Annals of Physics 325 (2010) 265–286

Contents lists available at ScienceDirect

Annals of Physics
journal homepage: www.elsevier.com/locate/aop

Numerical studies of a model fermion–boson system

T. Cheng, E.R. Gospodarczyk, Q. Su, R. Grobe *
Intense Laser Physics Theory Unit, Department of Physics, Illinois State University, Normal, IL 61790-4560, USA

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

Article history: We study the spectral and dynamical properties of a simplified

Received 28 April 2009 model system of interacting fermions and bosons. The spatial dis-
Accepted 3 September 2009 cretization and an effective truncation of the Hilbert space permit
Available online 11 September 2009
us to compute the distribution of the bare fermions and bosons in
the energy eigenstates of the coupled system. These states repre-
Keywords:
sent the physical particles and are used to examine the validity
Yukawa model system
Interacting boson–fermion model
of the analytical predictions by perturbation theory and by the
Numerical approach to quantum field Greenberg–Schweber approximation that assumes all fermions
theory are at rest. As an example of our numerical framework, we exam-
Space–time resolution for a model system ine how a bare electron can trigger the creation of a cloud of virtual
Virtual bosons bosons around. We relate this cloud to the properties of the asso-
Greenberg–Schweber approximation ciated energy eigenstates.
Ó 2009 Elsevier Inc. All rights reserved.

1. Introduction

Quantum electrodynamics is the fundamental description of all electromagnetic properties and the
associated photon [1–5] and electron [6] creation and annihilation processes. Using a variety of ana-
lytical and computational approaches often based on perturbative techniques for the S-matrix theory
[7], many accurate calculations for the energies of atomic sublevels, the scattering cross-sections, the
life times of metastable bound states, the spin and the magnetic moments of various particles have
been made. The predictions for these time-independent quantities have been confirmed experimen-
tally with astonishing accuracy. Furthermore, many difficulties associated with singularities and inter-
pretational problems of quantum field theories have been addressed.
In addition to these highly accurate calculations for static properties, quantum field theory pro-
vides also the framework to explore the temporal evolution of quantum systems. Due to many concep-
tual and especially computational problems, however, this aspect of the theory has received much less
attention. We are not aware of any studies that use interacting quantum field theories to compute the

* Corresponding author.
E-mail address: [email protected] (R. Grobe).

0003-4916/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.aop.2009.09.007
266 T. Cheng et al. / Annals of Physics 325 (2010) 265–286

evolution with space–time resolution. For the special case where the bosonic dynamical degrees of
freedom can be decoupled, some time-resolved studies of fermion creation have been reported re-
cently [8,9]. Here the back reaction of the fermions on the bosons has been neglected and the vector
potential has been approximated by an external field with a given space–time dependence [6].
Due to the technical and also conceptual difficulties, unambiguous time-dependent data that could
be used to either confirm or disprove theoretical predictions for the creation of particles are presently
lacking. To the best of our knowledge all experimental confirmations of interactions between elemen-
tary particles are based on their asymptotic properties such as cross-sections, energies and bound
state lifetimes. Furthermore, even from a theoretical point of view, the question of obtaining an unam-
biguous time evolution is by no means trivial. Calculations for the traditional S-operator are usually
much easier than those for the full time evolution and yet they fully satisfy the needs of the current
experiments in high energy physics. This situation could create the impression that a comprehensive
theory can be constructed that uses the S-operator as the fundamental quantity rather than the Ham-
iltonian and wave functions [10–12]. However, an S-matrix based theory connects only asymptotic
states from the infinite past to the infinite future and is therefore not complete to describe the time
evolution for all time intervals. If one knows the full Hamiltonian, the S-operator can be calculated
unambiguously, however, the reverse is not true as the same S-operator can be obtained from different
Hamiltonians. These Hamiltonians are called scattering equivalent in the sense that they have the
same energy spectrum. However, their eigenvectors are different and correspondingly the predicted
dynamics can be different, too. Therefore, scattering-equivalent theories may not be physically equiv-
alent [10–12]. A detailed account of the scattering equivalence and their relation to the point and in-
stant forms can be found in the excellent monograph by Stefanovich [12]. A trivial example where
scattering-equivalent Hamiltonians with different eigenvectors lead to different time-dependent pre-
dictions for the same initial state can be found if an atom is ionized by an external oscillatory electric
field of finite duration. The two corresponding non-relativistic Hamiltonians in the so-called pA and dE
gauges are unitarily equivalent under the dipole approximation, but the time-dependence of the
bound state population is different during the times the external field is turned on. These conceptual
problems are also related to the question whether one should use dressed and or bare bound states to
compute the appropriate time-dependent ionization probabilities during the interaction with the
external field. It is often not at all clear which of the associated Hamiltonians is the correct one to pre-
dict the true physical time evolution.
It is therefore relevant to examine these Hamiltonians with regard to their predicted time evolu-
tion. Following the approach of our previous works, we plan to obtain some first qualitative insight
into several fundamental questions by providing temporally and spatially resolved numerical solu-
tions. As an example, one of our long term goals is to provide a better microscopic picture of how
the repulsive as well as attractive Coulombic forces between charged particles such as electrons or
positrons can be visualized in terms of the exchange of bosons. The questions about the locality
and possible instantaneity of these interactions is not very well understood [12]. These questions
may also become important with regard to the controversial issue of the possibility of a spontaneous
breakdown of the electronic vacuum induced by a supercritical electric force field. This fundamental
process has been predicted in 1951 by Schwinger [13] but has not been confirmed experimentally
[6,14,15]. It is presently also not known how this electron–positron creation process would be affected
by the interfermionic (Coulombic) interactions that have been omitted in all studies so far.
On the one hand, addressing these fundamental questions requires the full theoretical framework
of quantum field theory, but on the other hand its usual accuracy is not really necessary for qualitative
explorations. This advantage permits us to introduce some approximations to make a computational
approach feasible. With this unconventional ‘‘qualitative” approach to quantum field theory we can
also address several ‘‘what-if” questions. For example, even though the numerical values for the par-
ticle masses are given by the specific value of the coupling constant, a different choice of this param-
eter might enhance and therefore reveal new phenomena, that exist for the ‘‘correct” parameters but
are harder to be observed.
Our unusual approach of using the framework of an – in principle – highly accurate theory to make
only phenomenological predictions based on numerous approximations for simplified model systems
might seem strange at first. For example, in order to make the theory of interacting quantum fields
 T. Cheng et al. / Annals of Physics 325 (2010) 265–286 267

computationally feasible, we will reduce the spatial dimension to only one. This reduction is similar to
an approach introduced in the 1980s in ionization physics [16,17]. Model based ab initio studies have
led in the past to first qualitative insights into several quantum mechanical phenomena such as the
occurrence of quantum collapse and revivals in quantum optics [18] or atomic stabilization in ioniza-
tion physics [19–22]. It is possible and unfortunately likely that the specific numerical values of these
simplified models vary significantly from experimental values and that the interactions might show
an unphysical and therefore purely mathematical scaling behavior. However, as our expertise and
computational power grow, we assume that these model simplifications can be gradually removed.
The goal of this paper is to lay the foundation for future numerical work. We will examine how the
time-dependent quantum field theory can be converted into a numerically tractable problem and
show how various numerical parameters can be chosen to lead to convergent results. As the easiest
model system to examine interacting quantum fields we study the scalar Yukawa theory that we re-
view in Section 2. This theory was used originally to model phenomenologically the interaction be-
tween nucleons and mesons [23]. In Section 3 we discretize this model based on several numerical
parameters, outline the numerical complexity and discuss those parameters that determine the
numerical convergence. In Section 4 we review an interesting limiting case of non-moving fermions
for which several analytical solutions can be found. These analytical solutions can be used to test
the numerical convergence for special cases. In addition the numerical data from moving fermions
can be used to examine the validity of these limiting cases. In Section 5 we show how a bare fermion
in vacuum can generate an oscillating cloud of bosons around it, a process that is sometimes called
virtual boson creation. In Section 6 we illustrate the spatial densities of the created virtual bosons
on a numerical grid. We summarize our findings in Section 7 concluding with an outlook on future
work.

2. The model system

We denote the quantum field operators for the spinless fermions (modeling electrons) and neutral
scalar bosons (modeling photons) with W b ðxÞ and u^ ðxÞ, where x denotes the (one-dimensional) spatial
coordinate. As customary in atomic physics and quantum optics, we will use from now on atomic units
with the speed of light c = 137 a.u., the electron’s mass and charge m = e = 1 a.u. and 
h = 1 a.u. We can
expand these operators in momentum states
Z
b ðxÞ 
W ^
dp bðpÞð2pÞ1=2 expðipxÞ ð2:1aÞ
Z
u
^ ðxÞ  dk f ðkÞcð2xðkÞÞ1=2 a
^ðkÞð2pÞ1=2 expðikxÞ þ h:c: ð2:1bÞ

where the fermionic and bosonic creation and annihilation operators fulfill the anticommutator and
h i h i
^
commutator relationships, bðpÞ; ^ 0 Þy ¼ dðp  p0 Þ and a
bðp ^ðk0 Þy ¼ dðk  k0 Þ. The unitless func-
^ðkÞ; a
þ 
tion f ðkÞ has been introduced to cut off high bosonic momenta in order to avoid any divergencies.
It is also related to the effective spatial extension of the fermion [23]. The interaction between the fer-
R
mions and bosons is given by the interaction part of the Hamiltonian, V  cc3=2 dx W b ðxÞu
b ðxÞy W ^ ðxÞ,
where we have included the factor c3=2 to make the coupling constant c unitless. The full Hamiltonian
to describe the fermion–boson dynamics is given by
Z Z
H¼ ^ y bðpÞ
dp EðpÞbðpÞ ^ ^ðkÞy a
þ dk f ðkÞ xðkÞ a ^ðkÞ þ V ð2:2aÞ
Z Z  
^ þ kÞy bðpÞ
V ¼ cc5=2 dp dk f ðkÞ½2p2xðkÞ1=2 bðp ^ a ^ðkÞy
^ðkÞ þ a ð2:2bÞ

p p 2
Here we denote the free energies by EðpÞ  ½c4 þ c2 p2  and xðkÞ  ½m2 c4 þ c2 k , where the bare
masses of the fermion and boson are 1 a.u. and m, respectively. This model Hamiltonian has been
introduced by Yukawa to study phenomenologically the nuclear interaction of nucleons with mesons
such as the pions.
268 T. Cheng et al. / Annals of Physics 325 (2010) 265–286

The dynamics is restricted due to the existence of two operators that commute with the Hamilto-
R ^ y^
nian, associated with the conservation of the total number of fermions, dpbðpÞ bðpÞ, and total
R ^ y ^ R y
momentum, dppbðpÞ bðpÞ ^ðkÞ a
þ dkf ðkÞka ^ðkÞ. These operators lead to invariant subspaces that
can reduce the required computer memory significantly as we will discuss below.
We should point out that due to the simple direct sum structure of the Hamiltonian of Eq. (2.2) for
c ¼ 0, its eigenstates provide a convenient set of basis states for our numerical analysis. However, the
physical meaning of these states is not clear as they do not take any interaction into account. They
describe what we would denote as bare particles. If we require that physical particle have a sharp en-
ergy, then they have to be described by the eigenvalues and eigenstates of H(c) for a nonzero value of
c. For our particular system, it turns out that there are invariant states for which neither the energy
nor the form of the states depend on c. These states are characterized by the property bðpÞk ^ Uii ¼ 0
for each value of p. These states include the vacuum state and states for which only the bosonic (bare)
energy levels are occupied. In other words, the states that describe zero fermions and one or more bare
bosons do not depend on c.
As we will discuss below and illustrate in a concrete analytical example in Section 4 for a simplified
model, this observation is not in contradiction with the fact that the corresponding creation and anni-
hilation operators for physical bosons actually do depend on c. In fact, each of the corresponding oper-
^
ators for physical fermions and bosons, denoted by BðpÞ ^
and AðkÞ, are complicated functions of the set
^
of all (bare) operators bðpÞ ^ðkÞ for all p and k. In order to continue to classify the physical particles
and a
as fermions and bosons we require the anticommutator and commutator relationships
h i h i
b
BðpÞ; b
b 0 Þy ¼ dðp  p0 Þ and AðkÞ;
Bðp b 0 Þy ¼ dðk  k0 Þ, and any fermionic and bosonic operators to
Aðk
þ 
^ðkÞy as well
commute. As a side remark, we note that while the application of both the bare operator a
^
as physical boson creation operator AðkÞy
on the vacuum state generates a physical boson, the two
operators a ^ðkÞy and AðkÞ ^
^ y differ by terms related to bðpÞ, whose action on the vacuum state vanishes.
A state with one physical fermion and no boson could be interpreted as a superposition of several
bare fermion states with various momenta and dressed by a cloud of bare bosons, but we consider this
particular viewpoint as less helpful. This perspective is especially confusing as in the absence of any
fermion the quantum states for bare and physical bosons are identical, even though the associated
operators, a b
^ðkÞ and AðkÞ, are different and agree only for c ¼ 0.
In order to examine the dynamics, we have to compute the time evolution of the quantum field
theoretical state kUðtÞii as a solution to the equation of motion, iokUðtÞii=ot ¼ HkUðtÞii for a given ini-
tial state kUð0Þii. The time evolved state will be used to calculate the spatial probability density of the
(bare) fermions via qðxÞ  hhUk W b ðxÞkUii and of the (bare) photons via qðyÞ  hhUku
b ðxÞy W ^ ðyÞy u
^ ðyÞkUii.
Another quantity of interest is the mixed density (correlation function) qðx; yÞ  hhUk W b ðxÞy Wb ðxÞ
^ ðyÞy u
u ^ ðyÞkUii. For example, qðx ¼ 0; yÞ is the probability to find a (bare) boson at a distance y from
the fermion if it is at location x = 0.

3. Computational considerations

3.1. The numerical Hamiltonian

To convert the continuous world of quantum field theory into a numerically treatable discrete and
finite form we introduce five independent numerical parameters L; N x ; P; K; and N. The particles are
confined spatially to a numerical box from L/2 to L/2 in which we discretize the spatial coordinate into
Nx points, corresponding to a grid spacing of Dx ¼ L=N x . As a result, the momenta are discretized with a
spacing Dp ¼ Dk ¼ 2p=L  D. The two integers P and K represent the largest fermion and boson mo-
R
menta, P D and K D, where P; K 6 ðN x  1Þ=2. We can therefore replace the momentum integrals dp
PP R PK
by D p¼P  DRp and dkf ðkÞ with D k¼K  DRk , assuming f ðkÞ ¼ 1 for jkj 6 K and f ðkÞ ¼ 0 other-
wise. The corresponding numerical annihilation operators are defined as ^p 
b
p ^ p
ðDÞbðpÞ and a^k  ðDÞa ^ðkÞ where in contrast to the continuous arguments k and p, the subscripts k
and p are now integers extending from P to P and K to K, respectively. These operators are automat-
 T. Cheng et al. / Annals of Physics 325 (2010) 265–286 269

h i h i
ically normalized based on the Kronecker symbols, b^p ; b
^y ¼ dpp0 and a ^y 0 ¼ dkk0 . Under these
^k ; a
p0 k
þ 
substitutions, the discretized version of the Hamiltonian of Eq. (2.2) becomes
^y b
^ ^y ^ 1=2 ^ y ^ p a ^yk

p p þ Rk xk ak ak þ g Rp Rk ð2xk Þ
H ¼ Rp Ep b bpþk b ^k þ a ð3:1Þ
1=2 5=2
where the new coupling constant g  L c c needs to be rescaled to adjust for various box sizes L to
predict universal data that are independent of the particular choices for L. While the fermionic occu-
pation numbers np can be only 0 or 1, the occupation numbers of the bosonic modes nk range from 0 to
N.
The dimension of the specific subspace with zero bosons (N = 0) and exactly j fermions is given by
the binomial coefficient Binð2P þ 1; jÞ. This corresponds to the one-dimensional vacuum for
j ¼ 0; ð2P þ 1Þ dimensions for the single-particle space for j = 1, and Pð2P þ 1Þ dimensions for the
P2Pþ1
two-particle space with j = 2. Obviously, for consistency we have j¼0 Binð2P þ 1; jÞ ¼ 2ð2Pþ1Þ , reflect-
ing that each of the 2P + 1 fermionic degrees has two possible occupation numbers.
The numerical Hilbert space for the most general state vector kUii ¼ kfnp gfnk gii has the dimension
of 2ð2Pþ1Þ ðN þ 1Þð2Kþ1Þ reflecting a maximum number of ð2P þ 1Þ fermions and Nð2K þ 1Þ bosons. Unfor-
tunately, this dimension increases exponentially with the truncation parameters P and K.
To give an idea about the numerical complexity, let us assume we permit the fermions to take se-
ven different momenta ð3D; . . . ; 3DÞ and allow the bosons to be in only three different modes
ðD; 0; DÞ with a maximum occupation number of N ¼ 3 ðnk ¼ 0; 1; . . . ; 3Þ. This would correspond to
a system of (at most) 16 (seven fermions plus nine bosons) fully interacting quantum particles. In this
case the vector space is 27  43 ¼ 8192 dimensional. As discussed in Section 2, due to the conservation
of the fermion number operator Rp b ^y b
^ ^ ^y ^ ^yk a
^k the Hamilto-
p p and the total momentum M  Rp pbp bp þ Rk ka
nian matrix can be decomposed into invariant subspaces.
In this work, we will restrict our analysis to the single-fermion subspace, corresponding to
Rp b^yp b^p ¼ 1.
^ For our above example (with P = 3, K = 1 and N = 3) this space is only 7  43 ¼ 448 dimen-
sional. This space can be broken down further into 13 subspaces, associated with total momen-
 
tum M b  6; 5; . . . 6. These 13 subspaces have a dimension of 4 (for M b ¼ 6Þ; 12 Mb ¼ 5 ;
         
24 M b ¼ 4 ; 40 Mb ¼ 3 ; 52 M b ¼ 2 ; 60 Mb ¼ 1 , and 64 Mb ¼ 0 . If we assume that the
momentum range of the bosonic modes [which is KðK þ 1ÞN=2] is smaller than or equal to the fermion’s
mode range P, then one can show that the dimension of the subspace with M b  0 is identical to the
dimension of the boson space itself, ðN þ 1Þ2Kþ1 ¼ 64. This is obvious as every possible bosonic state
can be balanced by the momentum of the fermion mode.
We will use the interaction-free energy eigenstates (g = 0) for a single fermion as the basis vectors
for our computation, as the action of the creation and annihilation operators is known analytically,
p ^p kfnp gfn gii ¼ kfn1 ; . . . np  1; . . .gfn gii. To sim-
^k kfnp gfnk gii ¼ nk kfnp gfn1 ; . . . nk  1; . . .gii and b
a k k
plify our notation for the scalar products, we use the abbreviation hhfnp gfnk gkfn0p gfn0k gii ¼
dnp;n0p dnk;n0k  dn;n0 . If all ð2P þ 1 þ 2K þ 1Þ integers are identical to their primed counterparts, we ob-
tain dn;n0 ¼ 1, and dn;n0 = 0 if any pair does not match. For the case where not all pairs agree, such as
hhfnp gfn1 ; . . . nk ; . . .gkfnp gfn1 ; . . . n0k þ 1; . . .gii, we use the notation dn;n0 ðnk ¼ n0k þ 1Þ. This means that
all integer pairs have to agree except the kth boson mode for which we require nk ¼ n0k þ 1 to obtain
dn;n0 ðnk ¼ n0k þ 1Þ ¼ 1. This short-hand notation allows us to single out those few modes, for which
the occupation numbers do not match. To have a consistent notation, we have dn;n0 ðnk ¼ n0k Þ ¼ dn;n0 .
Similarly, if we single out two modes we would generalize the notation to dn;n0 ðnk1 ¼
n0k1 þ 1; nk2 ¼ n0k2 þ 1Þ. In this basis, the 2ð2Pþ1Þ ðN þ 1Þð2Kþ1Þ  2ð2Pþ1Þ ðN þ 1Þð2Kþ1Þ matrix representation
of the Hamiltonian of Eq. (3.1) becomes

hhfnp gfnk gkHkfn0p gfn0k gii ¼ dn;n0 ½Rp00 np00 Ep00 þ Rk00 nk00 xk00  þ g Rp00 Rk00 ð2xk00 Þ1=2
 p  
 dn;n0 np ¼ n0p00  1; np ¼ n0p00 þk00 þ 1 ðnk00 Þdn;n0 nk ¼ n0k00  1
p  
þ ðnk00 þ 1Þdn;n0 nk ¼ n0k00 þ 1 ð3:2Þ
270 T. Cheng et al. / Annals of Physics 325 (2010) 265–286

We note that the truncation in the number of the fermion modes P is a potentially serious approx-
^y in the interaction term of Eq. (3.1) can excite modes of higher
imation as the operators of the form b pþk
momentum P þ k (that were truncated), while the set of equations with regard to the bosonic modes K
is closed. In other words, the Hamiltonian does not couple bosonic states with different momentum, it
just changes occupation of the states within the given bosonic manifold.

3.2. Numerical energy spectra of one-fermion states

Let us now analyze the energy spectrum of the Hamiltonian (3.1) as a function of the coupling
parameter c. If we choose L = 1 a.u., P = 3, K = 3 and N = 2 the subspace for momentum M = 0 is 1233
dimensional. From now on we express all energies in units of c2 . The corresponding seven bare fermi-
onic eigen energies are E0 ¼ 1; E1 ¼ 1:00105; E2 ¼ 1:00420 and E3 ¼ 1:00942. If we arbitrarily
choose the mass of the boson to be 1/100 of the fermion mass, m = 0.01, the corresponding seven bare
boson energies are x0 ¼ 0:01; x1 ¼ 0:0469; x2 ¼ 0:0922 and x3 ¼ 0:1379. The 1233 energies of
the combined system in this M = 0 subspace range from 1 to 2.12835 for c ¼ 0, corresponding to
the minimum of E0 ¼ 1 (zero bosons and the fermion at energy E0) to the largest energy
E0 þ 2x3 þ 2x2 þ 2x1 þ 2x0 þ 2x1 þ 2x2 þ 2x3 (14 bosons).
In Fig. 1a we show the lowest part of the spectrum as a function of the interaction strength from
c ¼ 0 to c ¼ 0:1. We have labeled the nine lowest states according to the quantum numbers ðp; nk Þ
they would take in the non-interacting limit c = 0. The lowest three eigenvalues (labeled a, b, c) cor-
respond to the state where (for c = 0) the fermion is in the p = 0 state together with n0 ¼ 0; 1 and 2
bosons in the k = 0 mode, respectively. Each of the next higher states (d, e, f) is doubly degenerate.
The two curves labeled (d) have the fermion in the state p ¼ 1, and a boson in the state with
k ¼ 1 for c = 0. The two two-boson states labeled by (e) have the same configuration as (d), but there
is an extra boson with momentum k ¼ 0; n0 ¼ 1. The two three-boson states (f) are characterized by
p ¼ 1; n1 ¼ 1 and n0 ¼ 2. We observe that for our numerical truncation parameters, the coupling c
seems to remove the degeneracy and leads to a complicated arrangements of states including several
crossings.
In order to examine the numerical convergence with respect to the truncation to only three differ-
ent fermion momenta (P = 3), we have repeated the diagonalization for 921 and 567 dimensional sub-
spaces with maximum momentum of only P = 2 and P = 1 for 10 values of c. While the spectra for the
p = 0 states (a), (b) and (c) are relatively nicely converged (for N = 2 and K = 3), we see that the states (d,
e, f) with p ¼ 1 change sufficiently if P is increased from P = 1 to P = 3. We also see that quite univer-
sally each of the less converged eigenvalues has higher energies than its converged value. We have
zoomed into the crossings of the curve (b) with the two curves (d) at c  0:051, but it did not reveal
any avoided crossing within our numerical resolution.
To illustrate the numerical convergence with respect to the truncation parameter K, the maximum
number of boson modes, we show in Fig. 1b the five lowest states obtained from diagonalizing the
Hamiltonian with a maximum momentum of K = 2 (with 195 states), K = 1 (27 states) and K = 0 (only
three states). The good convergence for K = 0 for the lowest state (a) in this weak coupling limit for
c < 0:02 suggests that this state is primarily a superposition of states with bosons in the k = 0 mode
only. This observation is also consistent with second order perturbation theory, as we will discuss
in Section 4. The states that are based on the boson with momentum k ¼ 1 cannot be predicted
for a truncation parameter K < 1.
It turns out, that the most difficult parameter to truncate is the maximum occupation number in
each bosonic mode N. To demonstrate this difficulty, we focus from now on the lowest energetic state
(a) in Fig. 1a. Due to the convergence problems for increasing c, the energies in Fig. 1a and b can pro-
vide only a rough qualitative estimate of the whole spectrum and one has to find a compromise be-
tween P, K and N to remain numerically feasible. In order to overcome these convergence problems
and to obtain more accurate energies for particular states, one has to optimize the choice of the basis
states. Using state (a) as an example for c ¼ 0:02, we have varied P, K, and N systematically. It turns out
that not every boson mode requires the same maximum occupation number. For example, for P = 2,
K = 2 and N = 6 (leading to 4949 states) all 10 significant digits of the energy remain the same if we
keep the maximum occupation number for the k = 0 mode at N 0 ¼ 6, but lower N 1 to 2 and N 2 to
 T. Cheng et al. / Annals of Physics 325 (2010) 265–286 271

A
1.10

f
e
1.05 d
c
b
1.00 a

0.95
K=3, N=2
P=3
0.90 P=2
P=1

0.85
0 0.02 0.04 0.06 0.08

B
1.05
d
c
b
1.00 a

0.95 P=3, N=2

K=3
K=2
0.90
K=1
K=0
0.85
0 0.02 0.04 0.06 0.08

C
1.0
N =1
0

N =3
0.8 0

N =5
0

0.6
N = 10
0

0.4 N =4 N = 25
±1 0

N = 50
0
0.2 N = 75
0
N=
0
0.0
0.00 0.04 0.08 0.12 0.16

Fig. 1. The lowest energy eigenvalues for the coupled fermion–boson system (Eq. (3.1)) as a function of the unitless coupling
parameter cð¼ L1=2 c5=2 gÞ in the sub-manifold of total momentum M = 0. [Numerical box size L = 1 a.u., boson mass 0.01 a.u.,
fermion mass = 1 a.u. and N x ¼ 7 grid points.] (a) The eigenvalues labeled a, b, c correspond to the (unperturbed) quantum
numbers p = 0 and n0 ¼ 0; 1; 2, respectively. The eigenvalues labeled d, e, f are doubly degenerate and correspond to
p ¼ 1; n1 ¼ 1, and n0 ¼ 0; 1; 2, respectively. The data obtained for K = 3 and N = 2 are compared to those for P = 1, 2, 3. (b) The
lowest four energies (for five states) as a function of c for P = 3, N = 2 and K = 0, 1, 2 and 3. (c) The lowest energy for P = 1 and
K = 1 as a function of c for several maximum occupation numbers N 0 of the k = 0 bosonic mode, with N 1 ¼ 4.

only 1 (leading to only 315 basis states). This reduction in the required number of basis states allows
us to increase N 0 to obtain convergent energies. For example, the two energies obtained for
272 T. Cheng et al. / Annals of Physics 325 (2010) 265–286

ðP ¼ 2; K ¼ 2; N 0 ¼ 10; N 1 ¼ 3 and N 2 ¼ 2Þ and ðP ¼ 2; K ¼ 2; N 0 ¼ 11; N 1 ¼ 4 and N 2 ¼ 3Þ differ

by less than 3  106 %. Our best estimate of the energy of this state for c ¼ 0:02 is E = 0.983656c2.
We use this value to estimate the accuracy of two approximate but analytical expressions derived
in the next section.
In Fig. 1c we show the convergence for the K = 1-model as a function of the maximum occupation
number N 0 in the k = 0 mode, while we keep the maximum occupations numbers in the k ¼ 1 modes
constant at N 1 ¼ 4. We display the energy of state (a) for the entire range of couplings c for which the
energy is positive. For the largest coupling c, the numerical energies for N 0 ¼ 165 and N 0 ¼ 175 are
identical. The graph suggests that while for c < 0:04 the truncation with N 0 ¼ 10 is sufficient, for
c < 0:1 more boson states are required and a truncation with N0 ¼ 50 is necessary.

4. Validity of the Greenberg–Schweber approximation

Under certain limits the Hamiltonian equation (3.1) has astonishing simplifying properties. In a
breakthrough paper in 1958, Greenberg and Schweber [24] assumed that all fermion modes take
p
the same energy, Ep  ½c4 þ c2 p2   c2  E, corresponding to all fermions at rest with p = 0. The
authors introduced a unitary transformation U GS to define new fermionic and bosonic operators
according to Bb p  Uy b^ b y
^
GS p U GS and A k  U GS ak U GS . If our Hamiltonian of Eq. (3.1) [with the short-hand
n o  
^ a
notation Hðb; ^Þ] is rewritten in terms of this set of new operators B b k , we obtain H b;
bp; A ^ a
^ ¼
        
H b^ B; b ;a
b A b A
^ B; b ¼ H1 B b þ H2 A b . In other words, when rewritten in these new operators the

same Hamiltonian decouples into the sum of purely fermionic and bosonic degrees of freedom, and
 
b is basically the free bosonic evolution, the Hamiltonian
there is no interaction term. While H2 A
 
H1 Bb contains quartic terms in B
b (and B
b y ) suggesting a direct coupling of the new fermionic modes.
This observation has interesting consequences with regard to our understanding of forces. The tra-
ditional and perturbation theory based view tries to explain the interaction between two (bare) fer-
mions based on the existence of intermediating (bare) ‘‘virtual” bosons, which intermediate these
forces. In this view, two (bare) fermions do not interact directly with each other. The picture provided
by Greenberg and Schweber provides a fundamentally different interpretation. The true and physically
detectable fermions are not described by the operators b ^ and a^, but by their dressed (or sometimes
b b
called ‘‘clothed”) counterparts B and A. As a result, two physical fermions can interact instantaneously
 
and directly with each other as described by H1 B b and are entirely decoupled from the (physical) bo-
b This view raises obviously several intriguing questions
sons, as described by the physical operators A.
about locality, instantaneity and causality of the interactions in view of special relativity. An interest-
ing discussion of these fundamental issues can be found in a recent book by Stefanovich [12].
In 1970, Walter [25] extended the GS-model to allow a non-relativistic momentum dependence of
the energy, EðpÞ ¼ c2 þ kp2 =2, and generalized the concept of the dressing transformation up to linear
terms in the parameter k. The Hamiltonian in the corresponding new variables contained new (phys-
ical) nucleon-meson coupling terms.
More specifically, the Greenberg–Schweber transformation is given by
h  1=2 ^y ^  y i
U GS ¼ exp g Rp Rk 2x3k ^k  a
bp bpþk a ^k ð4:1Þ

One can show that this form is related to the interaction Hamiltonian expressed in the interaction pic-
ture, however, the corresponding time integrals are extended only from t = 0 (instead of t ¼ 1) to
t ¼ 1. When re-expressed in the dressed variables, the Hamiltonian (3.1) becomes the sum of
   
H1 Bb and H2 A b :

 2 1 y by b
by B
b 2 b B b by b
H ¼ ERp B p p  g Rp Rq Rk 2 x k B pþk p B q B qþk þ Rk xk A k A k ð4:2Þ
 T. Cheng et al. / Annals of Physics 325 (2010) 265–286 273

A similar discretization as introduced in Section 3, Bb k  p ðDÞ BðkÞ,

b can also be applied in position
b p b
space, B x  ðDxÞ BðxÞ, where similar to the transition from the continuous momentum coordinate
k to an integer subscript k (Section 2) the continuous position coordinate x in BðxÞ b becomes an integer
subscript x in B b x extending from x ¼ N x =2 to x ¼ N x =2. As a result we have
b x  ð1=p N x ÞRp B
B b p ¼ ð1=p N x ÞRx B
b p expðipx2p=N x Þ and equivalently B b x expðipx2p=N x Þ. If we replace
b
the momentum operators B p by (their Fourier transformed) spatial representations and use
Rp expðipx2p=Nx Þ ¼ Nx d0x , we can rewrite the sum of the first two terms, as denoted above by H1 ðBÞ as
by B
b 2 by b by b
H1 ðBÞ ¼ ERx B x x þ c Rx Rx0 B x B x V xx0 B x0 B x ; ð4:3Þ

We use the notation V xx0  ðc5 =2pÞDRk x2 k exp½iðx  x Þk2p=N x  for the attractive one-dimensional
0

Yukawa-like potential. If we assume that all boson modes are weighted equally, f(k) = 1, we can
approximate the discrete sum (by using the continuum limit D ! 0) as DRk x2 k exp½izk2p=N x  
R
dkxðkÞ2 exp½izk ¼ ðp=mc3 Þ exp½cmjzj. In this limit the one-dimensional Yukawa-like potential
can be written as V xx0 ¼ ðc2 =2mÞ exp½cmjx  x0 jDx. The spatial range of the resulting force
V 0xx0 ¼ ðc3 =2Þ exp½cmjx  x0 jDx increases exponentially with decreasing boson mass m.
Furthermore, if we remove the self-energy (diagonal term) from the interaction term in Eq. (4.2), by
replacing B by B b by b by by b b
pþk p B q B qþk with  B pþk B q B p B pþk þ dp;q we obtain
 2 1 y
by B
b 2 b B by b b
H1 ðBÞ ¼ ðE  rÞRp B p p  g Rp Rq Rk 2xk B pþk q B p B pþk ð4:4Þ
 1
where r  g 2 Rk 2x2k denotes the energy correction, leading to what is related to the renormalized
mass.

4.1. GS predictions for the energy spectrum

The question that we will address here is whether it is possible in principle to generalize the GS
scheme to introduce new bosonic and fermionic operators such that also the Hamiltonian for moving
fermions can be decoupled entirely in new operator variables. If the Hamiltonian can be rewritten as
   
the sum of two decoupled portions, H ¼ H1 B b þ H2 Ab , then its energy spectrum must be given by

the direct sum of the fermionic and bosonic spectrum, Ea ¼ E1 ðiÞ þ E2 ðjÞ, with the fermionic quantum
numbers i ¼ 1; . . . ; 2ð2Pþ1Þ and the bosonic indices j ¼ 1; . . . ; ðN þ 1Þð2Kþ1Þ . In other words, the separabil-
ity of the spectrum Ea is a necessary criterion for the existence of a new set of operators that can sep-
arate the Hamiltonian. As we rewrite the same Hamiltonian just in a different set of new operators,
 
that were obtained by unitary transformations of the original one assumed by GS, H b; ^ a
^ ¼
    h    i
^ y ; Ua
U y H U bU ^U y U ¼ U y H B; b U ¼ U y H1 B
b A b þ H2 A b U, the spectrum is unchanged. That means
 
we can examine the spectrum of our original Hamiltonian (3.1) H b; ^ a
^ directly to test for a possible
separability. For example, if the spectrum Ea is separable, then among all differences jEa  Eb j there
must be 2ð2Pþ1Þ and ðN þ 1Þð2Kþ1Þ differences that take identical values. If the actual data jEa  Eb j do
not contain these groups of identical values then one can safely conclude that it is not even in principle
possible to find the set of new operators that would decouple the bosonic and fermionic degrees of
freedom.
To answer this question analytically we can determine the change of the eigenvalues. If we apply
the general (Raleigh) second-order perturbation theory in g we obtain:

EðgÞ ¼ Ep þ Rk nk xk þ Rp0 Rn0 jhhp; fnk gkVkp0 ; n0k iij2 =½Ep þ Rk nk xk  ðEp0 þ Rk n0k xk Þ ð4:5Þ
PN PN PN
where Rn denotes nK ¼0 nKþ1 ¼0 nK ¼0 . There is no linear term in g as V ¼
1=2 ^ y ^  y

g Rp Rk ð2xk Þ bpþk bp a ^k from Eq. (3.1) couples only states with different bosonic occupation
^k þ a
numbers. The summation in Eq. (4.5) extends in principle over all states of the Hilbert space. As V con-
274 T. Cheng et al. / Annals of Physics 325 (2010) 265–286

serves the number of fermions, only states with a single fermion are coupled to kp; fnk gii. If we eval-
uate each term in the double sum we obtain
    
EðgÞ ¼ Ep þ Rk nk xk þ g 2 Rk ð2xk Þ1 = ðnk þ 1Þ= Ep  Epk  xk þ nk = Ep  Epþk þ xk ð4:6Þ
2
We see immediately that in the test case of Greenberg–Schweber with Ep ¼ Epk ¼ c , the energy cor-
rection, g 2 Rk ð2xk Þ2 represents a global energy shift and thus preserves the separability of the spec-
trum as expected. We note that this term is identical to energy correction term denoted by r in Eq.
(4.4) associated with the correction to the renormalized mass.
We note also a second special case in which the separability of the spectrum is maintained. This is
the highly relativistic limit, Ep ¼ c2 þ cp, where each fermion moves with the speed of light c. Here the
energy differences in denominators also become independent of the quantum numbers,
Ep  Epk ¼ ckD. Except for the two special cases, Ep ¼ c2 and Ep ¼ c2 þ cp, Eq. (4.6) suggests that the
Hamiltonian (3.1) cannot be separated into decoupled fermionic and bosonic spaces.
In Fig. 2 we test the accuracy of the two analytical estimations for the energy of the state (a) as a
function of c. We see that for K ¼ 1 and c ¼ 0:02 the perturbative prediction from Eq. (4.6),
 1
c2 þ g 2 Rk ð2xk Þ1 =½c2  Epk  xk , and the GS-prediction, c2  g 2 Rk 2x2k , are very reliable. To be
more specific, while the exact energy of this state for c ¼ 0:02 is E ¼ 0:98356c2 , the perturbation the-
ory predicts E ¼ 0:98331c2 , and the GS prediction amounts to E ¼ 0:98319c2 .
Had we restricted the momentum range of the summation Rk to only 1 6 k 6 1, the perturbation
theory would predict E ¼ 0:98411c2 , and the GS prediction would be E ¼ 0:98408c2 , showing that
large k values are not necessary. This shows that despite the assumption of non-moving fermions,
the Greenberg–Schweber model can be remarkably accurate. We will postpone a discussion of the rea-
son for this agreement to Section 7.

4.2. The physical particle states

4.2.1. Momentum distribution of the bare fermions

The dressed Hamiltonian equation (4.2) separates the spaces for the physical fermions and bosons.
As a result, the eigenstates of (4.2) are direct products of the states for physical fermions and bosons.
The bosonic states are created by the consecutive application of the operator A b y on the vacuum state.
k
The two Hamiltonians of Eq. (3.1) for Ep ¼ E and (4.2) are unitarily equivalent and share the same vac-
uum state, denoted by kvacii. The single (physical) particle fermion state with momentum Q is defined

1.0

0.8

0.6

0.605 GS (K=1)
0.4
perturbative (K=1)
0.595 exact

0.2 0.585
perturbative (K= )

0.575
0.0996 0.1 0.1004
0.0
0.00 0.04 0.08 0.12 0.16

 1 2
Fig. 2. Comparison of the prediction of the Greenberg–Schweber energies (uppermost line, 1  g 2 Rk 2x2k =c ) and second-
2 1 2 2
order perturbation theory (dotted line, 1 þ g Rk ð2xk Þ =½c  Epk  xk =c ) with the exact numerically obtained energy
eigenvalue for state a as a function of the interaction strength c. The exact eigenvalue was computed for a 4275-dimensional
subspace with P = 4, K = 1 and mixed maximum occupation numbers, N 0 ¼ 170 and N 1 ¼ 4.
 T. Cheng et al. / Annals of Physics 325 (2010) 265–286 275

b y kvacii  kQ ii. This state is an energy eigenstate of Eq. (4.2) with eigenvalue E. Below we inves-
as B Q
tigate the (bare momentum) distribution of the state kQ ii. This is given by the average occupation
number of the mode with bare momentum p; qQ ðpÞ  hhQ kb ^
^y b
p p kQ ii.
This momentum density can be also related to the corresponding expansion coefficients C pn . We
can expand the state kQ ii in terms of the bare fermion and boson states kp; fnk gii, where the vector
n indicates the occupation number of each bosonic mode, kQ ii ¼ Rp Rn C pn kpii
kfnk gii. Here we used
the short-hand notation Rn  RnðKÞ RnðKÞ for the total of ð2K þ 1Þ consecutive summations, where
the occupation number of the kth momentum mode nðkÞ  nk increases from 0 to N. As the result, Rn
requires the summation of ðN þ 1Þð2Kþ1Þ individual terms. The sum over the (bare) bosonic subspace of
the square of the absolute values of the expansion coefficients C pn ¼ hhpk
hhfnk gkQ ii is identical to
qQ ðpÞ.
This can be seen as follows, Rn jC pn j2 ¼ Rn hhQ kpii
kfnk giihhpk
hhfnk gkQ ii ¼ hhQ kpii
hhpk
½Rn kfnk giihhfnk gkQ ii. Using the completeness in the bosonic sub-space this expression simpli-
fies to Rn jC pn j2 ¼ hhQ kpiihhpkQ ii. If we now consider that the (bare) single particle state kpii is created
from the (femionic) vacuum k0ii via b ^y , this expression reduces further to Rn jC pn j2 ¼ hhQ b^y k0iih
p p
^p Q ii. The unit operator in fermionic space 1 can be written as the sum of the projection operators
h0kb
on the vacuum and all other multi-particle states, 1 ¼ k0iihh0k þ Rp kpiihhpk. As the state kQ ii consists
entirely of single (bare particle states), we have hhpkb^p Q ii ¼ 0 as a result we can omit the projector
Rp kpiihhpk from the expression and obtain Rn jC pn j2 ¼ hhQ kb^yp b^p kQ ii.
In Appendix A we derive how this momentum density depends on Q, p and the coupling constant g
and find
h  1 i .
qQ ðpÞ ¼ exp g 2 Rk 2x3k Rj¼0 ðg 2j j!ÞF j ðQ ; pÞ ð4:4Þ

where the coefficients F j ðQ ; pÞ are independent of g and the first five are given by
F 0 ðQ; pÞ ¼ dQ ;p
. 
F 1 ðQ; pÞ ¼ 1 21 x3Q p
. 
F 2 ðQ; pÞ ¼ Rk 1 22 x3Q pk x3k ð4:5Þ
. 
F 3 ðQ; pÞ ¼ Rk Rk0 1 23 x3Q pkk0 x3k x3k0
. 
F 4 ðQ; pÞ ¼ Rk Rk0 Rk00 1 24 x3Qpkk0 k00 x3k x3k0 x3k00

As only the first factor in each summation depends on p, one can easily confirm that the total proba-
bility to find any momentum p is 1 ¼ Rp qQ ðpÞ. Furthermore, we see that the density depends only on
the difference between the bare and physical momentum, qQ ðpÞ ¼ qðQ  pÞ.
Let us now discuss the behavior of qðQ  pÞ. In Fig. 3 we display the (bare) fermion momentum
density qQ ðpÞ  hhQ kb ^
^y b
p p kQ ii for p ¼ 0; 1; 2; 3 for the physical fermion state labeled (a) in
Fig. 1a, as a function of the coupling strength c. For comparison the open circles are the predictions
of the GS-model of Eq. (4.4). The agreement is again astonishing. We note that while the exact data
can be obtained for larger couplings c, the convergence of the series in the analytical expression
(4.4) is very slow for large c. We therefore limit the comparison to 0 < c < 0:025.

4.2.2. Momentum distribution of the bare bosons

Next we can analyze the distribution of (bare) bosons for the physical fermion state kQ ii. An obvi-
ous quantity would be the average occupation number of the bosons with momentum k, defined as
qQ ðkÞ  hhQ ka^yk a^k kQ ii. Following a similar argumentation as for qQ ðpÞ in Section 4.2.1, one can show
that also this density can be related directly to the sum of the squared absolute values of the expan-
sion coefficients C pn of the state kQ ii leading to qQ ðkÞ ¼ Rp Rn nk jC pn j2 . In Appendix B we evaluate this
density and find the remarkably simple result:

 
qQ ðkÞ ¼ g 2 2x3k ð4:7Þ
276 T. Cheng et al. / Annals of Physics 325 (2010) 265–286

100
p=0

<bp+ bp>
10-1 GS

<bp+ bp>
exact
p = ±1
10-2

p = ±2
-3
10

p = ±3
-4
10

0 0.005 0.01 0.015 0.02 0.025

^y b
Fig. 3. The (bare) fermion momentum density, in the physical fermion state kQ ii defined as qQ ðpÞ  hhQ kb ^
p p kQ ii. The circles
are the prediction of the GS-model of Eq. (4.4).

This quantity does not depend on Q. As a result each momentum state kQ ii is characterized
  by 3
an iden-
tical momentum distribution of the (bare) bosons. The relative rapid fall-off for 1= 2x3k  k (for a
small boson mass m) suggests that the numerically required truncation of the higher momenta k
might not be a serious approximation. This is consistent with the results obtained in our numerical
convergence tests displayed in Fig. 1b.
By summing over all momentum modes in the continuum limit, we can compute the total number
R
of bare bosons in the physical fermion state as 1=D dkqQ ðkÞ ¼ c2 =ð2pm2 Þ. In other words, for a cou-
p
pling cc ¼ m ð2pÞ½¼ 0:0251 the distribution of fermions with different momenta p in the state kQ ii is
accompanied with exactly one (bare) boson. If we take the discreteness of the numerical momenta
 
into account, we obtain for our numerical discretization box (with L = 1) Rk g 2 = 2x3k ¼ 7:73
108 g 2 . If we require Rk qQ ðkÞ ¼ 1, we obtain g = 3597, equivalently to c ¼ 0:0164 and in perfect
agreement with our numerical finding for the Hamiltonian without the GS approximation. This is of
the same order as the prediction based on the continuum approximation cc .
This particular value of c can serve as a criterion to judge the strength of the coupling. Even though
the perturbation theory is quite accurate in predicting the energy for this value, this coupling is suf-
ficiently large to change the c ¼ 0 state with no boson to a new dressed state characterized by an equal
number of bosons and fermions. In this sense the coupling is more than just a perturbation.
In Fig. 4 we demonstrate the reliability of the GS prediction of Eq. (4.7). The agreement between the
GS prediction and the exact boson populations is again very good. While hhQ ka ^y0 a
^0 kQ ii approaches 10
for c ¼ 0:05, the other occupation numbers remain far less than unity. This is consistent with the find-
ings of Fig. 1c, that only the k = 0 mode with a large maximum occupation number must be taken into
account for accurate eigen energies while all the other modes are less relevant and can almost be ne-
glected for qualitative estimates.
There is another remarkable observation about an apparent fermion–boson symmetry in the eigen-
state kQ ii. We have included by the dots in the Figure also the fermionic occupation number we
showed already in Fig. 3, hhQ kb ^p kQ ii. Except for k = 0 and p = 0, the average number of bosons
^y b
p
and fermions for the same momentum k = p are nearly identical. Comparing the corresponding analyt-
ical expressions (4.4) and (4.7) we see that they both predict an identical scaling behavior up to second
order in the coupling constant: hhQ ka^y a
^ kQ ii ¼ hhQ kb ^ kQ ii ¼ g 2 =2x3 . It suggests that each fer-
^y b
k k p¼k p¼k k
mion is paired up with a boson with the same momentum, independent of the coupling strength in the
energy eigenstate. As the sum over all fermion momentum states is equal to 1, it means that also the
sum over all boson occupation numbers (except k = 0) remains under 1.
 T. Cheng et al. / Annals of Physics 325 (2010) 265–286 277

101
k=0
0
10 <ak+ ak>GS
<ak+ ak>exact
10-1
k, p = ±1
10-2
k, p = ±2
-3
10

<bp+ bp>
k, p = ±3
10-4

0 0.01 0.02 0.03 0.04 0.05

^yk a
Fig. 4. The (bare) boson momentum density, in the physical fermion state kQ ii defined as qQ ðpÞ  hhQ ka ^k kQ ii. The circles are
the prediction of the GS-model of Eq. (4.7) and the dots indicate the fermion momentum density from Fig. 3.

4.2.3. The spatial properties of the physical fermion state

As eigenstates kQ ii have to reflect the translational invariance of the Hamiltonian, the spatial prob-
ability densities in terms of bare bosons and fermions, defined as hhQ kb ^
^y b ^yx a
^x kQ ii, have
x x kQ ii and hhQ ka
to be independent of the position x. Here we have used again the spatial operators defined as
^x  ð1=p N x ÞRp b
b ^p expðipx2p=N x Þ and a p
^k expðikx2p=N x Þ. In order to learn something
^x  ð1= N x ÞRk a
about the spatial distribution we have to compute the correlation function qQ ðx; yÞ 
^ ^y a
^y b
hhQ kb x xay ^ y kQ ii. In Appendix C we derive the following analytical form:
  1=2 2
qQ ðx; yÞ ¼ ðg=Nx Þ2 Rk 2x3k 
exp ½ikðy  xÞ2p=Nx  ð4:8Þ

As expected, the density qðx; yÞ depends only on the fermion–boson distance, qQ ðx; yÞ ¼ qQ ðjx  yjÞ
and it is independent of Q. If we sum over all bosonic and fermionic coordinates, Ry Rx qQ ðx; yÞ we ob-
tain (in the continuum limit) c2 =ð2pm2 Þ, fully consistent with the total number of bosons Rk qQ ðkÞ as
shown in the previous Section 4.2.2. If we perform the continuum limit, the spatial density can be
written in terms of the modified Bessel function of order 1/4:

qQ ðx; yÞ ¼ ðg=ð2Nx ÞÞ2 ½1=D23=4 p1=2 jzj1=4 K 1=4 ðcmjzjÞc7=4 m1=4 C1 ð3=4Þ2 ð4:9Þ

5. Time dependence of a vertex reaction

As discussed in Section 1, the prediction of the time-evolution of quantum field theoretical observ-
ables is not yet experimentally confirmed and therefore controversial. As an example, we will analyze
here the temporal implications of the interaction term in Eq. (3.1) for a fermion, initially in the bare
energy eigenstate state for c ¼ 0. Our goal here is just to illustrate these implications within the
framework of the bare particles without necessarily arguing that the Hamiltonian equation (3.1) pre-
^y b
dicts the correct physical time evolution. The simple tri-linear vertex interaction b ^ ^y
pk p ak corresponds
to the creation of a boson with momentum k, while the fermion’s momentum p is decreased to p  k. If
^y b
^ ^yk a
we associate only arbitrarily chosen portions of the full Hamiltonian, such Rp Ep b p p and Rk xk a ^k ,
with the ‘‘energy”, then one can show that the above boson creation process triggered by single fer-
mion violates this kind of ‘‘energy conservation”. In other words, the required equality
Ep ¼ Epk þ xk cannot be satisfied, unless in the trivial case where m = 0 and k = 0. As a result, this pro-
cess has been called ‘‘virtual boson creation”.
Two main (in principle equivalent) approaches have been proposed to remove this process from
consideration. In the first approach [23] a new set of effective (dressed or sometimes called clothed)
278 T. Cheng et al. / Annals of Physics 325 (2010) 265–286

fermionic and bosonic creation and annihilation operators have been defined. When the Hamiltonian
is rewritten in terms of these new operators (as we showed in Section 4) the virtual boson creation is
absent. In other works, appropriate unitary transformations were constructed with the goal to elim-
inate the tri-linear interaction term from the new transformed Hamiltonian. An iterative scheme
where subsequently all virtual processes could be eliminated, was recently proposed by Kobayashi
et al. [26]. Using an infinite product of unitary operators, the process of the boson creation under
an inelastic fermion–fermion scattering occurred in higher order of the coupling constant. A physically
observed realization of this process is the neutral pi-meson creation in low-energy proton-proton
scattering. Its accurate theory [27] requires quantum chromodynamics, and it is not clear at all to
us, whether the simple and phenomenological Yukawa system is the correct theoretical approach.
Using the Hamiltonian of Eq. (3.1), we can construct the Heisenberg equation of motions
h i
^p =ot ¼ b
^ ; H and iob
^ =ot ¼ ½a
ioa ^p ; H for the bare operators.
k k

^k =ot ¼ xk a
ioa ^y b
^k þ g ð2xk Þ1=2 Rp b ^ ð5:1aÞ
pk p
^p =ot ¼ Ep b
^p þ g R ð2x Þ1=2 b^  
iob k k pk a ^yk
^k þ a ð5:1bÞ
The solution for the special case of the boson mode for k = 0 can be obtained analytically as the second
term on the right-hand side of Eq. (5.1a), Rp b^y b
^
p p , is a constant operator of motion representing the
conserved total number of fermions. The solution for our single-fermion initial state (with zero bo-
sons) is

a ^0 ðtÞ ¼ ð2g 2 =x30 Þ sin2 ðx0 t=2Þ
^y0 a ð5:2Þ
This solution shows a periodic growth and decrease of the (bare) photon number on a time scale
2p=x0 , which is independent of the coupling strength. It is interesting that neither the amplitude
nor the frequency of this temporal pattern depends on the initial state of the fermion, as long as it sat-
isfies hhUðt ¼ 0ÞkRp b^y b
^
p p kUðt ¼ 0Þii ¼ 1.
We note that (in the absence of any other ðk–0Þ modes) the free fermionic energy Rp Ep b ^y b
^
p p is con-
y
stant and the growth of the free  bosonic energy
 ^y ^ x ^
a
0 0 0^
a is exclusively associated with a decrease of the
interaction energy, g ð2x0 Þ1=2 a ^y0 Rp b
^0 þ a p bp . In this sense, the fermion acts as a catalyst for the peri-
odic production of the bare k = 0 bosons. It therefore does not violate the conservation of total energy
or the total momentum.
The analytical solution gives us also an opportunity to gauge the accuracy of our numerical evalu-
ations and also to test again analytical approximations. In Fig. 5 we have graphed the average photon
 y
number a ^0 ðtÞ as a function of time for three maximum occupation numbers N in this mode. It was
^0 a
obtained by the diagonalization of the Hamiltonian, expressing the initial state kUðt ¼ 0Þii in the

1.5
+
N=6 analytical
<a a >
0 0
N=5

1 N=3

0.5

0
0 0.05 0.1 0.2
time [a.u.]
 y
Fig. 5. The time evolution of the average boson number for the k = 0 mode a ^0 a
^0 ðtÞ for three maximum occupation numbers N.
2
For comparison, the dotted line is the analytical expression from Eq. (5.2), 2g 2 =x30 sin ðx0 t=2Þ. ½c ¼ 0:01; P ¼ 1; K ¼ 1.
 T. Cheng et al. / Annals of Physics 325 (2010) 265–286 279

energy eigenbasis kaii and then evolving this superposition in time. The time evolved  y state,

kUðtÞii ¼ Ra hhakUðt ¼ 0Þii expðiEa tÞkaii, was then used to compute the expectation value a ^0 a
^0 in
the Schrödinger picture. For comparison, the short-dashed line represents the analytical expression
(5.2). The effect of an insufficiently large truncation parameter N is twofold. As the entire Hilbert space
is not available, the state with largest occupation number N acts as a reflecting mirror, long before the
occupation number can reach its true maximum value ð2g 2 =x30 Þ, the occupation number shrinks again,
 y
leading to a reduced amplitude for a ^0 and also to an incorrect shorter period than 2p=x0 .
^0 a
For all other modes with k–0 the coupling operator Rp b ^y b ^ in Eq. (5.1a) is not conserved. The Hei-
pk p
senberg equations of motion couple the average boson number operator a^yk a
^k to all other momenta k
 y
and finding an exact analytical solution for a ^k a
^k seems hopeless.
 y
However, for short times xk t << 1, one can show that a ^k  g 2 =ð2xk Þt2 , independent of the fer-
^k a
mionic initial state. If we choose as an initial state an arbitrary superposition of fermionic momentum
states, kUðt ¼ 0Þii ¼ Rp C p kp; fnk ¼ 0gii, its time evolution for short times can be approximated as
kUðtÞii ¼ exp½iHtkUðt ¼ 0Þii  ð1  iHtÞkUðt ¼ 0Þii, leading to
h  i
^y 0 b
kUðtÞii ¼ Rp C p exp it Ep þ g Rp0 Rk0 ð2xk Þ1=2 b ^ 0a
^y 0 kp; fnk ¼ 0gii
0
p þk p k

 Rp C p ð1  itEp Þkp; fnk ¼ 0gii  igtRp C p Rk ð2xk Þ1=2 kp þ k; fnk ¼ 1gii ð5:3Þ
2 2
as a^k kp; fnk ¼ 0gii ¼ 0, we obtain hhU ^yk a
UðtÞii ¼ ðgtÞ =ð2xk ÞRp jC p j . For any single fermionic
ðtÞka ^k k
2
state we have the normalization Rp jC p j ¼ 1. As a result it is not possible to choose any superposition
of states kp; fnk ¼ 0gii that could coherently suppress the creation of bare bosons.
The next important question is whether this unavoidable creation is at least reversible (or periodic)
(as suggested by Eq. (5.2) for k = 0). If so, after a certain time the initial state of zero-bosons could be
recovered, and the growth would not evolve into a permanent state with a nonzero number of bosons.
In Appendix D we show that if the initial fermion state contains only sufficiently low momenta p, the
time evolution can be approximated in the limit of a small coupling by

a ^k ðtÞ  2g 2 =d2 xk Þ sin2 ðdt=2Þ
^yk a ð5:4Þ
The energy difference is d  xk  ðEpþk  Ep Þ where p is the initial fermion momentum. In Fig. 6 we
examine the accuracy of this approximation for several coupling strengths. We see that for small c
 y
the solutions seems to remain sinusoidal. For larger c, however, a ^k a
^k ðtÞ can not return to its original
value suggesting the reversibility of the virtual boson creation process.  y
2
We note that the time average (using sin ð Þ ¼ 1=2) of the occupation number a ^k a
^k ðtÞ is
2
g 2 =ðd xk Þ. In the Greenberg–Schweber approximation (for which d ¼ xk ) this scales similarly as the
analytical expression obtained for bare boson density qQ ðkÞ ¼ g 2 = 2x3k of Eq. (4.7). In other words,
except for a factor of 2, the dressed GS states from Section 4 could be viewed to represent approxi-
mately the properties of a time averaged evolution with regard to the bare bosons.

+
<a 0.3
a>
1 1

0.2
=0.05
0.1

0
+
exact analytical <a a >
1 1

0.01
=0.01
0
0 0.02 0.04 0.06 time [a.u.] 0.1
 y
Fig. 6. The time evolution of the average boson number a ^1 ðtÞ for two coupling strengths c. The straight line is the
^1 a
 y
^1 ðtÞ  2g 2 =d2 x1 Þ sin2 ðxt=2Þ whereas the dashed line is exact.
^1 a
approximate analytical expression from Eq. (5.4) a
280 T. Cheng et al. / Annals of Physics 325 (2010) 265–286

6. Spatial dependence of the vertex reaction

Let us now show how the numerical approach and the spectral information from the previous sec-
tions can be used to study the spatial evolution of the fermions and bosons. A detailed examination of
the physical significance is beyond the purpose and scope of this paper and we refer to future work on
this discussion.
As discussed in Section 4.2.3 the spatial probability density for the fermion and bosons for a general
state kUðtÞii is given by hhUðtÞkb ^
^y b ^y a
^ kUðtÞii, where we have the spatial opera-
x x kUðtÞii and hhUðtÞka
^x  ð1=p N x ÞRp b
^p expðipx2p=N x Þ and a p x x
tors as b ^k expðikx2p=N x Þ leading to
^x  ð1= N x ÞRk a

^yx a
a ^yp1 a
^x ðtÞ ¼ 1=N x Rp1 Rp2 hhUðtÞka ^p2 kUðtÞii exp½iðp2  p1 Þx2p=Nx  ð6:1aÞ
D E
^
^y b ^y ^
b x x ðtÞ ¼ 1=N x Rp1 Rp2 hhUðtÞkbp1 bp2 kUðtÞii exp½iðp2  p1 Þx2p=N x  ð6:1bÞ

As an initial state we take the coherent superposition of bare fermion states with momentum p, and
a
D Gaussian
E amplitude Gp ; kUðt ¼ 0Þii ¼2 Rp Gp kp; fn ¼ 0gii. As a result we have initially
^y b
^
  y k
b 
x x ðt ¼ 0Þ ¼ ð1=N x Þ Rp Gp expðipx2p=N x Þ
 and a ^x a
^x ðt ¼ 0Þ ¼ 0. In order to evolve the state k/ii
in time, we have to expand the bare states into the eigenstates of the Hamiltonian, kaii, with energy
Ea . As a result the time evolution is given by kUðtÞii ¼ Ra hhakUðt ¼ 0Þii expðiEa tÞkaii. The expansion
of each of these energy eigenstates in terms of the bare states is kaii ¼ Rp;n hhp; fnk gkaiikp; fnk gii. The
unitary matrix elements hhp; fnk gkaii are found numerically as eigenvectors of the Hamiltonian (3.1).
If we insert these expansions into Eq. (6.1) we obtain

hhUðtÞka ^k2 kUðtÞii ¼ Rb1 Rb2 Rp3 Rp4 Gp3 Gp4 Kb1;p3 ðtÞKb2;p4 ðtÞhhpðb1 Þfnk ðb1 Þgka
^yk1 a ^yk1 a
^k2 kpðb2 Þfnk ðb2 Þgii
ð6:2aÞ
^ y ^   ^ y ^
hhUðtÞkbp1 bp2 kUðtÞii ¼ Rb1 Rb2 Rp3 Rp4 Gp3 Gp4 Kb1;p3 ðtÞKb2;p4 ðtÞhhpðb1 Þfnk ðb1 Þgkbp1 bp2 kpðb2 Þfnk ðb2 Þgii
ð6:2bÞ

with the coupling matrix Kb;p ðtÞ  Ra expðiEa tÞhhp; fnk ¼ 0gkaiihhp; fnk gkaii . Numerically it is
advantageous to define an index matching vector that maps the index a (which numerates the eigen-
values of the coupled Hamiltonian), to the fermionic momentum pðaÞ as well as to the set of occupa-
tion numbers, fnk ðaÞg, that characterize the corresponding uncoupled state to which the eigenvector
kaii reduces to in the c ¼ 0 limit, i.e., limc!0 kaii ¼ kpðaÞ; fnk ðaÞgii.
As we have shown in Section 3.1, with P = 4, K = 4 and N = 1 the Hilbert space is 4608 dimensional.
In order to obtain convergent result, we can only describe very small coupling strength c such that the
truncation to only a single boson per mode is accurate. Even though we have to diagonalize a large
4608  4608 Hamiltonian matrix, the resulting densities can only be represented on nine grid points
to be consistent with nine momentum states.
In Fig. 7 we show the spatial evolution of the fermionic and bosonic spatial densities
D E  y p
b ^
^y b ^x a
^x ðtÞ for an initially spatially localized fermion state with Gp ¼ 1= N x , correspond-
x x ðtÞ and a

ing to a delocalized state in momentum space. The coupling was chosen to be c ¼ 13 and the 11 snap-
shots were taken at times t ¼ n3:5  105 a:u:, with n ¼ 0; 2; . . . ; 10. While the main purpose of this
work is to discuss the numerical aspects and we refer the reader to another work for a more detailed
physical analysis of this evolution, we just point out that the bosons grow only in those spatial regions
where the fermion is located. This highly localized growth of the bosons might be related to the obser-
vation of Section 4.2.2 that each fermion in the energy eigenstate was paired up with a boson of iden-
tical momentum. Due to the small mass of the boson compared to that of the fermion, after its creation
the boson distribution spreads out at a much faster rate than that of the fermion. The spreading and
the spatial evolution of the fermions cannot be predicted by the GS approximation. After a short period
of time the boson distribution hits the boundary of our numerical box at x ¼ L=2, and we obtain
D E
unphysical interferences. The total number of all fermions, Rx b ^y b
^
x x ðtÞ, remains equal to 1 for all
 y
times, while the increase of the total number of bosons, Rx a ^x a
^x ðtÞ, seems to be unaffected by the
reflection against the boundary.
 T. Cheng et al. / Annals of Physics 325 (2010) 265–286 281

<ax+ ax>
10-1 <bx+ bx>
-3
10
8 10-6 10-5

-4 10-7
t= 3.5x10 a.u. <ax+ ax>
6 10-6 10-9
-0.4 -0.2 0 x [a.u.] 0.4
4 10-6
-5
t= 3.5x10 a.u.
2 10-6

0
-0.4 -0.2 0 0.2 0.4
x [a.u.]
 y
Fig. 7. Spatial distribution of the virtual boson cloud a ^x a
^x ðtÞ around a spatially localized fermion at 10 times
t ¼ n  3:5  105 a:u: for
D n ¼
E 1; . . . ; 10. For comparison, the inset shows on a logarithmic scale the corresponding snapshots
^
^y b 3
of the fermion density b x x ðtÞ. The coupling strength is c ¼ 10 .

7. Summary and outlook

One of the long-term goals of this project is to gain a better understanding of how two fermions can
interact with each other via the exchange of bosons. In order to obtain an intuitive picture we plan to
analyze interacting quantum field theories with space–time resolution. As a first step we provided a
computational framework for these investigations, and discussed how the Hilbert space can be dis-
cretized and truncated to make a numerical analysis feasible.
We have focused our initial attention on the single-fermion system. We showed that in order to
obtain numerically converged eigenenergies for low fermion momentum, not all bosonic momentum
states are necessary. It turns out that the most important numerical parameter is the maximum occu-
pation number of the k = 0 boson mode, while large occupation numbers for other modes are less rel-
evant. The numerical data were used to examine the validity of the Greenberg–Schweber model that is
based on the assumption that the fermions are at rest and therefore have identical energies. While this
model obviously cannot be used to provide any dynamical information about the motion of the fermi-
ons, its analytical predictions with regard to the properties of the energy eigenstates of the fully cou-
pled system are of remarkable accuracy. We thus expect that this model might give a good analytical
guidance for the planned time dependent analysis.
One reason for this unexpected accuracy of the GS-approximation even for large coupling constants
might be related to the fact that in each energy eigenstate the contributions of the nearest two neigh-
boring momentum states are most important. If the bare fermionic energies are not identical (as in the
GS-model) but equidistant (which is naturally the case for only two coupled states), it is also possible
to find an analytical expression for all energies in the single-fermion subspace [28]. The energies for
the coupled system depend only weakly on the bare-energy spacings. As a result both models predict
the two spectra quite well.
As an example for this numerical approach to predict a time evolution we examined the virtual bo-
son creation process from a bare fermion. We analyzed the Heisenberg equation of motion for the
occupation operators and found that the bosons are created and annihilated on a time scale that is
independent of the coupling strength. The time-average of the boson numbers scales similarly as
the distribution of the bare bosons in the energy eigenstate.
In agreement with the prediction the GS model, the zero momentum (k = 0) boson mode seems to
be most dominant in determining the eigenenergies of the coupled system. Even though the time evo-
 y
lution of the population in this mode, a ^0 ðtÞ, is determined entirely by the constant operator
^0 a
Rp b^yp b^p , this mode cannot really be decoupled from the dynamics because the phase of a^0 ðtÞ is still
determined by the other modes. It might be interesting to examine its relevance to the fermion–fer-
mion interaction. The GS model predicts that all energies are down-shifted by the same amount that is
282 T. Cheng et al. / Annals of Physics 325 (2010) 265–286

proportional to the square of the coupling constant. It might also be interesting to eliminate this global
shift and remove the k = 0 boson mode from the dynamics.
The dynamical dominance of the k = 0 boson mode as suggested in Fig. 4 is a little bit surprising in
view of the effective coupling strength. This effective coupling in the interaction part of the Hamilto-
p
nian (3.1) for each mode scales like g= ð2xk Þ. The numerical value for the ratio of the k = 0 to the k = 1
p
mode, ðx1 =x0 Þ, is only 2.17 for our parameters, suggesting that the k = 0 coupling is only double that
of the k = 1 mode.
The GS model predicts that two fermions can interact instantaneously via the Yukawa-like force.
While this might not directly contradict causality within the re-coil free model of the GS framework,
it might be interesting to examine the modification of this instantaneity for the more realistic system
where the fermions can actually accelerate as a consequence of the creation and annihilation of the
virtual bosons.
Past works [9,29,30] on coherent superposition states of electron–positron systems have led to
interesting spatial localization phenomena. In principle, it is possible that a high degree of entangle-
ment between fermionic and bosonic states could also show similar coherence features. Our numer-
ical approach can be easily generalized to study also resonant-like conditions in which due to a
suitable choice of the boson mass the energies xk can be made comparable to the difference between
two neighboring fermionic bare states, xk  jEpþ1  Ep j. It might be interesting to examine whether
the theoretical framework including the description of coherent resonance phenomena or field in-
duced energy shifts can be transferred from quantum optics [1–5] for these cases.
Our first studies of the spatial distribution of the virtual boson cloud around a bare fermion were
severely limited to a very small numerical grid. While it showed clearly the locality of this creation
process, the numerical limitation due to the finite spatial box did not permit us to examine the spatial
evolution for long times. The time evolution of the boson population from Section 6 suggests that the
bosons cannot escape to arbitrary distances from the location of the fermion [31]. The bosons have to
stay close to the fermion to become annihilated after a certain time. We plan to visualize this intrigu-
ing process via a more model based approach in which only the very essential Hilbert states are taken
into account to allow for larger numerical grids.

Acknowledgments

We enjoyed several helpful discussions with Drs. A. Di Piazza, C.C. Gerry, S. Hassani and E.V. Stef-
anovich. This work has been supported by the NSF. We also acknowledge support from the Research
Corporation.

Appendix A

Here we derive the analytical expression for the average number of bare fermions with momentum
^y b
p in the physical fermion state kQ ii, defined as q ðpÞ ¼ hhQ kb ^p kQ ii. The physical single particle state
Q p

with momentum Q is defined as kQ ii ¼ B b y kvacii. Using the definition of B

b Q as B
b Q ¼ U GS b
^Q U y [where
Q GS
U GS is defined in Eq. (4.1)] this simplifies to
h i
B ^y U y ¼ exp g R 2x3 1=2 a
b y ¼ U GS b ^yk  a

^k SðkÞ b ^y
^ y ¼ Db ðA1Þ
Q Q GS k k Q Q

^y ¼ b
where SðkÞ  exp½ko=oQ  is the momentum shift operator for fermions, i.e., SðkÞb ^y . As the boso-
Q Q þk
nic operators for different modes commute, we can apply the Baker–Campbell–Hausdorff theorem and
decompose the dressing operator D into three products:
h  1=2 y i h  1=2 i
D ¼ K exp g Rk 2x3k ^k SðkÞ exp g Rk 2x3k
a ^SðkÞ
a ðA2Þ
h  1 i
where the normalization factor K  exp g 2 =2Rk 2x3k . As a result, the state kQ i simplifies to
h  3 1=2 y i
^ y
kQ ii ¼ DbQ kvacii ¼ K exp g Rk 2xk ^ y
^k SðkÞ bQ kvaci. In order to compute qQ ðpÞ  hhQ kb
a ^
^y b
p p kQ ii
 T. Cheng et al. / Annals of Physics 325 (2010) 265–286 283

^ p Db
we have to find the norm of the state b ^y kvacii by expanding the exponential operator around g = 0
Q
we obtain
. h ij h i
^ p Db
b ^p KRj ðgÞj j! R 2x3 1=2 a
^y kvacii ¼ b ^yk SðkÞ b ^y kvacii ¼ Kb ^y kvacii  Kg=1! R 2x3 1=2 a
^p b ^yk b^
Q k k Q Q k k
h   1=2
i2
 p SðkÞb^y kvacii þ Kg 2 =2! R 2x3 ^yk SðkÞ b
a ^y kvacii þ ¼ Kkvaciidp;Q
Q k k Q
h   1=2
i h    3 1=2 y z i
3 2 3 1=2
 Kg=1! Rk 2xk kkiidp;Q k Kg =2! Rk Rk 2xk 0 2x k 0 ^k a
a ^ 0 kvaciidp;Q kk0 þ
k
ðA3Þ

0 p
If we separate the diagonal term ðk ¼ k Þ from the double sum and use a ^yk a
^y 0 kvacii ¼ ð1 þ ð 2  1Þ
k
  1=2 
0 2
 3
 1=2 0
dk;k0 Þkk; k ii, the second order term in g simplifies Kg =2! Rk Rk0 2xk 2x3k0 kk; k iidp;Q 2k . As
^p Db
the last step, we compute the norm of the state b ^y kvacii and obtain
Q

h  3 1 i
^ y Þy b ^ ^y
^y b 2
Rj¼0 g 2j =j!F j ðQ ; pÞ
hhvackðDb Q p p DbQ kvacii ¼ exp g Rk 2xk ðA4Þ

where the coefficients are independent of g

Appendix B

Here we derive the analytical expression for the average number of photons with momentum k in
the physical fermion state kQ ii, defined as qQ ðkÞ ¼ hhQ ka ^yk a
^k kQ ii. This quantity is equivalent to the
norm of the state ^
a k kQ ii. We first use the bosonic commutator relationship for an arbitrary function
  y   y 
y  
f, a^k ; f a ^k ¼ of a ^k ^k kQ ii ¼ g 2x3k 1=2 kQ  ki:
^k and show that a
oa
h  1=2 y i
b y kvacii ¼ a
^k B
a ^k K exp g Rk0 2x3k0 ^ 0 Sðk0 Þ b
a ^y kvacii
Q k Q
 3 1=2 h  3 1=2 y i
¼ g 2xk SðkÞK exp g Rk0 2xk0 ^ 0 Sðk0 Þ b
a ^y kvacii
k Q
 3 1=2  3 1=2
¼ g 2xk SðkÞkQ ii ¼ g 2xk kQ  kii ðB1Þ
 
As a result, we obtain qQ ðkÞ ¼ g 2 = 2x3k . The total number of (bare) bosons in the state kQ ii can be
 
calculated as Rk g 2 = 2x3k . If we can evaluate this summation in the continuum limit ðD ! 0Þ, using
p R
xk ¼ ðm2 c4 þ k2 D2 c2 Þ we can approximate DRk x3 k  dkxðkÞ3 ¼ 2m2 c5 . As a result we obtain
Rk qQ ðkÞ ¼ c2 =ð2pm2 Þ.

Appendix C

^y b
Here we derive the expression for the joint spatial probability qQ ðx; yÞ  hhQ kb ^ ^y a
x xa y ^y kQ ii. Using a
similar approach as in Appendix B, we can compute equivalently the norm of the state b ^x a
^y kQ ii. Using
p ^x we express the operators in
^k expðiky2p=N x Þ and the corresponding expansion for b
^y  ð1= N x ÞRk a
a
their momentum representation

^x a
b ^p a
^y kQ ii ¼ ð1=Nx ÞRp Rk exp½iðpx þ kyÞ2p=Nx b ^k kQ ii
 3 1=2 ^p kQ  kii
¼ ð1=N x ÞRp Rk  g 2xk exp½iðpx þ kyÞ2p=N x b ðC1Þ
 
3 1=2
where we have used again a^k kQ ii ¼ g 2xk kQ  kii. Inserting the definition of kQ ii in terms of
the vacuum state, we obtain
 3 1=2 h i
^a ^ K exp g R 0 2x30 1=2 a
^y 0 Sðk0 Þ b^y kvacii
b x ^y kQii ¼ ð1=N x ÞRp Rk  g 2xk exp½iðpx þ kyÞ2p=N x b p k k k Q k

ðC2Þ
284 T. Cheng et al. / Annals of Physics 325 (2010) 265–286

If we expand the last term in this expression we obtain:

h  1=2 y i h i
exp g Rk0 2x3k0 ^ 0 Sðk0 Þ b
a ^y kvacii ¼ b
^y kvacii þ b^p g R 0 2x30 1=2 a ^y 0 Sðk0 Þ b^y kvacii
k Q k Q k k k k Q k
h  3 1=2 y i
0
þ ð1=2Þ g Rk0 2xk0 ^ 0 Sðk Þ
ak
h  3 1=2 y i
00 ^ y
 g Rk 2xk00
00 ^
ak00 Sðk Þ b Q k kvacii þ ðC3Þ

The action of the products of the type b ^y on the state kvacii can be performed,
^p b
q
^ ^ y
bp bq kvacii ¼ kvaciidpq , simplifying the expression (C3) to
 1=2 y ^y
kvaciidp;Qk  g Rk0 2x3k0 ^ 0b
ak Q kk0
kvaciidp;Q kk0
  1=2   1=2
þ ðg 2 =2ÞRk0 Rk00 2x3k0 2x3k00 ^y 0 a
a ^y kvaciidp;Q kk0 k00 þ ðC4Þ
k k00

In the original expression for b ^x a

^y kQ ii, we can now use the Kronecker symbols to eliminate the sum-
mation over p
 1=2 h  1=2 y
^x a
b ^y kQ ii ¼ ð1=Nx ÞRp Rk  g 2x3k exp½iðpx þ kyÞ2p=Nx K kvaciidp;Q k  g Rk0 2x3k0 ^ 0 kvacii
a k
 3 1=2  3 1=2 y y i
2
 dp;Q kk0 þ ðg =2ÞRk0 Rk00 2xk0 2xk00 ^ 0a
a ^ 00 kvaciidp;Q kk0 k00 þ
k k
 1=2 h  1=2 y
¼ exp½iQx2p=N x ð1=Nx ÞRk  g 2x3k exp½ikðy  xÞ2p=Nx K kvacii  g Rk0 2x3k0 ^ 0 kvacii
a k
 1=2   1=2
i
0 2 3 3 y y 0 00
 exp½ik x2p=Nx  þ ðg =2ÞRk Rk 2xk0
0 00 2xk00 ^ ^
ak0 ak00 kvaciiexp½iðk þ k Þx þ
ðC5Þ
 1=2  1=2
Let us now return to the norm of the state b ^x a ^y kQ ii. If we use Rk00 Rk0 2x300 2x3k0
k
 1
hhvacka ^y 0 kvacii ¼ Rk0 2x30
^k00 a
k k
and Rk0 Rk00 Rk00 0 Rk0000 f ðk0 ; k00 ; k000 ; k000 Þhhvacka^k0 a^k00 a^yk000 a^yk0000 kvacii ¼ 2Rk0
Rk00 f ðk0 ; k00 ; k0 ; k00 Þ the norm of the state simplifies to the final expression
  1=2 2 h  1  1
qQ ðx; yÞ ¼ ð1=N2x ÞRk  g 2x3k 
exp ½ikðy  xÞ2p=Nx  K2 1 þ g 2 Rk0 2x3k0 þ ðg 4 =2ÞRk0 Rk00 2x3k0
 1 i   1=2 2 h  1 i
 
 2x3k00 þ ¼ ð1=N 2x ÞRk  g 2x3k exp ½ikðy  xÞ2p=Nx  K2 xp g 2 Rk0 2x3k0
  1=2 2
2 
¼ ðg=Nx Þ Rk 2x3k exp ½ikðy  xÞ2p=N x  ðC6Þ

We can approximate this summation by using again the continuum limit ðD ! 0Þ; DRk exp
R
ðikz2p=N x Þx3
k  dk expðikzÞxðkÞ3=2 ¼ 23=4 p1=2 jzj1=4 K 1=4 ðcmjzjÞc7=4 m1=4 C1 ð3=4Þ, where K 1=4 ð. . .Þ
denotes the modified Bessel function of degree 1/4 and C (. . .) is the usual gamma function. We ob-
tain the final result
h i2
qQ ðx; yÞ ¼ ðg=ð2Nx ÞÞ2 1=D23=4 p1=2 jzj1=4 K 1=4 ðcmjzjÞc7=4 m1=4 C1 ð3=4Þ ðC7Þ

Appendix D

Here derive for the weak coupling limit an approximate solution for the temporal growth pattern of
 y
the average photon number in Eq. (5.3), a ^k ðtÞ  2g 2 =ðd2 xk Þ sin2 ðdt=2Þ, where the energy difference
^k a
is d  xk  ðEpþk  Ep Þ and p is the fermion momentum of the initial state kp; fnk ¼ 0gii. With our
Hamiltonian of Eq. (3.1), H ¼ H0 þ gV, the Heisenberg equation of motion for the bare boson number
 y 
 y   y 
operator, o a ^k ot ¼ i a
^k a ^k a ^k a
^k ; H ¼ ig a ^k ; V , leads to
 y   
^k =ot ¼ ig Rp ð2xk Þ1=2 b
^k a
o a ^y b^ ^ ^ y ^ ^y
pþk p ak  bp bpþk ak  gR1 ðD1Þ
 T. Cheng et al. / Annals of Physics 325 (2010) 265–286 285

We see that the operator a^yk a

^k is coupled to various products of time-dependent bosonic and fermionic
operators, and in principle an infinite set of coupled operator differential equations similar to Eq. (5.1)
needs to be solved. By taking higher temporal derivatives of Eq. (D1) and subsequently replacing the
derivatives on the right-hand side with the corresponding equations of motions for each operator, one
could try to combine the coupled first-order differential equations into a single equation for a ^yk a
^k of
higher order in time. If we take the second-order derivative of a ^yk a
^k , we obtain
 y  2
o2 a
^k a
^k =ot ¼ i½gR1 ; H0   i½gR1 ; gV
 
¼ g Rp ðEp  Epþk þ xk Þð2xk Þ1=2 b^y b^ ^ ^ y ^ ^y 2
pþk p ak þ bp bpþk ak  ig ½R1 ; V
2
 gR2  ig ½R1 ; V ðD2Þ
Similarly, we can obtain the third order equation via
 y 
3
o3 a ^k ot ¼ i½gR2  ig 2 ½R1 ; V; H
^k a
2
¼ ig½R2 ; H0   ig ½R2 ; V  g 2 ½½R1 ; V; H0   g 3 ½½R1 ; V; V ðD3Þ
As we are interested in the weak coupling limit, we omit the cubic term g 3 ½½R1 ; V; V. Furthermore, the
commutator ½R2 ; V as well as the double commutator ½½R1 ; V; H0  vanish. This simplifies the third
derivative to
 y 
3  
o3 a ^k ot ¼ i½gR2 ; H0  ¼ ig Rp ðEp  Epþk þ xk Þ2 ð2xk Þ1=2 b
^k a ^y b^ ^ ^ y ^ ^y
pþk p ak  bp bpþk ak ðD4Þ

If we assume that in the weak coupling limit only very small boson momenta k are important, the en-
ergy differences Ep  Epþk are very small compared to xk and the terms ðEp  Epþk þ xk Þ2 depend
therefore only weakly on p. In this limit, we factor the term ðEp  Epþk þ xk Þ2  d2 out of the sum
Rp , where we use the initial momentum p in d2 . The right-hand side can be now expressed in terms
of the right-hand side of Eq. (D2). This approximations lead to a remarkably simple equation that is
now completely decoupled from other time-dependent operators and contains only the term a ^yk a
^k
 y 
3    y 

o3 a ^k ot ¼ id2 g Rp ð2xk Þ1=2 b

^k a ^y b^ ^ ^ y ^ ^y 2
^ ^ ot
pþk p ak  bp bpþk ak ¼ d o ak ak ðD5Þ

From now on we consider the expectation value for the initial state kp; fnk ¼ 0gii. If we integrate
 y
2  y
this equation from t = 0 to t, we obtain o2 a ^k ot ¼ d2 a
^k a ^k a
^k þ hR3 i, where hR3 i denotes the initial


2
value at t = 0 of the operator o a y
^k ot that according to Eq. (D1) amounts to hig 2 ½R1 ; Vi, as the term
^k a 2

2
hi½gR1 ; H0 i is initially equal to zero. Evaluating this term we obtain hig ½R1 ; Vi ¼ g 2 =
D E
ðxk ÞRp1 Rp2 b ^y b ^ ^y ^ ^y ^ ^y ^
p1þk p1 bp2 bp3þk . As we have hhp; fnk ¼ 0gkbp1þk bp1 bp2 bp2þk kp; fnk ¼ 0gii ¼ dp1þk;p dp1;
2
p2dp2þk;p such that Rp1 dp1þk;p ¼ 1 and we obtain hig ½R1 ; Vi ¼ g 2 =xk . This reduces the third-order
equation (D5) to
 y 2  y
o2 a ^k =ot ¼ d2 a
^k a ^k þ g 2 =xk
^k a ðD6Þ

This (harmonic oscillator) equation can be solved leading to ^yk a
a ^k ðtÞ ¼ R3 cosðdtÞþ
R4 sinðdtÞ þ g 2 =ðxk d2 Þ, with new integration constants R3 and R4 . As the initial average boson number
in the state kp; fnk ¼ 0gii is zero, we obtain R3 ¼ 0, leading to the expression
 y
^k ðtÞ ¼ g 2 =ðxk d2 Þ½1  cosðdtÞ þ R4 sinðdtÞ. As the expectation value of the right-hand side of Eq.
^k a
a
 y
(D1) at t = 0 vanishes, o a ^k a
^k =ot ¼ 0, we obtain R4 ¼ 0, completing the derivation of the approximate
 y
solution a ^k ðtÞ ¼ 2g =ðxk d2 Þ sin2 ðdt=2Þ.
^k a 2

References

[1] L. Allen, J.H. Eberly, Optical Resonance and Two-Level Atoms, Dover, New York, 1975.
[2] L. Mandel, E. Wolf, Optical Coherence and Quantum Optics, Cambridge University Press, New York, 1995.
[3] P. Meystre, M. Sargent, Elements of Quantum Optics, Springer, Berlin, 1998.
286 T. Cheng et al. / Annals of Physics 325 (2010) 265–286

[4] W.P. Schleich, Quantum Optics in Phase Space, Springer, Berlin, 2001.
[5] C.C. Gerry, P.L. Knight, Introductory Quantum Optics, Cambridge University Press, Oxford, 2004.
[6] W. Greiner, B. Müller, J. Rafelski, Quantum Electrodynamics of Strong Fields, Springer, Berlin, 1985.
[7] See the article by I.P. Grant, J. Sapirstein, in: G.W.F. Drake (Ed.), Atomic Molecular & Optical Physics Handbook, AIP,
Woodbury, 1996.
[8] P. Krekora, Q. Su, R. Grobe, Phys. Rev. Lett. 92 (2004) 040406.
[9] P. Krekora, Q. Su, R. Grobe, Phys. Rev. Lett. 93 (2004) 043004.
[10] E.V. Stefanovich, Ann. Phys. 292 (2001) 139.
[11] E.V. Stefanovich, Renormalization and dressing in quantum field theory, 2005, hep-th/0503076.
[12] E.V. Stefanovich, Relativistic quantum dynamics: a non-traditional perspective on space, time, particles, fields, and action-
at-a-distance, Mountain View, 2004, physics.gen-ph/0504062v11.
[13] J. Schwinger, Phys. Rev. 82 (1951) 664.
[14] T. Cowan et al, Phys. Rev. Lett. 56 (1986) 444.
[15] I. Ahmad et al, Phys. Rev. Lett. 78 (1997) 618.
[16] J. Javanainen, J.H. Eberly, Q. Su, Phys. Rev. A 38 (1988) 3430.
[17] Q. Su, R. Grobe, Laser Phys. 15 (2005) 1381.
[18] N.B. Narozhny, J.J. Sanchez-Mondragon, J.H. Eberly, Phys. Rev. A 23 (1981) 236.
[19] J.H. Eberly, P. Lambropoulos (Eds.), Multiphoton Processes, Wiley, New York, 1978.
[20] M. Gavrila (Ed.), Atoms in Intense Laser Fields, Academic Press, Orlando, 1992.
[21] L.F. DiMauro, R.R. Freeman, K.C. Kulander (Eds.), Multiphoton Processes, AIP, New York, 2000.
[22] B. Piraux, K. Rzazewski (Eds.), Super-Intense Laser-Atom Physics, NATO Science Series, Kluwer Academic Publishers,
Dordrecht, 2000.
[23] For a review, see, e.g. S.S. Schweber, An Introduction to Relativistic Quantum Field Theory, Harper & Row, New York, 1962.
[24] O.W. Greenberg, S.S. Schweber, Nuovo Cimento 8 (1958) 378.
[25] R. Walter, Nuovo Cimento 68A (1970) 426.
[26] M. Kobayashi, T. Sato, H. Ohtsubo, Progr. Theor. Phys. 98 (1997) 927.
[27] For review, see, e.g. R. Machleidt, I. Slaus, J. Phys. G 27 (2001) R69.
[28] T. Cheng, Q. Su, R. Grobe, submitted for publication.
[29] P. Krekora, Q. Su, R. Grobe, J. Mod. Opt. 52 (2005) 489.
[30] M.V. Fedorov, M.A. Efremov, P.A. Volkov, Opt. Commun. 264 (2006) 413.
[31] D.J. Hearn, M. McMillan, A. Raskin, Phys. Rev. C 28 (1983) 2489.