Nuclear Physics A 744 (2004) 192–226

Improved convergence in the three-nucleon system

at very low energies
Harald W. Grießhammer a,b
a Institut für Theoretische Physik (T39), Physik-Department, Technische Universität München,
D-85747 Garching, Germany
b ECT*, Villa Tambosi, I-38050 Villazzano (Trento), Italy

Received 3 May 2004; received in revised form 12 July 2004; accepted 12 August 2004
Available online 9 September 2004

Abstract
Neutron–deuteron scattering in the context of “pion-less” effective field theory at very low
energies is investigated to next-to-next-to-leading order. Convergence is improved by fitting the
two-nucleon contact interactions to the tail of the deuteron wave-function, a procedure known as
Z-parameterisation and extended here to the three-nucleon system. The improvement is particularly
striking in the doublet–S-wave (triton) channel, where better agreement to potential-model calcula-
tions and better convergence from order to order in the power counting is achieved for momenta as
high as ∼ 120 MeV. Investigating the cut-off dependence of the phase-shifts, one confirms numeri-
cally the analytical finding that the first momentum-dependent three-body force enters at N2 LO. The
other partial waves converge also substantially faster. Effective-range parameters of the nd-system
are determined, e.g., for the quartet–S-wave scattering length aq = [6.35 ± 0.02] fm, which compares
favourably both in magnitude and uncertainty with recent high-precision potential-model determina-
tions. Differential cross-sections up to Elab ≈ 15 MeV agree with data.
 2004 Elsevier B.V. All rights reserved.

PACS: 11.80.Jy; 13.75.Cs; 14.20.Dh; 21.30.-x; 25.40.Dn; 27.10.+h

Keywords: Effective field theory; Three-body system; Three-body force; Faddeev equation; Partial waves

E-mail address: [email protected] (H.W. Grießhammer).

0375-9474/$ – see front matter  2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.nuclphysa.2004.08.012
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 193

1. Introduction

The few-nucleon system at very low energies is an important tool to understand key
physical questions. On the one hand, a new generation of high-precision experiments at
the lower end of the energy spectrum use both polarised targets and beams, and em-
ploy neutrino detectors or radioactive-beam facilities to extract, e.g., neutron properties,
neutrino masses and reactions relevant for nuclear astro-physics. On the other hand, a
plethora of pivotal physical processes is hard to access directly in experiments, like re-
action rates in big-bang nucleo-synthesis. In both cases, it is mandatory that binding
effects in light nuclei are taken into account with as few bias as possible towards a par-
ticular model of the few-nucleon system. Theory must provide such model-independent
extraction and calibration methods or predictions at very low energies, often well below
20 MeV.
While precision—the numerical stability of computations—can be controlled by com-
bining sophisticated algorithms with Moore’s law [1] and is hence (albeit sometimes
formidable) not a fundamental problem, the necessary accuracy—namely an estimate
of the systematic uncertainties of a theoretical ansatz—is harder to win and mandates
understanding the physical system at hand. In the three-nucleon sector, this calls in par-
ticular for a systematic understanding of the rôle three-nucleon forces play. Traditionally,
these were often introduced a posteriori to cure discrepancies between data and calcu-
lations, e.g., for the triton binding energy, but such a path is of course untenable when
data are scarce or absent. There is also a host of deviations between experiment and the-
ory for which a suitable three-body force could not yet be constructed, like the famed
Ay -problem [2].
The so-called “pion-less” version of Effective Field Theory in nuclear physics (EFT(/π ))
aspires to provide just such a systematic classification of all forces. At its heart lies the
tenet that physics at those very low energies can be described by point-like interactions
between nucleons only: one cannot identify pions as the lightest exchange-particles be-
tween nucleons as long as the typical external momentum ptyp in a reaction is below the
pion mass mπ because the Compton wave-lengths are not small enough to resolve the
nuclear forces as originating in part from one-pion exchange. With all particles but the
nucleons thus “integrated out” as heavy, one can identify a small, dimension-less parame-
p
ter Q = Λtypπ/
 1, where Λπ/ ∼ mπ is the typical momentum scale at which the one-pion
exchange is resolved and EFT(/ π ) must break down. The resultant power-counting orders
each process according to the power in Q at which it starts to contribute and establishes
therefore an ordering scheme which is used for systematically improvable, rigorous er-
ror estimates. This and the systematic, gauge invariant inclusion of external electro-weak
currents and relativistic effects, see, e.g., [3], distinguishes EFT(/π ) from its historical
roots, the effective range expansion [4] and the model-independent approach to three-body
physics [5].
EFT(/π ) is a mathematically well-defined, systematic low-energy theory of QCD. It is
computationally considerably simpler than potential models or the “pion-full” version of
EFT, which attempts to extend chiral perturbation theory to the few-nucleon system [6–8].
Conceptually, many problems which are also found when formulating a fully consistent
“pion-full” EFT are encountered in a simpler setting. It was used quite successfully to
194 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

provide model-independent results for a cornucopia of two-nucleon processes with exter-

nal electro-weak probes, see the reviews [9–11] and references therein, also for a sketch
of its development. Bedaque et al. [12,13] showed that a momentum-independent three-
body force must be LO for the triton, and sketched the path to include further three-body
forces systematically at higher orders. Recently, a rigorous power-counting to all orders
was developed for the three-nucleon forces in this approach, opening the path to new high-
accuracy extractions and predictions of nucleonic and nuclear properties [14]. Barford and
Birse confirmed this in an analysis of the renormalisation group flow of the position-space
version of the problem [15]. Now, convergence issues become interesting. They are the
subject of this article.
In EFT(/ π ), the strengthes of the contact interactions between the nucleons can be de-
termined from low-energy observables in various ways, which differ in principle only
by higher-order effects. Still, it is standard practise to improve the speed of conver-
gence by physical considerations. One would, for example, not try to describe deuteron
properties by starting from the Effective Range Expansion (ERE) in the 3 S1 -channel of
NN -scattering around zero momentum, in which the correct position of the deuteron
pole at Bt = 2.2246 . . . MeV is reached only perturbatively. Instead, it is prudent to put
the deuteron pole-position in the right place, and give the pole its correct strength. The
deuteron wave-function has then the proper asymptotic fall-off and normalisation. Phillips
et al. [16] showed that this Z-parameterisation is an effective way to sum up the dom-
inant effective-range contributions. It both simplifies calculations of deuteron properties
and improves convergence.
As the 3 S1 -channel of NN -scattering is an important sub-cluster of the three-nucleon
system, it is natural to consider the implications of this choice for the triton and other
partial waves. With a slight extension, it can be included into the standard EFT treat-
ment of the three-nucleon system in which an auxiliary two-nucleon field is introduced
to simplify computations. This article will show that not only is the convergence both
from order to order in the expansion and to experimental phase-shifts greatly improved;
its results support also the power counting for the three-body forces pertinent to the
triton—mathematically rigorously proven to all orders in [14]—by an error analysis of
the 2 S1/2 -channel.
This article is organised as follows: the next section merges Z-parameterisation with
the auxiliary-field method. After a brief review of the formalism to compute Nd-scattering
in EFT(/ π ) and the occurrence of three-body forces in the triton channel at the beginning
of Section 3, the results for the 2 S1/2 -channel, triton properties and the Phillips line are
examined in Section 3.3. I touch upon the pros and cons of including the effective range as
leading-order in Section 3.4. Finally, the effect of Z-parameterisation on the other partial
waves of the Nd-system is discussed in Section 3.5. Various effective-range parameters
of low partial waves are also listed, together with comparisons between experimental and
theoretical cross-sections. The conclusions in Section 4 are followed by Appendix A on the
derivation of the three-body equations in EFT(/ π ), defining also the spin–isospin projection
operators in the three-nucleon system.
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 195

2. Z-parameterisation and auxiliary-field-formalism in the two-nucleon system

2.1. Merging the auxiliary field formalism . . .

It is a standard technique to simplify calculations in EFT(/π ) both in the two- and three-
nucleon system by considering not directly contact interactions between four nucleons, but
to introduce auxiliary fields with the quantum numbers of the two-nucleon real and virtual
bound states, coupling to two nucleons,1 see, e.g., [17–21]:

  
L2N,t = −y dti† N T Pti N + H.c.
    ∞  
i† ∂ 2 γt2 ∂ 2 γt2 n+1 i
+ dt ∆t − c0t i∂0 + + − cnt i∂0 + + dt .
4M M 4M M
n=1
(2.1)
p 
Here, N = n is the nucleon iso-doublet and Pti = √1 τ2 σ2 σ i the projector onto the spin-
8
triplet iso-spin-singlet state with vector index i = 1, 2, 3. σ i (τ A ) are the spin (iso-spin)
Pauli matrices, A = 1, 2, 3 the iso-vector √ indices. The auxiliary field dt represents the
deuteron with binding momentum γt = MBt = 45.7025 MeV. Strictly speaking, one
should replace the (iso-scalar) nucleon mass M → M − Bt /2 to obtain the correct deuteron
mass, but this effect is negligible in what follows. Only the 3 S1 -channel of NN -scattering
is considered here; the 1 S0 -channel is discussed in Section 2.3. The choice of sign for cnt
is traditional [17].
That this Lagrangean is on-shell equivalent to the one containing only nucleon fields was
formally shown in [20] by a Gaussian integration over dt , followed by a field-redefinition
and disregarding terms with more than four nucleon fields. The advantage of this scheme—
stressed repeatedly, e.g., in [14,18–20,22]—is that the parameters follow naïve dimensional
analysis: as again discussed in Section 2.2, ∆t ∼ Q, with mass-dimension 1, is LO; the
dimension-less parameter c0t ∼ Q0 first appears at NLO since it comes with two powers of
momentum, c0t p2 ∼ Q2 ; while a dimension-full operator proportional to cnt ∼ Q0 enters
at N2n+1 LO because it is accompanied by 2n + 2 powers of the typical momentum.
All interactions which are not pure S-wave—like SD-mixing, P-wave scattering be-
tween two nucleons, etc.—are added either as interactions between the deuteron dt and
two nucleons, or between four nucleons [21,22]. They are not listed as they are of higher
order than necessary in the following—with the exception of SD mixing, whose effect will
be neglected in the three-nucleon system, see the discussion in Section 3.1. The only addi-
tional term at N2 LO is the kinetic energy of the nucleon, which up to relativistic corrections
is:
 
∂ 2
L1N = N i∂0 +
†
N. (2.2)
2M
The bare deuteron propagator i/∆t is thus dressed at LO by all interactions proportional
to y with an arbitrary number of loops, see Fig. 1. At NLO, one perturbative insertion

1 As by-product, I attempt to unify the notational cornucopia of which the present author is not completely
innocent. Throughout, the sub-script t (s) denotes quantities in the spin-triplet (singlet) channel of N N -scattering.
196 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

Fig. 1. Top: re-summation of the bare deuteron propagator (thick gray line) into the dressed deuteron propagator
(double line) at LO by dressing with two-nucleon bubbles. Bottom: the NLO, N2 LO and N3 LO corrections to the
deuteron propagator. The cross denotes an insertion of the deuteron kinetic-energy operator proportional to c0t ,
the star one proportional to c1t .

proportional to c0t is included, followed by two at N2 LO, and in general n at Nn LO. In

addition, one insertion of the operator proportional to cnt enters at N2n+1 LO, etc.
Choosing for the dt NN coupling constant without loss of generality
4π
y2 = , (2.3)
M
the auxiliary-field propagator with kinetic energy p0 and momentum p becomes particu-
larly simple in the low-energy expansion:
i
 =
iDt (p0 , p)
p2
∆t + µ − 4 − Mp0 − i
∞  p2 γ2  ∞ c p − p2 + γt n+1 m
2
 c0t p0 − 4M + Mt + n=1 nt 0
× 4M M
p2
m=0 ∆t + µ − 4 − Mp0 − i
i
→ .
p2  p2 γt2  ∞ c p − p2 + γt n+1
2
∆t + µ − 4 − Mp0 − i − c0t p0 − 4M + M − n=1 nt 0 4M M

(2.4)
Here, µ is the regulator of the linear divergence in the nucleon loop, regularised using
dimensional regularisation with the PDS subtraction scheme [23]. The scattering amplitude
between two non-relativistic nucleons with relative centre-of-mass (cm) momentum k in
the 3 S1 -channel is obtained by multiplying with −y 2 = −4π/M and setting the nucleons
on-shell (p0 = k 2 /M, p = 0 ):
4π 1
ANN (k) = −
M ∆t + µ + ik
∞  2 γ2   k2 γt2 n+1 m
 c0t kM + Mt + ∞ n=1 cnt M +
× M
∆t + µ + ik
m=0
4π 1
→− . (2.5)
M ∆ + µ + ik − c  k 2 + γt2  ∞  k2 γt2 n+1
t 0t M M − n=1 cnt M + M
To sum all effective-range corrections to all orders in the third line of (2.4) and (2.5) has no
advantage but to shorten the following determination of the low-energy coefficients; actual
calculations involve only a finite number of cnt ’s, in which case the difference between
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 197

the two lines is formally of higher order. Problems with re-summing the effective-range
corrections “to all orders” are discussed in Section 3.4.

2.2. . . . with Z-parameterisation

In Effective Range Expansion (ERE) around the deuteron pole [4], the scattering am-
plitude of two non-relativistic nucleons with relative cm momentum k in the 3 S1 -channel
is
  M 1
ANN 3 S1 , k = − ρ ∞ , (2.6)
4π γt − 2 (γt + k ) − n=1 ρnt (γt2 + k 2 )n+1 + ik
0t 2 2

where ρ0t = 1.764 fm is the effective range, ρ1t = 0.389 fm3 the shape parameter, etc.
Matching (2.5) to the ERE-result (2.6), one obtains the ERE-parameterisation of EFT(/ π ):
ρ0t
∆t,ERE + µERE = γt , c0t,ERE = M, cnt,ERE = ρnt M n+1 ∀n
1. (2.7)
2
While γt is clearly a typical low-momentum scale in EFT(/ π ), γt  Λπ/ ∼ mπ , the other pa-
rameters encode physics beyond the breakdown scale and hence one expects ρnt ∼ Λπ−2n−1 /
from dimensional analysis. Indeed, contributions to the scattering amplitude for example
from the shape-parameter are small, with the expansion parameter Q estimated as Q ∼
1/3 1/(2n+1)
γt ρ1t ≈ 0.17. Therefore, the amplitude can be expanded in powers of Q ∼ γt ρnt ,
and an error estimate follows, rendering effective range theory useful:
  
3  M 1 ρ0t γt2 + k 2 ρ0t γt2 + k 2 2  
ANN S1 , k = − 1+ + + O Q3 .
4π γt + ik 2 γt + ik 2 γt + ik
(2.8)
A problem with the expanded version (2.8) arises however because the effective range is
numerically somewhat larger than naïvely expected, ρ0 > m−1
π or

γt ρ0t ≈ 0.41. (2.9)

While the deuteron binding energy is in ERE-parameterisation reproduced immediately at
LO, the residue of the scattering amplitude at the deuteron pole is now only expanded in
this numerically not-so-small parameter,

  4π
Res ANN 3 S1 , k = iγt = − Zt with
M
1
Zt = = 1.690(3)
1 − γt ρ0t
= 1 + γt ρ0t + (γt ρ0t )2 + (γt ρ0t )3 + · · ·
= 1 + 0.409 + 0.167 + 0.068 + · · · , (2.10)

so that even including N2 LO-effects, seven percent are missed to close the gap to the exact
value. In contrast, not considering the N3 LO shape-parameter effects in A leads only to a
deviation on the order of γt3 ρ1t = 0.5%.
198 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

Clearly, the residue is an important characteristic of the two-body system as it deter-

mines the asymptotic normalisation of the deuteron wave-function at large distances r:

γ Zt e−γ r
Ψdeuteron(r → ∞) = . (2.11)
2π r
Phillips et al. [16] re-summed therefore the series in γt ρ0t partially by expressing ρ0t via
Zt . To discuss its advantage, re-write (2.6) in a superficially more complicated way as
1 + (Zt − 1) 1
γt + ik 1 + Zt −1   ∞
2 1 + γikt − Zt n=1 ρnt (γt − ik)(γt2 + k 2 )n
1

= 1 + (Zt − 1)
γt + ik
     2
Zt − 1 ik Zt − 1 ik  
× 1− 1+ + 1+ + O Q3 . (2.12)
2 γt 2 γt
Formally, Zt − 1 ≈ 0.69 is still treated as a small expansion parameter in the resulting
Z-parameterisation, but the deuteron residue is now restored already at NLO with no cor-
rections from higher orders as the second denominator in (2.12) has no residue at γt = −ik:
Zt = 
1 + (Zt − 1) + 
0 , n
2. (2.13)
  
LO NLO Nn LO

Outside the deuteron pole, the additional terms from Zt also converge faster because the
second term in the second denominator is a power series in (Zt − 1)/2 ≈ 0.3. A slight dis-
advantage is that the expansion parameter for the higher-order correction is now bigger by
70%: Zt ρnt γt2n+1 . Thus, the N3 LO-correction from the shape parameter is now ∼ 0.8%.
At this level of accuracy, however, other corrections (e.g., from P-wave interactions) must
be considered, whose scale is also set by the pion mass and which are estimated to be
stronger because (γt /(Λπ/ ∼ mπ ))3 ∼ 3%.
In Z-parameterisation, the expanded version of the EFT(/ π )-amplitude (2.5) is there-
fore first matched to reproduce the correct deuteron pole position: ∆t + µ = γt . Then, the
residue is found starting at NLO as
∞  
! 2γt n
Zt = 1 + c0t . (2.14)
M
n=1
At LO, no additional free parameter exists in EFT and the residue is one. At higher orders,
the residue is set equal to Zt when c0t is suitably chosen: With the expansion truncated at
max (n)
some finite order, c0t = nn=0 c0t contains now contributions from higher orders in the
(n)
Q-expansion, c0t being O(Q ). To summarise, the parameters in Z-parameterisation are
n

4π
y2 = ∼ Q0 , (2.15)
M
∆t + µ = γt ∼ Q, (2.16)
(n) M
c0t = (−) (Zt − 1)
n n+1
∼ Qn , (2.17)
2γt
cnt = ρnt M n+1 ∼ Q0 ∀n
1. (2.18)
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 199

In contradistinction, c0t receives in ERE-parameterisation only a contribution at order Q0 ,

and none at higher orders (2.7). However, such an expansion is with our choice of La-
grangean (2.1) only encountered for one parameter, and not—as previously—for all [14,
19,20,22].
The auxiliary-field propagator to N2 LO,

1
Dt (p0 , p ) =
p2
γt − 4 − Mp0 − i

 
Zt − 1 p2
1 +
×  γt + − Mp0 − i
2γt 4
LO   
NLO
 2  
Zt − 1 p2  
+ − Mp0 − γt2 +O Q3 , (2.19)
2γt 4
  
N2 LO

is thus a faster-converging alternative to the usual ERE of the NN -scattering amplitude

in the 3 S1 -channel. The wave-function renormalisation, i.e., the residue of the deuteron
propagator, is by construction exact at NLO:
  −1
∂ 1 
 2γt  
Zt := = 1 + (Zt − 1) + 0 , n
2 .
∂p0 Dt (p0 , 0 ) p0 =− γt2 M      n
M LO N LO
NLO
(2.20)
One could also replace directly the deuteron propagator in ERE by the result in Z-
parameterisation (2.12), as, e.g., argued to be computationally simpler by Beane and
Savage [21]. This would be mandatory if γt ρ0t ≈ 1. Equivalently, if one re-sums all orders
in c0t , one finds again the ERE-parameterisation (2.7), c0t = ZZt −1 M
t 2γt
= ρ20t M. However,
there are a number of dis-advantages of this approach, as will be discussed in Section 3.4.

2.3. Z-parameterisation for the 1 S0 -channel?

The obvious question is: why not also impose Z-parameterisation for the 1 S0 -channel
of NN -scattering? As its bound state is only virtual, one usually performs the ERE around
zero momentum,
  M 1
ANN 1 S0 , k = − r0s 2 ∞ , (2.21)
4π − a1 − 2 k − n=1 rns k
2n+2 + ik
s

with as = 23.714 fm, r0s = 2.73 fm, r1s = −0.48 fm3 the scattering length, effective range
and shape parameter. As r0s /as ≈ 0.11, the series-expansion of the residue of the virtual
bound-state converges much faster than in the deuteron channel. Still, in order to simplify
notation, consider a Lagrangean analogous to the 3 S1 -channel:
200 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

   
L2N,s = −y dsA† N T PsA N + H.c.
   ∞  
∂ 2 γs2 ∂ 2 γs2 n+1 A
+ ds ∆s − c0s i∂0 +
A†
+ − cns i∂0 + + ds .
4M M 4M M
n=1
(2.22)
The auxiliary field ds represents the spin-singlet iso-spin-triplet state, whose projector is
PsA = √1 σ2 τ2 τ A . One now first re-writes the ERE expansion (2.21) in the form analogous
8
to (2.6) in which the pole position does not change from order to order,
  M 1
ANN 1 S0 , k = − ρ0s ∞ (2.23)
4π γs − 2 (γs
2 + k2) − n=1 ρns (γs
2 + k 2 )n+1 + ik
and determines the coefficients by matching as
∞
1 r0s 2   2 n+1
γs = + γ − rns −γs , (2.24)
as 2 s
n=1
∞
  n
ρ0s = r0s + 2 (n + 1)rns −γs2 , (2.25)
n=1
∞ 
 
n+1  n−l
ρms = rms + rns −γs2 , m > 0. (2.26)
n=m
m+1

Truncation at ρ0s (ρ1s ) leads to the numerical values γs = −7.8904 MeV (−7.8902 MeV),
ρ0s = 2.730 fm (2.733 fm) (ρ1s = r1s ). The residue is indeed very close to unity:
1
Zs = = 0.9016 (0.9015). (2.27)
1 − γs ρ0s
At N2 LO, the difference between the perturbatively built residue 1 + γs ρ0s + (γs ρ0s )2 =
1 − 0.1092 + 0.0119 = 0.9027 and the exact values is with 0.1% considerably smaller
than leaving out relativistic and other effects. In the results presented in the following
section, the difference between the ERE-parameterisation and Z-parameterisation for the
1 S -channel cannot be discerned in the plots of the phase-shifts. Still, the re-formulation
0
serves the purpose to compactify and simplify formulae. The parameters of the Lagrangean
are determined in Z-parameterisation by the analogue to (2.15):
4π
y2 = ∼ Q0 , (2.28)
M
∆s + µ = γs ∼ Q, (2.29)
(n) M
c0s = (−)n (Zs − 1)n+1 ∼ Qn , (2.30)
2γs
cns = ρns M n+1 ∼ Q0 ∀n
1. (2.31)
Note that in contradistinction to previous work, y is chosen to be identical in the spin-triplet
and spin-singlet channel and thus carries no sub-script.
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 201

3. The three-body system in Z-parameterisation

3.1. Formalism
With the parameters of the two-nucleon Lagrangean fixed by Z-parameterisation, it is
now straight-forward to state the equations governing neutron–deuteron scattering. As they
were derived repeatedly in the literature (see, e.g., [13,22,24]), the following presentation
focuses mainly on notation. Appendix A contains a brief overview, defining also the perti-
nent projection operators on the various partial waves and three-nucleon configurations.
Two cluster-configurations exist in the three-nucleon system: the Ndt -cluster with total
spin S = 3/2 or S = 1/2, depending on whether the deuteron and nucleon spins are parallel
or anti-parallel; and the Nds -cluster which has total spin S = 1/2, as dsA is a scalar. The
leading-order three-particle amplitude is O(Q−2 ) (before wave-function renormalisation)
and includes all diagrams built out of the leading two-body interactions, i.e., the ones
proportional to y and ∆t , ∆s in the two-nucleon Lagrangeans (2.1)/(2.22). The resultant
Faddeev integral equation—first derived by Skorniakov and Ter-Martirosian [24] without
three-body force—is pictorially represented in Fig. 2.
As the Lagrangean up to N2 LO does not mix partial waves or flip the spin of the aux-
iliary fields, angular momentum is conserved in the quartet and doublet channels. Strictly
speaking, SD-mixing in the deuteron channel produces a splitting and mixing of the three-
body amplitudes with the same spin and angular momentum but different total angular
momentum. However, we limit ourselves to the averaged phase-shifts, as the spitting is at
this order not going to be realistic enough to describe spin-observables in the three-nucleon
system. This path was also pursued in [14] and [22] which used ERE-parameterisation, to
which the findings in Z-parameterisation will be compared. An analysis of these spitting
is postponed to a future presentation, which also deals with spin-observables in the three-
nucleon system.
The spin-quartet channel of the Nd-system is particularly simple. The lth partial wave
(l)
of the amplitude tq in the centre-of-mass (cm) frame is given by2
tq(l) (E; k, p) = −4πK(l)(E; k, p)
∞  
2 q2
+ dq q K (E; q, p)Dt E −
2 (l)
, q tq(l) (E; k, q). (3.1)
π 2M
0

Fig. 2. The Faddeev equation for N d-scattering to N2 LO. Thick solid line: propagator of the two intermediate
auxiliary fields ds and dt , denoted by D, see (3.6); K: propagator of the exchanged nucleon, see (3.2); H:
three-body force, see (3.11).

2 The sub-script q (d) denotes quantities in the spin-quartet (doublet) channel of the three-nucleon system.
202 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

2 γ2
The total non-relativistic energy is E := 34M
k
− Mt ; the incoming (outgoing) deuteron-

momentum k (p ); and the projected propagator of the exchanged nucleon on angular
momentum l 3
1
1 Pl (x)
K (E; q, p) :=
(l)
dx
2 p2 + q 2 − ME − i + pqx
−1
 2 
(−1)l p + q 2 − ME − i
= Ql , (3.2)
pq pq
where the lth Legendre polynomial of the second kind with complex argument is as in [25]
1
1 Pl (t)
Ql (z) = dt . (3.3)
2 z−t
−1

In the doublet channel, the Faddeev equation is two-dimensional in cluster-configuration

space as both Ndt - and Nds -configurations contribute:

td(l) (E; k, p)
    
1 1
= 2π K (E; k, p)
(l)
+ δ H(E; Λ)
l0
−3 −1
∞     
1 1 −3 1 −1
− dq q 2 K(l) (E; q, p) + δ l0 H(E; Λ)
π −3 1 −1 1
0
 
q2
, q t d (E; k, q).
(l)
×D E− (3.4)
2M
The three-body force H will be discussed in the next subsection. The vector
 (l) 
td,t t
t (l)
d := (l)
(3.5)
td,t s
(l)
is built out of the two amplitudes which get mixed: td,t t for the Ndt → Ndt -process, and
(l)
td,t s for the Ndt → Nds -process. Furthermore,
 
Dt (p0 , p) 0
D(p0 , p) := (3.6)
0 Ds (p0 , p)
is the propagator of the two intermediate auxiliary fields.
How to calculate higher-order corrections? The re-summation of all interactions pro-
portional to y and ∆s/t into a Faddeev equation is mandatory since the power-counting
classifies all diagrams which are built only out of them as contributing equally strong to

3 This corrects a typographical error in a previous paper on the higher partial waves in the N d-system [22],
affecting the odd partial waves.
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 203

the final amplitude. Ref. [14] proposed to calculate the higher-order corrections by ex-
panding the kernel and inhomogeneous part of the integral equation in powers of Q to the
desired order of accuracy, and iterate then by inserting it into a Faddeev equation ( partially
re-summed Nn LO-calculation). This re-summation of some higher-order effects does not
increase the accuracy of the calculation, which is still set by the accuracy to which the
kernel is expanded. It does however simplify the numerical treatment, as no divergences in
the amplitude or in three-body forces H are encountered as the numeric cut-off is removed.
It also leads to a simple, analytical argument at which orders three-body forces with deriv-
atives enter, as discussed below. In contradistinction, a strict perturbation around the LO
solution soon becomes cumbersome numerically, as full off-shell amplitudes need to be
computed and the numerical integrals soon start to diverge, making a numerical renormal-
isation necessary [26].
The computational effort to solve the integral equations (3.1)/(3.4) numerically is trivial,
as all potentials are separable. A simple M ATHEMATICA code can be down-loaded from
http://www.physik.tu-muenchen.de/~hgrie. Here, a step-function cut-off Λ in momentum
space was chosen. The cut-off dependence of the results is discussed below. To reduce
the numerical instabilities from the poles of the two-nucleon amplitudes and logarithmic
singularities of the projected nucleon propagator in the kernel, the integral equations are—
following Hetherington and Schick [27]—first solved on a contour in the complex plane.
The amplitudes on the real axis are then re-constructed by another use of the equations.
A grid of 70 points does more than suffice for numerical stability. For example, the imagi-
nary parts of k cot δ vanish below the deuteron-dis-integration threshold easily to 1 part in
108 .
Finally, the scattering phase-shift of the lth partial wave in the quartet and doublet chan-
nel is related to the renormalised on-shell amplitudes by
3π 1 (l) 3π 1
Tq(l) = Zt tq(l) = , Td,xy = , (3.7)
M k cot δq(l) − ik (l)
d,xy − ik
M k cot δ

where x, y = s, t label the matrix entries in cluster-configuration space, and

 
Zt √ 0
Td(l) = Z t (l) with Z := (3.8)
d 0 Zt Zs
is the renormalised doublet-amplitude and its wave-function renormalisation. In the dou-
blet channel, the only observable process is nucleon–deuteron scattering, Ndt → Ndt ,
i.e., x = y = t. Results are not discussed for the unphysical processes Ndt ↔ Nds and
Nds → Nds .
The elastic differential cross-section from Nd-scattering in the cm-frame is finally [28]
∞ 2 ∞ 2
dσ 1  Pl (cos θ )  
 Pl (cos θ ) 
=  (2l + 1)  + 2 (2l + 1)  . (3.9)
dΩ 3  k cot δ − ik 
(1)  k cot δq − ik 
(1)
l=0 d,t t l=0
A maximum angular momentum lmax = 4 serves to converge the differential cross-section.
The recipe to compute Nd-scattering up to N2 LO is hence as follows: expand the
auxiliary-field propagators and their wave-function renormalisations in the integral equa-
tions (3.1)/(3.4) to the desired level of accuracy in Z-parameterisation as in (2.19)/(2.20),
204 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

and then solve the Faddeev equation. To N2 LO, the projected nucleon propagator K(l) is
unchanged.
Nucleon–deuteron scattering is to N2 LO thus completely determined by four simple ob-
servables of NN -scattering: the deuteron binding energy and residue (or effective range),
and the scattering length and effective range of the 1 S0 -channel. Only the 2 S1/2 -channel
has further unknowns, namely the strength of the three-body interaction H0 at LO and
NLO, and in addition of H2 at N2 LO, as discussed momentarily. They are determined by
its measured scattering length ad [29] and the triton binding energy Bd , respectively:

ad = (0.65 ± 0.04) fm, Bd = 8.48 MeV. (3.10)

But why are two and only two three-body force needed at N2 LO?

3.2. Three-body force and Lagrangean

We now turn to the three-body force H already included in the doublet-channel Faddeev
equation (3.4). The Pauli principle in the quartet channel and the centrifugal barrier in all
partial waves l > 0 forbids three-body forces in nearly all partial waves at N2 LO; see also
[30]. However, neither rules out a three-body force without derivatives in the 2 S1/2 -wave,
the physically most interesting channel, which contains the triton and 3 He as real bound
states. While one would naïvely have guessed that such an energy-independent three-body
force scales as Q0 and hence enters only at N2 LO, an unusual renormalisation of the three-
nucleon system in the triton channel mandates its inclusion into the Faddeev equation as a
LO term [12,13]: as was first pointed out by Minlos and Faddeev [31], this channel suffers
from a peculiar cut-off sensitivity of the on-shell amplitudes if three-body forces are absent.
The kernel of (3.4) is for l = 0 at short distances identical to the one of an attractive 1/r 2 -
potential between the auxiliary fields and the nucleon [10]. Without a three-body force, the
wave function would hence collapse, and all observables would become sensitive to physics
at very short distances, a phenomenon well-known as Thomas effect [32]. Its mathematical
origin lies in the fact that in the absence of a three-body force, the solution to the integral
equation is not unique because it allows for a zero mode, as the kernel of the doublet-S-
wave Faddeev equation is not compact [33].
In praxi, the integral equation is solved numerically by imposing a cut-off Λ, which
should not be confused with the breakdown-scale Λπ/ of EFT(/ π ). In that case, a unique
solution exists in the 2 S1/2 -channel for each Λ and H = 0, but no unique limit as Λ → ∞.
The Nd-scattering length, for example, can be tuned as a function of Λ to have any value
between minus infinity and plus infinity. As long-distance phenomena must however be in-
sensitive to details of the short-distance physics (and in particular of the regulator chosen),
Bedaque et al. [12–14] showed that the system must be stabilised by a three-body force
∞  
2  ME + γt2 n
H(E; Λ) = H 2n (Λ)
Λ2 Λ2
n=0
2H0(Λ) 2H2 (Λ)  
= 2
+ 4
ME + γt2 + · · · (3.11)
Λ Λ
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 205

which absorbs all dependence on the cut-off as Λ → ∞. It is analytical in E and can be ob-
tained from a three-body Lagrangean, employing a three-nucleon auxiliary field analogous
to the treatment of the two-nucleon channels, see [14] and the end of this subsection.
H2n is dimension-less but depends on the cut-off Λ in a non-trivial way, as a renormali-
sation-group analysis reveals: instead of approaching a fixed-point as Λ → ∞, it shows an
oscillatory behaviour known as “limit cycle” [12,34,35].
As one needs a three-body force at LO, H0 ∼ Q−2 , to prevent the system from collapse,
all three-body forces obtained by expanding H in powers of E are also enhanced, with the
interactions proportional to H2n entering at N2n LO [14]. Since a numerical verification of
this analytical observation is one of the prime advantages Z-parameterisation has, the argu-
ment is recalled here: Bedaque et al. [13] noted that a re-parameterisation of the amplitudes
relates the problem for the LO three-body force to a particularly interesting symmetry, and
this was extended by Ref. [14]. Building the linear combination
 t (0)    (0) 
Wigner,− 1 1 −1 td,t t
t Wigner :=
(0)
(0)
= (0)
, (3.12)
tWigner,+ 2 1 1 td,t s

the Faddeev equation becomes in the 2 S1/2 -channel

t Wigner (E; k, p)
(0)

   1
1
= 4π K(0) (E; k, p) + H(E; Λ) 2
− 12 0
∞  
2 2K(0)(E; q, p) + H(E; Λ) 0
− dq q 2
π 0 −K(0) (E; q, p)
0
 q2 q2

Σ(E − 2M , q) ∆(E − 2M , q)
t Wigner(E; k, q).
(0)
× (3.13)
q2 q2
∆(E − 2M , q) Σ(E − 2M , q)

While Σ := 12 (Dt + Ds ) = 12 tr D is the “average” NN –S-wave scattering amplitude,

∆ := 12 (Dt − Ds ) = 12 str D parameterises the degree to which the 3 S1 - and 1 S0 -channel
differ. In the Wigner-SU(4)-limit of arbitrary combined spin- and iso-spin rotation [36],
the two amplitudes are identical, γt = γs , ρnt = ρns ∀n. In that case, ∆ = 0, and the two
(0)
equations for tWigner,∓ decouple, as first observed in [13]. The essential observation is now
that for the UV-behaviour of the amplitude, in which all scales are discarded except for
the off-shell momentum q ∼ Λ in the loop, the Wigner-SU(4)-limit is recovered automat-
(0)
ically. tWigner,+ obeys the same integral equation as the quartet-S-wave in (3.1), where the
(0)
Pauli principle forbids three-body forces without derivatives. Only tWigner,− is subject to
the Wigner–SU(4) symmetric three-body force H. Its integral equation is the same as for
three spin-less bosons, whose wave-function is well known to collapse in the absence of
three-body forces [12,31,33]. H0 therefore must be LO to ensure that the physical on-shell
amplitude is cut-off independent.
206 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

To determine the running of H2 , one expands the quantities of (3.13) for q

k, p, γ , . . . :
LO NLO N2 LO
  √
 q2  41 4γ Z − 1 4ME + 8γ 2 4 Z−1 3(Z − 1)2
Σ E− ,q → − − 2
− − √ − − q + ···
2M 3q 3q 2γ 3 3 q3 3 q 8γ 2
2
1 ME − 2q3
K(0) (E; q, p) → 2 + + ···
q q4
2
2H0LO (Λ) 2H0NLO (Λ) 2H0N LO (Λ) ME + γ 2 2H2 (Λ)
⇒ H(E; Λ) → + + + + ···
Λ2 Λ2 Λ2 Λ2 Λ2
(3.14)
At LO, H0 is independent of two-body observables and is fixed by a three-body datum.
At NLO, only Σ has non-vanishing contributions. They depend only on the low-energy
two-body observables γ ∼ Q and ρ0 ∼ Q0 , but not on the total cm-energy E ∼ Q2 —or
equivalently, not on the on-shell momentum k ∼ Q. The cut-off dependence they induce
can hence be absorbed by H0NLO(Λ), a momentum-independent correction to the LO three-
body force whose value now depends also on the two-body observables γ and ρ0 .
At N2 LO, more corrections arise which depend on γ and ρ0 . In addition, both the
projected nucleon propagator K and the NN -scattering amplitude Σ show corrections pro-
portional to ME. To match the behaviour of the off-shell (UV-)amplitude on the physical
on-shell momentum E, the inclusion of an energy-dependent three-body force H2 is thus
mandatory. Its value is not determined by two-body observables. The expansion in powers
of E ∼ Q2 proceeds straightforwardly: another three-body force H2n (Λ) proportional to
E n is needed every even order in Q from expanding both Σ and K(0) . H2n (Λ) first enters
at N2n LO, independent of two-body observables. To determine its strength, one additional
three-body datum is needed. The power-counting for the three-body forces is hence
H0 ∼ Q−2 , H2 ∼ Q−2 , H2n ∼ Q−2 . (3.15)
Corrections to the limit q → ∞ from the breaking of Wigner-SU(4)-symmetry can
only induce a momentum-dependent three-body force, because the only three-body force
without derivatives is necessarily Wigner-SU(4)-symmetric [13]. They can be shown to be
of higher order, see [14] for a more detailed discussion.
The chain of isotropic (S-wave) Wigner-SU(4)-symmetric three-body forces is also
found from a three-nucleon auxiliary-field Lagrangean, in complete analogy to the treat-
ment of the two-nucleon auxiliary fields in Sections 2.1–2.3 [13,14]:
y3 (Λ)
†  i  i  A  A  
L3N = − √ t σ N dt − τ N ds + H.c.
3
∞  
∂ 2 γt2 n
+t Ω −
†
h2n (Λ) i∂0 + + t, (3.16)
6M M
n=1

where the auxiliary field t has the quantum numbers of the triton/3He, i.e., spin and iso-
spin 1/2. The relative coupling strength between the processes t → Ndt and t → Nds is
fixed because there is only one three-body force without derivatives, which also happens to
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 207

be Wigner-SU(4)-symmetric [13,14]. This field is treated analogously to the two-nucleon

auxiliary fields: at LO, the propagator of a “triton” with kinetic energy E and momentum
p is proportional to 1/Ω, with insertions proportional to h2n (Λ) suppressed by n powers
of the triton kinetic energy:
  
i h2 (Λ) p2
1+ ME − + γt2 + · · · . (3.17)
Ω MΩ 6
Comparing the final result of the resultant three-body force in the 2 S1/2 -wave (A.13) with
(3.4), one finds that the couplings y3 (Λ) and h2n (Λ) are in the cm-frame related to H by
∞ 

2y 2(Λ) h2n (Λ)   m
2 n
H(E; Λ) = − 23 1+ ME + γ t (3.18)
y MΩ ΩM n
m,n=1

or matching order by order the dimension-less three-body strengthes H2n (Λ) of (3.11),
y32 (Λ) Λ2 y32 (Λ) Λ4
H0 (Λ) = − , H2 (Λ) = − h2 (Λ), ..., (3.19)
y 2 MΩ y 2 (MΩ)2
so that the couplings are particularly simple when one chooses
Λ2 H2 (Λ)
Ω= , y32 (Λ) = −y 2 H0 (Λ), h2 (Λ) = , (3.20)
M H0 (Λ)
as they follow then from (3.15) the natural scaling laws—with the exception of y3 (Λ):
y32 (Λ) ∼ Q−2 , Ω ∼ Q0 , h2n (Λ) ∼ Q0 . (3.21)
Notice that in contradistinction to Ds/t in Section 2.1, the power counting does not require
even a partial re-summation of the triton propagator.

3.3. The 2 S1/2 - (triton-) channel

That Z-parameterisation improves convergence is most importantly seen in the only

channel in which a bound state is found, and in which the three-body force enters already
at LO to stabilise the system from collapse. In particular, it can be instrumentalised to con-
firm that the first momentum-dependent three-body force enters at N2 LO, as the analytical
argument of the preceding section demands.
Previously, Hammer and Mehen calculated the 2 S1/2 -wave phase shifts below the
deuteron break-up to NLO by sandwiching the ERE-corrections to the two-particle propa-
gators between the LO half off-shell amplitudes [26]. ERE-parameterisation was also used
in the recent article which proposed partially re-summed Nn LO-calculations to simplify
numerics and showed that the momentum-dependent three-body force H2 is conceptually
necessary at N2 LO [14]. An estimate of NLO-effects in Z-parameterisation by Afnan and
Phillips [37] included the part of the two-particle propagator (2.19)/(2.20) which leads to
the correct residues Zs/t , but not the part which changes the off-shell propagation of the
deuteron field, iterating again the perturbatively expanded kernel.
In the following, one has to differentiate between two kinds of convergence: first, the
results should of course agree with available measurements, at the level of accuracy pre-
dicted by EFT. Second, it is vital for a reliable error-estimate and hence for predicting the
208 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

Fig. 3. Comparison of the real (top) and imaginary (bottom) parts of the neutron–deuteron 2 S1/2 -phase-shift
at LO (dotted), NLO (dashed) and N2 LO (dark band) in ERE-parameterisation (left) and Z-parameterisation
(right) as function of the cm-momentum k. The band in the N2 LO-curve shows the variation induced by shifting
the cut-off from Λ = 200 MeV to ∞. Dots: phase-shift analysis from 1967 [38]; crosses: Argonne V18 results
including the Urbana IX three-body force from [39] below, and from [40] above break-up. At LO and NLO, the
results for Λ = 900 MeV are shown. The dot-dashed curve in the imaginary parts of the ERE-result is a partial
N3 LO-calculation as described in the text.

accuracy of the calculation that EFT can demonstrate that contributions which are classi-
fied as higher-order are indeed suppressed, i.e., that corrections from order to order become
smaller in the range of validity. In particular whenever no or scarce data are available, this
a priori error-estimate is essential to minimise the theoretical bias in the description of few-
body properties. Z-parameterisation improves both variants over ERE-parameterisation.
The ERE- and Z-parameterisation results for the phase-shifts of the 2 S1/2 -channel at
LO, NLO and N2 LO are shown in Fig. 3, together with the only available phase-shift analy-
sis [38], dating from 1967, and results using the Argonne V18 potential supplemented with
the Urbana IX three-body force [39,40]. The bands in the N2 LO-results come from vary-
ing the cut-off in the Faddeev equation between 200 and 900 MeV, when Λ-independence
is achieved within drawing-accuracy, except at those points at which the strengthes of the
three-body forces diverge, see later Fig. 5.
At this order, considerably more sophisticated and involved potential-model calcula-
tions must agree with the EFT-predictions, if all are fitted to the same low-energy observ-
ables. For lack of a direct phase-shift analysis of the three-nucleon system, convergence to
experiment is discussed by comparison to such a potential-model calculation which repro-
duces the same two-nucleon observables, triton binding energy and scattering length in the
2S
1/2 -wave. At higher orders, EFT-calculations may be more accurate than those of poten-
tial models, because the EFT-result will depend on new three-body low-energy coefficients
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 209

Fig. 4. Convergence of k cot δ as function of the cm momentum. Left: deviation from the result of the poten-
tial-model calculation [39,40]. Right: dependence on cut-off variations between Λ = 200 and 900 MeV. Dotted:
LO; dashed: NLO; solid: N2 LO; thin lines: ERE-parameterisation; thick lines: Z-parameterisation. Thick dots on
the right: “N2 LO”-calculation in Z-parameterisation with H2 set to zero.

which do not enter in potential models, while a systematic treatment of three-body forces
is built into EFT(/ π ).
At LO, both parameterisations are identical. As expected from the discussion in Sec-
tion 2.2, the NLO corrections are larger for Z-parameterisation than for the ERE-version
due to the inclusion of the full strength of the deuteron residue. Higher-order corrections
in Z-parameterisation should be considerably smaller than in ERE-parameterisation, be-
cause their typical scale from NN -scattering is in the former set by (Zt − 1)/2 ≈ 0.3,
while in the latter, it is γt ρ0t ≈ 0.4. This amounts to a difference in the predicted accu-
racy of the N2 LO-calculation of (0.3)3 ≈ 3% in Z-parameterisation versus (0.4)3 ≈ 6.5%
in ERE-parameterisation for typical momenta in the three-body problem. Indeed, the cor-
rection from NLO to N2 LO is in general larger for the ERE-parameterisation than for
the Z-parameterisation, and the result closer to the potential-model values. The plot of
the logarithmic deviation of the on-shell point of the inverse K-matrix, k cot δ, from the
potential-model numbers at each order in both parameterisations verifies these findings on
the quantitative level, see left panel in Fig. 4.
In particular, the ERE-result for the imaginary part of the phase-shift has at large mo-
menta not converged to the potential-model results at N2 LO, while Z-parameterisation
agrees with these computations. As the effective-range effects are expected to be the dom-
inant corrections in ERE-parameterisation, the imaginary part of the ERE-result in Fig. 3
contains also a computation in which these are included in the deuteron propagator (2.4) to
order (γt ρ0t )3 . This is only a partial N3 LO-calculation, neglecting shape-parameter, rela-
tivistic, SD-mixing and other corrections, but improves the agreement to the level achieved
already at N2 LO in Z-parameterisation. Close to k = mπ ≈ Λπ/ , both parameterisations
become un-reliable as the NLO and N2 LO-corrections are comparable in size.
At least as important is that the internal convergence is drastically improved in Z-
parameterisation. This is evident from the right side of Fig. 4 which compares the un-
certainties induced into the N2 LO phase-shifts by cut-off variations. As outlined in the
Introduction, the main strength of the EFT-approach is that it allows for an a priori estimate
of the theoretical uncertainties of a calculation because contributions to the amplitudes are
210 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

ordered by a small expansion parameter Q. Nn LO corrections to k cot δ should typically

be of the order
 
ptyp n
∆(k cot δ) ∼ Q =
n
(3.22)
Λπ/
compared to the LO result. Typical low-momentum scales ptyp in the three-body sys-
tem are the binding momenta of the two-nucleon real and virtual bound states, γs ≈
−8.0 MeV, γt ≈ 45 MeV and the scattering momentum k. In addition, the three-body
forces
√ are determined in part by the typical three-nucleon bound-state momentum γd ∼
MBd ≈ 90 MeV, Bd the triton binding energy. The breakdown scale Λπ/ ≈ mπ of the
theory is the scale at which higher-order corrections become comparable in size. Like the
actual size of the expansion parameter Q, its value must be verified in actual calculations.
As observables at low energies must be independent of details of short-distance physics,
they must be independent of the arbitrary regulator Λ up to the order of the expansion. In
other words, the physical scattering amplitude (or more accurately its physically relevant
part, k cot δ) must be dominated by integrations over off-shell momenta q in the integral
equations (3.1)/(3.4) in the region in which the EFT is applicable, q  Λπ/ . As argued,
e.g., by Lepage [41], one can therefore estimate sensitivity to short-distance physics, and
hence provide a reasonable error analysis, by varying the cut-off Λ between the break-
down scale Λπ/ and ∞. In praxi, an upper limit Λmax = 900 MeV suffices since the phase-
shifts are essentially cut-off independent beyond Λ ≈ 600 MeV.
On the right in Fig. 4, the logarithmic cut-off variation is displayed as function of the
cm momentum, both for ERE- and Z-parameterisation. Several points are worth noticing.
First, the cut-off variation is substantially smaller in Z-parameterisation
√ over the whole
range. For example, at the typical three-body scale k ∼ MBd ≈ 90 MeV, the cut-off
variation is decreased from 15% and 4% at NLO and N2 LO in ERE-parameterisation to
12% and 1.3% in Z-parameterisation.4
With k the dominant low-energy scale for momenta above the break-up point (4γt2 /3)1/2
≈ 52 MeV, one can secondly verify the estimate (3.22) for the corrections by fitting n to
the nearly straight lines in the right panel of Fig. 4 in the momentum range between 70 and
100 to 130 MeV. At lower momenta, the slope does not change significantly from order
to order because the dominant corrections are ∼ (γs,t /Λ)n and hence nearly independent
of k. The power-law obtained for each order and parameterisation is listed in Table 1. The
slope increases in Z-parameterisation by one unit from LO to NLO, and by two units from
NLO to N2 LO. The latter may stem from the partial inclusion of higher-order graphs in
the partially re-summed integral equation. The ERE-parameterisation follows a weaker
power-law than Z-parameterisation, indicating again greater cut-off sensitivity and less
accuracy.
Thirdly, extrapolating the relative errors from the fit region, the cut-off variations
at NLO and N2 LO become comparable at k ∼ 150–200 MeV, matching the estimate
Λπ/ ∼ mπ .
Finally, one confirms numerically the analytical finding in [14], recalled in Section 3.1,
that a momentum-dependent three-body force H2 enters at N2 LO to increase accuracy.

4 The deviation at N2 LO to the potential-model results is at the same scale improved from 9.1% to 5.7%.
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 211

Table 1
δ(Λ=200 MeV)
Fit of the logarithmic cut-off variation |1 − kk cot
cot δ(Λ=900 MeV)
| of k cot δ, shown in Fig. 4, for
momenta higher than 70 and lower than 100 to 130 MeV to a power-law (k/Λπ/ )n for ERE- and
Z-parameterisation at LO, NLO and N2 LO, and at N2 LO without momentum-dependent three-body
force H2 . Varying the upper limit changes only the numbers for LO
Order n (Z-param.) n (ERE-param.)
LO 1.8–2.1
NLO 2.9 2.6
N2 LO 4.8 3.7
“N2 LO”, H2 = 0 3.1 2.8

Fig. 5. Dependence of the dimension-less three-body couplings at N2 LO on the sharp momentum cut-off
Λ ∈ [140; 2800] MeV in Z-parameterisation. Triangles: H0 (Λ); crosses: H2 (Λ).

Setting H2 = 0 does not only substantially increase the cut-off dependence of the N2 LO-
calculation, putting it close to the NLO dependence. As the slope at large momenta is also
practically unchanged compared to the NLO value, H2 is necessary to improve cut-off
independence.
Notice that observables are at N2 LO converged to be practically cut-off independent as
low as 600 MeV, while the dimension-less three-body forces H0 and H2 vary dramatically
even in a small window of cut-off variations, see Fig. 5. Only when the absolute value of H0
or H2 is unnaturally large ( 10) is cut-off sensitivity felt. In that case, more of the string
of three-nucleon forces would be needed. Such singular points (e.g., at Λ ∼ 1000 MeV)
are not considered here. The actual numerical values of H0 and H2 are quite sensitive to the
regulator chosen, and to the precise values for γs/t and Zs/t used. It is thus not the three-
body forces which are large but the effect of variations in the cut-off Λ on observables.
No simple connection between H0 and H2 exists at N2 LO, except that they both show a
limit-cycle with the same period, which depends on γs/t and Zs/t .
To complete the discussion, Fig. 6 displays the NLO-prediction for the Phillips line
in Z-parameterisation, compared to various old and modern potential model calculations
212 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

Fig. 6. Prediction for the Phillips line at LO (dotted; lightly shaded region) and NLO (dashed; dark shaded region)
in Z-parameterisation when the cut-off is varied in the region Λ ∈ [140; 900] MeV. Left: the dots correspond to
the predictions for the triton binding energy and doublet scattering length in different models with the same
two-body scattering lengths and effective ranges as inputs [42]. ×: experimental result. The outcome in various
modern high-precision N N -potentials is indicated by crosses [43]. Right: larger range, the rectangle showing the
region covered on the left.

which share the same two-body on-shell physics. In the three-body system, their widely
different off-shell behaviour predicts however different three-body bound state energies
and Nd-scattering lengthes. The correlation between scattering length and bound state en-
ergy was first discussed by Phillips [44] and explained by Efimov [42]. The spread induced
by varying the cut-off between 140 and 900 MeV covers the off-shell dependence of all
potential models. Note the pole in the scattering length at Bd ≈ 36 MeV at NLO for a
cut-off Λ = 900 MeV. This intrusion of another bound triton state is however outside the
π ). As ktyp ≈ 180 MeV Λπ/ and the widthes of the LO and
range of applicability of EFT(/
NLO band are comparable, the EFT-calculation has not converged there.

3.4. Including the effective range to all orders?

Why not re-sum all effective-range parameters up to a given order and replace the ex-
panded version of the two-body scattering amplitudes (2.12) with the version (2.6)/(2.21),
? 1
 →
Ds/t (p0 , p) ,
 2   n+1
γs/t + p4 − Mp0 − p2 p2
ρ 2 ∞ ρ
γs/t − 0s/t
2 n=1 ns/t γs/t + 4 − Mp0 − 4 − Mp0

(3.23)
possibly with all coefficients after ρns/t set to zero at N2n+1 LO?
This would immediately
restore all effective-range corrections, give the correct pole residues, and is mandatory if
ρ0 γ ≈ 1. With this re-summation, the scattering length in the 4 S3/2 -channel and phase-
shifts of the higher partial waves [18–20,22,45] were found with an accuracy of  4%. It
was later also shown to simplify analytical calculations in the two-nucleon system [21].
However, the improvement in accuracy is only superficial: the higher ERE-parameters
ρns/t are known with less and less accuracy, and other higher-order terms like P-wave
interactions and relativistic corrections are still not included.
More severely, the propagator (3.23) has additional spurious poles, see Table 2: one
for each S-wave at NLO (ρ0s/t = 0), and two more at N3 LO (ρ1s/t = 0). Their positions
change as more parameters ρns/t are included, and their residues add with the residue
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 213

Table 2 
Binding momentum γspur = MBspur and dimension-less residue analogous to (2.10)/(2.20) of the additional
spurious bound states in the 3 S1 - and 1 S0 -scattering amplitudes when the effective-range parameters are re-
summed as in (3.23)
3 S -channel 1 S -channel
1 0
Order γspur,t [MeV] Zspur,t γspur,s [MeV] Zspur,s
NLO 178 −1.69 152 −0.90
N3 LO −385 −0.14 133 −0.71
170 ± 108i −0.77 ∓ 0.60i −62 ± 351i −0.10 ∓ 0.12i

of the low-lying real and virtual bound state to zero. Albeit they all lie strictly speaking
outside the range of validity of EFT(/ π ), γspur,s/t > Λπ/ , they are, in particular, in the 1 S0 -
channel close enough to influence observables at momenta k above the break-up threshold,
in particular as the closest of the spurious bound states have rather large residues. Recall
that the half-off-shell amplitude goes to zero in the 2 S1/2 -channel only slowly, A(k, p) ∝
1/p [12,31].
In the Nd-system, these additional poles generate new cuts in the solution to the in-
tegral equations: in the doublet channels, they conspire to pose an additional technical
problem when one performs a rotation of the integration contour into the complex plane in
the Hetherington–Schick procedure. On the other hand, expanding the two-particle propa-
gator to a given order as in (2.19), only the physical pole with γs/t  Λπ/ is found in the
two-particle scattering amplitude, and all problems with spurious bound-states are easily
avoided.
Most severely, the asymptotic behaviour of the two-nucleon propagator is changed from
1/p to 1/p2 at large p 1/ρ0s/t . This leads to a non-trivial modification of the argument
for the inclusion of a three-body force in the 2 S1/2 -channel, as the resulting integral equa-
tion is then not analytically soluble in the UV-limit. Therefore, it is not straight-forward
to construct analytically the cut-off dependence of the three-body force as in [13,14] and
demonstrate that it solves the Thomas problem of an attractive 1/r 2 -potential. Gabbiani
argued that with the kernel becoming less attractive, behaving only like 1/p2 (i.e., in po-
sition space like the Coulomb potential 1/r), the cut-off sensitivity of the Wigner-SU(4)
symmetric component (3.13) of the integral equation in this channel disappears [46]. Thus,
he claims that no limit cycle is found, and no three-body force is needed to achieve cut-off
independence. According to him, the resulting phase-shifts without three-body forces in
the 2 S1/2 -channel fail however miserably. I cannot confirm these findings but see that the
phase-shifts in the 2 S1/2 -channel are markedly cut-off sensitive even at high Λ, and that
three-body forces are necessary when the effective range is included to all orders. These
findings will be discussed in a subsequent article [47].
Re-summing the effective-range corrections poses therefore numerically as well as tech-
nically non-trivial problems most importantly in the 2 S1/2 -wave (but also in the higher
doublet-channels) while no increase in accuracy is achieved. To insert on the other hand the
expanded two-nucleon amplitudes (2.19) in the kernel of the integral equation, as proposed
in [14] and done here is a straightforward way to calculate phase-shifts to high accuracy,
free of the problems mentioned above. In the end, all regularisation and re-summation
214 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

schemes—including those which generate spurious bound-states outside the range of va-
lidity of the EFT—must agree to the order of accuracy of the calculation. They must also
have the same number of independently fixed three-body forces. To set up a consistent
power-counting scheme and computations with good numerical convergence in a particu-
lar version can however be cumbersome and time-consuming.

3.5. More partial waves

After discussing the merits of Z-parameterisation in the only channel in which a three-
body force enters already at LO and which exhibits a bound state of three nucleons, we now
turn to the other partial waves. The solutions of the integral equations (3.1)/(3.4) for the
real and imaginary parts of the phase-shifts are displayed in Fig. 7 for the doublet channel,
and in Fig. 9 for the quartet channel. The quartet-S-wave is shown separately in Fig. 8. In
the following, they are compared to potential-model calculations, and their convergence
from order to order is discussed and juxtaposed with the phase-shifts obtained when the
effective range is included to all orders. Table 3 lists low-energy parameters of lower partial
waves.
Previous work on the higher phase-shifts in the Nd-system in EFT(/ π ) focused mainly
on the 4 S3/2 -channel, where pd-scattering [45] and the scattering length [18–20] and
phase-shifts [19,20,22] of nd-scattering were computed up to N2 LO. The phase-shifts of
the other partial waves were presented in Ref. [22] to N2 LO. The NLO-results of [20] and
[22] were obtained in “strict perturbation”, i.e., sandwiching the NLO-correction to the
deuteron propagator proportional to c0s/t once between the LO-solution to the half-off-
shell amplitude. Ref. [22] used already the Z-parameterisation to fix c0t at NLO. In each
case, the N2 LO-calculations were carried out with the re-summed deuteron propagator
(3.23).
Here, we use as in the 2 S1/2 -channel Z-parameterisation at each order and include per-
turbatively all corrections up to NLO and N2 LO, respectively, into the kernel of the integral
equation. As already discussed in Section 3.1, this contains some higher-order effects and
leads to technical simplifications but—like a re-summed deuteron propagator—not to in-
creased accuracy. The goal is again to show that convergence and agreement with results
from potential-model calculations is improved.
Like in the discussion of the 2 S1/2 -wave in Section 3.3, the N2 LO-results are compared
to phase-shifts of a modern high-precision NN -potential. Numbers for nd-scattering us-
ing the AV18-potential with the Urbana IX-three-body-force were reported by Kievsky
et al. [39,40] and Hüber et al. [48,49]. Within the drawing accuracy, they agree with the
phase-shifts reported for the Nijmegen potentials 93, I and II, and for CDBonn [49]. As the
partial waves do not mix in EFT at N2 LO, the potential-model results for the 2s+1 lj -partial
waves were grouped into bins of fixed spin s and angular momentum l, with total angular
momenta j = l ± 12 in the doublet channel, and j = l ± { 12 ; 32 } in the quartet.
The higher the partial wave, the better the agreement with the potential model calcu-
lations. This comes as no surprise, since the strong centrifugal barrier for large angular
momenta eliminates sensitivity on short-distance physics and the solution to the integral
equation approaches the result of the Born approximation.
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 215

Fig. 7. Real (left) and imaginary (right) part of the phase-shifts of the higher partial waves in the doublet channel
in Z-parameterisation as function of the cm-momentum k, Λ = 1000 MeV. Dotted: LO; dashed: NLO; solid:
N2 LO; dot-dashed: effective-range corrections summed to all orders. Result of AV18 + U IX from [39] below
break-up, from [48] above, and at k = 104 MeV from [49]. 2 lj -wave denoted by ×: j = l − 12 ; +: j = l + 12 .
216 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

Fig. 8. Phase-shift in Z-parameterisation (left) and comparison of ERE- and Z-parameterisation for k cot δ (right)
of the 4 S3/2 -wave as function of the cm-momentum k; Λ = 1000 MeV. Notation as in Figs. 7 and 4, and in
addition on the right: boxes/stars: partial wave analysis [38] (for points with error-bars as reported in [44]); dot at
k = 0: measured scattering length [29].

Table 3
The effective-range parameters (3.24) of the low partial waves in the N d-system from Z-parameterisation in the
first three orders; theoretical accuracy discussed in the text
Quartet-S a [fm] r0 [fm]
Order Z-param. ERE-param. Z-param.
LO 5.091 −0.1
NLO 6.410 5.938 1.95
N2 LO 6.354 ± 0.020 6.204 1.8 ± 0.1
Experiment [29] 6.35 ± 0.02
Modern N N -potentials [43,50] 6.34 . . . 6.35
N2 LO re-summed, cf. [18–20] 6.365 1.8

Doublet-P Doublet-D Quartet-P Quartet-D

Order κ0 [fm3 ] κ1 [fm−1 ] κ0 [fm5 ] κ0 [fm3 ] κ1 [fm−1 ] κ0 [fm5 ]
LO 33.1 −3.2 −268 −76.7 1.0 531
NLO 53.9 −2.1 −456 −139 0.45 894
N2 LO 53.3 ± 0.2 −2.1 −456 −140 0.43 ± 0.01 894
N2 LO re-summed −140 0.43 894

The EFT-calculation reproduces of course the well-known behaviour of the phase shifts
at small momenta, e.g., [28],
(l)
1 κ1 2  
k 2l+1 cot δ (l) → − (l)
+ k + O k 4 ∀k → 0, (3.24)
κ0 2

where the S-wave parameters κ0(0) and κ1(0) are the scattering length a and effective range
(1)
r0 , the P-wave parameter κ0 the scattering volume, etc. Table 3 lists parameters for low
partial waves of the Nd-system. The theoretical accuracy may, for example, be estimated
conservatively by Q ∼ γt /mπ ≈ 1/3 of the difference between the NLO- and N2 LO-result,
or by the difference to the N2 LO-result with re-summed effective-range corrections as in
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 217

Fig. 9. Real (left) and imaginary (right) part of the phase-shifts of the higher partial waves in the quartet channel in
Z-parameterisation as function of the cm-momentum k, Λ = 1000 MeV. Notation as in Fig. 7, and 2: j = l − 32 ;
♦: j = l + 32 for the 4 lj -wave.

(3.23), since all N2 LO-calculations must agree within the N2 LO-accuracy. The numerical
uncertainties are easily covered by these error-bars. In the doublet channel, no re-summed
results are given due to the non-trivial cut-structure induced by the spurious poles of the
two-nucleon propagators, see Section 3.4. Furthermore, no numbers are presented for the
218 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

2S
1/2 -channel as the measured scattering length (3.10) is used to determine the strength of
the three-body force, and the effective range is very large, ∼ 500 fm, rendering an effective-
range expansion useless for this partial wave. In the other partial waves, the parameters
involve the typical length scale ∼ 2 . . . 5 fm, so that an effective-range expansion is often
useful for momenta up to the deuteron break-up, kmax  20 . . . 52 MeV. More effective-
range parameters are obtained without difficulty, but they are less likely to be measurable
in the upcoming partial wave analysis of the nd-system [51].
The quartet–S-wave scattering length has drawn substantial interest recently as its
knowledge sets at present the experimental uncertainty in a recent, indirect determination
of the doublet scattering length (3.10) [52]. At N2 LO, the result of Z-parameterisation in
EFT(/ π ) (using a perturbatively expanded kernel which is the inserted into the integral equa-
tion), [6.35 ± 0.02] fm, agrees very well with experiment [29] and modern high-precision
potential-model calculations (which give aq = [6.34 . . .6.35] fm) [43,50], albeit partial-
wave mixing, iso-spin breaking and electro-magnetic effects are not present in EFT(/ π)
at N2 LO. aq is obviously to a very high degree sensitive only to the correct asymptotic
tail of the deuteron wave function, as comparison between the NLO result with its correct
deuteron residue, the N2 LO-result and the result of the re-summed deuteron propagator
shows. As the amplitude decays at large off-shell momenta as 1/p3.17... [19], this is not
surprising.
Comparing k cot δ in this partial wave at low energies in Fig. 8, the ERE-version
converges much slower than the result of Z-parameterisation. For example, the N2 LO-
scattering length differs from experiment and the Z-parameterisation result of the same
order still by 2%, and the correction from NLO to N2 LO is with 5% much larger than the
correction in Z-parametrisation. Indeed, partial N3 LO- and N4 LO-calculations in ERE-
parameterisation which include only effective-range effects bring one slowly closer to the
experimental number:
 
a 4 S ERE = (5.09
 + 
0.84 + 
0.27 + 
0.10 + 
0.04 + · · ·) fm ≈ 6.34 fm. (3.25)
LO NLO N2 LO “N3 LO” “N4 LO”

The re-summed effective-range result of EFT(/ π ) at N2 LO was first reported in [18,19].

The ERE-result at NLO in [20], 6.7 fm, was found not by re-summing the expanded kernel
as here, but by inserting the NLO-correction only once between the LO amplitudes. Thus,
convergence is also sped up by iterating the perturbed kernel.
EFT shares also the rather complex structure found even at low k, e.g., in the doublet-
and quartet-D- and -F-waves. Some disagreement with the potential-model results is seen
in some imaginary parts at rather high momenta k  120 MeV, close to the breakdown
scale of EFT(/ π ), where the expansions parameter Q is approaching 1. Given that the
imaginary parts are very small compared to the real parts and also most sensitive to de-
tails of the break-up reaction nd → nnp, the overall agreement is not dis-encouraging.
The most notable deviation occurs in the doublet–P-wave, where internal convergence of
the EFT-result is achieved up to more that k ≈ 100 MeV, while significant deviations from
potential-model calculations start around break-up. In this as in all other channels, cut-off
dependence is negligible above Λ ∼ 900 MeV (indicating also that a three-body force is
not needed for convergence, see [30]), and numerics is stable. The discrepancy—already
found in [22]—has therefore a physical origin. We may expect that it is cured by cancella-
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 219

tion with the well-known, rather strong P-wave interaction in the NN -system. It enters at
N3 LO, where the splitting between the partial waves will also be reproduced more realis-
tically. Clearly, more work is needed here.
While the NLO-corrections to the LO EFT-result are as expected rather large in Z-
parameterisation, the N2 LO-modifications are in the real parts of the phase-shifts tiny, and
do not exceed 40% of the NLO corrections in the imaginary parts even as k → Λπ/ ∼ mπ ,
the 2 P-wave being again the only notable exception. The main difference between the
strictly perturbative NLO-calculation of [22] which already utilised Z-parameterisation
and the partially re-summed version used here is that now the NLO and N2 LO-results
lie practically on top of each other, while they noticeably differ in strictly perturbative
Z-parameterisation above break-up. The difference does however never exceed 5%, and
hence lies within the accuracy of the NLO calculation. The N2 LO-results of [19,20,22]
using the fully re-summed deuteron and 1 S0 -propagators (3.23) are nearly in-discernible
from the partially re-summed version (2.19) in the real parts even at hight k. In the imag-
inary parts, they do usually not deviate by more than 10% of the correction from NLO to
N2 LO at k ∼ 140 MeV. This lies again well within the power-counting prediction that par-
tial inclusion of N3 LO-effects will not increase the accuracy of the N2 LO-result. It gives
however a band within which one expects a full N3 LO-calculation to lie.
Finally, an unsolved technical issue has to be mentioned. The problem that the spurious
poles of the re-summed two-body propagators induce additional cuts in the solution of
the integral equation in the doublet channel, was already discussed in Section 3.4. Such
breaches of unitarity below break-up, induced by naïve integration over a cut at high off-
shell momenta, are the reason why the phase-shifts of the doublet-P- and doublet-D-wave
have a small non-zero value at the break-up point. As the half-off-shell amplitudes of higher
partial waves converge substantially faster than of the doublet-S-wave, these violations
become however less and less noticeable. For the higher partial waves, they disappear in
the numerical noise.
To summarise, the overall convergence of the phase-shifts is good in all channels, both
internally and to the available potential-model calculations. The only noticeable excep-
tion is the doublet–P-wave, in which the radius of convergence is limited to momenta k
below break-up. Neglecting partial-wave splitting, simple observables like the differential
cross-section do therefore agree with potential-model results and direct experimental mea-
surements, within the level of accuracy of the EFT-prediction. As an illustration, the elastic
differential cross-section is shown in Fig. 10 for three neutron energies: just below the nd-
break-up (k = 50 MeV), and at two momenta close to the breakdown scale of EFT(/ π ),
k ≈ 90; 120 MeV. The accuracy of the calculation is estimated by varying the N2 LO-
correction to k cot δ by Q ∼ γt /mπ ≈ 1/3 around its central value. The agreement with
experiments [53,54] and potential-model calculations [39,55] is quite satisfactory even
close to the breakdown scale k ∼ mπ . The main source of deviation is of course the in-
sufficient description of the doublet-P-wave.
More complex variables need of course more refinement. For example, one obtains a
zero-result for the nucleon–deuteron vector analysing power Ay in EFT(/ π ) at N2 LO be-
cause partial-wave splitting is absent. The well-known under-prediction of this observable
in all modern, high-precision NN -potentials [2]—even when supplemented with three-
body forces—is a challenge a higher-order EFT-calculation has to meet.
220 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

Fig. 10. Elastic differential cross-sections of nd-scattering at N2 LO in the cm-frame in Z-parameterisation

(solid) with estimated accuracy (band). Dot-dashed: AV18 + Urbana IX from [39] at neutron kinetic energy
Elab = 3 MeV, AV18 from [55] at Elab = 10; 18 MeV; diamonds: nd-data [53] (taken at 3; 10.25; 18 MeV);
stars: pd-data [54].
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 221

4. Conclusions and outlook

Z-parameterisation [16] fixes the parameters of “pion-less” effective field theory at very
low energies up to N2 LO, i.e., up to an estimated theoretical accuracy of  3%, to the
correct deuteron pole position and residue, instead of the effective-range parameters. As
the asymptotic fall-off and normalisation of the deuteron wave-function at large distances
is correctly reproduced, this improves drastically convergence in all processes in which the
deuteron is either found as in- or out-state, or as sub-cluster of a more complex few-body
system. The reason for the improvement is found in a rather large expansion parameter of
γt ρ0t ≈ 0.4 for effective-range corrections, so that the deuteron residue is regained only up
to 7% at N2 LO when fitting to the effective range, while it is exact already at NLO in Z-
parameterisation. Thus, the speed of convergence is faster, while the radius of convergence,
set by the first scale on which new physics enters, is of course not changed.
In this article, Z-parameterisation was extended and applied to deuteron–nucleon scat-
tering, see Sections 2 and 3.1. To perform the calculations, a computationally simple and
convenient scheme was again employed which first expands the kernel to the desired order
of accuracy in the power counting and then iterates it to all orders [14].
Numerical problems which origin from additional, spurious poles in the NN -amplitudes
when the effective-range corrections are re-summed to all orders in the deuteron prop-
agator are avoided in Z-parameterisation, Section 3.4. Therefore, standard techniques
are readily applicable, and the calculation of scattering-observables also above break-up
becomes computationally trivial, see a M ATHEMATICA code at http://www.physik.tu-
muenchen.de/~hgrie.
In all partial waves, the agreement of the phase-shifts with sophisticated potential-model
calculations is increased, in particular, in the real and imaginary parts above the deuteron
break-up point, see Section 3.5. What is more, the corrections from one order to the next
in the EFT(/ π )-expansion are smaller in Z-parameterisation, so that a partial re-summation
of effective-range effects increases indeed the theoretical accuracy of the calculation. This
allowed also for high-accuracy predictions of effective-range parameters in the low partial
waves of the nd-system, which can be compared to an upcoming partial-wave analysis
of the nd-system [51]. Most notably, the quartet–S-wave scattering length is predicted as
aq = [6.35 ± 0.02] fm at N2 LO, using only four observables from the NN -system as in-
put, namely the asymptotic form and normalisation of the deuteron and virtual 1 S0 -bound
state. This result compares both in magnitude and uncertainty favourably with the most ad-
vanced potential-model calculations ([6.34 . . .6.35] fm [43,50]), and with the experimental
value ([6.35 ± 0.02] fm [29]). Simple observables like differential cross-sections are well
reproduced as high as Elab ∼ 15 MeV.
The most striking success lies however in the 2 S1/2 -channel, which contains the triton
as real bound state, Section 3.3. The numerical analysis of the cut-off dependence of its
phase-shift in Z-parameterisation clearly supports the analytic finding in Ref. [14] that one
and only one momentum-dependent three-body force enters at N2 LO, namely the Wigner-
SU(4)-symmetric one with two derivatives. Fixing its strength to the triton binding energy,
the accuracy of the calculation—deduced from the cut-off dependence of the answer—is
improved at N2 LO from ∼ 4% to ∼ 1% at a momentum scale of ∼ 90 MeV typical for
the three-body system. The cut-off dependence follows in Z-parameterisation the pattern
222 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

predicted by the power-counting. The difference to potential-model calculations is also

substantially diminished.
The major source of uncertainty in the triton channel stems thus not from the theo-
retical accuracy of the EFT(/ π )-calculation, but from the error in the determination of
its nd-scattering length, which is known only to ∼ 7% accuracy. To reduce this un-
comfortably large error bar to an accuracy of ∼ 0.7% is the goal of an ongoing direct
measurement of the incoherent scattering length at the Paul Scherer Institute [56]. A re-
cent measurement of the bound coherent scattering length [52] relied for the extraction
of ad = [0.645 ± 0.003(exp) ± 0.007 (theor)] fm on theoretical input for the 4 S3/2 -wave
scattering length [50].
Future work includes a complete N3 LO-calculation, including iso-spin breaking ef-
fects and partial-wave splitting and mixing (which should improve in particular the
doublet-P-waves), the extension to include Coulomb interactions for the 3 He- and pd-
system, and the coupling of electro-weak probes to the three-nucleon system. Not only
will this allow for direct comparison with a cornucopia of data and help to shed light on
long-standing puzzles like the Ay -problem. It will also lead to predictions with an accuracy
relevant for nuclear astro-physics and neutrino physics at very low energies, contributing
to the ongoing efforts to improve our knowledge about big-bang nucleo-synthesis, stellar
evolution, neutrino-mass determinations and other fundamental processes of the Standard
Model.

Acknowledgements

It is my pleasure to thank P.F. Bedaque, H.-W. Hammer, D.R. Phillips and G. Rupak
for intense discussions and encouragement, and A. Kievsky and T.C. Black for communi-
cations about their results. N. Kaiser and W. Weise provided critical companionship. The
warm hospitality and financial support for stays at the Nuclear Theory Group of Lawrence
Berkeley National Laboratory, at the INT in Seattle and at the ECT* in Trento was instru-
mental for this research. In particular, I am grateful to the organisers and participants of
the “Berkeley Visitors Programme on Effective Field Theories 2003” and of the “INT Pro-
gramme 03-3: Theories of Nuclear Forces and Nuclear Systems”. This work was supported
in part by the Bundesministerium für Forschung und Technologie, and by the Deutsche
Forschungsgemeinschaft under contracts GR1887/2-1, 2-2 and 3-1.

Appendix A. Deriving the Faddeev equation

The Faddeev integral equation in the kinematics defined by Fig. 11 is

 j B bβ  
t iA aα (E; k,  p ) = K j B iA bβ aα (E; k, p )

d q  j B bβ
3
− K lC cγ (E; q , p )
(2π)3
 
q 2  cγ
×D E− , q t lC iA aα (E; q , p ), (A.1)
2M
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 223

Fig. 11. The Faddeev equation for N d-scattering to N2 LO. Thick solid line: propagator of the two intermediate
auxiliary fields ds and dt , denoted by D, see (3.6); triple line: propagator of the triton auxiliary field t, see (3.17);
K: interaction to N2 LO.

where E = (3k2 /(4M)) − (γt2 /M) is the total non-relativistic energy in the cm-frame, k
 the momentum of the incoming (outgoing) deuteron/spin-zero field with spin–isospin
(p)
indices i/A (j/B), and the incoming (outgoing) nucleon has spin–isospin αa (βb). The
integration over the loop-energy was already performed, setting q0 = q 2 /2M under the
assumption that no additional poles above the real q0 -axis are hidden in the interaction K.
Sub-scripts (super-scripts) denote quantum numbers for incoming (outgoing) particles.
One decomposes now the integral equation into the pertinent spin–isospin and angular-
momentum channels.

A.1. Projectors

As two cluster-configurations exist, namely Ndt and Nds , it is convenient to decompose

each operator (interactions, amplitudes, etc.) as
 bβ  
†  † † 
j j
O(Ndt → Ndt )i O(Nds → Ndt )A dti
O = Nbβ dt,j , ds,B N aα .
O(Ndt → Nds )B
i O(Nds → Nds )B
A aα dsA
(A.2)

In other words, each operator is represented in cluster-configuration space by a (2 × 2)-

matrix which in addition carries spin- and iso-spin indices. It is understood in the following
that the spin–isospin indices on the (2 × 2)-matrix are applied to each entry separately,
and that Kronecker-δ’s in spin or iso-spin are not displayed. Thus, for example, an entry
O(Ndt → Ndt ) = σi σ j in the above matrix is written out as
 j bβ  
j β b
O(Ndt → Ndt )i aα = σi σ α δa . (A.3)

O is usually not symmetric in any pair of spin–isospin indices (ij ), (AB), (αβ), (ab).
224 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

The projection onto a state with angular momentum l, connecting momenta q , p with
q · p = pq cos θ and p = |p | is as usual given by
1
1
O (q, p) =
(l)
d cos θ O(
q , p ). (A.4)
2
−1
Finally, the projectors onto the possible spin-isospin states of the three-nucleon system
are: combining the auxiliary fields with spin-index i or iso-spin index A with a nucleon of
spin-isospin α, a into the spin-doublet channel with spin µ and iso-spin m
 
1 σi 0 mµ
(Pd,iA )mµaα = √ , (A.5)
3 0 τA aα
and into the spin-quartet channel with spin (µj ) and iso-spin m
 j 1 j mµ
 j mµ δi − 3 σ σi 0
Pq,i aα
= . (A.6)
0 0 aα
The projectors are related to their Hermitean conjugates by
 j †
PdiA = (Pd,iA )† = “Pd,iA ”, Pq,i = Pq,j
i
, (A.7)
and are ortho-normalised:
j
Pd,iA PdiA = 1, Pq,j
k
Pq,i = Pq,i
k
, Pd,iA Pq,j
i
= 0. (A.8)
For a complete set of spin–isospin projectors, supplement these by the (ortho-normalised)
projector onto the iso-spin quartet channel with iso-spin (mB) and spin µ, which is not
found in nucleon–deuteron scattering:
 mµ
 B mµ 0 0  B †
Piso-q,A aα = B − 1τBτ , Piso-q,A = PisoA
-q,B . (A.9)
0 δA 3 A aα

A.2. Projecting the interaction terms

The exchange of a nucleon is from the Lagrangeans (2.1)/(2.22)/(2.2) up to N2 LO:

 jB bβ
KN -ex iA aα (E; q , p )
 bβ
−My 2 1 σi σ j σ j τA
= . (A.10)
2 p2 + q 2 + p · q − ME − i σi τ B τA τ B aα
The projection of the propagator of the exchanged nucleon onto angular momentum l is
obtained by combining (A.4) with (3.2). There is no mixture between the doublet- and
quartet-channels, and one finds in the doublet-channel
   
σi σ j σ j τA −1 3
Pd,j B Pd =
iA
(A.11)
σi τ B τA τ B 3 −1
and in the quartet
 
σi σ j σ j τA
Pq,j
l
Pq,k
i
= 2Pq,k
l
. (A.12)
σi τ B τA τ B
 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226 225

In cluster-configuration space, one obtains for the three-body interaction from (3.16)
 j B bβ
KH iA aα
(E; q , p )
 ∞   j bβ
−My32 (Λ) h2n (Λ)   m
2 n σ σi −σ j τA
= 1+ ME + γt .
3MΩ ΩM n −σi τ B τ B τA aα
m,n=1
(A.13)
By construction, the only non-zero contribution is in the doublet-S-wave:
 j   
σ σi −σ j τA 1 −1
Pd,j B P iA
= 3 . (A.14)
−σi τ B τ B τA d −1 1

A.3. Result

 × 2)-matrix
Putting these results together and multiplying in the doublet-case the (2
t (l) in cluster-configuration space from the right with the column vector 10 , one projects
finally onto the nucleon–deuteron system. There is no mixture or breaking between in- or
out-states of different individual nucleon spin and iso-spin. One arrives thus finally at the
integral equations for the quartet (3.1) and doublet (3.4) channels quoted in the main text.

References

[1] G.E. Moore, Electronics 38 (8) (1965).

[2] See, e.g., D. Hüber, J.L. Friar, Phys. Rev. C 58 (1998) 674, nucl-th/9803038.
[3] J.W. Chen, G. Rupak, M.J. Savage, Nucl. Phys. A 653 (1999) 386, nucl-th/9902056.
[4] J. Schwinger, Hectographed Notes on Nuclear Physics, Harvard University, 1947;
G.F. Chew, M.L. Goldberger, Phys. Rev. 75 (1949) 1637;
F.C. Barker, R.E. Peierls, Phys. Rev. 75 (1949) 3122;
H.A. Bethe, Phys. Rev. 76 (1949) 38.
[5] V. Efimov, Nucl. Phys. A 362 (1981) 45;
V. Efimov, Phys. Rev. C 44 (1991) 2303;
V. Efimov, E.G. Tkachenko, Phys. Lett. B 157 (1985) 108.
[6] S. Weinberg, Nucl. Phys. B 363 (1991) 3.
[7] See, e.g., C. Ordoñez, L. Ray, U. van Kolck, Phys. Rev. C 53 (1996) 2086, hep-ph/9511380.
[8] See, e.g., E. Epelbaum, nucl-th/0309019, and references therein.
[9] U. van Kolck, Prog. Part. Nucl. Phys. 43 (1999) 337, nucl-th/9902015.
[10] S.R. Beane, P.F. Bedaque, W.C. Haxton, D.R. Phillips, M.J. Savage, in: M. Shifman (Ed.), At the Frontier
of Particle Physics, World Scientific, Singapore, 2001, nucl-th/0008064.
[11] P.F. Bedaque, U. van Kolck, Annu. Rev. Nucl. Part. Sci. 52 (2002) 339, nucl-th/0203055.
[12] P.F. Bedaque, H.-W. Hammer, U. van Kolck, Phys. Rev. Lett. 82 (1999) 463, nucl-th/9809025;
P.F. Bedaque, H.-W. Hammer, U. van Kolck, Nucl. Phys. A 646 (1999) 444, nucl-th/9811046.
[13] P.F. Bedaque, H.-W. Hammer, U. van Kolck, Nucl. Phys. A 676 (2000) 357, nucl-th/9906032.
[14] P.F. Bedaque, G. Rupak, H.W. Grießhammer, H.-W. Hammer, Nucl. Phys. A 714 (2003) 589, nucl-
th/0207034.
[15] T. Barford, PhD Thesis, Manchester University, 2004 nucl-th/0404072;
T. Barford, M.C. Birse, nucl-th/0406008.
[16] D.R. Phillips, G. Rupak, M.J. Savage, Phys. Lett. B 473 (2000) 209, nucl-th/9908054.
[17] In EFT, this was first considered by D.B. Kaplan, Nucl. Phys. B 494 (1997) 471, nucl-th/9610052.
226 H.W. Grießhammer / Nuclear Physics A 744 (2004) 192–226

[18] P.F. Bedaque, U. van Kolck, Phys. Lett. B 428 (1998) 221, nucl-th/9710073.
[19] P.F. Bedaque, H.-W. Hammer, U. van Kolck, Phys. Rev. C 58 (1998) R641, nucl-th/9802057.
[20] P.F. Bedaque, H.W. Grießhammer, Nucl. Phys. A 671 (2000) 357, nucl-th/9907077.
[21] S.R. Beane, M.J. Savage, Nucl. Phys. A 694 (2001) 511, nucl-th/0011067.
[22] F. Gabbiani, P.F. Bedaque, H.W. Grießhammer, Nucl. Phys. A 675 (2000) 601, nucl-th/9911034.
[23] D.B. Kaplan, M.J. Savage, M.B. Wise, Nucl. Phys. B 534 (1998) 329, nucl-th/9802075.
[24] G.V. Skorniakov, K.A. Ter-Martirosian, Sov. Phys. JETP 4 (1957) 648.
[25] I.S. Gradshteyn, I.M. Ryzhik, Table of Integrals, Series and Products, fifth ed., Academic Press, San Diego,
CA, 1994.
[26] H.-W. Hammer, T. Mehen, Phys. Lett. B 516 (2001) 353, nucl-th/0105072.
[27] J.H. Hetherington, L.H. Schick, Phys. Rev. B 137 (1965) 935;
R.T. Cahill, I.H. Sloan, Nucl. Phys. A 165 (1971) 161;
R. Aaron, R.D. Amado, Phys. Rev. 150 (1966) 857;
E.W. Schmid, H. Ziegelmann, The Quantum Mechanical Three-Body Problem, Vieweg Tracts in Pure and
Applied Physics, vol. 2, Pergamon Press, 1974.
[28] M.L. Goldberger, K.M. Watson, Collision Theory, Wiley, New York, 1964.
[29] W. Dilg, L. Koester, W. Nistler, Phys. Lett. B 36 (1971) 208.
[30] H.W. Grießhammer, in: U.-G. Meißner, H.-W. Hammer, A. Wirzba (Eds.), Mini-Proceedings of Chiral Dy-
namics: Theory and Experiment, 2003, hep-ph/0311212;
H.W. Grießhammer, in preparation.
[31] R.A. Minlos, L.D. Faddeev, Sov. Phys. JETP 14 (1962) 1315.
[32] L.W. Thomas, Phys. Rev. 47 (1935) 903.
[33] G.S. Danilov, Sov. Phys. JETP 13 (1961) 349.
[34] K.G. Wilson, Phys. Rev. D 3 (1971) 1818;
S.D. Głazek, K.G. Wilson, Phys. Rev. D 47 (1993) 4657;
S.D. Głazek, K.G. Wilson, Phys. Rev. Lett. 89 (2002) 230401, hep-th/0203088.
[35] E. Braaten, H.W. Hammer, Phys. Rev. Lett. 91 (2003) 102002, nucl-th/0303038.
[36] E. Wigner, Phys. Rev. 51 (1937) 106.
[37] I.R. Afnan, D.R. Phillips, Phys. Rev. C 69 (2004) 034010, nucl-th/0312021.
[38] W.T.H. van Oers, J.D. Seagrave, Phys. Lett. B 24 (1967) 562.
[39] A. Kievsky, S. Rosati, W. Tornow, M. Viviani, Nucl. Phys. A 607 (1996) 402.
[40] A. Kievsky, private communication.
[41] G.P. Lepage, How to renormalize the Schrödinger equation, in: Lectures given at 9th Jorge Andre Swieca
Summer School: Particles and Fields, Sao Paulo, Brazil, 16–28 February 1997, nucl-th/9706029.
[42] V. Efimov, E.G. Tkachenko, Few-Body Systems 4 (1988) 71.
[43] H. Witała, A. Nogga, H. Kamada, W. Glöckle, J. Golak, R. Skibinski, nucl-th/0305028.
[44] A.C. Phillips, G. Barton, Phys. Lett. B 28 (1969) 378.
[45] G. Rupak, X.W. Kong, Nucl. Phys. A 717 (2003) 73, nucl-th/0108059.
[46] F. Gabbiani, nucl-th/0104088.
[47] H.W. Grießhammer, in preparation.
[48] D. Hüber, J. Golak, H. Witała, W. Glöckle, H. Kamada, Few-Body Systems 19 (1995) 175.
[49] W. Glöckle, H. Witała, D. Hüber, H. Kamada, J. Golak, Phys. Rep. 274 (1996) 107.
[50] J.L. Friar, D. Hüber, H. Witała, G.L. Payne, Acta Phys. Pol. B 31 (2000) 749, nucl-th/9908058.
[51] T.C. Black, et al., in preparation.
[52] T.C. Black, et al., Phys. Rev. Lett. 90 (2003) 192502;
K. Schön, et al., Phys. Rev. C 67 (2003) 044005.
[53] P. Schwarz, et al., Nucl. Phys. A 398 (1983) 1.
[54] K. Sagara, et al., Phys. Rev. C 50 (1994) 576.
[55] A. Kievsky, M. Viviani, S. Rosati, Phys. Rev. C 64 (2001) 024002, nucl-th/0103058.
[56] B. van den Brandt, H. Glättli, H.W. Grießhammer, P. Hautle, J. Kohlbrecher, J.A. Konter, O. Zimmer, Nucl.
Instrum. Methods A 526 (2004) 91, nucl-ex/0401029.