IQuS@UW-21-090

Scalable Quantum Simulations of Scattering in Scalar Field Theory on 120 Qubits

∗
Nikita A. Zemlevskiy
InQubator for Quantum Simulation (IQuS), Department of Physics,
University of Washington, Seattle, WA 98195, USA.
(Dated: November 6, 2024)
Simulations of collisions of fundamental particles on a quantum computer are expected to have
an exponential advantage over classical methods and promise to enhance searches for new physics.
Furthermore, scattering in scalar field theory has been shown to be BQP-complete, making it a
representative problem for which quantum computation is efficient. As a step toward large-scale
quantum simulations of collision processes, scattering of wavepackets in one-dimensional scalar field
arXiv:2411.02486v1 [quant-ph] 4 Nov 2024

theory is simulated using 120 qubits of IBM’s Heron superconducting quantum computer ibm fez.
Variational circuits compressing vacuum preparation, wavepacket initialization, and time evolution
are determined using classical resources. By leveraging physical properties of states in the theory,
such as symmetries and locality, the variational quantum algorithm constructs scalable circuits that
can be used to simulate arbitrarily-large system sizes. A new strategy is introduced to mitigate errors
in quantum simulations, which enables the extraction of meaningful results from circuits with up to
4924 two-qubit gates and two-qubit gate depths of 103. The effect of interactions is clearly seen, and
is found to be in agreement with classical Matrix Product State simulations. The developments that
will be necessary to simulate high-energy inelastic collisions on a quantum computer are discussed.

∗ [email protected]
 2

CONTENTS

I. Introduction 3

II. Lattice scalar field theory 4

A. Qubit representation 6

III. Scalable variational circuits 7

IV. Quantum simulation of scalar field theory using scalable variational circuits 10
A. Vacuum preparation 10
B. Wavepacket creation 12
C. Time evolution 14
D. Summary of variational methods and error budget 16

V. Results 17

VI. Summary and outlook 22

Acknowledgments 22

A. Comparison of exact and variational time evolution 24

B. Optimization 25

C. Error mitigation 26

D. Variational parameters 29

E. Tables 35

References 45
 3

I. INTRODUCTION

The rapid growth of the computational capabilities of quantum devices over the last several years has fueled the
pursuit of quantum advantage. Despite the steady improvement of quantum hardware, both in qubit counts and error
rates, its inherent noise is a key challenge for the realization of general-purpose quantum computation, and is expected
to persist until the development of fault-tolerant devices. In the Noisy Intermediate-Scale Quantum (NISQ) regime [1],
the implementation of many quantum algorithms with provable speedups over their classical counterparts is out of
reach, due to their assumption of fault tolerance enabled by error correction [2]. Driven by the chase for practical
quantum advantage, known as quantum utility, sophisticated error mitigation techniques have been developed [3–5].
These methods are used to account for the effects of hardware noise without real-time error correction, enabling the
extraction of expectation values from noisy quantum states.
With these error mitigation techniques and quantum devices available today, digital calculations are becoming
possible beyond the capacity of brute-force or approximate classical methods [5–21]. Because an achievement of
quantum utility requires a computation done at scale, the development of scalable methods is critical to fully use the
power of state-of-the-art quantum computers. Variational algorithms have recently emerged as a promising avenue for
such computations [22–24]. These hybrid quantum-classical techniques use classical resources to determine the circuits
that are executed on quantum computers. Variational algorithms have been developed for a variety of tasks, famously
for ground-state preparation using the Variational Quantum Eigensolver (VQE) [25], but also for combinatorial
optimization [26], quantum machine learning [27], linear algebra problems [28], quantum simulation [29], and other
applications. The power of these algorithms stems from their ability to significantly compress circuits that would
otherwise be too deep for current hardware capabilities.
Quantum simulation has been identified as a uniquely-positioned computational task capable of demonstrating
scientifically-useful quantum advantage [30]. Unlike problems that assume error correction, quantum simulations have
a built-in tolerance of errors, coming from the nature of computing expectation values in a systematically-improvable
manner. As a result of this, error mitigation is an ideal method to deal with noise in the setting of quantum simulation
on near-term devices. Furthermore, quantum simulation is able to take advantage of physical properties of target
systems to increase the scale of the computations, while maintaining control over approximations and uncertainties.
For these reasons, quantum simulation is a well-suited problem to exercise quantum machines available today.
Quantum simulation holds the potential for predictive power in a wide class of fundamental physics problems
that are out of reach for classical methods. In particular, quantum simulations of high-energy inelastic scattering of
relativistic particles, such as those that take place in extreme cosmological events or particle colliders, promise to shed
light onto the structure and properties of the Standard Model and what may lie beyond [31]. Standard techniques in
quantum field theory predict the outputs of scattering events in regimes where perturbation theory is valid. Although
finite-volume lattice methods [32, 33] are able to compute non-perturbative quantities for specific processes (for
example, see Refs. [34–38]), studying arbitrary scattering events is challenging. Even in the perturbative regime,
conventional analytical methods only give static quantities, or results that are valid at infinitely late times. While
this approximation is valid for many quantities of interest, the dynamics of collision events are generally inaccessible.
Numerical lattice techniques suffer from an exponential increase of resources due to the sign problem when attempting
to simulate real-time dynamics. Furthermore, classical simulations of highly-inelastic scattering using methods based
on truncations of entanglement, such as Matrix Product States (MPS), are expected to fail because of the large
amount of entanglement that is believed to be generated in these events [39–44]. As such, the dynamics of scattering of
fundamental particles is a prime example of a physically-relevant problem that is challenging for classical computation,
making it a promising target for quantum simulation.
Much work has been done toward the goal of full-scale quantum simulations of scattering in theories and regimes
where analytic results are scarce and classical computations are intractable. Proposals of wavepacket creation techniques
[39, 45–49] and strategies to extract scattering observables from dynamics [50, 51] have complemented demonstrations
of quantum simulations of scattering in various models [52–63]. One of the simplest relativistic field theories in which
scattering may be studied is the scalar field theory in one dimension.1 This is the theory of coupled harmonic oscillators
at each point in space with a quartic potential that governs interactions between particles. Scalar and pseudoscalar
particles play key roles in many areas, including nuclear physics [65, 66], condensed matter physics [67, 68], high energy
physics [69–71], as well as cosmology [72] and physics beyond the Standard Model [73–75]. Scalar fields have also been
used as a venue to study phase transitions and spontaneous symmetry breaking [76–78]. Moreover, scattering in scalar
field theory has been shown to be BQP-complete [79], meaning all “quantum-efficient” problems may be solved by

1 While it has been proven that the scalar field theory is trivial in five spacetime dimensions, and believed to be trivial in four, the existence
of a non-trivial continuum limit has been rigorously established in the one- and two-dimensional cases [64]. The continuum limit can be
recovered from lattice simulations by studying systems with successively finer lattices and carefully tuning the system parameters as
functions of the lattice spacing.
 4

mapping them to quantum simulations of scattering with various initial conditions. As such, this “simple” theory is a
good sandbox for demonstrations of scalable quantum simulation methods. Following the seminal work of Jordan, Lee,
and Preskill (JLP) [79–81], there have been numerous efforts targeting aspects of quantum simulations of scattering in
scalar field theory [82–91].
In this work, variational algorithms on the basis of brickwall [92–102] circuits2 and other ansatze are used to
simulate the scattering of two particle wavepackets on IBM’s quantum computers. While the application of variational
algorithms to quantum simulations has been proposed and executed on devices in the past [103–107], this work serves
as the first example of a quantum simulation using physics-informed brickwall circuits at scale for state preparation
and time evolution. Furthermore, it is the first quantum simulation of scattering of relativistic particles. These proof-
of-concept simulations are a demonstration of the techniques that will be foundational for simulations of high-energy,
high-inelasticity scattering events in more complicated theories, where rich physical phenomena are expected. The
simulations proceed in several steps:
1. Vacuum preparation
2. Particle creation
3. Time evolution
4. Measurement
By incorporating the symmetries and physical structure of the scalar field theory into the circuit design, the simulation
is carried out in a fully-scalable way, unlocking the full capabilities of state-of-the-art devices. These techniques guide
the design of variationally-optimized circuits to minimize the circuit depth in order to maximize the utility of devices
available today. While previous quantum simulations have taken a hybrid approach, (e.g., state preparation was done
variationally, but time evolution was implemented using an algorithm with rigorous performance guarantees), this work
establishes the application of physics-informed variational algorithms to all parts of the simulation. The variational
methods developed in this work are expected to have broad relevance to simulating relevant physical models.
This paper is organized as follows. An overview of lattice scalar field theory in one dimension and its mapping to
qubits is given in Sec. II. The algorithm for implementing scalable variational circuits (SVCs) for a class of quantum
simulation problems is outlined in Sec. III. This algorithm may be seen as a general framework capable of compressing
physical simulation circuits at scale. Sec. IV describes how knowledge of the physics of the theory is used to determine
SVCs both for state preparation and time evolution. The errors stemming from these approximate circuit compression
methods are quantified. In Sec. V, SVCs are used to simulate scattering in scalar field theory using 120 qubits of
IBM’s superconducting quantum computer ibm fez. Both the free and the interacting cases are studied. A new error
mitigation strategy on the basis of simulations of vacuum evolution is introduced. With this technique, the results are
found to be in qualitative agreement with classical MPS simulations, and the effect of the interaction strength on the
nature of the collisions is clearly seen. Improvements and comments on the extensibility of the developed methods are
presented in Sec. VI.

II. LATTICE SCALAR FIELD THEORY

The Hamiltonian of a real scalar field ϕ in one dimension with periodic boundary conditions (PBCs) defined on a
lattice of L spatial sites is written as
L−1
X 1 2 1 2 2 1 λ
H= Π + m ϕj + (ϕj − ϕj+1 )2 + ϕ4j (1)
j=0
2 j 2 2 4!
≡ HΠ + Hϕ + Hkin + Hint .

Here m is the bare mass, λ is the coupling controlling the strength of ϕ4 interactions, Πj is the conjugate momentum
operator obeying the canonical commutation relation [ϕi , Πj ] = iδij , and the nearest-neighbor finite-difference operator
∇ϕ = (ϕj − ϕj+1 ) is used in the lattice representation of the kinetic term. The lattice spacing is set to a = 1 in this
work.

2 “Brickwall”,“brickwork”, and “hardware-efficient” are commonly used to refer to dense variational ansatze using native single- and
two-qubit gates. “Brickwall” is used to describe this class of circuits in this work.
 5

With PBCs, the eigenstates of H are labeled by their spatial momenta ki = 0, ± 2π 4π

L , ± L , . . . , +π. The vector of
⃗ = (ϕ0 , ϕ1 , . . . , ϕL−1 ) is related to the vector of spatial momentum eigenstates χ
fields ϕ ⃗ by a spatial Fourier transform:
⃗ where V = √1 ei⃗k·⃗j is the Fourier transform matrix. In the free theory (λ = 0), the energy of a single-particle
⃗ = V ϕ,
χ L
eigenstate with spatial momentum k is
 
2 2 2 k
Ek = m + 4 sin . (2)
2
The spectrum of H in the free theory with PBCs can be constructed by introducing creation and annihilation
operators
r r !
X
−ikj Ek 1
ak = e ϕj + i Πj . (3)
j
2 2Ek

This can be inverted to find expressions for ϕ and Π in terms of ak and a†k . Written in terms of the creation and
annihilation operators, H takes the following form:
 
X † 1
H= Ek ak ak + . (4)
2
k

The vacuum of the free theory can be written down analytically. In the ak Fock basis, the ground state of the
system can be written as a tensor product of the L harmonic oscillators χ
⃗ , each in its ground state (denoted by the
subscript 0):
|ψvac ⟩ = |ψχ0 ⟩0 ⊗ |ψχ1 ⟩0 ⊗ · · · ⊗ ψχL−1 0
. (5)
The eigenfunctions of the harmonic oscillator can be used to construct wavefunctions in the free scalar field theory.
The momentum-space representation of the nth harmonic oscillator eigenfunction is given by
  14
1 En 1 2
p 
⟨χ|n⟩ = √ e − 2 E n χ Hn En χ , (6)
2n n! π

where Hn is the nth Hermite polynomial. Since H0 = 1, the ground-state wavefunction of the scalar field theory in this
basis can be written as a product of Gaussians:

(detE)1/4 − 1 χ⃗ T E χ⃗
⟨⃗
χ|ψvac ⟩ = e 2 , (7)
π L/4
where E is the matrix consisting of the energies in Eq. (2) on the diagonal. The position-space wavefunction can be
found using the Fourier transform χ ⃗
⃗ = V ϕ.
Similar to the vacuum, wavepackets in the free theory can be specified analytically. Using the vacuum state of
Eq. (5), the operator a†k defined in Eq. (3) creates a single-particle excitation with momentum k:

|k⟩ = a†k |ψvac ⟩ = |ψχ0 ⟩0 ⊗ · · · ⊗ |ψχk ⟩1 ⊗ · · · ⊗ |ψχL−1 ⟩0 . (8)

A general single-particle state is a superposition of these excitations:
X
|ψwp (p)⟩ = N g(k, p)|k⟩, (9)
k

where g(k, p) is an envelope in k-space that specifies the form of the state, and N is a normalization constant.
1 2 2
1
A localized wavepacket state can be created by choosing a Gaussian profile g(k, p; σp ) = √2πσ e− 2 (k−p) /σp . The
p
wavefunction of a Gaussian wavepacket with momentum p and spread σp is found using Eq. (6) to be
1 T
X p 
χ|ψwp (p; σp )⟩ = N e− 2 χ⃗ E χ⃗
⟨⃗ g(k, p; σp ) H1 Ek χk . (10)
k

This specifies a wavepacket centered at j = 0; it can be moved elsewhere using a spatial translation operator. As before,
the position-space wavefunction is found by the Fourier transform χ ⃗ Multiparticle states can be constructed in
⃗ = V ϕ.
a similar fashion.
 6

Time-evolved observables in the free theory can be computed analytically by expanding the state in the appropriate
basis of multiparticle excitations. For example, the time evolution of a single-wavepacket state can be written as
|ψwp ⟩(t) = k e−itEk g(k, p)a†k |ψvac ⟩. By rewriting ϕ2j in terms of ak and a†k , ⟨ϕ2j ⟩(t) can be obtained using Wick
P

contractions. In the interacting (λ ̸= 0) theory, Ek , a†k , and |ψvac ⟩ are all functions of λ. In cases where 4!
λ
is small,
4
observables can be obtained by treating the ϕ term as a perturbation.
In the context of quantum simulation, the interacting theory can be studied provided that the required states can be
prepared. JLP introduced a procedure to create interacting particle wavepackets from their free counterparts [80, 81].
By the adiabatic theorem, if the ϕ4 interactions are turned on in a manner that is slow enough for states to adjust,
then eigenstates in the free theory will flow into the corresponding eigenstates in the interacting theory. However,
because wavepackets are superpositions of eigenstates, they will propagate and broaden during this adiabatic evolution.
To counter this, the adiabatic evolution turning on λ can be broken up into several steps, and the unwanted evolution
can be “undone” at each step by evolving backwards. Let s parameterize the interactions in the Hamiltonian

H(s) = Hϕ + HΠ + Hkin + sHint , s ∈ [0, 1]. (11)

The adiabatic turn-on is then implemented in N steps by interleaving forward time evolution with increasing s at each
step with reverse time evolution at fixed s:
N −1
( Z (n+1) !)    ( Z (2n+1) !)
Y t N 2n + 1 t 2N
Uadiabatic = T exp −i H(s)ds exp itH T exp −i H(s)ds , (12)
n=0
2 (2n+1)
2N
2N 2 Nn

where T is the time-ordering operator. With a proper choice of N and t depending on the parameters m and λ
[80, 81, 108], it can be shown that unwanted evolution can be sufficiently suppressed.

A. Qubit representation

At each lattice site j, the field takes continuous values ϕj ∈ (−∞, ∞). Since the Hilbert space at each j is infinite,
the field must be digitized to represent it on a quantum computer. The JLP digitization is used to map this system
onto qubits [80, 81]. Given a register of nq qubits, a maximum magnitude ϕmax is chosen,3 and the field at each point
in space is evenly discretized in 2nq steps of size δϕ in the range [−ϕmax , ϕmax ]. At each site j, the digitized field can
take on the values
2ϕmax
ϕj = −ϕmax + ℓδϕ , δϕ = , ℓ ∈ [0, 2nq − 1]. (13)
2nq − 1
Twisted boundary conditions in field space [82] are used to preserve the symmetry in the digitizations of ϕ- and
Π-space. The allowed conjugate momenta kϕ 4 are distributed symmetrically around 0:
 
π 1 2π
kϕ,j = − + ℓ − , ℓ ∈ [0, 2nq − 1]. (14)
δϕ 2 2 nq δ ϕ
The allowed state of the field at each site is represented by qubits using a binary encoding of the 2nq allowed values
from Eqs. 13 and 14:

|ϕj = −ϕmax + ℓδϕ ⟩ = |ℓ⟩, (15)

(e.g., |ϕj = −ϕmax + δϕ ⟩ = |1⟩ = |01⟩ for nq = 2).

This digitization is particularly convenient because the form of the operators in the Hamiltonian is simple when
expressed in terms of their action on qubits:
nq −1
ϕmax X ℓ
ϕj = − 2 Znq j+ℓ , (16)
2nq − 1
ℓ=0

3 There exists an optimal choice of ϕmax for a given nq as a result of the Nyquist-Shannon Sampling Theorem [109–111]. The optimal
values of ϕmax are determined numerically for small nq in Ref. [82], and analytically in Refs. [112, 113].
4 The conjugate momenta kϕ , which are used to construct the Πj operator, should not be confused with the spatial momenta k, which
label eigenstates of the Hamiltonian.
 7

FIG. 1: Mapping the one-dimensional lattice scalar field theory to qubits. For a system of L spatial sites, nq qubits
are used to represent the state of the field at each spatial site. The infinite-dimensional bosonic Hilbert space of ϕ at
each site is truncated to 2nq values. A maximum magnitude of the field ϕmax is chosen, and the values of the field in
increments of δϕ are assigned to the possible qubit states.

and in Π-space:
nq −1
π X ℓ
Πj = − n q 2 Znq j+ℓ . (17)
2 δϕ
ℓ=0
The representation of the lattice scalar field by qubits is shown in Fig. 1. This construction efficiently recovers the
low-energy physics in a way that is a systematically extendable to include higher-energy contributions. In Ref. [82] it
was shown that, aside from device noise and errors stemming from approximations to the time evolution operator, the
error in local observables decays double-exponentially with the amount of qubits used in the site-wise digitization.5
The JLP prescription gives an efficient implementation of the operators using quantum gates [80, 81]. Operators in
the Hamiltonian of Eq. (1) can be broken into two non-commuting groups: those that contain ϕ, and those that contain
Π. The operators made up of powers and tensor products of ϕ are diagonal in the ϕ basis, and are straightforward to
implement. Aside from factors of the identity, ϕ2j and (ϕj − ϕj+1 )2 contain only ZZ operators and can be implemented
using O(n2q ) two-qubit gates. Similarly, ϕ4j can be implemented with O(n4q ) two-qubit gates, as it has additional ZZZZ
terms for nq ≥ 4. Π2j is translated into gates by using a site-wise Quantum Fourier Transform (QFT) to switch into
the conjugate momentum basis where Π is diagonal. Using this, Π2j can be implemented with O(n2q ) two-qubit gates
as well.6
Conventionally, the time evolution operator is implemented in quantum simulations using first-order Trotterization:
  n  2 !
−itH (1) t t (1)
e = Utrot +O , Utrot (t) = e−it(Hϕ +Hkin +Hint ) e−itHΠ , (18)
n n
where the terms in the exponential are organized into commuting groups. Higher-order product formulas can be
(2n) (n) (n)
recursively constructed by symmetric combinations of Utrot via the formula: Utrot (t) = Utrot (t/2)Utrot (−t/2)† [116].
These formulas have more favorable error scaling with the step size t/n, at the cost of a larger gate depth. In this work,
(2)
Utrot is used as a point of comparison against variational methods. After two-qubit gate cancellations and reordering
terms, one second-order Trotter step can be implemented with a two-qubit gate depth of 20 for nq = 2.

III. SCALABLE VARIATIONAL CIRCUITS

The preparation of an arbitrary state on a quantum computer requires a number of gates that grows exponentially
with system size [117]. Fortunately, states in physical theories have properties, such as symmetries, that can be

5 That is, for a local observable, the percent error ϵ in the low-energy eigenvalues of the digitized observable compared to their undigitized
nq
counterparts scales as ϵ ∼ 2−2 . In fact, it was found in Ref. [82] that nq = 4 provides sufficient precision in local observables, so that
other sources of error dominate. Furthermore, it was also found that the ϕ basis is advantageous over other representations, such as
the basis of harmonic oscillator eigenstates, and strategies associated with improved actions that are often used in lattice QCD. This
advantage stems from the simplicity and efficiency of the implementations of operators in the theory (Eqs. (16), (17)).
6 Note that for a device with nearest-neighbor (NN) two-qubit gate connectivity, as is the case for most superconducting quantum
computers, there is an extra two-qubit gate overhead to convert circuits from arbitrary connectivity to NN. To reduce the two-qubit gate
depth, efficient methods for the decomposition of specific circuit elements to NN connectivity have been developed [15, 114, 115].
 8

utilized to simplify the state preparation task. This was studied in the context of using the locality of correlations
to systematically truncate exact scalar field theory state preparation circuits in Refs. [83, 85]. As was recently
demonstrated for state preparation in Refs. [10, 15], the combination of these ideas with variational methods proves to
be a powerful way to create compressed scalable circuits, while not sacrificing accuracy.
There always exists a unitary that takes one particular state |ψansatz ⟩ to another target state |ψtarg ⟩ in the Hilbert
space. Implementing this unitary is the goal of state preparation (e.g., |ψansatz ⟩ = |ψvac ⟩, |ψtarg ⟩ = |ψwp ⟩) and time
evolution (e.g., |ψansatz ⟩ = |ψwp ⟩, |ψtarg ⟩ = e−itH |ψwp ⟩). While many methods exist to approximately construct these
unitaries, they often produce circuits that are too deep to implement on currently-available hardware. Furthermore,
these methods are often general, even though only a single or several matrix elements are of interest in a simulation.
In this sense, the problem may be reduced from finding an implementation of an operator that properly transforms all
states in the Hilbert space, to that of finding an operator that implements |ψansatz ⟩ → |ψtarg ⟩. To take advantage of this
simplification and compress unitaries given by conventional methods, variational circuits can be used to approximate
the action of these unitaries in a systematically-improvable way.
Suppose a quantum circuit U (θ) ⃗ is parameterized by θ. ⃗ The task of the variational algorithm is to choose the
⃗
parameters θopt that implement |ψansatz ⟩ → |ψtarg ⟩ as closely as possible. To measure the quality of the prepared state
⃗ = U (θ)|ψ
|ψansatz (θ)⟩ ⃗ ansatz ⟩ against |ψtarg ⟩, the local infidelity is used:
 q 2

⃗
 √ ⃗ √
Id ρtarg , ρansatz (θ) = 1 − tr ρtarg ρansatz (θ) ρtarg . (19)

Here, ρtarg and ρansatz are the density matrices of the target and prepared states, respectively, and the subscript d
indicates that reduced states spanning d spatial sites are used.7 Since the global fidelity |⟨ψtarg |ψansatz ⟩|2 will become
arbitrarily small with increasing system size, the reduced state infidelity is used instead to measure the local quality of
|ψansatz ⟩. This quantity is particularly relevant for translationally-invariant states, or translationally-invariant states
with local perturbations, since these have a repeating local structure.
The mass gap and locality of interactions appearing in physical systems with Hamiltonians such as Eq. (1) enable
scaling up quantum simulations of these systems. Correlations in the ground states of such systems decay exponentially
with separation beyond the correlation length (i.e., for |i − j| > ξ, ⟨ϕi ϕj ⟩ ∼ e−|i−j|/ξ ) [118]. Here, the correlation length
ξ is defined to be the inverse of the gap (equivalently, the inverse of the mass of the lightest particle), ξ = 1/mparticle .8
Because of this, the finite-volume ground-state wavefunction for L ≫ ξ is exponentially close to its infinite-volume
form. This, in turn, means that local observables measured in the ground state agree with their infinite-volume values
up to corrections that are exponentially small in the system size, ⟨A⟩L = ⟨A⟩∞ + O(e−L/ξ ).
The exponential convergence of ground states with system size implies that the structure of the circuits that create
these states also converges exponentially [15]. In turn, this means that the circuits that act on these ground states
also exhibit the same exponential convergence. Importantly, this applies not only to ground states, but generally to
states whose structure does not change as the system size increases (provided the correlations decay as above). In
particular, fixed-width wavepackets of single particles meet these criteria due to their locality. This property enables
the determination of circuits that can be scaled to arbitrary system sizes. If the target state |ψtarg ⟩ is known for a
small system size, the equivalent state can be created in a larger system with the following algorithm to create SVCs:
⃗ The ansatz must have a structure that
1. Create a parametrized circuit ansatz that implements the unitary U (θ).
can be scaled up to arbitrary system sizes.
⃗
2. Optimize the parameters in U (θ):
(a) Prepare the target state |ψtarg ⟩ to a desired precision using a known method.9
(b) Initialize the initial state |ψansatz ⟩.
 
(c) Optimize the parameters θ⃗ by minimizing the local infidelity: min Id ρtarg , ρansatz (θ)
⃗ , with |ψansatz (θ)⟩
⃗ =
⃗
θ
⃗ ansatz ⟩.
U (θ)|ψ
3. Repeat steps 1, 2 for a set of increasing system sizes L. Fit an exponential to the parameters θ⃗ and extrapolate
to L of interest.

7 For pure states ρtarg and ρansatz (i.e., d = L), this expression reduces to IL = 1 − |⟨ψtarg |ψansatz ⟩|2 .
8 In the free theory, mparticle → m as δϕ → 0, while in the interacting theory, mparticle = mparticle (m, λ). mparticle =
̸ m in this work as a
result of the digitization used (nq = 2). This digitization may also induce interactions.
9 The classical methods to determine |ψtarg ⟩ for ground-state preparation, particle creation, and time evolution are given in Sec. II. Their
quantum counterparts are discussed in Sec. IV.
 9

FIG. 2: The algorithm to create variational circuits that can be scaled up to arbitrary system sizes [10]. A parametrized
ansatz with a scalable structure is chosen by considering symmetries of the system and properties specific to the target
unitary. The infidelity between the target state and the produced state is minimized to optimize the variational circuit.
This process is repeated for a series of increasing system sizes. The parameters are then extrapolated to an arbitrary
system size of interest.

This algorithm is shown schematically in Fig. 2. Depending on the convergence properties of the system being studied,
this algorithm can be carried out using simulations on a classical computer, or on a small register of a quantum
computer.10
⃗ ansatz ⟩, there
Because this algorithm creates a variational unitary by optimizing a single matrix element ⟨ψtarg |U (θ)|ψ
is no guarantee that the implemented U will behave as expected for different starting states. However, since only
initial states |ψansatz ⟩ with very specific properties are of interest in quantum simulations, this algorithm is ideal for
the simulation task. If an approximation to the unitary is needed that is valid for more general states, step 2c of the
algorithm can be modified by minimizing the average of the infidelity over several states instead:
N −1
1 X  ⃗ .

min Id ρtarg,i , ρansatz,i (θ) (20)
⃗
θ N i=0

The set of training states states {|ψtarg,i ⟩, |ψansatz,i ⟩} can be chosen to cover all the states of interest, or some
representative subset of them [96, 120, 121]. This process takes advantage of the simplification of the problem of
approximating a general unitary in the case of quantum simulation. Creating variational operators by optimizing
matrix elements makes this algorithm much more tractable both for classical and quantum computing, than if the
whole operator were optimized.
The power of this method lies within the choice of the ansatz circuit. The ansatz must be carefully chosen to have
enough degrees of freedom and expressivity to have a chance at replicating the action of its target unitary. Greedy
algorithms such as ADAPT-VQE have recently emerged as efficient and powerful ways to build the ansatz circuit
[122, 123]. This method, in the form it is used in this work, is summarized below.
1. Begin with an initial state |ψansatz ⟩, a target state |ψtarg ⟩, and a pool of operators {Oi } made up of commutators
of terms in the Hamiltonian. These operators respect some or all of the symmetries of the system.
2. Choose the next operator to append to the ansatz:
(a) For each ⃗
 operator in the pool, optimize the local infidelity as a function of all parameters θ in the ansatz,
⃗
min Id ρtarg , ρprep (θ) .11
⃗
θ
(b) Append the operator that decreases the infidelity the most to the ansatz, and update all the parameters
|ψansatz ⟩ → e−iθi Oi |ψansatz (θ0 , θ1 , . . . , θi−1 )⟩ = |ψansatz (θ0 , θ1 , . . . , θi )⟩.
3. Repeat the step 2 until the required convergence criteria is reached, such as a combination of circuit depth and
infidelity. Use the parameters determined at the previous step as an initial guess for the next step.

10 If running on a quantum computer, methods such as those described in Ref. [119] may be used to estimate the infidelity.
11 ADAPT-VQE was originally developed for general problems where |ψtarg ⟩ is not known a-priori, so the optimization was implemented by
minimizing the energy to find the ground state.
 10

This algorithm corresponds to steps 1-2 of the SVC framework as one way to determine the ansatz and optimize the
variational parameters at a fixed system size. When used with SVC, this algorithm is known as SC-ADAPT-VQE
[10, 15]. Incorporating symmetries directly into the ansatz makes it more optimizable, and guarantees that the
resulting state obeys the symmetries of the system (provided the input state also has the symmetries). This method
and its numerous variants have been shown to perform extremely well for ground-state preparation. In particular, it
has been used for ground-state preparation and particle creation in the context of translationally-invariant systems
[10, 15, 123–125]. Impressively, it was also shown that although greedy methods may not find the best sequence of
operators, almost all other sequences perform substantially worse, which removes the need to optimize over all possible
sequences.
It is important to consider the tradeoff between ease of optimization and circuit depth when selecting a variational
ansatz. The simplest shallow and expressive ansatz is the brickwall ansatz, where each layer consists of parameterized
single-qubit rotations followed by entangling gates. Compared to ADAPT-VQE, this approach trades the gate depth
on the quantum device for the classical computational complexity involved in optimizing a large set of parameters for
a set of circuits of increasing size. Optimizing a circuit with many parameters is a computationally intensive task
that is susceptible to the well-known problem of barren plateaus [126, 127]. Properties of the target state or target
operator can be used to simplify the ansatz structure and reduce the number of parameters that need to be optimized.
To ease the classical optimization task, new brickwall layers can be added and optimized iteratively using the same
algorithm as described above.
As such, finding the right balance between circuit depth and structure is imperative to create a trainable variational
operator that mimics the target unitary with a shallow circuit. Using the physics of the system and the operator
being approximated to inform the design of the ansatz proves to be a powerful way to increase the performance of this
variational method [128–132].

IV. QUANTUM SIMULATION OF SCALAR FIELD THEORY USING SCALABLE VARIATIONAL

CIRCUITS

The scalable variational circuits procedure can be carried out on a classical computer or a small quantum computer.
Throughout this work, classical computational resources are leveraged to determine parameters that are extrapolated
to system sizes of interest to be run on quantum computers. This work uses nq = 2 qubits to represent the state of ϕ
at each site, for the theory with bare mass m = 1/2. The method of Ref. [82] is used to choose the cutoff ϕmax = 1.5
to minimize field digitization effects. The dynamics of wavepackets in the free (λ = 0) theory is compared to those
in the interacting theory with coupling λ = 2. For the parameters chosen, the correlation lengths ξλ of the free and
interacting theory are ξ0 = 2.4503 and ξ2 = 1.5817, respectively. These are determined, as in the previous section, by
computing the inverse gap for small but increasing system sizes L, fitting an exponential, and extrapolating to the
infinite-volume limit. The circuit elements that are used to compose the variational circuits throughout this section
are shown in Fig. 3. Details of the optimization strategies and choices are given in App. B. The resulting parameters
θ⃗opt for all components of the variational circuit are given in App. D.

A. Vacuum preparation

As a first step toward simulating a scattering process, the SVC framework is used to prepare the ground state of
the scalar field theory. Together with a variational input state, SC-ADAPT-VQE is used to prepare the vacuum of
the theory both in the free and interacting cases. Compared to other methods for preparing ground states, such as
Quantum Imaginary Time Evolution [134] or subspace methods [135], SC-ADAPT-VQE gives shallow circuits with
few variational parameters, making optimization easy and enabling implementation on modern-day devices.
In the free theory, |ψvac ⟩ defined in Eq. (7) can be used as |ψtarg ⟩. Approximations to the vacuum of the interacting
theory can be found analytically using standard perturbative methods for small λ as is discussed in Sec. II. Fortunately,
for the chosen system parameters m and λ, the variational parameters of the circuits initializing these states show
convergence for small-enough system sizes where exact diagonalization (ED) is viable. Taking advantage of this,
interacting ground state determined from ED is used as |ψtarg ⟩.
It is important to choose a good initial input for the SC-ADAPT-VQE algorithm. There has been recent work
using MPS or other ansatze as input to variational algorithms to improve their performance [136–138]. Time reversal
symmetry (implying the reality of the ground-state wavefunction), ϕ → −ϕ symmetry, and translational invariance can
be used to choose a judicious input state to the greedy algorithm. The circuit to initialize the input state, shown in
blue in Fig. 3a, uses one variational parameter to create an arbitrary real, symmetric state on each spatial site. This
product state can be viewed as the vacuum |ψevac ⟩ of the theory with the correlation matrix K = V † EV truncated to
 11

FIG. 3: The circuit elements that are used for state preparation and time evolution in simulations of scattering. (a):
The circuit elements used to prepare the vacuum. The input state to SC-ADAPT-VQE is a real, symmetric tensor
product state over the spatial sites prepared by a single-site variational operator (blue). The circuit implementing
e−iθO1 (red) is used to couple neighboring sites. The symmetric version of the QFT defined in Ref. [82], SQFT, is used
to account for the symmetric digitization of ϕ and Π around 0. (b): The translationally-invariant circuit to prepare
the vacuum. One layer of SC-ADAPT-VQE is Trotterized into separate terms coupling even-odd and odd-even spatial
sites. This circuit has depth 25 and 3 variational parameters (one for the input state, and one for each application
of O1 ). (c): One layer of the brickwall circuit that is used to create particle wavepackets with negative momentum
(indicated by the up arrow) on top of the vacuum. To create the corresponding positive-momentum particles, the
circuit is flipped and site-wise SWAP gates are added. Each layer has 20 parameters and is depth 2. The building block
of this circuit is the variational gate introduced in Ref. [133]. (d): The circuit elements that are used to implement
time evolution. Both single- (cyan) and two-site (yellow) operators are used, mimicking the terms present in the
present in the Hamiltonian of Eq. (1). The building block of these circuits is the same as (c). (e): Two layers of the
translationally-invariant circuit used to implement time evolution. Each layer has depth 2 and 12 parameters. Single-
and two-site operators are applied in an iterative fashion.

only contain elements on its diagonal:

1
⃗ ψevac ⟩ = N e− 2 ϕ ⃗ T diag(K)ϕ
⃗
⟨ϕ| . (21)

By choosing this state, or a state like this, the variational algorithm only has to build out the correlations between
spatial sites.
The Hamiltonian of the theory is real, so it must have real eigenstates. By using exponentials of commutators
of terms in the Hamiltonian for the operator pool, SC-ADAPT-VQE ensures that only real unitaries are used, and
maintains the symmetries of the system. The lowest order operator in ϕ and Π that acts over more than P one site and
is not included in the Hamiltonian is derived from the commutator of terms in Hkin and HΠ , O1 = 12 j [ϕj ϕj+1 , Π2j ].
Because of the chosen form of Πj , the canonical commutation relation [ϕi , Πj ] = iδij only holds in the limit δϕ → 0
(i.e., only for the undigitized operators ϕj , Πj ). Nevertheless, this relation can still be used at nonzero δϕ to simplify
the operators in the SC-ADAPT-VQE pool because the requisite symmetries are maintained. This makes operators
such as O1 easier to implement using quantum gates. Using this, O1 can be rewritten as
X
O1 = i Πi ϕi+1 , (22)
i

which can readily be implemented on a quantum device using the techniques described in Sec. II A. To build out longer
 12

3 × 10−2 One layer L=8 λ = 0, one layer

10−1 1.3
λ=0 λ=0 θ
λ=2 λ=2 θ∞
2 × 10−2 1.2
Exp. fit

θ1
I4

I4
10−2 1.1

10−2
1.0

10−3
6 8 10 12 0 1 2 3 4 2 4 6 8 10 12
L Layers L
(a) (b) (c)

FIG. 4: Convergence of the wavefunction and state preparation parameters in the vacuum preparation circuit shown
in Fig. 3b for both the free and interacting theory. (a): I4 taken over a region of four contiguous lattice sites (eight
qubits) as a function of system size. One layer of SC-ADAPT-VQE is used. (b): I4 as a function of number of
SC-ADAPT-VQE layers used in the ansatz for a L = 8 system. Each layer uses a longer-range operator Od given in
Eq. 23 to create correlations that span more lattice sites. (c): An example of the exponential convergence of one of
the angles parameterizing the λ = 0 vacuum preparation circuit with one layer of SC-ADAPT-VQE, as a function of
system size. The remainder of the angles are given numerically in App. D.

correlations, the separation between the sites that O1 acts on can be increased:
X
Od = i Πi ϕi+d . (23)
i

This operator can similarly be obtained from the commutator of Π2j with a longer-range verion of Hkin . The pool used
for SC-ADAPT-VQE consists of the operators {Od } for a range of increasing d.
The circuit implementing e−iθO1 on a pair of sites is given in red in Fig. 3a. Because O1 acts on all pairs of
neighboring sites, it must be Trotterized into an operator acting on even-odd sites, and an operator acting on odd-even
sites. This is shown in Fig. 3b, where there are two layers of the O1 circuit. This Trotterization breaks translational
invariance, so one variational parameter is used for each Trotter layer to partially mitigate this effect.
Figure 4 shows the convergence of the variationally-prepared vacuum. The local infidelity over four sites, I4 , is
used to measure the quality of prepared state, spanning roughly 2-3× the correlation length of the infinite-volume
vacuum. With just one SC-ADAPT-VQE layer, I4 can be seen to be non-increasing in Fig. 4a, meaning that the
local quality of the prepared function is not degrading with system size for both λ = 0 and λ = 2. The infidelity is
systematically improved by applying more layers of Od with successively larger d. The result of longer-range operators
creating correlations that span more lattice sites is seen in the plots of I4 in Fig. 4b. In both of these figures, it can be
seen that the λ = 2 wavefunction is converging faster as a result of the smaller correlation length in the interacting
theory. Figure 4c is a representative example of the convergence of the parameters with increasing system size. These
plots indicate that the L-extrapolation is applicable and the parameters to initialize the ground state for an arbitrary
system size can be determined using modest L, where classical simulation is possible.12
The circuit shown in Fig. 3b, consisting of the circuit preparing the symmetric input state and one layer of SC-
ADAPT-VQE, is used to initialize the vacuum. It has a two-qubit gate depth of 25, making it shallow enough to save
circuit depth for wavepacket preparation and time evolution. The extrapolated parameters θ⃗opt that implement the
vacuum preparation are given in Table V, and numerical I4 values are given in Table I.

B. Wavepacket creation

The creation of particles on top of the prepared vacuum requires initializing local excited states. JLP proposed a
method of creating wavepackets in the free theory by applying unitary evolution under a†k and ak defined in Eq. (3),

12 If the SVC method is run using classical computation, the correlation length of ground states that can be prepared with the presented
method is limited by the maximum number of qubits that can be simulated classically.
 13

λ = 0 continuum λ = 0 lattice L = 8, nq = 3
λ = 0 lattice L = ∞ λ = 2 lattice L = 8, nq = 3

1.0 100
2.5 nq = 2
λ=0
0.5 λ=2
2.0

10−1
Ek

vk
0.0

I3
1.5

1.0 −0.5

0.5 −1.0 10−2

−π −π/2 0 π/2 π −π −π/2 0 π/2 π 1 2 3 4 5 6
k k Layers
(a) (b) (c)

FIG. 5: (a): The dispersion relation Ek and (b): the group velocity vk for m = 1/2 and λ = 0, 2. Exact computation
is possible in the continuum (faint solid lines) and in the free lattice theory (dashed lines). By exactly diagonalizing
the Hamiltonian of Eq. (1) in the single-particle sector at a fixed lattice size, it is possible to keep track of the lowest
eigenvalues as a function of λ, matching them onto the exact values at λ = 0 to construct Ek and vk numerically (solid
lines with circle and square markers). The numerical computations are done on a lattice of L = 8 spatial sites with
nq = 3 qubits per site. Discrete gradients are used for the numerical computations of vk . (c): The convergence of the
prepared particle wavepackets occupying three spatial sites with momentum k = −π/3, as a function of number of
layers of the circuit from Fig. 3c. The local infidelity spanning the width of the wavepacket, I3 , is used to measure the
quality of the prepared wavefunction for λ = 0, 2. Compared to (a) and (b), (c) uses nq = 2 qubits to represent the
field at each site.

and then adiabatically turning on the interaction as described in Sec. II. Instead of this, SVC is used to compress
the wavepacket preparation circuit by taking advantage of classical computation. This step of the algorithm prepares
a wavepacket whose size remains fixed as the total system size increases. While this method does not allow for the
scalable determination of the variational parameters to prepare wavepackets of increasing size, it utilizes the scalability
in total system size, including vacuum regions.13 This is useful because a finer spacing in k-space increases the spatial
momentum resolution of the wavepacket, building a more accurate representation of wavepackets in the continuum.
To reduce the impact of lattice artifacts, the momenta of the wavepackets must be √ carefully chosen. The infinite-
volume lattice dispersion relation (Eq. (2)) mimics that of the continuum, Ek,cont. = m2 + k 2 , for small k. These are
shown by the faint solid and dashed lines respectively in Fig. 5a. The group velocity of a particle wavepacket with
momentum k is found in the infinite-volume lattice theory to be
∂Ek sin k
vk = =q , (24)
∂k m + 4 sin2
2 k
2

and in the continuum to be

∂Ek,cont. k
vk,cont. = =√ . (25)
∂k m + k2
2

These are plotted as the dashed and faint solid lines respectively in Fig. 5b. Ek and vk can be computed numerically
in the digitized theory for any value of λ by using ED and projecting onto each momentum sector. These numerically-
calculated values are shown in Figs. 5a and b by the solid lines with circle and square markers.
To simulate a high-energy collision, the highest momenta should be selected where the (digitized) lattice dispersion
relation is still close to that of the continuum. With L = 8 spatial sites, the resolution is poor and the digitized vk
only match the continuum values at small k. The agreement can be improved with finer resolution by increasing the

13 In this case, the classical computing capability limits the maximum size of the wavepacket.
 14

number of lattice sites. Finite-volume effects are reduced by using many lattice sites outside the interaction region.
Based on this and on considerations of circuit depth and spread in k-space, wavepackets with k = ±π/3 are chosen,
with spread σk = π/3.14 A final choice in the setup of the simulation is the initial location of the wavepackets. The
wavepackets should be well-separated at t = 0 so they can be approximately treated as asymptotic “in” states. While
this is strictly true only for infinite separation, the wavepackets are local, and their extent is fixed by the width of the
circuit preparing them. Taking advantage of the locality of the ϕ4 interaction, the wavepackets are initialized to be
separated by one spatial site. This construction may need to be modified for theories where interactions have less
locality.
Leveraging the analytic knowledge of the free theory, the wavefunction of the wavepacket |ψwp ⟩ defined in Eq. (10)
is used as |ψtarg ⟩ for the variational optimization. The adiabatic turn-on of interactions given by Eq. (12) is simulated
classically, producing interacting wavepacket wavefunctions that are then used for |ψtarg ⟩ in SVC for the interacting
theory. With this method, interacting wavepackets are prepared directly on top of the interacting vacuum. This way,
in addition to reducing the circuit depth, SVC also removes the necessity for implementing the adiabatic evolution to
turn on interactions.
Because the region where the desired wavepacket has support does not grow with increasing system size L, the
amount of qubits over which the variational circuit has to act is fixed by the spatial width of the wavepacket. Scaling
these circuits is simple and straightforward since the number of parameters and structure do not change with system
size. This is another example of using physical properties of the state, in this case locality of the wavepacket, to guide
the design of the variational circuit. One layer of the brickwall ansatz used for this step of the state preparation is
shown in Fig. 3c. The building blocks of this circuit are the variational gates defined in Ref. [133]. This block is
particularly useful because any product of gates from the universal gate set {CNOT, RX , RY , RZ } can be written in
terms of products of the block with different parameters. Each layer of the brickwall ansatz has a two-qubit gate depth
of 2 and uses 20 variational parameters.
Since spatially-separated particles are required for the “in” states of the scattering simulation, the wavepacket
creation circuits are executed for both wavepackets in parallel. The same set of parameters is used to initialize both the
left and right wavepackets to maintain symmetry. The circuit preparing the left (k = +π/3) wavepacket is created in
such a way that the resulting wavepacket is the mirror image of the right (k = −π/3) wavepacket. This is indicated by
the up (down) arrows for k = −π/3 (k = +π/3) respectively, seen in the full circuit shown in Fig. 8. Site-wise SWAP
gates are used to flip the qubit mapping in Eq. 13, increasing the two-qubit gate depth of this part of the circuit by six.
Figure 5c shows the convergence of the prepared wavepacket state as a function of the number of layers of the
brickwall ansatz of Fig. 3c. It is seen that the local infidelity spanning the wavepacket width, I3 , is systematically
improved by adding more layers. Four layers of the circuit shown in Fig. 3c are used to prepare the free and interacting
wavepackets.15 These brickwall circuits have two-qubit gate depths of 14 (including extra site-wise SWAP gates),
and 80 parameters in total. While it is possible to run an extrapolation of the wavepacket parameters, this is not
necessary. The angles parameterizing the vacuum creation circuits are observed to change on the scale of 10−4 − 10−3
for L = 10 − 12 (see Table V), which is at the precision tolerance of IBM devices. The extrapolated angles for the
vacuum preparation are used to prepare the equivalent vacuum state on a smaller system size for purposes of training
the wavepacket circuit. The vacuum preparation angles extrapolated to L = 60 are used in an L = 12 system to
optimize the wavepacket preparation parameters. The resulting parameters θ⃗opt are given in Table VI, and numerical
I3 values are given in Table I.

C. Time evolution

Scalable variational circuits can also be used to produce compressed time evolution circuits in place of non-variational
methods such as Trotterization, QDrift [139], or Linear Combination of Unitaries [140]. There has been much work
in recent years in this direction. The structure of the ansatze that have been used varies widely, from circuits with
parameterized Trotter terms [96, 99, 120, 141], to brickwall ansatze [104, 142–147]. The related fast-forwarding
techniques [148, 149] involve using variational circuits to approximately diagonalize a Hamiltonian to implement
evolution in a number of gates that is constant with t but scales with L because of the diagonalization step.
In this work, a translationally-invariant brickwall ansatz is used to compress the time evolution operator. Note
that while there is no guarantee that a shallow brickwall circuit can approximate the time evolution operator well,
time evolution under any Hamiltonian with local interactions is subject to Lieb-Robinson bounds [150]. Moreover,

14 The spatial extent of the wavepackets is fixed to three spatial sites in this work. Similar to the field digitization, choosing the extent of
the wavepackets in position and momentum space according to the Nyquist-Shannon Sampling Theorem will ensure a double-exponential
convergence in precision for future calculations.
15 Note that with the parameters set to zero, inserting CNOTs between layers of this circuit will collapse the circuit to the identity. This is
used in vacuum simulations to maintain the structure of the circuit.
 15

FIG. 6: One variational step of the time evolution circuit. It consists of six forward layers followed by six reverse
layers of the circuits of Fig. 3d. Single- and two-body terms are used in an alternating brickwall fashion, mimicking
the terms present in the system Hamiltonian.

there is a maximum value for the group velocity of excitations from the lattice dispersion relation (Eq. (24)) [51] so
it is reasonable to expect that low-depth brickwall circuits can reliably compress evolution under the Hamiltonian
of Eq. (1) for sufficiently early times. The ansatz is made up of layers of the circuit shown in Fig. 3e, arranged in a
symmetric way, seen in Fig. 6. Each variational step is made up of six forward layers and six reverse layers, with the
reverse layers having the same parameters as the forward layers. This structure is chosen to mimic the structure of a
second-order Trotter step. Each layer uses 12 parameters and has a two-qubit gate depth of two. The time evolution
of the two-wavepacket state e−itH |ψ2wp ⟩ on an L = 12 system is used as |ψtarg ⟩ to determine the variational circuits.
Similar to the wavepacket preparation step, the L-extrapolation is not necessary for time evolution. However, in
this case this simplification is only valid for early times, because the time evolution circuits act on the whole lattice.
Since the vacuum is not prepared exactly, and the time evolution is not implemented exactly, the vacuum regions
will evolve in time [20]. This evolution will be captured by the variational optimization procedure, up to finite-size
effects that scale as O(t/L).16 To get around this limitation, the optimization can be run on larger systems for later
times, or an L-extrapolation can be implemented. Furthermore, propagating particles in a scattering simulation show
more drastic effects from finite-size effects, such as wrapping around the lattice due to PBCs and re-scattering. This
also necessitates running the optimization for later simulation times on a larger lattice. While this is a limitation of
the method in general, the scattering simulations are set up to require minimal time before scattering is observed
(see Sec. IV B for details). This places a limit both on the time required in the simulation, as well as the size of the
interaction region that needs to be used for training the variational circuits.
One promising way to use variational circuits for time evolution without running into these finite-size effects is
to compress a single time evolution step, and use that step repeatedly in place of a Trotter step. Similar methods
are investigated in Refs. [99, 104]. While the resulting circuits will be less compressed than those produced by the
current method, they will be more scalable with L. Another interesting direction is using MPS simulations of the time
evolution operator for training [104, 144, 147]. This would remove the limitations placed on this method by finite-size
effects. In this setup, the time evolution of one or many states may be used for training, as described in Sec. III.
Methods that target time evolution operators in specific subspaces, such as restricted to the low-energy subspace, are
currently being developed [108, 145]. The combination of subspace methods with variational time evolution has the
potential to simplify the optimization task while also improving the quality of the produced circuits.
Figure 7 shows the convergence of the time evolution variational circuits, compared to second-order Trotter circuits
as a function of circuit depth. The plots show the infidelity as a function of two-qubit gate depth for the time evolution
of a single wavepacket on an L = 6 system. The Trotter time evolution circuits are constructed by adding more
Trotter steps as described in Sec. II A. Variational circuits fix the number of variational steps to two while increasing
the depth of each step by adding layers of the circuit elements in Fig. 3d following the structure of Fig. 6. The
results for t = 1 − 9 show that the brickwall circuits are able to consistently compress the Trotter circuits. This can
especially be seen at late times, where considerably deeper Trotter circuits are required to accurately implement the
time evolution operator, whereas the variational circuits show a modest growth in depth. Sharp dips in the curves of
both the variational and Trotter infidelities are results of finite-size effects, as well as of numerical optimization (for
the variational lines). A difference in the infidelity between λ = 0 (top row) and λ = 2 (bottom row) can be seen,
which is attributed to using fixed-depth circuits to describe states with different correlation lengths. These plots show

16 These effects are estimated and the vacuum evolution is used for error mitigation. See App. C for details.
 16

t=1 t=2 t=3 t=4 t=5 t=6 t=7 t=8 t=9

1.0
Trotter
I6 (λ = 0)

Var.
0.5

0.0
1.0
I6 (λ = 2)

0.5

0.0
0 150 300 0 150 300 0 150 300 0 150 300 0 150 300 0 150 300 0 150 300 0 150 300 0 150 300
CNOT depth

FIG. 7: The convergence of the time-evolved single-wavepacket state on an L = 6 system with λ = 0 (top row) and
λ = 2 (bottom row) for t = 1 − 9. The second-order Trotter time evolution (circles) is compared to the variational
steps (squares) as a function of two-qubit gate depth. Two variational steps consisting of an increasing number of
layers shown in Fig. 3d are used. The infidelity of the whole L = 6 system with the exact time-evolved state is used to
measure the quality of the wavepacket state as it propagates through the lattice.

that by increasing the number of layers (and so, the number of parameters), the variational ansatz is able to represent
the time-evolved state with increasing accuracy.
For the runs on the device, the number of variational steps is increased with the simulation time: t = {1, 2, 3} use
one variational step, t = {4, 5, 6} use two, and t = {7, 8, 9} use three. The number of layers in each step is kept fixed
to six forward and six reverse layers, as in Fig. 6. The steps are optimized separately for each simulation time to find
better quality solutions for the variational circuit. They are applied in a Trotter-like fashion, so that the number of
parameters is the same for all time steps. The time evolution portion of the variational circuit uses 72 parameters in
total. The parameters θ⃗opt that implement the time evolution are given in Tables VII-XV and the local infidelities I10
for each time step are shown in Table I.

D. Summary of variational methods and error budget

The choices in the design of the variational circuits and methods of determining parameters for a system of L = 60
lattice sites are summarized below.
• The vacuum |ψvac ⟩ is created using an input state preparation circuit and one step of SC-ADAPT-VQE, shown
in Figs. 3a and b. The circuits are trained on L = 2 − 12, and parameters are extrapolated to L = 60. This step
of the state preparation has two-qubit gate depth 25 and three variational parameters.
• The two-wavepacket state |ψ2wp ⟩ is initialized using two layered-brickwall circuits in parallel (one layer is shown
in Fig. 3c). One set of parameters is used for wavepackets with opposite momenta by spatially flipping the circuit
and introducing site-wise SWAP gates. These circuits are determined using the L = 60 vacuum parameters on
an L = 12 system, and the angles are not extrapolated. This step of the state preparation has a two-qubit gate
depth of 14 and 80 parameters.
• Time evolution for t = 1 − 9 is implemented using the translationally-invariant brickwall circuit shown in Fig. 6.
It has a second-order Trotter step-like structure, and the number of steps is increased with time. The circuits are
optimized using time evolution of the two-wavepacket state |ψ2wp ⟩. They are trained on a lattice of size L = 12
with the vacuum preparation parameters for L = 60 and the wavepacket preparation parameters determined in
the previous step. The parameters are not extrapolated. This part of the circuit has a two-qubit gate depth of
22⌈ 3t ⌉ and 72 variational parameters.
The systematic errors introduced by the approximations used in this work can be quantified by running exact
classical simulations using qiskit [151]. The local infidelity Id (Eq. (19)), which is observed to be independent of
system size as a result of the SVC framework, is used to quantify the local quality of the prepared state compared
 17

Id # of parameters Parameter values

Simulation step λ=0 λ=2
Vacuum preparation (d = 4) 0.0181 0.0071 3 Table V
Wavepacket preparation (d = 3) 0.0555 0.0550 80 Table VI
t=1 0.1680 0.1275 72 Table VII
t=2 0.3307 0.2024 72 Table VIII
t=3 0.5412 0.3157 72 Table IX
t=4 0.5692 0.3674 72 Table X
Time t=5 0.6930 0.3675 72 Table XI
evolution
(d = 10) t=6 0.6447 0.4214 72 Table XII
t=7 0.6106 0.4032 72 Table XIII
t=8 0.6087 0.3937 72 Table XIV
t=9 0.5932 0.4207 72 Table XV

TABLE I: Values of the local infidelity Id for all steps of the simulation algorithm for both the free and interacting
theories. For vacuum preparation, a range of 4 lattice sites is used, which is roughly 2-3× the correlation length of the
state. This number is uniform throughout the lattice because of translational invariance. For wavepacket preparation,
the reduced state of the wavepacket (occupying three spatial sites) is used. For time evolution, the reduced state on 10
spatial sites is used, which contains the region where the interaction takes place. The number of variational parameters
used for each step is shown in the third column. The table where the corresponding parameters are specified is given
in the rightmost column.

to the state of interest. The width of the region, d, over which the reduced state infidelity is computed, is chosen
to be representative of the given step of the algorithm. For vacuum preparation, a subsystem spanning lattice sites
covering several correlation lengths is used. Because of translational invariance of the vacuum, this region can be
located anywhere on the lattice, and Id is observed to be uniform throughout.17 For wavepacket preparation, the
wavepacket width is used. A subsystem spanning the area where the interaction takes place from t = 1 − 9 is used for
all steps of time evolution. The values of Id at each step of the simulation algorithm are shown in Table I.
It can be seen that the infidelities for λ = 0 are consistently larger than those for λ = 2. This is a result of the
interacting theory having a smaller correlation length than the free theory. Because circuits with equal numbers of
parameters and equal two-qubit gate depths are used to initialize and time-evolve states for both values of the coupling
λ, this difference in quality is expected. This is an indication that the circuit depth needs to be increased for states
with larger ξ in order to spread correlations over a greater number of lattice sites.
It should be noted that Id , as well as other measures on the wavefunction, is a conservative estimate of the quality
of observables calculated from these simulations. In fact, states with a modest overlap with the target state are seen to
reproduce observables in the target state quite well. Appendix A shows the similarity of observables computed using
exact and variational methods.

V. RESULTS

Using the SVC framework, the circuit shown in Fig. 8 is scaled up to L = 60 (120 qubits) and executed on ibm fez
to simulate the scattering of two wavepackets in the ϕ4 theory. nq = 2 qubits and a cutoff of ϕmax = 1.5 are used
to represent the state of the field at each spatial site with a bare mass of m = 1/2. Error mitigation is vital for the
successful extraction of observables from noisy quantum simulations [9]. An overview of the error mitigation methods
used in this work is given here, and the implementation details and choices can be found in App. C. Once the bare
circuits are created, several error mitigation layers are added. Dynamical decoupling [152, 153] is used to mitigate
idle errors and crosstalk between qubits. Pauli twirling (PT) [154] is added to all two-qubit gates (CZ for ibm fez)
to convert coherent errors to incoherent stochastic noise, after which a depolarizing noise model is assumed for each

17 Translational invariance is seen over pairs of sites. This is a result of the vacuum preparation scheme using only operators spanning two
neighboring sites. See Sec. IV A for a discussion.
 18

FIG. 8: The physics-informed SVC used to simulate wavepacket scattering in scalar field theory. The individual circuit
elements are defined in Fig. 3. First, the vacuum of the theory is created using SC-ADAPT-VQE. Wavepackets are
initialized on top of this vacuum using a local brickwall ansatz. Then, a translationally-invariant brickwall ansatz is
used to implement the time evolution.

# of variational # of
# of two-qubit Two-qubit # of TREX twirls # of shots
t steps PTs/layouts
gates gate depth (per circuit) (per circuit)
(per t) (per circuit)
1-3 1 2284 59 80 2 8000
4-6 2 3604 81 80 2 8000
7-9 3 4924 103 80 2 8000

TABLE II: Details of the quantum simulations performed on 120 qubits of IBM’s ibm fez superconducting quantum
computer. For a given simulation time t (first column), the number of variational steps is given by ⌈ 3t ⌉ (second column).
The total number of two-qubit gates and corresponding two-qubit gate depth are given in columns three and four
respectively, accounting for gate cancellations from different parts of the circuits and possible optimizations. Each
circuit is executed using a number of Pauli-twirled circuits, with each one being mapped onto randomly-chosen rotation
of the selected qubit layout; this number is given in column five. Measurement mitigation is applied using two TREX
twirls for each Pauli-twirled circuit (sixth column). The number of shots per circuit is given in the rightmost column.

observable of interest. Similar to PT for two-qubit gates, Twirled Readout Error eXtinction (TREX) [155] is used to
mitigate measurement errors. Observables are then estimated using Operator Decoherence Renormalization (ODR)
[10, 15, 156, 157], which involves running a “mitigation” circuit for every Pauli- and TREX-twirled “physics” circuit.
Circuits used for mitigation must be classically simulable, but have similar error profiles to the physics circuits. In
this work, circuits implementing the time evolution of the vacuum state e−itH |ψvac ⟩ are used as mitigation circuits.
Although the vacuum preparation and time evolution are implemented approximately and e−itH |ψvac ⟩ = ̸ |ψvac ⟩, the
vacuum time evolution can be determined classically via an exponential extrapolation. See App. C for details on the
classical calculation of this time evolution. Compared to methods that have been used in the past, it is found that the
vacuum evolution circuits more accurately reflect the noise in the wavepacket circuits. This is especially noticeable
for the case of brickwall circuits, where the gates are densely packed and noise spreads faster than in Trotter time
evolution circuits. Using the mitigation circuits, the depolarization parameters are computed using the ratio of the
measured outcome ⟨Oj ⟩meas to the expected noiseless outcome ⟨Oj ⟩true for each local observable Oj

⟨Oj ⟩meas ⟨ψvac |eitH Oj e−itH |ψvac ⟩meas

pj = = . (26)
⟨Oj ⟩true ⟨ψvac |eitH Oj e−itH |ψvac ⟩true
 19

ibm fez
9 1.0
λ=0
6
0.75
t

hφ2j i2wp − hφ2j ivac

3
0 0.5
MPS simulator
9
6 0.25
t

3
0.0
0
0 10 20 30 40 50 59
Spatial site j

(a)
ibm fez
9 1.0
λ=2
6
0.75
t

hφ2j i2wp − hφ2j ivac

3
0 0.5
MPS simulator
9
6 0.25
t

3
0.0
0
0 10 20 30 40 50 60
Spatial site j

(b)

FIG. 9: The time evolution of the vacuum-subtracted ⟨ϕ2j ⟩ for a L = 60 (120 qubits) lattice for (a) the free (λ = 0)
and (b) interacting (λ = 2) theories. Two wavepackets are initialized with opposite momenta (k = ±π/3) at t = 0
and centered at sites j = 28, 32. The particles travel toward each other, collide, and propagate as time goes on.
Error-mitigated results from IBM’s quantum computer ibm fez (top panels) are compared to classical noiseless MPS
simulations (bottom panels). The vacuum of the theory is built using variational circuits determined by SC-ADAPT-
VQE, while the wavepacket preparation and time evolution is implemented by variational brickwall circuits. In the
device data, the vacuum evolution is evaluated via a classically-determined exponential extrapolation. A broken-down
view for each time slice is shown in Fig. 10, and a zoomed-in comparison of the scattering region between the free and
interacting theories is given in Fig. 11. Details on the error mitigation strategies are given in the main text and in
App. C.

The pj are then used to adjust for the noise in the physics circuits using the same equation in reverse. They are also
used to filter out outlying measurements caused by local device imperfections. In addition to these methods, layout
randomization is used to drive the noise closer to depolarizing noise.
The two-qubit gate depths and counts, as well as other parameters of the runs on ibm fez for times t = 1 − 9, are
shown in Table II. Circuits with a maximum two-qubit gate depth of 103 using 4924 two-qubit gates (for t = 7, 8, 9)
are used. For time t, ⌈ 3t ⌉ variational steps are used (the variational step is defined in Sec. IV C and shown in Fig. 6).
For each time t, four circuits are executed: e−itH |ψ2wp ⟩ and e−itH |ψvac ⟩ for λ = 0 and λ = 2. 80 Pauli-twirled circuits
are run for each physics and mitigation circuit. For each Pauli-twirled circuit, two TREX twirled circuits are run for
measurement mitigation. Each circuit executed on the device is evaluated using 8000 shots.
In these simulations, fluctuations of ⟨ϕ2j ⟩ above the vacuum are measured on the quantum device.18 The results

18 Note that it is possible to efficiently measure ⟨Hj ⟩ using the same setup. Besides Π2j , all terms in the Hamiltonian of Eq. (1) are in the ϕ
basis, so their expectation values can be computed using measurements in the ϕ basis. To measure in the Π basis, a second set of circuits
can be run with the kϕ local Fourier transform appended. The results can then be added together to compute ⟨Hj ⟩. From this, it is
straightforward to estimate quantities that are accessible to analytic computations, such as cross sections computed using perturbative
quantum field theory.
 20

λ=0 λ=2
0.8 t = 1 t=1 ibm_fez
0.4 MPS simulator
0.0
0.8 t = 2 t=2
0.4

0.0
0.8 t = 3 t=3
0.4

0.0
0.8 t = 4 t=4
0.4

0.0
hφ2j i2wp − hφ2j ivac

0.8 t = 5 t=5
0.4

0.0
0.8 t = 6 t=6
0.4

0.0
0.8 t = 7 t=7
0.4

0.0
0.8 t = 8 t=8
0.4

0.0
0.8 t = 9 t=9
0.4

0.0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Spatial site j Spatial site j

FIG. 10: A detailed view of the vacuum-subtracted ⟨ϕ2j ⟩ for the scattering of two wavepackets, as shown in Fig. 9,
by time step. Noiseless classical MPS simulations (grey dashed lines) are compared to data from IBM’s quantum
computer ibm fez after error mitigation (points) for λ = 0 (left column) and λ = 2 (right column). In the device data,
the vacuum evolution is evaluated via a classically-determined exponential extrapolation. The error bars represent one
standard deviation of the bootstrap-resampled data.

from the device and from MPS simulations for ⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac for λ = 0 and λ = 2 are shown in Figs. 9a and
9b respectively. A breakdown of the results by time step is shown in Fig. 10, and numerical values for these runs
are given in Tables XVI-XXIV. The uncertainties in the results from the quantum computer are estimated using
bootstrap resampling. The expected results are determined using the qiskit [151] MPS simulator with a maximum
bond dimension of 100. For the circuits considered, it is seen that the results of these simulations are converged to
10−2 − 10−3 .
The scattering process is seen by examining ⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac in Fig. 9. The propagation of particles is clearly
identified as disturbances in the vacuum-subtracted ⟨ϕ2j ⟩ in the interaction region in the center (spatial sites 24-36).
These are indicated as bright spots on the plots. The wavepackets, centered to the left and right of the center of the
lattice at t = 0, are initialized with opposite momenta. They travel toward each other and collide in the center. After
 21

λ=0 λ=2
9 1.00

6
ibm fez

0.75
t

hφ2j i2wp − hφ2j ivac

0 0.50
9

6 0.25
MPS
t

3
0.00
0
24 26 28 30 32 34 36 24 26 28 30 32 34 36
Spatial site j Spatial site j

FIG. 11: A zoomed-in view of the interaction region of Fig. 9 highlighting the effect of interactions. The vacuum-
subtracted ⟨ϕ2j ⟩ in the free theory (left panels) is compared to the interacting theory (right panels). The top panels
show error-mitigated results from IBM’s quantum computer ibm fez, while the bottom panels show classical noiseless
MPS simulations.

the collision takes place, a light cone is seen to develop, indicating that the particles have propagated past each other
and are travelling away from the point of the collision.

The effect of interactions is clearly seen when comparing results in the interaction region between the free and
interacting theories as in Fig 11. In the free theory, the collision is seen to peak around t = 4, after which the particles
travel away from each other. When interactions are included in the theory, the peak of the collision is observed later
(around t = 5), and a time delay due to the interaction is visible. The difference between λ = 0 and λ = 2 can also
be seen at late times (t = 9), where the free particles can be seen having travelled one more spatial site than the
interacting particles. Because the ϕ4 potential is repulsive, particles with overlapping wavefunctions incur an energy
penalty. In this sense, the time delay can be viewed as the interacting particles encountering a potential step and thus
losing kinetic energy when their wavepackets overlap. As a result, the particles spend a longer time in the central
region where they are able to interact. It is expected that this effect will be even more prominent at greater interaction
strengths.

The results in Figs. 9, 10, and 11 show that by using a combination of variational circuit methods (SC-ADAPT-VQE
and brickwall), together with error mitigation techniques, quantum computers are able to simulate scattering processes
of interacting particles. The scalability of each step of this simulation algorithm enables the simulation of collision
events on arbitrarily large system sizes in a systematically-improvable manner. It should be noted that the observation
of ⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac = 0 in the vacuum regions is a nontrivial result, as it requires the wavepacket simulations from
the quantum computer to agree closely with extrapolations from classical computations. In Figs. 9 and 10, the data
from the quantum computer is seen to qualitatively agree with the MPS simulations. Increasing levels of noise (as
well as uncertainties) can be seen in the results for later times, which is expected as a result of deeper circuits. There
exist points where the error-mitigated results still disagree from the MPS data by several standard deviations (see
Fig. 10). This is particularly noticeable in the interaction region during the collision of the particles, but occurs in
vacuum regions as well. This highlights a weakness of the chosen error mitigation method: because the state chosen for
error mitigation is e−itH |ψvac ⟩, its evolution does not capture potential changes to wavepackets. Because the noise is
state-dependent, these mitigation circuits cannot mimic the noise in the circuits simulating scattering perfectly. This
reflects the simplicity of the assumed error model, and is an area where much improvement is expected in the future.
Although errors persist even after the application of error mitigation, it is clear that quantum computers today are
capable of probing scattering events, potentially shedding light on inelastic collisions in more complicated theories in
the future.
 22

VI. SUMMARY AND OUTLOOK

Quantum simulations hold the potential of reliably uncovering the dynamics of complex inelastic collision events,
such as those that take place in extremely dense and hot environments. As a first step toward such computations, this
work simulates elastic scattering in one-dimensional lattice scalar field theory. Scalable variational quantum algorithms
are developed for both state preparation and time evolution, paving the path for future simulations at larger scales.
The simulation begins by using a translationally-invariant SC-ADAPT-VQE ansatz to prepare the vacuum of the
theory. Wavepackets representing particles are then created on top of the vacuum using a brickwall circuit, and a
translationally-invariant brickwall ansatz is used to implement time evolution of the system. The scalability of all
circuits in these simulations is a result of the physical properties of states in the system, namely their translational
invariance, locality, and the presence of a mass gap. The circuits are determined classically on modest system sizes,
and extrapolated to arbitrarily large system sizes to be run on quantum devices. These methods are used to simulate
the collision of two wavepackets using 120 qubits of IBM’s 156 qubit superconducting quantum computer ibm fez.
The use of error mitigation and suppression techniques is crucial for the extraction of results that qualitatively agree
with classical MPS simulations. Signatures of scattering and effects of the interaction strength are clearly identified in
the vacuum-subtracted expectation value of the field determined from data from the quantum computer.
Overall, the utility of variational quantum algorithms in systematically compressing circuits is clear; the specialization
of these methods to quantum simulations enables the use of physics to provide further performance improvements.
These newly-developed scalable variational methods, together with existing techniques and new modifications to error
mitigation strategies, enable the first simulation of a relativistic particle scattering event on a quantum computer.
Furthermore, these results establish the first demonstration of the extraction of qualitatively correct observables from
quantum devices using variational brickwall circuits at scale. This work establishes variational algorithms as tools that
are useful for the compression of all parts of a quantum simulation. In particular, it is demonstrated that the problem
of barren plateaus that plagues many variational quantum computations is reduced through knowledge of the system
being simulated (e.g., its symmetries and mass gap). Efficient variational circuits are determined by using physics to
remove redundancies in the circuit parameterization. The development of variational circuit compression methods
that are both scalable and amenable to error mitigation is useful for minimizing noise in the quantum device through
reduced two-qubit gate depth. These methods and results constitute an important stepping stone toward the quantum
simulation of inelastic scattering processes at high energies, where known classical methods are expected to fail.
Improvements to the optimization step of the variational algorithm, as well as more efficient ansatze, are expected in
the near future. This work uses the most simple iterative gradient-based optimization method. Recent work [158, 159]
has shown that it is possible to incorporate more sophisticated optimization protocols to improve the quality of the
approximate quantum state. In parallel, ansatze for state preparation that take advantage of mid-circuit measurements
[11, 12, 160–162] can be immediately used in simulations today; the combination of these methods with variational
quantum algorithms is a promising direction. These modifications would allow for greater precision in variational
calculations, while maintaining the same circuit depth, or even reducing it.
Highly-inelastic scattering events are believed to generate a large amount of entanglement, and as a result are prime
candidates for demonstrations of quantum utility. Toward the simulation of inelastic collisions, several improvements
to the current algorithm will be necessary. Particles with energy higher than the two-particle threshold will need to
be created, requiring both a finer lattice spacing a, and more tightly peaked wavepackets in k-space. In turn, both
of these will require wavepackets that occupy a larger number of qubits. The methods presented will need to be
extended to the scalable creation of wavepackets of increasing sizes and their time evolution. The presence of a large
number of lattice sites occupied by the vacuum will continue to provide the resolution in lattice momentum that will
be necessary to observe inelastic effects. More qubits per site (i.e., increasing nq ) will be needed to achieve higher
precision in these calculations. All of these improvements will require circuits with greater two-qubit gate depths. The
scalability of the presented methods can also be applied by extrapolating the variational parameters to continuum
physics, through taking the continuum limit in the process of the variational optimization. As the capabilities of state
of the art quantum computers to realize large-scale system sizes improve, more sophisticated techniques for particle
detection will be necessary. Further in the future, the extension of these methods to include more complex interactions,
such as those described by non-Abelian gauge theories, will be required for the simulation of realistic particle collisions.

ACKNOWLEDGMENTS

The author thanks Roland Farrell, Marc Illa, Zhiyao Li, Henry Froland, Anthony Ciavarella, and Martin Savage for
helpful discussions and insightful comments. This work was supported in part by the U.S. Department of Energy,
Office of Science, Office of Nuclear Physics, InQubator for Quantum Simulation (IQuS) [163] under Award Number
DOE (NP) Award DE-SC0020970 via the program on Quantum Horizons: QIS Research and Innovation for Nuclear
 23

Science. This work was also supported, in part, through the Department of Physics [164] and the College of Arts and
Sciences [165] at the University of Washington. This work has made extensive use of Wolfram Mathematica [166],
python [167], julia [168], jupyter [169] notebooks in the conda [170] environment, IBM’s quantum programming
environment qiskit [151], and iTensor [171, 172] in this work. This work was enabled, in part, by the use of advanced
computational, storage and networking infrastructure provided by the Hyak supercomputer system at the University
of Washington [173]. This research was done using services provided by the OSG Consortium [174–177], which is
supported by the National Science Foundation awards #2030508 and #1836650. The author acknowledges the use of
IBM Quantum Credits for this work. The views expressed are those of the author, and do not reflect the official policy
or position of IBM or the IBM Quantum team.
 24

1.0
hφ2j i2wp − hφ2j ivac t=0 Exact t=1 t=2 t=3 t=4
0.8
Variational
0.6
0.4
0.2
0.0
1.0
t=5 t=6 t=7 t=8 t=9
hφ2j i2wp − hφ2j ivac

0.8
0.6
0.4
0.2
0.0
0 3 6 9 12 0 3 6 9 12 0 3 6 9 12 0 3 6 9 12 0 3 6 9 12
Spatial site j Spatial site j Spatial site j Spatial site j Spatial site j

(a)
1.0
t=0 Exact t=1 t=2 t=3 t=4
hφ2j i2wp − hφ2j ivac

0.8
Variational
0.6
0.4
0.2
0.0
1.0
t=5 t=6 t=7 t=8 t=9
hφ2j i2wp − hφ2j ivac

0.8
0.6
0.4
0.2
0.0
0 3 6 9 12 0 3 6 9 12 0 3 6 9 12 0 3 6 9 12 0 3 6 9 12
Spatial site j Spatial site j Spatial site j Spatial site j Spatial site j

(b)

FIG. 12: The vacuum-subtracted ⟨ϕ2j ⟩ for different times t = 0 − 9 computed exactly using matrix exponentiation
(dashed lines) compared to its approximation using SVC (solid lines) for (a) the free (λ = 0) theory and (b) the
interacting (λ = 2) theory. An L = 12 (24 qubits) system is used.

Appendix A: Comparison of exact and variational time evolution

In this appendix, noiseless simulations of an L = 12 system are used to compare exact time evolution in the digitized
theory with nq = 2 to time evolution approximated by SVC. Exact calculations are done using matrix exponentiation
applied to the t = 0 state prepared via SVC methods. Figure 12 shows the comparison for λ = 0 and λ = 2. The SVC
variational results can be seen to qualitatively track the exact time evolution, but degrade with time. The quality of
the λ = 0 variationally-determined observables is seen to be consistently worse than those for λ = 2, which is again
a consequence of the λ = 0 theory having a longer correlation length. Although the local infidelity spanning the
interaction region I10 is high (see Table I), the observables calculated from the SVC states still follow the exact values,
suggesting that fidelity is a rough bound on the quality of prepared variational states. Since quantum devices compute
observables, even states that have high infidelity may be used to approximately simulate the desired process.
In the last two λ = 0 plots for t = 8, 9, the SVC observables are seen to deviate significantly from the exactly-
computed observables. This is an indication that the circuit used to simulate these time steps is not deep enough and
more variational layers should be used. The number of steps is incremented at t = 4 and t = 7, and it can be seen in
Fig. 12 that the variational approximations get closer to the true values at these times, than for the preceding time
 25

t # of variational Initial guess t # of variational Initial guess t # of variational Initial guess

steps steps steps
1 1 - 4 2 t=2 7 3 t=5
2 1 t=1 5 2 t=4 8 3 t=7
3 1 t=2 6 2 t=5 9 3 t=8

TABLE III: The map specifying the parameters from completed variational optimizations used as initial guesses for
optimizations of circuits implementing later simulation times.

steps.

Appendix B: Optimization

Optimization of the vacuum and wavepacket preparation circuits proceeds in a straightforward way. The SC-ADAPT-
VQE ansatz has only three parameters, and its symmetry-preserving structure results in fast convergence to the true
vacuum. The locality of the wavepacket state enables the general brickwall circuit used in state preparation to function
well. Because of its expressivity, it is able to initialize the wavepacket with high accuracy using a small number of
layers. Both of these optimization steps are implemented in an iterative manner, where additional layers are added as
needed until desired criteria are met, such as a combination of maximum circuit depth and desired value of Id .
The time evolution circuit optimization is done in a sequential manner over the simulation times. Because of the
Trotter step-like nature of the ansatz, results from optimizations of previous steps are used as initial guesses for future
optimizations. For example, since t = 4 uses two variational steps, and t = 2 uses one step, the angles implementing
the t = 2 evolution are used as initial guesses for the t = 4 optimization. Table III specifies how the initial guesses
utilize parameters that have been optimized for past simulation times. The BFGS gradient-based optimization method
[178] is used, with a tolerance of 10−5 .
Gradient-based circuit optimization methods such as BFGS encounter challenges when the amount of parameters is
increased due to the problem of barren plateaus. Barren plateaus occur as a result of circuit overparameterization,
when gradients of the cost function with respect to the parameters become vanishingly small, preventing gradient-based
local optimization algorithms from effectively making progress.19 The barren plateau problem has been carefully
studied (for a review, see Ref. [127]), as well as methods to overcome it in special cases [179–183]. In the context of
the optimizations run to approximate the time evolution operator using the ansatz circuit of Fig. 6, examination of the
optimization landscape for small systems shows many flat regions. Despite this, it can be seen (e.g., in Fig. 7), that by
using knowledge of the physics of the system to design variational circuits, these circuits are able to capture the time
evolution well.
Initial guesses are a crucial ingredient to tame barren plateaus when using gradient-based optimization methods.
So-called “warm starts” to standard optimization methods have been investigated in the context of parametrized circuit
optimization [105, 184]. In this light, it may be beneficial to use gradient-free methods, such as those taking advantage
of the periodic dependence of the cost function on the parameters, to produce initial guesses for gradient-based
optimizers. The barren plateau problem can be further reduced by introducing more structure into the circuit, such as
through physical considerations as described in Sec. III. This can be viewed as a way to remove redundancy in the
circuit.
The ansatz for time evolution for a specific simulation time is also built up in a greedy way, iteratively adding layers
and using the parameters found for the previous layer as initial guesses. This way, with each layer of the variational
ansatz, the algorithm creates a more precise approximation to the true time evolution. An important property of the
circuit in Fig. 6 is that it implements the identity operator when all parameters are set to 0. Because of this, the
layer-wise iterative optimization method ensures that additional layers increase the quality of the solution, since the
optimizer can set all parameters to 0 if no other choice increases the overlap with the target state. This method helps
to address the problem of barren plateaus by providing optimized initial guesses for future layers of the circuit. This is
found to perform much better than optimizing all parameters from scratch, where most initial guesses land in a part of
parameter space with vanishing gradients. This method is used as a heuristic when selecting the number of layers per
variational step and the number of steps for each simulation time.

19 Interestingly, it was recently shown that the absence of barren plateaus indicates that there exists an efficient way to simulate the problem
classically [126].
 26

In addition to using variational parameters optimized for earlier simulation times as initial guesses, parameters
optimized for Hamiltonians with different λ may be used. It can be seen in Table I that the λ = 0 optimizations
consistently perform worse than the λ = 2 ones due to the larger correlation length in the free theory. Parameters
implementing the time evolution in theories with smaller correlation lengths may be used as initial guesses for those
with larger correlation lengths. For instance, the λ = 2 parameters may help drive the λ = 0 optimization to a better
region of parameter space and produce a higher quality result. Overall, it is expected that there will be improvements
to the optimization techniques, both based on the structure of the circuits, and on the hyperparameters and initial
guesses.

Appendix C: Error mitigation

A suite of error mitigation techniques is used in this work, enabling extraction of meaningful results from noisy data
from quantum devices. The central ingredient of the error mitigation strategy used in this work is ODR, which is a
local version of decoherence renormalization [156, 185, 186]. In this method, known noise-free results of circuits that
are easily simulated classically are compared to results of the same circuits from the quantum device. This ratio is
then used in a spatially-local (operator-wise) fashion to measure the noise that has impacted each local expectation
value. ODR assumes a depolarizingPnoise model. The standard method of PT is used to convert the coherent noise
in each circuit to Pauli noise, ρ → i pi Pi ρPi , where ρ is a reduced state used to measure a local observable. As is
standard for similar error mitigation strategies, depolarizing noise is assumed, which can be seen as an special case of
Pauli noise where all Pauli error terms are equally likely, ρ → (1 − p)ρ + p1. The error mitigation scheme works by
estimating the parameter pj for this channel for each local observable Oj .
In this work, ODR is run at the level of ⟨ZZ⟩ measurements, from which ⟨ϕ2j ⟩ is computed using Eq. (16). In the
specific case of nq = 2 and ϕmax = 1.5, ⟨ϕ2j ⟩ is given by

⟨ϕ2j ⟩ = 1.25 + ⟨Z2j Z2j+1 ⟩. (C1)

ODR estimates expectation values in the physics circuit ⟨ZZ⟩phys twirl by twirl, using knowledge of the noiseless
expectation values of the mitigation circuit ⟨ZZ⟩mit,true and its measurements from the device ⟨ZZ⟩mit,meas using
Eq. (26):
1 ⟨Z2j Z2j+1 ⟩mit,true
⟨Z2j Z2j+1 ⟩phys = ⟨Z2j Z2j+1 ⟩phys,meas = ⟨Z2j Z2j+1 ⟩phys,meas . (C2)
pj ⟨Z2j Z2j+1 ⟩mit,meas
ODR hinges on selecting mitigation circuits whose noise is as similar to the noise in the physics circuits as possible.
The brickwall circuits used in this work are much more dense than those used in previous works where ODR was
successfuly applied. This results in errors spreading much faster throughout the qubits. As a result, it is found in this
work that previous choices of mitigation circuits, such as applying the time evolution operator U (−t/2)U (t/2) = 1
(self-mitigation [157]) or U (0) = 1 were not able to reflect the noise present in the physics circuit well. Similarly, using
various “Cliffordized” versions of the physics circuits as mitigation circuits did not capture the noise accurately for use
in ODR.20 This is particularly evident at late times, where the circuits are deeper and the mitigation circuits used
previously deviate more from the physics circuit being implemented.
Instead, the circuits simulating the vacuum time evolution are used as mitigation circuits in this work. If the
vacuum were prepared exactly, and if the time evolution operator were implemented exactly, time evolution would
have no effect as the vacuum is an eigenstate of the Hamiltonian, e−itH |ψvac ⟩ = |ψvac ⟩. Because both the vacuum state
preparation and the time evolution are implemented approximately using variational circuits, the prepared |ψvac ⟩ has
some time evolution. However, this vacuum evolution can be computed classically, owing to the translation-invariant
nature of |ψvac ⟩. Up to finite-size effects, the time evolution of translationally-invariant states does not change with
system size. The vacuum time evolution for a given system size can be estimated following the same procedure as
for the convergence of the variational parameters described in the SVC algorithm. Due to the fact that correlations
in the vacuum decay exponentially in this theory, finite-size effects have an exponentially-small contribution. As a
result, the infinite-volume vacuum expectation values ⟨ψvac |ZZ|ψvac ⟩∞ (as well as those for any L) can be estimated
by extrapolating from an exponential fit to the vacuum expectation values at a series of L. This provides a way to
compute the ⟨ZZ⟩mit,true . The values of these extrapolations for all simulation times in both the free and interacting
theories, as well as the values at L = 60 determined by MPS, are provided in Table IV. For the parameters and

20 Proposals to combine Zero Noise Extrapolation [3] with ODR have seen fruitful results [17].
 27

λ=0 λ=2
Even sites Odd sites Even sites Odd sites
t
MPS Extrapolation MPS Extrapolation MPS Extrapolation MPS Extrapolation
1 0.3949 0.3944 0.3882 0.3877 0.3580 0.3576 0.3469 0.3465
2 0.4677 0.4670 0.4678 0.4671 0.3692 0.3688 0.3621 0.3617
3 0.4970 0.4955 0.5073 0.5058 0.3477 0.3471 0.3551 0.3546
4 0.4678 0.4627 0.4768 0.4717 0.3959 0.3941 0.3767 0.3748
5 0.5339 0.5251 0.5621 0.5528 0.3620 0.3617 0.3672 0.3671
6 0.5275 0.5331 0.5404 0.5201 0.4027 0.4017 0.3995 0.3985
7 0.4349 0.4292 0.4227 0.4170 0.3693 0.3686 0.4010 0.4004
8 0.4620 0.4596 0.4617 0.4551 0.3910 0.3902 0.3768 0.3757
9 0.4780 0.4606 0.4987 0.4844 0.4143 0.4136 0.3946 0.3939

TABLE IV: ⟨ψvac |eitH ϕ2j e−itH |ψvac ⟩ that are used as “true” values in the error mitigation scheme. Expectation values
determined from classical simulations by using an exponential extrapolation are compared to those determined by
MPS. A maximum system size of L = 16 (32 qubits) was used to compute the extrapolation. The even and odd sites
have slightly different values because of the Trotterized state preparation, see Sec. IV A for details.

0.8
hφ2j i2wp − hφ2j ivac

0
Raw

hZZi2wp

0.4 MPS simulator

Pure noise
−1 ibm fez
0 40 80
Pauli twirl/layout
0.0
0.8
hφ2j i2wp − hφ2j ivac
ODR + Filter

0
hZZi2wp

0.4
−1
0 40 80
Pauli twirl/layout
0.0
0 10 20 30 40 50 60
Spatial site j

FIG. 13: The effect of ODR and filtering on the data from ibm fez for λ = 0, t = 8. The points show data collected
using 80 PTs, two TREX twirls, and 8000 shots for each circuit. The expectation value ⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac is shown
before (top panel) and after (bottom panel) ODR and filtering. The grey dashed line represents classical noiseless
MPS simulations, and the grey dotted line shows the results if the state had fully decohered to the completely mixed
state. The error bars represent one standard deviation of the bootstrap-resampled data. ODR and filtering act on the
⟨ZZ⟩2wp expectation value twirl by twirl (insets).

simulation times considered, this method is seen to give much better estimates of the noise, as a result of the same
time evolution being applied in both the physics and mitigation circuits. The effect of ODR can be seen in Fig. 13.
The unmitigated data shown in the upper panel is much closer to the completely decohered result (grey dotted line),
than the true result from MPS (grey dashed line). After the error mitigation scheme is applied (bottom panel) to
measure the noise and adjust the results, measurements of ⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac are much closer to the expected results
from MPS simulations. The action of ODR on ⟨ZZ⟩2wp measurements, from which ⟨ϕ2j ⟩2wp is calculated, is shown in
the insets of Fig. 13 for lattice site j = 4.
This method, similar to self-mitigation and other methods used earlier, is limited to early simulation times. In the
case of the previous methods, the limiting factor was the growth of the difference in the noise between the physics and
 28

mitigation circuits with increasing circuit depth. Mitigation using the extrapolated vacuum evolution is limited by
the quality of extrapolation. Finite-size effects grow as O(t/L), so larger system sizes must be used to get reliable
extrapolated values at later times. In previous choices of mitigation circuit, the physics circuit was modified to
implement the identity and applied to the same initial state. In this work it is found that applying the same exact
unitary mitigates errors in dense circuits (like the brickwall circuits of Fig. 8) much better, even though the starting
state is different. This method is expected to work especially well in settings where perturbations on top of the vacuum
are small, so that the state-dependent noise is similar in the physics and mitigation circuits. This is the case for
wavepackets in scalar field theory (as seen in Fig. 10 for example).
Note that it is possible to directly calculate an approximation to the vacuum evolution using MPS. In this work,
results from classical MPS simulations are used as ground truth to compare to results from ibm fez. For the system
parameters chosen and the times considered, it can be seen that the extrapolated vacuum evolution is quite close to the
MPS predictions (see Table IV), with the differences only noticeable at late times. It is expected that highly-inelastic
collisions will generate large quantities of entanglement and magic, thus being genuinely out of reach for simulations
using classical methods such as MPS or Clifford+T. In this setting, the MPS values for the vacuum evolution may
be used directly for mitigation. This is a reasonable expectation, as the ground states of one-dimensional gapped
systems, such as the scalar field theory considered in this work, are known to obey area law entanglement [187], so
their short-time evolution may be simulated with an MPS with a relatively small bond dimension.21
The ratio pj calculated from the results of the mitigation circuit is used to remove observations that have been
particularly affected by noise. This procedure, referred to as filtering [15], discards measurements of a particular local
observable if the ratio of the measured expectation value to the true expectation value is below a certain threshold; in
this work the threshold is set to pj ≥ 0.01 for all runs. Once the ratio is applied to adjust the measurements of the
physics circuit, a second round of filtering is done. This filter removes values of ⟨ZZ⟩phys ≥ 1. Since the maximum
possible value of ⟨ZZ⟩ = 1, any values above 1 are the result of the more decoherence being present for that observable
in the mitigation circuits; these observations are discarded as well. These two filters work by removing outliers and
extremely noisy data points, causing the size of the bootstrap samples for each local observable to be different.
A loop of qubits connected by two-qubit gates is necessary because of the PBCs used in this work. As a result of
the scalability of the simulation methods outlined in previous sections, it is possible to run circuits on the largest
possible loop on ibm fez (120 qubits). A specific loop is chosen based on a combination of the smallest errors and
longest coherence time throughout the device. It is found in this work that the randomization of mapping from circuit
qubits to device qubits (layout randomization) helps to tailor the noise closer to depolarizing noise, in addition to
averaging out effects from noise due to localized imperfections on the device, such as specific noisy qubits, gates, and
readout channels. In this work, a loop is fixed for all runs and the starting point of the loop is chosen at random for
each circuit. For each run, the circuits are executed in an order that alternates between the mitigation (vacuum) and
physics (wavepacket) circuits, for the same PT and layout. Since the noise on the device is expected to fluctuate with
time, this is intended to more accurately measure the noise by running the mitigation circuits directly before the
corresponding physics circuits. The same set of twirls and layouts is used for the physics and mitigation circuits.
To mitigate measurement errors, TREX is used. This method involves applying Pauli gates on all qubits at random
prior to measurement, flipping the value of the classical result if the Pauli applied is X or Y , and doing nothing in the
case of Z or I. Two TREX twirls are used for each Pauli-twirled circuit in this work.
The IBM Heron processors have seen a 3-5× increase in performance over their Eagle predecessors, and have
significantly decreased qubit crosstalk [188]. Despite this, it is still seen in this work that crosstalk remains a non-
negligible source of error. DD is used to minimize this effect and remove erroneous evolution during qubit idle times.
To mitigate these effects, the crosstalk-robust CRXY4 DD sequence [189] is used. This sequence consists of staggered
XY4 sequences applied to neighboring qubits to cancel out collective evolution caused by crosstalk.22
In particular, the effects of crosstalk are noticeable in regions where qubits are idle for numerous circuit layers
directly adjacent to ones that have gates acting on them. These types of errors are most prevalent in the wavepacket
creation step of the simulation algorithm, where the wavepacket creation circuits are acting in limited regions of the
quantum device (green blocks of Fig. 8). To mitigate these effects, barriers are inserted to isolate the times where both
qubits are idle or single-qubit gates are being run on the wavepacket qubits, from the times where two-qubit gates are
acting in the wavepacket region. The barriers instruct the circuit transpilation to add DD separately for these two
cases because the noise is different during each process.

21 The approximate variational methods may yield states that are more entangled than the true, non-variational states. This may be
averted by using techniques similar to Refs. [104, 144, 147] to train the variational circuits with MPS in the first place. Furthermore,
MPS approximations to the |ψ2wp ⟩ evolution circuit may be used as mitigation circuits in the future as well, provided they can be
constructed in a way that preserves the structure of the noise.
22 Recent work found that asymmetric implementations of Y gates can have an adverse impact on the effectiveness of DD sequences
[190]. As a result, all Y gates used in this work, including the DD stage, are converted to the native gate set using the symmetric
implementation Y = RZ (π/2)XRZ (−π/2).
 29

Appendix D: Variational parameters

In this section, the variational parameters determined using the SVC framework presented in Sec. III are given for
vacuum preparation, wavepacket excitation, and time evolution. Exponential extrapolation is run for the vacuum
preparation parameters. The L-extrapolated vacuum parameters are used at L = 12 to optimize the wavepacket
preparation and time evolution circuits.

λ=0 λ=2
L θ0 θ1 θ2 θ0 θ1 θ2
2 2.5517 1.3354 0.5659 2.6685 1.0626 0.4370
4 2.6346 1.0111 0.9131 2.7164 0.7923 0.7438
6 2.6462 0.9469 0.8703 2.7206 0.7547 0.7141
8 2.6490 0.9312 0.8580 2.7213 0.7491 0.7088
10 2.6497 0.9278 0.8536 2.7215 0.7478 0.7087
12 2.6499 0.9248 0.8536 2.7214 0.7466 0.7088
∞ 2.6495 0.9273 0.8534 2.7214 0.7476 0.7085

TABLE V: The variational parameters to prepare the vacuum as a function of system size, for both the free and
interacting theory. They correspond to the angles θ0 , θ1 , θ2 defined in the circuit elements of Fig. 3a. These parameters
are determined using classical simulations and optimization. The infinite-volume values shown in the last row are
calculated by an exponential extrapolation.
 30

λ=0 λ=2
j i=0 i=1 i=2 i=3 i=0 i=1 i=2 i=3
0 -0.1531 0.8023 0.5352 -0.6361 3.0113 -0.2089 1.8433 1.2508
1 -2.6753 1.7051 -0.0746 1.5072 -0.1898 1.9292 0.0173 -1.3358
2 0.9219 0.4939 0.0057 0.3665 2.0861 1.3400 0.1340 1.8122
3 0.2726 -0.3087 0.5203 -0.2217 2.9159 1.3803 -0.9289 0.8222
4 1.6515 1.5371 1.3394 1.4023 1.4625 -1.0540 1.5245 1.5592
5 -0.3017 -0.2190 -0.7053 -0.0079 1.0102 1.5805 1.4121 -0.0310
6 0.4914 0.0186 -0.0630 1.5343 -0.1876 1.6883 0.8825 1.9420
7 -1.7130 -0.2557 -1.1006 -0.7642 1.1893 -0.0476 -0.8028 1.2294
8 -1.5369 1.5928 1.5733 -1.2954 1.3134 1.2069 1.5725 2.1823
9 1.2792 -0.0988 -1.7023 -0.0211 -0.2964 -1.7139 -0.3014 1.7205
10 1.5154 0.0565 -1.0869 3.1620 -0.6159 1.6691 1.4801 -1.4323
11 -1.5202 0.8090 -0.5202 -1.5110 -1.3265 1.2461 1.5431 2.1283
12 0.5463 -0.7434 -0.8032 0.9535 0.6183 0.6455 -1.4778 -1.3675
13 -1.0278 -0.5927 1.1085 1.7822 1.7852 1.7364 1.3890 0.0582
14 1.7048 1.5187 1.3373 1.5012 -0.9282 -1.1880 2.7412 1.6371
15 0.7122 -0.5207 -0.0962 0.0393 1.0079 -0.7267 1.3845 0.2163
16 1.7216 -0.4310 0.7563 1.0700 1.2270 1.4292 0.9514 1.7011
17 -1.5542 -0.0394 -0.0974 1.5916 -0.9657 -0.2391 0.0401 1.7943
18 1.5395 1.5849 1.1416 1.6014 2.9777 -1.2030 1.0188 1.5363
19 -0.1785 1.1545 -0.0179 -1.6719 1.8511 0.5998 0.0393 -0.5372

TABLE VI: The parameters θij used to initialize Gaussian wavepackets of momenta k = ±π/3 and spread in k-space
σk = π/3 in the free and interacting theories. The θij correspond to the angles defined in the brickwall layer shown in
Fig. 3c: i labels the layer, and j labels the parameter within layer i. These parameters are determined using classical
simulations and optimization.

λ=0 λ=2
j i=0 i=1 i=2 i=3 i=4 i=5 i=0 i=1 i=2 i=3 i=4 i=5
0 0.0843 -0.0609 0.6955 0.2961 -0.048 -0.0191 0.1727 0.0329 -1.1632 0.0605 0.6813 0.1975
1 -0.2860 -3.4066 -2.9043 0.0167 2.2841 -0.0292 0.8868 -2.6254 1.6950 -0.1890 -3.8622 0.1428
2 1.5618 0.1327 1.4740 -0.1091 -1.4579 -1.3838 -0.8353 0.1261 -1.1718 -0.1111 1.0946 -0.7012
3 0.2173 1.3044 -1.1338 -1.4324 -1.3960 -1.6492 -0.2904 -2.0721 1.6594 0.5606 1.0016 -0.4645
4 0.0121 -1.1901 1.9423 -1.4593 0.1350 -1.1644 -0.6027 3.2926 0.2786 -0.8616 -1.5139 0.5588
5 -1.5731 2.2241 1.3656 0.1695 -0.6027 -1.1536 1.4649 -1.5636 0.7272 1.4701 2.0278 1.0915
6 -0.1173 0.7041 -1.0110 1.5841 -0.6495 -0.0152 0.0666 -1.5045 0.8148 0.6901 -0.1011 0.0370
7 1.8137 1.2280 1.5580 -0.1221 0.3547 -0.0793 1.7586 -1.2382 0.9962 -0.5063 0.1100 0.0887
8 0.0636 -1.0966 1.9644 -1.4114 3.3265 1.0599 -0.6402 2.2965 -0.0883 -1.1814 -2.0575 0.8626
9 -1.6840 2.0991 1.5450 0.3095 0.5833 -1.1253 1.3265 -2.0403 0.9642 1.9077 1.3393 1.1265
10 -0.0975 0.7022 -1.0872 1.6051 -0.6503 -0.0324 -0.1070 -1.1265 1.2493 -0.1006 -0.5110 -0.0454
11 1.8584 1.1814 1.594 -0.0322 0.3009 -0.0683 1.6193 -1.1086 0.9842 -0.5982 -0.0874 0.0748

TABLE VII: The parameters θij used to implement time evolution for t = 1 in the free and interacting theories.
The θij correspond to the angles defined in the brickwall layer shown in Fig. 3d: i labels the layer, and j labels the
parameter within layer i. These parameters are determined using classical simulations and optimization.
 31

λ=0 λ=2
j i=0 i=1 i=2 i=3 i=4 i=5 i=0 i=1 i=2 i=3 i=4 i=5
0 -0.0127 -0.0303 0.8574 0.2545 -0.1172 0.0181 0.0398 0.0648 -0.8169 0.3516 0.7947 0.1075
1 -0.7944 -4.0568 -3.6008 0.0228 3.4077 0.7981 1.1739 -1.9846 1.7857 0.4659 -4.8039 -0.2786
2 1.5237 0.3437 1.6710 -0.1623 -1.6623 -1.6914 -1.5506 0.1435 -0.4705 -0.0216 1.8417 -1.0534
3 0.0042 1.5494 -1.0016 -1.4038 -1.9429 -1.6069 -0.0645 -0.9712 -1.0871 1.1405 1.3022 -0.4437
4 -0.2237 -0.8308 1.5671 -1.4545 -0.1409 -0.5384 -0.4929 2.1439 -0.2443 -1.7539 -1.5463 -0.2068
5 -1.5910 2.5192 1.4323 0.2501 -0.8664 -0.5929 1.2547 -1.7316 -0.1444 2.1785 1.9204 1.2651
6 -0.0115 0.6926 -0.7056 2.0844 -0.5686 0.0572 0.2956 -1.9191 0.9858 -0.5653 0.1543 0.0046
7 2.6181 1.3091 1.4943 0.0006 -0.3561 -0.0597 0.4242 1.1723 1.2054 -0.0798 0.1706 0.0134
8 -0.2804 -0.9134 1.7664 -1.2446 3.1131 0.5317 -0.6196 2.0314 -0.3388 -1.9148 -1.7189 -0.1011
9 -1.5676 2.3869 1.5213 0.3573 0.8461 -0.5576 1.1632 -1.6832 0.0979 1.9924 2.1347 1.2702
10 -0.0285 0.6670 -0.6825 2.0468 -0.5351 0.0372 0.1732 -2.0512 0.9512 -0.7355 0.0777 0.0141
11 2.5855 1.3686 1.4927 0.0083 -0.3019 -0.0333 0.3224 1.0980 1.3002 -0.0688 0.3608 0.0228

TABLE VIII: The parameters θij used to implement time evolution for t = 2 in the free and interacting theories.
The θij correspond to the angles defined in the brickwall layer shown in Fig. 3d: i labels the layer, and j labels the
parameter within layer i. These parameters are determined using classical simulations and optimization.

λ=0 λ=2
j i=0 i=1 i=2 i=3 i=4 i=5 i=0 i=1 i=2 i=3 i=4 i=5
0 0.0021 -0.0586 1.5776 0.0003 0.529 0.4828 0.1081 0.0477 -1.1605 0.5728 1.1499 0.2074
1 -1.5784 -4.6994 -3.1345 0.0540 3.4056 0.0368 0.1861 -2.7631 2.0690 0.1181 -3.8915 2.7622
2 1.4004 0.6131 0.7863 -0.1804 -1.6037 -0.9163 -1.5665 0.1879 -0.5834 0.0316 1.8911 -1.1214
3 0.1212 1.6162 -1.5701 -1.7525 -1.5181 -0.8125 -0.0984 -1.5692 -0.2535 3.0750 1.6547 -0.5287
4 -0.3352 -1.5919 1.2529 -0.4186 1.6014 -1.5573 -0.1697 1.5929 -0.2217 -1.5676 -1.8083 -0.4096
5 -1.5505 3.1467 1.4259 0.3900 -1.4477 0.2438 1.6468 -1.5093 -0.3895 2.8569 1.8389 1.1812
6 -0.0018 1.1808 -0.0031 1.5923 -2.6329 -0.9154 -0.0441 -1.8221 0.9875 1.0029 -0.2364 0.0088
7 3.1182 1.5498 1.5514 0.3567 -0.0363 -0.0068 1.0255 0.1220 1.6884 -0.0175 -1.8634 -0.0213
8 -0.4566 -1.4424 1.4231 -0.5025 4.7225 1.5587 -0.1532 1.6630 -0.1710 -1.5943 -1.7456 -0.4261
9 -1.5682 3.1396 1.4702 0.0032 1.4704 -0.2056 1.6425 -1.5363 -0.3713 2.9048 1.9027 1.1827
10 0.0028 0.9779 0.0008 1.5751 -0.6327 0.9388 -0.0570 -1.7169 0.9733 0.9879 -0.1879 -0.0017
11 3.1055 1.5655 1.6030 0.2539 -0.0241 -0.0084 1.0308 0.1463 1.670 -0.0398 -1.8732 -0.0112

TABLE IX: The parameters θij used to implement time evolution for t = 3 in the free and interacting theories. The θij
correspond to the angles defined in the brickwall layer shown in Fig. 3d: i labels the layer, and j labels the parameter
within layer i. These parameters are determined using classical simulations and optimization.
 32

λ=0 λ=2
j i=0 i=1 i=2 i=3 i=4 i=5 i=0 i=1 i=2 i=3 i=4 i=5
0 -0.1073 0.0847 0.4879 0.3093 -0.2566 -0.0338 -0.0025 -0.0237 -0.7062 0.2692 0.5673 -0.2870
1 -0.5637 -5.1000 -2.6777 -0.0863 3.3271 1.2604 1.0773 0.7424 2.4129 0.3060 -4.2993 0.5121
2 1.5109 0.4379 1.0296 -0.1679 -1.6439 -0.8366 -1.6468 0.3683 -0.4062 0.0612 1.7473 -1.9365
3 0.0692 1.5616 -1.0566 -2.0557 -2.3746 -1.5466 -0.0042 -1.4611 0.0559 0.7209 0.9852 -0.7805
4 -0.4893 -1.0224 2.3117 -1.6204 0.2698 -0.5523 -0.4652 2.3685 -0.7806 -1.5357 -1.5061 0.0209
5 -1.6233 3.2011 1.8399 -0.7357 -0.8480 -0.2008 1.5544 -1.6221 0.9499 2.2305 2.2893 1.1904
6 -0.0555 1.5751 -1.0267 1.3182 -1.8387 0.1214 -0.0479 -1.2671 0.6778 -0.0874 -0.0155 -0.0129
7 3.6119 1.2176 1.5505 0.2732 -0.9677 -0.0461 -2.3391 0.0142 1.3246 -0.0462 -0.7364 0.0004
8 -0.5405 -0.9474 2.3937 -1.5040 3.4139 0.5829 -0.4848 2.2950 -0.7543 -1.5413 -1.4418 0.0648
9 -1.4556 3.2751 1.6913 -0.9322 0.7875 -0.3782 1.4431 -1.6935 0.9909 2.1455 2.0475 1.1984
10 -0.0087 1.5019 -1.0804 1.3444 -1.9556 0.0493 0.0243 -1.2489 0.7149 0.0038 -0.0335 -0.0282
11 3.4239 1.1735 1.5976 0.2317 -1.0280 -0.0450 -2.4800 -0.0288 1.4062 -0.0953 -0.7767 0.0393

TABLE X: The parameters θij used to implement time evolution for t = 4 in the free and interacting theories. The θij
correspond to the angles defined in the brickwall layer shown in Fig. 3d: i labels the layer, and j labels the parameter
within layer i. These parameters are determined using classical simulations and optimization.

λ=0 λ=2
j i=0 i=1 i=2 i=3 i=4 i=5 i=0 i=1 i=2 i=3 i=4 i=5
0 -0.1326 0.0372 0.5814 0.3058 -0.2071 -0.1162 0.0000 -0.0105 -1.3237 0.6746 0.6885 1.3835
1 -0.7267 -5.7599 -3.2273 -0.0727 3.6763 1.1732 -0.0002 1.0968 1.5905 1.5804 -5.7955 1.5715
2 1.5629 0.4892 1.0416 -0.1845 -1.7186 -1.1372 -1.6402 0.3005 -0.4687 -0.0184 2.7857 -1.6367
3 -0.0209 1.5764 -1.1759 -2.1123 -2.2658 -1.4966 0.0821 -1.513 0.0071 0.2494 1.5453 -1.5708
4 -0.5286 -1.0685 2.2987 -1.7233 0.1708 -0.5050 -0.0255 1.5761 0.0117 -3.1069 -1.1159 0.2522
5 -1.6923 3.3742 2.0409 -0.7240 -0.6956 -0.1600 1.4987 -1.0560 -0.2535 3.5926 1.5549 0.2101
6 -0.1063 1.4380 -0.8177 1.2692 -1.6112 0.0894 0.0001 -0.9552 0.4211 -1.2032 0.2553 0.4862
7 4.3371 1.3833 1.5239 0.3667 -0.7124 0.0020 -2.6630 -0.0029 1.5707 -0.0319 -1.5476 0.1618
8 -0.6487 -0.9463 2.4319 -1.6415 3.3190 0.5682 0.0047 1.6030 0.0016 -3.1070 -0.9698 0.2296
9 -1.3430 3.6764 1.7624 -1.0195 0.8142 -0.4189 1.3910 -1.0438 -0.2541 3.6141 1.5336 0.2097
10 -0.0081 1.1308 -1.0171 1.4978 -1.9940 0.0378 -0.0011 -0.9338 0.4016 -1.1540 0.2346 0.6202
11 3.9799 1.1996 1.6125 0.2875 -1.0267 -0.0509 -2.6706 -0.0246 1.5695 -0.0200 -1.5884 0.1631

TABLE XI: The parameters θij used to implement time evolution for t = 5 in the free and interacting theories. The θij
correspond to the angles defined in the brickwall layer shown in Fig. 3d: i labels the layer, and j labels the parameter
within layer i. These parameters are determined using classical simulations and optimization.
 33

λ=0 λ=2
j i=0 i=1 i=2 i=3 i=4 i=5 i=0 i=1 i=2 i=3 i=4 i=5
0 0.0411 -0.1769 0.5015 0.8762 0.3464 -0.0139 0.0030 -0.0092 -1.4508 0.6506 0.8001 1.1527
1 -0.0942 -4.5314 -4.1438 -0.0091 2.9710 2.2259 -0.0019 1.1037 1.8273 1.5725 -4.2827 1.5692
2 1.5761 0.5747 0.6118 -0.5092 -1.5826 -1.5449 -1.5997 0.2543 -0.3165 -0.0120 2.8019 -1.7114
3 -0.0601 1.5681 -1.4277 -1.6061 -1.5133 -1.8533 0.0943 -1.5557 0.0137 0.2642 1.6118 -1.5714
4 -0.3745 -0.7460 3.0021 -2.1116 -0.0530 -1.3761 -0.0057 1.5674 -0.0024 -3.1086 -0.9774 0.1754
5 -1.5873 4.5537 1.6295 -1.0644 -0.4700 -0.8731 1.5394 -0.9324 -0.2696 3.3930 1.5974 0.0565
6 0.0100 1.0453 0.2529 1.4969 -0.9874 0.0076 -0.0109 -0.6916 0.6663 -1.5556 0.1951 0.6161
7 3.0271 0.2209 1.6037 0.1586 -0.9758 0.1598 -2.6591 -0.0200 1.5562 0.0239 -1.5693 0.1799
8 -0.3817 -0.7077 3.0607 -2.039 3.1999 1.3527 0.0047 1.5856 0.0037 -3.1055 -1.0376 0.2258
9 -1.5157 4.5203 1.6534 -1.0739 0.4792 -0.8984 1.4434 -0.9179 -0.2720 3.3773 1.5878 0.0779
10 0.0023 1.0685 0.2252 1.4997 -0.9731 0.0103 -0.0042 -0.7239 0.5917 -1.5104 0.2549 0.5399
11 3.0216 0.1465 1.5569 0.2173 -0.9865 0.1577 -2.6715 -0.0153 1.5556 0.0135 -1.6080 0.1801

TABLE XII: The parameters θij used to implement time evolution for t = 6 in the free and interacting theories.
The θij correspond to the angles defined in the brickwall layer shown in Fig. 3d: i labels the layer, and j labels the
parameter within layer i. These parameters are determined using classical simulations and optimization.

λ=0 λ=2
j i=0 i=1 i=2 i=3 i=4 i=5 i=0 i=1 i=2 i=3 i=4 i=5
0 -0.0038 0.0528 -0.0417 0.2083 0.0104 -0.0076 0.3432 0.0404 -1.0961 0.4267 1.1327 -0.0044
1 -1.5722 -4.7573 -4.6618 -0.0137 -0.1573 -0.2374 1.5457 -0.2010 1.5569 -0.0334 -4.7501 -0.5418
2 1.5930 1.9531 0.8304 -1.6783 -1.5760 -1.5841 -1.5684 0.0604 -0.4015 -0.0202 1.5155 -1.8278
3 0.0228 1.5732 -1.5732 -1.5716 -2.7976 -1.5621 -0.0520 0.0594 -1.5664 1.7422 1.5781 -0.0860
4 -1.9226 -0.2589 2.7112 -2.0992 0.0877 -0.4603 -1.2895 2.7726 -0.1389 -2.6832 -2.0356 -0.0042
5 -1.5758 4.6627 1.5745 3.3032 -0.0037 -0.7178 0.4848 -1.4714 -0.1673 1.6008 3.0597 1.2465
6 0.0072 0.3859 -0.2240 3.0196 -0.4894 -0.0041 0.0162 -2.2549 0.8874 -0.7982 0.3035 -0.0108
7 3.1959 1.5730 1.5746 -0.0129 0.2335 0.0013 -1.4587 1.5969 1.6607 -0.0131 0.5083 0.0036
8 -1.8641 -0.2125 2.7557 -2.0107 3.1658 0.5005 -1.4565 2.5604 0.0662 -2.0192 -1.5055 -0.4916
9 -1.5729 4.6746 1.5732 3.2918 0.0080 -0.7120 1.3067 -1.5503 -0.0798 1.5939 1.2361 1.2577
10 0.0017 0.3691 -0.2857 2.6193 -0.1122 -0.0086 0.4963 -1.9117 0.8412 -0.6049 0.8032 0.0027
11 3.1826 1.5762 1.5861 -0.0231 0.2348 0.0014 -0.7771 1.5899 1.0078 0.0033 0.5246 -0.0099

TABLE XIII: The parameters θij used to implement time evolution for t = 7 in the free and interacting theories.
The θij correspond to the angles defined in the brickwall layer shown in Fig. 3d: i labels the layer, and j labels the
parameter within layer i. These parameters are determined using classical simulations and optimization.
 34

λ=0 λ=2
j i=0 i=1 i=2 i=3 i=4 i=5 i=0 i=1 i=2 i=3 i=4 i=5
0 0.0017 0.1202 -0.2994 0.2670 0.008 -0.0089 2.0762 0.0757 -1.9467 0.5305 1.2240 -0.0050
1 -1.5720 -4.7144 -4.7091 0.0071 -0.1362 -0.1595 1.5288 -0.7801 1.5855 -0.1109 -4.7125 -0.4436
2 1.5624 2.4975 0.7148 -2.4083 -1.5697 -1.5723 -1.5883 -0.0486 -0.6387 0.0399 1.5001 -1.5977
3 0.0355 1.5617 -1.5725 -1.5728 -3.3672 -1.5360 -0.0729 0.8390 -1.5779 -0.1046 1.6059 -0.0781
4 -2.9469 -0.5245 2.2046 -2.2773 0.0940 -0.5461 -1.6336 1.2492 0.6364 -1.0464 -1.9926 -0.3007
5 -1.5805 4.7094 1.5692 3.2779 -0.0022 -1.0045 -0.2026 -1.4905 1.9586 1.5602 3.1292 1.2552
6 0.0011 0.7103 -0.2456 2.9032 0.0792 -0.0064 -0.1564 -2.5584 1.6904 -1.5897 -0.2588 0.0067
7 3.1444 1.5738 1.5574 -0.0258 0.1597 0.0009 -1.3099 1.5913 2.0877 -0.0052 0.4419 -0.0027
8 -2.8916 -0.5562 2.2471 -2.2285 3.0843 0.5374 -1.6450 1.2560 0.5729 -1.2107 -1.8533 -0.3861
9 -1.5770 4.7101 1.5731 3.2801 0.0088 -0.9555 -0.1530 -1.5376 1.9699 1.5432 2.9588 1.2599
10 -0.0034 0.5683 -0.3675 3.2224 -0.2362 -0.0142 -0.0932 -2.7019 1.7268 -1.6101 -0.4026 0.0064
11 3.1386 1.5736 1.5699 -0.0124 0.1559 -0.0004 -1.2813 1.5986 2.0409 -0.0089 0.4543 -0.0007

TABLE XIV: The parameters θij used to implement time evolution for t = 8 in the free and interacting theories.
The θij correspond to the angles defined in the brickwall layer shown in Fig. 3d: i labels the layer, and j labels the
parameter within layer i. These parameters are determined using classical simulations and optimization.

λ=0 λ=2
j i=0 i=1 i=2 i=3 i=4 i=5 i=0 i=1 i=2 i=3 i=4 i=5
0 -0.0008 0.2800 -0.2194 0.3617 -0.0039 0.1371 1.6507 0.0406 -1.6259 0.6184 1.3403 -0.0024
1 -1.5719 -4.7144 -4.7125 0.0032 0.1182 0.0038 1.2220 -1.5253 1.5760 -0.0568 -4.719 -0.4047
2 1.4799 2.7295 0.8456 -2.5016 -1.5347 -1.5740 -1.5872 -0.0099 -0.6879 0.0060 0.5950 -1.6816
3 0.0767 1.5663 -1.5709 -1.5705 -3.1967 -1.3495 -0.0371 0.7480 -1.5209 -0.5640 1.5786 -0.0048
4 -3.3481 -1.0828 1.7347 -2.9603 0.2967 0.5598 -1.5525 1.5285 0.6970 -1.5498 -1.9947 -0.3791
5 -1.5795 4.7124 1.5657 3.0253 -0.0025 -0.8377 -0.2366 -1.1993 2.1882 1.5553 3.1469 1.2310
6 -0.0013 0.7131 -0.3294 3.1614 -0.0275 0.0018 -0.0657 -2.5103 1.6019 -1.6033 -0.0107 0.0000
7 3.1395 1.5733 1.5699 0.0879 -0.0030 -0.0001 -0.623 1.5356 2.0720 0.0043 0.4026 -0.0003
8 -3.4245 -0.9444 1.8114 -2.9586 0.2792 0.5453 -1.5796 1.4988 0.6103 -1.5431 -1.9636 -0.3887
9 -1.5741 4.7115 1.5722 3.0286 -0.0033 -0.7946 -0.1634 -1.2214 2.1575 1.5569 3.1081 1.2298
10 -0.0042 0.6035 -0.423 3.1141 0.0180 -0.0166 0.0585 -2.6936 1.5923 -1.5523 -0.0084 -0.0008
11 3.1352 1.5725 1.5752 0.1049 -0.0049 0.0013 -0.5291 1.5413 1.9455 0.0013 0.4044 0.0008

TABLE XV: The parameters θij used to implement time evolution for t = 9 in the free and interacting theories.
The θij correspond to the angles defined in the brickwall layer shown in Fig. 3d: i labels the layer, and j labels the
parameter within layer i. These parameters are determined using classical simulations and optimization.
 35

Appendix E: Tables

In this section, expectation values ⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac are tabulated for λ = 0 and λ = 2. Error mitigated results from
ibm fez are compared to classical MPS data. Larger uncertainties in the data from the quantum computer can be
seen as a result of increasing circuit depth: the circuits for t = 1, 2, 3 have a depth of 59 (Tables XVI, XVII, XVIII),
those for t = 4, 5, 6 have a depth of 81 (Tables XIX, XX, XXI), and those for t = 7, 8, 9 have a depth of 103 (Tables
XXII, XXIII, XXIV).
 36

⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac

Spatial site λ=0 λ=2
j MPS ibm fez MPS ibm fez
0 0.000 0.014(10) 0.000 0.018(23)
1 0.000 0.017(10) 0.000 -0.008(09)
2 0.000 -0.005(11) 0.000 0.009(10)
3 0.000 -0.011(10) 0.000 0.015(10)
4 0.000 -0.015(10) 0.000 -0.005(12)
5 0.000 -0.003(09) 0.000 0.010(09)
6 0.000 -0.029(14) 0.000 0.014(13)
7 0.000 0.002(10) 0.000 0.009(15)
8 0.000 0.035(11) 0.000 0.003(11)
9 0.000 0.016(11) 0.000 -0.021(13)
10 0.000 -0.001(09) 0.000 -0.008(12)
11 0.000 0.001(09) 0.000 0.016(09)
12 0.000 0.006(10) 0.000 0.006(14)
13 0.000 -0.012(08) 0.000 0.017(15)
14 0.000 0.011(11) 0.000 -0.018(16)
15 0.000 0.011(09) 0.000 0.025(13)
16 0.000 0.008(11) 0.000 0.010(17)
17 0.000 0.007(12) 0.000 -0.032(11)
18 0.000 -0.017(15) 0.000 -0.008(14)
19 0.000 -0.007(11) 0.000 0.007(11)
20 0.000 -0.001(13) 0.000 -0.013(09)
21 0.000 0.000(12) 0.000 -0.004(10)
22 0.000 0.003(12) 0.000 0.005(16)
23 0.000 -0.009(11) 0.000 0.009(11)
24 0.000 0.031(15) 0.000 -0.022(15)
25 0.000 -0.005(13) 0.000 -0.013(13)
26 0.018 0.023(10) 0.004 0.003(11)
27 0.149 0.186(08) 0.085 0.164(12)
28 0.497 0.494(09) 0.376 0.447(15)
29 0.331 0.317(07) 0.232 0.268(12)
30 0.118 0.093(09) 0.069 0.074(11)
31 0.407 0.377(11) 0.243 0.256(12)
32 0.507 0.498(11) 0.349 0.396(11)
33 0.127 0.158(13) 0.110 0.171(12)
34 0.007 0.003(11) 0.006 0.012(16)
35 0.001 -0.009(09) 0.000 -0.016(12)
36 0.000 0.003(09) 0.000 -0.004(08)
37 0.000 -0.000(11) 0.000 0.013(10)
38 0.000 -0.002(11) 0.000 -0.003(10)
39 0.000 0.008(09) 0.000 -0.023(11)
40 0.000 0.015(09) 0.000 0.001(09)
41 0.000 0.018(08) 0.000 -0.027(15)
42 0.000 0.002(16) 0.000 -0.012(10)
43 0.000 0.007(10) 0.000 0.009(16)
44 0.000 0.007(11) 0.000 -0.009(15)
45 0.000 0.004(12) 0.000 -0.003(12)
46 0.000 0.002(12) 0.000 0.004(09)
47 0.000 0.009(09) 0.000 0.013(11)
48 0.000 0.008(08) 0.000 -0.006(12)
49 0.000 0.025(10) 0.000 0.000(10)
50 0.000 0.004(07) 0.000 0.014(09)
51 0.000 0.026(11) 0.000 0.003(09)
52 0.000 0.011(11) 0.000 -0.006(12)
53 0.000 -0.021(11) 0.000 0.001(11)
54 0.000 -0.008(12) 0.000 -0.001(10)
55 0.000 0.012(13) 0.000 0.024(15)
56 0.000 -0.009(11) 0.000 0.011(10)
57 0.000 0.010(11) 0.000 0.004(11)
58 0.000 -0.018(12) 0.000 -0.003(10)
59 0.000 0.005(10) 0.000 0.002(08)

TABLE XVI: Numerical values for the expectation value ⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac at t = 1 from MPS and ibm fez for the
free and interacting theories, as shown in Figs. 9 and 10.
 37

⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac

Spatial site λ=0 λ=2
j MPS ibm fez MPS ibm fez
0 0.000 0.007(10) 0.000 0.003(09)
1 0.000 0.013(08) 0.000 -0.004(12)
2 0.000 0.021(11) 0.000 -0.003(11)
3 0.000 0.008(08) 0.000 0.004(09)
4 0.000 -0.013(11) 0.000 0.007(11)
5 0.000 -0.004(12) 0.000 0.005(08)
6 0.000 0.004(10) 0.000 -0.006(10)
7 0.000 0.005(09) 0.000 -0.018(09)
8 0.000 0.009(09) 0.000 -0.004(10)
9 0.000 0.015(09) 0.000 0.009(09)
10 0.000 -0.004(11) 0.000 0.016(10)
11 0.000 0.011(08) 0.000 0.008(09)
12 0.000 -0.008(09) 0.000 0.021(12)
13 0.000 -0.014(11) 0.000 -0.007(12)
14 0.000 -0.005(11) 0.000 0.006(09)
15 0.000 -0.003(10) 0.000 -0.009(09)
16 0.000 -0.001(12) 0.000 -0.003(10)
17 0.000 0.019(13) 0.000 -0.021(10)
18 0.000 0.010(14) 0.000 0.022(11)
19 0.000 0.004(10) 0.000 0.007(08)
20 0.000 0.010(12) 0.000 0.002(09)
21 0.000 0.008(13) 0.000 0.020(09)
22 0.000 0.010(11) 0.000 0.008(09)
23 0.000 0.026(11) 0.000 0.000(08)
24 -0.001 -0.014(11) 0.000 -0.011(09)
25 0.003 0.013(09) 0.001 0.003(10)
26 0.048 0.049(10) 0.027 0.039(11)
27 0.135 0.141(10) 0.060 0.140(11)
28 0.336 0.367(11) 0.278 0.359(13)
29 0.345 0.322(10) 0.311 0.362(12)
30 0.228 0.180(08) 0.179 0.129(11)
31 0.401 0.322(11) 0.328 0.360(10)
32 0.286 0.333(09) 0.248 0.299(10)
33 0.046 0.025(10) 0.055 0.099(11)
34 0.045 0.026(08) 0.023 0.016(09)
35 0.001 0.007(09) -0.001 -0.018(16)
36 0.000 0.003(09) 0.000 -0.003(09)
37 0.000 0.002(10) 0.000 0.010(11)
38 0.000 -0.020(13) 0.000 -0.007(09)
39 0.000 0.003(09) 0.000 -0.015(10)
40 0.000 -0.015(10) 0.000 0.002(11)
41 0.000 -0.002(10) 0.000 0.008(08)
42 0.000 0.015(09) 0.000 0.001(07)
43 0.000 -0.017(10) 0.000 0.011(08)
44 0.000 0.031(12) 0.000 -0.005(09)
45 0.000 0.011(08) 0.000 -0.004(10)
46 0.000 -0.003(10) 0.000 0.010(13)
47 0.000 0.011(11) 0.000 -0.002(11)
48 0.000 -0.002(09) 0.000 0.015(09)
49 0.000 -0.014(11) 0.000 -0.003(09)
50 0.000 0.003(09) 0.000 -0.029(16)
51 0.000 -0.022(09) 0.000 0.008(09)
52 0.000 0.001(12) 0.000 -0.011(12)
53 0.000 -0.016(12) 0.000 -0.001(09)
54 0.000 -0.004(12) 0.000 -0.023(08)
55 0.000 -0.009(09) 0.000 0.012(10)
56 0.000 0.006(10) 0.000 -0.005(08)
57 0.000 -0.007(10) 0.000 0.005(08)
58 0.000 0.009(10) 0.000 -0.001(11)
59 0.000 -0.010(09) 0.000 0.005(09)

TABLE XVII: Numerical values for the expectation value ⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac at t = 2 from MPS and ibm fez for the
free and interacting theories, as shown in Figs. 9 and 10.
 38

⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac

Spatial site λ=0 λ=2
j MPS ibm fez MPS ibm fez
0 0.000 -0.013(11) 0.000 0.007(09)
1 0.000 -0.007(09) 0.000 0.007(09)
2 0.000 0.002(11) 0.000 0.002(08)
3 0.000 0.010(08) 0.000 0.028(12)
4 0.000 0.004(08) 0.000 0.018(07)
5 0.000 -0.006(10) 0.000 0.002(08)
6 0.000 0.019(11) 0.000 0.014(15)
7 0.000 0.025(17) 0.000 0.007(12)
8 0.000 -0.002(10) 0.000 -0.026(14)
9 0.000 -0.012(15) 0.000 0.001(12)
10 0.000 -0.013(10) 0.000 -0.010(12)
11 0.000 -0.016(10) 0.000 0.004(07)
12 0.000 -0.006(10) 0.000 -0.007(10)
13 0.000 -0.001(11) 0.000 -0.002(08)
14 0.000 -0.005(09) 0.000 0.021(10)
15 0.000 -0.017(09) 0.000 0.006(13)
16 0.000 0.003(14) 0.000 0.013(10)
17 0.000 -0.000(09) 0.000 0.009(09)
18 0.000 0.005(10) 0.000 0.013(12)
19 0.000 0.010(09) 0.000 -0.013(11)
20 -0.001 0.001(10) 0.000 -0.005(09)
21 -0.001 0.006(11) 0.000 0.002(09)
22 -0.001 0.005(09) 0.000 -0.011(10)
23 -0.001 -0.001(09) 0.000 -0.001(12)
24 -0.004 0.013(09) 0.000 -0.004(10)
25 0.006 -0.003(10) 0.008 0.004(10)
26 0.052 0.079(12) 0.028 0.033(19)
27 0.034 0.079(11) 0.051 0.138(13)
28 0.140 0.182(10) 0.214 0.292(11)
29 0.360 0.322(11) 0.392 0.368(10)
30 0.437 0.336(12) 0.391 0.259(13)
31 0.309 0.253(12) 0.366 0.354(09)
32 0.128 0.123(10) 0.162 0.254(12)
33 0.069 0.052(08) 0.056 0.121(10)
34 0.042 0.053(11) 0.029 0.051(12)
35 -0.004 0.007(09) 0.007 0.015(14)
36 -0.001 -0.003(10) -0.001 0.020(12)
37 0.000 0.010(09) 0.000 -0.008(09)
38 0.000 0.011(10) 0.000 -0.000(09)
39 0.000 -0.002(08) 0.000 -0.002(10)
40 0.000 0.006(09) 0.000 -0.003(09)
41 0.000 -0.007(11) 0.000 -0.008(15)
42 0.000 0.011(10) 0.000 0.003(08)
43 0.000 -0.020(13) 0.000 -0.001(09)
44 0.000 0.017(10) 0.000 0.007(08)
45 0.000 -0.017(10) 0.000 0.018(09)
46 0.000 0.006(11) 0.000 0.006(12)
47 0.000 0.006(08) 0.000 0.011(09)
48 0.000 0.002(11) 0.000 0.020(10)
49 0.000 0.006(10) 0.000 -0.006(09)
50 0.000 0.011(08) 0.000 -0.023(11)
51 0.000 0.013(08) 0.000 -0.010(07)
52 0.000 0.004(10) 0.000 0.004(09)
53 0.000 0.004(13) 0.000 0.002(11)
54 0.000 -0.013(12) 0.000 -0.013(11)
55 0.000 0.006(10) 0.000 -0.010(10)
56 0.000 -0.003(09) 0.000 0.011(08)
57 0.000 0.011(11) 0.000 -0.003(09)
58 0.000 0.002(10) 0.000 -0.003(10)
59 0.000 -0.008(10) 0.000 0.009(07)

TABLE XVIII: Numerical values for the expectation value ⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac at t = 3 from MPS and ibm fez for the
free and interacting theories, as shown in Figs. 9 and 10.
 39

⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac

Spatial site λ=0 λ=2
j MPS ibm fez MPS ibm fez
0 0.000 0.023(12) 0.000 0.008(15)
1 0.000 0.012(11) 0.000 0.001(10)
2 0.000 0.004(11) 0.000 0.014(10)
3 0.000 0.005(11) 0.000 0.019(11)
4 0.000 -0.020(13) 0.000 0.003(12)
5 0.000 0.001(12) 0.000 0.015(12)
6 0.000 -0.027(12) 0.000 0.003(09)
7 0.000 0.011(12) 0.000 0.016(12)
8 0.000 0.019(15) 0.000 0.025(13)
9 0.000 0.019(12) 0.000 0.007(14)
10 0.000 -0.017(14) 0.000 0.016(10)
11 0.000 -0.006(11) 0.000 0.016(11)
12 0.000 -0.014(12) 0.000 0.009(13)
13 0.000 0.008(15) 0.000 0.018(12)
14 0.000 0.001(13) 0.000 0.014(14)
15 0.000 0.019(12) 0.000 0.020(09)
16 -0.001 0.007(12) 0.000 0.024(08)
17 -0.001 0.001(12) 0.000 -0.001(11)
18 -0.002 0.006(12) -0.001 0.003(13)
19 -0.002 0.024(13) -0.001 -0.010(14)
20 -0.003 0.030(12) -0.001 0.003(13)
21 -0.003 0.028(11) -0.001 0.011(10)
22 -0.004 0.026(13) -0.001 0.005(11)
23 -0.002 0.022(14) -0.001 0.011(10)
24 0.011 0.006(14) 0.008 0.015(10)
25 0.049 0.043(10) 0.035 0.047(11)
26 0.072 0.083(14) 0.053 0.084(11)
27 0.049 0.076(12) 0.059 0.141(10)
28 0.149 0.175(11) 0.133 0.215(11)
29 0.470 0.364(09) 0.419 0.367(09)
30 0.813 0.559(11) 0.682 0.429(12)
31 0.436 0.371(10) 0.380 0.364(08)
32 0.117 0.158(13) 0.102 0.203(11)
33 0.066 0.084(11) 0.047 0.088(14)
34 0.071 0.091(12) 0.034 0.073(10)
35 0.022 0.053(11) 0.018 0.054(14)
36 0.001 0.014(12) 0.002 0.014(10)
37 0.000 0.006(11) 0.000 0.006(12)
38 0.000 -0.002(16) 0.000 0.021(11)
39 0.000 0.018(11) 0.000 0.023(12)
40 0.000 0.016(13) 0.000 0.026(13)
41 0.000 0.017(12) 0.000 0.018(11)
42 0.000 -0.016(13) 0.000 0.015(11)
43 0.000 0.006(15) 0.000 -0.001(11)
44 0.000 0.021(12) 0.000 0.004(14)
45 0.000 0.029(13) 0.000 -0.006(11)
46 0.000 0.030(16) 0.000 0.009(10)
47 0.000 0.009(12) 0.000 0.003(09)
48 0.000 0.005(12) 0.000 0.011(13)
49 0.000 0.002(10) 0.000 0.022(11)
50 0.000 0.020(12) 0.000 0.018(12)
51 0.000 0.001(19) 0.000 -0.005(09)
52 0.000 -0.020(11) 0.000 -0.008(10)
53 0.000 0.016(12) 0.000 0.015(13)
54 0.000 -0.003(12) 0.000 0.008(10)
55 0.000 -0.007(11) 0.000 -0.001(10)
56 0.000 0.000(12) 0.000 -0.000(14)
57 0.000 0.005(12) 0.000 0.018(11)
58 0.000 0.017(13) 0.000 0.010(09)
59 0.000 0.007(13) 0.000 0.004(10)

TABLE XIX: Numerical values for the expectation value ⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac at t = 4 from MPS and ibm fez for the
free and interacting theories, as shown in Figs. 9 and 10.
 40

⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac

Spatial site λ=0 λ=2
j MPS ibm fez MPS ibm fez
0 0.001 0.005(10) 0.000 0.008(13)
1 0.000 -0.016(15) 0.000 -0.011(13)
2 0.001 0.019(12) 0.000 -0.021(14)
3 0.000 -0.030(14) 0.000 -0.010(14)
4 0.001 -0.038(15) 0.000 0.003(11)
5 0.000 -0.020(13) 0.000 -0.008(14)
6 0.001 -0.019(11) 0.000 0.003(14)
7 0.000 -0.016(12) 0.000 0.018(12)
8 0.001 -0.023(13) 0.000 0.023(10)
9 0.000 0.002(15) 0.000 0.008(11)
10 0.000 -0.001(10) 0.000 -0.023(12)
11 0.000 -0.011(13) 0.000 -0.009(12)
12 0.000 -0.015(12) 0.000 0.011(11)
13 0.000 -0.013(12) 0.000 0.029(13)
14 0.000 -0.006(13) 0.000 0.004(09)
15 0.000 -0.005(13) 0.000 -0.007(10)
16 -0.001 -0.006(12) 0.000 -0.014(14)
17 -0.001 -0.000(13) -0.001 -0.009(12)
18 -0.002 -0.001(11) -0.001 0.001(21)
19 -0.003 -0.004(13) -0.001 -0.002(14)
20 -0.004 -0.002(13) -0.001 0.011(09)
21 -0.004 -0.012(13) -0.002 -0.002(09)
22 -0.004 -0.002(11) -0.001 -0.004(12)
23 -0.001 0.011(12) 0.001 -0.001(13)
24 0.019 0.014(13) 0.010 0.012(14)
25 0.036 0.026(11) 0.033 0.037(13)
26 0.077 0.055(13) 0.043 0.042(13)
27 0.031 0.058(11) 0.055 0.114(12)
28 0.153 0.123(12) 0.145 0.212(11)
29 0.438 0.323(10) 0.493 0.410(09)
30 0.661 0.444(10) 0.690 0.475(12)
31 0.315 0.248(12) 0.408 0.358(10)
32 0.079 0.054(11) 0.119 0.162(10)
33 0.093 0.068(10) 0.092 0.125(11)
34 0.053 0.039(14) 0.036 0.055(11)
35 0.016 0.001(13) 0.015 0.025(11)
36 0.002 -0.015(13) 0.008 0.011(17)
37 0.000 -0.017(13) 0.001 -0.006(11)
38 0.000 -0.002(12) 0.000 0.006(13)
39 0.000 -0.021(14) 0.000 0.009(10)
40 0.001 0.016(14) 0.000 0.005(12)
41 0.001 -0.013(11) 0.000 -0.012(13)
42 0.001 -0.027(13) 0.000 -0.006(16)
43 0.001 -0.039(14) 0.000 0.008(10)
44 0.001 -0.021(12) 0.000 -0.011(12)
45 0.000 -0.019(11) 0.000 -0.015(13)
46 0.001 -0.010(14) 0.000 -0.017(13)
47 0.000 -0.008(13) 0.000 -0.035(14)
48 0.001 -0.008(12) 0.000 -0.030(14)
49 0.000 -0.012(14) 0.000 -0.018(12)
50 0.000 -0.014(10) 0.000 -0.003(10)
51 0.000 -0.020(11) 0.000 0.001(10)
52 0.000 -0.018(13) 0.000 0.007(12)
53 0.000 -0.024(11) 0.000 -0.012(16)
54 0.001 -0.001(11) 0.000 -0.012(12)
55 0.000 -0.032(15) 0.000 0.003(13)
56 0.001 -0.018(12) 0.000 -0.005(12)
57 0.000 0.006(12) 0.000 0.013(12)
58 0.001 0.008(12) 0.000 0.019(12)
59 0.000 -0.012(12) 0.000 0.027(11)

TABLE XX: Numerical values for the expectation value ⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac at t = 5 from MPS and ibm fez for the free
and interacting theories, as shown in Figs. 9 and 10.
 41

⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac

Spatial site λ=0 λ=2
j MPS ibm fez MPS ibm fez
0 0.001 0.022(10) 0.000 0.025(18)
1 0.001 -0.010(11) 0.000 -0.036(16)
2 0.001 0.033(10) 0.000 -0.020(17)
3 0.001 0.010(10) 0.000 -0.021(25)
4 0.001 0.036(11) 0.000 -0.003(17)
5 0.001 0.020(10) 0.000 -0.020(13)
6 0.001 0.027(11) 0.000 0.004(16)
7 0.001 -0.021(12) 0.000 -0.026(23)
8 0.001 0.039(13) 0.000 -0.001(14)
9 0.001 -0.000(11) 0.000 0.002(19)
10 0.001 0.056(12) 0.000 0.019(12)
11 0.001 0.012(11) 0.000 0.013(17)
12 0.000 0.038(10) 0.000 -0.006(22)
13 0.000 0.024(12) 0.000 -0.017(13)
14 0.000 0.033(12) 0.000 -0.016(18)
15 -0.001 0.040(13) 0.000 -0.052(16)
16 -0.001 0.050(12) 0.000 0.000(19)
17 -0.003 0.044(12) 0.000 -0.034(15)
18 -0.003 0.065(09) -0.001 -0.059(29)
19 -0.005 0.046(14) -0.001 -0.051(22)
20 -0.006 0.074(13) -0.001 -0.021(19)
21 -0.007 0.038(13) -0.001 -0.004(14)
22 -0.006 0.062(11) -0.001 -0.010(15)
23 0.005 0.040(12) 0.003 -0.034(22)
24 0.025 0.054(10) 0.012 -0.011(23)
25 0.028 0.051(13) 0.023 0.012(20)
26 0.041 0.067(13) 0.058 0.037(17)
27 0.121 0.094(12) 0.044 0.066(12)
28 0.198 0.177(09) 0.110 0.110(23)
29 0.298 0.204(11) 0.257 0.200(30)
30 0.338 0.254(12) 0.426 0.282(13)
31 0.168 0.140(12) 0.177 0.236(14)
32 0.123 0.122(10) 0.089 0.138(13)
33 0.049 0.052(11) 0.069 0.069(21)
34 0.024 0.053(11) 0.018 0.033(16)
35 0.020 -0.009(11) 0.009 0.012(16)
36 0.010 0.054(11) 0.007 0.008(11)
37 0.000 0.033(10) 0.001 0.004(12)
38 0.000 0.038(10) 0.000 -0.009(19)
39 0.001 0.024(12) 0.000 0.019(15)
40 0.001 0.008(10) 0.000 -0.017(20)
41 0.001 0.005(12) 0.000 0.010(19)
42 0.001 0.042(12) 0.000 0.025(15)
43 0.001 -0.025(10) 0.000 -0.010(16)
44 0.001 0.010(12) 0.000 -0.011(16)
45 0.001 -0.000(14) 0.000 -0.010(10)
46 0.001 0.004(12) 0.000 0.010(14)
47 0.001 -0.006(13) 0.000 0.010(12)
48 0.001 0.029(14) 0.000 -0.008(14)
49 0.001 -0.009(14) 0.000 0.000(22)
50 0.001 0.015(13) 0.000 0.007(15)
51 0.001 -0.011(14) 0.000 0.006(20)
52 0.001 0.045(13) 0.000 0.003(12)
53 0.001 0.023(12) 0.000 -0.021(18)
54 0.001 0.043(10) 0.000 -0.004(13)
55 0.001 -0.023(11) 0.000 0.007(19)
56 0.001 0.052(10) 0.000 -0.033(23)
57 0.001 -0.016(14) 0.000 -0.007(21)
58 0.001 0.032(10) 0.000 -0.023(17)
59 0.001 0.020(11) 0.000 -0.017(13)

TABLE XXI: Numerical values for the expectation value ⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac at t = 6 from MPS and ibm fez for the
free and interacting theories, as shown in Figs. 9 and 10.
 42

⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac

Spatial site λ=0 λ=2
j MPS ibm fez MPS ibm fez
0 0.000 0.023(14) 0.000 -0.003(10)
1 0.000 0.012(16) 0.000 0.008(15)
2 0.000 0.015(17) 0.000 0.014(12)
3 0.000 0.025(16) 0.000 -0.001(13)
4 0.000 0.002(16) 0.000 -0.003(11)
5 0.000 -0.002(14) 0.000 -0.001(14)
6 0.000 -0.003(13) 0.000 0.008(14)
7 0.000 0.003(16) 0.000 -0.003(17)
8 0.000 0.007(12) 0.000 0.001(12)
9 0.000 0.025(13) 0.000 -0.002(11)
10 0.000 0.007(13) 0.000 0.011(17)
11 0.000 0.032(15) 0.000 0.004(11)
12 0.000 0.010(17) 0.000 -0.006(15)
13 0.000 0.004(13) 0.000 0.001(12)
14 0.000 -0.005(17) 0.000 -0.006(14)
15 -0.001 0.013(15) 0.000 -0.017(16)
16 -0.001 0.023(15) 0.000 0.005(12)
17 -0.002 0.015(12) 0.000 -0.002(10)
18 -0.003 0.023(17) -0.001 -0.012(14)
19 -0.003 0.016(18) -0.001 -0.023(13)
20 -0.004 0.042(15) -0.001 0.001(11)
21 -0.004 -0.017(14) -0.001 0.015(13)
22 -0.003 0.012(14) -0.001 -0.004(11)
23 0.013 0.036(16) 0.004 0.011(13)
24 0.019 0.023(13) 0.007 0.004(13)
25 0.034 0.022(15) 0.016 0.033(11)
26 0.053 0.053(15) 0.015 0.028(11)
27 0.144 0.118(16) 0.055 0.098(13)
28 0.298 0.238(15) 0.127 0.176(13)
29 0.281 0.228(17) 0.210 0.222(12)
30 0.087 0.103(15) 0.113 0.159(11)
31 0.212 0.175(14) 0.220 0.211(13)
32 0.256 0.213(14) 0.107 0.177(13)
33 0.149 0.131(13) 0.052 0.097(15)
34 0.035 0.047(16) 0.018 0.019(12)
35 0.024 0.054(15) 0.016 0.031(13)
36 0.013 0.023(15) 0.010 0.042(12)
37 0.010 0.030(19) 0.004 -0.012(11)
38 0.000 0.015(14) 0.000 -0.017(13)
39 0.000 0.006(17) 0.000 -0.001(15)
40 0.000 0.013(14) 0.000 0.014(12)
41 0.000 -0.001(16) 0.000 -0.007(12)
42 0.000 -0.004(12) 0.000 0.001(11)
43 0.000 -0.004(15) 0.000 -0.007(11)
44 0.000 0.015(18) 0.000 0.012(13)
45 0.000 0.010(12) 0.000 0.004(12)
46 0.000 0.008(17) 0.000 0.012(13)
47 0.000 0.003(15) 0.000 -0.015(14)
48 0.000 -0.008(15) 0.000 -0.022(14)
49 0.000 0.008(12) 0.000 -0.004(15)
50 0.000 0.030(16) 0.000 0.012(11)
51 0.000 0.014(14) 0.000 -0.012(12)
52 0.000 0.041(18) 0.000 -0.009(12)
53 0.000 -0.003(15) 0.000 -0.001(11)
54 0.000 0.027(16) 0.000 0.013(11)
55 0.000 0.006(15) 0.000 -0.005(13)
56 0.000 0.013(14) 0.000 -0.006(11)
57 0.000 0.004(20) 0.000 0.011(14)
58 0.000 0.041(13) 0.000 -0.032(12)
59 0.000 0.004(16) 0.000 -0.010(11)

TABLE XXII: Numerical values for the expectation value ⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac at t = 7 from MPS and ibm fez for the
free and interacting theories, as shown in Figs. 9 and 10.
 43

⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac

Spatial site λ=0 λ=2
j MPS ibm fez MPS ibm fez
0 0.000 -0.004(10) 0.000 -0.024(16)
1 0.000 0.011(12) 0.000 -0.014(15)
2 0.000 -0.007(13) 0.000 -0.013(10)
3 0.000 0.012(10) 0.000 -0.028(16)
4 0.000 0.010(12) 0.000 -0.019(13)
5 0.000 -0.004(10) 0.000 0.013(19)
6 0.000 0.016(13) 0.000 -0.012(13)
7 0.000 -0.001(14) 0.000 0.009(15)
8 0.000 0.008(08) 0.000 0.045(14)
9 0.000 0.007(10) 0.000 -0.027(13)
10 0.000 -0.002(11) 0.000 0.013(14)
11 0.000 0.038(12) 0.000 0.022(15)
12 0.000 0.014(12) 0.000 0.009(11)
13 0.000 -0.012(12) 0.000 0.001(14)
14 -0.001 -0.003(12) 0.000 -0.009(13)
15 -0.001 0.007(11) -0.001 -0.013(13)
16 -0.002 0.015(09) -0.001 -0.024(16)
17 -0.002 0.020(10) -0.001 -0.005(12)
18 -0.003 0.014(13) -0.001 -0.001(12)
19 -0.004 -0.003(11) -0.002 -0.015(15)
20 -0.006 0.001(09) -0.002 -0.009(13)
21 -0.005 -0.003(11) -0.002 -0.002(13)
22 -0.001 0.010(11) 0.000 0.006(11)
23 0.021 0.004(12) 0.006 0.009(16)
24 0.025 0.037(11) 0.011 0.003(15)
25 0.033 0.022(13) 0.015 0.022(13)
26 0.078 0.076(12) 0.036 0.023(17)
27 0.196 0.132(11) 0.096 0.087(17)
28 0.241 0.207(12) 0.265 0.210(12)
29 0.186 0.172(15) 0.199 0.206(16)
30 0.082 0.106(13) 0.092 0.086(19)
31 0.172 0.174(10) 0.188 0.190(14)
32 0.179 0.183(12) 0.232 0.223(13)
33 0.148 0.104(11) 0.111 0.114(12)
34 0.044 0.034(13) 0.039 0.032(13)
35 0.028 0.010(09) 0.017 0.011(12)
36 0.014 0.048(08) 0.014 0.023(16)
37 0.013 0.030(12) 0.009 0.022(13)
38 0.003 0.021(11) 0.003 -0.001(15)
39 0.000 0.002(10) 0.000 -0.001(13)
40 0.000 -0.005(11) 0.000 0.015(15)
41 0.000 -0.009(11) 0.000 0.006(18)
42 0.000 0.019(10) 0.000 -0.008(14)
43 0.000 0.023(11) 0.000 -0.001(14)
44 0.000 -0.007(10) 0.000 -0.011(14)
45 0.000 0.018(13) 0.000 -0.009(14)
46 0.000 0.010(12) 0.000 -0.012(12)
47 0.000 0.003(10) 0.000 -0.025(13)
48 0.000 -0.004(10) 0.000 -0.011(13)
49 0.000 -0.022(13) 0.000 -0.016(15)
50 0.000 0.027(11) 0.000 0.025(15)
51 0.000 0.015(12) 0.000 -0.003(15)
52 0.000 -0.012(13) 0.000 -0.009(14)
53 0.000 0.010(12) 0.000 -0.002(11)
54 0.000 -0.012(11) 0.000 0.002(15)
55 0.000 -0.002(15) 0.000 0.009(16)
56 0.000 0.013(14) 0.000 0.005(14)
57 0.000 0.014(11) 0.000 0.007(14)
58 0.000 0.018(12) 0.000 -0.005(16)
59 0.000 0.000(11) 0.000 -0.004(14)

TABLE XXIII: Numerical values for the expectation value ⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac at t = 8 from MPS and ibm fez for the
free and interacting theories, as shown in Figs. 9 and 10.
 44

⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac

Spatial site λ=0 λ=2
j MPS ibm fez MPS ibm fez
0 0.000 0.007(12) 0.000 -0.017(18)
1 0.000 0.017(11) 0.000 -0.003(15)
2 0.000 0.033(11) 0.000 0.022(15)
3 0.000 0.020(14) 0.000 0.007(13)
4 0.000 -0.001(13) 0.000 0.036(14)
5 0.000 0.033(12) 0.000 -0.002(19)
6 0.000 0.015(12) 0.000 0.007(14)
7 0.000 0.025(11) 0.000 0.006(15)
8 0.000 0.030(12) 0.000 -0.001(14)
9 0.000 0.013(10) 0.000 0.003(13)
10 0.000 0.013(12) 0.000 0.043(19)
11 0.000 0.027(12) 0.000 0.011(14)
12 0.000 0.033(12) 0.000 0.004(18)
13 0.000 0.017(11) 0.000 0.028(13)
14 -0.001 0.013(11) 0.000 0.005(19)
15 -0.001 0.015(12) -0.001 0.047(16)
16 -0.002 0.023(14) -0.001 -0.008(16)
17 -0.002 0.009(10) -0.001 0.008(12)
18 -0.002 0.031(12) -0.001 0.008(13)
19 -0.001 0.031(13) -0.002 0.034(15)
20 -0.002 0.043(13) -0.001 -0.001(18)
21 0.002 0.027(13) 0.000 0.011(13)
22 0.018 0.035(10) 0.005 0.007(16)
23 0.028 0.047(12) 0.009 0.023(15)
24 0.022 0.007(12) 0.011 0.009(13)
25 0.042 0.043(11) 0.028 0.024(18)
26 0.127 0.080(12) 0.067 0.032(14)
27 0.221 0.170(12) 0.140 0.096(17)
28 0.134 0.173(10) 0.215 0.183(13)
29 0.029 0.103(12) 0.106 0.166(14)
30 0.087 0.080(15) 0.058 0.054(11)
31 0.025 0.123(12) 0.121 0.180(15)
32 0.125 0.142(10) 0.179 0.220(15)
33 0.229 0.141(13) 0.112 0.099(11)
34 0.098 0.053(10) 0.066 0.024(12)
35 0.029 0.041(12) 0.025 0.012(15)
36 0.014 0.049(11) 0.012 0.027(13)
37 0.017 0.055(12) 0.009 0.027(14)
38 0.011 0.032(12) 0.006 0.008(14)
39 0.000 0.039(12) 0.001 -0.000(15)
40 -0.001 0.016(13) 0.000 0.006(11)
41 0.000 0.019(11) 0.000 0.002(15)
42 0.000 0.016(11) 0.000 0.001(13)
43 0.000 0.029(10) 0.000 0.035(16)
44 0.000 0.027(13) 0.000 0.037(23)
45 0.000 0.028(10) 0.000 0.023(14)
46 0.000 0.032(11) 0.000 -0.002(15)
47 0.000 0.020(11) 0.000 0.026(17)
48 0.000 0.031(11) 0.000 0.024(12)
49 0.000 0.008(11) 0.000 0.016(14)
50 0.000 0.025(11) 0.000 0.005(11)
51 0.000 0.009(12) 0.000 -0.010(14)
52 0.000 0.008(13) 0.000 -0.002(12)
53 0.000 0.029(13) 0.000 0.004(17)
54 0.000 0.019(10) 0.000 0.017(15)
55 0.000 0.015(12) 0.000 -0.001(12)
56 0.000 0.019(11) 0.000 0.018(13)
57 0.000 0.029(10) 0.000 0.015(16)
58 0.000 0.013(13) 0.000 0.013(17)
59 0.000 0.046(13) 0.000 0.009(17)

TABLE XXIV: Numerical values for the expectation value ⟨ϕ2j ⟩2wp − ⟨ϕ2j ⟩vac at t = 9 from MPS and ibm fez for the
free and interacting theories, as shown in Figs. 9 and 10.
 45

[1] J. Preskill, Quantum Computing in the NISQ era and beyond, Quantum 2, 79 (2018), arXiv:1801.00862 [quant-ph].
[2] S. P. Jordan, Quantum algorithm zoo (2024).
[3] K. Temme, S. Bravyi, and J. M. Gambetta, Error Mitigation for Short-Depth Quantum Circuits, Phys. Rev. Lett. 119,
180509 (2017), arXiv:1612.02058 [quant-ph].
[4] Z. Cai, R. Babbush, S. C. Benjamin, S. Endo, W. J. Huggins, Y. Li, J. R. McClean, and T. E. O’Brien, Quantum error
mitigation, Rev. Mod. Phys. 95, 045005 (2023), arXiv:2210.00921 [quant-ph].
[5] H. Liao, D. S. Wang, I. Sitdikov, C. Salcedo, A. Seif, and Z. K. Minev, Machine Learning for Practical Quantum Error
Mitigation, (2023), arXiv:2309.17368 [quant-ph].
[6] F. Arute et al., Quantum supremacy using a programmable superconducting processor, Nature 574, 505 (2019),
arXiv:1910.11333 [quant-ph].
[7] H. Yu, Y. Zhao, and T.-C. Wei, Simulating large-size quantum spin chains on cloud-based superconducting quantum
computers, Phys. Rev. Res. 5, 013183 (2023), arXiv:2207.09994 [quant-ph].
[8] O. Shtanko, D. S. Wang, H. Zhang, N. Harle, A. Seif, R. Movassagh, and Z. Minev, Uncovering Local Integrability in
Quantum Many-Body Dynamics, (2023), arXiv:2307.07552 [quant-ph].
[9] Y. Kim et al., Evidence for the utility of quantum computing before fault tolerance, Nature 618, 500 (2023).
[10] R. C. Farrell, M. Illa, A. N. Ciavarella, and M. J. Savage, Scalable Circuits for Preparing Ground States on Digital
Quantum Computers: The Schwinger Model Vacuum on 100 Qubits, PRX Quantum 5, 020315 (2024), arXiv:2308.04481
[quant-ph].
[11] E. Bäumer, V. Tripathi, D. S. Wang, P. Rall, E. H. Chen, S. Majumder, A. Seif, and Z. K. Minev, Efficient Long-Range
Entanglement Using Dynamic Circuits, PRX Quantum 5, 030339 (2024), arXiv:2308.13065 [quant-ph].
[12] E. H. Chen et al., Realizing the Nishimori transition across the error threshold for constant-depth quantum circuits,
(2023), arXiv:2309.02863 [quant-ph].
[13] D. Bluvstein et al., Logical quantum processor based on reconfigurable atom arrays, Nature 626, 58 (2024), arXiv:2312.03982
[quant-ph].
[14] E. Pelofske, A. Bärtschi, L. Cincio, J. Golden, and S. Eidenbenz, Scaling Whole-Chip QAOA for Higher-Order Ising Spin
Glass Models on Heavy-Hex Graphs, (2023), arXiv:2312.00997 [quant-ph].
[15] R. C. Farrell, M. Illa, A. N. Ciavarella, and M. J. Savage, Quantum simulations of hadron dynamics in the Schwinger
model using 112 qubits, Phys. Rev. D 109, 114510 (2024), arXiv:2401.08044 [quant-ph].
[16] T. A. Chowdhury, K. Yu, M. A. Shamim, M. L. Kabir, and R. S. Sufian, Enhancing quantum utility: Simulating large-scale
quantum spin chains on superconducting quantum computers, Phys. Rev. Res. 6, 033107 (2024), arXiv:2312.12427
[quant-ph].
[17] A. N. Ciavarella and C. W. Bauer, Quantum Simulation of SU(3) Lattice Yang-Mills Theory at Leading Order in Large-Nc
Expansion, Phys. Rev. Lett. 133, 111901 (2024), arXiv:2402.10265 [hep-ph].
[18] K. Shinjo, K. Seki, T. Shirakawa, R.-Y. Sun, and S. Yunoki, Unveiling clean two-dimensional discrete time quasicrystals
on a digital quantum computer, (2024), arXiv:2403.16718 [quant-ph].
[19] Z. Bao et al., Creating and controlling global Greenberger-Horne-Zeilinger entanglement on quantum processors, Nature
Commun. 15, 8823 (2024), arXiv:2401.08284 [quant-ph].
[20] T. Hayata and Y. Hidaka, Floquet evolution of the q-deformed SU(3)1 Yang-Mills theory on a two-leg ladder, (2024),
arXiv:2409.20263 [hep-lat].
[21] T. Hayata, K. Seki, and A. Yamamoto, Floquet prethermalization of Z2 lattice gauge theory on superconducting qubits,
(2024), arXiv:2408.10079 [hep-lat].
[22] J. R. McClean, J. Romero, R. Babbush, and A. Aspuru-Guzik, The theory of variational hybrid quantum-classical
algorithms, New J. Phys. 18, 023023 (2016), arXiv:1509.04279 [quant-ph].
[23] M. Cerezo et al., Variational quantum algorithms, Nature Rev. Phys. 3, 625 (2021), arXiv:2012.09265 [quant-ph].
[24] J. Biamonte, Universal variational quantum computation, Phys. Rev. A 103, L030401 (2021).
[25] A. Peruzzo, J. McClean, P. Shadbolt, M.-H. Yung, X.-Q. Zhou, P. J. Love, A. Aspuru-Guzik, and J. L. O’Brien, A
variational eigenvalue solver on a photonic quantum processor, Nature Commun. 5, 4213 (2014), arXiv:1304.3061 [quant-ph].
[26] E. Farhi, J. Goldstone, and S. Gutmann, A Quantum Approximate Optimization Algorithm, (2014), arXiv:1411.4028
[quant-ph].
[27] J. Biamonte, P. Wittek, N. Pancotti, P. Rebentrost, N. Wiebe, and S. Lloyd, Quantum machine learning, Nature 549, 195
(2017), arXiv:1611.09347 [quant-ph].
[28] C. Bravo-Prieto, R. LaRose, M. Cerezo, Y. Subasi, L. Cincio, and P. J. Coles, Variational Quantum Linear Solver, Quantum
7, 1188 (2023), arXiv:1909.05820 [quant-ph].
[29] X. Yuan, S. Endo, Q. Zhao, Y. Li, and S. Benjamin, Theory of variational quantum simulation, Quantum 3, 191 (2019),
arXiv:1812.08767 [quant-ph].
[30] A. J. Daley, I. Bloch, C. Kokail, S. Flannigan, N. Pearson, M. Troyer, and P. Zoller, Practical quantum advantage in
quantum simulation, Nature 607, 667 (2022).
[31] C. W. Bauer, Z. Davoudi, N. Klco, and M. J. Savage, Quantum simulation of fundamental particles and forces, Nature
Rev. Phys. 5, 420 (2023), arXiv:2404.06298 [hep-ph].
[32] M. Lüscher, Volume Dependence of the Energy Spectrum in Massive Quantum Field Theories. 1. Stable Particle States,
Commun. Math. Phys. 104, 177 (1986).
 46

[33] M. Lüscher, Volume Dependence of the Energy Spectrum in Massive Quantum Field Theories. 2. Scattering States,
Commun. Math. Phys. 105, 153 (1986).
[34] S. R. Beane, K. Orginos, and M. J. Savage, Hadronic Interactions from Lattice QCD, Int. J. Mod. Phys. E 17, 1157 (2008),
arXiv:0805.4629 [hep-lat].
[35] S. R. Beane, W. Detmold, K. Orginos, and M. J. Savage, Nuclear Physics from Lattice QCD, Prog. Part. Nucl. Phys. 66,
1 (2011), arXiv:1004.2935 [hep-lat].
[36] T. Luu and M. J. Savage, Extracting Scattering Phase-Shifts in Higher Partial-Waves from Lattice QCD Calculations,
Phys. Rev. D 83, 114508 (2011), arXiv:1101.3347 [hep-lat].
[37] R. A. Briceno, Z. Davoudi, T. C. Luu, and M. J. Savage, Two-Baryon Systems with Twisted Boundary Conditions, Phys.
Rev. D 89, 074509 (2014), arXiv:1311.7686 [hep-lat].
[38] M. T. Hansen, Multi-Hadron Observables from Lattice Quantum Chromodynamics, Ph.D. thesis, Washington U., Seattle
(2014).
[39] A. Milsted, J. Liu, J. Preskill, and G. Vidal, Collisions of False-Vacuum Bubble Walls in a Quantum Spin Chain, PRX
Quantum 3, 020316 (2022), arXiv:2012.07243 [quant-ph].
[40] B. Vanhecke, F. Verstraete, and K. Van Acoleyen, Entanglement scaling for λϕ24, Phys. Rev. D 106, L071501 (2022),
arXiv:2104.10564 [hep-lat].
[41] M. Van Damme, L. Vanderstraeten, J. De Nardis, J. Haegeman, and F. Verstraete, Real-time scattering of interacting
quasiparticles in quantum spin chains, Phys. Rev. Res. 3, 013078 (2021).
[42] M. Rigobello, S. Notarnicola, G. Magnifico, and S. Montangero, Entanglement generation in (1+1)D QED scattering
processes, Phys. Rev. D 104, 114501 (2021), arXiv:2105.03445 [hep-lat].
[43] R. Belyansky, S. Whitsitt, N. Mueller, A. Fahimniya, E. R. Bennewitz, Z. Davoudi, and A. V. Gorshkov, High-Energy
Collision of Quarks and Mesons in the Schwinger Model: From Tensor Networks to Circuit QED, Phys. Rev. Lett. 132,
091903 (2024), arXiv:2307.02522 [quant-ph].
[44] I. Papaefstathiou, J. Knolle, and M. C. Bañuls, Real-time scattering in the lattice Schwinger model, (2024), arXiv:2402.18429
[hep-lat].
[45] E. Gustafson, Y. Meurice, and J. Unmuth-Yockey, Quantum simulation of scattering in the quantum Ising model, Phys.
Rev. D 99, 094503 (2019), arXiv:1901.05944 [hep-lat].
[46] Y.-C. Huang, F.-M. He, and S.-Y. Lin, Quantum mechanical gaussian wavepackets of single relativistic particles (2023),
arXiv:2309.09429 [quant-ph].
[47] I. Tutunnikov, C. Chuang, and J. Cao, Coherent spatial control of wave packet dynamics on quantum lattices, J. Phys.
Chem. Lett. 14, 51 (2023), arXiv:2311.07254 [quant-ph].
[48] C. F. Kane, N. Gomes, and M. Kreshchuk, Nearly-optimal state preparation for quantum simulations of lattice gauge
theories, (2023), arXiv:2310.13757 [quant-ph].
[49] Z. Davoudi, C.-C. Hsieh, and S. V. Kadam, Scattering wave packets of hadrons in gauge theories: Preparation on a
quantum computer, (2024), arXiv:2402.00840 [quant-ph].
[50] T. Li, W. K. Lai, E. Wang, and H. Xing (QuNu), Scattering amplitude from quantum computing with reduction formula,
Phys. Rev. D 109, 036025 (2024), arXiv:2301.04179 [hep-ph].
[51] R. C. Farrell, M. Illa, and M. J. Savage, Steps Toward Quantum Simulations of Hadronization and Energy-Loss in Dense
Matter, (2024), arXiv:2405.06620 [quant-ph].
[52] P. Dreher, E. Gustafson, Y. Zhu, N. M. Linke, and Y. Meurice, Real-time Quantum Calculations of Phase Shifts On NISQ
Hardware Platforms Using Wavepacket Time Delay, PoS LATTICE2021, 464 (2022).
[53] M. Turco, G. M. Quinta, J. a. Seixas, and Y. Omar, Quantum Simulation of Bound State Scattering, PRX Quantum 5,
020311 (2024), arXiv:2305.07692 [quant-ph].
[54] M. Kreshchuk, J. P. Vary, and P. J. Love, Simulating Scattering of Composite Particles, (2023), arXiv:2310.13742
[quant-ph].
[55] Y. Chai, A. Crippa, K. Jansen, S. Kühn, V. R. Pascuzzi, F. Tacchino, and I. Tavernelli, Fermionic wave packet scattering:
a quantum computing approach, (2023), arXiv:2312.02272 [quant-ph].
[56] R. A. Briceño, R. G. Edwards, M. Eaton, C. González-Arciniegas, O. Pfister, and G. Siopsis, Toward coherent quantum
computation of scattering amplitudes with a measurement-based photonic quantum processor, Phys. Rev. Res. 6, 043065
(2024), arXiv:2312.12613 [quant-ph].
[57] S. Sharma, T. Papenbrock, and L. Platter, Scattering phase shifts from a quantum computer, (2023), arXiv:2311.09298
[nucl-th].
[58] F. Turro et al., Demonstration of a quantum-classical coprocessing protocol for simulating nuclear reactions, Phys. Rev. A
108, 032417 (2023).
[59] P. Wang, W. Du, W. Zuo, and J. P. Vary, Nuclear scattering via quantum computing, (2024), arXiv:2401.17138 [nucl-th].
[60] S. Wu, W. Du, X. Zhao, and J. P. Vary, Ultra-relativistic quark-nucleus scattering on quantum computers, (2024),
arXiv:2404.00819 [quant-ph].
[61] E. R. Bennewitz, B. Ware, A. Schuckert, A. Lerose, F. M. Surace, R. Belyansky, W. Morong, D. Luo, A. De, K. S. Collins,
O. Katz, C. Monroe, Z. Davoudi, and A. V. Gorshkov, Simulating Meson Scattering on Spin Quantum Simulators, (2024),
arXiv:2403.07061 [quant-ph].
[62] F. Turro, K. A. Wendt, S. Quaglioni, F. Pederiva, and A. Roggero, Evaluation of phase shifts for non-relativistic elastic
scattering using quantum computers, (2024), arXiv:2407.04155 [quant-ph].
[63] M. Yusf, L. Gan, C. Moffat, and G. Rupak, Elastic scattering on a quantum computer, (2024), arXiv:2406.09231 [nucl-th].
 47

[64] D. C. Brydges, J. Frohlich, and A. D. Sokal, A NEW PROOF OF THE EXISTENCE AND NONTRIVIALITY OF THE
CONTINUUM PHI**4 in two-dimensions AND PHI**4 in three-dimensions QUANTUM FIELD THEORIES, Commun.
Math. Phys. 91, 141 (1983).
[65] S. Weinberg, Phenomenological Lagrangians, Physica A 96, 327 (1979).
[66] E. T. Holland, K. A. Wendt, K. Kravvaris, X. Wu, W. Erich Ormand, J. L. DuBois, S. Quaglioni, and F. Pederiva, Optimal
Control for the Quantum Simulation of Nuclear Dynamics, Phys. Rev. A 101, 062307 (2020), arXiv:1908.08222 [quant-ph].
[67] R. Shankar, Renormalization group approach to interacting fermions, Rev. Mod. Phys. 66, 129 (1994), arXiv:cond-
mat/9307009.
[68] T. Schäfer, Effective Theories of Dense and Very Dense Matter, Lect. Notes Phys. 852, 193 (2012), arXiv:nucl-th/0609075.
[69] P. W. Higgs, Broken Symmetries and the Masses of Gauge Bosons, Phys. Rev. Lett. 13, 508 (1964).
[70] P. W. Higgs, Broken symmetries, massless particles and gauge fields, Phys. Lett. 12, 132 (1964).
[71] I. Caprini, G. Colangelo, and H. Leutwyler, Mass and width of the lowest resonance in QCD, Phys. Rev. Lett. 96, 132001
(2006), arXiv:hep-ph/0512364.
[72] A. H. Guth, Inflationary universe: A possible solution to the horizon and flatness problems, Phys. Rev. D 23, 347 (1981).
[73] R. D. Peccei and H. R. Quinn, CP Conservation in the Presence of Instantons, Phys. Rev. Lett. 38, 1440 (1977).
[74] R. D. Peccei, The Strong CP problem and axions, Lect. Notes Phys. 741, 3 (2008), arXiv:hep-ph/0607268.
[75] E. Rinaldi, X. Han, M. Hassan, Y. Feng, F. Nori, M. McGuigan, and M. Hanada, Matrix-Model Simulations Using
Quantum Computing, Deep Learning, and Lattice Monte Carlo, PRX Quantum 3, 010324 (2022), arXiv:2108.02942
[quant-ph].
[76] S. Rychkov and L. G. Vitale, Hamiltonian truncation study of the ϕ4 theory in two dimensions, Phys. Rev. D 91, 085011
(2015), arXiv:1412.3460 [hep-th].
[77] S. Rychkov and L. G. Vitale, Hamiltonian truncation study of the ϕ4 theory in two dimensions. II. The Z2 -broken phase
and the Chang duality, Phys. Rev. D 93, 065014 (2016), arXiv:1512.00493 [hep-th].
[78] S. Thompson and G. Siopsis, Quantum computation of phase transition in interacting scalar quantum field theory, Quant.
Inf. Proc. 22, 396 (2023), arXiv:2303.02425 [quant-ph].
[79] S. P. Jordan, H. Krovi, K. S. M. Lee, and J. Preskill, BQP-completeness of scattering in scalar quantum field theory,
Quantum 2, 44 (2018).
[80] S. P. Jordan, K. S. M. Lee, and J. Preskill, Quantum Computation of Scattering in Scalar Quantum Field Theories, Quant.
Inf. Comput. 14, 1014 (2014), arXiv:1112.4833 [hep-th].
[81] S. P. Jordan, K. S. M. Lee, and J. Preskill, Quantum Algorithms for Quantum Field Theories, Science 336, 1130 (2012),
arXiv:1111.3633 [quant-ph].
[82] N. Klco and M. J. Savage, Digitization of scalar fields for quantum computing, Phys. Rev. A 99, 052335 (2019),
arXiv:1808.10378 [quant-ph].
[83] N. Klco and M. J. Savage, Systematically Localizable Operators for Quantum Simulations of Quantum Field Theories,
Phys. Rev. A 102, 012619 (2020), arXiv:1912.03577 [quant-ph].
[84] N. Klco and M. J. Savage, Minimally entangled state preparation of localized wave functions on quantum computers, Phys.
Rev. A 102, 012612 (2020), arXiv:1904.10440 [quant-ph].
[85] N. Klco and M. J. Savage, Fixed-point quantum circuits for quantum field theories, Phys. Rev. A 102, 052422 (2020),
arXiv:2002.02018 [quant-ph].
[86] D. M. Kurkcuoglu, M. S. Alam, J. A. Job, A. C. Y. Li, A. Macridin, G. N. Perdue, and S. Providence, Quantum simulation
of ϕ4 theories in qudit systems, (2021), arXiv:2108.13357 [quant-ph].
[87] A. Macridin, A. C. Y. Li, S. Mrenna, and P. Spentzouris, Bosonic field digitization for quantum computers, Phys. Rev. A
105, 052405 (2022), arXiv:2108.10793 [quant-ph].
[88] J. Liu, Z. Li, H. Zheng, X. Yuan, and J. Sun, Towards a variational Jordan–Lee–Preskill quantum algorithm, Mach. Learn.
Sci. Tech. 3, 045030 (2022), arXiv:2109.05547 [quant-ph].
[89] M. Illa and M. J. Savage, Basic elements for simulations of standard-model physics with quantum annealers: Multigrid
and clock states, Phys. Rev. A 106, 052605 (2022), arXiv:2202.12340 [quant-ph].
[90] A. C. Y. Li, A. Macridin, S. Mrenna, and P. Spentzouris, Simulating scalar field theories on quantum computers with
limited resources, Phys. Rev. A 107, 032603 (2023), arXiv:2210.07985 [quant-ph].
[91] Z. Ozzello and Y. Meurice, Bootstrap methods for digitized scalar field theory, PoS LATTICE2023, 163 (2024),
arXiv:2312.10559 [hep-th].
[92] R. Haghshenas, J. Gray, A. C. Potter, and G. K.-L. Chan, Variational Power of Quantum Circuit Tensor Networks, Phys.
Rev. X 12, 011047 (2022), arXiv:2107.01307 [quant-ph].
[93] L. Leone, S. F. E. Oliviero, L. Cincio, and M. Cerezo, On the practical usefulness of the Hardware Efficient Ansatz,
Quantum 8, 1395 (2024), arXiv:2211.01477 [quant-ph].
[94] S. Filippov, B. Sokolov, M. A. C. Rossi, J. Malmi, E.-M. Borrelli, D. Cavalcanti, S. Maniscalco, and G. Garcı́a-Pérez,
Matrix product channel: Variationally optimized quantum tensor network to mitigate noise and reduce errors for the
variational quantum eigensolver, (2022), arXiv:2212.10225 [quant-ph].
[95] K. N. Okada, K. Osaki, K. Mitarai, and K. Fujii, Classically optimized variational quantum eigensolver with applications
to topological phases, Phys. Rev. Res. 5, 043217 (2023), arXiv:2202.02909 [quant-ph].
[96] M. S. J. Tepaske, D. Hahn, and D. J. Luitz, Optimal compression of quantum many-body time evolution operators into
brickwall circuits, SciPost Phys. 14, 073 (2023), arXiv:2205.03445 [cond-mat.str-el].
[97] T. Barthel and Q. Miao, Absence of barren plateaus and scaling of gradients in the energy optimization of isometric tensor
network states, (2023), arXiv:2304.00161 [quant-ph].
 48

[98] Q. Miao and T. Barthel, Convergence and Quantum Advantage of Trotterized MERA for Strongly-Correlated Systems,
(2023), arXiv:2303.08910 [quant-ph].
[99] M. S. J. Tepaske, D. J. Luitz, and D. Hahn, Optimal compression of constrained quantum time evolution, (2023),
arXiv:2311.06347 [quant-ph].
[100] N. F. Robertson, A. Akhriev, J. Vala, and S. Zhuk, Approximate Quantum Compiling for Quantum Simulation: A Tensor
Network based approach, (2023), arXiv:2301.08609 [quant-ph].
[101] Y. Zhang, R. Wiersema, J. Carrasquilla, L. Cincio, and Y. B. Kim, Scalable quantum dynamics compilation via quantum
machine learning, (2024), arXiv:2409.16346 [quant-ph].
[102] L. Wright, C. M. Keever, J. T. First, R. Johnston, J. Tillay, S. Chaney, M. Rosenkranz, and M. Lubasch, Noisy
intermediate-scale quantum simulation of the one-dimensional wave equation, (2024), arXiv:2402.19247 [quant-ph].
[103] M. M. Ibrahim, H. Mohammadbagherpoor, C. Rios, N. T. Bronn, and G. T. Byrd, Evaluation of Parameterized Quantum
Circuits With Cross-Resonance Pulse-Driven Entanglers, IEEE Trans. Quantum Eng. 3, 1 (2022), arXiv:2211.00350
[quant-ph].
[104] L. Causer, F. Jung, A. Mitra, F. Pollmann, and A. Gammon-Smith, Scalable simulation of nonequilibrium quantum
dynamics via classically optimized unitary circuits, Phys. Rev. Res. 6, 033062 (2024), arXiv:2312.14245 [quant-ph].
[105] Q. Wang, L. Zhukas, Q. Miao, A. S. Dalvi, P. J. Love, C. Monroe, F. T. Chong, and G. S. Ravi, Demonstration of a
CAFQA-bootstrapped Variational Quantum Eigensolver on a Trapped-Ion Quantum Computer, (2024), arXiv:2408.06482
[quant-ph].
[106] J. Z. Zhuang, Y. K. Wu, and L. M. Duan, Hardware-efficient variational quantum algorithm in trapped-ion quantum
computer, (2024), arXiv:2407.03116 [quant-ph].
[107] N. Sachdeva et al., Quantum optimization using a 127-qubit gate-model IBM quantum computer can outperform quantum
annealers for nontrivial binary optimization problems, (2024), arXiv:2406.01743 [quant-ph].
[108] Z. Li, D. M. Grabowska, and M. J. Savage, Sequency Hierarchy Truncation (SeqHT) for Adiabatic State Preparation and
Time Evolution in Quantum Simulations, (2024), arXiv:2407.13835 [quant-ph].
[109] R. D. Somma, Quantum simulations of one dimensional quantum systems, Quantum Info. Comput. 16, 1125–1168 (2016).
[110] A. Macridin, P. Spentzouris, J. Amundson, and R. Harnik, Digital quantum computation of fermion-boson interacting
systems, Phys. Rev. A 98, 042312 (2018), arXiv:1805.09928 [quant-ph].
[111] A. Macridin, P. Spentzouris, J. Amundson, and R. Harnik, Electron-Phonon Systems on a Universal Quantum Computer,
Phys. Rev. Lett. 121, 110504 (2018), arXiv:1802.07347 [quant-ph].
[112] C. W. Bauer and D. M. Grabowska, Efficient representation for simulating U(1) gauge theories on digital quantum
computers at all values of the coupling, Phys. Rev. D 107, L031503 (2023), arXiv:2111.08015 [hep-ph].
[113] C. Kane, D. M. Grabowska, B. Nachman, and C. W. Bauer, Efficient quantum implementation of 2+1 U(1) lattice gauge
theories with Gauss law constraints, (2022), arXiv:2211.10497 [quant-ph].
[114] A. G. Fowler, S. J. Devitt, and L. C. L. Hollenberg, Implementation of shor’s algorithm on a linear nearest neighbour
qubit array, Quantum Info. Comput. 4, 237–251 (2004).
[115] B. Park and D. Ahn, Reducing CNOT count in quantum Fourier transform for the linear nearest-neighbor architecture,
Sci. Rep. 13, 8638 (2023).
[116] M. Suzuki, Generalized trotter’s formula and systematic approximants of exponential operators and inner derivations with
applications to many-body problems, Communications in Mathematical Physics 51, 183–190 (1976).
[117] E. Knill, Approximation by Quantum Circuits, (1995), arXiv:quant-ph/9508006.
[118] M. B. Hastings and T. Koma, Spectral gap and exponential decay of correlations, Commun. Math. Phys. 265, 781 (2006),
arXiv:math-ph/0507008.
[119] H.-Y. Huang, J. Preskill, and M. Soleimanifar, Certifying almost all quantum states with few single-qubit measurements,
(2024), arXiv:2404.07281 [quant-ph].
[120] R. Mansuroglu, T. Eckstein, L. Nützel, S. A. Wilkinson, and M. J. Hartmann, Variational Hamiltonian simulation for
translational invariant systems via classical pre-processing, Quantum Sci. Technol. 8, 025006 (2023), arXiv:2106.03680
[quant-ph].
[121] Y. Guo, T. Angelides, K. Jansen, and S. Kühn, Concurrent VQE for Simulating Excited States of the Schwinger Model,
(2024), arXiv:2407.15629 [quant-ph].
[122] H. R. Grimsley, S. E. Economou, E. Barnes, and N. J. Mayhall, An adaptive variational algorithm for exact molecular
simulations on a quantum computer, Nature Communications 10, 10.1038/s41467-019-10988-2 (2019).
[123] C. Feniou, M. Hassan, D. Traoré, E. Giner, Y. Maday, and J.-P. Piquemal, Overlap-ADAPT-VQE: practical quantum
chemistry on quantum computers via overlap-guided compact Ansätze, Commun. Phys. 6, 192 (2023), arXiv:2301.10196
[quant-ph].
[124] J. S. Van Dyke, K. Shirali, G. S. Barron, N. J. Mayhall, E. Barnes, and S. E. Economou, Scaling adaptive quantum
simulation algorithms via operator pool tiling, Phys. Rev. Res. 6, L012030 (2024), arXiv:2206.14215 [quant-ph].
[125] E. Gustafson et al., Surrogate Constructed Scalable Circuits ADAPT-VQE in the Schwinger model, (2024), arXiv:2408.12641
[quant-ph].
[126] M. Cerezo et al., Does provable absence of barren plateaus imply classical simulability? Or, why we need to rethink
variational quantum computing, (2023), arXiv:2312.09121 [quant-ph].
[127] M. Larocca, S. Thanasilp, S. Wang, K. Sharma, J. Biamonte, P. J. Coles, L. Cincio, J. R. McClean, Z. Holmes, and
M. Cerezo, A Review of Barren Plateaus in Variational Quantum Computing, (2024), arXiv:2405.00781 [quant-ph].
[128] S. Sim, P. D. Johnson, and A. Aspuru-Guzik, Expressibility and Entangling Capability of Parameterized Quantum Circuits
for Hybrid Quantum-Classical Algorithms, Adv. Quantum Technol. 2, 1900070 (2019), arXiv:1905.10876 [quant-ph].
 49

[129] K. Seki, T. Shirakawa, and S. Yunoki, Symmetry-adapted variational quantum eigensolver, Phys. Rev. A 101, 052340
(2020).
[130] L. Funcke, T. Hartung, K. Jansen, S. Kühn, and P. Stornati, Dimensional Expressivity Analysis of Parametric Quantum
Circuits, Quantum 5, 422 (2021), arXiv:2011.03532 [quant-ph].
[131] E. R. Anschuetz and B. T. Kiani, Quantum variational algorithms are swamped with traps, Nature Commun. 13, 7760
(2022), arXiv:2205.05786 [quant-ph].
[132] J. Gibbs, Z. Holmes, and P. Stevenson, Exploiting symmetries in nuclear Hamiltonians for ground state preparation,
(2024), arXiv:2402.10277 [quant-ph].
[133] L. Madden and A. Simonetto, Best Approximate Quantum Compiling Problems, ACM Trans. Quant. Comput. 3, 7 (2022),
arXiv:2106.05649 [quant-ph].
[134] M. Motta, C. Sun, A. T. K. Tan, M. J. O. Rourke, E. Ye, A. J. Minnich, F. G. S. L. Brandão, and G. K.-L. Chan,
Determining eigenstates and thermal states on a quantum computer using quantum imaginary time evolution, Nature
Phys. 16, 205 (2019), arXiv:1901.07653 [quant-ph].
[135] N. Yoshioka et al., Diagonalization of large many-body Hamiltonians on a quantum processor, (2024), arXiv:2407.14431
[quant-ph].
[136] G. S. Ravi, P. Gokhale, Y. Ding, W. M. Kirby, K. N. Smith, J. M. Baker, P. J. Love, H. Hoffmann, K. R. Brown, and
F. T. Chong, CAFQA: A classical simulation bootstrap for variational quantum algorithms, (2022), arXiv:2202.12924
[quant-ph].
[137] A. Khan, B. K. Clark, and N. M. Tubman, Pre-optimizing variational quantum eigensolvers with tensor networks, (2023),
arXiv:2310.12965 [quant-ph].
[138] S. Fomichev et al., Initial state preparation for quantum chemistry on quantum computers, (2023), arXiv:2310.18410
[quant-ph].
[139] E. Campbell, Random Compiler for Fast Hamiltonian Simulation, Phys. Rev. Lett. 123, 070503 (2019).
[140] A. M. Childs and N. Wiebe, Hamiltonian Simulation Using Linear Combinations of Unitary Operations, Quant. Inf.
Comput. 12, 0901 (2012), arXiv:1202.5822 [quant-ph].
[141] A. Kotil, R. Banerjee, Q. Huang, and C. B. Mendl, Riemannian quantum circuit optimization for Hamiltonian simulation,
J. Phys. A 57, 135303 (2024), arXiv:2212.07556 [quant-ph].
[142] K. Mizuta, Y. O. Nakagawa, K. Mitarai, and K. Fujii, Local Variational Quantum Compilation of Large-Scale Hamiltonian
Dynamics, PRX Quantum 3, 040302 (2022), arXiv:2203.15484 [quant-ph].
[143] C. M. Keever and M. Lubasch, Classically optimized Hamiltonian simulation, Phys. Rev. Res. 5, 023146 (2023),
arXiv:2205.11427 [quant-ph].
[144] S. Miyakoshi, T. Sugimoto, T. Shirakawa, S. Yunoki, and H. Ueda, Diamond-shaped quantum circuit for real-time quantum
dynamics in one dimension, (2023), arXiv:2311.05900 [quant-ph].
[145] S. Kanasugi, Y. Hidaka, Y. O. Nakagawa, S. Tsutsui, N. Matsumoto, K. Maruyama, H. Oshima, and S. Sato, Subspace-
Based Local Compilation of Variational Quantum Circuits for Large-Scale Quantum Many-Body Simulation, (2024),
arXiv:2407.14163 [quant-ph].
[146] J. S. Gacon, Scalable Quantum Algorithms for Noisy Quantum Computers, Ph.D. thesis, Ecole Polytechnique, Lausanne
(2024), arXiv:2403.00940 [quant-ph].
[147] J. Gibbs and L. Cincio, Deep Circuit Compression for Quantum Dynamics via Tensor Networks, (2024), arXiv:2409.16361
[quant-ph].
[148] B. Commeau, M. Cerezo, Z. Holmes, L. Cincio, P. J. Coles, and A. Sornborger, Variational Hamiltonian Diagonalization
for Dynamical Quantum Simulation, (2020), arXiv:2009.02559 [quant-ph].
[149] S. Gu, R. D. Somma, and B. Şahinoğlu, Fast-forwarding quantum evolution, Quantum 5, 577 (2021), arXiv:2105.07304
[quant-ph].
[150] E. H. Lieb and D. W. Robinson, The finite group velocity of quantum spin systems, Commun. Math. Phys. 28, 251 (1972).
[151] A. Javadi-Abhari, M. Treinish, K. Krsulich, C. J. Wood, J. Lishman, J. Gacon, S. Martiel, P. D. Nation, L. S. Bishop,
A. W. Cross, B. R. Johnson, and J. M. Gambetta, Quantum computing with Qiskit (2024), arXiv:2405.08810 [quant-ph].
[152] L. Viola and S. Lloyd, Dynamical suppression of decoherence in two state quantum systems, Phys. Rev. A 58, 2733 (1998),
arXiv:quant-ph/9803057.
[153] N. Ezzell, B. Pokharel, L. Tewala, G. Quiroz, and D. A. Lidar, Dynamical decoupling for superconducting qubits: A
performance survey, Physical Review Applied 20, 10.1103/physrevapplied.20.064027 (2023).
[154] J. J. Wallman and J. Emerson, Noise tailoring for scalable quantum computation via randomized compiling, Phys. Rev. A
94, 052325 (2016), arXiv:1512.01098 [quant-ph].
[155] E. v. d. Berg, Z. K. Minev, and K. Temme, Model-free readout-error mitigation for quantum expectation values, Phys.
Rev. A 105, 032620 (2022), arXiv:2012.09738 [quant-ph].
[156] M. Urbanek, B. Nachman, V. R. Pascuzzi, A. He, C. W. Bauer, and W. A. de Jong, Mitigating depolarizing noise on
quantum computers with noise-estimation circuits, Phys. Rev. Lett. 127, 270502 (2021), arXiv:2103.08591 [quant-ph].
[157] S. A Rahman, R. Lewis, E. Mendicelli, and S. Powell, Self-mitigating Trotter circuits for SU(2) lattice gauge theory on a
quantum computer, Phys. Rev. D 106, 074502 (2022), arXiv:2205.09247 [hep-lat].
[158] J. Stokes, J. Izaac, N. Killoran, and G. Carleo, Quantum Natural Gradient, Quantum 4, 269 (2020), arXiv:1909.02108
[quant-ph].
[159] A. W. Harrow and J. C. Napp, Low-Depth Gradient Measurements Can Improve Convergence in Variational Hybrid
Quantum-Classical Algorithms, Phys. Rev. Lett. 126, 140502 (2021).
 50

[160] D. Malz, G. Styliaris, Z.-Y. Wei, and J. I. Cirac, Preparation of Matrix Product States with Log-Depth Quantum Circuits,
Phys. Rev. Lett. 132, 040404 (2024), arXiv:2307.01696 [quant-ph].
[161] E. Bäumer and S. Woerner, Measurement-Based Long-Range Entangling Gates in Constant Depth, (2024), arXiv:2408.03064
[quant-ph].
[162] L. Piroli, G. Styliaris, and J. I. Cirac, Approximating many-body quantum states with quantum circuits and measurements,
(2024), arXiv:2403.07604 [quant-ph].
[163] https://iqus.uw.edu (2023).
[164] https://phys.washington.edu (2023).
[165] https://artsci.washington.edu (2023).
[166] W. R. Inc., Mathematica, Version 13.2, champaign, IL, 2022.
[167] G. Van Rossum and F. L. Drake, Python 3 Reference Manual (CreateSpace, Scotts Valley, CA, 2009).
[168] J. Bezanson, A. Edelman, S. Karpinski, and V. B. Shah, Julia: A fresh approach to numerical computing, SIAM Review
59, 65 (2017).
[169] F. Perez and B. E. Granger, Ipython: A system for interactive scientific computing, Computing in Science & Engineering
9, 21 (2007).
[170] Anaconda software distribution (2020).
[171] M. Fishman, S. R. White, and E. M. Stoudenmire, The ITensor Software Library for Tensor Network Calculations, SciPost
Phys. Codebases , 4 (2022).
[172] M. Fishman, S. R. White, and E. M. Stoudenmire, Codebase release 0.3 for ITensor, SciPost Phys. Codebases , 4 (2022).
[173] https://hyak.uw.edu/ (2023).
[174] R. Pordes, D. Petravick, B. Kramer, D. Olson, M. Livny, A. Roy, P. Avery, K. Blackburn, T. Wenaus, F. Würthwein,
I. Foster, R. Gardner, M. Wilde, A. Blatecky, J. McGee, and R. Quick, The open science grid, in J. Phys. Conf. Ser., 78,
Vol. 78 (2007) p. 012057.
[175] I. Sfiligoi, D. C. Bradley, B. Holzman, P. Mhashilkar, S. Padhi, and F. Wurthwein, The pilot way to grid resources using
glideinwms, in 2009 WRI World Congress on Computer Science and Information Engineering, 2, Vol. 2 (2009) pp. 428–432.
[176] OSG, Ospool (2006).
[177] OSG, Open science data federation (2015).
[178] R. Fletcher, Practical methods of optimization (Wiley, 1986).
[179] M. Ostaszewski, E. Grant, and M. Benedetti, Structure optimization for parameterized quantum circuits, Quantum 5, 391
(2021), arXiv:1905.09692 [quant-ph].
[180] K. M. Nakanishi, K. Fujii, and S. Todo, Sequential minimal optimization for quantum-classical hybrid algorithms, Phys.
Rev. Res. 2, 043158 (2020).
[181] D. Wierichs, J. Izaac, C. Wang, and C. Y.-Y. Lin, General parameter-shift rules for quantum gradients, Quantum 6, 677
(2022), arXiv:2107.12390 [quant-ph].
[182] S. Barison, F. Vicentini, and G. Carleo, An efficient quantum algorithm for the time evolution of parameterized circuits,
Quantum 5, 512 (2021).
[183] J. Nádori, G. Morse, Z. Majnay-Takács, Z. Zimborás, and P. Rakyta, The promising path of evolutionary optimization to
avoid barren plateaus, (2024), arXiv:2402.05227 [quant-ph].
[184] R. Puig, M. Drudis, S. Thanasilp, and Z. Holmes, Variational quantum simulation: a case study for understanding warm
starts, (2024), arXiv:2404.10044 [quant-ph].
[185] R. C. Farrell, I. A. Chernyshev, S. J. M. Powell, N. A. Zemlevskiy, M. Illa, and M. J. Savage, Preparations for
quantum simulations of quantum chromodynamics in 1+1 dimensions. I. Axial gauge, Phys. Rev. D 107, 054512 (2023),
arXiv:2207.01731 [quant-ph].
[186] R. C. Farrell, I. A. Chernyshev, S. J. M. Powell, N. A. Zemlevskiy, M. Illa, and M. J. Savage, Preparations for quantum
simulations of quantum chromodynamics in 1+1 dimensions. II. Single-baryon β-decay in real time, Phys. Rev. D 107,
054513 (2023), arXiv:2209.10781 [quant-ph].
[187] M. B. Hastings, An area law for one-dimensional quantum systems, J. Stat. Mech. 0708, P08024 (2007), arXiv:0705.2024
[quant-ph].
[188] J. Gambetta, Ibm quantum system two: The era of quantum utility is here: Ibm quantum computing blog (2023).
[189] Z. Zhou, R. Sitler, Y. Oda, K. Schultz, and G. Quiroz, Quantum crosstalk robust quantum control, Phys. Rev. Lett. 131,
210802 (2023).
[190] A. Vezvaee, V. Tripathi, D. Kowsari, E. Levenson-Falk, and D. A. Lidar, Virtual z gates and symmetric gate compilation
(2024), arXiv:2407.14782 [quant-ph].