20 (5)
[N2 VH0 ] (ppb) < 0.1 22 (3) obscured 24 (5)
[Vn H- ] (ppb) 3.1 (1) ≤ 0.1 25 (3) 41 (4)

TABLE 10 Concentrations of quantifiable defects in sample GG1 in the as-grown state and after each treatment stage. From
Ref. (Hartland, 2014).

dephasing times are achieved primarily by decreasing the

effective gyromagnetic ratio (i.e., the ratio relating B⊥ to
an energy level shift) relative to the native NV- electronic
gyromagnetic ratio. Although the scheme enables vector
sensing from a single NV- center and may be compatible
with NV- spin ensembles, the method presently precludes
sensing from multiple NV- orientations. So far there has
been no experimental demonstration of improved sensitivity
using this method relative to that of an optimized Ramsey-
type equivalent. The requirement for ≈ 1000 gauss axial
fields is also disadvantageous and likely prevents utilization
of off-axis NV- centers for sensing.

FIG. 31 Comparison of dynamic and geometric phase magne- D. Techniques for the strong NV- -NV- interaction regime
tometry. For dynamic phase magnetometry (i.e., Ramsey), the
Bloch vector (blue arrow), is optically prepared and then rotated Dipolar interactions among NV- spins contribute to
by a π/2-pulse to the equator. The Bloch vector then precesses
ensemble-NV- dephasing, as described in Sec. III.G. When
about the fixed Larmor vector [orange arrow (→)] before being
mapped into a population difference by a second π/2-pulse and NV- centers comprise the majority of spin defects in dia-
read out optically. b) For geometric-phase magnetometry, the mond, or when a different majority spin species is decoupled
Bloch vector is optically prepared and then rotated to the equa- from the NV- centers via spin bath driving, NV- -NV- inter-
tor. Additional off-resonant driving then rotates the Larmor actions may degrade relaxation times T2∗ , T2 , and T1 (Choi
vector about the z-axis. As the spins precess, a geometric phase et al., 2017a), limiting the sensitivity of both DC and AC
proportional to the product of the solid angle (orange disk) and magnetometers. Measurement protocols that decouple or
the number of Larmor vector rotations is acquired in addition
leverage these like-spin interactions while retaining sensi-
to the dynamic phase. To cancel the dynamic phase while con-
tinuing geometric phase accrual, a π-pulse and a reversal of the tivity to magnetic signals offer an avenue to surpass this
off-resonant drive are inserted at the sequence midpoint. Lastly, sensitivity limit.
the Bloch vector is mapped onto a population difference by a Proposed techniques to improve sensitivity in the limit
second π/2-pulse and read out optically. From Ref. (Arai et al., of strong NV- -NV- interactions may be separated into two
2018). categories. Protocols in the first category mitigate dipolar
interactions between like spins to extend either the dephas-
ing time T2∗ (O’Keeffe et al., 2019) for DC sensing or the
the electronic spin energy level at the nuclear spin preces- coherence time T2 (Choi et al., 2017b) for AC sensing. How-
sion frequency. The modulation deviation is proportional ever, these techniques partially decouple the spins from the
to B⊥ . Thus, by performing standard AC magnetometry fields to be sensed, which may counteract the sensitivity
at the nuclear spin precession frequency, the magnitude of enhancement from T2∗ or T2 extension. Protocols in the
the perpendicular magnetic field B⊥ can be detected. second category harness like-spin interactions (Raghunan-
The technique is intriguing because (i), it allows the effec- dan et al., 2018) and may generate entangled many-body
tive gyromagnetic ratio of the sensor to be tuned and (ii), it states (Choi et al., 2017). Measurements of an entangled
enables the use of AC magnetometry techniques including spin state comprising N spins can beat the√standard quan-
dynamical-decoupling protocols to sense DC fields for dura- tum limit for spin projection noise (η ∝ 1/ N , see Eqn. 9),
tions on the order of T2 or longer (see Sec. IV.A). However, and may approach the Heisenberg limit (η ∝ 1/N ) (Choi
the method is expected (and observed) to upmodulate both et al., 2017; Gammelmark and Mølmer, 2011).
magnetic signals and magnetic11noise, including spin bath To illuminate the promise and challenges associated with
noise, to the AC measurement band. Further, the improved
 47

0'"1"&,"2&1"3' 4$&(5%$#$'1 6$&-7351 to the standard quantum limit. Even without improved
785,($(
:";'&, !!! !!! !!! AC magnetic sensitivity, the scheme is expected to provide
!"#$ an increased measurement bandwidth by enabling faster
field sampling than conventional sensing. When the domi-
nant noise limiting the NV- ensemble’s spin coherence time
!%&'()$%($*+"$,-

is instead set by spatially-correlated noise, such as dipo-

lar interactions with nearby magnetic dipoles of a differ-
ent species (Choi et al., 2017), enhanced AC magnetic field
!"#$
sensitivity from reduced spin projection noise may again
be possible. Although the protocol may also be compati-
FIG. 32 Schematic diagram of entanglement-enhanced sensing
ble with broadband DC magnetometry, the scaling of the
protocol proposed in Ref. (Choi et al., 2017). During the ini-
tialization, measurement, and readout steps, the amplitude of correlated ensemble’s effective T2∗ with entangled number
a transverse magnetic field Ω and the repetition frequency of of spins N remains unclear. Further investigation is re-
additional transverse-magnetic-field π-pulses are tuned. In the quired to determine if this protocol could yield a sensitivity
initialization stage, a correlated many-body spin state is gener- improvement over conventional DC magnetometry.
ated as Ω is decreased toward a quantum critical point at ΩC . While the approach proposed in Ref. (Choi et al., 2017)
At the end of the measurement period, an axial AC magnetic represents an important milestone towards magnetome-
field signal is mapped onto the total magnetization of the ensem-
try enhanced by NV- -NV- dipolar interactions, the pro-
ble, which for NV- centers can be detected using conventional
readout. From Ref. (Choi et al., 2017). tocol is expected to be challenging to execute. First,
the mean NV- -NV- separation distance hrNV- ,NV- i must
be small compared to the average distance to the nearest
entanglement-enhanced techniques, we focus on the specific paramagnetic defect hrNV- ,other i, but large compared to the
protocol proposed in Ref. (Choi et al., 2017). The tech- thickness L of any (quasi) two-dimensional NV- layer, i.e.,
nique, which is expected to be applicable to NV- centers, hrNV- ,other i>hrNV- ,NV- i>L. This hierarchy indicates that
utilizes strong like-spin interactions to create quantum cor- for a typical NV- -rich diamond with hrNV- ,NV- i ∼ 10 nm, the
related states sensitive to AC magnetic fields. The proposed NV- layer thickness L should be less than a few nanome-
scheme, outlined schematically in Fig. 32 generates entan- ters. Shallow nitrogen implantation into diamond (Glenn
glement within a 2D array of spins by first polarizing the et al., 2017; Pham et al., 2011) or nitrogen delta-doping
individual spins along a transverse magnetic field (which for during CVD growth (Ohno et al., 2012; Osterkamp et al.,
NV- centers may be a MW-frequency field) and then adi- 2015) can yield layers that approach the appropriate thick-
abatically decreasing the field toward a quantum critical ness. Crucially, this 2D requirement restricts the practical
point. For the measurement to be compatible with global NV- ensemble size, which may limit achievable sensitivity
NV- ensemble readout, the system approaches the quantum (see Sec. II.C and Appendix A.3) when considering wide-
critical point, generating entanglement, without crossing field imaging and bulk magnetometry applications. Second,
over the quantum phase transition to a Greenberger-Horne- since the preparation and readout steps require a slow adia-
Zeilinger (GHZ) state. In the measurement step, periodic batic field ramp, the practical requirement that these steps
transverse-magnetic-field π-pulses are applied to the ensem- occur within time T 2 limits the degree of achievable en-
ble, allowing axial AC magnetic fields to excite the many- tanglement. Consequently, the protocol will mostly likely
body system. The number of excitations detected after the entangle sub-ensembles much smaller than the total ensem-
transverse field is increased to its original value provides a ble size. Disorder (i.e., static field inhomogeneity) in the
measure of the strength of the AC magnetic field. ensemble, e.g., from the random positioning of NV- cen-
Importantly, the entangled state’s coherence time, de- ters, also restricts the maximum entangled sub-ensemble
noted T 2 , is no longer limited by like-spin interactions, but size. Both of these mechanisms are expected to increase the
by external noise. That is, if the coherence time T2 is sepa- measurement’s spin projection noise above the Heisenberg
rated into contributions from NV- -NV- dipolar interactions limit, further restricting the parameter regime where sen-
and from other noise (including spin-lattice relaxation) as sitivity enhancements are possible. In spite of the serious
challenges and limitations, the proposed technique remains
1 1 1 a promising first step toward practical schemes harnessing
= - - + , (35)
T2 T2 {NV -NV } T2 {other} the full quantum nature of NV- ensembles for sensitivity
enhancement.
then the entangled state’s coherence time T 2 is only a func- Additionally, with entangled and individually address-
tion of T2 {other}. Therefore, when NV- -NV- interactions able spins, a wealth of proposed quantum error correction
dominate, T 2 may be comparable to or exceed T2 , yield- (QEC) sensing schemes may become more experimentally
ing improved AC magnetic field sensitivity via both the feasible (Bonato et al., 2015; Cramer et al., 2016; Kessler
increased coherence time and reduced readout noise. et al., 2014; Layden and Cappellaro, 2018; Layden et al.,
However, when the noise on each spin in the entangled en- 2019; Unden et al., 2016; Waldherr et al., 2014). QEC pro-
semble is independent, T 2 is expected to diminish linearly tocols mitigate certain relaxation mechanisms by encod-
with the number of entangled spins N (i.e., T 2 ∝ 1/N ), ing quantum information into redundant degrees of free-
which at best
√ cancels the sensitivity enhancement obtained dom (Degen et al., 2017). Certain QEC implementations
from the 1/ N reduction in spin projection noise compared
 48

have already shown extension of NV- spin coherence (Unden in NV-rich diamonds. Among the factors limiting T2∗ are
et al., 2016), although all demonstrations so far have em- magnetic-field, electric-field, and strain gradients. Exter-
ployed at most a handful of spins. To date, most QEC stud- nal bias-magnetic-field gradients may be mitigated through
ies concentrate on methods to correct noise along a different experimental design. Whereas internal strain and electric-
axis from the signal, (e.g., errors that cause spin flips rather field gradients can be more difficult to eliminate outright,
than dephasing), limiting their use for extending T2∗ or T2 . the NV- ensemble can be made insensitive to such gradi-
Recently, however, QEC schemes to correct dephasing-type ents through operation at sufficiently strong bias magnetic
errors have been proposed (Layden and Cappellaro, 2018; fields (Sec. IV.D) and employment of double-quantum co-
Layden et al., 2019). herence magnetometry (Sec. IV.B). Ensemble-NV- T2∗ val-
In the near future, ensemble-NV- sensing at the stan- ues may also be limited by dipolar interactions with the
dard quantum limit is likely to outperform entanglement- diamond’s inhomogeneous paramagnetic spin bath. We de-
enhanced schemes, as argued in Ref. (Braun et al., 2018). termine the individual contributions to T2∗ from substitu-
Nonetheless, further development of these techniques re- tional nitrogen N0S electronic spins (Sec. III.D), 13 C nu-
mains an important endeavor toward enabling long-term clear spins (Sec. III.F), and NV- spins (Sec. III.G). Recent
sensitivity improvements approaching fundamental limits. experiments determine T2 - and T2∗ -dependencies on nitro-
gen concentration to better than 10% (Bauch et al., 2018;
Bauch et al., 2019). We suggest reducing the unwanted
VIII. CONCLUSION AND OUTLOOK bath-spin concentrations through (i) diamond growth us-
ing isotopically-purified 12 C (Sec. III.F), and (ii) diamond
The recent excitement accompanying quantum sensing treatment via optimized electron irradiation and annealing
with nitrogen-vacancy centers in diamond is well motivated, procedures (Sec. VI). We also identify spin bath driving us-
as NV-diamond sensors promise many advantages over al- ing strong, resonant RF fields as an effective measure to de-
ternative sensing technologies. NV- centers provide preci- couple N0S and other impurity spins from the NV- ensemble
sion and repeatability similar to atomic systems in a ro- (Sec. IV.C). Recent work implementing spin bath driving
bust solid-state package with less experimental complexity. combined with double-quantum coherence magnetometry
Furthermore, NV- -based devices can operate under ambi- in NV- ensembles demonstrates T2∗ extension by more than
ent conditions and record spatial variations at length scales 16× (Bauch et al., 2018). We expect continued progress on
inaccessible to most other quantum sensors. this front; one avenue opened up when T2∗ is increased to
Efforts to optimize performance of ensemble-NV- sensors the NV- -NV- dipolar interaction limit is the exploration of
are particularly warranted, as these devices at present have enhanced sensing techniques harnessing quantum entangle-
greater potential for improvement than other NV- sensing ment (Choi et al., 2017) (Sec. VII.D).
platforms. Historically, the NV-diamond community has In Sec. V we survey existing techniques to improve
focused on optimizing few- or single-NV- sensors, while the ensemble-NV- readout fidelity F = 1/σR , which, for con-
best demonstrated ensemble-NV- devices exhibit sensitivi- ventional fluorescence-based readout, is currently limited
ties orders of magnitude away from theoretical limits (Tay- to ∼ 0.015 (see Table 2). We analyze methods that allow
lor et al., 2008). readout fidelities for single NV- centers and small ensem-
Consequently, this work provides a comprehensive survey bles in nanodiamonds to approach the spin projection limit,
of methods for optimizing broadband magnetometry from including spin-to-charge conversion readout (Sec. V.A) and
DC to ∼ 100 kHz using ensembles of NV- centers. We ex- ancilla-assisted repetitive readout (Sec. V.C). However, no
plore strategies to enhance sensitivity toward physical lim- demonstrated method has substantially outperformed con-
its, both through highlighting key parameters (Sec. II.C) ventional fluorescence-based readout for large NV- ensem-
and through evaluating proposed methods to improve those bles (Table 2). Nonetheless, we anticipate that with careful
parameters. After identifying Ramsey magnetometry as experimental design and advances in diamond-sample en-
the most promising sensing protocol (Secs. II.A and II.B), gineering, fidelity-enhancement methods so far limited to
we focus on understanding and improving the spin dephas- single spins or small ensembles may be extended to large
ing time T2∗ (Secs. III and IV), the spin readout fidelity NV- ensembles. Additionally, given that any method em-
F ≡ 1/σR (Sec. V), and the host diamond material prop- ploying optical readout benefits from increased collection
erties (Sec. VI). Below, we summarize our analyses within efficiency, such optimizations (Sec. V.E) remain worthwhile
these broad categories, and we recommend areas where fu- for improving magnetometer sensitivity.
ture study could lead to improvements in magnetometer As optimal sensing techniques require co-development
sensitivity and performance. with diamond samples tailored to these techniques, this
Measurements employing Ramsey-type protocols with work reviews diamond fabrication and relevant material
NV- ensembles are limited by T2∗ , which presently remains properties in Sec. VI. In particular, we focus on meth-
orders of magnitude shorter than the physical limit of 2T1 . ods to engineer lab-grown diamond samples optimized
The magnetic field sensitivity improves nearly linearly with for ensemble-NV- magnetometry. We analyze growth via
T2∗ extension when the measurement overhead time is sig- chemical vapor deposition, high-pressure-high-temperature
nificant (tO & T2∗ ), as is common for present-day ensemble- synthesis, and mixed-synthesis methods (Sec. VI.C). We ex-
NV- magnetometers. Therefore, this work focuses on un- amine how diamond synthesis and treatment can be used
derstanding limitations to T2∗ and methods to extend T2∗ to engineer high N-to-NV- conversion efficiencies Econv , and
 49

we investigate methods to improve and stabilize the charge vancements possible before fundamental limits are reached.
state efficiency ζ = [NV- ]/[NVT ] (Sec. VI.B). We also in- By combining the knowledge collected here with likely fu-
vestigate undesired defects commonly found in NV-rich di- ture advances, we expect further expansion of applications
amond samples (Sec. VI.F). These defects, including multi- of quantum sensors based on NV- ensembles in diamond.
vacancy clusters and hydrogen-related impurities, may both
trap charges in the diamond and contribute to the dipolar
spin bath, reducing both Econv and T2∗ . Acknowledgments
Although present understanding of diamond synthesis,
treatment, and characterization is extensive and spans mul- The authors thank Eisuke Abe, Victor Acosta, Scott
tiple decades, further work is needed to reproducibly create Alsid, Ashok Ajoy, Keigo Arai, Nithya Arunkumar, Ania
NV- -rich diamond samples with low strain, low concentra- Bleszynski-Jayich, Dolev Bluvstein, Danielle Braje, Do-
tions of unwanted impurities, and high NV- concentrations. minik Bucher, Alexei Bylinskii, Paola Cappellaro, Soon-
In particular, advancing diamond materials science to en- won Choi, Alexandre Cooper, Colin Connolly, Andrew Ed-
able longer native T2∗ values is a worthwhile pursuit; e.g., monds, Michael Geis, David Glenn, Ben Green, Kohei
although the NV- center’s sensitivity to strain can be re- Itoh, Jean-Christophe Jaskula, Pauli Kehayias, Mark Ku,
duced (Secs. IV.B and IV.D), employing low-strain host di- Junghyun Lee, David LeSage, Igor Lovchinsky, Matthew
amonds is preferable regardless. Importantly, a robust and Markham, Claire McLellan, Idan Meirzada, Gavin Morley,
optimized protocol for diamond irradiation and annealing Mark Newton, Michael O’Keeffe, David Phillips, Emma
that takes nitrogen concentration into account should be es- Rosenfeld, Brendan Shields, Swati Singh, Matthew Stei-
tablished (Secs. VI.D and VI.E). Furthermore, widespread necker, and Daniel Twitchen for helpful comments and dis-
access to high-quality scientific diamonds is imperative and cussions. This material is based upon work supported by,
would greatly accelerate advances in NV-diamond-related or in part by, the United States Army Research Laboratory
research. Presently, diamonds with natural carbon isotopic and the United States Army Research Office under Grant
abundance, suboptimal nitrogen concentrations, and unde- No. W911NF-15-1-0548; the National Science Foundation
sired strain and surface characteristics are widely employed Electronics, Photonics and Magnetic Devices (EPMD),
by the community solely because most research groups lack Physics of Living Systems (PoLS), and Integrated NSF Sup-
access to optimized diamond samples. port Promoting Interdisciplinary Research and Education
In addition, many aspects of NV physics, and charge dy- (INSPIRE) programs under Grants No. ECCS-1408075,
namics for ensembles in particular, remain poorly under- No. PHY-1504610, and No. EAR-1647504, respectively; Air
stood and warrant further investigation. We anticipate that Force Office of Scientific Research award FA9550-17-1-0371;
additional knowledge could be harnessed to improve sensor Defense Advanced Research Projects Agency Quantum As-
performance, similar to how the study of NV- and NV0 ion- sisted Sensing and Readout (DARPA QuASAR) program
ization characteristics under low optical intensity by Aslam under Contract No. HR0011-11-C-0073; and Lockheed Mar-
et al. (Aslam et al., 2013) prompted the development of tin under Contract No. A32198. J. M. S. was partially sup-
spin-to-charge conversion readout. Further examination of ported by a Fannie and John Hertz Foundation Graduate
charge dynamics under magnetometer operating conditions Fellowship and a NSF Graduate Research Fellowship under
(e.g., high optical intensity) is expected to yield fruitful in- Grant No. 1122374.
sights. For NV-rich diamonds, systematic studies of (i) NV-
ionization (both from the singlet and triplet excited states),
Appendix A
and (ii) recombination from the NV0 excited state versus
optical wavelength and intensity, would be particularly use- 1. Derivations
ful. Such studies would address present knowledge gaps
and could inform diamond-engineering protocols to better a. Ramsey DC magnetic field measurement
stabilize the NV- charge state in ensemble-based devices.
These investigations could also lay the groundwork for new The following is a derivation of a Ramsey-type pulsed
sensitivity-enhancement techniques tailored to ensembles. magnetometry sequence (see Fig. 7) using a magnetic dipole
In addition, continued basic research into the NV- center moment. Here the magnetic moment is taken to be an NV-
is warranted. For example, while four electronic states of center’s ground-state electronic spin, although this discus-
NV- have been observed, two additional predicted states sion applies to any two-level system sensitive to magnetic
have not yet been experimentally confirmed (Jensen et al., fields, including atomic vapors and other solid state defects.
2017). Although the NV- ground state spin is a triplet with S = 1,
We also expect unanticipated creative ideas to emerge a bias magnetic field B0 can be applied along the NV- sym-
that further enhance readout fidelity, dephasing time T2∗ , metry axis to split the ms = +1 and ms = −1 energy levels
and overall magnetic field sensitivity. Ensemble-NV- mag- so that resonant MWs may selectively drive the ms = 0 to
netometers are already relevant in wide-varying sensing ms = +1 (or ms = 0 to ms = −1) transition. Any off-
applications, thanks to key advances made over the past axis magnetic field component B⊥ can be ignored so long
decade, which we have summarized here. Moreover, NV- as (γe B⊥ )2 /[(2πD)2 ± (γe B0 )2 ]  1, where D = 2.87 GHz
diamond quantum sensing is a quickly developing platform, is the zero-field splitting and γe = ge µB /~ is the gyromag-
well positioned to continue improving, with significant ad- netic ratio of the NV- electronic spin. Here the NV- center’s
 50

nuclear spin is also ignored, as well as static electric fields This Hamiltonian causes the spin system to undergo Rabi
or strain. We describe this two-level subspace as a pseudo- oscillations at angular frequency Ω = γe B1 /2. The oscil-
spin-1/2 system with |ms = +1i = |↓i, |ms = 0i = |↑i, and ~ 1 (t) is turned off abruptly after a duration
lating field B
Hamiltonian τ 2 = 2Ω = γeπB1 , so that
π
π

H = (2πD + γe B)Sz 
γe B1 σy τ π2

!
(A1) |ψ̃(τ π2 )i = exp −i |ψ̃(0)i
~ 2πD + γe B 0 4
= ,  π 
2 0 −2πD − γe B = exp −i σy | ↑i
4 ! !
expressed in the {|↓i, |↑i} basis, where here we take Sz to (A6)
1 1 −1 0
be the spin-1/2 z-projection operator with units of ~/2; and =√
2 1 1 1
B = B0 + Bsense is the total magnetic field projection along
the NV- symmetry axis (the z-axis), which is the sum of the 1
= √ (−| ↓i + | ↑i) ,
applied bias field and an unknown DC field to be sensed. 2
Here terms in the Hamiltonian proportional to the identity
matrix have been dropped, as they introduce only a global which uses the identity e−iθn̂·~σ = cos (θ) I − i sin (θ) (n̂ ·
phase to the states’ time evolution. In the bias field B0 ~σ ) where n̂ is a unit vector on the Bloch sphere. This
the spin resonance frequency is ω0 = 2πD + γe B0 . Spin constitutes a π/2-pulse on the spin.
operators are expressed in the Sz basis in terms of the Pauli Next, the magnetic moment undergoes free precession in
~ = ~ ~σ , yielding
matrices S the absence of B~1 (t) for a sensing time τ . During this time
2
the system Hamiltonian returns to H from Eqn. A1. We
~ω0 ~ continue to use the interaction picture with H0 = ~ω2 0 σz ,
H= σz + γe Bsense σz . (A2) and with new interaction Hamiltonian H10 determined by
2 2
~ sense = Bsense ẑ as
B
As described herein, a Ramsey sequence consists of two
π/2-pulses of an oscillating magnetic field resonant with
the transition between | ↑i and | ↓i, which are separated ~
H10 = γe Bsense σz . (A7)
by a free precession time τ . The sequence begins at time 2
t = 0, with the spin polarized to |ψ(0)i = |↑i. An oscillating
Recognizing that H10 commutes with H0 , the transformed
magnetic field oriented perpendicular to the NV- symmetry
axis B~ 1 (t) = B1 cos(ωt)ŷ with angular frequency ω ≈ ω0 interaction Hamiltonian H̃10 ≡ U0† (t)H10 U0 (t) = H10 , and
~ 1 is thus the interaction-picture state vector |ψ̃(t)i evolves un-
is turned on abruptly. Without loss of generality B
der H10 into
assumed to be polarized along the y-axis. For B1  Bsense ,
the second term in H can be dropped, thereby ignoring 0

effects of the unknown DC sensing field while the oscillating |ψ̃(τπ/2 + τ )i = e−iH1 τ /~ |ψ̃(τπ/2 )i
field is on. The Hamiltonian for the system driven by this 1 (A8)
= √ (−e−iφ/2 | ↓i + eiφ/2 | ↑i),
oscillating field, denoted Hdriv , becomes 2
~ω0 ~ where
Hdriv = σz + γe B1 cos(ωt)σy . (A3)
2 2
φ = γe Bsense τ (A9)
We proceed in the interaction picture, with H0 = ~ω2 0 σz
and H1 = ~2 γe B1 cos(ωt)σy . This step is equivalent to is the phase accumulated due to Bsense in the interaction
transforming into a rotating frame with angular frequency picture. (If Bsense = 0, the state vector |ψ̃(t)i accumulates
ω0 . The interaction-picture state vector |ψ̃(t)i is defined no phase, as H10 vanishes and the entire Hamiltonian H =
in terms of the Schrödinger-picture state vector |ψ(t)i as H0 .)
|ψ̃(t)i = U0† (t)|ψ(t)i with U0 (t) = e−iH0 t/~ . This state To complete the sequence, a second oscillating field
evolves according to |ψ̃(t)i = Ũ1 (t)|ψ̃(0)i where Ũ1 (t) = B~2 (t) = B~2 cos(ωt), again with ω = ω0 , is applied for a
e−iH̃1 t/~ , with π/2-pulse. As with the first π/2-pulse, Bsense  B2 is as-
sumed so that additional spin state evolution due to Bsense
H̃1 = U0† (t)H1 U0 (t) can be ignored. The polarization of B~2 (t) is chosen to be
! along n̂ in the x-y plane at an angle ϑ with respect ŷ, the po-
~ 0 −i(ei(ω0+ω)t +ei(ω0−ω)t )
= γe B1 . larization direction of the first π/2-pulse B~1 (t). After again
4 i(e−i(ω0−ω)t +e−i(ω0+ω)t ) 0
making the rotating wave approximation, the transformed
(A4)
interaction Hamiltonian, H̃100 is given by
The transformed interaction Hamiltonian H̃1 is simplified ~
by assuming resonant driving of the spin with ω = ω0 and H̃100 ≈ γe B2 (cos (ϑ) σy − sin (ϑ) σx ) (A10)
4
by making the rotating wave approximation, dropping off-
resonant terms rotating at 2ω0 , to yield
~
H̃1 ≈ γe B1 σy . (A5)
4
 51

and b. Spin-projection-noise-limited sensitivity

|ψ̃(τ π2 + τ + τ π2 )i The spin-projection-noise-limited magnetic field sensitiv-

−iH̃100 τ π /~ ity is defined as the field δB at which the size of the signal
=e 2 |ψ̃(τ π2 + τ )i
! ! (A11) δhSz i due to δB is equal to the uncertainty in the signal,
1 −e−iϑ −e−iφ/2
p
1 1 i.e., when δhSz i = ∆Sz , where ∆Sz = hSz2 i − hSz i2 is the
=√ iϑ
·√ ,
2 e 1 2 eiφ/2 standard deviation of a series of identical measurements of
δB. From Eqn. A14 assuming precession time τ , this min-
which, up to a global phase, is equal to imum field is

φ−ϑ
 
φ−ϑ
 1 ∆Sz
|ψ̃i = cos | ↓i + ieiϑ sin | ↑i. (A12) δBsp = . (A16)
2 2 γe τ | dhSz i |
dφ

The phase accumulated during τ is thus mapped on to a When M uncorrelated consecutive measurements are
population difference between the |↓i and |↑i states. The taken, each with precession time τ over a total measure-
population difference is detected by measuring the rotating- mentptime tmeas
p, the minimum field is modified by the fac-
frame observable S̃z , which is equal to the fixed-frame spin tor 1/M = τ /tmeas , yielding
projection operator Sz , as Sz commutes with H0 . The value
of Bsense is determined by relating this measured observable 1 1 ∆Sz
δBsp = √ . (A17)
to φ: γe τ tmeas | z i |
dhS
dφ

~ The spin-projection-noise-limited sensitivity of a Ramsey

hSz i = hψ̃|σz |ψ̃i
2     magnetometry measurement is then
~ 2 φ−ϑ 2 φ−ϑ
= cos − sin √ 1 ∆Sz
2 2 2 ηsp = δBsp tmeas = √ . (A18)
(A13)
~ γe τ | dhSz i |
dφ
= cos(φ − ϑ)
2
∆Sz
~ The quotient dhSz i is calculated:
= cos(γe Bsense τ − ϑ). | dφ |
2
~
The cosinusoidal fluctuations in hSz i are termed Ramsey hSz i = cos(φ − ϑ), (A19)
fringes. Common choices of ϑ are 0 and π/2. The case 2
where ϑ = 0 (respectively, ϑ = π/2) is commonly called
cosine (sine) magnetometry, as the observable hSz i varies dhSz i ~ (A20)
= − sin(φ − ϑ),
as the cosine (sine) of Bsense for fixed τ . For ensembles dφ 2
of NV- centers, hSz i is measured by reading out the spin-
state-dependent fluorescence over a predetermined readout
~2
window of several hundred nanoseconds (see Fig. 6), as dis- hSz2 i = hψ̃|σ 2 |ψ̃i
cussed in Sec. II.A and later in Appendix A.1.c. 4  z 
~2 (A21)
  
For small Bsense such that φ  2π, Eqn. A13 can be 2 φ−ϑ 2 φ−ϑ
= cos + sin
linearized about φ = 0 for any value of ϑ except ϑ = 0. 4 2 2
The values of Bsense and φ can then be related to a small ~2
change in the observable δhSz i = hSz i|φ − hSz i|0 as follows: = ,
4
φ 1 δhSz i
Bsense = ≈ ∆Sz =
p
hSz2 i − hSz i2
γe τ γe τ dhSz i |0
dφ
(A14) r
2 ~2 (A22)
1 ~ hSz i|φ− cos(ϑ) = (1 − cos2 (φ − ϑ))
≈ . 4
γe τ sin(ϑ)
~
= | sin(φ − ϑ)|,
For ϑ = π/2, the slope of the Ramsey fringe is maximized, 2
and Eqn. A14 reduces to
∆Sz (A23)
2 = 1.
Bsense ≈ hSz i. (A15) dhSz i
| dφ |
~γe τ
For ϑ = 0, where the linearization method fails, the linear Equation A23 illustrates that the signal-to-noise ratio of
term δhSz i vanishes, as the slope of the Ramsey fringe goes a spin-projection-noise-limited measurement is independent
to zero; a small Bsense produces to lowest order a quadratic of the value of φ or ϑ. The projection noise ∆Sz is exactly
change in hS̃z i. equal to the slope of the Ramsey fringe | dhSzi
dφ |. As a result,
a magnetometer limited by spin projection noise has the
 52

same sensitivity regardless of where on the Ramsey fringe Recalling the operator commutation relation [â, ↠] = 1,
the measurement is taken, which is given by ∆N̂ is calculated:
√ 1
√ . hNˆ2 i = h(↠â + b̂† b̂)(↠â + b̂† b̂)i
ηsp = δBsp tmeas = (A24)
γe τ = hψph |(↠â↠â + b̂† b̂b̂† b̂)|ψph i
For sensing with an ensemble √ of N independent spins, the = hψph |(↠(↠â + 1)â + b̂† (b̂† b̂ + 1)b̂)|ψph i
ensemble
sensitivity ηsp = ηsp / N such that 
1 + cos(ϕ)
 
1 − cos(ϕ)

= b(b + 1) + a(a + 1) ,
1 2 2
ensemble
ηsp = √ . (A25)
γe N τ (A30)

   
1/2 +cos(ϕ) 1/2 −cos(ϕ)
c. Photon-shot-noise-limited sensitivity 2
hN̂ i = b 2
+a 2
2 2 (A31)
The above discussion considered a direct measurement 2
b a 2

ba
of Sz . The measurement technique for NV- spins - optical + + cos2 (ϕ) + sin2 (ϕ),
4 4 2
readout - instead indirectly probes the spin through mea-
suring the spin-state-dependent fluorescence. Shot noise q
in the collected fluorescence must be incorporated into the ∆N̂ = hN̂ 2 i − hN̂ i2
measurement uncertainty and sensitivity. s     
To phenomenologically introduce Poisson fluctuations b2 ba a2 1+cos(ϕ) 1−cos(ϕ)
= − + sin2 (ϕ)+b +a
from the fluorescence photons into the sensitivity, the op- 4 2 4 2 2
tical readout procedure is treated as a mapping of the spin
r
(a − b) 2  ϕ   ϕ 
eigenstates onto two light field modes: |ms = +1i = |↓i → = sin2 (ϕ) + b cos2 + a sin2 .
4 2 2
|βi and |ms = 0i = |↑i → |αi, where |αi and |βi are co- (A32)
herent states defined by â|αi = α|αi and b̂|βi = β|βi. We
define a = |α|2 as the mean number of photons in |αi and Using Eqns. A32 and A29, the sensitivity reduces to
b = |β|2 as the mean number of photons in |βi. Since the v
u (a−b)2
sin2 (ϕ) + b cos2 ϕ2 + a sin2 ϕ2



|ms = 0i state produces more fluorescent photons during ∆N̂ u
4
= t .
readout than the |ms = +1i state, a > b. The final spin | dhN i (a−b)2
sin2 (ϕ)
dφ | 4
state |ψ̃i from Eqn. A12 is mapped onto the photon field
(A33)
state
For the case ϕ = π/2, the sensitivity is optimized, yield-
ing
   
φ−ϑ iϑ φ−ϑ
|ψph i = cos |βi + ie sin |αi. (A26)
2 2 v
u (a−b)2 s
a+b
∆N̂ u + 2(a + b)
A measurement of the spin state has become a measurement = t 4(a−b)2 2 = 1 + . (A34)
| dhN̂ i
| (a − b)2
of the number of photons collected from the two light fields dφ 4
N̂ = ↠â + b̂† b̂. Defining ϕ = φ − ϑ,
ϕ ϕ We identify C = a−ba+b as the measurement contrast, (i.e.,
hN̂ i = hψph |(↠â + b̂† b̂)|ψph i = b cos2 + a sin2 the fringe visibility), and navg = a+b
2 as the average num-
2 2
   ber of photons collected per measurement (per spin, if the
1 + cos(ϕ) 1 − cos(ϕ)
=b +a . measurement is on an ensemble). The contrast C depends
2 2 on the degree of initial polarization of the spin state and
(A27) the readout duration. Measurement contrast also dimin-
ishes with increased free precession time due to spin de-
where the two light fields are assumed to be noninterfering
phasing and docoherence, parameterized by T2∗ , as shown
so that â|βi = b̂|αi = 0 and hα|βi = hβ|αi = 0.
in Eqn. 13. However, since this degradation affects both the
The sensitivity of a magnetometer employing optical
shot-noise and spin-projection-noise terms in the measure-
readout is written in the same way as the spin-projection-
ment sensitivity ηopt , it is included explicitly rather than
noise-limited sensitivity given in Eqn. A18, but with the
incorporated into C. Thus, the sensitivity (neglecting over-
observable Sz replaced by N̂ :
head time) for a Ramsey measurement on a single spin with
√ 1 ∆N̂ both photon shot noise and spin-projection noise is given
ηopt = δBopt tmeas = √ , (A28) by
γe τ | dhN̂ i |
dφ
q s
√ 1 1
where ∆N̂ = hN̂ 2 i − hN̂ i2 . The derivative of hN̂ i with ηopt = δBopt tmeas = ∗ p √ 1+ .
respect to φ is γe e−(τ /T2 ) τ C 2 navg
(A35)
dhN̂ i dhN̂ i (a − b)
= = sin(ϕ). (A29)
dφ dϕ 2
 53

(See Appendix A.7 for discussion of the stretched exponen- 3.0 0.575
0.675
/T2*
tial parameter p.) When sensing with an ensemble of N 0.9
independent spins, the sensitivity is given by
0.625
ensemble ηopt
ηopt =√ . (A36) 2.5
N 0.525 0.65
0.8

In conventional NV- optical readout, measurement con-

trast is low (. 15%), and the number of photons navg col-
lected per spin at most order unity (see Table 2), and is 2.0 0.6 0.7

p
often much less due to imperfect collection efficiency. Thus,
C 2 navg  1, and shot noise becomes the dominant contri- 0.7
bution to the magnetic field sensitivity, which in the absence 0.6
0.5
of overhead time is given by 1.5
0.725

0.775
1 1 0.825
0.5
ηshot ≈ −(τ /T ∗ )p √ . (A37) 0.75
0.8
γe Ce 2 navg τ 0.85
1.0 0.55 0.875 0.9

and 0 1 2 3 4 5
Normalized overhead time - tO /T2*
ensemble 1 1
ηshot ≈ . (A38)
γe Ce−(τ /T2∗ )p N navg τ
p
FIG. 33 Optimal precession time τ for a pulsed magnetometry
protocol. Contour plot shows precession time τ to achieved opti-
mal sensitivity, in units of T2∗ , for different stretched exponential
d. Overhead time parameters p and different overhead times tO .

The sensitivity equations above have neglected any op-

tical initialization or readout time, as well as the finite SQ). Equation A40, and thus Fig. 33, also apply for Hahn
duration of the two π/2-pulses of the Ramsey sequence. echo (Eqn. 27) with T2∗ replaced by T2 (see Section IV.A).
Grouping all of these factors into an experimental dead In practice, additional experimental factors warrant con-
time tO , we find the sensitivity factor for M measure- sideration when choosing the Ramsey free precession time
τ . For example, because time-varying electric and magnetic
ments
p pwith sensing time τ over a total time tmeas is
each
fields and temperature may mask as dephasing mechanisms,
1/M = (τ + tO )/tmeas , yielding a sensitivity limited by
shot noise and spin-projection noise of the measured value of T2∗ depends on the measurement du-
ration. Thus, if the time required to measure the value of
√ s
T2∗ is significantly longer than the duration of a magnetic
√ 1 τ + tO 1
ηopt = δBopt tmeas = ∗ p 1+ 2 . field measurement, field fluctuations may artificially reduce
γe e−(τ /T2 ) τ C navg
the measured value of T2∗ compared to the value relevant
(A39) for sensing. This spoiled T2∗ measurement could lead to
a suboptimal choice of τ (Bauch et al., 2018). Therefore,
care should be taken when choosing the appropriate free
2. Optimal precession time
precession time τ for a magnetometry experiment.
The optimal precession time τ to achieve best Ramsey
magnetometry sensitivity (Eqn. 15) depends on the value 3. Considerations for increasing sensor number
of the stretched exponential parameter p, the initialization
time tI , and the readout time tR . By defining the overhead Increasing the number N of interrogated NV- centers by
time per measurement as tO = tI + tR , Eqn. 15 reduces to increasing either the interrogation volume or the NV- den-
√ sity may be partially effective to improve magnetic field sen-
1 τ + tO
η ∝ −(τ /T ∗ )p . (A40) sitivity. In this instance, the number of photons detected
e 2 τ per measurement N increases with the number of sensors
For tO  T2∗ , sensitivity is optimized when τ ≈ T2∗ /2 for N . However, a series of practical factors may hinder this
1 ≤ p ≤ 2. Particularly, for tO = 0 and p = 1 or p = strategy. First, sensitivity enhancement exhibits sublinear
2 (see Appendix A.7), sensitivity is exactly optimized for scaling with N and the associated number of photons de-
τ = T2∗ /2. As tO increases from zero, the optimal precession tected per measurement N, i.e., η ∝ √1N , making signifi-
time increases as well, asymptotically approaching τ = T2∗ cant sensitivity improvements from increasing N difficult.
when tO  T2∗ for p = 1. Figure 33 shows the optimal Additional technical difficulties may arise when increas-
precession time τ for various combinations of p and tO . ing N , such as the photon number requirement for optical
For clarity the optimal precession time is normalized to the initialization. Assuming that each interrogated NV- center
dephasing time in the employed measurement basis (DQ or requires m photons for optical initialization, each measure-
 54

ment is expected to require an energy of where we have ignored typically less common defects in
fully treated diamond (i.e., irradiated, annealed, etc.) such
hc as NVH- , N2 V- , etc. (Hartland, 2014); and T2∗ {other} de-
Einit = N m , (A41)
λ notes the T2∗ limit from all non-nitrogen-related dephasing
where h is Planck’s constant, c is the speed of light and λ is mechanisms. The above equation can be rewritten as
the excitation wavelength. If measurements are performed 1 N+
every T2∗ , the required mean power is ∗ = AN0S [NT ][1−Econv −E0conv −EconvS
] + ANV- [NT ][Econv ]
T2
N m hc (A44)
Pinit = . (A42) 1
T2∗ λ + ANV0 [NT ][E0conv ] + ∗
T2 {other}
For example, initialization of all 1.76 × 1014 NV- centers
in a 1 mm3 diamond with 1 ppm [NV- ] would require where Econv ≡ [NV- ]/[NT ], E0conv ≡ [NV0 ]/[NT ], and
Einit = 200 µJ, using a crude guess of m = 3 (see Table 13). N+
Econv
S
≡ [N+ T
S ]/[N ] are the conversion efficiencies from the
Assuming T2∗ = 1 µs, the required power is Pinit = 200 W.
total nitrogen concentration [NT ] to [NV- ], [NV0 ], and [N+
S]
Eqn. A42 illustrates that achieving a sensitivity improve-
respectively. The AX coefficients characterize the magnetic
ment by increasing the NV- ensemble size will increase Pinit
dipole interaction strength between NV- spins and spin
unless T2∗ is increased as well. For experimental approaches
species X. The value of AN0S is defined in Eqn. 22, the
employing an acousto-optic modulator to gate a CW laser,
value of ANV- is defined in Sec. III.G, and in this section
the required CW laser power will be higher as many pho-
for reasons of compactness we do not differentiate between
tons are wasted.
ANV-k and ANV-∦ . The value of ANV0 is defined so that the
Another difficulty encountered when increasing the num-
ber of interrogated NV- centers N (and thus detected pho- NV- dephasing from NV0 satisfies 1
T2∗ {NV0 }
= ANV0 [NV0 ].
ton number N) is that reaching the shot noise limit can Under the assumption that
N+
Econv , E0conv , Econv
S
are indepen-
become challenging for large values of N. For example, the dent of [NT ], consolidation yields
absolute noise contributed by some systematic (not stochas-
tic) noise sources scales linearly with the number of photons 1 1
= κ[NT ] + ∗ , (A45)
detected, i.e., ∝ k1 N, where k1  1. In √ comparably pro- T2∗ T2 {other}
portional units, shot noise scales as ∝ N. For N > k12 ,
1
N+
the systematic noise will be larger than shot noise. Primary where κ = AN0S [1 − Econv − E0conv − Econv
S
] + ANV- [Econv ] +
examples of such noise sources include laser intensity noise 0
ANV0 [Econv ]. The detected number of PL photons per mea-
in all implementations, timing jitter in the readout pulse surement is N ∝ [NT ]Econv V navg where V is the interro-
length for Ramsey and pulsed ODMR, and MW amplitude gation volume. For simplicity we consider the limit where
noise in CW-ODMR. initialization and readout times tI and tR are negligible, so
Lastly, increases in interrogation volume or NV- density that sensitivity is
are both accompanied by unique challenges independent of
those associated with increases in sensor number N . For
s s s
1 1 1
example, larger sensing volumes require better engineering η∝ = × κ+ T ∗ ,
NT2 ∗ Econv V navg [N ]T2 {other}
to ensure the bias magnetic field and MW field are uniform
over the sensing volume (Abe and Sasaki, 2018; Eisenach (A46)
et al., 2018). Alternatively, increasing NV- density neces- which suggests that for [NT ]  κ T ∗ {other}
1
, sensitivity is
2
sarily positions NV- spins (and all other nitrogen-related independent of [NT ]. Qualitatively, this can be interpreted
paramagnetic spins) closer together, which results in in- as follows: when T2∗ is limited by nitrogen-related dephas-
creased dipolar dephasing and shorter associated T2∗ val- ing mechanisms (i.e., NV- , NV0 , N0S ), and again assum-
ues, canceling the sensitivity improvement from addressing N+
ing Econv , E0conv , and Econv
S
are independent of [NT ], de-
more NV- spins. T
creasing [N ] increases T2∗ by the same fractional quantity
that the NV- ensemble photoluminescence N is decreased.
4. Choosing nitrogen concentration in diamond samples However, when T2∗ is limited by other broadening mecha-
nisms unrelated to nitrogen, decreasing [NT ] decreases the
The following discussion parallels the clear analysis pre- collected fluorescence N without any corresponding T2∗ in-
sented in Ref. (Kleinsasser et al., 2016), which the reader crease. The implications here are significant: this analysis
is encouraged to review. Equation 18 can be simplified by suggests that while there is not a unique value of [NT ] for
grouping all non-nitrogen-related broadening mechanisms maximal sensitivity, there is a minimum value. In other
together, yielding words, if nitrogen-related broadening is a small contribu-
tor to T2∗ , the nitrogen content should be increased; the
increased resulting PL will favorably offset the increase in
1 1 1 1 1 T2∗ , resulting in overall enhanced sensitivity.
= ∗ 0 + ∗ - + + ,
T2∗ T2 {NS } T2 {NV } T2∗ {NV0 } T2∗ {other} A few points are in order regarding the above analysis.
(A43) Experimental considerations can also set an upper bound
 55

on the most desirable total nitrogen concentration [NT ]. EPR linewidths are commonly tabulated by their peak-
For example, the larger detected photon number N associ- to-peak widths ∆B, where ∆B denotes the magnetic field
ated with higher values of [NT ] can present technical chal- spacing between extrema of the resonance line first deriva-
lenges (see Appendix A.3). Moreover, the above analysis tive (Poole, 1996). In (linear) frequency units, this peak-to-
considers the simple limit where the initialization and read- peak width is δ = gµhB ∆B. Accurately relating δ and T2∗ re-
out times are negligible; accounting for this fixed overhead quires the resonance lineshape to be known (Kwan and Yen,
time (see Eqns. 10, 14, and 15) favors trading off nitro- 1979). For example, a Lorentzian profile with full width at
gen concentration density for longer values of T2∗ , in order half maximum √ (FWHM) Γ, expressed in frequency units,
to reduce the fractional overhead time devoted to initial- has δ = Γ/ 3 and Γ = πT1 ∗ (see Appendix A.5). Com-
2
ization and readout. Overall, combined experimental and bining these relations yields T2∗ Lor = √3πδ 1
. A Gaussian
theoretical considerations suggest that for best sensitivity lineshape with the same measured peak-to-peak linewidth
nitrogen content should be decreased until nitrogen-related δ has standard deviation σ = δ/2 and σ = √2πT 1
∗ (see Ap-
broadening is similar to broadening unrelated to nitrogen, √ √2

i.e., κ[NT ] ≈ T ∗ {other}

1
. pendix A.5). Thus, T2∗ Gau = πδ2 , which is 6× longer
2
than T2∗ Lor . A visual comparison of these relationships is
displayed in Fig. 34.
5. Spin resonance linewidth and T2∗ Diamond EPR literature results may report values of δ
without giving the associated resonance lineshape, prevent-
The quantity T2∗ , which characterizes the time scale of the ing accurate determination of T2∗ from δ. For example,
free induction decay (FID), is inversely proportional to the linewidth measurements in Ref. (van Wyk et al., 1997) on
natural spin resonance linewidth in the absence of power substitutional nitrogen defects N0S in diamond indicate a
broadening. Exact conversion between T2∗ and linewidth scaling 1/T2∗ {N0S } = AN0S [N0S ] with varying nitrogen con-
requires knowledge of the functional form of the FID or centration [N0S ], but the scaling factor AN0S cannot be accu-
the resonance lineshape (Abragam, 1983c; Kwan and Yen, rately determined without knowledge of the lineshape.
1979). Ramsey fringes decaying with an FID envelope ∝ Theoretical and experimental results on dipolar-coupled
∗
e−t/T2 indicate a Lorentzian spin resonance profile with full spin systems suggest a Lorentzian resonance lineshape when
width at half maximum (FWHM) Γ = πT1 ∗ , as shown by spin-bath interactions are the dominant source of line-
2
the Fourier transform pair: broadening (Abragam, 1983c; Dobrovitski et al., 2008; Hall
et al., 2014; Kittel and Abrahams, 1953). Furthermore,
1 Ramsey measurements with NV- spin ensembles show FID
∗ p
2πif0 t −t/T2∗ 1 2πT2∗ envelopes well fit by e−(t/T2 ) with p ∼ 1, corresponding
Ft [e e ](f ) =  2
π 1 to a Lorentzian lineshape (see Appendix A.7, Fig. 34 and
2πT2∗ + (f − f0 )2 (A47) Ref. (Bauch et al., 2018)) when T2∗ is expected to be spin-
1 Γ/2 bath limited.
= ,
π (Γ/2) + (f − f0 )2
2 Assuming a Lorentzian profile when converting δ val-
ues from Ref. (van Wyk et al., 1997) to T2∗ values yields
valid for t ≥ 0, where f0 is the Ramsey fringe frequency. AN0S ≈ 130 ms-1 ppm-1 for nitrogen spins in a nitrogen
∗ 2
A Gaussian decay envelope ∝ e−(t/T2 ) corresponds to a spin bath (see Fig. 35). This calculated scaling factor is
resonance with a Gaussian profile, with standard deviation considered to be an upper bound because (i), a Gaussian
1
σ = √2πT ∗ as shown by the Fourier transform pair: or Voigt profile would result in a smaller value of AN0S
2
than that calculated
√ by assuming a Lorentzian profile, as
∗ 2 √ ∗ 2 1/T2∗ Gau = 1/( 6×T2∗ Lor ); and (ii), other sources of broad-
Ft [e2πif0 t e−(t/T2 ) ](f ) = πT2∗ e−(πT2 (f −f0 ))
ening may contribute to the EPR linewidths observed in
1 2 2 (A48)
= √ e−(f −f0 ) /(2σ ) . Ref. (van Wyk et al., 1997). In the latter case the true con-
σ 2π tribution to dephasing from dipolar interactions between
N0S spins would be smaller than that estimated from the
measured δ. Nitrogen-spin-bath induced dephasing of N0S
6. Estimating T2∗ from spin resonance linewidths of N0S
and of NV- are expected to be similar, as the dipolar cou-
pling between two N0S spins is similar to the dipolar coupling
Although sensor performance is dictated by T2∗ of the
between a N0S and an NV- for equivalent separation (Han-
NV- ensemble, T2∗ values of other paramagnetic defects
son et al., 2008). Thus, the spin-bath-limited linewidth of
within the diamond, such as substitutional nitrogen defects,
nitrogen defects in diamond measured via EPR can serve as
can provide useful information on sources of NV- spin de-
a proxy for the spin-bath limited linewidth of NV- centers.
phasing. Such T2∗ values can be extracted from linewidth
The value of AN0S ≈ 130 ms-1 ppm-1 for N0S from the data
measurements, for example from electron paramagnetic res-
in Ref. (van Wyk et al., 1997) serves as an independent
onance (EPR). Accurate conversion from EPR linewidth to
estimate of AN0S for NV- centers in a nitrogen spin bath.
paramagnetic-defect T2∗ enables leveraging of existing dia-
This value is in reasonable agreement with the measured
mond EPR data (van Wyk et al., 1997) to better understand
AN0S ≈ 101 ms-1 ppm-1 for NV- ensembles from Ref. (Bauch
the contributions of different noise sources to NV- ensemble
T2∗ values. et al., 2018).
 56

!"
3$,"%)%/$45$#(6)&(6$
%" !" !"
3$,"%)%/$47#"8(9$
%" !"
-#$$4:%;+/&("%45$/)0
$"&
$" $"%
$"%
%") %")
%"&
%"( %"(
!"#$%&'()% %
%"% δ !"# $ .
%"' %"' *+, " %
!%"& -# $
%"# %"#
!$"%
%" %"%
!$"&
!!" !#" !$" %" $" #" !" !!" !#" !$" %" $" #" !" %"% %"# %"' %"( %") $"%
-#$.+$%/0 -#$.+$%/0 1(2$
$"&
$" $"%
$"%
%") %")
%"&
%"( %"(
*)+,,()% % %
%"% δ ! " # &'(&) &
%"' %"' *+, "
!%"& -#
%"# %"#
!$"%
%" %"%
!$"&
!!" !#" !$" %" $" #" !" !!" !#" !$" %" $" #" !" %"% %"# %"' %"( %") $"%
-#$.+$%/0 -#$.+$%/0 1(2$

FIG. 34 Resonance derivatives ( ), resonance profiles ( ), and free induction decay (FID) envelopes ( ) for Lorentzian and
Gaussian lineshape profiles with the same peak-to-peak widths δ. Full-width-at-half-max linewidths Γ and FID decay envelope
times T2∗ are indicated and expressed in terms of the peak-to-peak width δ, a commonly reported parameter characterizing linewidth
in electron paramagnetic resonance (EPR) data.

100
of p are well characterized for single spins under a vari-
1/T2* from Gaussian Profile ety of environmental conditions (Dobrovitski et al., 2008;
1/T2* from Lorentzian Profile Hall et al., 2014; Hanson et al., 2008; Maze et al., 2012;
Fit
de Sousa, 2009). For example, a single spin experienc-
ing dipolar coupling to a surrounding bath of spins dis-
plays an FID envelope with stretched exponential parame-
10 ter p = 2 (Dobrovitski et al., 2008; Hall et al., 2014; Maze
et al., 2012; de Sousa, 2009) (Gaussian ODMR lineshape,
1/T2* (µs )
-1

see Table 11 and Fig. 34). Meanwhile, NV- ensembles with

linewidth limited by dipolar coupling to a spin bath are
predicted (Dobrovitski et al., 2008; Hall et al., 2014) and
measured (Bauch et al., 2018; MacQuarrie et al., 2015) to
1 exhibit FID envelopes with p = 1 (Lorentzian ODMR line-
shape, see Table 11 and Fig. 34). However, experimental
Ramsey measurements on NV- ensembles may sometimes
exhibit decay envelopes with p 6= 1, suggesting the pres-
ence of other broadening mechanisms such as strain gra-
0.01 0.1 1 10 100 1000 dients, magnetic field gradients, or temperature fluctua-
Nitrogen concentration (ppm)
tions (Bauch et al., 2018). A noninteger p for an ensemble
FIG. 35 Bounds on 1/T2∗ associated with EPR linewidth mea- may also indicate the presence of more complex dephasing
surements of N0S defects (P1 centers) from Ref. (van Wyk et al., and decoherence dynamics, including spatial inhomogene-
1997) in diamonds with a range of nitrogen impurity concentra- ity, than can be encompassed by a single decay time con-
tions, calculated assuming Gaussian ( ) and Lorentzian (N) EPR stant. In some cases the decay may be better described by
lineshapes. A fit to the function 1/T2∗ = AN0 [N0S ] + b assum- a sum (Cao, 1994) or a product of multiple decay curves
S
ing a Lorentzian lineshape ( ) yields AN0 ≈ 130 ms-1 ppm-1 for with different values of T2∗ and p. For example, a product
S
nitrogen spins in a nitrogen spin bath (see main text). of two FID decays, one with p = 1 and one with and p = 2,
corresponds to a Voigt profile lineshape. Allowing p to vary
when fitting FID envelopes crudely accounts for these sorts
7. Stretched exponential parameter of lineshape variations while only requiring a single addi-
tional fit parameter. Therefore, Ramsey FID measurements
Equations A47 and A48 show that the spin resonance exhibiting 1 ≤ p ≤ 2 for some NV- ensembles may suggest
lineshape can be parameterized by the stretched exponen- contributions to the ODMR lines from both Lorentzian and
tial parameter p of the free induction decay (FID) envelope Gaussian broadening mechanisms (Bauch et al., 2018).
∗ p
e−(t/T2 ) . We note that for the idealized case of a purely Hahn echo T2 decays of single NV- spins have been pre-
Lorentzian lineshape, p = 1, and for a purely Gaussian dicted (de Sousa, 2009) and measured (de Lange et al.,
lineshape, p = 2. The exact ODMR lineshape and value
 57

Ramsey T2∗ decay p Reference (experiment) Reference (theory)

Single NV- 2 (Maze et al., 2012) (Dobrovitski et al., 2008; Hall et al., 2014; de Sousa, 2009)
NV- ensemble 1 (Bauch et al., 2018; MacQuarrie et al., 2015) (Dobrovitski et al., 2008; Hall et al., 2014)

TABLE 11 Stretched exponential parameters p associated with free induction decay envelopes for single NV- centers and NV-
ensembles in dipolar-coupled spin baths

2010) to exhibit a stretched exponential parameter p = 3 Mizuochi et al. to overestimate the contribution of 13 C to
when T2 is limited by spin-bath noise. In contrast, Hahn T2∗ and draw incorrect conclusions on the scaling of T2∗ with
echo decay envelopes for ensembles of NV- spins have been [13 C].
seen to exhibit p varying from ∼ 0.5 to 3, depending on the Reference (Balasubramanian et al., 2009) report
dominant contributors to the spin bath and the bias mag- linewidths of 210 kHz and 55 kHz for diamonds with 1.1%
netic field angle (Bauch et al., 2019; Stanwix et al., 2010). and 0.3% 13 C respectively, data which is clearly consistent
with Eqn. 24. However, the authors of Ref. (Balasubrama-
nian et al., 2009) interpret their data using formalism ap-
8. Isotopic purity confusion in the literature propriate for 13 C & 10% (Abragam, 1983c), which results
in them employing Eqn. A50.
In Sec. III.F, we discussed the T2∗ limit imposed by [13 C], As discussed in Sec. III.F, Eqn. 24 has been experimen-
which is described by an inverse linear scaling in Eqn. 24, tally verified in the dilute limit in a similar system (Abe
reproduced below, et al., 2010). Given that the mean single-NV- FID time
1 is longer than the ensemble FID time by ∼ 2× in dia-
= A13 C [13 C], (A49) mond with natural abundance 13 C (see Section III.F and
T2∗ {13 C}
Ref. (Maze et al., 2012)), if Eqn. A50 were also correct,
where A13 C ≈ 0.100 ms-1 ppm-1 . Although such inverse lin- then at sufficiently low 13 C concentration, an NV- ensem-
ear scaling with [13 C] is predicted by several theoretical ble would dephase more slowly than its constituent spins.
calculations (see Refs. (Abragam, 1983c; Dobrovitski et al., This prediction conflicts with the present understanding
∗{single} 13
2008; Hall et al., 2014; Kittel and Abrahams, 1953)), some that T2 { C} > T2∗ {13 C} regardless of concentration
experiments based on single NV- centers (incorrectly we (see Sec. III.B).
believe) suggest an inverse square root scaling (Balasubra-
manian et al., 2009; Mizuochi et al., 2009), i.e.,
9. Linear Stark and Zeeman regimes
1 {single}
p
∗{single} 13
= A13 C [13 C]. (A50) Here we describe coupling of electric fields, strain, and
T2 { C}
magnetic fields to the NV- spin resonances in the regimes of
for single NV- centers in the dilute limit ([13 C]/[12 C]  both low and high axial bias magnetic field B0,z . This treat-
0.01). In Ref. (Mizuochi et al., 2009), the data were derived ment draws heavily on equations and analysis in Ref. (Ja-
from mean T2∗ values taken from many single NV- defects monneau et al., 2016). While understanding of strain’s ef-
in the diamond. However, Eqn. A50 conflicts with theo- fect on the NV- spin continues to evolve (Barfuss et al.,
retical calculations by both Dobrovitski et al. (Dobrovitski 2018; Barson et al., 2017; Doherty et al., 2013; Udvarhelyi
et al., 2008) and Hall et al. (Hall et al., 2014) explicitly for et al., 2018), we take the NV- ground state spin Hamilto-
single NV- centers. Both sources instead suggest that for nian in the presence of a bias magnetic field B ~ 0 , an electric
the mean single NV- center, ~
field E, and intrinsic crystal strain to be (Doherty et al.,
2013; Udvarhelyi et al., 2018)
1 {single} 13
= A13 C [ C], (A51)
H/h = D + Mz + dk Ez Sz2


∗{single} 13
T2 { C}
ge µB
+ (B0,z Sz + B0,x Sx + B0,y Sy )
similar to Eqn. 24. h
We hypothesize that the origin of this discrepancy + (d⊥ Ex + Mx ) Sy2 − Sx2

(A52)
is omission of nitrogen broadening in the study from
+ (d⊥ Ey + My ) (Sx Sy + Sy Sx )
Ref. (Mizuochi et al., 2009), as summarized in Table 12.
Using the relation 1/T2
∗{single} {single}
{N0S } = AN0 [N0S ] (see +Nx (Sx Sz + Sz Sx ) + Ny (Sy Sz + Sz Sy ) .
S
{single}
Sec. III.D), we roughly estimate AN0 = 56 ms−1 ppm-1 Here Si with i = x, y, z are the dimensionless spin-1 projec-
S
from Ref. (Zhao et al., 2012). For the lowest C sample 13 tion operators; D is the NV- zero field splitting (≈ 2.87 GHz
in the data from Ref. (Mizuochi et al., 2009), which has at room temperature); dk = 3.5 × 10−3 Hz/(V/m) and
[N0S ] ∼ 1 ppm, this estimate predicts a nitrogen-limited d⊥ = 0.17 Hz/(V/m) are the axial and transverse elec-
∗{single} tric dipole moments (Dolde et al., 2011; Michl et al., 2019;
T2 {N0S } of ∼ 18 µs, close to the actual reported T2∗
Van Oort and Glasbeek, 1990), see Table 16; and Mz , Mx ,
measurement. Neglecting the additional nitrogen contribu-
∗{single} My , Nx , and Ny are spin-strain coupling parameters.
tion to T2 for the lowest 13 C sample likely caused
 58

∗{single} ∗{single} ∗{single}

[13 C] (ppm) Measured T2 (µs) Synthesis [NS ] (ppm) Calculated T2 {13 C} (µs) Calculated T2 {N0S } (µs)
(Mizuochi et al., 2009) (This work) (This work)
10700 3.3 CVD < 0.001 2.3 & 18000
3500 6.2 CVD < 0.001 7 & 18000
300 18 HPHT ∼1 82 ∼ 18

∗{single}
TABLE 12 The three diamonds used in Ref. (Mizuochi et al., 2009). The calculated value of T2 {13 C} is derived using the
mean value of T2∗ = 2.3 µs for single NV- centers in a natural abundance 13 C sample measured with a bias field of 20 G from
∗{single} 13 ∗{single}
Ref. (Maze et al., 2012), so that T2 { C} = 2.3 µs × 0.0107
[13 C]
. The calculated value of T2 {N0S } is estimated using the
∗{single}
simulation in Fig. 1 of Ref. (Zhao et al., 2012) which predicts T2 {N0S } = 18 ± 1 ps/[N0S ].

The Hamiltonian can be simplified when D is large com- In the linear Stark regime, characterized by βz  ξ⊥ ,
pared to all other coupling terms, i.e., in the regime of low the spin eigenstates become, approximately, equal super-
magnetic field, electric field, and strain. In particular, en- positions of |+1i and |−1i, and the transition frequencies
ergy level shifts associated with transverse magnetic field exhibit maximal sensitivity to variations in ξ⊥ :
components B0,x and B0,y (Jamonneau et al., 2016), and  2 "  #
4
with spin-strain coupling parameters Nx and Ny , are sup- ∂ν± 1 βz βz
=1− +O . (A58)
pressed by D and thus may be neglected from the Hamil- ∂ξ⊥ βz ξ⊥ 2 ξ⊥ ξ⊥
tonian (Kehayias et al., 2019). This low-field Hamiltonian
HLF is given by In contrast, in the linear Zeeman regime, characterized by
ge µB βz  ξ⊥ , the spin eigenstates become, approximately, |+1i
HLF /h = D + Mz + dk Ez Sz2 +


B0,z Sz and | − 1i, and sensitivity to strain/electric fields is sup-
h
(A53) pressed by the ratio ξβ⊥z :
+ (d⊥ Ex + Mx ) Sy2 − Sx2

+ (d⊥ Ey + My ) (Sx Sy + Sy Sx ) .  3 " 5 #

∂ν± ξ⊥ 1 ξ⊥ ξ⊥
= − +O . (A59)
We focus on the interplay between different terms in HLF ∂ξ⊥ βz ξ⊥ βz 2 βz βz
that shift the NV- spin resonance frequencies in opposite di-
rections, including B0,z , Ex , Ey , Mx , and My . In contrast, By performing magnetic sensing in the linear Zeeman
dephasing associated with variations in terms that shift the regime, spatial and temporal variations in transverse elec-
resonance frequencies in common-mode (D, Ez , and Mz ), tric fields and strain couple less strongly to the NV- spin,
can be mitigated by employing double-quantum coherence and thus their contribution to T2∗ is diminished. The linear
magnetometry (see Sec. IV.B) and are ignored herein. In Zeeman regime is best suited for high-sensitivity magne-
addition to shifting the spin resonance frequencies, trans- tometry not only because of the T2∗ extension from sup-
verse electric fields, Ex and Ey , and transverse spin-strain pressed sensitivity to variations in ξ⊥ , but also because
coupling terms, Mx and My , mix the ms = ±1 spin states magnetic field changes couple most strongly to ν± in this
into regime:
   
θ iφ θ
|+i = cos |+1i + e sin |−1i,
" 4 #
(A54) ∂ν± 1

ξ⊥
2
ξ⊥
2 2 =1− +O . (A60)
 
θ
 
θ ∂βz βz ξ⊥ 2 βz βz
|−i = sin |+1i − eiφ cos |−1i, (A55)
2 2
Experiments that must operate at near-zero B ~ 0 for other
where tan(φ) = (d⊥ Ey + My )/(d⊥ Ex + Mx ) and reasons, such as to protect ferromagnetic samples, should
tan(θ) = ξ⊥ /βz . Here βz = (ge µB /h)B0,z represents use low-strain diamonds to avoid operating in the unfavor-
- able regime where βz  ξ⊥ (Backlund et al., 2017; Fu et al.,
p magnetic field coupling to the NV spin and ξ⊥ =
the
(d⊥ Ex + Mx ) + (d⊥ Ey + My ) combines the effects of
2 2 2014; Glenn et al., 2017; Zheng et al., 2019). In this lin-
transverse strain and electric fields. The transition frequen- ear Stark regime, not only is sensitivity to magnetic signals
cies |0i ↔ |+i and |0i ↔ |−i are suppressed by the ratio ξβ⊥z ,
q 3 " 5 #
ν± = D + Mz + dk Ez ± ξ⊥ 2 + β2,

z (A56) ∂ν± βz 1 βz βz
= − +O , (A61)
∂βz βz ξ⊥ ξ⊥ 2 ξ⊥ ξ⊥
and the coupling strength of transverse strain and electric
fields to the NV- spin resonance frequencies is given by but also T2∗ may be shortened by electric field and strain
variations.
∂ν± ±1
=r  2 . (A57)
∂ξ⊥
1 + ξβ⊥z 10. Example annealing calculations

We present some calculations to estimate parameters nec-

essary for LPHT annealing to form NV- centers. Using
 59

D0 = 1.6 × 10−3 cm2 /s (Fletcher and Brown, 1953) for In this nitrogen-based diamond classification system, all
the diffusion constant, Ea = 2.3 eV for the activation en- diamonds are categorized into one of two primary types:
ergy, and T = 800 ◦ C for the annealing temperature, we Type I diamonds contain measurable quantities of nitrogen
expect D = 2.5 nm2 /s. When annealing at T = 800 ◦ C and while Type II diamonds do not, as shown in Fig. 36. There
tanneal = 12 × 3600 s, a single vacancy in a perfect lattice is no wide consensus on what constitutes “measurable” in
is expected (based on the model presented here) to have an age of ever-advancing characterization tools, although a
made ∼ 2.7 × 107 lattice jumps, visited 1.5 × 107 distinct common definition is a quantity detectable with an FTIR
lattice sites (Fastenau, 1982; Vineyard, 1963), and diffused spectrometer (Breeding and Shigley, 2009). Most sources
a root-mean-square
√ distance of hrrms i ≈ .8 µm, assuming suggest a delineation somewhere between 0.5 ppm (Zaitsev,
hrrms i = 6Dtanneal . The uncertainties in these estimates 2001) and 20 ppm (Dischler, 2012; Gaillou et al., 2012).
are dominated by the ±0.3 eV uncertainty in Ea (Davies This delineation uncertainty is particularly unfortunate for
et al., 1992; Mainwood, 1999), which can lead to an order the NV- community, as many diamonds employed for en-
of magnitude variation in D for T = 800 ◦ C. Ignoring small semble NV- experiments fall in this range.
repulsive forces between substitutional nitrogen and mono- Type I diamonds can be further classified by the spe-
vacancies (Davies et al., 1992), a vacancy is expected to cific nitrogen complexes incorporated into the carbon lat-
visit ∼ 106 /4 lattice sites in a 1 ppm [NS ] diamond to form tice. For example, Type Ia diamonds contain aggregated
an NV. The factor 4 arises from the four closest sites to a nitrogen impurities, and describe the vast majority of nat-
substitutional nitrogen while the 106 arises because only 1 ural diamonds (& 95%, depending on the delineation nitro-
out of every 106 lattice sites is occupied by a substitutional gen concentration (Breeding and Shigley, 2009; Zaitsev,
nitrogen. Because the number of distinct lattice sites vis- 2001)). Typical nitrogen concentrations in natural Type Ia
ited is substantially greater than the number of sites needed diamonds are in the hundreds of ppm (e.g., 500 ppm (Za-
7
to form an NV center (i.e., 1.5×10
106 /4  1), the chosen values
itsev, 2001)) but can be as high as 3000 ppm (Neves and
of T and tanneal are expected to ensure adequate NV center Nazaré, 2001). If the aggregated nitrogen predominantly
formation. The simple analysis stated here is complicated forms A centers consisting of two substitutional nitrogens
by the uncertainty in D0 and Ea , as well as the presence of located adjacent in the diamond lattice, the diamond is
other vacancies, vacancy aggregates, dislocations, surfaces, classified as Type IaA. If the aggregated nitrogen predom-
etc., which which can also trap vacancies, but are beyond inantly forms B centers consisting of four substitutional
the scope of this paper. For additional detail and discussion nitrogens surrounding a lattice vacancy, the diamond is
see Ref. (Alsid et al., 2019). classified as Type IaB. In contrast, diamonds containing
The analysis presented above is derived only from first- predominantly isolated single nitrogen impurities are clas-
principles calculations and the measured value of Ea . More sified as Type Ib and make up about 0.1% of all natural
accurate behavior may be predicted by employing measured diamonds (Zaitsev, 2001). As higher nitrogen density pro-
values of D at a given temperature, such as D ≈ 1.1 nm2 /s motes aggregation, Type Ib diamonds typically exhibit ni-
at 750 ◦ C (Martin et al., 1999) and D ≈ 1.8 nm2 /s at trogen concentrations at or below the 100 ppm level (Zait-
850 ◦ C (Alsid et al., 2019). sev, 2001), less than typical for Type Ia diamonds (Zaitsev,
2001).
Type II diamonds containing no “measurable” nitrogen
11. The diamond type classification system can be additionally classified as well. Type IIa diamonds
contain no other measurable impurities and make up the
We briefly overview the “diamond type” classification sys- majority of gem-grade diamonds in spite of comprising only
tem introduced in the 1930s and outlined in Refs. (Robert- 1 to 2% of natural diamonds. These diamonds are the most
son et al., 1933, 1936). In spite of the system’s shortcom- optically transparent diamonds: while Type IIa diamonds
ings, it has been widely adopted by the gemstone commu- with low levels of impurities may exhibit pale shades of yel-
nity and is partially used by the scientific community today. low, pink, or purple; extremely pure Type IIa diamonds
In the mid-1930’s the authors of Refs. (Robertson et al., are colorless (Zaitsev, 2001). Nearly all single NV- exper-
1933, 1936) noted that although the vast majority of natu- iments employ Type IIa diamonds. As boron is another
ral diamonds exhibited absorption lines in the 225−300 nm common impurity in natural diamond, Type II diamonds
band and near 8 µm, these same absorption features were with “measurable” boron are categorized as Type IIb. These
absent in a small minority of diamonds. The authors fur- diamonds make up about 0.1% of all natural diamonds and
ther observed that diamonds lacking these same absorption may exhibit a bluish or greyish hue.
features tended to exhibit lower birefringence and higher Although the diamond type classification system was de-
photoconductivity relative to their peers (Robertson et al., veloped for natural diamonds, it appropriately describes
1933, 1936). In 1959 the authors of Ref. (Kaiser and Bond, synthetic diamonds as well. CVD-grown diamonds without
1959) attributed the observed infrared absorption features nitrogen doping are Type IIa. Man made HPHT diamonds
to carbon-nitrogen molecular vibrations, signaling the pres- of Type IaA, Ib, IIa, and IIb have been created. Further
ence of nitrogen. Nitrogen was found to be the most com- diamond types exist: see Ref. (Zaitsev, 2001).
mon impurity occurring in natural diamonds, which made
its presence or absence a logical basis for diamond classifi-
cation.
 60

Reference (Tetienne et al., 2012) (Gupta et al., 2016) (Robledo et al., 2011) (Acosta et al., 2010b)
NV- centers probed 4 3 2 ensemble units
Values reported avg. (max, min) avg. (max, min) avg.
3
E(ms = 0) →3 A2 (ms = 0) 67.9 (63.2, 69.1) 66.16 (66.08, 66.43) 64.2 - µs-1
3
E(ms = ±1) →3 A2 (ms = ±1) 67.9 (63.2, 69.1) 66.16 (66.08, 66.43) 64.9 - µs-1
3
E(ms = 0) →1 A1 5.7 (5.2, 10.8) 11.1 (10.9, 11.2) 11.2 - µs-1
3
E(ms = ±1) →1 A1 49.9 (48.6, 60.7) 91.8 (89.3, 92.9) 80.0 - µs-1
1
E→3 A2 (ms = 0) 1.0 (0.7, 1.5) 4.87 (4.75, 4.90) 3.0 - µs-1
1
E→3 A2 (ms = ±1) 0.75 (0.4, 1.4) 2.04 (2.03, 2.13) 2.6 - µs-1
1
E lifetime - 144.5 (144.3, 145.3) 178 ± 6 219 ± 3 ns

TABLE 13 NV- decay rates measured at room temperature. Averages over measured NV- centers are weighted by reported
uncertainties. Dashed lines (-) indicate values not reported. Branching ratios can be derived from the given data. Although not
tabulated, vibrational decay within the 3 E state is fast, with Ref. (Huxter et al., 2013) observing a ∼ 4 ps timescale, and Ref. (Ulbricht
et al., 2018) observing a ∼ 50 fs timescale. The 1 A1 lifetime was measured to be ≈ 100 ps at 78K (Ulbricht and Loh, 2018) and is
likely shorter at room temperature.

Type I Type II
Nitrogen impurites No "measurable" nitrogen impurites

Type Ia Type Ib Type IIa Type IIb

Aggregated Isolated single No "measurable" Boron impurities
nitrogen nitrogen impurities
impurites impurities

C C N C C C C C C C C C
Type IaA Type IaB C C C C C C C C C B C C
Aggregated Aggregated
A center B center C N C C C C C C C C C C
impurites impurites
C C C C C C C C C C C B
N N C C C C N C

C C C C C N V N
C = carbon atom B = boron atom
C C N C C C N C
N = nitrogen atom V = lattice vacancy
C C N C C C C C

FIG. 36 The diamond type classification system as described in the main text. Adapted from Ref. (Breeding and Shigley, 2009)
.
 61

Acronym Description
CPMG Carr-Purcell-Meiboom-Gill (pulse sequences)
CW Continuous wave
CVD Chemical vapor deposition
DEER Double electron-electron resonance
DQ Double-quantum
EPR Electron paramagnetic resonance
ESLAC Excited-state level anti-crossing
ESR Electron spin resonance
FID Free induction decay
GSLAC Ground state level anti-crossing
HPHT High pressure high temperature
LAC Level anti-crossing
LPHT Low pressure high temperature
MW Microwave
NIR Near-infrared
NMR Nuclear magnetic resonance
NQR Nuclear quadrupole resonance
ODMR Optically detected magnetic resonance
PDMR Photoelectrically detected magnetic resonance
PE Photoelectric (readout)
PL Photoluminescence
QND Quantum non-demolition
RF Radiofrequency
SCC Spin-to-charge conversion
SQ Single quantum (standard basis)

TABLE 14 Frequently used acronyms

 62

Quantity Symbol Units Notes

Longitudinal (spin-lattice) relaxation time T1 s
Coherence time (transverse relaxation time) T2 s
Dephasing time (free induction decay time) T2∗ s
Single-NV- dephasing time T2∗ {single} s
√
Magnetic field sensitivity η T/ Hz
Interrogation time (free-precession time for Ramsey) τ s
Initialization, readout, and overhead time tI , tR , tO s tO ≡ tI + tR
Stretched exponential parameter p -
Static (bias) & microwave magnetic field B0 & B1 T
Electronic spin gyromagnetic ratio γe s-1 /T ≡ ge µB /~
Readout fidelity F - ≡ 1/σR
Factor above spin projection noise σR - ≡ 1/F
Rabi frequency ΩR s-1
ODMR center frequency & linewidth ν & ∆ν Hz
Dephasing or decay rate Γ s-1
Measurement contrast (fringe visibility) C -
CW-ODMR contrast, pulsed ODMR contrast CCW , Cpulsed -
Number of sensors (NV- centers in ensemble) N -
Average collected photons per readout per NV- navg -
Average collected photons per readout from an NV- ensemble N -

Concentration of species X [X] cm-3 or ppm

Negative, neutral & total NV concentration [NV- ], [NV0 ], [NVT ] cm-3 or ppm
Total nitrogen concentration in the lattice [NT ] cm-3 or ppm
Neutral, positive, total substitutional nitrogen concentration [N0S ], [N+ T
S ], [NS ] cm-3 or ppm

Contribution to T2∗ from mechanism X T2∗ {X} s

Dipolar interaction strength between N0S and NV- AN0 s-1 /ppm
S
Dipolar interaction strength between 13 C and NV- A13 C s-1 /ppm
Dipolar interaction strength between NV- spins in the same group ANV-k s-1 /ppm
(same resonance frequency)
Dipolar interaction strength between NV- spins in different groups ANV-∦ s-1 /ppm
(different resonance frequencies)
Proportionality factor for N0S contribution to T2 BN0 s-1 /ppm
S

Hamiltonian H J
Electronic spin, electronic spin projection S, ms -
Nuclear spin, nuclear spin projection I, mI -
NV- ground state spin eigenstates {| 0i, |−1i, |+1i} -
Zero field splitting parameter D Hz ≈ 2.87 GHz
Spin-strain coupling parameters Mz , Mx , My , Nx , Ny Hz
Electric field components Ex , Ey , Ez V/m
NV- transverse, axial (longitudinal) electric dipole moment d⊥ , dk Hz/(V/m)
Transverse strain and electric field coupling parameter ξ⊥ Hz
gµB
Axial magnetic field coupling parameter βz Hz ≡ h
Bz

Total N-to-NV- conversion efficiency Econv - ≡ [NV- ]/[NT ]

N-to-NV conversion efficiency χ - ≡ [NVT ]/[NT ]
NV-to-NV- charge state efficiency ζ - ≡ [NV- ]/[NVT ]

TABLE 15 Frequently used symbols

 63

Const. Description Value Ref.

gk Axial g-factor 2.0028 ± 0.0003 (Loubser and van Wyk, 1978)
2.0029 ± 0.0002 (Felton et al., 2009)

g⊥ Transverse g-factor 2.0028 ± 0.0003 (Loubser and van Wyk, 1978)

2.0031 ± 0.0002 (Felton et al., 2009)

14
Ak N axial magnetic hyperfine constant ±2.32 ± 0.01 MHz (Loubser and van Wyk, 1978)
2.30 ± 0.02 MHz (He et al., 1993)
−2.14 ± 0.07 MHz (Felton et al., 2008)
−2.166 ± 0.01 MHz (Steiner et al., 2010)
−2.162 ± 0.002 MHz (Smeltzer et al., 2009)

15
N axial magnetic hyperfine constant −3.1 MHz (Rabeau et al., 2006)
3.01 ± 0.05 MHz (Fuchs et al., 2008)
3.03 ± 0.03 MHz (Felton et al., 2009)

14
A⊥ N transverse magnetic hyperfine constant +2.10 ± 0.10 MHz (He et al., 1993)
−2.70 ± 0.07 MHz (Felton et al., 2008)

15
N transverse magnetic hyperfine constant −3.1 MHz (Rabeau et al., 2006)
3.01 ± 0.05 MHz (Fuchs et al., 2008)
3.65 ± 0.03 MHz (Felton et al., 2009)

14
P N nuclear electric quadrupole parameter −5.04 ± 0.05 (He et al., 1993)
−5.01 ± 0.06 (Felton et al., 2008)
−4.945 ± 0.01 (Steiner et al., 2010)
−4.945 ± 0.005 (Smeltzer et al., 2009)

dk Axial dipole moment 3.5 ± 0.02 × 10−3 Hz/(V/m) (Van Oort and Glasbeek, 1990)

d⊥ Transverse dipole moment 0.165 ± 0.007 Hz/(V/m) (Michl et al., 2019)

0.175 ± 0.030 Hz/(V/m) (Dolde et al., 2011)
0.17 ± 0.025 Hz/(V/m) (Van Oort and Glasbeek, 1990)

TABLE 16 Compiled constants for the electronic ground state of the NV- center in diamond. Data are reproduced in part from
Ref. (Doherty et al., 2013).

Cross section Value Reference

σ3 A2 →3 E (λ = 532 nm) (3.1 ± 0.8) × 10−17 cm2 (Wee et al., 2007)
σ3 A2 →3 E (λ = 532 nm) (9.5 ± 2.5) × 10−17 cm2 (Chapman and Plakhotnik, 2011)
σ3 A2 →3 E (λ = 532 nm) (2.6 ± 0.5) × 10−17 cm2 (Fraczek et al., 2017)
σ3 A2 →3 E (λ = 532 nm) 2.4 × 10−17 cm2 (Subedi et al., 2019)

TABLE 17 Absorption cross section at 532 nm for the NV- 3 A2 →3 E transition. The value from Ref. (Fraczek et al., 2017) was
calculated from their data under the assumption that the NV- and NV0 absorption cross sections are equal to within 2×.
 64

REFERENCES Babinec, T. M., H. Fedder, J. T. Choy, I. Bulu, M. W. Doherty,

P. R. Hemmer, J. Wrachtrup, and M. Loncar (2012), in Con-
Abe, E., and K. Sasaki (2018), J. Appl. Phys. 123 (16), 161101. ference on Lasers and Electro-Optics 2012 (Optical Society
Abe, E., A. M. Tyryshkin, S. Tojo, J. J. L. Morton, W. M. of America) p. JW1I.6.
Witzel, A. Fujimoto, J. W. Ager, E. E. Haller, J. Isoya, S. A. Babinec, T. M., H. J. M., M. Khan, Y. Zhang, J. R. Maze, P. R.
Lyon, M. L. W. Thewalt, and K. M. Itoh (2010), Phys. Rev. Hemmer, and M. Loncar (2010), Nat. Nanotechnol. 5 (3),
B 82, 121201. 195.
Abragam, A. (1983a), The Principles of Nuclear Magnetism Backlund, M. P., P. Kehayias, and R. L. Walsworth (2017),
(Clarendon Press, Oxford) Chap. 4. Phys. Rev. Applied 8, 054003.
Abragam, A. (1983b), The Principles of Nuclear Magnetism Baker, J. (2007), Diam. Relat. Mater. 16 (2), 216.
(Clarendon Press, Oxford) Chap. 3. Balasubramanian, G., I. Y. Chan, R. Kolesov, M. Al-Hmoud,
Abragam, A. (1983c), The Principles of Nuclear Magnetism J. Tisler, C. Shin, C. Kim, A. Wojcik, P. R. Hemmer,
(Clarendon Press, Oxford). A. Krueger, T. Hanke, A. Leitenstorfer, R. Bratschitsch,
Achard, J., A. Tallaire, V. Mille, M. Naamoun, O. Brinza, F. Jelezko, and J. Wrachtrup (2008), Nature 455, 648.
A. Boussadi, L. William, and A. Gicquel (2014), Phys. Status Balasubramanian, G., P. Neumann, D. Twitchen, M. Markham,
Solidi A 211 (10), 2264. R. Kolesov, N. Mizuochi, J. Isoya, J. Achard, J. Beck,
Acosta, V. M. (2011), Optical Magnetometry with Nitrogen- J. Tissler, V. Jacques, P. R. Hemmer, F. Jelezko, and
Vacancy Centers in Diamond, Ph.D. thesis (University of Cal- J. Wrachtrup (2009), Nat. Mater. 8 (5), 383.
ifornia Berkeley). Balmer, R. S., J. R. Brandon, S. L. Clewes, H. K. Dhillon,
Acosta, V. M., E. Bauch, A. Jarmola, L. J. Zipp, M. P. Led- J. M. Dodson, I. Friel, P. N. Inglis, T. D. Madgwick, M. L.
better, and D. Budker (2010), Appl. Phys. Lett. 97 (17), Markham, T. P. Mollart, N. Perkins, G. A. Scarsbrook, D. J.
174104. Twitchen, A. J. Whitehead, J. J. Wilman, and S. M. Wool-
Acosta, V. M., E. Bauch, M. P. Ledbetter, C. Santori, K.-M. C. lard (2009), J. Phys.: Condens. Matter 21 (36), 364221.
Fu, P. E. Barclay, R. G. Beausoleil, H. Linget, J. F. Roch, Bangert, U., R. Barnes, M. H. Gass, A. L. Bleloch, and I. S.
F. Treussart, S. Chemerisov, W. Gawlik, and D. Budker Godfrey (2009), J. Phys.: Condens. Matter 21 (36), 364208.
(2009), Phys. Rev. B 80, 115202. Bar-Gill, N., L. M. Pham, C. Belthangady, D. Le Sage,
Acosta, V. M., E. Bauch, M. P. Ledbetter, A. Waxman, L.-S. P. Cappellaro, J. R. Maze, M. D. Lukin, A. Yacoby, and
Bouchard, and D. Budker (2010a), Phys. Rev. Lett. 104, R. Walsworth (2012), Nat. Commun. 3, 858.
070801. Bar-Gill, N., L. M. Pham, A. Jarmola, D. Budker, and R. L.
Acosta, V. M., A. Jarmola, E. Bauch, and D. Budker (2010b), Walsworth (2013), Nat. Commun. 4, 1743.
Phys. Rev. B 82 (20), 201202. Baranov, P. G., H. J. von Bardeleben, F. Jelezko, and
Ahmadi, S., H. A. R. El-Ella, J. O. B. Hansen, A. Huck, and J. Wrachtrup (2017), “Retrospectives: Magnetic resonance
U. L. Andersen (2017), Phys. Rev. Applied 8, 034001. studies of intrinsic defects in semiconductors,” in Magnetic
Ahmadi, S., H. A. R. El-Ella, A. M. Wojciechowski, T. Gehring, Resonance of Semiconductors and Their Nanostructures: Ba-
J. O. B. Hansen, A. Huck, and U. L. Andersen (2018a), in sic and Advanced Applications (Springer, Vienna) pp. 179–
Conference on Lasers and Electro-Optics (Optical Society of 211.
America) p. JF3B.2. Barfuss, A., M. Kasperczyk, J. Kölbl, and P. Maletinsky (2018),
Ahmadi, S., H. A. R. El-Ella, A. M. Wojciechowski, T. Gehring, ArXiv e-prints arXiv:1810.07190 [cond-mat.mes-hall].
J. O. B. Hansen, A. Huck, and U. L. Andersen (2018b), Phys. Barry, J. F., M. J. Turner, J. M. Schloss, D. R. Glenn, Y. Song,
Rev. B 97, 024105. M. D. Lukin, H. Park, and R. L. Walsworth (2016), Proc.
Aiello, C. D., M. Hirose, and P. Cappellaro (2013), Nat. Com- Natl. Acad. Sci. 113 (49), 14133.
mun. 4, 1419. Barson, M. S. J., E. Krausz, N. B. Manson, and M. W. Doherty
Alsid, S. T., J. F. Barry, L. M. Pham, J. M. Schloss, M. F. (2019), ArXiv e-prints arXiv:1905.09037 [cond-mat.mes-hall].
O’Keeffe, P. Cappellaro, and D. A. Braje (2019), ArXiv e- Barson, M. S. J., P. Peddibhotla, P. Ovartchaiyapong, K. Gane-
prints 1906.11406. san, R. L. Taylor, M. Gebert, Z. Mielens, B. Koslowski, D. A.
Ammerlaan, C. A. J., and E. A. Burgemeister (1981), Phys. Simpson, L. P. McGuinness, J. McCallum, S. Prawer, S. On-
Rev. Lett. 47 (13), 954. oda, T. Ohshima, A. C. Bleszynski Jayich, F. Jelezko, N. B.
Angerer, A., T. Nöbauer, G. Wachter, M. Markham, A. Stacey, Manson, and M. W. Doherty (2017), Nano Lett. 17 (3), 1496.
J. Majer, J. Schmiedmayer, and M. Trupke (2015), ArXiv Bauch, E. (2010), Nitrogen-Vacancy Defects in Diamond for sub-
e-prints arXiv:1509.01637 [quant-ph]. millimeter Magnetometry, Master’s thesis (Technische Uni-
Angus, J. C., H. A. Will, and W. S. Stanko (1968), J. Appl. versität Berlin and University of California Berkeley).
Phys. 39 (6), 2915. Bauch, E., C. A. Hart, J. M. Schloss, M. J. Turner, J. F. Barry,
Arai, K., C. Belthangady, H. Zhang, N. Bar-Gill, S. J. DeVience, P. Kehayias, S. Singh, and R. L. Walsworth (2018), Phys.
P. Cappellaro, A. Yacoby, and R. L. Walsworth (2015), Nat. Rev. X 8, 031025.
Nanotechnol. 10 (10), 859. Bauch, E., S. Singh, J. Lee, C. A. Hart, J. M. Schloss, M. J.
Arai, K., J. Lee, C. Belthangady, D. R. Glenn, H. Zhang, and Turner, J. F. Barry, L. Pham, N. Bar-Gill, S. F. Yelin,
R. L. Walsworth (2018), Nat. Commun. 9 (1), 4996. and R. L. Walsworth (2019), arXiv e-prints arXiv:1904.08763
Ariyaratne, A., D. Bluvstein, B. A. Myers, and A. C. B. Jayich [quant-ph].
(2018), Nat. Commun. 9 (1), 2406. Beha, K., A. Batalov, N. B. Manson, R. Bratschitsch, and
Arroyo-Camejo, S., A. Lazariev, S. W. Hell, and G. Balasubra- A. Leitenstorfer (2012), Phys. Rev. Lett. 109, 097404.
manian (2014), Nat. Commun. 5, 4870. Belthangady, C., N. Bar-Gill, L. M. Pham, K. Arai, D. Le Sage,
Ashbaugh III, C. E. (1988), Gems Gemol. 24 (4). P. Cappellaro, and R. L. Walsworth (2013), Phys. Rev. Lett.
Aslam, N., M. Pfender, P. Neumann, R. Reuter, A. Zappe, 110, 157601.
F. Fávaro de Oliveira, A. Denisenko, H. Sumiya, S. Onoda, Bernien, H., B. Hensen, W. Pfaff, G. Koolstra, M. S. Blok,
J. Isoya, and J. Wrachtrup (2017), Science 357 (6346), 67. L. Robledo, T. H. Taminiau, M. Markham, D. J. Twitchen,
Aslam, N., G. Waldherr, P. Neumann, F. Jelezko, and L. Childress, and R. Hanson (2013), Nature 497, 86.
J. Wrachtrup (2013), New J. Phys. 15 (1), 013064. Berry, M. V. (1984), Proc. R. Soc. London, Ser. A 392 (1802),
 65

45. 507 (1), 190.

Bluvstein, D., Z. Zhang, and A. C. B. Jayich (2019), Phys. Rev. Charles, S. J., J. E. Butler, B. N. Feygelson, M. E. Newton, D. L.
Lett. 122, 076101. Carroll, J. W. Steeds, H. Darwish, C.-S. Yan, H. K. Mao, and
Boesch, B. P., G. S. Bruce, J. D. Cammerata, D. N. Coar, L. N. R. J. Hemley (2004), Phys. Status Solidi A 201 (11), 2473.
Egan, B. N. Fisk, W. Lew, A. Manickam, S. M. Sekelsky, Chatzidrosos, G., A. Wickenbrock, L. Bougas, N. Leefer, T. Wu,
J. B. Stetson, P. G. Kaup, J. L. Miller, J. C. Russo, and K. Jensen, Y. Dumeige, and D. Budker (2017), Phys. Rev.
E. S. Stockman (2016), “DNV magnetic field detector,” WO Applied 8, 044019.
Patent 2016/118791 Al. Chen, X.-D., S. Li, A. Shen, Y. Dong, C.-H. Dong, G.-C. Guo,
Bogdanov, S., M. Y. Shalaginov, A. Akimov, A. S. Lagutchev, and F.-W. Sun (2017), Phys. Rev. Applied 7, 014008.
P. Kapitanova, J. Liu, D. Woods, M. Ferrera, P. Belov, J. Iru- Childress, L., M. V. Gurudev Dutt, J. M. Taylor, A. S. Zibrov,
dayaraj, A. Boltasseva, and V. M. Shalaev (2017), Phys. Rev. F. Jelezko, J. Wrachtrup, P. R. Hemmer, and M. D. Lukin
B 96, 035146. (2006), Science 314 (5797), 281.
Bonato, C., M. S. Blok, H. T. Dinani, D. W. Berry, M. L. Choi, J., S. Choi, G. Kucsko, P. C. Maurer, B. J. Shields,
Markham, D. J. Twitchen, and R. Hanson (2015), Nat. Nan- H. Sumiya, S. Onoda, J. Isoya, E. Demler, F. Jelezko, N. Y.
otechnol. 11, 247. Yao, and M. D. Lukin (2017a), Phys. Rev. Lett. 118, 093601.
Boss, J. M., K. Chang, J. Armijo, K. Cujia, T. Rosskopf, J. R. Choi, S., N. Y. Yao, and M. D. Lukin (2017b), Phys. Rev. Lett.
Maze, and C. L. Degen (2016), Phys. Rev. Lett. 116, 197601. 119, 183603.
Boss, J. M., K. S. Cujia, J. Zopes, and C. L. Degen (2017), Choi, S., N. Y. Yao, and M. D. Lukin (2017), ArXiv e-prints
Science 356 (6340), 837. arXiv:1801.00042 [quant-ph].
Botsoa, J., T. Sauvage, M.-P. Adam, P. Desgardin, E. Leoni, Choy, J. T., I. Bulu, B. J. M. Hausmann, E. Janitz, I.-C. Huang,
B. Courtois, F. Treussart, and M.-F. Barthe (2011), Phys. and M. Loncar (2013), Appl. Phys. Lett. 103 (16), 161101.
Rev. B 84, 125209. Choy, J. T., B. J. M. Hausmann, T. M. Babinec, I. Bulu,
Bourgeois, E., A. Jarmola, P. Siyushev, M. Gulka, J. Hruby, M. Khan, P. Maletinsky, A. Yacoby, and M. Loncar (2011),
F. Jelezko, D. Budker, and M. Nesladek (2015), Nat. Com- Nat. Photonics 5 (12), 738.
mun. 6, 8577. Chrostoski, P., H. R. Sadeghpour, and D. H. Santamore (2018),
Bourgeois, E., E. Londero, K. Buczak, J. Hruby, M. Gulka, Phys. Rev. Applied 10, 064056.
Y. Balasubramaniam, G. Wachter, J. Stursa, K. Dobes, Chu, Y., N. de Leon, B. Shields, B. Hausmann, R. Evans, E. To-
F. Aumayr, M. Trupke, A. Gali, and M. Nesladek (2017), gan, M. J. Burek, M. Markham, A. Stacey, A. Zibrov, A. Ya-
Phys. Rev. B 95, 041402. coby, D. Twitchen, M. Loncar, H. Park, P. Maletinsky, and
Boyd, S. R., I. Kiflawi, and G. S. Woods (1994), Philos. Mag. M. Lukin (2014), Nano Lett. 14 (4), 1982.
B 69 (6), 1149. Clevenson, H., M. E. Trusheim, C. Teale, T. Schroder, D. Braje,
Braun, D., G. Adesso, F. Benatti, R. Floreanini, U. Marzolino, and D. Englund (2015), Nat. Phys. 11 (5), 393.
M. W. Mitchell, and S. Pirandola (2018), Rev. Mod. Phys. Collins, A. T. (2002), J. Phys.: Condens. Matter 14 (14), 3743.
90, 035006. Collins, A. T. (2007), New Diamond and Frontier Carbon Tech-
Breeding, C. M., and J. E. Shigley (2009), Gems Gemol. 45 (2), nology 17.
96. Collins, A. T., H. Kanda, and H. Kitawaki (2000), Diam. Relat.
Breeze, J. D., E. Salvadori, J. Sathian, N. M. Alford, and Mater. 9 (2), 113.
C. W. M. Kay (2018), Nature 555, 493. Cook, R. J., and D. H. Whiffen (1966), Proc. R. Soc. London,
Breuer, S. J., and P. R. Briddon (1995), Phys. Rev. B 51, 6984. Ser. A 295 (1441), 99.
Bube, R. (1960), Photoconductivity of solids (Wiley). Cramer, J., N. Kalb, M. A. Rol, B. Hensen, M. S. Blok,
Buchan, J. T., M. Robinson, H. J. Christie, D. L. Roach, D. K. M. Markham, D. J. Twitchen, R. Hanson, and T. H.
Ross, and N. A. Marks (2015), J. Appl. Phys. 117 (24), Taminiau (2016), Nat. Commun. 7, 11526.
245901. Cruddace, R. (2007), Magnetic resonance and optical studies
Bucher, D. B., D. R. Glenn, H. Park, M. D. Lukin, and of point defects in single crystal CVD diamond, Ph.D. thesis
R. L. Walsworth (2018), ArXiv e-prints arXiv:1810.02408 (University of Warwick).
[physics.app-ph]. Cui, S., and E. L. Hu (2013), Appl. Phys. Lett. 103 (5), 051603.
Budker, D., and M. Romalis (2007), Nat. Phys. 3 (4), 227. Cywiński, L., R. M. Lutchyn, C. P. Nave, and S. Das Sarma
Bursill, L. A., and R. W. Glaisher (1985), Am. Mineral. 70, (2008), Phys. Rev. B 77, 174509.
608. Dale, M. W. (2015), Colour Centres on Demand in Diamond,
Bylander, J., S. Gustavsson, F. Yan, F. Yoshihara, K. Harrabi, Ph.D. thesis (University of Warwick).
G. Fitch, D. G. Cory, Y. Nakamura, J.-S. Tsai, and W. D. Dannefaer, S., T. Bretagnon, and D. Kerr (1993), Diam. Relat.
Oliver (2011), Nature Physics 7, 565. Mater. 2 (12), 1479.
Cai, J., F. Jelezko, M. B. Plenio, and A. Retzker (2013), New Dannefaer, S., P. Mascher, and D. Kerr (1992), Diam. Relat.
J. Phys. 15 (1), 013020. Mater. 1 (5), 407.
Campbell, B., W. Choudhury, A. Mainwood, M. Newton, and Davies, G. (1976), J. Phys. C: Solid State Phys. 9 (19), L537.
G. Davies (2002), Nucl. Instrum. Methods Phys. Res., Sect. Davies, G. (1979), J. Phys. C: Solid State Phys. 12 (13), 2551.
A 476 (3), 680. Davies, G. (1981), Rep. Prog. Phys. 44 (7), 787.
Campbell, B., and A. Mainwood (2000), Phys. Status Solidi A Davies, G., B. Campbell, A. Mainwood, M. Newton,
181 (1), 99. M. Watkins, H. Kanda, and T. Anthony (2001), Phys. Status
Cann, B. L. (2009), Magnetic resonance studies of point defects Solidi A 186 (2), 187.
in diamond, Ph.D. thesis (University of Warwick). Davies, G., and T. Evans (1972), Proc. R. Soc. London, Ser. A
Cao, G. Z., F. A. J. M. Driessen, G. J. Bauhuis, L. J. Giling, 328 (1574), 413.
and P. F. A. Alkemade (1995), J. Appl. Phys. 78 (5), 3125. Davies, G., and M. F. Hamer (1976), Proc. R. Soc. London,
Cao, J. (1994), J. Magn. Reson., Ser B 103 (3), 296. Ser. A 348 (1653), 285.
Casola, F., T. van der Sar, and A. Yacoby (2018), Nat. Rev. Davies, G., S. C. Lawson, A. T. Collins, A. Mainwood, and S. J.
Mater. 3, 17088. Sharp (1992), Phys. Rev. B 46, 13157.
Chapman, R., and T. Plakhotnik (2011), Chem. Phys. Lett. Davis, H. C., P. Ramesh, A. Bhatnagar, A. Lee-Gosselin, J. F.
 66

Barry, D. R. Glenn, R. L. Walsworth, and M. G. Shapiro dace, M. E. Newton, K.-M. C. Fu, C. Santori, R. G. Beau-
(2018), Nat. Commun. 9 (1), 131. soleil, D. J. Twitchen, and M. L. Markham (2012), Phys.
Deák, P., B. Aradi, M. Kaviani, T. Frauenheim, and A. Gali Rev. B 86, 035201.
(2014), Phys. Rev. B 89, 075203. Eichhorn, T. R., C. A. McLellan, and A. C. B. Jayich (2019),
Degen, C. L. (2008), Appl. Phys. Lett. 92 (24), 243111. ArXiv e-prints 1901.11519 [cond-mat.mtrl-sci].
Degen, C. L., F. Reinhard, and P. Cappellaro (2017), Rev. Mod. Eisenach, E. R., J. F. Barry, L. M. Pham, R. G. Rojas, D. R.
Phys. 89, 035002. Englund, and D. A. Braje (2018), Rev. Sci. Instrum. 89 (9),
DeGiorgio, V., and M. O. Scully (1970), Phys. Rev. A 2, 1170. 094705.
DeVience, S. J., L. M. Pham, I. Lovchinsky, A. O. Sushkov, El-Ella, H. A. R., S. Ahmadi, A. M. Wojciechowski, A. Huck,
N. Bar-Gill, C. Belthangady, F. Casola, M. Corbett, and U. L. Andersen (2017), Opt. Express 25 (13), 14809.
H. Zhang, M. Lukin, H. Park, A. Yacoby, and R. L. Epstein, R. J., F. M. Mendoza, Y. K. Kato, and D. D.
Walsworth (2015), Nat. Nanotechnol. 10, 129 . Awschalom (2005), Nat. Phys. 1, 94.
D’Haenens-Johansson, U. F., A. Katrusha, P. Johnson, Fang, K., V. M. Acosta, C. Santori, Z. Huang, K. M. Itoh,
W. Wang, et al. (2015), Gems Gemol. 51 (3). H. Watanabe, S. Shikata, and R. G. Beausoleil (2013), Phys.
D’Haenens-Johansson, U. F., K. S. Moe, P. Johnson, S. Y. Rev. Lett. 110, 130802.
Wong, R. Lu, and W. Wang (2014), Gems Gemol. 50 (1). Farchi, E., Y. Ebert, D. Farfurnik, G. Haim, R. Shaar, and
Dischler, B. (2012), Handbook of Spectral Lines in Diamond: N. Bar-Gill (2017), Spin 07 (03), 1740015.
Volume 1: Tables and Interpretations, Vol. 1 (Springer). Farfurnik, D., N. Alfasi, S. Masis, Y. Kauffmann, E. Farchi,
Dobrinets, I. A., V. G. Vins, and A. M. Zaitsev (2013), HPHT- Y. Romach, Y. Hovav, E. Buks, and N. Bar-Gill (2017),
Treated Diamonds: Diamonds Forever (Springer). Appl. Phys. Lett. 111 (12), 123101.
Dobrovitski, V. V., A. E. Feiguin, D. D. Awschalom, and Farfurnik, D., A. Jarmola, L. M. Pham, Z. H. Wang, V. V.
R. Hanson (2008), Phys. Rev. B 77 (24), 245212. Dobrovitski, R. L. Walsworth, D. Budker, and N. Bar-Gill
Dodson, J., J. R Brandon, H. Dhillon, I. Friel, S. L Geoghe- (2015), Phys. Rev. B 92, 060301.
gan, T. Mollart, P. Santini, G. Scarsbrook, D. Twitchen, Farrer, R. (1969), Solid State Commun. 7 (9), 685.
A. J Whitehead, J. J Wilman, and H. de Wit (2011), Proc. Fastenau, R. (Jun 1982), Diffusion limited reactions in crys-
of SPIE 8016, 80160L. talline solids, Ph.D. thesis (Technische Hogeschool Delft).
Doherty, M. W., F. Dolde, H. Fedder, F. Jelezko, J. Wrachtrup, Fávaro de Oliveira, F., D. Antonov, Y. Wang, P. Neumann, S. A.
N. B. Manson, and L. C. L. Hollenberg (2012), Phys. Rev. B Momenzadeh, T. Häußermann, A. Pasquarelli, A. Denisenko,
85, 205203. and J. Wrachtrup (2017), Nat. Commun. 8, 15409.
Doherty, M. W., N. B. Manson, P. Delaney, F. Jelezko, Fávaro de Oliveira, F., S. A. Momenzadeh, D. Antonov,
J. Wrachtrup, and L. C. Hollenberg (2013), Phys. Rep. J. Scharpf, C. Osterkamp, B. Naydenov, F. Jelezko,
528 (1), 1. A. Denisenko, and J. Wrachtrup (2016), Nano Lett. 16 (4),
Doherty, M. W., C. A. Meriles, A. Alkauskas, H. Fedder, M. J. 2228.
Sellars, and N. B. Manson (2016), Phys. Rev. X 6, 041035. Fedder, H., F. Dolde, F. Rempp, T. Wolf, P. Hemmer, F. Jelezko,
Doherty, M. W., V. V. Struzhkin, D. A. Simpson, L. P. McGuin- and J. Wrachtrup (2011), Appl. Phys. B 102 (3), 497.
ness, Y. Meng, A. Stacey, T. J. Karle, R. J. Hemley, N. B. Felton, S., B. L. Cann, A. M. Edmonds, S. Liggins, R. J. Crud-
Manson, L. C. L. Hollenberg, and S. Prawer (2014), Phys. dace, M. E. Newton, D. Fisher, and J. M. Baker (2009), J.
Rev. Lett. 112, 047601. Phys.: Condens. Matter 21, 364212.
Doi, Y., T. Fukui, H. Kato, T. Makino, S. Yamasaki, Felton, S., A. M. Edmonds, M. E. Newton, P. M. Martineau,
T. Tashima, H. Morishita, S. Miwa, F. Jelezko, Y. Suzuki, D. Fisher, and D. J. Twitchen (2008), Phys. Rev. B 77,
and N. Mizuochi (2016), Phys. Rev. B 93, 081203. 081201.
Doi, Y., T. Makino, H. Kato, D. Takeuchi, M. Ogura, H. Okushi, Fescenko, I., A. Laraoui, J. Smits, N. Mosavian, P. Kehayias,
H. Morishita, T. Tashima, S. Miwa, S. Yamasaki, P. Neu- J. Seto, L. Bougas, A. Jarmola, and V. M. Acosta (2018),
mann, J. Wrachtrup, Y. Suzuki, and N. Mizuochi (2014), ArXiv e-prints arXiv:1808.03636 [cond-mat.mes-hall].
Phys. Rev. X 4, 011057. Fisher, D., D. J. F. Evans, C. Glover, C. J. Kelly, M. J. Sheehy,
Dolde, F., H. Fedder, M. W. Doherty, T. Nobauer, F. Rempp, and G. C. Summerton (2006), Diam. Relat. Mater. 15, 1636.
G. Balasubramanian, T. Wolf, F. Reinhard, L. C. L. Hollen- Fletcher, R. C., and W. L. Brown (1953), Phys. Rev. 92, 585.
berg, F. Jelezko, and J. Wrachtrup (2011), Nat. Phys. 7 (6), Forneris, J., S. Ditalia Tchernij, A. Tengattini, E. Enrico,
459. V. Grilj, N. Skukan, G. Amato, L. Boarino, M. Jakšić, and
Dolde, F., I. Jakobi, B. Naydenov, N. Zhao, S. Pezzagna, P. Olivero (2017), Carbon 113, 76.
C. Trautmann, J. Meijer, P. Neumann, F. Jelezko, and Fraczek, E., V. G. Savitski, M. Dale, B. G. Breeze, P. Diggle,
J. Wrachtrup (2013), Nat. Phys. 9 (3), 139. M. Markham, A. Bennett, H. Dhillon, M. E. Newton, and
Drake, M., E. Scott, and J. A. Reimer (2016), New J. Phys. A. J. Kemp (2017), Opt. Mater. Express 7 (7), 2571.
18 (1), 13011. Fu, K.-M. C., C. Santori, P. E. Barclay, and R. G. Beausoleil
Dréau, A., M. Lesik, L. Rondin, P. Spinicelli, O. Arcizet, J.-F. (2010), Appl. Phys. Lett. 96 (12), 121907.
Roch, and V. Jacques (2011), Phys. Rev. B 84 (19), 195. Fu, R. R., B. P. Weiss, E. A. Lima, R. J. Harrison, X.-N. Bai,
Dréau, A., J.-R. Maze, M. Lesik, J.-F. Roch, and V. Jacques S. J. Desch, D. S. Ebel, C. Suavet, H. Wang, D. Glenn,
(2012), Phys. Rev. B 85, 134107. D. Le Sage, T. Kasama, R. L. Walsworth, and A. T. Kuan
Dumeige, Y., M. Chipaux, V. Jacques, F. Treussart, J.-F. Roch, (2014), Science 346 (6213), 1089.
T. Debuisschert, V. M. Acosta, A. Jarmola, K. Jensen, P. Ke- Fuchs, G. D., V. V. Dobrovitski, R. Hanson, A. Batra, C. D.
hayias, and D. Budker (2013), Phys. Rev. B 87, 155202. Weis, T. Schenkel, and D. D. Awschalom (2008), Phys. Rev.
Dwyer, K. J., J. M. Pomeroy, and D. S. Simons (2013), Appl. Lett. 101, 117601.
Phys. Lett. 102 (25), 254104. Fujita, N., R. Jones, S. Öberg, and P. Briddon (2009), Diam.
Edmonds, A. M. (2008), Magnetic resonance studies of point Relat. Mater. 18 (5–8), 843.
defects in single crystal diamond, Ph.D. thesis (University of Fukui, T., Y. Doi, T. Miyazaki, Y. Miyamoto, H. Kato, T. Mat-
Warwick). sumoto, T. Makino, S. Yamasaki, R. Morimoto, N. Tokuda,
Edmonds, A. M., U. F. S. D’Haenens-Johansson, R. J. Crud- M. Hatano, Y. Sakagawa, H. Morishita, T. Tashima, S. Miwa,
 67

Y. Suzuki, and N. Mizuochi (2014), Appl. Phys. Express Hadden, J. P., J. P. Harrison, A. C. Stanley-Clarke, L. Marseglia,
7 (5), 055201. Y. L. D. Ho, B. R. Patton, J. L. O’Brien, and J. G. Rarity
Gaillou, E., J. E. Post, D. Rost, and J. E. Butler (2012), Am. (2010), Appl. Phys. Lett. 97 (24), 241901.
Mineral. 97 (1), 1. Hahn, E. L. (1950), Phys. Rev. 80, 580.
Gammelmark, S., and K. Mølmer (2011), New J. Phys. 13 (5), Hall, L. T., J. H. Cole, and L. C. L. Hollenberg (2014), Phys.
053035. Rev. B 90, 075201.
Gaukroger, M., P. Martineau, M. Crowder, I. Friel, S. Williams, Hall, L. T., P. Kehayias, D. A. Simpson, A. Jarmola, A. Stacey,
and D. Twitchen (2008), Diam. Relat. Mater. 17 (3), 262. D. Budker, and L. C. L. Hollenberg (2016), Nat. Commun.
Glenn, D. R., D. B. Bucher, J. Lee, M. D. Lukin, H. Park, and 7, 10211.
R. L. Walsworth (2018), Nature 555 (7696), 351. Hämäläinen, M., R. Hari, R. J. Ilmoniemi, J. Knuutila, and
Glenn, D. R., R. R. Fu, P. Kehayias, D. Le Sage, E. A. Lima, O. V. Lounasmaa (1993), Rev. Mod. Phys. 65 (2), 413.
B. P. Weiss, and R. L. Walsworth (2017), Geochem. Geophys. Hanson, R., V. V. Dobrovitski, A. E. Feiguin, O. Gywat, and
Geosyst. 18 (8), 3254. D. D. Awschalom (2008), Science 320 (5874), 352.
Glenn, D. R., K. Lee, H. Park, R. Weissleder, A. Yacoby, M. D. Hanson, R., F. M. Mendoza, R. J. Epstein, and D. D.
Lukin, H. Lee, R. L. Walsworth, and C. B. Connolly (2015), Awschalom (2006), Phys. Rev. Lett. 97, 087601.
Nat. Methods 12 (8), 736. Hartland, C. B. (2014), A Study of Point Defects in CVD Di-
Glover, C. (2003), A Study of Defects in Single Crystal CVD amond Using Electron Paramagnetic Resonance and Optical
Diamond, Ph.D. thesis (University of Warwick). Spectroscopy, Ph.D. thesis (University of Warwick).
Glover, C., M. E. Newton, P. Martineau, D. J. Twitchen, and Hauf, M. V., B. Grotz, B. Naydenov, M. Dankerl, S. Pezzagna,
J. M. Baker (2003), Phys. Rev. Lett. 90, 185507. J. Meijer, F. Jelezko, J. Wrachtrup, M. Stutzmann, F. Rein-
Glover, C., M. E. Newton, P. M. Martineau, S. Quinn, and hard, and J. A. Garrido (2011), Phys. Rev. B 83, 081304.
D. J. Twitchen (2004), Phys. Rev. Lett. 92, 135502. Hauf, M. V., P. Simon, N. Aslam, M. Pfender, P. Neumann,
Goldman, M. L., M. W. Doherty, A. Sipahigil, N. Y. Yao, S. D. S. Pezzagna, J. Meijer, J. Wrachtrup, M. Stutzmann, F. Rein-
Bennett, N. B. Manson, A. Kubanek, and M. D. Lukin hard, and J. Garrido (2014), Nano Lett. 14 (5), 2359.
(2015a), Phys. Rev. B 91, 165201. He, X.-F., N. B. Manson, and P. T. H. Fisk (1993), Phys. Rev.
Goldman, M. L., A. Sipahigil, M. W. Doherty, N. Y. Yao, B 47, 8816.
S. D. Bennett, M. Markham, D. J. Twitchen, N. B. Man- Hensen, B., H. Bernien, A. E. Dréau, A. Reiserer, N. Kalb,
son, A. Kubanek, and M. D. Lukin (2015b), Phys. Rev. Lett. M. S. Blok, J. Ruitenberg, R. F. L. Vermeulen, R. N.
114, 145502. Schouten, C. Abellán, W. Amaya, V. Pruneri, M. W. Mitchell,
Goss, J. P., P. R. Briddon, V. Hill, R. Jones, and M. J. Rayson M. Markham, D. J. Twitchen, D. Elkouss, S. Wehner, T. H.
(2014), J. Phys.: Condens. Matter 26 (14), 145801. Taminiau, and R. Hanson (2015), Nature 526, 682.
Goss, J. P., P. R. Briddon, S. Papagiannidis, and R. Jones Heremans, F. J., G. D. Fuchs, C. F. Wang, R. Hanson, and
(2004), Phys. Rev. B 70, 235208. D. D. Awschalom (2009), Appl. Phys. Lett. 94 (15), 152102.
Goss, J. P., P. R. Briddon, M. J. Rayson, S. J. Sque, and Hernández-Gómez, S., F. Poggiali, P. Cappellaro, and N. Fabbri
R. Jones (2005), Phys. Rev. B 72, 035214. (2018), Phys. Rev. B 98, 214307.
Green, B. L., B. G. Breeze, and M. E. Newton (2017), J. Phys.: Himics, L., S. Tóth, M. Veres, Z. Balogh, and M. Koós (2014),
Condens. Matter 29 (22), 225701. Appl. Phys. Lett. 104 (9), 093101.
Green, B. L., M. W. Dale, M. E. Newton, and D. Fisher (2015), Hoa, L. T. M., T. Ouisse, D. Chaussende, M. Naamoun, A. Tal-
Phys. Rev. B 92, 165204. laire, and J. Achard (2014), Cryst. Growth Des. 14 (11),
Grezes, C., B. Julsgaard, Y. Kubo, W. L. Ma, M. Stern, 5761.
A. Bienfait, K. Nakamura, J. Isoya, S. Onoda, T. Ohshima, Hobbs, P. C. (2011), Building electro-optical systems: making it
V. Jacques, D. Vion, D. Esteve, R. B. Liu, K. Mølmer, and all work (Wiley).
P. Bertet (2015), Phys. Rev. A 92, 020301. Hopper, D. A., R. R. Grote, A. L. Exarhos, and L. C. Bassett
Grinolds, M. S., M. Warner, K. De Greve, Y. Dovzhenko, (2016), Phys. Rev. B 94, 241201.
L. Thiel, R. L. Walsworth, S. Hong, P. Maletinsky, and A. Ya- Hopper, D. A., R. R. Grote, S. M. Parks, and L. C. Bassett
coby (2014), Nat. Nanotechnol. 9, 279 . (2018a), ACS Nano 12 (5), 4678.
Groot-Berning, K., N. Raatz, I. Dobrinets, M. Lesik, P. Spini- Hopper, D. A., H. J. Shulevitz, and L. C. Bassett (2018b),
celli, A. Tallaire, J. Achard, V. Jacques, J.-F. Roch, A. M. Micromachines 9 (9).
Zaitsev, J. Meijer, and S. Pezzagna (2014), Phys. Status So- Horsley, A., P. Appel, J. Wolters, J. Achard, A. Tallaire,
lidi A 211 (10), 2268. P. Maletinsky, and P. Treutlein (2018), Phys. Rev. Applied
Grosz, A., M. J. Haji-Sheikh, and S. C. Mukhopadhyay (2017), 10, 044039.
High Sensitivity Magnetometers (Springer). Hounsome, L. S., R. Jones, P. M. Martineau, D. Fisher, M. J.
Grotz, B., M. V. Hauf, M. Dankerl, B. Naydenov, S. Pezzagna, Shaw, P. R. Briddon, and S. Öberg (2006), Phys. Rev. B 73,
J. Meijer, F. Jelezko, J. Wrachtrup, M. Stutzmann, F. Rein- 125203.
hard, and J. A. Garrido (2012), Nat. Commun. 3, 729. Hrubesch, F. M., G. Braunbeck, M. Stutzmann, F. Reinhard,
Gulka, M., E. Bourgeois, J. Hruby, P. Siyushev, G. Wachter, and M. S. Brandt (2017), Phys. Rev. Lett. 118, 037601.
F. Aumayr, P. R. Hemmer, A. Gali, F. Jelezko, M. Trupke, Hsieh, S., P. Bhattacharyya, C. Zu, T. Mittiga, T. J. Smart,
and M. Nesladek (2017), Phys. Rev. Applied 7, 044032. F. Machado, B. Kobrin, T. O. Höhn, N. Z. Rui, M. Kamrani,
Gullion, T., D. B. Baker, and M. S. Conradi (1990), J. Magn. S. Chatterjee, S. Choi, M. Zaletel, V. V. Struzhkin, J. E.
Reson. 89 (3), 479. Moore, V. I. Levitas, R. Jeanloz, and N. Y. Yao (2018),
Gupta, A., L. Hacquebard, and L. Childress (2016), J. Opt. ArXiv e-prints arXiv:1812.08796 [cond-mat.mes-hall].
Soc. Am. B 33 (3), B28. Hu, X., Y. Dai, R. Li, H. Shen, and X. He (2002), Solid State
Häberle, T., T. Oeckinghaus, D. Schmid-Lorch, M. Pfender, Commun. 122 (1), 45.
F. F. de Oliveira, S. A. Momenzadeh, A. Finkler, and Huxter, V. M., T. A. A. Oliver, D. Budker, and G. R. Fleming
J. Wrachtrup (2017), Rev. Sci. Instrum. 88 (1), 013702. (2013), Nat. Phys. 9, 744.
Hacquebard, L., and L. Childress (2018), Phys. Rev. A 97, Iakoubovskii, K., G. J. Adriaenssens, and M. Nesladek (2000),
063408. J. Phys.: Condens. Matter 12 (2), 189.
 68

Iakoubovskii, K., and A. Stesmans (2002), Phys. Rev. B 66 (4), Katagiri, M., J. Isoya, S. Koizumi, and H. Kanda (2004), Appl.
045406. Phys. Lett. 85 (26), 6365.
Ishikawa, T., K.-M. C. Fu, C. Santori, V. M. Acosta, R. G. Kato, H., M. Wolfer, C. Schreyvogel, M. Kunzer, W. Müller-
Beausoleil, H. Watanabe, S. Shikata, and K. M. Itoh (2012), Sebert, H. Obloh, S. Yamasaki, and C. Nebel (2013), Appl.
Nano Lett. 12 (4), 2083. Phys. Lett. 102 (15), 151101.
Itano, W. M., J. C. Bergquist, J. J. Bollinger, J. M. Gilligan, Kaupp, H., T. Hümmer, M. Mader, B. Schlederer, J. Benedikter,
D. J. Heinzen, F. L. Moore, M. G. Raizen, and D. J. Wineland P. Haeusser, H.-C. Chang, H. Fedder, T. W. Hänsch, and
(1993), Phys. Rev. A 47, 3554. D. Hunger (2016), Phys. Rev. Applied 6, 054010.
Itoh, K. M., and H. Watanabe (2014), MRS Commun. 4 (4), Kehayias, P., M. W. Doherty, D. English, R. Fischer, A. Jar-
143. mola, K. Jensen, N. Leefer, P. Hemmer, N. B. Manson, and
Ivády, V., T. Simon, J. R. Maze, I. A. Abrikosov, and A. Gali D. Budker (2013), Phys. Rev. B 88, 165202.
(2014), Phys. Rev. B 90, 235205. Kehayias, P., A. Jarmola, N. Mosavian, I. Fescenko, F. M. Ben-
Jacques, V., P. Neumann, J. Beck, M. Markham, D. Twitchen, ito, A. Laraoui, J. Smits, L. Bougas, D. Budker, A. Neumann,
J. Meijer, F. Kaiser, G. Balasubramanian, F. Jelezko, and S. R. J. Brueck, and V. M. Acosta (2017), Nat. Commun.
J. Wrachtrup (2009), Phys. Rev. Lett. 102, 057403. 8 (1), 188.
Jakobi, I., S. A. Momenzadeh, F. F. de Oliveira, J. Michl, Kehayias, P., M. J. Turner, R. Trubko, J. M. Schloss, C. A.
F. Ziem, M. Schreck, P. Neumann, A. Denisenko, and Hart, M. Wesson, D. R. Glenn, and R. L. Walsworth (2019),
J. Wrachtrup (2016), J. Phys.: Conf. Ser. 752 (1), 012001. arXiv e-prints arXiv:1908.09904 [physics.optics].
Jamonneau, P., M. Lesik, J. P. Tetienne, I. Alvizu, L. Mayer, Kessler, E. M., I. Lovchinsky, A. O. Sushkov, and M. D. Lukin
A. Dréau, S. Kosen, J.-F. Roch, S. Pezzagna, J. Meijer, (2014), Phys. Rev. Lett. 112, 150802.
T. Teraji, Y. Kubo, P. Bertet, J. R. Maze, and V. Jacques Khan, R. U. A., B. L. Cann, P. M. Martineau, J. Samartseva,
(2016), Phys. Rev. B 93, 024305. J. J. P. Freeth, S. J. Sibley, C. B. Hartland, M. E. Newton,
Jarmola, A., V. M. Acosta, K. Jensen, S. Chemerisov, and H. K. Dhillon, and D. J. Twitchen (2013), J. Phys.: Condens.
D. Budker (2012), Phys. Rev. Lett. 108, 197601. Matter 25 (27), 275801.
Jaskula, J.-C., E. Bauch, S. Arroyo-Camejo, M. D. Lukin, S. W. Khan, R. U. A., P. M. Martineau, B. L. Cann, M. E. Newton,
Hell, A. S. Trifonov, and R. L. Walsworth (2017), Optics and D. J. Twitchen (2009), J. Phys.: Condens. Matter 21 (36),
Express 25 (10), 11048. 364214.
Jaskula, J.-C., B. J. Shields, E. Bauch, M. D. Lukin, A. S. Kiflawi, I., A. T. Collins, K. Iakoubovskii, and D. Fisher (2007),
Trifonov, and R. L. Walsworth (2017), ArXiv e-prints J. Phys.: Condens. Matter 19 (4), 046216.
arXiv:1711.02023 [quant-ph]. Kim, E., V. M. Acosta, E. Bauch, D. Budker, and P. R. Hemmer
Jayakumar, H., S. Dhomkar, J. Henshaw, and C. A. Meriles (2012), Appl. Phys. Lett. 101 (8), 082410.
(2018), Appl. Phys. Lett. 113 (12), 122404. Kim, M., H. J. Mamin, M. H. Sherwood, K. Ohno, D. D.
Jensen, K., V. M. Acosta, A. Jarmola, and D. Budker (2013), Awschalom, and D. Rugar (2015), Phys. Rev. Lett. 115,
Phys. Rev. B 87, 014115. 087602.
Jensen, K., P. Kehayias, and D. Budker (2017), in Smart Kirschner, F. K. K., F. Flicker, A. Yacoby, N. Y. Yao, and S. J.
Sensors, Measurement and Instrumentation: High Sensitiv- Blundell (2018), Phys. Rev. B 97, 140402.
ity Magnetometers, Vol. 19, pp. 553–576. Kirui, J., J. van Wyk, and M. Hoch (2013), Diam. Relat. Mater.
Jensen, K., N. Leefer, A. Jarmola, Y. Dumeige, V. M. Acosta, 39, 78.
P. Kehayias, B. Patton, and D. Budker (2014), Phys. Rev. Kitching, J. (2018), Appl. Phys. Rev. 5 (3), 031302.
Lett. 112 (16), 160802. Kittel, C., and E. Abrahams (1953), Phys. Rev. 90, 238.
Jeske, J., J. H. Cole, and A. D. Greentree (2016), New J. Phys. Kleinsasser, E. E., M. M. Stanfield, J. K. Q. Banks, Z. Zhu, W.-
18 (1), 013015. D. Li, V. M. Acosta, H. Watanabe, K. M. Itoh, and K.-M. C.
Jeske, J., D. W. M. Lau, X. Vidal, L. P. McGuinness, P. Reineck, Fu (2016), Appl. Phys. Lett. 108 (20), 202401.
B. C. Johnson, M. W. Doherty, J. C. McCallum, S. Onoda, Knowles, H. S., D. M. Kara, and M. Atatüre (2013), Nat. Mater.
F. Jelezko, T. Ohshima, T. Volz, J. H. Cole, B. C. Gibson, 13, 21.
and A. D. Greentree (2017), Nat. Commun. 8, 14000. Koike, J., D. M. Parkin, and T. E. Mitchell (1992), Appl. Phys.
Ji, P., and M. V. G. Dutt (2016), Phys. Rev. B 94, 024101. Lett. 60 (12), 1450.
Jiang, L., J. S. Hodges, J. R. Maze, P. Maurer, J. M. Taylor, Kolkowitz, S., A. Safira, A. A. High, R. C. Devlin, S. Choi,
D. G. Cory, P. R. Hemmer, R. L. Walsworth, A. Yacoby, A. S. Q. P. Unterreithmeier, D. Patterson, A. S. Zibrov, V. E.
Zibrov, and M. D. Lukin (2009), Science 326 (5950), 267. Manucharyan, H. Park, and M. D. Lukin (2015), Science
Jones, R., J. Goss, H. Pinto, and D. Palmer (2015), Diam. 347 (6226), 1129.
Relat. Mater. 53, 35. Kucsko, G., S. Choi, J. Choi, P. C. Maurer, H. Zhou, R. Landig,
Jones, R., L. S. Hounsome, N. Fujita, S. Öberg, and P. R. H. Sumiya, S. Onoda, J. Isoya, F. Jelezko, E. Demler, N. Y.
Briddon (2007), Phys. Status Solidi A 204 (9), 3059. Yao, and M. D. Lukin (2018), Phys. Rev. Lett. 121, 023601.
Jones, R., J. E. Lowther, and J. Goss (1996), Appl. Phys. Lett. Kucsko, G., P. C. Maurer, N. Y. Yao, M. Kubo, H. J. Noh, P. K.
69, 2489. Lo, H. Park, and M. D. Lukin (2013), Nature 500 (7460),
Kageura, T., K. Kato, H. Yamano, E. Suaebah, M. Kajiya, 54.
S. Kawai, M. Inaba, T. Tanii, M. Haruyama, K. Yamada, Kwan, C. L., and T. F. Yen (1979), Anal. Chem. 51 (8), 1225.
S. Onoda, W. Kada, O. Hanaizumi, T. Teraji, J. Isoya, de Lange, G., T. van der Sar, M. Blok, Z.-H. Wang, V. Dobrovit-
S. Kono, and H. Kawarada (2017), Appl. Phys. Express ski, and R. Hanson (2012), Sci. Rep. 2, 382 .
10 (5), 055503. de Lange, G., Z. H. Wang, D. Ristè, V. V. Dobrovitski, and
Kaiser, W., and W. Bond (1959), Phys. Rev. 115 (4), 857. R. Hanson (2010), Science 330 (6000), 60.
Kalish, R. (1999), Carbon 37 (5), 781. Lawson, S. C., D. Fisher, D. C. Hunt, and M. E. Newton (1998),
Kanda, H. (2000), Braz. J. Phys. 30, 482. J. Phys.: Condens. Matter 10 (27), 6171.
Karaveli, S., O. Gaathon, A. Wolcott, R. Sakakibara, O. A. Layden, D., and P. Cappellaro (2018), npj Quantum Informa-
Shemesh, D. S. Peterka, E. S. Boyden, J. S. Owen, R. Yuste, tion 4 (1), 30.
and D. Englund (2016), Proc. Natl. Acad. Sci. 113 (15), 3938. Layden, D., S. Zhou, P. Cappellaro, and L. Jiang (2019), Phys.
 69

Rev. Lett. 122, 040502. icon Carbide, Diam. Relat. Mater. 17 (7), 1212.
Lazariev, A., S. Arroyo-Camejo, G. Rahane, V. K. Kavata- Mamin, H. J., M. H. Sherwood, M. Kim, C. T. Rettner, K. Ohno,
mane, and G. Balasubramanian (2017), Scientific Reports D. D. Awschalom, and D. Rugar (2014), Phys. Rev. Lett.
7 (1), 6586. 113 (3), 030803.
Le Sage, D., and K. Arai (2011), personal communication. Manson, N. B., and J. P. Harrison (2005), Diam. Relat. Mater.
Le Sage, D., K. Arai, D. R. Glenn, S. J. DeVience, L. M. Pham, 14, 1705.
L. Rahn-Lee, M. D. Lukin, A. Yacoby, A. Komeili, and R. L. Manson, N. B., M. Hedges, M. S. J. Barson, R. Ahlefeldt, M. W.
Walsworth (2013), Nature 496 (7446), 486. Doherty, H. Abe, T. Ohshima, and M. J. Sellars (2018), New
Le Sage, D., L. M. Pham, N. Bar-Gill, C. Belthangady, M. D. J. Phys. 20 (11), 113037.
Lukin, A. Yacoby, and R. L. Walsworth (2012), Phys. Rev. Marblestone, A., B. Zamft, Y. Maguire, M. Shapiro, T. Cy-
B 85, 121202. bulski, J. Glaser, D. Amodei, P. B. Stranges, R. Kalhor,
Lea-wilsonf, M. A., J. N. Lomer, and J. A. V. Wyk (1995), D. Dalrymple, D. Seo, E. Alon, M. Maharbiz, J. Carmena,
Philos. Mag. B 72 (1), 81. J. Rabaey, E. Boyden, G. Church, and K. Kording (2013),
Lesik, M., J.-P. Tetienne, A. Tallaire, J. Achard, V. Mille, Front. in Comput. Neurosci. 7, 137.
A. Gicquel, J.-F. Roch, and V. Jacques (2014), Appl. Phys. Markham, M., J. Dodson, G. Scarsbrook, D. Twitchen, G. Bal-
Lett. 104 (11), 113107. asubramanian, F. Jelezko, and J. Wrachtrup (2011), Diam.
Levitt, M. H. (2008), Spin Dynamics: Basics of Nuclear Mag- Relat. Mater. 20 (2), 134.
netic Resonance, 2nd ed. (John Wiley & Sons) Chap. 11. Marseglia, L., J. P. Hadden, A. C. Stanley-Clarke, J. P. Harrison,
Li, L., E. H. Chen, J. Zheng, S. L. Mouradian, F. Dolde, B. Patton, Y. L. D. Ho, B. Naydenov, F. Jelezko, J. Meijer,
T. Schröder, S. Karaveli, M. L. Markham, D. J. Twitchen, P. R. Dolan, J. M. Smith, J. G. Rarity, and J. L. O’Brien
and D. Englund (2015), Nano Lett. 15 (3), 1493. (2011), Appl. Phys. Lett. 98 (13), 133107.
Liggins, S. (2010), Identification of point defects in treated single Martin, J., R. Wannemacher, J. Teichert, L. Bischoff, and
crystal diamond, Ph.D. thesis (University of Warwick). B. Köhler (1999), Appl. Phys. Lett. 75 (20), 3096.
Lim, T.-S., J.-L. Chern, and K. Otsuka (2002), Opt. Lett. Martineau, P. M., M. P. Gaukroger, K. B. Guy, S. C. Lawson,
27 (12), 1037. D. J. Twitchen, I. Friel, J. O. Hansen, G. C. Summerton,
Liu, Y.-X., A. Ajoy, and P. Cappellaro (2019), Phys. Rev. Lett. T. P. G. Addison, and R. Burns (2009), J. Phys.: Condens.
122, 100501. Matter 21 (36), 364205.
Lobaev, M., A. Gorbachev, S. Bogdanov, A. Vikharev, D. Radi- Maurer, P. C., G. Kucsko, C. Latta, L. Jiang, N. Y. Yao,
shev, V. Isaev, V. Chernov, and M. Drozdov (2017), Diam. S. D. Bennett, F. Pastawski, D. Hunger, N. Chisholm,
Relat. Mater. 72, 1. M. Markham, D. J. Twitchen, J. I. Cirac, and M. D. Lukin
Lomer, J. N., and A. M. A. Wild (1973), Radiat. Eff. 17 (1-2), (2012), Science 336 (6086), 1283.
37. Maurer, P. C., J. R. Maze, P. L. Stanwix, L. Jiang, A. V. Gor-
Londero, E., E. Bourgeois, M. Nesladek, and A. Gali (2018), shkov, A. A. Zibrov, B. Harke, J. S. Hodges, A. S. Zibrov,
Phys. Rev. B 97, 241202. A. Yacoby, D. Twitchen, S. W. Hell, R. L. Walsworth, and
Loretz, M., S. Pezzagna, J. Meijer, and C. L. Degen (2014), M. D. Lukin (2010), Nat. Phys. 6, 912 .
Appl. Phys. Lett. 104 (3), 033102. Maze, J. R., A. Dréau, V. Waselowski, H. Duarte, J.-F. Roch,
Loretz, M., T. Rosskopf, and C. L. Degen (2013), Phys. Rev. and V. Jacques (2012), New J. Phys. 14 (10), 103041.
Lett. 110, 017602. Maze, J. R., P. L. Stanwix, J. S. Hodges, S. Hong, J. M. Taylor,
Loubser, J. H. N., and J. A. van Wyk (1978), Rep. Prog. Phys. P. Cappellaro, L. Jiang, M. V. G. Dutt, E. Togan, A. S. Zi-
41 (8), 1201. brov, A. Yacoby, R. L. Walsworth, and M. D. Lukin (2008),
Lovchinsky, I., J. D. Sanchez-Yamagishi, E. K. Urbach, S. Choi, Nature 455 (7213), 644.
S. Fang, T. I. Andersen, K. Watanabe, T. Taniguchi, A. Bylin- McCloskey, D., D. Fox, N. O’Hara, V. Usov, D. Scanlan,
skii, E. Kaxiras, P. Kim, H. Park, and M. D. Lukin (2017), N. McEvoy, G. S. Duesberg, G. L. W. Cross, H. Z. Zhang,
Science 355 (6324), 503. and J. F. Donegan (2014), Appl. Phys. Lett. 104 (3), 031109.
Lovchinsky, I., A. O. Sushkov, E. Urbach, N. P. de Leon, S. Choi, McLellan, C. A., B. A. Myers, S. Kraemer, K. Ohno, D. D.
K. De Greve, R. Evans, R. Gertner, E. Bersin, C. Müller, Awschalom, and A. C. Bleszynski Jayich (2016), Nano Lett.
L. McGuinness, F. Jelezko, R. L. Walsworth, H. Park, and 16 (4), 2450.
M. D. Lukin (2016), Science 351 (6275), 836. Meiboom, S., and D. Gill (1958), Rev. Sci. Instrum. 29 (8),
Low, G. H., T. J. Yoder, and I. L. Chuang (2014), Phys. Rev. 688.
A 89, 022341. Meirzada, I., Y. Hovav, S. A. Wolf, and N. Bar-Gill (2018),
Lyons, J. L., and C. G. V. de Walle (2016), J. Phys.: Condens. Phys. Rev. B 98, 245411.
Matter 28 (6), 06LT01. Meirzada, I., S. A. Wolf, A. Naiman, U. Levy, and N. Bar-Gill
Ma, Z., S. Zhang, Y. Fu, H. Yuan, Y. Shi, J. Gao, L. Qin, (2019), ArXiv e-prints 1906.05055.
J. Tang, J. Liu, and Y. Li (2018), Opt. Express 26 (1), 382. Michl, J., J. Steiner, A. Denisenko, A. Buelau, A. Zimmermann,
Maclaurin, D., M. W. Doherty, L. C. L. Hollenberg, and A. M. K. Nakamura, H. Sumiya, S. Onoda, P. Neumann, J. Isoya,
Martin (2012), Phys. Rev. Lett. 108, 240403. and J. Wrachtrup (2019), ArXiv e-prints arXiv:1901.01614
MacQuarrie, E. R., T. A. Gosavi, A. M. Moehle, N. R. Jung- [cond-mat.mes-hall].
wirth, S. A. Bhave, and G. D. Fuchs (2015), Optica 2 (3), Michl, J., T. Teraji, S. Zaiser, I. Jakobi, G. Waldherr, F. Dolde,
233. P. Neumann, M. W. Doherty, N. B. Manson, J. Isoya, and
Maertz, B. J., A. P. Wijnheijmer, G. D. Fuchs, M. E. J. Wrachtrup (2014), Appl. Phys. Lett. 104 (10), 102407.
Nowakowski, and D. D. Awschalom (2010), Appl. Phys. Lett. Mita, Y. (1996), Phys. Rev. B 53, 11360.
96 (9), 092504. Mitchell, E. W. J. (1965), “Radiation damage in diamond,”
Mainwood, A. (1999), Phys. Status Solidi A 172 (1), 25. in Physical Properties of Diamond , edited by R. Berman
Malinauskas, T., K. Jarasiunas, E. Ivakin, V. Ralchenko, (Clarendon Press, Oxford).
A. Gontar, and S. Ivakhnenko (2008), Proceedings of Di- Miyazaki, T., Y. Miyamoto, T. Makino, H. Kato, S. Yamasaki,
amond 2007, the 18th European Conference on Diamond, T. Fukui, Y. Doi, N. Tokuda, M. Hatano, and N. Mizuochi
Diamond-Like Materials, Carbon Nanotubes, Nitrides and Sil- (2014), Appl. Phys. Lett. 105 (26), 261601.
 70

Mizuochi, N., P. Neumann, F. Rempp, J. Beck, V. Jacques, D. D. Awschalom (2012), Appl. Phys. Lett. 101 (8), 082413.
P. Siyushev, K. Nakamura, D. J. Twitchen, H. Watanabe, O’Keeffe, M. F., L. Horesh, J. F. Barry, D. A. Braje, and I. L.
S. Yamasaki, F. Jelezko, and J. Wrachtrup (2009), Phys. Chuang (2019), New J. Phys. 21 (2), 023015.
Rev. B 80, 041201. Onoda, S., K. Tatsumi, M. Haruyama, T. Teraji, J. Isoya,
Mkhitaryan, V. V., and V. V. Dobrovitski (2014), Phys. Rev. W. Kada, T. Ohshima, and O. Hanaizumi (2017), Phys. Sta-
B 89, 224402. tus Solidi A 214 (11), 1700160.
Mkhitaryan, V. V., F. Jelezko, and V. V. Dobrovitski (2015), Orwa, J., K. Ganesan, J. Newnham, C. Santori, P. Barclay,
Sci. Rep. 5, 15402. K. Fu, R. Beausoleil, I. Aharonovich, B. Fairchild, P. Olivero,
Momenzadeh, S. A., R. J. Stöhr, F. F. de Oliveira, A. Brun- A. Greentree, and S. Prawer (2012), Diam. Relat. Mater. 24,
ner, A. Denisenko, S. Yang, F. Reinhard, and J. Wrachtrup 6.
(2015), Nano Lett. 15 (1), 165. Osterkamp, C., J. Lang, J. Scharpf, C. Müller, L. P. McGuin-
Morishita, H., E. Abe, W. Akhtar, L. S. Vlasenko, A. Fujimoto, ness, T. Diemant, R. J. Behm, B. Naydenov, and F. Jelezko
K. Sawano, Y. Shiraki, L. Dreher, H. Riemann, N. V. Abrosi- (2015), Appl. Phys. Lett. 106 (11), 113109.
mov, P. Becker, H.-J. Pohl, M. L. W. Thewalt, M. S. Brandt, Palyanov, Y. N., I. N. Kupriyanov, A. F. Khokhryakov, and
and K. M. Itoh (2011), Appl. Phys. Express 4 (2), 021302. V. G. Ralchenko (2015), “Crystal growth of diamond,” in
Morishita, H., S. Kobayashi, M. Fujiwara Hiromitsu Kato Toshi- Handbook of Crystal Growth: Bulk Crystal Growth: Second
haru Makino Satoshi Yamasaki, and N. Mizuochi (2018), Edition, Vol. 2, pp. 671–713.
ArXiv e-prints arXiv:1803.01161 [cond-mat.mes-hall]. Pelliccione, M., B. A. Myers, L. M. A. Pascal, A. Das, and A. C.
Mrózek, M., D. Rudnicki, P. Kehayias, A. Jarmola, D. Budker, Bleszynski Jayich (2014), Phys. Rev. Applied 2, 054014.
and W. Gawlik (2015), EPJ Quantum Technology 2 (1), 22. Pfender, M., N. Aslam, P. Simon, D. Antonov, G. Thiering,
Murai, T., T. Makino, H. Kato, M. Shimizu, T. Murooka, E. D. S. Burk, F. Fávaro de Oliveira, A. Denisenko, H. Fedder,
Herbschleb, Y. Doi, H. Morishita, M. Fujiwara, M. Hatano, J. Meijer, J. A. Garrido, A. Gali, T. Teraji, J. Isoya, M. W.
S. Yamasaki, and N. Mizuochi (2018), Appl. Phys. Lett. Doherty, A. Alkauskas, A. Gallo, A. Grüneis, P. Neumann,
112 (11), 111903. and J. Wrachtrup (2017), Nano Lett. 17 (10), 5931.
Myers, B. A., A. Ariyaratne, and A. C. B. Jayich (2017), Phys. Pham, L. M. (2013), Magnetic Field Sensing with Nitrogen-
Rev. Lett. 118, 197201. Vacancy Color Centers in Diamond, Ph.D. thesis (Harvard
Myers, B. A., A. Das, M. C. Dartiailh, K. Ohno, D. D. University).
Awschalom, and A. C. Bleszynski Jayich (2014), Phys. Rev. Pham, L. M., N. Bar-Gill, C. Belthangady, D. Le Sage, P. Cap-
Lett. 113, 027602. pellaro, M. D. Lukin, A. Yacoby, and R. L. Walsworth
Naydenov, B., F. Reinhard, A. Lämmle, V. Richter, R. Kalish, (2012a), Phys. Rev. B 86, 045214.
U. F. S. D’Haenens-Johansson, M. Newton, F. Jelezko, and Pham, L. M., N. Bar-Gill, D. Le Sage, C. Belthangady,
J. Wrachtrup (2010), Appl. Phys. Lett. 97 (24), 242511. A. Stacey, M. Markham, D. J. Twitchen, M. D. Lukin, and
Neu, E., P. Appel, M. Ganzhorn, J. Miguel-Sánchez, M. Lesik, R. L. Walsworth (2012b), Phys. Rev. B 86 (12), 121202.
V. Mille, V. Jacques, A. Tallaire, J. Achard, and Pham, L. M., S. J. DeVience, F. Casola, I. Lovchinsky, A. O.
P. Maletinsky (2014), Appl. Phys. Lett. 104 (15), 153108. Sushkov, E. Bersin, J. Lee, E. Urbach, P. Cappellaro, H. Park,
Neumann, P., J. Beck, M. Steiner, F. Rempp, H. Fedder, P. R. A. Yacoby, M. Lukin, and R. L. Walsworth (2016), Phys. Rev.
Hemmer, J. Wrachtrup, and F. Jelezko (2010a), Science B 93, 045425.
329 (5991), 542. Pham, L. M., D. L. Sage, P. L. Stanwix, T. K. Yeung, D. Glenn,
Neumann, P., I. Jakobi, F. Dolde, C. Burk, R. Reuter, G. Wald- A. Trifonov, P. Cappellaro, P. R. Hemmer, M. D. Lukin,
herr, J. Honert, T. Wolf, A. Brunner, J. H. Shim, D. Suter, H. Park, A. Yacoby, and R. L. Walsworth (2011), New J.
H. Sumiya, J. Isoya, and J. Wrachtrup (2013), Nano Lett. Phys. 13 (4), 045021.
13 (6), 2738. Pinto, H., R. Jones, D. W. Palmer, J. P. Goss, P. R. Briddon,
Neumann, P., R. Kolesov, V. Jacques, J. Beck, J. Tisler, and S. Öberg (2012), Phys. Status Solidi A 209 (9), 1765.
A. Batalov, L. Rogers, N. B. Manson, G. Balasubramanian, Plakhotnik, T., M. W. Doherty, J. H. Cole, R. Chapman, and
F. Jelezko, and J. Wrachtrup (2009), New J. Phys. 11 (1), N. B. Manson (2014), Nano Lett. 14 (9), 4989.
013017. Poggiali, F., P. Cappellaro, and N. Fabbri (2018), Phys. Rev.
Neumann, P., R. Kolesov, B. Naydenov, J. Beck, F. Rempp, X 8, 021059.
M. Steiner, V. Jacques, G. Balasubramanian, M. L. Markham, Poole, C. P. (1996), Electron Spin Resonance: A Comprehensive
D. J. Twitchen, S. Pezzagna, J. Meijer, J. Twamley, Treatise on Experimental Techniques, edited by 2nd (Courier
F. Jelezko, and J. Wrachtrup (2010b), Nat. Phys. 6 (4), 249. Corporation).
Neves, A., and M. H. Nazaré (2001), Properties, growth and Popa, I., T. Gaebel, M. Domhan, C. Wittmann, F. Jelezko, and
applications of diamond , 26 (Institution of Engineering and J. Wrachtrup (2004), Phys. Rev. B 70 (20), 201203.
Technology). Pu, A., V. Avalos, and S. Dannefaer (2001), Diam. Relat. Mater.
Newell, A. N., D. A. Dowdell, and D. H. Santamore (2016), J. 10 (3), 585.
Appl. Phys. 120 (18), 185104. Pu, A., T. Bretagnon, D. Kerr, and S. Dannefaer (2000), Diam.
Newton, M., B. Campbell, D. Twitchen, J. Baker, and T. An- Relat. Mater. 9 (8), 1450.
thony (2002), Diam. Relat. Mater. 11 (3-6), 618. Rabeau, J. R., P. Reichart, G. Tamanyan, D. N. Jamieson,
Newton, M. E. (2007), EPR, ENDOR and EPR Imaging of De- S. Prawer, F. Jelezko, T. Gaebel, I. Popa, M. Domhan, and
fects in Diamond (Royal Society of Chemistry) Chap. 6. J. Wrachtrup (2006), Appl. Phys. Lett. 88 (2), 023113.
Nöbauer, T., A. Angerer, B. Bartels, M. Trupke, S. Rotter, Rabi, I. I. (1937), Phys. Rev. 51, 652.
J. Schmiedmayer, F. Mintert, and J. Majer (2015), Phys. Raghunandan, M., J. Wrachtrup, and H. Weimer (2018), Phys.
Rev. Lett. 115, 190801. Rev. Lett. 120, 150501.
Nöbauer, T., K. Buczak, A. Angerer, S. Putz, G. Steinhauser, Rajendran, S., N. Zobrist, A. O. Sushkov, R. Walsworth, and
J. Akbarzadeh, H. Peterlik, J. Majer, J. Schmiedmayer, and M. Lukin (2017), Phys. Rev. D 96, 035009.
M. Trupke (2013), ArXiv e-prints arXiv:1309.0453 [quant-ph]. Ramsey, N. F. (1950), Phys. Rev. 78, 695.
Ohno, K., F. Joseph Heremans, L. C. Bassett, B. A. Myers, Riedel, D., D. Rohner, M. Ganzhorn, T. Kaldewey, P. Appel,
D. M. Toyli, A. C. Bleszynski Jayich, C. J. Palmstrøm, and E. Neu, R. J. Warburton, and P. Maletinsky (2014), Phys.
 71

Rev. Applied 2, 064011. Schwartz, J., P. Michaelides, C. D. Weis, and T. Schenkel

Riedel, D., I. Söllner, B. J. Shields, S. Starosielec, P. Appel, (2011), New J. Phys. 13 (3), 035022.
E. Neu, P. Maletinsky, and R. J. Warburton (2017), Phys. Shao, L., M. Zhang, M. Markham, A. M. Edmonds, and M. Lon-
Rev. X 7, 031040. car (2016), Phys. Rev. Applied 6, 064008.
Robertson, R., J. J. Fox, and A. E. Martin (1933), Philos. Trans. Shaw, M. J., P. R. Briddon, J. P. Goss, M. J. Rayson, A. Ker-
R. Soc. London, Ser. A 232 (707-720), 463. ridge, A. H. Harker, and A. M. Stoneham (2005), Phys. Rev.
Robertson, R., J. J. Fox, and A. E. Martin (1936), Proc. R. Lett. 95, 105502.
Soc. Lond. A 157 (892), 579. Shi, F., Q. Zhang, P. Wang, H. Sun, J. Wang, X. Rong, M. Chen,
Robledo, L., H. Bernien, T. v. d. Sar, and R. Hanson (2011), C. Ju, F. Reinhard, H. Chen, J. Wrachtrup, J. Wang, and
New J. Phys. 13 (2), 025013. J. Du (2015), Science 347 (6226), 1135.
Rogers, L. J., S. Armstrong, M. J. Sellars, and N. B. Manson Shields, B. J., Q. P. Unterreithmeier, N. P. de Leon, H. Park,
(2008), New J. Phys. 10 (10), 103024. and M. D. Lukin (2015), Phys. Rev. Lett. 114 (13), 136402.
Romach, Y., A. Lazariev, I. Avrahami, F. Kleißler, S. Arroyo- Shin, C. S., C. E. Avalos, M. C. Butler, D. R. Trease, S. J.
Camejo, and N. Bar-Gill (2019), Phys. Rev. Applied 11, Seltzer, J. Peter Mustonen, D. J. Kennedy, V. M. Acosta,
014064. D. Budker, A. Pines, and V. S. Bajaj (2012), J. Appl. Phys.
Romach, Y., C. Müller, T. Unden, L. J. Rogers, T. Isoda, K. M. 112 (12), 124519.
Itoh, M. Markham, A. Stacey, J. Meijer, S. Pezzagna, B. Nay- Simpson, D. A., J.-P. Tetienne, J. M. McCoey, K. Ganesan,
denov, L. P. McGuinness, N. Bar-Gill, and F. Jelezko (2015), L. T. Hall, S. Petrou, R. E. Scholten, and L. C. L. Hollenberg
Phys. Rev. Lett. 114, 017601. (2016), Sci. Rep. 6, 22797.
Rondin, L., G. Dantelle, A. Slablab, F. Grosshans, F. Treussart, Siyushev, P., M. Nesladek, E. Bourgeois, M. Gulka, J. Hruby,
P. Bergonzo, S. Perruchas, T. Gacoin, M. Chaigneau, H.-C. T. Yamamoto, M. Trupke, T. Teraji, J. Isoya, and F. Jelezko
Chang, V. Jacques, and J.-F. Roch (2010), Phys. Rev. B 82, (2019), Science 363 (6428), 728.
115449. Slepetz, B., and M. Kertesz (2014), Phys. Chem. Chem. Phys.
Rondin, L., J.-P. Tetienne, T. Hingant, J.-F. Roch, 16, 1515.
P. Maletinsky, and V. Jacques (2014), Rep. Prog. Phys. Slichter, C. P. (1990), Principles of Magnetic Resonance
77 (5), 056503. (Springer, Berlin, Heidelberg).
Rong, X., J. Geng, F. Shi, Y. Liu, K. Xu, W. Ma, F. Kong, Smeltzer, B., J. McIntyre, and L. Childress (2009), Phys. Rev.
Z. Jiang, Y. Wu, and J. Du (2015), Nat. Commun. 6, 8748. A 80, 050302.
Rose, A. (1963), Concepts in photoconductivity and allied prob- Smith, W. V., P. P. Sorokin, I. L. Gelles, and G. J. Lasher
lems (Interscience Publishers). (1959), Phys. Rev. 115 (6), 1546.
Rosskopf, T., A. Dussaux, K. Ohashi, M. Loretz, R. Schirhagl, Smits, J., J. T. Damron, P. Kehayias, A. F. McDowell, N. Mosa-
H. Watanabe, S. Shikata, K. M. Itoh, and C. L. Degen (2014), vian, I. Fescenko, N. Ristoff, A. Laraoui, A. Jarmola, and
Phys. Rev. Lett. 112, 147602. V. M. Acosta (2019), Science Advances 5 (7).
Rugar, D., H. J. Mamin, M. H. Sherwood, M. Kim, C. T. Ret- de Sousa, R. (2009), “Electron spin as a spectrometer of nuclear-
tner, K. Ohno, and D. D. Awschalom (2015), Nat. Nanotech- spin noise and other fluctuations,” in Electron Spin Reso-
nol. 10, 120. nance and Related Phenomena in Low-Dimensional Struc-
Rutledge, K. M., and K. K. Gleason (1998), “Characterization tures (Springer Berlin Heidelberg) pp. 183–220.
methods,” in Handbook of Industrial Diamonds and Diamond Stanwix, P. L., L. M. Pham, J. R. Maze, D. Le Sage, T. K.
Films, Chap. 9 (CRC Press). Yeung, P. Cappellaro, P. R. Hemmer, A. Yacoby, M. D. Lukin,
Sakaguchi, I., M. Nishitani-Gamo, K. P. Loh, H. Haneda, and and R. L. Walsworth (2010), Phys. Rev. B 82 (20), 201201.
T. Ando (1999), J. Appl. Phys. 86 (3), 1306. Staudacher, T., F. Shi, S. Pezzagna, J. Meijer, J. Du, C. A.
Santori, C., P. E. Barclay, K.-M. C. Fu, and R. G. Beausoleil Meriles, F. Reinhard, and J. Wrachtrup (2013), Science
(2009), Phys. Rev. B 79, 125313. 339 (6119), 561.
Savitski, V. G. (2017), J. Phys. D: Appl. Phys. 50 (47), 475602. Steiner, M., P. Neumann, J. Beck, F. Jelezko, and J. Wrachtrup
Schirhagl, R., K. Chang, M. Loretz, and C. L. Degen (2014), (2010), Phys. Rev. B 81, 035205.
Annu. Rev. Phys. Chem. 65 (1), 83. Steinert, S., F. Dolde, P. Neumann, A. Aird, B. Naydenov,
Schloss, J. M. (2019), Optimizing Nitrogen-Vacancy Diamond G. Balasubramanian, F. Jelezko, and J. Wrachtrup (2010),
Magnetic Sensors and Imagers for Broadband Sensitivity, Rev. Sci. Instrum. 81 (4), 043705.
Ph.D. thesis (Massachusetts Institute of Technology). Steinert, S., F. Ziem, L. T. Hall, A. Zappe, M. Schweikert,
Schloss, J. M., J. F. Barry, M. J. Turner, and R. L. Walsworth N. Götz, A. Aird, G. Balasubramanian, L. Hollenberg, and
(2018), Phys. Rev. Applied 10, 034044. J. Wrachtrup (2013), Nat. Commun. 4, 1607 .
Schmitt, S., T. Gefen, F. M. Stürner, T. Unden, G. Wolff, Stepanov, V., and S. Takahashi (2016), Phys. Rev. B 94 (2),
C. Müller, J. Scheuer, B. Naydenov, M. Markham, S. Pez- 024421.
zagna, J. Meijer, I. Schwarz, M. Plenio, A. Retzker, L. P. Stürner, F. M., A. Brenneis, J. Kassel, U. Wostradowski,
McGuinness, and F. Jelezko (2017), Science 356 (6340), 832. R. Rölver, T. Fuchs, K. Nakamura, H. Sumiya, S. Onoda,
Schoenfeld, R. S., and W. Harneit (2011), Phys. Rev. Lett. J. Isoya, and F. Jelezko (2019), Diam. Relat. Mater. 93, 59.
106 (3), 030802. Subedi, S. D., V. V. Fedorov, J. Peppers, D. V. Martyshkin,
Schreyvogel, C., V. Polyakov, R. Wunderlich, J. Meijer, and S. B. Mirov, L. Shao, and M. Loncar (2019), Opt. Mater.
C. E. Nebel (2015), Sci. Rep. 5, 12160 . Express 9 (5), 2076.
Schreyvogel, C., M. Wolfer, H. Kato, M. Schreck, and C. E. Sushkov, A. O., N. Chisholm, I. Lovchinsky, M. Kubo, P. K.
Nebel (2014), Sci. Rep. 4, 3634. Lo, S. D. Bennett, D. Hunger, A. Akimov, R. L. Walsworth,
Schröder, T., F. Gädeke, M. J. Banholzer, and O. Benson H. Park, and M. D. Lukin (2014), Nano Lett. 14 (11), 6443.
(2011), New J. Phys. 13 (5), 055017. Tahara, K., H. Ozawa, T. Iwasaki, N. Mizuochi, and M. Hatano
Schwander, M., and K. Partes (2011), Diam. Relat. Mater. (2015), Appl. Phys. Lett. 107 (19), 193110.
20 (9), 1287. Tallaire, A., L. Mayer, O. Brinza, M. A. Pinault-Thaury,
Schwartz, J., S. Aloni, D. F. Ogletree, and T. Schenkel (2012), T. Debuisschert, and J. Achard (2017a), Appl. Phys. Lett.
New J. Phys. 14 (4), 043024. 111 (14), 143101.
 72

Tallaire, A., V. Mille, O. Brinza, T. N. T. Thi, J. Brom, Y. Logu- Phys. 76, 1037.
inov, A. Katrusha, A. Koliadin, and J. Achard (2017b), Diam. Vanier, J., and C. Audoin (1989), The Quantum Physics of
Relat. Mater. 77, 146. Atomic Frequency Standards (A. Hilger).
Tang, C., A. Neves, and A. Fernandes (2004), Diam. Relat. Vineyard, G. H. (1963), J. Math. Phys. 4 (9), 1191.
Mater. 13 (1), 203. Vins, V. (2007), “Technique of production of fancy red dia-
Taylor, J. M., P. Cappellaro, L. Childress, L. Jiang, D. Budker, monds,” U.S. Patent 2007/0053823 A1.
P. R. Hemmer, A. Yacoby, R. Walsworth, and M. D. Lukin Waldermann, F., P. Olivero, J. Nunn, K. Surmacz, Z. Wang,
(2008), Nat. Phys. 4 (10), 810. D. Jaksch, R. Taylor, I. Walmsley, M. Draganski, P. Reichart,
Teraji, T., T. Taniguchi, S. Koizumi, Y. Koide, and J. Isoya A. Greentree, D. Jamieson, and S. Prawer (2007), Diam.
(2013), Appl. Phys. Express 6 (5), 055601. Relat. Mater. 16 (11), 1887.
Teraji, T., T. Yamamoto, K. Watanabe, Y. Koide, J. Isoya, Waldherr, G., J. Beck, M. Steiner, P. Neumann, A. Gali,
S. Onoda, T. Ohshima, L. J. Rogers, F. Jelezko, P. Neumann, T. Frauenheim, F. Jelezko, and J. Wrachtrup (2011), Phys.
J. Wrachtrup, and S. Koizumi (2015), Phys. Status Solidi A Rev. Lett. 106, 157601.
212 (11), 2365. Waldherr, G., Y. Wang, S. Zaiser, M. Jamali, T. Schulte-
Tetienne, J.-P., N. Dontschuk, D. A. Broadway, A. Stacey, D. A. Herbrüggen, H. Abe, T. Ohshima, J. Isoya, J. F. Du, P. Neu-
Simpson, and L. C. L. Hollenberg (2017), Sci. Adv. 3 (4). mann, and J. Wrachtrup (2014), Nature 506, 204.
Tetienne, J.-P., R. W. de Gille, D. A. Broadway, T. Teraji, S. E. Walsworth, R. (2017), “Absorption-based detection of spin im-
Lillie, J. M. McCoey, N. Dontschuk, L. T. Hall, A. Stacey, purities in solid-state spin systems,” U.S. Patent 9,658,301
D. A. Simpson, and L. C. L. Hollenberg (2018), Phys. Rev. B2.
B 97, 085402. Wang, Q., C. P. Smith, M. S. Hall, C. M. Breeding, and T. M.
Tetienne, J.-P., T. Hingant, L. Rondin, A. Cavaillès, L. Mayer, Moses (2005), Gems Gemol. 41 (1).
G. Dantelle, T. Gacoin, J. Wrachtrup, J.-F. Roch, and Wang, Z.-H., G. de Lange, D. Ristè, R. Hanson, and V. V.
V. Jacques (2013), Phys. Rev. B 87, 235436. Dobrovitski (2012), Phys. Rev. B 85, 155204.
Tetienne, J.-P., L. Rondin, P. Spinicelli, M. Chipaux, T. De- Wang, Z.-H., and S. Takahashi (2013), Phys. Rev. B 87, 115122.
buisschert, J.-F. Roch, and V. Jacques (2012), New J. Phys. Wee, T.-L., Y.-K. Tzeng, C.-C. Han, H.-C. Chang, W. Fann, J.-
14 (10), 103033. H. Hsu, K.-M. Chen, and Y.-C. Yu (2007), J. Phys. Chem.
Thiering, G., and A. Gali (2018), Phys. Rev. B 98, 085207. A 111 (38), 9379.
Tokuda, N. (2015), “Homoepitaxial diamond growth by plasma- Weerdt, F. D., and A. Collins (2008), Diam. Relat. Mater.
enhanced chemical vapor deposition,” in Novel Aspects of Di- 17 (2), 171.
amond: From Growth to Applications, edited by N. Yang Wickenbrock, A., H. Zheng, L. Bougas, N. Leefer, S. Afach,
(Springer International Publishing, Cham) pp. 1–29. A. Jarmola, V. M. Acosta, and D. Budker (2016), Appl.
Toyli, D. M., C. F. de las Casas, D. J. Christle, V. V. Dobrovit- Phys. Lett. 109 (5), 053505.
ski, and D. D. Awschalom (2013), Proc. Natl. Acad. Sci. Wieser, M. E., N. Holden, T. B. Coplen, J. K. Böhlke,
110 (21), 8417. M. Berglund, W. A. Brand, P. De Bièvre, M. Gröning,
Toyli, D. M., D. J. Christle, A. Alkauskas, B. B. Buckley, C. G. R. D. Loss, J. Meija, T. Hirata, T. Prohaska, R. Schoenberg,
Van de Walle, and D. D. Awschalom (2012), Phys. Rev. X 2, G. O Connor, T. Walczyk, S. Yoneda, and X.-K. Zhu (2013),
031001. Pure Appl. Chem. 85 (5), 1047.
Trusheim, M. E., and D. Englund (2016), New J. Phys. 18 (12), Witzel, W. M., M. S. Carroll, A. Morello, Ł. Cywiński, and
123023. S. Das Sarma (2010), Phys. Rev. Lett. 105 (18), 187602.
Tucker, O. D., M. E. Newton, and J. M. Baker (1994), Phys. Wolf, S. A., I. Rosenberg, R. Rapaport, and N. Bar-Gill (2015a),
Rev. B 50, 15586. Phys. Rev. B 92, 235410.
Twitchen, D., D. Hunt, M. Newton, J. Baker, T. Anthony, and Wolf, T., P. Neumann, K. Nakamura, H. Sumiya, T. Ohshima,
W. Banholzer (1999a), Physica B: Condens. Matter 273–274, J. Isoya, and J. Wrachtrup (2015b), Phys. Rev. X 5, 041001.
628. Wood, J. D. A., D. A. Broadway, L. T. Hall, A. Stacey, D. A.
Twitchen, D. J., S. L. Geoghegan, and N. Perkins (2010), Simpson, J.-P. Tetienne, and L. C. L. Hollenberg (2016),
“Method for treating single crystal CVD diamond and prod- Phys. Rev. B 94, 155402.
uct obtained,” WO Patent 2010/149775 Al. Wort, C. J., and R. S. Balmer (2008), Mater. Today 11 (1), 22.
Twitchen, D. J., M. E. Newton, J. M. Baker, T. R. Anthony, Wu, Y., F. Jelezko, M. B. Plenio, and T. Weil (2016), Angew.
and W. F. Banholzer (1999b), Phys. Rev. B 59, 12900. Chem. Int. Ed. 55 (23), 6586.
Udvarhelyi, P., V. O. Shkolnikov, A. Gali, G. Burkard, and van Wyk, J. A., E. C. Reynhardt, G. L. High, and I. Kiflawi
A. Pályi (2018), Phys. Rev. B 98, 075201. (1997), J. Phys. D: Appl. Phys. 30 (12), 1790.
Uedono, A., K. Mori, N. Morishita, H. Itoh, S. Tanigawa, S. Fu- Yafet, Y. (1963), in Solid state physics, Vol. 14 (Elsevier) pp.
jii, and S. Shikata (1999), J. Phys.: Condens. Matter 11 (25), 1–98.
4925. Yale, C. G., F. J. Heremans, B. B. Zhou, A. Auer, G. Burkard,
Ulbricht, R., S. Dong, A. Gali, S. Meng, and Z.-H. Loh (2018), and D. D. Awschalom (2016), Nat. Photonics 10, 184.
Phys. Rev. B 97, 220302. Yamamoto, T., C. Müller, L. P. McGuinness, T. Teraji, B. Nay-
Ulbricht, R., and Z.-H. Loh (2018), Phys. Rev. B 98, 094309. denov, S. Onoda, T. Ohshima, J. Wrachtrup, F. Jelezko, and
Unden, T., P. Balasubramanian, D. Louzon, Y. Vinkler, M. B. J. Isoya (2013a), Phys. Rev. B 88 (20), 201201.
Plenio, M. Markham, D. Twitchen, A. Stacey, I. Lovchinsky, Yamamoto, T., T. Umeda, K. Watanabe, S. Onoda, M. L.
A. O. Sushkov, M. D. Lukin, A. Retzker, B. Naydenov, L. P. Markham, D. J. Twitchen, B. Naydenov, L. P. McGuin-
McGuinness, and F. Jelezko (2016), Phys. Rev. Lett. 116, ness, T. Teraji, S. Koizumi, F. Dolde, H. Fedder, J. Honert,
230502. J. Wrachtrup, T. Ohshima, F. Jelezko, and J. Isoya (2013b),
van der Sar, T., F. Casola, R. Walsworth, and A. Yacoby (2015), Phys. Rev. B 88, 075206.
Nat. Commun. 6, 7886. Yamano, H., S. Kawai, K. Kato, T. Kageura, M. Inaba,
Van Oort, E., and M. Glasbeek (1990), Chem. Phys. Lett. T. Okada, I. Higashimata, M. Haruyama, T. Tanii, K. Ya-
168 (6), 529. mada, S. Onoda, W. Kada, O. Hanaizumi, T. Teraji, J. Isoya,
Vandersypen, L. M. K., and I. L. Chuang (2005), Rev. Mod. and H. Kawarada (2017), Jpn. J. Appl. Phys. 56 (4S),
 73

04CK08. 90, 032319.

Yeung, T. K., D. L. Sage, L. M. Pham, P. L. Stanwix, and R. L. Zheng, H., J. Xu, G. Z. Iwata, T. Lenz, J. Michl,
Walsworth (2012), Appl. Phys. Lett. 100 (25), 251111. B. Yavkin, K. Nakamura, H. Sumiya, T. Ohshima, J. Isoya,
Zaitsev, A., K. Moe, and W. Wang (2017), Diam. Relat. Mater. J. Wrachtrup, A. Wickenbrock, and D. Budker (2019), Phys.
71, 38. Rev. Applied 11, 064068.
Zaitsev, A. M. (2001), Optical Properties of Diamond (Springer, Zhou, B. B., P. C. Jerger, V. O. Shkolnikov, F. J. Heremans,
Berlin). G. Burkard, and D. D. Awschalom (2017), Phys. Rev. Lett.
Zhang, X., and E. L. Hu (2016), Appl. Phys. Lett. 109 (8), 119, 140503.
081101. Zu, C., W. B. Wang, L. He, W. G. Zhang, C. Y. Dai, F. Wang,
Zhao, N., S.-W. Ho, and R.-B. Liu (2012), Phys. Rev. B 85, and L. M. Duan (2014), Nature 514, 72.
115303.
Zhao, N., J. Wrachtrup, and R.-B. Liu (2014), Phys. Rev. A