Ultramicroscopy 184 (2018) 199–208

Contents lists available at ScienceDirect

Ultramicroscopy
journal homepage: www.elsevier.com/locate/ultramic

Contact atomic force microscopy using piezoresistive cantilevers in

load force modulation mode
P. Biczysko a, A. Dzierka a, G. Jóźwiak a,∗, M. Rudek a, T. Gotszalk a, P. Janus b, P. Grabiec b,
I.W. Rangelow c
a
Wroclaw University of Technology, Faculty of Microsystem Electronics and Photonics, Wroclaw, Poland
b
Institute of Electron Technology, Warsaw, Poland
c
Ilmenau University of Technology, Institute of Micro and Nanoelectronics, Department of Micro- and Nanoelectronic Systems, Faculty of Electrical
Engineering and Information Technology, Gustav-Kirchhoff-Straße 1, Ilmenau 98693, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Scanning probe microscopy (SPM) encompasses several techniques for imaging of the physical and chem-
Received 17 June 2016 ical material properties at nanoscale. The scanning process is based on the detection of the deflection of
Revised 19 June 2017
the cantilever, which is caused by near field interactions, while the tip runs over the sample’s surface. The
Accepted 18 September 2017
variety of deflection detection methods including optical, piezoresistive, piezoelectric technologies has
Available online 20 September 2017
been developed and applied depending on the measurement mode and measurement environment. There
Keywords: are many advantages (compactness, vacuum compatibility, etc.) of the piezoresistive detection method,
Atomic force microscopy which makes it very attractive for almost all SPM experiments. Due to the technological limitations the
Piezoresistive probe stiffness of the piezoresistive beams is usually higher than the stiffness of the cantilever detected using
Drift free scanning optical methods. This is the basic constraint for the application of the piezoresistive beams in contact
Load force modulation mode (CM) atomic force microscopy (AFM) investigations performed at low load forces (usually less than
20 nN). Drift of the deflection signal, which is related to thermal fluctuations of the measurement setup,
causes that the microscope controller compensates the fluctuations instead of compensating the strength
of tip-surface interactions. Therefore, it is quite difficult to keep near field interaction precisely at the
setpoint level during the whole scanning process. This can lead to either damage of the cantilever’s tip
and material surface or loosing the contact with the investigated sample and making the measurement
unreliable.
For these reasons, load force modulation (LoFM) scanning mode, in which the interaction at the tip
is precisely controlled at every point of the sample surface, is proposed to enable precise AFM surface
investigations using the piezoresistive cantilevers. In this article we describe the developed measurement
algorithm as well as proposed and introduced hardware and software solutions. The results of the exper-
iments confirm strong reduction of the AFM entire setup drift. The results demonstrating contactless tip
lateral movements are presented. It is common knowledge that such a scanning reduces tip wear.
© 2017 Elsevier B.V. All rights reserved.

1. Introduction ments, due to troublesome optics adjustment. Moreover in the ba-

sic OBD setup the detection of the microprobe-surface interaction
Atomic force microcopy (AFM) is a well known technology used is done, whereas the quantitative (in other words metrological)
in the nanoscale investigations of the physical and chemical sur- investigations are possible only when more complex calibration
face properties. In the AFM technology the strength of the inter- procedure is applied. From that point of view, solutions in which
action between microprobe and surface is detected basing on the the AFM cantilevers integrate a deflection sensor are very useful,
observation of the AFM cantilever’s static or resonance deflection. especially when the interactions should be measured metrologi-
The most popular method used in this task is the optical beam de- cally. The cantilevers with integrated deflection sensors are also
flection (OBD) technology [1]. It provides high sensitivity but it is very suitable in applications, where the AFM technology is com-
difficult to be applied in the vacuum and low temperature experi- bined with another experimental techniques like scanning electron
microscopy (SEM), optical microscopy (OM) or focused ion beam
(FIB) solutions. In these applications quartz tuning forks (QTF) and
∗
Corresponding author. piezoresistive cantilevers have been sometimes utilized [2,3].
E-mail address: [email protected] (G. Jóźwiak).

http://dx.doi.org/10.1016/j.ultramic.2017.09.002
0304-3991/© 2017 Elsevier B.V. All rights reserved.
200 P. Biczysko et al. / Ultramicroscopy 184 (2018) 199–208

It has been already presented that the piezoresistive cantilevers be scanned precisely in contact, even if the cantilevers are used
are very versatile microprobe sensors used in different applica- with stiffness of around 100 N/m. The LoFM is based on the solu-
tions [4–12]. The main advantages of the piezoresistive technology tion called Pulse Mode AFM proposed by Rosa-Zeiser et al [12]. In
is the possibility of the sensor calibration independently on the this setup the analog PID controller was driven by a sample and
system, in which the probe is applied [13,10]. Moreover, various hold (SH) circuit responsible for measurements in and out of the
piezoresistive cantilever devices for almost every scanning probe tip-surface contact. More advanced system was proposed by Pablo
microscopy (SPM) mode can be fabricated, which enabled inves- et al. [23]. In the described architecture a digital signal processor
tigations of not only mechanical, but electrical or thermal surface (DSP) controller was applied to analyze the interaction between
properties as well [14]. Finally, the integration with the piezoresis- the tip and the investigated surface. In the mentioned above se-
tive beam makes it possible to increase the scanning speed and tups soft cantilevers with stiffness less than 5 N/m detected only
surface imaging [15]. In this way, it was possible to design and by the OBD were utilized. The main purpose of the described ex-
fabricate massively parallel systems, which may speed up the sur- periments was to investigate the mechanical surface properties and
face characterization to the level required by the industrial pro- record the force-distance curves.
duction processes. The high sensitivity of the piezoresistive probes In our setup we applied special algorithms to detect the tip
was shown in many papers, in which images of atomic structures surface distance by smart analysis of the tip-surface contact in
of various crystals observed in the contact and non contact AFM the attraction and retraction phase. In this way we are able to
modes (CM AFM and NC AFM) were shown. distinguish between the snap-in and adhesion phenomena, whose
Tortonese et al. presented the CM AFM images of subnanometer dynamics differs dramatically. Moreover the applied piezoresistive
images of graphite, boron nitride, molybden disulfide and tanta- cantilevers were calibrated basing on the thermomechanical noise
lum diselenide crystals recorded with cantilevers of stiffness rang- analysis, which made it possible to quantify the load force and
ing from 5 to 100 N/m. No indication as to the load force was given the tip deflection detection sensitivity [10,24]. In the LoFM tech-
and as no crystal edges were observed it can be assumed that the nology the XY movements are done only when the piezoresistive
presented images correspond with the atomic corrugations rather cantilever is withdrawn. It is not an easy task for the piezoresis-
than the true atomic topography [3]. Jumpertz et al. presented CM tive cantilevers, due to thermal drift present in the signal. Digital
AFM images of highly orientated pyrolytic graphite (HOPG) but as analysis of the deflection signal is required in order to identify the
very little information on the tip-sample interaction were given, “out of contact” state. In this way the tip wear is reduced and the
they should be interpreted only as qualitative ones [16]. The true load force is precisely controlled. In the developed scheme we are
atomic resolution on silicon surface was shown for the first time able to detect the load force in the range from 2 nN up to 100 nN
by Giessibl et al. [17], who presented images recorded in the fre- with the resolution of 2.4 nN and the time period in which signals
quency modulation (FM) AFM mode in ultra high vacuum (UHV). are analyzed at every point on the surface varies from 3 ms till
Recently a report on application of the piezoresistive cantilevers 20 ms. The described resolution is limited by the range of observ-
in the investigations of biological samples was published in which able deflections for a cantilever with stiffness equal to 75 N/m. The
high resolution images of biological samples recorded in amplitude increase or decrease of the stiffness scales this range up or down.
modulation (AM) AFM mode were shown [18]. The proposed technology is also suitable for the surface mea-
Basing on our over two decade activities on the piezoresistive surements performed using scanning thermal microscopy (SThM)
cantilever research, we feel obliged to note, that application of and conductive atomic force microscopy (C AFM) modes. In both
the piezoresistive cantilevers in the CM AFM is still a challenging modes the LoFM technology offers the possibility to change and/or
task [19]. In contrast to the resonance AFM technologies in the CM monitor the measurement conditions at the tip “in contact” and
AFM the interactions between the tip and the investigated surface “out of contact” state. The constant load force maintenance is also
can be determined with very high accuracy, which constitutes the a very important factor in this kind of measurements.
biggest advantage of this technology. Moreover, if the load force In order to implement the LoFM algorithms we designed a ded-
acting at the tip is low, the obtained in this way measurement re- icated module, enabling to control the entire tip approach-retract
sults reflect directly the surface properties. The reason, why the process, signal acquisition and analysis and finally the load force
CM AFM investigations using piezoresistive cantilevers are quite definition. The core of the LoFM controller is formed by an ARM
seldom reported is the high stiffness of the piezoresistive can- microprocessor, which ensures the needed software and hardware
tilevers stemming from relatively huge thickness of the beam in flexibility. The designed solution makes it also possible to scan the
which the piezoresistors are integrated. The surface imaging was surface in the classical way when the proportional integral deriva-
usually done at load force bigger than 50 nN, which was limitation tive (PID) controller defines the load force while sample scanning.
of many surface measurements. There a reports on piezorestive
probes, whose stiffness less than 5 N/m [20,21]. However, it should
be noted, that the described in these papers solutions do not inte- 2. Methods
grate tips for topography surface scanning. In work [22] it has been
shown that adding only of a conductive tip reduced deflection res- 2.1. Piezoresistive cantilever
olution to 2,4 nm (with force resolution ranging from 1,9 nN to
36 nN were probe stiffness was not reported). Moreover, in such In our experiments we used cantilevers which integrate a
a setup even the smallest thermal, mechanical or electrical drift of piezoresistive Wheatstone bridge deflection sensor and a 4-wire
the piezoresistive cantilever and/or measurement head may lead to electrically connected metallic tip - Fig. 1. The presented struc-
huge variation of the load force acting at the tip, which in turn can ture is foreseen to be applied in the SThM experiments, in which
cause damage of the tip or sample surface modification. the thermal flux between the resistive tip and the surface is mon-
Here, we present so called load force modulation (LoFM) mode itored [25,26]. The architecture of the cantilever is optimized so
in which the interaction at the tip is precisely controlled at ev- that the structure openings improve thermal and electrical shield-
ery point at the sample surface. This is done by cyclic approach ing between the resistive probe, the tip wiring and the Wheatstone
and retraction of the tip to and from the investigated sample. In bridge deflection sensor. When the cantilever is bent the mechan-
the performed cycle the piezoresistive deflection detector output is ical stress occurs in the whole cantilever volume, including the
analyzed and the feedback setpoint is calculated so that to main- middle structure beam, in which the Wheatstone bridge is inte-
tain precisely the defined load force. In this way the surface may grated. The signal UD of the bridge diagonal voltage is proportional
 P. Biczysko et al. / Ultramicroscopy 184 (2018) 199–208 201

Fig. 2. ARMscope architecture.

Fig. 1. Piezoresistive cantilever applied in LoFM investigations. • System central processing unit (SCPU) formed by ARM7 LPC
2478 microprocessor with USB interface enabling communica-
tion with the host PC,
to the cantilever deflection (see Eq. (1)). • Data acquisition units – formed by four channel 18-bit, 5 MSps
analogue-digital converter (ADC) cards, two channel 16-bit 500
UD (z ) = UGz, (1) kSps ADC cards, and eight channel 14-bit 350 ksps ADCcards.
U is the bias voltage of the Wheatstone bridge, G is constant re-
• Scanning field controller – formed by four 16-bit digital to
lated to the microcantilever structure (see references [5,13] for de- analogue converters (DAC) channels with 12-bit digital poten-
tails). If G is scaled by the cantilever stiffness k, the diagonal volt- tiometers, for zooming and rotating of the scanfield. The con-
age UD and force F acting at the cantilever end can be connected troller contains four 16-bit DAC channels for probe and/or sam-
by Eq. (2) ple biasing.
• Control platform – formed by stepper motor drivers and aux-
G iliary analoque drivers.
UD (F ) = U F , (2)
k
The ARMScope system is supported by the so called ARMScan-
As it can be seen from the above formulas, the response of the
ner software and Smart Embedded Software Kit (SESK) controlling
piezoresistive beam depends only on the structure and measure-
the SCPU.
ment electronics properties and therefore can be calibrated with
The ARMScanner software enables simultaneous acquisition of
high accuracy, which is the main strength of the piezoresistive can-
up to 8 analogue signals, surface scanning with the frequency of
tilever technology.
up to 20 lines/s. In order to speed up the measurement process
Fabrication process of the applied cantilevers is based on the
the XY scanners can be controlled using triangle and sine signals.
modified technology described in details in our previous papers
Moreover, the developed software routines make it possible to con-
[27]. At the beginning, the diffusion paths, piezoresistors, contacts
trol tip-surface spectroscopy routines and perform lithography pro-
and metal paths, typical for the CMOS technology, are fabricated.
cess.
In post-CMOS MEMS sequences, the tip, the shape and the thick-
The ARMScanner software is dedicated to operating systems,
ness of cantilevers are defined. The last stage includes the device
which are based on 3.5 .NET platform. Among its features the
separation using planarization with sacrificial layer and dry plasma
most important are: very convenient and precise positioning sys-
etching. In an optional fabrication step, the tip can be sharpened in
tem with intelligent navigation, built-in online image processing
a FIB postprocessing step.
module for trend removal, multithreading support and dedicated
data transmission protocol. The standardized structure of software
2.2. Scanning and data acquisition controller modules facilitates implementation of new measurement modes.
In Fig. 3 the ARMScanner user interface with control panel, im-
In the described experiments we used a custom designed open age views and intelligent navigation (compass window) is shown.
architecture scanning probe microscope controller (called ARM- In the right image window the effect of real-time trend removal
Scope), whose function is to control scanning process, perform sig- is presented. Such a configuration helps in the navigation on the
nal acquisition and data analysis. The designed, fabricated and ap- investigated surface with low topography and substantial slope.
plied system is formed by the hardware and software modules. The The SESK modern layered structure increases portability to
proposed setup ensures the necessary flexibility enabling easy im- novel hardware solutions. The microprocessor dependent devices
plementation of various SPM measurement modes. are abstracted by a device driver software layer. Each card used
The most important feature of the ARMScope is its modularity in the system has its own software library of hardware dependent
of both hardware and software modules. Specific functions were functions. Implemented functional blocks reduce the cost of func-
spread across easily interchangeable extension cards communicat- tionality extension.
ing with the ARM processor via External Memory Interface (EMIF) The data flow diagram of SESK is shown in Fig. 4. The ARM-
– Fig. 2. Thanks to the system modularity it is possible to adopt Scanner settings are transmitted via USB interface following stan-
the controller setup to the requirements of the planned experi- dardized data transmission scheme. The protocol block of the sys-
ment. Application of the ARM EMIF standardizes communication tem controller is responsible for decoding incoming data frames
with the extension cards. It allows also modifications and install- and verifying theirs integrity (control sum). Depending on the
ment of the system upgrades. The entire scanning and data acqui- command type the particular block of the controller is initialized
sition ARMscope controller consists of: and configured.
202 P. Biczysko et al. / Ultramicroscopy 184 (2018) 199–208

Fig. 3. ARMScanner software user interface.

The scanning process requires simultaneous signal generation providing the needed programmable signal attenuation. The out-
and sampling. These functions are divided in two independent put DAC21 and DAC22 signals are set by 12bit two channel DAC2
software blocks. The time control block uses high resolution hard- (AD5722R IC), which operates at 100 kSps speed. The DAC21 and
ware timers to provide necessary synchronization between other DAC22 offset signals can be adjusted in the ±10 V range. These sig-
software modules. Blocks behavior is set with respect to user de- nals are set at the beginning of each measurement in order to fit to
fined data that determines positioning signal shape, timing as well the full dynamic range of the DAC and ADC converters and achieve
as a sampling type. When properties of all blocks are set, execu- their full scale resolution.
tion module can be triggered which initiates a selected procedure The flexible design of the LoFM controller allows to implement
(scanning, surface approaching, spectroscopic measurement). not only the standard PID scheme, but also the innovative LoFM
algorithm. The main purpose of the LoFM mode implies that the
cantilever tip remains in contact with the sample only at discrete
2.3. LoFM controller and defined scanning points. Movement of the tip and the can-
tilever between these points is performed in safe distance above
In order to enable the LoFM operation the dedicated control the sample (in other words in the “out of contact” state). Leaving
unit, called LoFM controller, was designed and integrated with the the contact at each point gives the unique possibility to gain the
ARMScope controller. The LoFM controller is built around digital reference zero point, for which the cantilever does not respond
signal processing system with the ARM microcontroller, which en- to the force interaction. From the theoretical point of view simi-
sured the necessary controller flexibility - Fig. 5. lar effect can be also obtained in the double scan measurements
The level of the output signal from the piezoresistive deflection [28] but in the LoFM mode the XY misalignment errors are com-
sensor is amplified by a programmed gain amplifier (PGA) (AD8250 pletely avoided. Based on the reference zero point the setpoint
integrated circuit (IC)) and offset additionally by the first chan- value of the control loop can be successively updated in such a
nel of the digital analogue converter no.2 (DAC2 )(DAC21 signal, see way that the load force can be maintained precisely at the re-
Fig. 5). In this way the input signal from the piezoresistive deflec- quested level.
tion detector is adjusted to the full scale of the main LoFM con- Leaving contact between measurement points also allows to
troller 20 bits analogue to digital converter (ADC) (LTC2378-20 IC). eliminate influence of constant lateral forces responsible for the
Since the bandwidth of the input signal is limited to 10 kHz and sample wear and tip fracturing [11]. Additionally, measurements
to avoid signal aliasing, the LoFM ADC operates at 20 kHz speed. done at each “in contact” and “out of contact” phases give a possi-
The acquired samples are processed by the embedded software run bility to compensate the disturbance connected with temperature
by the LoFM ARM processor and converted by the 16 bits output change of the microscope chamber, cantilever heat capacity, elec-
DAC1 (AD5542 IC, see Fig. 5), which operates at 20 kSps speed. tronics offset and drift etc. This is the main characteristics of the
The output of DAC1 is offset by the second channel of the DAC2 LoFM technology improving the precision of the contact AFM mea-
(DAC22 signal, see Fig. 5) and connected to the 10bit potentiometer
 P. Biczysko et al. / Ultramicroscopy 184 (2018) 199–208 203

Fig. 4. Smart Embedded Software Kit data flow diagram.

of contact” state. Next the probe is moved into contact by coarse Z

axis controller based on DC motors. Approaching is secured by the
PID controller as in standard CM AFM.
The LoFM scan process consists of five independent steps -
Fig. 6. In step no. 1 the tip is in the contact with the investigated
sample and the surface is measured like in standard CM-AFM. In
step no. 2, the cantilever is being continuously withdrawn from
the contact with the investigated surface at the defined speed. At
the beginning of this process, due to the adhesive forces, the tip
remains contacted with the investigated surface, but after a while,
Fig. 5. LoFM controller architecture.
when it is released the cantilever starts to oscillate. At the defined
distance above the investigated surface the withdrawing process
is stopped and the LoFM controller waits for stabilization of the
surements, which is related to the differential principle presented deflection signal –step no. 3. The distance is defined by the sys-
in the so called null-point method [28]. tem operator. In step no. 4 the interaction at the tip and output of
In our solution, in contrast to other scanning methods, where the piezoresistive deflection detector in the “out of contact” state
the load force is pulsed [11,12,23], it is much more difficult to as- are analyzed. When the analysis in the “out of contact” state is
sure that the XY scanning is frozen, when the surface is inves- finished, system moves to the next point, the approach procedure
tigated. Unambiguity of the zero force detection hinders precise starts and the tip is being approached to the sample.
synchronization of the LoFM process with the XY surface scanning. In order to perform the LoFM surface measurement in a reli-
The precise synchronization is achieved thanks to the novel proce- able and precise way identification of the “in contact” and “out of
dure for zero signal calibration in the “out of contact” state. In this contact” cantilever state must be done with the highest possible
procedure the AFM system computes the mean value of derivative precision.
of the deflection signal in the “out of contact” state. Basing on this This results from the fact that the tip-surface interactions are
value it is possible to determine the upper and lower limit of the completely different in the approach and retraction phase. From
derivative of the deflection signal. The limits are used in the proce- the physical point of view this phenomenon is related to quite
dure that determines whether the AFM probe is in the stable “out weak force interaction when the tip approaches the surface and
204 P. Biczysko et al. / Ultramicroscopy 184 (2018) 199–208

Fig. 6. LoFM control algorithm diagram: (1) measurement of topography and/or other signals, (2) leaving the contact state-tip retraction, (3) waiting for stable out of contact
state, (4) measurements of deflection signal and/or other signals, moving to the next point and correction of the set point, (5) going into contact at the next point-tip
approach.

set point for the PID algorithm is calculated on the basis of the
zero point value and the required load force. Then, the measure-
ments in the “in contact” state can be peroformed (e.g. topography,
current, temperature, spreading resistance etc.).
The main difference between the LoFM and currently known
modes (e.g. Pulse Mode, Peak Force etc.) is its resistance to the
drift present when a piezoresistive cantilever is used. If the devel-
oped procedure is not applied it may happen that at the low load
force the tip is not withdrawn to the “out of contact” state, when
the XY controller moves the sample. Thanks to the proposed solu-
tion it is certain that the XY sample movements are done in the
“out of contact” state.
The flow of the measurement is governed by the SCPU and the
ARMScope software.

3. Results and discussion

The mechanical structure of the piezoresistive cantilever is

Fig. 7. Oscillogram of Z-axis controller output, deflection signal from piezoresistive
bridge, which is used for “in contact” state detection, numerical derivative of de-
quite complex, which makes the reliable calculation of the can-
flection signal used for “out of contact” state detection. All signals were recorded in tilever stiffness almost impossible. Therefore, the spring constant
the LoFM mode at a single point during one full measurement cycle (from the 1st and sensitivity of piezoresistive detector are calibrated by parallel
step up to the 5th step). optical and electrical measurements, as it was introduced in [10].
The SP-S 120 SIOS MessTechnik GmbH interferometer was applied
for optical measurement of the cantilever thermal vibrations. Si-
relatively strong adhesion forces when the tip is retracted. The de- multaneously the electrical signal from the piezoresistive bridge
scribed problem becomes even more difficult as not only the value was recorded. The Welch method [29] based on overlapped pe-
of the “in contact” and “out of contact” interactions is different but riodograms averaging was used to estimate the power spectra of
theirs dynamics as well. Therefore, in the developed LoFM technol- both signals - Fig. 8. The mathematical model of this spectrum is
ogy the mentioned calibration procedure has been implemented, defined by the Nyquist fluctuation – dissipation theorem [30] and
which allows to perform the LoFM procedure irrespectively of the frequency response of the simple harmonic oscillator (SHO) [10].
sample surface type and dynamics of the force interaction. The spectra were fitted by the following equation
The stable “out of contact” state is detected, if the predefined  2  2 −1
2
number of consecutive samples of derivative of the deflection sig- S( f ) = A f0 − f 2 + f f0 Q −1 + B, (3)
nal is observed in the range defined by calibration procedure (see
Fig. 7). where S(f) is the estimated power spectrum, A and B are fitting
Then, the controller calculates mean value of the deflection sig- parameters, f0 and Q are fitted eigenfrequency and the quality fac-
nal and uses it as the new reference zero point. The contact state is tor. The constant B is introduced to remove the impact of white
detected by similar procedure but the samples of deflection signal noise, which is also recorded in the analyzed frequency range. The
itself are counted. The center of the range used in the “in contact” cantilever stiffness was calculated on the basis of the values of f0 ,
state detection is positioned on the basis of the zero point value Q, and AO estimated on the basis of the optical signal (see refer-
and the required load force (see Fig. 7). When the “in contact” state ence [10] for details). The sensitivity of the Wheatstone bridge in-
is detected, the standard PID control procedure is turned on. The tegrated with the cantilever structure was calculated on the basis
 P. Biczysko et al. / Ultramicroscopy 184 (2018) 199–208 205

Table 1
Low frequency noise level.

Piezoresistive bridge bias Deflection noise level Force noise level Bandwidth
[V] [pm] [nN] [kHz]

±0.25 V 22 1.650 0.1

±0.25 V 59 4.425 1
±0.25 V 179 13.425 10
±0.5 V 14 1.050 0.1
±0.5 V 32 2.400 1
±0.5 V 92 6.900 10
±1.0 V 13 0.975 0.1
±1.0 V 22 1.650 1
±1.0 V 50 3.750 10
±1.5 V 14 1.050 0.1
±1.5 V 20 1.500 1
±1.5 V 38 2.850 10
±2.0 V 14 1.050 0.1
±2.0 V 20 1.500 1
±2.0 V 32 2.400 10

Fig. 9. Power spectral density of low frequency noise for different biases of piezore-
Fig. 8. Power spectra estimated for thermomechanical noise measured optically sistive bridge.
and piezoresistive signal measured electrically.

The ARMScope system with the implemented LoFM control

of Eq. (4): mode was applied in the surface investigations. To give a proof
 of concept, we present the system response with and without the
AE
c= , (4) LoFM stabilization control. In Figs. 10 and 11 we present the scan-
AO
ning lines, which were recorded while scanning the NGR-11010-
where AE and AO are the parameters fitted to the power spectrum F120 Veeco calibration grating (10 μm pitch, 200 nm deep). The red
of electrical and optical signal respectively. solid line presents the recorded single scan line deflection signal in
For the applied cantilever the effective spring constant is “out of contact” state, while the dashed one presents cantilever de-
75 N/m and sensitivity 14.3 μV/nm for the bridge bias voltage of flection signal acquired “in contact” with the sample. During these
±2 V. experiments the applied AFM system was thermally unshielded
The noise properties of the piezoresistive sensor were deter- and exposed to normal lab air currents in order to induce the elec-
mined by the noise analysis of the piezoresistive deflection detec- trical and thermal drifts.
tor. The measurements were performed for the bridge bias varying In the “out of contact” state the cantilever does not respond to
from ±0.25 V to ±2 V. The power spectral densities of noise reg- the near field interactions, so the observed change in the deflection
istered with the selected bridge bias (see Fig. 9) were estimated signal corresponds to unwanted drift of the measurement system.
by Welch method [29]. The root mean square values of the low The difference between dashed and solid red signal can be inter-
frequency noise level were calculated by numerical integration of preted as the force that the tip imposes on the surface. If the dif-
the power spectra in the bandwidth from 0.1 Hz up to 10 kHz. The ference is maintained constant during the whole scanning process,
results are summarized in Table 1. Increase of the bridge bias im- we assume that the load force is precisely controlled.
proves noise performance, which is visible especially in the band- In Fig. 10 during the standard CM AFM imaging with XY scan-
width of 10 kHz. In the bandwidth of 100 Hz 1/f noise dominates ner switched off the deflection signal measured above the sample
the spectrum and bias increase does not affect noise level strongly. rises, while the contact interactions (dashed red line) are kept at
The LoFM experiments were carried out for ±2 V in 10 kHz band- the same level (set point) by the control loop. The difference be-
width, so the deflection noise level was about 32pm and force tween red lines (black line) corresponds to the varying load force
noise level about 2.4 nN. These results provides the information that translates to the additional measurement error, tip wear and
about resolution limits of the AFM system based on the cantilever or sample modification.
structure described in Section 2.1. These limits are related only to In Fig. 11 we present how the AFM system operates when the
the piezoresistive cantilever and its electronics and they do not LoFM mode was active. The system was able to compensate the
take into account mechanical or acoustic noise sources. variation of the load force so that signal difference was minimized.
206 P. Biczysko et al. / Ultramicroscopy 184 (2018) 199–208

Fig. 10. Deflection signal in “out of contact” (a) and “in contact” (b) states and theirs comparison (c) during standard CM AFM imaging. (For interpretation of the references
to color in this figure legend, the reader is referred to the web version of this article).

Fig. 11. Deflection signal “in out of contact” (a) and “in contact” (b) states and theirs comparison (c) during LoFM AFM imaging. (For interpretation of the references to color
in this figure legend, the reader is referred to the web version of this article).

With the use of the LoFM mode it was possible to neglect the drift ground (root mean square value of error signal) calculated basing
of the deflection signal and provide precise force measurements. In on the recorded images of noise was 0.15 nm and 0.32 nm for the
this case the fluctuations of the load force were reduced from 35% CM AFM and LoFM AFM techniques respectively. It should be noted
to 1,5%. that the presented noise levels were registered by the experimen-
In order to describe the noise performance quantitatively we tal system, whose mechanical architecture was not optimized and
analyzed the noise background in both imaging modes, when the set only for test purposes. The noise level in the LoFM mode is
XY scanner was switched off and only the Z actuator was applied slightly bigger than in the standard CM AFM technology, which is
to maintain the load force at the tip. In this case the noise back- the consequence of the removal of the drift impact. In this case
 P. Biczysko et al. / Ultramicroscopy 184 (2018) 199–208 207

Fig. 12. Images of HOPG surface measured in CM AFM using stiff piezoresistive cantilevers (a,c) and LoFM mode (b,d).

scanned with the standard speed of 2 lines/s, time duration needed

to record 512 × 512 line image was about 5 min, in which the sys-
tem was sensitive to parasitic drifts. The same surface was mea-
sured in the LoFM mode with the load force of 10 nN.
The 2 nm high steps are seen in images recorded in both modes.
The contrast of the crystal edges in the LoFM image is bigger than
in the image measured in the CM AFM technology. In our opinion
this results from the fact that in the LoFM mode the tip does not
twist during scanning over the HOPG sample.
In Fig. 13 the deflection signal from the piezoresistive detector
(red line) is shown. The green circles indicate time, when the tip
is in the “out of contact” state. The signals controlling XY scanner
(black and cyan line) reveals, that the sample moves only when the
cantilever is in the stable “out of contact” state. The stable “out of
contact” state is confirmed also by the plot of the signal controlling
the Z-axis movement (blue line).
Fig. 13. LoFM control scheme signals: LoFM controller output signals driving XYZ
scanner, LoFM deflection signal measured at the piezoresistive bridge built-in the
cantilever structure. XY signals did not change during the move from one measure- 4. Conclusion
ment point to another one, so the cantilever’s tip was not worn. (For interpretation
of the references to color in this figure legend, the reader is referred to the web
version of this article).
In this paper we presented the LoFM method for contact AFM
experiments in which stiff piezoresistive cantilevers are applied.
The proposed method allowed to define the load force acting at
the error of the estimation of the setpoint at each of the LoFM the tip with very high precision by cyclic approach and retract pro-
cycle adds up to the output of the LoFM controller. The further cess. The algorithm was implemented in the universal SPM plat-
improvements are still possible. It seems that the application of form called - ARMScope and used in series of investigations that
Kalman filtering [31] could improve the variance of estimation of demonstrated the opportunities given by the proposed technique.
the zero reference force. Nevertheless, the obtained results confirm Application of precisely calibrated piezoresistive cantilever made it
that the proposed LoFM technology is attractive especially when possible to determine the load force while surface scanning. It was
precise control of the load force is needed. shown that it was possible to compensate online the drift of the
The resolution possibilities of the contact AFM using stiff AFM setup. The resolution limited by the cantilever its electronics
piezoresistive cantilevers done in the LoFM mode was qualitatively was about 2.4 nN in the bandwidth of 10 kHz. The implemented
tested by imaging of the high orientated pyrolytic graphite (HOPG) control algorithms were described and results of measurements of
sample. Freshly cleaved HOPG crystal surface is usually used in the calibration sample as well as HOPG crystal were presented. The
experiments in which the sensitivity and resolution of the AFM quality of images of HOPG was comparable to the quality of images
system is verified. When the crystal terraces are seen with high measured in the standard CM AFM but the force acting on the sur-
contrast it can be assumed that the system resolution is satisfy- face did not change during the scanning. The registered control and
ing. Fig. 12 shows the HOPG topography recorded in standard CM deflection signals confirmed that the AFM scanner moved laterally
AFM technology with the load force of 30 nN. The surface was only when the probe was in the “out of contact”. The system is
208 P. Biczysko et al. / Ultramicroscopy 184 (2018) 199–208

also capable of performing experiments requiring measurements, [13] T. Gotszalk, P. Grabiec, I.W. Rangelow, Calibration and examination of piezore-
when the tip was “in contact” and “out of contact” states with the sistive Wheatstone bridge cantilevers for scanning probe microscopy, Ultrami-
croscopy 97 (1-4) (2003) 385–389.
investigated surface, which opens new measurement opportunities [14] G. Wielgoszewski, P. Sulecki, T. Gotszalk, P. Janus, P. Grabiec, M. Hecker, Y. Ritz,
in the SPM technologies enabling electrical and thermal investiga- E. Zschech, Scanning thermal microscopy: a nanoprobe technique for studying
tion of surface properties. For these reasons, the LoFM scanning is the thermal properties of nanocomponents, Phys. Status Solidi B 248 (2011)
370–374.
the advanced technology, which should be used for the investiga- [15] M. Kaestner, T. Ivanov, A. Schuh, A. Ahmad, T. Angelov, Y. Krivoshapkina,
tion of delicate samples, when the probe pressure must be pre- M. Budden, M. Hofer, S. Lenk, J-P. Zoellner, I.W. Rangelow, A. Reum, E. Guliyev,
cisely controlled. M. Holz, N. Nikolov, Scanning probes in nanostructure fabrication, J. Vac. Sci.
Technol. B 32 (2014) 06F101-1-8.
[16] R. Jumpertz, A. v. d. Hart, O. Ohlsson, F. Saurenbach, J. Schelten, Piezoresistive
Acknowledgments sensors on AFM cantilevers with atomic resolution, Microelectron. Eng. 41–42
(1998) 441–444.
[17] F.J. Giessibl, B.M. Trafas, Piezoresistive cantilevers utilized for scanning tunnel-
This research has been supported by the European Union (EU)
ing and scanning force microscope in ultrahigh vacuum, Rev. Sci. Instrum. 65
FP7 grant no. 3186253, NANOHEAT - MultidomaiN plAtform for (1994) 1923–1929.
iNtegrated MOre-tHan-MoorE/Beyond CMOS systems charActerisa- [18] M. Dukic, J.D. Adams, G. Fantner, Piezoresistive AFM cantilevers surpasssing
tion & diagnosTics. standard optical beam deflection in low noise topography imaging, Sci. Rep. 5
(16393) (2015) 1–11.
[19] I.W. Rangelow, F. Shi, P. Hudek, T. Gotszalk, P.B. Grabiec, P. Dumania, Fabrica-
References tion of piezoresistive sensed AFM cantilever probe with integreted tip, Proc.
SPIE 2879 (1996) 56–64.
[1] C.A.J. Putman, B.G. de Grooth, N. van Hulst, G.J. Greve, A theoretical com- [20] J.C. Doll, S.J. Park, B.L. Pruitt, Design optimization of piezoresistive cantilevers
parison between interferometric and optical beam deflection technique for for force sensing in air and water, J. Appl. Phys. 106 (2009) 064310.
the measurement of cantilever displacement in AFM, Ultramicroscopy 42–44 [21] J.C. Doll, E.A. Corbin, W.P. King, Beth, L. Pruitt, Self-heating in piezoresistive
(1992) 1509–1513. cantilevers, Appl. Phys. Lett. 98 (2011) 223103.
[2] F.J. Giessibl, F. Pielmeier, T. Eguchi, Y. Hasegawa, An Comparison of force sen- [22] N. Harjee, A.G.F. Garcia, M. König, J.C. Doll, D. Goldhaber-Gordon, B.L. Pruitt,
sors for atomic force microscopy based on quartz tuning forks and length-ex- Coaxial tip piezoresistive scanning probes for high-resolution electrical imag-
tensional resonators, Phys. Rev. B 4 (2011) 125409–125424. ing, in: IEEE Proc. of Int. Conf. on MEMS, 2010, pp. 344–347.
[3] M. Tortonese, R.C. Barrett, C.F. Quate, Atomic resolution with an atomic force [23] P.J. de Pablo, J. Colchero, J. Gómez-Herrero, A.M. Baró, Jumping mode scanning
microscope using piezoresistive detection, Appl. Phys. Lett. 62 (1993) 834–836. force microscopy, App. Phys. Lett. 73 (1998) 3300–3302.
[4] I.W. Rangelow, T. Gotszalk, P. Grabiec, K. Edinger, N. Abedinov, Thermal [24] M. Woszczyna, P. Zawierucha, M. Światkowski ˛ , T. Gotszalk, P. Grabiec,
nanoprobe, Microectron. Eng. 57-58 (2001) 737–748. N. Nikolov, J. Mielczarski, E. Mielczarska, N. Glezose, T. Ivanow, K. Ivanowa,
[5] I.W. Rangelow, P. Grabiec, T. Gotszalk, K. Edinger, Piezoresistive SXM sensors, Y. Sarov, I.W. Rangelow, Quantitative force and mass measurements using the
Surf. Interface Anal. 33 (2) (2002) 59–64. cantilever with integrated actuator and deflection detector, Microelectron. Eng.
[6] R. Pedrak, T. Ivanov, K. Ivanova, T. Gotszalk, N. Abedinov, I.W. Rangelow, 86 (2009) 1043–1045.
K. Edinger, E. Tomerov, T. Schenkel, P. Hudek, Micromachined atomic force mi- [25] G. Wielgoszewski, P. Sulecki, P. Janus, P. Grabiec, E. Zschech, T. Gotszalk, A
croscopy sensor with integrated piezoresistive sensor and thermal bimorph ac- high-resolution measurement system for novel scanning thermal microscopy
tuator for high-speed tapping-mode atomic force microscopy phase-imaging in resistive nanoprobes, Meas. Sci. Technol. 22 (2011) 094023-1-7.
higher eigenmodes, J. Vac. Sci. Technol. B 21 (6) (2003) 3102–3107. [26] G. Jóźwiak, G. Wielgoszewski, T. Gotszalk, L. Kepi ˛ ński, Thermal mapping of a
[7] M. Joodaki, P. Janus, T. Gotszalk, G. Kompa, K. Edinger, I.W. Rangelow, Applica- scanning thermal microscopy tip, Ultramicroscopy 133 (2013) 80–87.
tion of a scanning thermal nano-probe for thermal imaging of high frequency [27] P. Janus, P. Grabiec, A. Sierakowski, T. Gotszalk, M. Rudek, D. Kopiec, W. Ma-
active devices, Jpn. J. Appl. Phys. 44 (9A) (2005) 6823–6825. jstrzyk, G. Boetsch, B. Koehler, Design, technology, and application of in-
[8] A. Gaitas, S. Gianchandani, W. Zhu, A piezo-thermal probe for thermomechan- tegrated piezoresistive scanning thermal microscopy (SThM) microcantilever,
ical analysis, Rev. Sci. Instrum. 82 (5) (2011) 053701. Proc. SPIE 9236 (2014).
[9] A. Gaitas, T. Li, W. Zhu, A probe with ultrathin film deflection sensor for scan- [28] K. Kim, J. Chung, J. Won, O. Kwon, J.S. Lee, S.H. Park, Y.K. Choi, Quantitative
ning probe microscopy and material characterization, Sensor. Actuat. A-Phys. scanning thermal microscopy with double scan technique, Appl. Phys. Lett. 93
168 (2) (2011) 229–232. (2008) 203115.
[10] G. Jóźwiak, D. Kopiec, P. Zawierucha, T. Gotszalk, P. Janus, P. Grabiec, [29] P.W. Welch, The use of fast Fourier transform for the estimation of power spec-
I.W. Rangelow, The spring constant calibration of the piezoresistive cantilever tra: a method based on time averaging over short, modified periodograms,
based biosensor, Sensor. Actuat. B-Chem. 170 (2012) 201–206. IEEE Trans. Audio Electroacoustics 15 (2) (1967) 70–76.
[11] B. Pittenger, N. Erina, C. Su, Quantitative Mechanical Property Mapping at the [30] H. Nyquist, Thermal agitation of electric charge in conductors, Phys. Rev. 32
Nanoscale with PeakForce QNM, Application Notes Veeco Instruments Inc., (1928) 110–113.
2010. [31] R.E. Kalman, A new approach to linear filtering and prediction problems, Trans.
[12] A. Rosa-Zeiser, E. Weilandt, S. Hild, O. Marti, The simultaneous measurement ASME–J. Basic Eng. 82 (Series D) (1960) 35–45.
of elastic, electrostatic and adhesive properties by scanning force microscopy:
pulsed-force mode operation, Meas. Sci. Technol. 8 (1997) 1333–1338.