UMD-PP-025-02, IQuS@UW-21-102

Quantum computation of hadron scattering in a lattice gauge theory

Zohreh Davoudi,1, 2, 3, 4, ∗ Chung-Chun Hsieh,1, 2, 3, † and Saurabh V. Kadam5, ‡

1
Department of Physics and Maryland Center for Fundamental Physics (MCFP),
University of Maryland, University of Maryland, College Park, MD 20742 USA
2
Joint Center for Quantum Information and Computer Science (QuICS),
National Institute of Standards and Technology (NIST) and
University of Maryland, College Park, MD 20742 USA
3
The NSF Institute for Robust Quantum Simulation,
University of Maryland, College Park, Maryland 20742, USA
arXiv:2505.20408v1 [quant-ph] 26 May 2025

4
National Quantum Laboratory (QLab),
University of Maryland, College Park, MD 20742 USA
5
InQubator for Quantum Simulation (IQuS) and Department of Physics,
University of Washington, Seattle, WA 98195, USA
We present a digital quantum computation of two-hadron scattering in a Z2 lattice gauge
theory in 1+1 dimensions. We prepare well-separated single-particle wave packets with
desired momentum-space wavefunctions, and simulate their collision through digitized time
evolution. Multiple hadronic wave packets can be produced using the efficient, systematically
improvable algorithm of this work, achieving high fidelity with the target initial state.
Specifically, employing a trapped-ion quantum computer (IonQ Forte), we prepare up to
three meson wave packets using 11 and 27 system qubits, and simulate collision dynamics of
two meson wave packets for the smaller system. Results for local observables are consistent
with numerical simulations at early times, but decoherence effects limit evolution into
long times. We demonstrate the critical role of high-fidelity initial states for precision
measurements of state-sensitive observables, such as S-matrix elements. Our work establishes
the potential of quantum computers in simulating hadron-scattering processes in strongly
interacting gauge theories.

∗
[email protected]
†
[email protected]
‡
[email protected]; corresponding author.
 2

CONTENTS

I. Introduction 2

II. Theoretical formalism 5

A. Z2 LGT in (1+1)D with dynamical matter 5
B. Ansatz for meson creation operators in the interacting theory 8
C. States’ fidelities with the ansatz meson creation operators 12

III. Quantum algorithm and circuit design 14

A. QGS : Preparing the interacting vacuum 15
B. QInit : Preparing the initial scattering state 17
C. QTrott : Trotterized time evolution 18

IV. Quantum hardware and emulator results 20

A. Variational-quantum-eigensolver optimization 20
B. Preparing the initial hadronic wave packets 21
C. Time-evolved observables 25
D. Return probability 28

V. Conclusion and outlook 29

Acknowledgment 31

A. Verifying the ansatz validity for a larger system using tensor networks 31

B. Circuit for QWP in the minimal-gauge formalism 34

C. Three-wave-packet preparation 35

D. Hadamard test for calculating the return probability 36

E. Time evolution of local observable for NP =13 37

F. Pauli twirling and operator decoherence renormalization 38

G. Tables of variational-quantum-eigensolver results 39

References 39

I. INTRODUCTION

A plethora of scattering experiments continue to serve as an excellent testing ground for

verifying theoretical predictions for the structure of hadrons and nuclei [1–4], searching for new
particles and interactions beyond the Standard Model [5–12], and investigating phases of matter
in and out of equilibrium [13–17]. Many of these experiments involve atomic nuclei, which are
composed of nucleons that interact via the strong force. Quantum chromodynamics (QCD) is
the quantum theory of the strong force, which is a non-Abelian gauge theory that describes
the interactions between the constituents of nucleons, quarks and gluons [18]. QCD becomes
strongly interacting at low energies, hence non-perturbative physics takes effect. Furthermore, the
theory exhibits confinement at low energies: the interacting degrees of freedom are bound states
 3

of quarks and gluons. Classical computers [19], in conjunction with the lattice-QCD method [20],
have enabled first-principles non-perturbative studies of QCD observables [21–23]. In the lattice-
QCD approach, the spacetime is discretized and time is rendered imaginary. Quantities are
evaluated in a path-integral formalism using Monte Carlo methods, where in most scenarios, the
action leads to a well-defined probability distribution [24–26]. Lattice QCD is most suitable for
calculating static observables since it uses an Euclidean-time action. Nonetheless, it can also be
employed to extract dynamical observables like scattering and decay amplitudes [27–30], thanks to
theoretical developments that utilize the finite-volume nature of lattice-QCD calculations [31, 32].
However, the application of such methods is limited to low-energy and low-inelasticity processes.
Currently, first-principles predictions for general real-time scattering processes, e.g., those relevant
for descriptions of high-energy particle collisions and the early universe, remain beyond the reach
of the state-of-the-art lattice-QCD calculations [33, 34].
An alternative to the Euclidean-time Lagrangian formulation of lattice gauge theories (LGTs)
is the Hamiltonian formulation [35]. In this formulation, one constructs the Hamiltonian, and the
associated Hilbert space, of a gauge theory on a discretized space but with a continuous time. Real-
time processes are then computed from time evolution of states. Interest in the Hamiltonian LGTs
has revived in the recent years thanks to novel computational paradigms that offer polynomial,
instead of exponential, scaling of the computational resources with respect to system’s volume.
One paradigm is the tensor-network methods that achieve polynomial scaling by systematically
controlling the entanglement generation in the state when possible, using suitable tensor-network
ansatzes [36–39]. These calculations are gaining attention for studying real-time dynamics in
LGTs [40–49], including scattering problems [50–53]. Nonetheless, utilizing this tool for simulating
high-dimensional models and non-Abelian gauge theories remains limited [54–56]. The second
novel computational paradigm uses quantum computers, leveraging the principles of quantum
superposition and entanglement [57–60], and offering exponential scaling in its (more conventional)
building blocks of qubits [61] or its rapidly emerging alternatives of qudits [62, 63], fermions [64],
or bosons [65]. LGTs subsequently can be simulated by mapping their Hilbert space to that of the
constituent degrees of freedom of a quantum computer [66–73]. One way to simulate a Hamiltonian
time evolution is by expressing the time-evolution operator in terms of a universal set of single-
qubit gates, and two-qubit entangling gates. This approach, which is known as digital quantum
simulation, is quantum-hardware agnostic, and will be the basis of this work.
In a pioneering work by Jordan, Lee, and Preskill [74, 75], an algorithm for digital quantum
simulation of scattering was proposed for a scalar field theory with quartic interactions. This
algorithm adopts the S-matrix approach in the scattering theory, where the “in” (“out”) state is
taken to be single-particle wave packets long in the past (future) and far separated in space. The
initial scattering state of well-separated single-particle wave packets is prepared from the adiabatic
evolution of a known state. The known state is taken to be two well-separated single-particle
wave packets in the non-interacting theory, and adiabatic evolution amounts to gradually turning
on the interaction strength. Once the wave packets in the interacting theory are prepared, the
state is evolved under the full Hamiltonian until they collide. Finally, measurements are performed
by simulating detectors in the late-time regions. Adiabatic state preparation has, nonetheless,
several practical and theoretical drawbacks. For example, it requires an overhead to compensate
for wave-packet broadening during the evolution [74]. Furthermore, adiabatic state preparation
becomes computationally infeasible when a phase transition occurs on the evolution path [76, 77]—
a situation that incurs, e.g., when one tries to prepare the ground state of the confined phase of a
model from that of the deconfined phase. In the context of quantum field theories, improvements
to the adiabatic state preparation have been proposed by using more efficient bases [78], by
traversing efficient paths within the adiabatic trajectory [79], or by combining variational [80–83]
and adiabatic [84, 85] methods in an scalable manner [86–88]. Recently, algorithms for wave-packet
 4

preparation and two-particle scattering in (1+1)D scalar field theory have been demonstrated on
quantum computers [89, 90], albeit limited to highly truncated fields and short-time evolutions,
due to decoherence. Finally, a first attempt at observing inelastic scattering in a (1+1)D Ising field
theory was reported in Ref. [91].
Proposals exist for preparing the initial interacting wave packets by circumventing the adiabatic
state preparation [92–96], including extraction of the scattering observables from appropriate
asymptotic matrix elements directly [97–105]. Nonetheless, these approaches are either limited to
massive excitations in the process or are challenging to apply to studies of high-energy particle
collisions, including subsequent hadronization, fragmentation, and thermalization processes.
Another method that requires no adiabatic interpolation and, instead, creates the wave packets
directly in the interacting theory is that proposed in Ref. [50] in a (1+1)D U(1) LGT using tensor-
network methods, and used in Ref. [106] to study fermion-antifermion scattering. We recently
provided an improvement to this approach in Ref. [107] by using an efficient hybrid classical-
quantum variational algorithm that creates wave packets of bound excitations in confining gauge
theories in (1+1)D. We further demonstrated the algorithm’s viability in the noisy intermediate-
scale quantum (NISQ) era by implementing it on a trapped-ion quantum computer.
In the present work, we further optimize our state-preparation method by parameterizing a
position-space rather than a momentum-space form, which proves to be more efficient. We employ
this algorithm together with a time-evolution algorithm to simulate scattering in a (1+1)D Z2 LGT
with dynamical fermions. This LGT exhibits non-perturbative features, such as confinement. It,
thus, serves as a testing ground for simulating the complex theory of QCD, and at the same time, it
may provide insights into the physics of confinement [108–118]. First, we show a resource-efficient
method for preparing the initial state of well-separated single-particle wave packets with tunable
parameters, like their spatial width and central momentum. The new approach of this work allows
for various controlled approximations to the target state, and in the worst case, relies on an ansatz
that uses only a polynomial number of parameters in the system size. Moreover, for theories with
finite correlation length, the algorithm’s cost scales polynomially with the correlation length rather
than with the system size. We demonstrate our state-preparation algorithm by preparing multiple
single-particle wave packets on an IonQ Forte quantum computer [119] for two different lattice
sizes, and compare it against the very high fidelity (and thus quantum-resource intensive) state
obtained using the noiseless emulator, observing good agreement.
Importantly in this work, we proceed with evolving the two wave packets till they collide.
Various methods have been proposed for simulating time evolution in the recent years [120–124].
Trotter product expansion of the unitary time-evolution operator, nonetheless, was found to be
the most resource-efficient algorithm for our NISQ-era application. Our goal is to investigate the
type of scattering observables accessible using near-term NISQ-era devices. We first quantum
compute on the same hardware the real-time evolution of local observables. We show that these
observables agree reasonably well with the ideal result at early times, and employing a symmetry-
based noise mitigation during post-processing has little effect. We further show, with the aid of the
noisy emulator provided by IonQ, that additional noise-mitigation circuits could be implemented
to retrieve the signal for observables that are more deviated from their idea values. Finally, we
highlight the importance of preparing high-fidelity initial states by comparing the probability of
returning to the initial state for different degrees of initial-state infidelities. This quantity is
ultimately related to the scattering S-matrix, and for unstable single-particle states, it plays a
role in calculating the decay width [105, 125, 126]. We find that such a non-local observable, as
expected, is highly sensitive to the approximations made in the initial state.
The paper is organized as follows. In Sec. II, we provide the necessary theoretical background
for constructing the single-particle wave packets and the ansatz used to create them. Section III
describes the different components of the quantum algorithm employed to prepare and evolve the
 5

wave packets. Section IV presents both the quantum hardware and emulator results, along with
the noise-mitigation methods used. Finally, we conclude, and provide an outlook, in Sec. V. Several
appendices supplement our results and analyses.

II. THEORETICAL FORMALISM

This section contains the required theoretical background for implementing the scattering
protocol of this work on a quantum hardware. First, an efficient formulation of a Z2 LGT in
(1+1)D is derived in Sec. II A, then an ansatz for building the lowest-energy particle excitation in
each momentum sector is developed and optimized in Sec. II B. Finally, numerical evidence for the
performance of the optimized ansatz, and its accuracy in regimes of arbitrary coupling and mass,
are presented in Sec. II C.

A. Z2 LGT in (1+1)D with dynamical matter

Consider a Z2 LGT coupled to a single flavor of fermions on a spatial lattice with  a periodic
boundary condition (PBC). The spatial lattice of N lattice points is denoted by Γ = 0, a, · · · , (N −
1)a , where a is the lattice spacing. Matter contents in this theory are staggered fermions [35, 127]:
fermions and antifermions live on the even and odd sites of the lattice, respectively, implying that
N must be an even number, and the staggered lattice corresponds to a physical system with
NP := N/2 lattice sites. The fermionic creation (annihilation) operator at site n is denoted by
ξn† (ξn ). The (spin- 21 ) hardcore bosons hosted on the lattice links are the gauge-boson degrees of
freedom, represented by the gauge-link operator σ̃nx for the link connecting sites n and n + a, and
its non-commuting conjugate variable, the electric-field operator, σ̃nz .
The local Hilbert space at site n is spanned by |fn ⟩ with fn = 0, 1, the staggered-fermion
occupation. In the Dirac-sea picture, fn = 0 (1) at even lattice sites corresponds to the absence
(presence) of a particle, while at odd lattice sites it corresponds to the presence (absence) of an
antiparticle. Note that, ξn |0n ⟩ = ξn† |1n ⟩ = 0, ξn |1n ⟩ = |0n ⟩, and ξn† |0n ⟩ = |1n ⟩ , ∀n ∈ Γ. The
local Hilbert space at the link originating from site n in the electric-field basis is spanned by |sn ⟩
with sn =↑n , ↓n , the two spin projections of a hardcore boson along the z-axis at site n. The
gauge-link and electric-field operators in this basis are thus given by the Pauli x and z operators
σ̃nx = |↑n ⟩ ⟨↓n | + |↓n ⟩ ⟨↑n | and σ̃nz = |↑n ⟩ ⟨↑n | − |↓n ⟩ ⟨↓n |, respectively.
A generic basis state in this Hilbert space is given by

|ψ⟩ = |f0 , fa , · · · , f(N −1)a ⟩ ⊗ |s0 , sa , · · · , s(N −1)a ⟩ . (1)

Only a portion of this Hilbert space is physically relevant because the physical states |ψ⟩phys must
satisfy the local Gauss’s laws

Gn |ψ⟩phys = |ψ⟩phys , ∀n, (2)

where

Gn = σ̃nz σ̃n−1
z
eiπQ̄n . (3)

Here,

1 − (−1)n/a
Q̄n = ξn† ξn − . (4)
2
 6

such that Q̄n = 1 (−1) when a fermion (antifermion) is present at n, and Q̄n = 0 otherwise.
The system’s Hamiltonian H is given by
1 X † x  X X
aH = ξn σ̃n ξn+a + H.c. + amf (−1)n/a ξn† ξn + aϵ σ̃nz , (5)
2
n∈Γ n∈Γ n∈Γ

where mf ≥ 0 is the fermion mass and ϵ is the strength of the electric-field Hamiltonian. Here,
ξN a is identified with ξ0 due to the PBC. From here onward, we set a = 1, hence quantities are
expressed in units of lattice spacing. The PBC yields lattice translational invariance and allows
for specifying well-defined momentum quantum numbers.
The Gauss’s law operator in Eq. (3) commutes with the Hamiltonian in Eq. (5). The
Hamiltonian also commutes with a global operator Q given by the total fermionic occupation
in the lattice:
N
X −1
Q= ξn† ξn . (6)
n=0

Following our previous work [107], we restrict this study to the subspace of the Hilbert space
spanned by states with Q = NP . A special state in this subspace, called the strong-coupling
vacuum (SCV), is the ground state of the Hamiltonian in the limit of mf , ϵ ≫ 1, and is given by

|Ω⟩SCV = |0, 1, · · · , 0, 1⟩ ⊗ |s, s, · · · , s, s⟩ . (7)

Here, s = ↑ (↓) for ϵ < 0 (ϵ > 0).

The gauge degrees of freedom in (1+1)D LGTs are not dynamical in nature, and are often
rotated away when open boundary condition (OBC) are in place. In this case, only the fermionic-
matter degrees of freedom survive but at the expense of yielding a non-local Hamiltonian. Such
an elimination of the gauge bosons is achieved by transforming the fermionic fields as
n−1
!
Y
x
ξn → ψn = σ̃i ξn , (8a)
i=0
n−1
!
Y
ξn† → ψn† = σ̃ix ξn† . (8b)
i=0

With the PBC, the gauge degrees of freedom cannot be completely rotated away. However, they
are reduced to a single spin degree of freedom on one link, which can be seen by examining the
Hamiltonian in Eq. (5) under the field re-definitions in Eq. (8):
N −2
1 X †  1
†
 X X
H= ψn ψn+1 + H.c. + ψN −1
e x ψ0 + H.c. + mf
Σ (−1)n ψn† ψn + ϵ e zn .
Σ (9)
2 2
n=0 n∈Γ n∈Γ

Here,
N
Y −1
e x :=
Σ σ̃nx , (10)
i=0

e zn is evaluated
and the second term in Eq. (9) resulted from the PBCs. The electric-field operator Σ
using the Gauss’s law in Eq. (2):
Pn
e zn := σ̃nz = eiπ
Σ m=0 Q̄m z
σ̃N −1 , (11)
 7

where Q̄n is obtained by substituting Eq. (8) in Eq. (4):

1 − (−1)n
Q̄n = ψn† ψn − . (12)
2
The factor nm=0 Q̄m can be either an even or an odd integer, resulting in Σ e z = σ̃ z ez
P
n N −1 or Σn =
z
P N −1 † P N −1 † P N −1
−σ̃N −1 , respectively. (The restriction n=0 ξn ξn = n=0 ψn ψn = NP implies n=0 Q̄n = 0,
z z z
yielding ΣN −1 = σ̃N −1 .) As a result, the conjugate-variable field to the electric field ±σ̃N −1 at the
e
x x
last link is ±σ̃N −1 . The action of Σ on |ψ⟩ defined in Eq. (1) is, therefore, equivalent to the action
e
of ±σ̃Nx
−1 on |sN −1 ⟩. (Either of these options yields the same spectrum so we choose a positive
sign.) In other words, we use the convention that the action of the transformed gauge-link operator
e x is now restricted to the link connecting lattice sites N − 1 and 0.1
Σ
One can then reduce the bosonic Hilbert space to the local Hilbert space at the last link, and
a generic state in this Hilbert space is now given by

|ψ⟩phys = |f0 , f1 , · · · , fN −1 ⟩ ⊗ |sN −1 ⟩ . (13)

The ψn and ψn† field operators act on the fermionic Hilbert space in the same manner as ξn and ξn†
operators act on the state in Eq. (1). The action of the gauge-link and the electric-field operator
at site n is restricted to the reduced bosonic Hilbert space. The electric Hamiltonian, nonetheless,
has become non-local, where the non-locality originates from the dependence of Σ e zn on the fermion-
number occupations up to site n. Non-locality is the cost to pay to render all states physical in this
Hilbert space (there is no local Gauss’s law constraint anymore in the new formulation). When
subjected to the restricted subspace that satisfies Q = NP , the dimension of the Hilbert space is
2 × (2NP )!/(NP !)2 , and the SCV state is given by

|Ω⟩SCV = |0, 1, · · · , 0, 1⟩ ⊗ |sN −1 ⟩ , (14)

with sN −1 =↑ (↓) for ϵ < 0 (ϵ > 0).

We refer to the Hamiltonian in Eq. (5) and the Hilbert space spanned by states in Eq. (1) as the
explicit gauge-link formulation (EGF), and to the Hamiltonian in Eq. (9) and its corresponding
Hilbert space spanned by states in Eq. (13) as the minimal gauge-link formulation (MGF) of a Z2
LGT in (1+1)D with PBCs.2 Quantum simulation of MGF requires fewer qubits than the EGF,
as can be seen from mapping of fermionic and bosonic degrees of freedom to qubits. The fermions
in this work are mapped to qubits via the Jordan-Wigner transformation
n−1
!
Y
†
ψn = σm σn− ,
z
(15a)
m=0
n−1
!
Y
z
ψn = σm σn+ , (15b)
m=0

to satisfy the anti-commuting nature of fermionic field operators. Here, σiz is the Pauli-z operator
and σi+ (σi− ) is the Pauli-raising (lowering) operator acting on the ith qubit. Hence, in both
formulation, N qubits are needed to represent the fermionic Hilbert space. The hardcore bosons
are mapped to qubits directly with the s state denoting the corresponding qubit eigenstate in the
z-basis. The qubit-resource savings is evident: to express the bosonic Hilbert space, the MGF needs

1
While the Hamiltonian in Eq. (9) appears to distinguish the last link from the rest of the links, lattice translational
symmetry of the original formulation is still preserved in this merely gauge-transformed formulation.
2
Such alternative formulations have been studied in the past for different LGTs with PBCs, e.g., in Ref. [128–130].
 8

only one qubit while the EGF requires N qubits. In this work, we use the MGF for simulating
scattering in a Z2 LGT, which requires N + 1 qubits for mapping the Hilbert space of N (NP )
staggered (physical) lattice sites.
The qubit Hamiltonian used for the unitary time evolution is obtained by substituting Eq. (15)
in Eq. (9), resulting in

H = H h + H m + H ϵ, (16)

with
N −2
h 1X x x y
αN x y y

H = σn σn+1 + σny σn+1 + x
σN −1 σ̃N x x
−1 σ0 + σN −1 σ̃N −1 σ0 , (17a)
4 4
n=0

mf X
Hm = (−1)n+1 σnz , (17b)
2
n∈Γ
 
N
X −2 n
Y
H ϵ = ϵ σ̃N
z
−1 + ϵ
z
γn σ̃N −1
 σjz  . (17c)
n=0 j=0

Here, H h obtains the interaction (or the hopping) energy, H m obtains the fermion-mass energy,
and H ϵ obtains the electric-filed energy. The factor αN := (−1)NP +1 in Eq. (17a) arises from pulling
to the right a string of σ z Paulis that span the entire lattice and evaluating
Pn it for the subspace
−iπ (1−(−1) m )/2
considered here, i.e., Q = NP . The factor γn in Eq. (17c) is γn := e m=0
PN −1
= in for
even n and in+1 for odd n. The first term in the same equation results from eiπ m=0 Q̄m = 1 in
the Q = NP sector. Note that, the operators acting on the gauge-boson qubit are distinguished
from the fermion qubits with an overhead tilde notation.
In the next subsection, we discuss how to build the single-particle wave-packet state that
comprises the initial scattering state.

B. Ansatz for meson creation operators in the interacting theory

To simulate two-meson scattering in a (1+1)D Z2 LGT on a quantum computer, one first needs
to prepare an initial state made up of two single-particle wave packets, ideally separated far away
from each other. The form of the single-particle wave packets |Ψ⟩ follows our earlier work in
Ref. [107]. The state |Ψ⟩ is defined as a collection of the lowest-energy single-particle states, |k⟩,
in each momentum sector k, smeared by the wave-packet profile Ψ(k):
X
|Ψ⟩ = Ψ(k) |k⟩, (18)
k∈Γ
e

e := π − NP , −NP + 1, · · · , NP − 1 ∩ [− π , π ) corresponds to the physical momenta on


where Γ NP 2 2
the lattice. In Ref. [107], each |k⟩ was obtained by optimizing an ansatz for the creation operator
b†k . Explicitly, b†k excites |k⟩ when acted on the interacting vacuum |Ω⟩:

|k⟩ = b†k |Ω⟩ , (19)

with ⟨Ω|Ω⟩ = ⟨k|k⟩ = 1. This optimization can be done using a variational quantum eigensolver
(VQE). The ansatz in Ref. [107] was motivated by a similar ansatz proposed earlier in Ref. [50] in
the context of tensor-network calculations of scattering in a U(1) gauge theory.
 9

Case EGF MGF M

fm,n

† † 1 z
m=n ξm ξm ψm ψm 2 (1 − σm )

n−1
! n−1
!
Y Y
|m − n| < NP †
ξm σ̃lx ξn †
ψm ψn −
σm σlz σn+
m<n l=m l=m+1
−1 −1
N m−1
! m−1
! N
!
Y Y Y Y
x x † †
e x ψm NP − +
|m − n| > NP ξn σ̃l σ̃l ξm ψn Σ (−1) σlz σm σn σlz x
σ̃N −1
l=n l=0 l=0 l=n+1

m−1
! m−1
!
Y Y
|m − n| < NP ξn σ̃lx †
ξm †
ψn ψm (−1) σn+ σlz −
σm
m>n l=n l=n+1

−1 −1
N n−1
! n−1
! N
!
Y Y Y Y
|m − n| > NP †
ξm σ̃lx σ̃lx ξn †
ψm e x ψn (−1)NP +1
Σ σlz −
σn+ σm σlz x
σ̃N −1
l=m l=0 l=0 l=m+1

TABLE I. Definition of Mm,n for different cases of m and n values. The second and third column summarize
the form of Mm,n in the EGF and MGF, respectively. The last column denotes its form in the MGF after
performing the Jordan-Wigner transformation in Eq. (15), denoted here by M fm,n . This is later used for
†
mapping the bk operator to operations on qubits in a quantum circuit. The forward-(backward-)wrapped
meson operators are identified when the creation operator at m appears to the left (right) of the annihilation
operator at n in the EGF and MGF columns. The |m − n| = NP case is not explicitly shown in the table
for brevity: it can be easily
√ obtained by adding the forward- and backward-wrapped mesons of length NP ,
each with a coefficient 1/ 2.

In this work, we propose a closely-related alternative ansatz whose form is motivated by the
translation symmetry of the physical lattice. It uses kinematic factors similar to the ansatz in
Refs. [50, 107], which are obtained by solving the free fermionic theory on a staggered lattice. The
translation symmetry of the staggered lattice, where a unit translation on the original (physical)
lattice turns into a translation by two units on the staggered sites, is incorporated in the ansatz
by considering the gauge-invariant operators Mm,n . The operator Mm,n for the simple case of
†
m = n reduces to the fermion number operator ξm ξm . However, when m ̸= n, Mm,n is given by a
non-local operator called a bare-meson creation operator. In the EGF, this operator is composed
†
of ξm , ξn , and a string of σ̃lx operators connecting them. For m < n and the periodic lattice, there
† Qn−1 x Q −1 x Qm−1 x †
can be two ways of constructing such an operator: ξm ( l=m σ̃l ) ξn or ξn ( N
l=n σ̃l )( l=0 σ̃l ) ξm .
The former is referred to as a forward-wrapped (n − m)-length meson creation operator and the
latter a backward-wrapped (N − n + m)-length meson creation operator. When m > n, the
forward-wrapped (backward-wrapped) meson creation operators are given by similar expressions
with the fermionic operators ordered accordingly, such that the forward-wrapped meson is given
† QN −1 x Qn−1 x Qm−1 †
by ξm ( l=m σ̃l )( l=0 σ̃l ) ξn and the backward-wrapped meson is given by ξn ( l=n σ̃lx ) ξm . We
require that the ansatz only include the shorter meson depending on m and n, and if the meson
length is equal for both forward and backward wrapping, i.e., |m − n| = NP , each operator is added
with a coefficient √12 such that each bare meson is created with a probability of 12 .
The Mm,n operators translate by two staggered lattice sites under one unit translation of
the physical lattice, as mentioned above. Thus, for a given bare meson length l, the operators
Mm,n that start at even (odd) m are mapped to each other via the physical translations. Upon
performing the gauge transformation in Eq. (8), the bare-meson creation operator reduces to a
 10

fermion hopping term between m and n if the meson does not contain the remnant gauge link,
otherwise, the fermion hopping term includes an additional Σ e x operator defined in Eq. (10) (whose
x
action reduces to σ̃N −1 on the bosonic Hilbert space in the MGF). A summary of Mm,n definitions
in both EGF and MGF is given in Table I, along with their mappings to the qubit space in the
MGF after the Jordan-Wigner transformations in Eq. (15).
In this confined theory, it is expected that the operators which have a support over an extent
much greater than the confinement scale will be suppressed exponentially. Motivated by this
intuition, we introduce an ansatz for b†k that can be improved order by order, and takes the
following form:
N −1
1 X (j)
b†k = ηk . (20)
N
j=0

Here, N is a normalization factor that will be discussed later. The superscript j denotes the
j th order at which only the bare-meson creation operators with length less than or equal to j
(j)
contribute. The expression for ηk is given by
(j),k (j),k
(j) |m−n|2 k |m−n|2 k
X X
ηk = e−α0 C m,n Mm,n + e−α1 C m,n Mm,n , (21)
(j) (j)
m,n∈Γ0 m,n∈Γ1

(j)
where Γi = {m, n ∈ Γ || |m − n| = j and m mod 2 = i}. Here, |m − n| is taken to be the shortest
k
distance between m and n on a periodic lattice. C m,n are kinematical factors given by
k X
C m,n = δk,p+q C (p, m)D(q, n), (22)
p,q∈Γ
e

with the momentum sums in p and q running over the Brillouin zone of the staggered lattice, Γ,
e
and
s
mf + ωp ipm
C (p, m) = e (Pm,0 + vp Pm,1 ) , (23a)
2πωp
s
mf + ωq iqn
D(q, n) = e (−vq Pn,0 + Pn,1 ) , (23b)
2πωq

which are obtained by solving

q for the plane-wave solutions in the free fermion theory onn+0(1)
a staggered
2 2 sin(k) 1+(−1)
lattice [50]. Here, ωk = mf + sin (k) and vk = mf +ωk . Moreover, Pn,0(1) = 2 is the
projection operator to
P the even (odd) staggered sites. One can rewrite Eq. (21) using Eqs. (22)
and (23) in terms of m,n∈Γ C (p, m)D(q, n)Mm,n . The ansatz is, therefore, built in such a way to
impart momentum p and q to the m and n end of the bare-meson creation operator Mm,n . The
Kronecker delta δk,p+q then ensures that the total momentum of the composite object is the desired
(j),k
−α |m−n|2
momentum k. Finally, the exponential factors e 0/1 control the strength of contribution of
each bare-meson creation operator according to their lengths, motivated by the discussions above.
At any given order j, the operator b†k can be approximated by bk , which using Eqs. (20)-(23)
(j)†

has the form:

j
(j)† ′
X X
(j ),k
bk = Cm,n Mm,n , (24)
j ′ =0 m,n∈Γ(j ′ )
 11

(j) (j)
where Γ(j) = Γ0 ∪ Γ1 and

(j),k 1 −α(j),k |m−n|2 k (j)

Cm,n = e i C m,n for m, n ∈ Γi . (25)
N
P (j),k 2
We choose the normalization constant N at each order j such that m,n∈Γ(j) |Cm,n | = 1.
(j),k (j),k
For a given k, there are two parameters to be optimized at each order j, α0
Given and α1 .
the natural hierarchy of operators based on their lengths, the order-by-order optimization scheme
(j),k
spares one from optimizing all the α0/1 parameters simultaneously. One can start the optimization
process with j = 1 that restricts the ansatz to length-1 bare-meson creation operators.3 A
(1),k (1),k ∗
variational minimization algorithm can then be run for parameters α0/1 to obtain α0/1 that
(1)†
minimizes the energy of the state bk |Ω⟩. To improve the overlaps with the true lowest-energy
state with a momentum k, one can then choose to go to the next order in j and repeat the variational
(1),k (2),k
minimization process on α0/1 and α0/1 . Nonetheless, the parameter space can be restricted to
(1),k ∗
a narrow region around the optimized α0/1 from the previous order of optimization to ease the
next optimization. The process is iterated for progressively larger j values, while restricting the
(ℓ),k (ℓ),k ∗
parameter space for α0/1 with ℓ < j to a narrow range near the already optimized values α0/1 .
One can then expect to obtain the form of b†k that is very close to the true momentum creation
operator of the interacting theory, which excites the |k⟩ state when acted on the interacting vacuum
|Ω⟩.
Once the ansatzes for operators b†k are optimized to reach a target accuracy, the list of optimized
(ℓ),k ∗
α0/1 can be used in Eqs. (18) and (19) to define an operator b†Ψ . This operator creates the desired
wave-packet state |Ψ⟩ from the interacting vacuum:
 

Ψ(k)b†k  |Ω⟩ ≡ b†Ψ |Ω⟩ .

X
|Ψ⟩ =  (26)
k∈Γ
e

Furthermore, using the ansatzes for b†k from Eq. (20) and the Jordan-Wigner transformed definitions
of Mm,n , denoted by Mfm,n in Table I, the operator b† can be approximated by the operator b(j)†
Ψ Ψ
(j)†
built of bk operators:

j
(j)† (j ′ ) f
X X
bΨ = Cm,n Mm,n , (27)
j ′ =0 m,n∈Γ(j ′ )

where
X
(j) (j),k
Cm,n = Ψ(k) Cm,n . (28)
k∈Γ̃

(j) (ℓ),k ∗
Thus, the coefficients Cm,n depend on the optimized parameters α0/1 , the kinematic factors C
(j)†
and D defined in Eqs. (23), and the wave-packet profile Ψ(k). This form of bΨ ≈ b†Ψ will be used
in Sec. IV for preparing the initial scattering state.

3 (j=0),k
Note that, α0/1 do not enter the optimization process since the ansatz does not depend on them when m = n.
 12

FIG. 1. The j th -order ansatz infidelity 1 − F (in logarithmic scale) as a function of Hamiltonian parameters
ϵ and mf for N = 10. Each column corresponds to a non-negative momentum k in the Brillouin zone,
(j)
and each row corresponds to the order of the ansatz j. Higher fidelity is observed for |kop ⟩ when larger
values of mf , |ϵ|, and a higher ansatz order j are used. For example, F = 0.99 is achieved at j = 3 in the
entire parameter space for all k. The region to the left of the cyan contour corresponds to the existence of
a low-energy non-mesonic excitation that is not captured by our mesonic ansatz.

C. States’ fidelities with the ansatz meson creation operators

In this section, we demonstrate the accuracy with which the optimized ansatz represents the
desired states at a given order. In small systems, the accuracy can be checked by calculating the
fidelity between the momentum eigenstates obtained by exact diagonalization and the optimized
ansatz. We define the fidelity of a hadron state in the momentum sector k as
2
(j)
F := op ⟨k |k⟩ex . (29)

(j)†
Here, |k (j) ⟩op is prepared by acting on |Ω⟩ex with the optimized ansatz for bk , where |Ω⟩ex is the
ground state obtained by numerically diagonalizing the Hamiltonian. Similarly, |k⟩ex is the lowest-
energy mesonic eigenstate with momentum k obtained from the exactly diagonalized Hamiltonian.
 13

Figure 1 shows the fidelity scan in a 10-(staggered) site theory over a range of Hamiltonian
parameters, mf and ϵ, for different k-momentum sectors, and at various ansatz orders. The
order-by-order optimization is conducted in the following manner: at the j th order, a total of 2j
(j ′ ),k
parameters α0/1 with j ′ ≤ j are optimized. The initial values of the first 2j − 2 parameters
(j ′ ),k
α0/1 with j ′ < j have been set by the optimized results at the previous (j − 1)th order, and their
(j ′ ),k
variation are limited to a window w = 0.1 × max(|α0/1 |, 1) to aid the optimization at the j th
order.
As mentioned earlier, the j th order ansatz consists of bare mesons with length up to j in
position space. Therefore, it is reasonable to expect that a higher-order ansatz can better capture
the interacting momentum eigenstate |k⟩. Also, as the order j is increased, the ansatz is expected
to improve in capturing longer-range correlations. These expectations are in agreement with our
numerical results. The fidelity can reach 0.98 (0.99) for all the parameter range considered if the
first-(third-)order ansatz is used, and the optimization at all orders works better in the strong-
coupling regime where the mass gap is larger. We also note that the infidelities in |k (j) ⟩op are
associated with contamination from the excited states in the same momentum sector k, since our
ansatz transforms properly under translation by construction.
In the k = 0 sector, there is one non-mesonic configuration that can be constructed by flipping
the single bosonic link, which cannot be captured by the ansatz. Such an excitation corresponds to
one unit of electric flux across the periodic lattice in the EGF, and has an energy proportional to N .
It, thus, does not generally appear in the low-energy spectrum. Nonetheless, for sufficiently small ϵ
at a given mf , i.e., the area to the left of the cyan contour in Fig. 1, such a non-mesonic excitation
can have energies lower than the single-particle momentum states. In this region, |k = 0⟩ex is
defined as the first mesonic excitation in the k = 0 sector, and the fidelity is calculated accordingly.
Compared to the results in our recent work [107], which adopts a momentum-space ansatz
containing only two parameters overall, the order-by-order translationally invariant position-space
ansatz developed here performs better, even at the edge of the Brillouin zone (i.e., for large
momenta). The ability to create accurate momentum eigenstates with large k yields faster-moving
wave packets. This, first of all, injects more energy into the process and leads more interesting
outgoing scattering channels. Second, it reduces the total evolution time needed to bring the wave
packets together in a scattering simulation; therefore, fewer Trotter steps and shallower circuits
are needed for a Trotterized time evolution.
(j),k
Finally, we note that the optimized values of α0/1 are not unique with different initializations
of the optimization routine. However, we observe that the normalization condition for N renders
(j) (j)
the resulting Cm,n coefficients out of these initializations approximately identical. (The Cm,n
coefficients determine the gate parameters in the quantum circuit.) It is also seen that the optimized
(j),k
α0/1 can be different between k and −k momentum sectors, as opposed to being symmetric like
the ansatz in Ref. [107].4 Only the positive-k fidelities are plotted in Fig. 1 for brevity. (Similar
fidelities are obtained for the negative-k states).
The order-by-order ansatz is expected to work for larger systems, as demonstrated in Appendix A
for a 26-(staggered) site theory. For lattice sizes that exceed the regime of viability of exact
diagonalization, we compare the ansatz against matrix-product states (MPS), which can accurately
approximate the exact eigenstates in (1+1)D, upon employing a density-matrix-renormalization-
group (DMRG) method. A parameter scan similar to Fig. 1, but in the 26-(staggered) site theory,
is presented in Fig. 11, demonstrating high fidelities achieved even with the lowest-order ansatzes.
4
We have tried a symmetric ansatz with respect to k → −k, which is identical to ansatz in Eq. (21) except one
needs to include a factor of i|m−n| in both sums. We have checked that, despite restoring the symmetry, this ansatz
overall returns lower fidelities than the anstaz based on Eq. (21). We, therefore, do not consider such an ansatz
any further.
 14

FIG. 2. Shown is the circuit used in this work for simulating scattering in a (1+1)D Z2 LGT in the MGF.
The circuit acts on N + 1 system qubits representing the lattice associated with N staggered sites, and
one or more ancilla qubits denoted by a1,2,t . The system qubits are initialized in the SCV state |Ω⟩SCV , as
shown by the first vertical dotted line, while each ancilla starts in the |0⟩ state. The circuit consists of three
subcircuit modules. The vertical dotted lines indicate the state of the system qubits after the application
of each module. The first module prepares the ground state, |Ω⟩, of the Hamiltonian in Eq. (16) using the
circuit block QGS with parameters θh∗ and θm∗ . The second module prepares the initial scattering state
composed of two well-separated input wave packets, resulting in the state |Ψ1 , Ψ2 ⟩. It requires at least one
ancilla qubit to prepare the initial state, shown here with a black solid line. Alternatively, it can also be
applied using an extra ancilla qubits, shown with a dotted black line, to improve the accuracy of preparing
the target state, as explained in the text. Finally, the last module, QTrott , performs the unitary time
evolution U (t) = e−itH under the Hamiltonian H in Eq. (16). The circuit components in red are to perform
a Hadamard test to compute the return probability of the initial state as a function of time; they can be
omitted when measuring only the expectation values of diagonal operators. The Hadamard test requires an
additional ancilla and a controlled application of QTrott with control on the ancilla. Here, H denotes the
i x
Hadamard gate and Rx (θ) := e− 2 θσ . To compute the return probability, as shown in Appendix D, one
needs to separately implement either the Rx ( π4 ) or the H gate as the last operation on the at ancilla, which
is denoted in the circuit by Rx ( π4 )/H. Details of each module and their constituent circuits are discussed in
Sec. III.

For the remainder of the main text, to reduce the clutter, we drop the superscript (j) from
quantities, since the order at which they are assumed will be clear from the context.

III. QUANTUM ALGORITHM AND CIRCUIT DESIGN

In this section, we use the Hamiltonian and the ansatz for hadron states defined in Sec. II to
lay out a digital quantum algorithm for performing multi-hadron scattering. We choose to work
with the MGF for a more efficient circuit implementation.
The overall protocol used in this work starts from the SCV state |Ω⟩SCV in Eq. (14), which is
one of the computational basis states. The protocol is composed of three circuit modules:

1. QGS prepares the interacting vacuum |Ω⟩ from |Ω⟩SCV .

2. QInit constructs the initial scattering state comprised of well-separated wave packets.

3. QTrott performs the Trotterized time evolution under the Hamiltonian in Eq. (16).

These three modules are depicted in Fig. 2, with each module separated by a dotted line, indicating
the corresponding quantum state prepared at the end of each stage. Each circuit module will be
 15

FIG. 3. Shown is the circuit QGS that prepares the interacting ground state |Ω⟩ of the Hamiltonian in
Eq. (16). This circuit is parameterized by two parameters, θh and θm . The circuit QGS acts on the strong-
coupling vacuum |Ω⟩SCV , which is given by alternating |0⟩ and |1⟩ states on qubits that represent fermion
lattice sites, labeled here with a subscript fi for the ith site, and |0⟩ state for the qubit that represents
the bosonic link in the MGF, denoted here by the subscript bN −1 . The two-qubit gate are defined as
− i θ σx σx − i θ σy σy
Rxx (θ) := e 2 fi1 fi2 and Ryy (θ) := e 2 fi1 fi2 , where i1 and i2 denote fermion lattice sites each gate
involves. The three-qubit gates Rxx̃x (θ) and Ryx̃y (θ) are defined in the similar manner with the overhead
tilde indicating the qubit operation on the bosonic link qubit bN −1 , and αN is defined below Eq. (17).
Finally, the parameters θh∗ and θm∗ are obtained using a VQE method that minimize the energy of the
state prepared by the circuit with respect to the Hamiltonian in Eq. (16).

discussed in the following. Computing observables may require its own circuit module, which we
will partly discuss in Sec. IV D and Appendix D.
Before proceeding, we emphasize that our procedure constitutes a hybrid classical-quantum
algorithm, where both the ground state in module 1 and the k-momentum eigenstates required in
module 2 are prepared using the VQE method. In this work, the VQE results for the ground
state and k-momentum eigenstates are verified only through classical evaluation, in order to
save quantum-computing resources. However, in theory, it is possible to implement both VQE
calculations on a quantum computer provided that the expectation value of the Hamiltonian
operator can be determined with sufficient accuracy from the hardware results (which generally
require a large number of measurements).

A. QGS : Preparing the interacting vacuum

The circuit for QGS is similar to the one given in Ref. [107] (which was inspired by a VQE
ground-state circuit in the case of a (2+1)D Z2 LGT in Ref. [131]). The strategy is to prepare the
ground state via iterative evolution of an initial state using the terms in the Hamiltonian. Thus,
QGS is parameterized by θxh , θxm , and θxϵ , which are related to H h , H m , and H ϵ in Eqs. (17a)- (17c),
respectively. The form of QGS is given by

N
! ! !
GS
− 2i θxh Hn,n+1
h
− 2i θxm Hn
m − 2i θxϵ Hn
ϵ
Y Y Y Y
QGS = e e e , (30)
x=1 n∈Γ n∈Γ n∈Γ
 16

FIG. 4. Two different ways of creating the two wave-packet initial scattering state are shown here in (a)
and (b). In the former case, the QWP circuit that prepares a single-particle wave packet is acted twice using
the same ancilla a1 with different wave-packet profiles, Ψ1 and Ψ2 , as input. In the latter case, the two
applications of QWP circuit use two different ancilla a1 and a2 . The non-participating ancilla in a QWP is
indicated with a line crossing over the corresponding circuit block in (b). The method in (a) yields a larger
systematic error compared to (b), leading to a lower-fidelity wave-packet state, as discussed in the text.
Detailed decomposition of each QWP (Ψ) is presented in Appendix B.

where NGS denotes the number of iterations and

+ σny σn+1
y

σnx σn+1
x if n ̸= N − 1,
h
Hn,n+1 = (31a)
x x x y x y

αN σN −1 σ̃N −1 σ0 + σN −1 σ̃N −1 σ0 if n = N − 1,


Hnm = (−1)n+1 σnz , (31b)

 Q 
z n z
γn σ̃N

−1 σ
j=0 j if n ̸= N − 1,
Hnϵ = (31c)
σ̃ z

if n = N − 1.
N −1

It is found that it suffices to set NGS = 1 and θ1ϵ = 0 to prepare high fidelity |Ω⟩ for the
parameters mf and ϵ, and the lattice sizes considered later in Sec. IV. The circuit for QGS is
shown in Fig. 3. It is parameterized by θh and θm , dropping subscript x = 1 for brevity. In
i h h
QGS (θh , θm ), we use the first-order Trotter product formula for each term in e− 2 θ Hn,n+1 . Explicitly,
i h h i h h
we first implement e− 2 θ H2k,2k+1 for k = 0, · · · , N/2 − 1 before implementing e− 2 θ H2k+1,2k+2 for
k = 0, · · · , N/2 − 1 (with the identification HNh h
−1,N ≡ HN −1,0 ). This scheme preserves the global-
Q symmetry. Each Trotter layer requires 8NP + 4 CNOT gates. The parameters θh∗ and θm∗
that prepare the ground state are obtained using a VQE algorithm. The algorithm minimizes
the energy of the state prepared by QGS (θh , θm ) with respect to the Hamiltonian in Eq. (16).
Computing ⟨H h ⟩ from Eq. (17) requires independent measurements in the σ x and σ y basis for
qubits encoding fermionic fields, while the qubit encoding the boson field needs to be measured
only in the σ̃ x basis. The values of ⟨H m ⟩ and ⟨H ϵ ⟩ in the state prepared by QGS (θh , θm ) can be
measured simply in the computational basis of the qubits.
 17

B. QInit : Preparing the initial scattering state

The second module, QInit (Ψ1 , Ψ2 ), prepares the initial scattering state of two wave packets,
|Ψ1 , Ψ2 ⟩. Each wave packet is defined in Eq. (18) in terms of non-unitary operators b†k s. The
wave packets thus need to be created by implementing the b†k operators in an extended Hilbert
space using one or two ancilla qubits, as shown in Fig. 4(a) and (b), respectively. Here, QWP (Ψi )
creates a single wave packet state |Ψi ⟩ using an ancilla qubit that is initiated in |0⟩. The method in
Fig. 4(a) uses only one ancilla qubit for both QWP (Ψ1 ) and QWP (Ψ2 ), with a σ x operator acting on
the ancilla in between the two circuits. In Fig. 4(b), on the other hand, each QWP uses a different
ancilla qubit. Compared to the latter, the former method uses fewer quantum resources but at a
cost of lower accuracy of achieving the target state, as will be discussed below.
The circuit for QWP is based on that in Ref. [75] for the momentum creation operators b†k in
a non-interacting scalar field theory, in which case hb†k ’s exact
i expressions are known. The method
relies on two properties of b†k : the operator obeys bk , b†k′ = δk,k′ , and bk |Ω⟩ = 0. It is expected
that the b†k operators in the interacting theory, obtained via the ansatz in Eq. (20), obey these
relations only approximately. Then from Eq. (26), bΨ |Ω⟩ ≈ 0 and [bΨ , b†Ψ ] ≈ 1 for a wavefunction
profile Ψ(k) that is normalized as (Ψ|Ψ) = 1, where
X
(Ψ2 |Ψ1 ) := Ψ∗2 (k)Ψ1 (k). (32)
k∈Γ
e

With this, one can introduce an ancilla qubit, a, and define the following Hermitian operator:

ΘΨ := b†Ψ ⊗ |1a ⟩⟨0a | + bΨ ⊗ |0a ⟩⟨1a | (33)

such that

π
e−i 2 ΘΨ |Ω⟩ ⊗ |0a ⟩ = −i |Ψ⟩ ⊗ |1a ⟩ . (34)

The circuit block QWP (Ψ1 ) in Fig. 4 implements Eq. (34) for the wave-packet profile Ψ1 (k).
The circuit acts upon qubit a1 initialized in |0a1 ⟩, and on the system qubits holding the interacting
vacuum |Ω⟩. Recall that |Ω⟩ is prepared by QGS (θh∗ , θm∗ ) in the previous step. Furthermore, The
π
unitary operator e−i 2 ΘΨ is implemented via a Trotter expansion, where the terms constituting
ΘΨ are exponentiated separately a number of times with the angle π/2 divided into a number
of Trotter steps ñt . Design and implementation of QWP is presented in our earlier work [107],
although within the EGF of the Z2 LGT. The methods used there can still be directly translated
to the MGF, as summarized in Appendix B. We have restricted our discussion to the case of a j = 1
order ansatz, and using a second-order product formula with one Trotter step only. In general,
preparing a single wave packet at order j (once the ansatz optimization is done classically or via
VQE) requires [4(j 2 + 9j + 1)NP + 2j 2 + 2j] × 2ñt CNOT gates for an N -site theory.
After preparing the first wave packet |Ψ1 ⟩ in the initial scattering state, the remaining part of
the second module is achieved by applying the second QWP either using the same ancilla, following
the action of σax1 , as shown in Fig. 4(a), or by using a second ancilla, a2 , initiated in |0a2 ⟩, as shown
in Fig. 4(b). In both cases, the state becomes −i |Ψ1 ⟩ ⊗ |0a′ ⟩ with |0a′ ⟩ being the corresponding
ancilla qubit. The second wave packet Ψ2 can then be prepared by applying QW P (Ψ2 ) on the state
|Ψ1 ⟩ ⊗ |0a′ ⟩, provided that the overlap between two wave packets is negligible, i.e., (Ψ2 |Ψ1 ) ≈ 0, a
 18

condition that holds for far separated wave packets. This statement can be proved by noting that

e−iθΘΨ2 |Ψ1 ⟩ ⊗ |0a′ ⟩ ≈ cos(θ) |Ψ1 ⟩ ⊗ |0a′ ⟩ − i sin(θ) |Ψ1 , Ψ2 ⟩ ⊗ |1a′ ⟩

h √ i
+ (Ψ2 |Ψ1 ) cos( 2θ) − cos(θ) |Ψ2 ⟩ ⊗ |0a′ ⟩
√
 
1
− i(Ψ2 |Ψ1 ) √ sin( 2θ) − sin(θ) |Ψ2 , Ψ2 ⟩ ⊗ |1a′ ⟩ , (35)
2

where |Ψ1 , Ψ2 ⟩ := b†Ψ2 b†Ψ1 |Ω⟩ and |Ψ2 , Ψ2 ⟩ := b†Ψ2 b†Ψ2 |Ω⟩. The right-hand side of Eq. (35) is derived
h i
†
by expanding the left-hand side, repeatedly using bk1 , bk2 ≈ δk1 ,k2 and bk |Ω⟩ ≈ 0, and imposing
the Ψ1 and Ψ2 normalizations (Ψ1 |Ψ1 ) = (Ψ2 |Ψ2 ) = 1. Thus, upon taking θ = π/2 and assuming
(Ψ2 |Ψ1 ) ≈ 0, one arrives at

QWP (Ψ2 ) |Ψ1 ⟩ ⊗ |0a′ ⟩ ≈ −i |Ψ1 , Ψ2 ⟩ ⊗ |1a′ ⟩ . (36)

The two-wave-packets state is, therefore, expected to be prepared by QInit (Ψ1 , Ψ2 ) if the ancilla
qubit(s) is (are) measured in the state |1a1 ⟩ (|1a1 1a2 ⟩) when prepared using one (two) ancilla(s).
However, the imperfection in ansatz leads to deviation from this expectation, i.e., there will be
a non-zero probability of preparing states other than the wave-packet state when the ancilla(s)
measurement turn in the outcome |1a1 ⟩ (|1a1 1a2 ⟩). Furthermore, the hardware implementation of
QInit (Ψ1 , Ψ2 ) will incur the device noise, which could contribute to an error compounded with the
systematic error. We refer to this combined error as the ancilla-violation error, and denote it with
a
/ in the following.
Using two ancilla qubits in QInit (Ψ1 , Ψ2 ) results in a higher-fidelity state than using only one
ancilla qubit. Consider a noiseless scenario, where a / denotes purely the systematic error. In the
case of using only a1 for QInit (Ψ1 , Ψ2 ), the probability of ancilla being measured in |0a1 ⟩ counts
towards a /. Thus, if there is an error in the application of the first wave-packet preparation circuit
QWP (Ψ1 ), that is the ancilla-qubit state is a superposition of |0a1 ⟩ and |1a1 ⟩, this state (upon
being hit by σax1 ) propagates through the second wave-packet preparation circuit QWP (Ψ2 ), and
its |1a1 ⟩ component contributes to the probability amplitude of |1a1 ⟩ in the end of the circuit. This
yields a non-zero probability for preparing states other than the two-wave-packet state if ancilla
is measured in |1a1 ⟩, and such an error, clearly, remains undetected. On the other hand, the a /
error for QInit (Ψ1 , Ψ2 ) circuit with a1 and a2 is given by the combined probabilities of ancilla qubits
being in states |0a1 0a2 ⟩, |0a1 1a2 ⟩, and |1a1 0a2 ⟩. In this case, such an error can be detected, resulting
in higher fidelity for the final state. We numerically demonstrate these conclusions in Sec. IV B
using a noiseless quantum simulation of QInit (Ψ1 , Ψ2 ) using the two schemes.

C. QTrott : Trotterized time evolution

The circuit block QTrott (t) in the final module in Fig. 2 performs the unitary time evolution,
U (t) |Ψ1 , Ψ2 ⟩, where U (t) = e−itH with the Hamiltonian H in Eq. (16). The part of the circuit
shown in red color can be used for computing the return probability, R(t) := | ⟨Ψ1 , Ψ2 |U (t)|Ψ1 , Ψ2 ⟩ |2 ,
by performing the Hadamard test with an additional ancilla, at . It requires a controlled application
of the time-evolution operator U (t) with control on the ancilla, as shown in Fig 2. To study time-
dependent expectation values of an operator that is diagonal in the computational basis, one only
needs to evolve the system qubits using U (t). The circuit implementation of U (t) will be discussed
here, while the details of the Hadamard test and its extension to controlled operation are left to
Appendix D.
 19

FIG. 5. (a) Shown is the Trotterized time-evolution module QTrott from Fig. 2, in terms of its constituents
according to Eq. (37). The circuit blocks denote one Trotter step and the dots inside the circuit indicate
i h i m
repeated application of this structure for nt Trotter step. Circuits for e− 2 H δt and e− 2 H δt can be obtained
ϵ
from Fig. 3 as discussed in the text, and the circuit for e−iϵH δt is given in (b). The qubit labels for both
(a) and (b), and the single-qubit gate Rz have been defined in the caption of Fig. 3. The dotted rows in (b)
indicate similar circuit structure for qubits from f2 to fN −3 . The filled (unfilled) CNOT control indicates
that the target σ̃ x gate on the boson-link qubit is applied if the control even (odd) matter-site qubit is in
|1⟩ (|0⟩) implying a presence of fermion (anti-fermion); the Rz rotation that follows this operation adds the
appropriate phase to the state. A detailed explanation of its structure is given in the text.

We use the second-order Trotter product formula to approximate U (t), with δt as the Trotter
time step:
nt  
Y i h i m ϵ i m i h
U (t) ≈ e− 2 δtH e− 2 δtH e−iδtH e− 2 δtH e− 2 δtH , (37)
j=1

where t = nt δt, and H h , H m and H ϵ are defined in Eq. (17). Circuits for terms containing H m
and H ϵ can be realized without further approximation. Terms with H h , on the other hand, are
implemented by expanding them using the first-order Trotter product formula to separate the terms
i h
containing σ x operators from the terms containing σ y operators. The circuit blocks for e− 2 δtH and
i m
e− 2 δtH are identical to the subcircuit composed of two-qubit gates and the subcircuit composed
of single-qubit gates in QGS in Fig. 3, with θh = δt/4 and θm = mf δt/2, respectively. The circuit
i ϵ i ϵ
for e− 2 δtH is shown in Fig. 5, which uses the fact that the phase from e− 2 δtH is related to the
boson link and the matter distribution across the lattice. If the fermion occupation is 1 (0) at even
(odd) fermion sites, the bosonic qubit is acted with a σ̃ x such that the evolved phase is consistent
 20

with the Hamiltonian H ϵ in Eq. (17c). Furthermore, the conserved charge Q = NP implies that
after the controlled operation on the last fermion-site qubit, the boson-link qubit is restored to its
original value, and thus its phase can be evaluated according to the first term in Eq. (17c). Each
layer of second-order Trotterized time evolution requires 18NP + 8 CNOT gates.
This concludes the discussion on all the circuit elements involved in preparing the initial
scattering state and its time evolution. In the next section, we present the quantum-emulator
and hardware results arising from implementing these circuits to study a scattering process.

IV. QUANTUM HARDWARE AND EMULATOR RESULTS

The circuit layout shown in Fig. 2 was employed to study scattering in a (1+1)D Z2 LGT with
dynamical matter. The viability of this circuit was demonstrated by executing it on the IonQ Forte
quantum computer—a Ytterbium-ion-based quantum computer with 32 qubits, and high-fidelity
single- and two-qubit gates [119]. Due to limited device-time availability, the VQE optimizations
required for the parameters in the QGS circuit, and for the parameters that characterize the ansatz
in Eq. (20), as well as the Hadamard test for computing the return probability, were performed
using a noiseless quantum-circuit emulator. The quantum hardware was instead used to implement
the vacuum and wave-packet preparation circuits with known parameters, and to evolve the wave
packets in a Trotterized manner. A few noise-mitigation strategies are also employed to improve
the outcome of the simulations. All these results are presented and discussed in this section.
We work with two different system sizes, NP = 5 and NP = 13, corresponding to N = 10 and
N = 26, or equivalently 11 and 27 qubits, respectively. The Hilbert space with the symmetry
restriction Q = NP has 504 states for NP = 5, and 20, 801, 200 states for NP = 13. The former
system size is within the reach of the exact-diagonalization technique on a standard computer, while
the low-energy states of the latter system size could be obtained using the DMRG method [132]
applied to an MPS ansatz for the states [133], as discussed in Appendix A. We further set mf = 1.0
and ϵ = −0.3, which are the parameters used in our previous work [107]. This parameter set ensures
that the ratio of the contribution to the vacuum energy from the non-diagonal Hamiltonian, H h ,
to that from the diagonal Hamiltonian, H m + H ϵ , is approximately 0.221 in the ground state for
both NP = 5 and NP = 13. This choice, therefore, puts us in a non-trivial regime of parameters,
away from the strong- or weak-coupling limits.

A. Variational-quantum-eigensolver optimization

The VQE optimization for θh∗ and θm∗ parameters can proceed via the quantum circuit
introduced in Sec. III A. The results for these parameters, as well as the fidelity of the prepared
ground state, are summarized in Table V in Appendix G. These results clearly indicate that the
ground state prepared after the first module in Fig. 2 is a very good approximation of the true
ground state |Ω⟩. Next, the parameters that characterize the ansatz for the low-lying momentum-
eigenstate creation operator b†k in Eq. (20) are obtained for each target momentum kt ∈ Γ. e One
performs a VQE energy minimization for the states arising from QWP (Ψ) acted on |Ω⟩ with
Ψ(k) = δk,kt in Eq. (18). We use the second-order Trotter formula on the angle π/2 in Eq. (34)
with ten and two Trotter steps for NP = 5 and NP = 13, respectively. These choices lead to a small
Trotter error while keeping the VQE optimization time manageable on the noiseless emulator. The
results of the VQE optimization and the corresponding fidelities are summarized in Tables VI
and VII in Appendix G. These results indicate that the target states are captured with a very high
fidelity.
 21

NP σ1 k̄1 µ1 σ2 k̄2 µ2 (Ψ2 |Ψ1 )

7π 2π 7π 2π
5 2 − 7 0.0666
20 5 20 5
3π 2π 3π 2π
13 6 − 19 0.0104
13 13 13 13

TABLE II. Shown are the parameters of the initial-state wavefunctions defined in Eq. (38) for two different
lattice sizes. The last column quantifies the overlap of the wavefunctions, with (Ψ2 |Ψ1 ) defined in Eq. (32).

B. Preparing the initial hadronic wave packets

We consider the ith wave packet to have a Gaussian profile

(k − k̄i )2
 
Ψi (k) = NΨi exp (−ikµi ) exp − , (38)
4σi 2

with width σi centered at k̄i in momentum space, and at µi in position space. µi , σi , and k̄i
are all real parameters. The normalization constant Nψi is chosen such that (Ψi |Ψi ) = 1. The
two wave packets for the scattering simulation are centered around opposite momenta with the
same magnitude, i.e., k̄1 = −k̄2 , such that they move towards each other under time evolution.
Moreover, we set |µ1 − µ2 | = NP such that the wave packets are far separated in position space.
This choice ensures that (Ψ2 |Ψ1 ) is sufficiently small as required for the repeated application of
QWP , see Eq. (35). The parameters considered for the NP = 5 and NP = 13 simulations and the
corresponding (Ψ2 |Ψ1 ) values are shown in Table II.
The optimized parameters from Tables VI and VII, together with the wave-packet-profile
(j)
parameters in Table II, can now be used in Eq. (28) to compute the Cm,n (≡ Cm,n ) coefficients.
The j = 1 ansatz considered here contains terms with m = n and 1-length mesons, and the
corresponding coefficients determine the rotation angles for the one-qubit gates in QInit (Ψ1 , Ψ2 ).
We use the second-order Trotter formula with ñT Trotter steps for the angle π/2 in Eq. (34).
Furthermore, to reduce the circuit depth, we choose to implement only the angles θ such that
θ > θc . The choice of Trotter order, the number of Trotter steps, and θ > θc impact the accuracy of
each wave-packet preparation step. The implications of such approximations have been thoroughly
studied for a single-wave-packet preparation in our previous work [107]. Moreover, the number of
ancillary qubits used in QInit (Ψ1 , Ψ2 ), and the two wave-packets overlap (Ψ2 |Ψ1 ), introduce further
systematic errors in the two-wave-packet state preparation, as discussed in Sec. III B.
The various levels of approximations introduce systematic errors in the state preparation.
Nonetheless, they offer control over the quantum resources used, i.e., the number of qubits, single-
qubit, and two-qubit gates. We consider two different approximations (Appxs) to enable the
computations given the quantum resources available to us:

I. Using two ancilla qubits, second-order Trotter expansion, ñT = 10 for NP = 5 and ñT = 2
for NP = 13, and θc = 0,

II. Using one ancilla qubit, second-order Trotter expansion, ñT = 1 for both system sizes, and
θc = 0.1.
 22

FIG. 6. Shown are rotation angles for single-qubit gates appearing in the QInit (Ψ1 , Ψ2 ) circuit, plotted
against their corresponding lattice-site index n for (a) NP = 5 and (b) NP = 13. The wave-packet parameters
for the Gaussian wavefunctions Ψ1 (k) and Ψ2 (k) are given in Table II. Together with the optimized ansatz
parameters given in Table VI and VII, they determine the coefficients Cm,n with |m − n| ≤ 1. The real
(imaginary) parts of Cn,n coefficients, up to a proportionality constant, are shown in filled (unfilled) circles.
Similarly, the magnitude of forward (backward) 1-meson coefficients Cn,n+1 (Cn−1,n ), up to a proportionality
constant, are shown in filled (unfilled) stars. The proportionality constants arise from the second-order
Trotter expansion used to implement Eq. (34) with one Trotter step. Two different colors are used for each
wavefunction to denote the corresponding filled and unfilled markers for better visual differentiability. The
color used for filled (unfilled) marker is depicted as the color of the inner square (outer border) in the legend.
The green line shows the cutoff value θc = 0.1; all rotations angles that fall below this value are discarded
when when executing QInit (Ψ1 , Ψ2 ) on the IonQ Forte quantum computer.

The magnitudes of the rotation angles corresponding to Appx II are plotted in Fig. 6. As is
observed from the plotted values, only four Cm,n values per wave packet contribute to rotation
angles with θ > θc = 0.1.5
We compare observables in the state |Ψ1 , Ψ2 ⟩ prepared with either of these approximations.
Two observables, diagonal in the computational basis, are studied. One is the staggered (fermion)
density,
(
⟨ψn† ψn ⟩ if n ∈ even,
χn = (39)
1 − ⟨ψn† ψn ⟩ if n ∈ odd,
and the other is the electric-field value at the qubit encoding the hardcore boson,
E = ⟨σ̃ z ⟩ . (40)
The quantum resources required to implement QInit (Ψ1 , Ψ2 ) for the wave-packet parameters listed
in Table II are summarized in Table III considering both approximations.
Results for χn and E are displayed in Fig. 7. These computations are performed using the
noiseless Aer simulator using 5 × 105 shots (yielding negligible uncertainty from the shot noise).

5
As is seen from the figure, for the NP = 13 system, θc can be set to even smaller values without increasing the
number of Cm,n coefficients contributing to the θ > θc set.
 23

Single-qubit gates CNOT gates

NP Appx Qubits (raw/transpiled) (raw/transpiled) a
/

I 13 12757/13343 11324/7315 13.25 %

5
II 12 310/369 236/167 6.96 %

I 29 6925/7660 5948/3934 11.08 %

13
II 28 494/572 300/197 6.07 %

TABLE III. Shown are the number of qubits, single-qubit gates, and CNOT gates required for preparing
the initial scattering state |Ψ1 , Ψ2 ⟩ in Fig. 2 with the wave-packet parameters in Table II. The raw gate
counts can be obtained from the circuits described in the main text, while the transpiled gate counts are
obtained with the Qiskit transpiler. (Version 1.1.1 was used for Appx I while version 1.0.2 was used for
Appx II.) The two approximations, Appx I and Appx II, yield different accuracy as described in Sec IV B.
Only circuits for Appx II were implemented on the IonQ Forte quantum computer. Finally, the last column
denotes the ancilla-violation error a/, calculated using the noiseless Aer simulator (with 5 × 105 shots).

They are compared against the ideal results obtained from exact numerical results (for NP = 5) and
using an MPS ansatz for the states (for NP = 13). The Appx I results are in better agreement with
the ideal results. Appx II results, nonetheless, are not far off. Adopting Appx II, therefore, is not
unreasonable especially since its required resources are significantly lower than Appx I. These results
are obtained by discarding measurements for which the ancilla qubit(s) do not have the correct
value(s) |1a1 , 1a2 ⟩ for Appx I and |1a1 ⟩ for Appx II. The percentage of discarded measurements are
shown in Table III under the column a /, the ancilla-violation error. More states are discarded in
Appx I than Appx II, since a part of the erroneous states are hidden in the valid ancilla values in
Appx II, see discussions in Sec. III B.
The two-wave-packet preparation circuits within Appx II were executed on the IonQ Forte
quantum computer, with 1000 shots per circuit. The χn and E values obtained from these runs are
shown in Fig. 7. The hardware results contain various errors that are sourced from the trapped-
ion quantum devices. Similar to Ref. [107], we employ a simple post-processing error-mitigation
scheme based on the global symmetry of the fermion-qubits configurations. Explicitly, we discard
the states from the final results that exhibit Q ̸= NP . Such states indicate the occurrence of at
least one error in the fermion qubits. We refer to this error as the symmetry-violation error, and
denote it by Q. / We find Q / = 49.40 % for NP = 5 and Q / = 71.20 % for NP = 13. The states
in the set complementary to Q / are not necessarily contamination free, since the errors that do
not change Q are not filtered away. Part of these errors percolate into the ancilla-violation error:
a
/ = 14.82 % for NP = 5 and a / = 18.40 % for NP = 13. These are larger than their respective
noiseless-simulator results in Table III, due to additional hardware errors. The error a / is calculated
over the shots remained after discarding those with a Q / error. States with the incorrect values
of ancilla measurement are also discarded before evaluating the observables’ expectation-values.
Finally, the error bars in Fig. 7 are obtained from the standard deviation of the mean of the
bootstrap samples of the physical events with 100 resampled configurations (at which value the
bootstrap-sample mean distributions is stabilized).
The hardware results are in a good agreement with their noiseless-simulator counterparts for
both observables. The deviation from noiseless simulator is generally more prominent at the center
of the wave packet. This feature can be attributed to the fact that the qubits at the center of
 24

FIG. 7. Shown are the staggered density χn across lattice-site index n (left) and the electric field at the
boson qubit E (right) for the two-wave-packet scattering states, with the wave-packet parameters in Table II.
The results for the lattice sizes NP = 5 (top, 11 system qubits) and NP = 13 (bottom, 27 system qubits)
are compared against the ideal results, which are obtained by exact numerical solution (empty green circles)
and by using the MPS-ansatz states (solid green circles), respectively. The dashed green line is depicted for
visual guidance and is not a fit to data. Two approximations are implemented: Appx I (yellow stars) and
Appx II (red stars), both obtained using Aer noiseless simulator using 5 × 105 shots. Additionally, Appx II is
implemented on the IonQ Forte quantum computer (cyan triangles) with 1000 shots for each NP . The error
bars are obtained from bootstrap resampling of the global-symmetry-based noise-mitigated hardware results.
The ranges for the y-axes in all plots are taken to be over all possible values the corresponding observable
can take. The axis for E values is broken and rescaled appropriately to resolve the closely located data
points near its maximum allowed value.

the wave packets are acted on by gates with larger angles (i.e., larger |Cm,n | values as seen in
Fig. 6). Larger gate angles indicate longer gate-implementation times, making the gates more
susceptible to quantum decoherence. In summary, the ability to systematically control the various
levels of approximation allows us to compromise marginally on the accuracy of preparing the initial
scattering state while benefiting from significant reduction in quantum-resource requirements. This
is especially useful if the observable under investigation is local, and is, hence, robust against
extensive error accumulation under time evolution [134, 135]. We further investigate this point in
 25

t 1 2 3 4 5 6 7 8

Single-qubit gates 573 771 969 1167 1365 1563 1761 1959

CNOT gates 227 287 347 407 467 527 587 647

Q
/ 56.07 % 62.43 % 67.13 % 70.77 % 71.73 % 73.47 % 73.63 % 76.73 %

a
/ 15.78 % 17.66 % 16.02 % 19.61 % 18.28 % 19.47 % 20.61 % 20.77 %

TABLE IV. Shown are the values of the single- and two-qubit gate counts, symmetry-violation error Q, /
and the ancilla-violation error a
/, for different Trotter steps t in time evolution of two meson wave packets,
resulted from the IonQ Forte device. The gate counts are associated with the transpiled quantum circuit
by the Qiskit transpiler. (The raw gate counts can be obtained from those provided in Table III for the
state preparation and the cost of each Trotter step of evolution given in the text.)

Secs. IV C and IV D by studying different observables under time evolution.

Finally, our method can be extended to prepare multiple spatially spaced wave packets. One
can repeat the application of QWP (Ψi ) with one or more ancilla qubits as many times as needed.
We have demonstrated the preparation of three wave packets in an NP = 13 system on the IonQ
Forte, see Appendix C, observing good agreement with the expected results.

C. Time-evolved observables

We use the circuit for Trotterized time evolution via a second-order Trotter product formula
shown in Sec. III C to evolve an initial two-wave-packet state on an IonQ Forte quantum computer.
The NP = 5 parameters in Table II are chosen for this purpose since each Trotter time step involves
a shallower circuit than the larger-system counterpart: each Trotter-step circuit block constitutes
204 single-qubit gates and 60 CNOT gates. The three modules in Fig. 2 are implemented with
Appx II for 3000 shots and δt = 1. The number of shots and the size of the Trotter step required
to match the theoretical prediction are estimated using the noisy emulator provided for this device
by IonQ.6
The device results are analyzed using the methods described in the previous subsection, and
the Q/ and a/ values calculated for each Trotter step are shown in Table IV. The growing Q / and a/
errors with each Trotter step indicate increasing noise in the device results with deeper circuits. As
mentioned before, the a / error in the noiseless simulation denotes the systematic error due to the
approximate nature of b†k operators. In this case, this error does not change with time evolution
as the ancilla qubit used for the initial-state preparation does not participate in the time-evolution
circuit. Thus, the increasing a/ error with each Trotter step in Table IV purely reflects the increasing
hardware error.
The results for the time-evolved staggered density are shown in the Fig. 8. The values obtained
from the quantum circuits are compared against the exact calculations that do not exhibit the
Trotter error (i.e., they are obtained by calculating U (t) using exact matrix exponentiation). From
these results, the wave packets can be seen to be moving towards each other with their peak

6
For future simulations involving large system sizes, a viable strategy is to repeat experiment for a range of Trotter-
step sizes and shot numbers to reach convergence in measured observables.
 26

FIG. 8. Shown are the expectation values of the staggered density χn across lattice-site index n, in a
Trotterized time-evolved scattering state of two meson wave packets for NP = 5 with 11 system qubits.
Each column (row) shares the x-axis (y-axis) label. The plot legends and error bars are the same as in
Fig. 7. The exact results correspond to the evaluation of time-evolution unitary matrix U (t) = e−itH
upon exact exponentiation of the Hamiltonian matrix for a given time t, acting on the corresponding initial
state in Fig. 7(a). The noiseless-simulator and hardware results correspond to 5 × 105 and 3000 shots,
respectively. The quantum circuits for the Trotter time evolution are taken with time steps of δt = 1. The
meshed squares for the t = 7 and t = 8 plots denote the hardware-noise-dominated results. The number of
single-qubit and CNOT gates implemented for the combined state-preparation and time-evolution circuits
for each t is provided in Table IV.
 27

FIG. 9. Shown are the expectation values of the electric field E at the boson qubit, in a Trotterized time-
evolved scattering state of two wave packets for NP = 5 with 11 system qubits. The hardware results (cyan
triangles) are dominated by noise. Results obtained using the IonQ Forte noisy emulator are shown in
Appendix F to demonstrate a possible recovery of the signal using additional noise-mitigation tools, such
as Pauli twirling and operator decoherence renormalization, at the cost of additional quantum-processing
time.

values decreasing with time. The noiseless results (using 5 × 105 shots) in both approximations
(Appxs I and II) follow the exact wave-packet profile in the early times, but deviate from the
exact values with increasing evolution time due to accumulated Trotter error. In fact, in some
instances, the cruder, hence less resource-intensive, Appx II displays less Trotter error than Appx
II. This observation makes it clear that small deviations in the initial states may be insignificant in
subsequent simulation steps given the effect of other errors during the evolution. Nonetheless, the
shape of the profile for both initial-state approximations resembles the exact wave-packet evolution,
and still offers qualitative description of the physical process.
The hardware results also show the qualitative features of time-evolved staggered density,
however, the device noise starts to dominate after Trotter time t = 6. Furthermore, our results are
obtained without performing any additional error-mitigation-circuit runs and only by discarding
the a/ and Q / errors during post-processing. Nonetheless, we have checked that discarding the a /
and Q/ data does not significantly improve the outcome, and only leads to reduced statistics, hence
larger shot noise. The reason can be attributed to the fact that most of symmetry-violated errors
are associated with a few (mostly one) bit-flip errors in primarily random locations in the qubit
register, whose effect becomes insignificant when computing expectation value of local operators.
Our observation is consistent with the conclusions of Ref. [136], which finds that the effect of
symmetry-based noise mitigation is both quantity dependent and time dependent.
We further compute the time evolution of the electric-field expectation value at the boson qubit,
E, as shown in Fig. 9. Here, the noiseless result agrees with the exact result up to Trotter errors,
and almost retains its value at t = 0 throughout the evolution. The reason is that flipping the
electric field at the boson qubit costs an energy proportional to the system size, see Eq. (17c).
The hardware result for this quantity, on the other hand, significantly deviates from the exact
value. Thus, obtaining the Trotter time evolution of E requires further noise-mitigation techniques.
 28

FIG. 10. Shown is the return probability, R(t), of the initial two wave-packet state for NP = 13 (27 system
qubits) against time. The Trotter time step is δt = 0.25, but results for only integer times are plotted. The
MPS results are obtained using the TDVP algorithm, and the values for Appxs I and II cases are calculated
using the circuit for the Hadamard test.

We have demonstrated this in Appendix F using the IonQ Forte noisy emulator where Pauli
twirling [137], along with operator decoherence renormalization [86, 138, 139], are employed to
recover the time-evolved value of E.7
The hardware runs’ moderate coherence time, and time evolution’s large circuit depth, prohibit
accessing interesting long-time scattering dynamics in this hardware study. Another issue is the
small system size, leading to boundary affects in the simulation outcome in the long-time limit. One
would, therefore, need to simulate evolution of wave packets in larger systems, but the associated
circuit depths would increase considerably. Nonetheless, the results presented here marks the first
hadron-scattering simulation on a quantum computer;8 it has pushed the limits of what is possible
for such an involved simulation problem on any quantum hardware to date.

D. Return probability

Computing the scattering S-matrix, which relates scattering states in early and late times, is
a critical observable in nuclear and high-energy physics. In this section, we compute the return
probability (also known as the survival probability and Loschmidt echo):
R(t) := |⟨Ψ1 , Ψ2 |U (t)|Ψ1 , Ψ2 ⟩|2 , (41)
which is a diagonal entry of the scattering S-matrix. Here, we restrict our discussion to the
computation of return probability for the initial state consisting of the two wave packets prepared
in Sec. IV B, rather than its phenomenological implications, or more interesting, but significantly
more involved, final-state-momentum-dependent overlaps.
7
We relied on the noisy emulator for this analysis because such mitigation techniques require more quantum
processing, and our access to the device was limited.
8
See also the parallel submission by Schuhmacher et al. in the same arXiv listing.
 29

Consider the case of NP = 13 with the initial state |Ψ1 , Ψ2 ⟩ shown in Fig. 7(b), using both
approximations, Appxs I and II. Recall that the former (latter) approximation requires more
(less) quantum resources, and prepares a state with higher (lower) fidelity with respect to the
target state. The return probability for each approximation is computed using the Hadamard test,
shown schematically in Fig. 2 and described in more detail in Appendix D, using the noiseless Aer
emulator. The Trotter step size is taken to be δt = 0.25 to avoid a large Trotter error, and the results
are shown in Fig. 10 only for integer times to avoid clutter. The emulator results are compared
against the ideal results obtained from optimizing an MPS ansatz for the state and performing the
time evolution using the time-dependent variational-principle (TDVP) algorithm [140–142].
Figure 10 indicates that the return probability associated with Appx II deviates significantly
from the ideal case, but that associated with Appx I agrees with the ideal result up to Trotter
errors. This can attributed to the non-local nature of the observable R(t) which is sensitive to the
state vector. Large deviations from the target state vector, like in Appx II, cause interference effects
from the contamination when computing the probability in Eq. (41). This is in contrast to local
observables that are rather insensitive to small deviations from the target state, see Appendix E
for the example of local staggered density for the same simulation.
Preparing a high-fidelity initial state is, therefore, of crucial importance in future precision-
physics applications of quantum computers, especially for observables like the S-matrix elements
that are highly sensitive to the state vector. On the other hand, understanding qualitative behavior
of local observables under time evolution is a more tractable problem for the NISQ devices, as
demonstrated in Sec. IV C.

V. CONCLUSION AND OUTLOOK

We developed and implemented a digital algorithm for simulating scattering of hadrons on a

quantum computer. Focusing on a (1+1)D Z2 lattice gauge theory with fermionic matter, we first
prepared the initial scattering state of well-separated single-hadron wave packets with well-defined
localized momenta. Our hybrid classical-quantum state-preparation algorithm offers control over
various levels of approximations for preparing the initial state, such that the fidelity with the target
state can be systematically improved at the cost of increasing quantum resources. We prepared
the two- and three-wave-packet state on the IonQ Forte trapped-ion-based quantum computer for
systems with 10 and 26 fermionic sites, using one of these approximations, requiring significantly
lower circuit depths, see Table III. Even in the presence of hardware noise, results agree well with
the ideal results, see Figs. 7 and 14, and a symmetry-based noise mitigation yielded only small
improvement.
We then performed a quantum time evolution using the second-order Trotter product formula
for a system of 10 fermionic sites, which allowed for hadrons to evolve till they collide, see Fig. 8.
Unfortunately, the device noise washed out the signal for the local fermion-density observable
before the hadronic wave packets could fully collide. Nonetheless, the early-evolution time data
are encouraging, despite a large number of entangling gates used and no major noise-mitigation
strategy implemented. The electric-field energy stored in the remnant gauge link (after gauge
fixing) exhibited significant decoherence even at the early steps of time evolution, see Fig. 9.
Nonetheless, using a noisy emulator, we demonstrated the effectiveness of two noise-mitigation
schemes, Pauli twirling together with operator decoherence renormalization, in recovering the
electric-field value.
Finally, we emphasized the importance of preparing high-fidelity initial scattering states for
future precision computations of state-sensitive observables, such as phase shifts, S-matrix, and
decay width. Our computation of the return probability, which is a diagonal entry of the S-matrix,
 30

demonstrates that approximations that may appear benign when measuring local observables,
appear to exacerbate the error in overlap quantities such as return probability, see Fig. 10.
In the following, we present an outlook of this study, including potential avenues for future
research:

⋄ Our ansatz is both effective and efficient in creating single-particle hadronic states with
high fidelity in (1+1)D Abelian theories (see our previous work for a U(1) example [107]).
Given the ultimate goal of preparing hadronic states in QCD on a quantum computer, the
natural next step is to extend the ansatz to non-Abelian LGTs, and to LGTs in more than
one spatial dimensions. These involve introducing additional gauge-invariant bare-meson
operators, both arising from non-Abelian components of the fields, and from Wilson-loop
operators in (2+1) and higher dimensions. One hopes that the ansatz can approximate the
target hadronic state with high fidelity using only a polynomial number of parameters, even
toward the continuum limit.

⋄ Quantum simulation of non-Abelian LGT dynamics in more than one spatial dimensions is
a resource-intensive task [143–153]. It is perceivable that various alternative formulations
of the Kogut-Susskind Hamiltonian formulation of LGTs [145, 154–179] will alleviate some
of the resource requirements. Among these, the loop-string-hadron (LSH) formulation [155–
157], is particularly appealing since its degrees of freedom are generated by fully local and
gauge-invariant operators in the electric basis, and only an algebraic Abelian constraint
remains to be enforced on each link of the lattice. This formulation, therefore, may present
a cost-efficient framework for extending our ansatz. Bare-meson operators can be simply
mapped to gauge-invariant states of the LSH formulation, and the Abelian constraints can
aid noise mitigation in quantum-circuit implementations.

⋄ The single-hadron momentum eigenstates prepared using the algorithm of this work, and
extended to more complex gauge theories, are crucially needed in studying hadron structure
using quantum computers. Ultimately, one aims to constrain, from real-time correlators of
hadrons, various non-perturbative QCD matrix elements of relevance to collider physics [180],
such as various distribution functions [181–188] for large-hadron and electron-ion colliders.
The expressive form of our ansatz, and its potential extensions, may prove valuable in gaining
more insights into such structure quantities in the future.

⋄ Our method is applicable to preparation of any number of hadronic wave-packets (as long
as systems size is large enough to contain well-separated wave packets). This feature is
particularly useful when one aims to compute exclusive scattering amplitudes. A second
qubit register is used to prepare the desired multi-hadron final state. Then a swap test [189]
can be used to compute the overlap between the time-evolved initial state and the desired final
state. Upon insertion of spacetime-separated electroweak currents, one can further access
quantities such as hadron tensor. Thus, the optical theorem can be invoked to obtain the
inclusive scattering cross section for a given center-of-mass energy. However, both of these
analyses require larger lattice sizes and evolution times, hence wider and deeper quantum
circuits.

⋄ The scattering protocol presented here allows one to access the outgoing scattering state
at any moment past collision. Therefore, not only can one access asymptotic observables
such as the S-matrix, but also can measure other local and non-local observables in early
moments after the collision and beyond. A novel quantity of interest, not readily accessible
in collider experiments, is entanglement, in form of bipartite entanglement entropy or
 31

entanglement Hamiltonian. There is significant interest in quantum simulations of string

breaking, hadronization, fragmentation, and thermalization in the aftermath of high-energy
hadronic collisions [83, 190–203]. Nonetheless, many of early experimental studies are limited
to simple models and scenarios, and do not involve high-energy hadronic collisions. Future
quantum-simulation studies, built upon efficient high-energy hadronic-state preparation
schemes (such as that presented in this work), can begin to shed light on such important
non-equilibrium phenomenology.

Note: A parallel submission by J. Schuhmacher, G. X. Su, J. J. Osborne, A. Gandon, J. C. Halimeh,

and I. Tavernelli is to appear in the same arXiv listing as our paper, where authors perform a
quantum-simulation experiment of hadron scattering in a quantum-link-model formulation of the
U(1) LGT on an IBM quantum computer.

ACKNOWLEDGMENT

Quantum-emulator and quantum-hardware computations of this work were enabled by access

to IonQ systems provided by the University of Maryland’s National Quantum Laboratory (QLab).
We thank support from IonQ scientists and engineers, especially Daiwei Zhu, during the execution
of our runs. We further thank Franz Klein at QLab for the help in setting up access and
communications with IonQ contacts. S.K. acknowledges valuable discussions with Roland Farrell
and Francesco Turro. 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. The numerical results in this work were obtained using Qiskit [204], ITensors [205],
Julia [206] and Jupyter Notebook [207] software applications within the Conda [208] environment.
Z.D. and C-C.H. acknowledge support from the National Science Foundation’s Quantum
Leap Challenge Institute on Robust Quantum Simulation (award no. OMA-2120757); the U.S.
Department of Energy (DOE), Office of Science, Early Career Award (award no. DESC0020271);
and the Department of Physics, Maryland Center for Fundamental Physics, and College of
Computer, Mathematical, and Natural Sciences at the University of Maryland, College Park. Z.D.
is further grateful for the hospitality of Nora Brambilla, and of the Excellence Cluster ORIGINS
at the Technical University of Munich, where part of this work was carried out. The research
at ORIGINS is supported by the Deutsche Forschungsgemeinschaft (DFG, German Research
Foundation) under Germany’s Excellence Strategy (EXC-2094—390783311). S.K. acknowledges
support by the U.S. DOE, Office of Science, Office of Nuclear Physics, InQubator for Quantum
Simulation (IQuS) (award no. DE-SC0020970), and by the DOE QuantISED program through the
theory consortium “Intersections of QIS and Theoretical Particle Physics” at Fermilab (Fermilab
subcontract no. 666484). This work was also supported, in part, through the Department of
Physics and the College of Arts and Sciences at the University of Washington.

Appendix A: Verifying the ansatz validity for a larger system using tensor networks

In this appendix, we study the applicability of our wave-packet ansatz to a larger system
than considered in Sec. II C. We consider the Z2 LGT on a lattice with 26 staggered sites (13
physical sites) with the Brillouin zone Γ̃ = {− 6π 5π 6π
13 , − 13 , · · · , 13 }. The true momentum eigenstates,
while cannot be obtained by exact diagonalization due to the size of the Hilbert space, can be
approximated with a matrix product states (MPS) ansatz on a tensor network. Using the density-
matrix-renormalization-group (DMRG) optimization, one can obtain |Ω⟩MPS that recovers the
exact interacting ground state |Ω⟩ with high precision in (1+1)D. Excited states can also be built
 32

FIG. 11. The parameter scan for the 26-site (1+1)D Z2 LGT demonstrates the performance of the ansatz
at a large system size. The infidelity 1 − F is shown for non-negative momenta in the Brillouin zone in each
columns (on a logarithmic scale). Results using the j th -order ansatz is shown in the j th row. In general,
the ansatz performs better in the strong-coupling regime, which agrees with the results in a smaller system
in Sec. II C. F > 0.99 is achieved at the 3rd order except for |k = 6π
13 ⟩ with mf = 0.1, 0.3 and ϵ = −0.1. At
these couplings, there are other excited states (above the single-particle state) not belonging to the k = 6π 13
sector but with energies very similar to that of |k = 6π13 ⟩, reducing the effectiveness of the MPS ansatz to
discern excited states reliably. The MPS states with these coupling are thus less reliable for the fidelity test,
and the corresponding parameter range is hashed in the figure.

successively using the same DMRG process upon constraining the output to be orthogonal to the
ground state and all other excited states with lower energy, if any. The MPS states are further
constrained such that the total fermionic excitation remains in the Q = NP sector, and thus the
gauge symmetry is guaranteed on a periodic lattice. All DMRG calculations in this work are carried
out using the ITensors library [205].
The DMRG calculations for the ground state and the 13 excited states with the lowest energy in
each momentum sector are performed for a range of (mf , ϵ) values. In each DMRG optimization,
the maximum allowed bond dimension is set to 600, and the Schmidt coefficients under 10−12 are
truncated. The number of sweeps is set to O(100) to ensure convergence in energy. To further
verify the quality of an output DMRG state |ψ⟩MPS , the variance of energy is calculated:

MPS ⟨ψ|H
2 |ψ⟩ − (MPS ⟨ψ|H|ψ⟩MPS )2
MPS
Var := . (A1)
(MPS ⟨ψ|H|ψ⟩MPS )2
If |ψ⟩MPS is exactly an eigenstate of H, the variance should be zero. For all pairs (mf , ϵ), the
DMRG parameters listed above can achieve Var = O(10−10 ).
Similar to Sec. II C, the fidelity between the order-by-order ansatz states |k⟩op and the DMRG
states |k⟩MPS , now assumed to approximate the exact eigenstates with very high accuracy, is
presented for each k value belonging to the Brillouin zone. However, one caveat is that the DMRG
momentum states |k⟩MPS for k ̸= 0 are not accessible due to the energy degeneracy between
 33

FIG. 12. (a) Shown is the circuit decomposition of QWP (Ψ) that acts on the system and the ancilla qubits
as denoted in Fig. 4. Each inner light-blue circuit block represents a Trotter step for the Trotterized angle
π/2 in Eq. (34). The dots and the green circuit block within each light-blue block denote that each term
after the Trotter expansion is of the form e−iθΘm,n for m, n ∈ Γ, where θ depends on the Trotter order and
Trotter step and Θm,n is defined in Eq. (B1). (b) Shown is the circuit for e−iθΘm,n with m ̸= n following
Eq. (B3). Explicit circuits for the unitary operation Um,n (pink) and the diagonal operation e−iθDm,n (light
green) are presented in Fig. 13 for the choice of j = 1 in Eq. (20), which we have used in this work. Circuits
for general j can be constructed similarly.

|k⟩MPS and |−k⟩MPS . The momentum quantum number is not specified during the DMRG sweeps.
We assume the output DMRG states with the same energy, labeled |ψ1 ⟩MPS and |ψ2 ⟩MPS , are
approximately a superposition of the actual momentum eigenstate |±k⟩MPS :
−
|ψ1 ⟩MPS ≈ c+
1 |k⟩MPS + c1 |−k⟩MPS ,
−
|ψ2 ⟩MPS ≈ c+
2 |k⟩MPS + c2 |−k⟩MPS , (A2)
2 2 2 2
with c+ 1 + c−1 ≈ c+2 + c−
2 ≈ 1. Since the orthogonality between the degenerate states holds
for the DMRG algorithm (up to numerical precision), i.e., MPS ⟨ψ1 |ψ2 ⟩MPS = MPS ⟨ k| − k⟩MPS = 0,
2 2 2 2
it is easily seen that c+
1 + c+2 ≈ c− 1 + c−2 ≈ 1. Furthermore, since the ansatz is built with
specified momentum, op ⟨k| − k⟩MPS = 0. As a result, one can still access the fidelity of |k⟩op using
|ψ1/2 ⟩MPS , since
2 2 2 2 2 2
op ⟨k|ψ1 ⟩MPS + op ⟨k|ψ2 ⟩MPS ≈ c+
1 op ⟨k|k⟩MPS + c+
2 op ⟨k|k⟩MPS
2
= op ⟨k|k⟩MPS = F. (A3)
The fidelity of the ansatz scanning over different mf and ϵ values is demonstrated in Fig. 11.
Similar to the parameter scan on a smaller system in Sec. II C, only non-negative momenta in
 34

the Brillouin zone are plotted for brevity. Similar fidelities are observed for negative-momentum
eigenstates. Again, the ansatz performs better in the strong-coupling regime, and the order-by-
order improvement is visible. Here, the system size is large enough such that the non-mesonic
excitations have energy well above that of the |k = 0⟩ eigenstate, and thus the cyan contour in
Fig. 1 is not present. The 3rd order ansatz is able to achieve F > 0.99, except for the highest
momentum |k = 6π 13 ⟩ for a small parameter range. Specifically, for mf = 0.1, 0.3 and ϵ = −0.1,
there are excited states in other momentum sectors (above the single-particle eigenstate) with
energy very similar to that of ||k| = 6π 13 ⟩. It is difficult for the DMRG to discern these from the
eigenstates in the |k| = 6π
13 sector. As a result, the orthogonality originally assumed for the states
in Eq. (A2) is not strictly valid, i.e., |ψ1/2 ⟩ have overlap to states not in the k = 6π
13 sector. Ideally,
a translationally invariant MPS ansatz based on quasiparticle excitations can do a better job at
yielding the momentum eigenstates [51, 209–211].
We conclude that the momentum eigenstates can be faithfully be built from a finite-order ansatz
of this work even in larger systems, although for verification purposes, better MPS ansatzes may
be needed for large values of momenta to allow a fidelity test.

Appendix B: Circuit for QWP in the minimal-gauge formalism

In this appendix, we describe the circuit for QWP (Ψ) that creates the initial single-particle
wave packet with wavefunction profile Ψ. In other words, we aim to implement the operation in
Eq. (34), where the sum in Eq. (33) is exponentiated using the second-order Trotter formula with
one Trotter step for the angle π/2. Each term in the Trotter expansion is denoted by e−iθΘm,n for
θ ∈ R, where Θm,n is given by
∗ f†m,n ⊗ |0a ⟩⟨1a |,
Θm,n := Cm,n M
fm,n ⊗ |1a ⟩⟨0a | + Cm,n M (B1)

such that,
X
ΘΨ = Θm,n . (B2)
m,n∈Γ

Thus, QW P (Ψ) is composed of circuits for e−iθΘm,n , as shown in Fig. 12(a).

To implement each e−iθΘm,n , as in our previous work [107], we apply an efficient circuit design
based on a singular-value-decomposition (SVD) method [149]. Consider an operator e−iθΘm,n with
Θm,n = A† ⊗|1a ⟩⟨0a |+A⊗|0a ⟩⟨1a | such that A2 = A†2 = 0 and the SVD decomposition A = V SW † .
Then,

e−iθΘm,n = U † e−iθDm,n U, (B3)

where U = Ha (V † ⊗ |0a ⟩⟨0a | + W † ⊗ |1a ⟩⟨1a |) and Dm,n = S ⊗ σaz . The operator A can be read off
from Eq. (B1). It can be easily verified from Table I that due to the presence of the σ ± operators
in Mfm,n for m ̸= n, A2 = A†2 = 0.
The circuit for implementing Eq. (B3) is shown in Fig. 12, with its elements specified in Fig. 13
for the case j = 1 in Eq. (20), i.e., the bare-meson creation operators Mm,n and their Jordan-
Wigner-transformed forms M fm,n are restricted to create at most 1-length mesons, which we have
used in Sec. IV. In this scenario, |m − n| ≤ 1 in Eq. (B1), where |m − n| is defined below Eq. (21).
The circuit for implementing terms with m = n is trivial, and thus, in not discussed here.
Finally, the circuit for QWP (Ψ) requires the coefficients Cm,n . Recall that, each Cm,n depends
(ℓ),k ∗
on the input wavefunction Ψ(k) as well as the ansatz parameters α0/1 which optimize the b†k
 35

FIG. 13. When the ansatz for b†k is restricted to have operators smaller or equal to 1-length mesons, i.e.,
j = 1 in Eq. (20), the circuits for Um,n and e−iθDm,n for |m − n| = 1 can be obtained from the example
shown in (a) and (b), respectively. Here, m = N −1 and n = 0 is considered with CN −1,0 = eiϕN −1,0 |CN −1,0 |
and using a second-order Trotter expansion on the angle π/2 in Eq. (34) with one Trotter step. This term
includes the gauge-boson qubit bN −1 in the MGF. The circuit for UN −1,0 is given in (a), where the system
qubits and the ancilla qubit are denoted according the convention explained in the caption of Fig. 12. All
CNOT gates are controlled on the ancilla qubit with the filled circle denoting the control state |1⟩. The
CNOT gates colored in black are targeted on qubits f0 and fN −1 , while the one in red is targeted on the
qubit representing the boson link. The single-qubit gates are the the Hadamard gate H and rotation gate
Rz (defined in the caption of Fig. 3). The circuit for e−iθDN −1,0 is shown in (b) with β := θ |CN −1,0 | /2 with
θ = π/4 for a single-step second-order product-formula implementation. The circuits in (a) and (b) remain
the same for the operators that do not include the boson link but still satisfy |m − n| = 1, after making the
following changes: 1) adjusting the CNOT-gate positions appropriately, 2) using the corresponding values
for β and ϕ, and 3) omitting the red CNOT gate in (a).

operator. The former component is an input based on the desired scattering state one wants
to prepare. The latter are obtained using a VQE method that minimizes the energy of the state
prepared by acting b†k from Eq. (20) on |Ω⟩. To obtain the corresponding VQE circuit, note that the
b†kt operator for a target momentum sector with momentum kt , which is parameterized by α0/1 t ,
(ℓ),k

is obtained by taking Ψ(k) = δk,kt in Eq. (26). Thus, the circuit for variational optimization of
b†kt in Eq. (20) is given by the same circuit that prepares the wave packet Ψ(k), i.e., QWP (Ψ) with
Ψ(k) = δk,kt .

Appendix C: Three-wave-packet preparation

In this appendix, we demonstrate the preparation of an initial scattering state containing three
wave packets for an NP = 13 system. Here, the third wave packet is produced with µ3 = 13,
σ3 = 3π/13 and k̄3 = 2π/13 in Eq. (38), and only Appx II is considered for the QInit module.
The wave packets do not significantly overlap, with the overlap factors, defined in Eq. (32), being
(Ψ2 |Ψ1 ) = 0.0104, (Ψ3 |Ψ1 ) = −0.0306, and (Ψ3 |Ψ2 ) = 0.0059. The QInit circuit requires 267
CNOT gates, and the circuit was implemented with 2000 shots. The rotation angles, along with
the results for the staggered density, are shown in Fig. 14. The hardware results yield Q
/ = 75.7 %
and a/ = 17.70 % (compared to a / = 6.25 % using the Aer noiseless-simulator with 5 × 105 shots).
Overall, the three-wave-packet state exhibits the features of the target state obtained from the
MPS ansatz for the state. Lastly, the observable E obtained from the hardware result is in within
two standard deviations of the noiseless and the ideal results.
 36

FIG. 14. (a) Shown are rotation angles for single-qubit gates appearing in the QInit (Ψ1 , Ψ2 , Ψ2 ) circuit,
plotted against their corresponding lattice-site index n for a system of NP = 13 physical sites, using Appx II
for the preparation circuit. The wave-packet parameters for the first two wave packets are listed in Table II.
The third wave packet is produced with µ3 = 13, σ3 = 3π/13 and k̄3 = 2π/13 in Eq. (38). The optimized
ansatz parameters are provided in Table VII. For other details, see the caption of Fig. 6. (b) Shown are the
staggered density χn (left) and the electric field at the boson qubit E (right) for an initial state containing
three wave packets on a NP = 13 lattice. Only Appx II is used to execute the three-wave-packet version
of the QInit circuit. The Aer noiseless-simulator results corresponds to 5 × 105 shots while the IonQ Forte
quantum-computer results correspond to 2000 shots.

Appendix D: Hadamard test for calculating the return probability

Consider the amplitude

A(t) := ⟨Ψ1 , Ψ2 |U (t)|Ψ1 , Ψ2 ⟩ . (D1)

The return probability R(t) is then given by

R(t) := |A(t)|2 = Re(A(t))2 + Im(A(t))2 , (D2)

 37

FIG. 15. Shown are the staggered density χn corresponding to four instances of time during the evolution,
plotted against the lattice-site index n. The values of χn agree well with the ideal result for both Appxs
I and II. These are to be compared with the return-probability results plotted in Fig. 10, which exhibit a
significant deviation for Appx II.

The real (Re) and imaginary (Im) parts of A(t) can be computed using a standard Hadamard
test. The state |Ψ1 , Ψ2 ⟩ ⊗ |0at ⟩ is first acted with a Hadamard gate on the ancilla at , followed by
a controlled U (t) operation upon the |1⟩ state of the ancilla. Finally, a Hadamard or an Rx ( π4 )
gate is acted on the ancilla to obtain the real or imaginary parts, respectively. Denote by p0 and
p1 the probability for the ancilla to be measured in state |0at ⟩ and |1at ⟩, respectively. Then the
quantity p0 − p1 obtains the real or the imaginary part out of the respective circuits. The circuit
for controlled U (t) can be obtained by controlling each block in Fig. 5 upon the state of the ancilla
h m
at every Trotter step. The controlled e−iδtH and e−iδtH circuits are given by controlling every
ϵ
constituent gate, while the controlled e−iδtH is obtained by replacing the Rz gates with their
controlled versions. This is because in the absence of the Rz gates in the circuit in Fig. 5(b), the
collective action of the entangling gates is an identity on the state of fermion and the boson qubits.

Appendix E: Time evolution of local observable for NP =13

In Fig. 10 of the main text, we presented the results for the return probability R(t) for an
NP = 13 system obtained from the MPS ansatz, as well as Appxs I and II using the noiseless
Aer emulator. We observed significant deviation in this non-local quantity for the cruder Appx
II, possibly due to large interference effects. In this appendix, we consider time evolution of the
staggered density χn instead, to investigate if the same deviation is seen for a local quantity.
To enable this comparison, we set all the simulation parameters identical to those in Fig. 10,
i.e., Trotter time step δt = 0.25 with 5 × 105 shots for the noiseless emulator. The results for χn
 38

FIG. 16. Shown are the noise-mitigated result obtained from Pauli twirling followed by operator decoherence
renormalization (ODR) for the electric field at the boson qubit, E, after two-wave-packet state evolution for
time t in the NP system (compare with Fig. 9). The inverted magenta triangles are the results from the Pauli-
twirled U (t) circuits. The black crosses are the Pauli-twirled identity circuits, U (0), with similar structure as
U (t). Both of these results were obtained using the noisy emulator for the IonQ Forte quantum computer
using 3000 shots. The error bars obtained from 1000 bootstrap samples are smaller than the markers, and
thus, they are not shown here. The latter is used to scale the values associated with the inverted magenta
triangles to the ODR results, denoted here by blue stars, and the error bars are obtained through error
propagation. The results from the IonQ Forte device (cyan stars), which were calculated after removing Q /
and a/ errors during post-processing, are also shown for comparison.

are shown in Fig. 15 at four different values of time, t ∈ {1, 4, 6, 11}. These sample values of t are
chosen to compare χn when return probability for Appx II shows relatively large deviation from
its ideal result (exhibiting 15% to 30% relative errors). As is observed, this local observable shows
very small deviation from the ideal result for both Appxs I and II, confirming that local observables
are more robust to small differences in initial states.

Appendix F: Pauli twirling and operator decoherence renormalization

In this appendix, we describe and implement a noise-mitigation strategy for obtaining the values
of E = ⟨σ̃ z ⟩ under the Trotter time evolution. The results for this quantity are obtained from runs
on the IonQ Forte quantum computer using Appx II, and are displayed in cyan triangles in Fig. 9
and below in Fig. 16. These results clearly show significant error in this quantity. Even at t = 1,
the value deviates from the ideal result, and it quickly diverges further away from the ideal values
during the evolution. On the other hand, as was seen in in Fig. 7, there exists reasonable agreement
between hardware and simulator results for this quantity at t = 0 (i.e., after the wave-packets’
preparation). We, therefore, conjuncture that the error in the time-evolved observable results from
the larger number of entangling gates applied to the boson qubit compared to fermionic qubits,
see Fig. 5(b).
The desired result could be recovered, nonetheless, by executing a few additional noise-
mitigating circuits, as shown in Fig. 16, and described below:
 39

Pauli twirling: Using the noisy emulator provided by IonQ for their Forte class devices,
we execute 3000 shots for each Trotter step. These are then split over 10 equivalent
“Pauli-twirled” circuits, and the final result is computed over the aggregated set of shots.
Each Pauli-twirled circuit is obtained by randomly picking one of the following equivalent
ϵ
operations for each CNOT gate in the e−iδtH block:9 CNOT, (σ z ⊗ I) ˜ CNOT (σ z ⊗ I), ˜
x ˜ x x x y y z
(σ ⊗ I) CNOT (σ ⊗ σ̃ ), and (σ ⊗ σ̃ ) CNOT (σ ⊗ σ̃ ). Here, tilde operators act on
the boson qubit. This process converts the coherent noise into a decoherent noise. As shown
in Fig. 16, Pauli-twirled results (inverted magenta triangles) improve the results and, in
particular, fix the shifted value at t = 1. This implies that the uniform offset for all time
steps likely is caused by coherent errors in the device.

Operator decoherence renormalization (ODR): Next, we apply a procedure called ODR,

which involves scaling the value of the observable E as EODR (t) = E(t)/(1 − ρ(t)). The
scaling factor ρ(t) is obtained by executing a circuit for each Trotter time t with a similar
structure as the target circuit but whose output is known. We obtain ρ(t) by executing the
Pauli-twirled circuits for the U (0) operator, and demand that the value of E remains the
same as its value E0 at t = 0 [shown in Fig. 7(a)]. The U (0) circuit can be executed by
setting δt → −δt in the last two circuit blocks in Fig. 5(a) and using ϵ = 0 for the middle
block. The U (0) circuits blocks are implemented nt times, where nt is the number of Trotter
steps up to time t. The value of E ′ (t) = ⟨σ̃ z (t)⟩ for these circuits are shown by the black plus
sign in Fig. 16. Note that since the electric-field value at the boson qubit does not change
significantly during the time evolution, E ′ (t) is nearly identical to E(t). The scaling factor
ρ(t) is now obtained from these results, and from E0 , via the relation ρ(t) = 1 − E ′ (t)/E0 .
We then apply the same scaling factor ρ(t) to obtain EODR out of E(t), the results of which
are shown as blue stars in Fig. 16. These results are in much better agreement with the ideal
values than unmitigated results, proving the effectiveness of these mitigation strategies in
this simulation.
This analysis is performed on the IonQ Forte emulator instead of the IonQ Forte quantum system
due to the limited hardware access. It would be interesting to see if this procedure also works for
the actual hardware data.

Appendix G: Tables of variational-quantum-eigensolver results

The VQE-optimization results are summarized in this appendix. The VQE was performed
using the noiseless Aer simulator. Table V contains the optimized values of the parameters that
characterize QGS for NP = 5 and NP = 13. Table VI and VII contain the optimization results for
the b†k ansatz with j = 1 for NP = 5 and NP = 13, respectively.

[1] A. Accardi et al., “Electron Ion Collider: The Next QCD Frontier: Understanding the glue that binds
us all,” Eur. Phys. J. A 52, 268 (2016), arXiv:1212.1701 [nucl-ex].
[2] Daniele P. Anderle et al., “Electron-ion collider in China,” Front. Phys. (Beijing) 16, 64701 (2021),
arXiv:2102.09222 [nucl-ex].
[3] Abhay Deshpande, Richard Milner, Raju Venugopalan, and Werner Vogelsang, “Study of the
fundamental structure of matter with an electron-ion collider,” Ann. Rev. Nucl. Part. Sci. 55, 165–228
(2005), arXiv:hep-ph/0506148.
ϵ
9
We chose these CNOT gates because the majority of entangling gates acting on the boson qubit belong to e−iδtH .
 40

∗
NP θh∗ θm∗ EGS EGS FGS

5 0.1705 0.7846 -8.8739 -8.8747 0.9998

13 0.1707 0.7855 -23.0727 -23.0743 0.9996

TABLE V. Shown are the values of θh∗ and θm∗ in Eq. (3) for NP = 5 and NP = 13, obtained from
minimizing the energy of the state QGS (θh , θm ) |Ω⟩SCV with respect to the Hamiltonian in Eq. (16). We
∗
employ the VQE method and simulate the circuit using a noiseless simulator. EGS is the energy of the
optimized state which agrees, up to sub-percent level, with the exact value EGS . The fidelity of the prepared
ground state is defined as FGS := |⟨Ω|QGS (θh∗ , θm∗ )|ΩSCV ⟩|2 , where |Ω⟩ is the target ground state. EGS and
|Ω⟩ for NP = 5 were calculated by diagonalizing the Hamiltonian, while for NP = 13, they were obtained by
finding the lowest-energy state using the DMRG method on an MPS ansatz for the state, see Appendix A
for details.

(1),k ∗ (1),k ∗
k α0 α1 Ek∗ Ek Fk

0 -0.0957 1.1112 -6.1048 -6.1260 0.9843

π
-3.2695 -3.0880 -6.0687 -6.0825 0.9894
5
π
− -1.0565 -1.0013 -6.0687 -6.0825 0.9893
5
2π
-1.7754 1.1020 -5.9780 -5.9800 0.9982
5
2π
− -1.5139 -1.4590 -5.9779 -5.9800 0.9981
5

(1),k ∗
TABLE VI. Shown are the optimized parameters α0/1 in Eq. (20) for Np = 5. Here, Ek∗ is the Hamiltonian
b†k
expectation value of the state |Ω⟩ using the optimized parameters, and Ek is the corresponding target
energy associated with the single-particle eigenstate |k⟩ with momentum k. The fidelity measure Fk is given
by Fk := |⟨k|b†k |Ω⟩|2 . Both Fk and Ek are computed by diagonalizing the Hamiltonian exactly.

[4] P. Achenbach et al., “The present and future of QCD,” Nucl. Phys. A 1047, 122874 (2024),
arXiv:2303.02579 [hep-ph].
[5] Meenakshi Narain et al., “The Future of US Particle Physics - The Snowmass 2021 Energy Frontier
Report,” (2022), arXiv:2211.11084 [hep-ex].
[6] O. Bruning, H. Burkhardt, and S. Myers, “The Large Hadron Collider,” Prog. Part. Nucl. Phys. 67,
705–734 (2012).
[7] Vladimir Shiltsev and Frank Zimmermann, “Modern and Future Colliders,” Rev. Mod. Phys. 93,
015006 (2021), arXiv:2003.09084 [physics.acc-ph].
[8] R. L. Workman et al. (Particle Data Group), “Review of Particle Physics,” PTEP 2022, 083C01
(2022).
[9] Sophie Kadan, “Searches for Supersymmetry (SUSY) at the Large Hadron Collider,” (2024),
arXiv:2404.16922 [hep-ex].
 41

(1),k ∗ (1),k ∗
k α0 α1 Ek′∗ Ek′ Fk′

0 -3.1579 -1.9231 -20.3043 -20.3255 0.9843

π
-1.9443 -1.7946 -20.2987
13
-20.3187 0.9852
π
− -1.9447 1.1538 -20.2986
13
2π
-3.1078 -2.5162 -20.2824
13
-20.2992 0.9876
2π
− -3.6310 -3.7909 -20.2826
13
3π
-1.2193 0.7456 -20.2567
13
-20.2688 0.9906
3π
− -1.7016 -1.7705 -20.2570
13
4π
-3.7347 -3.2568 -20.2236
13
-20.2306 0.9945
4π
− -2.7878 0.0941 -20.2236
13
5π
-3.8328 -3.8252 -20.1854
13
-20.1882 0.9978
5π
− -3.3866 1.8048 -20.1856
13
6π
1.0874 0.5975 -20.1431
13
-20.1452 0.9753
6π
− -1.2578 -2.7840 -20.1449
13

(1),k ∗
TABLE VII. Shown are the optimized parameters α0/1 in Eq. (20) for Np = 13. Here, Ek′∗ , Ek′ and Fk′
have the same definitions as their unprimed counterparts in Table VI, however, they were computed using
the DMRG method applied on an MPS ansatz for the states. Ek′ are the degenerate energy eigenvalues for
the single-particle eigenstates with k = ±|k|. The fidelity measure Fk′ is computed using the DMRG states
that are degenerate in energy, as described in Appendix A.

[10] B. Abi et al. (DUNE), “Prospects for beyond the Standard Model physics searches at the Deep
Underground Neutrino Experiment,” Eur. Phys. J. C 81, 322 (2021), arXiv:2008.12769 [hep-ex].
[11] S. Andringa et al. (SNO+), “Current Status and Future Prospects of the SNO+ Experiment,” Adv.
High Energy Phys. 2016, 6194250 (2016), arXiv:1508.05759 [physics.ins-det].
[12] Yasuo Takeuchi (Super-Kamiokande), “Recent oscillation results and future prospects of Super-
Kamiokande,” PoS NOW2022, 004 (2023).
[13] László P Csernai, Introduction to relativistic heavy ion collisions, Vol. 1 (Wiley New York, 1994).
 42

[14] T. Ludlam and S. Aronson (BRAHMS, STAR, PHOBOS, PHENIX), “Hunting the quark gluon
plasma,” (2005), 10.2172/15015225.
[15] Anton Andronic, Peter Braun-Munzinger, Krzysztof Redlich, and Johanna Stachel, “Decoding
the phase structure of QCD via particle production at high energy,” Nature 561, 321–330 (2018),
arXiv:1710.09425 [nucl-th].
[16] Jürgen Berges, Michal P. Heller, Aleksas Mazeliauskas, and Raju Venugopalan, “QCD thermalization:
Ab initio approaches and interdisciplinary connections,” Rev. Mod. Phys. 93, 035003 (2021),
arXiv:2005.12299 [hep-th].
[17] Alessandro Lovato et al., “Long Range Plan: Dense matter theory for heavy-ion collisions and neutron
stars,” (2022), arXiv:2211.02224 [nucl-th].
[18] Franz Gross et al., “50 Years of Quantum Chromodynamics,” Eur. Phys. J. C 83, 1125 (2023),
arXiv:2212.11107 [hep-ph].
[19] Bálint Joó, Chulwoo Jung, Norman H. Christ, William Detmold, Robert Edwards, Martin Savage, and
Phiala Shanahan (USQCD), “Status and Future Perspectives for Lattice Gauge Theory Calculations
to the Exascale and Beyond,” Eur. Phys. J. A 55, 199 (2019), arXiv:1904.09725 [hep-lat].
[20] Kenneth G. Wilson, “Confinement of Quarks,” Phys. Rev. D 10, 2445–2459 (1974).
[21] Y. Aoki et al. (Flavour Lattice Averaging Group (FLAG)), “FLAG Review 2024,” (2024),
arXiv:2411.04268 [hep-lat].
[22] Zohreh Davoudi et al., “Report of the Snowmass 2021 Topical Group on Lattice Gauge Theory,” in
Snowmass 2021 (2022) arXiv:2209.10758 [hep-lat].
[23] Andreas S. Kronfeld et al. (USQCD), “Lattice QCD and Particle Physics,” (2022), arXiv:2207.07641
[hep-lat].
[24] I. Montvay and G. Munster, Quantum fields on a lattice, Cambridge Monographs on Mathematical
Physics (Cambridge University Press, 1997).
[25] Heinz J. Rothe, Lattice Gauge Theories : An Introduction (Fourth Edition), Vol. 43 (World Scientific
Publishing Company, 2012).
[26] Christof Gattringer and Christian B. Lang, Quantum chromodynamics on the lattice, Vol. 788
(Springer, Berlin, 2010).
[27] Raúl A. Briceño, Zohreh Davoudi, and Thomas C. Luu, “Nuclear Reactions from Lattice QCD,” J.
Phys. G 42, 023101 (2015), arXiv:1406.5673 [hep-lat].
[28] Raul A. Briceno, Jozef J. Dudek, and Ross D. Young, “Scattering processes and resonances from
lattice QCD,” Rev. Mod. Phys. 90, 025001 (2018), arXiv:1706.06223 [hep-lat].
[29] Maxwell T. Hansen and Stephen R. Sharpe, “Lattice QCD and Three-particle Decays of Resonances,”
Ann. Rev. Nucl. Part. Sci. 69, 65–107 (2019), arXiv:1901.00483 [hep-lat].
[30] Zohreh Davoudi, William Detmold, Kostas Orginos, Assumpta Parreño, Martin J. Savage, Phiala
Shanahan, and Michael L. Wagman, “Nuclear matrix elements from lattice QCD for electroweak and
beyond-Standard-Model processes,” Phys. Rept. 900, 1–74 (2021), arXiv:2008.11160 [hep-lat].
[31] M. Luscher, “Volume Dependence of the Energy Spectrum in Massive Quantum Field Theories. 2.
Scattering States,” Commun. Math. Phys. 105, 153–188 (1986).
[32] Martin Luscher, “Two particle states on a torus and their relation to the scattering matrix,” Nucl.
Phys. B 354, 531–578 (1991).
[33] Christof Gattringer and Kurt Langfeld, “Approaches to the sign problem in lattice field theory,” Int.
J. Mod. Phys. A 31, 1643007 (2016), arXiv:1603.09517 [hep-lat].
[34] Keitaro Nagata, “Finite-density lattice QCD and sign problem: Current status and open problems,”
Prog. Part. Nucl. Phys. 127, 103991 (2022), arXiv:2108.12423 [hep-lat].
[35] John B. Kogut and Leonard Susskind, “Hamiltonian Formulation of Wilson’s Lattice Gauge Theories,”
Phys. Rev. D 11, 395–408 (1975).
[36] Mari Carmen Bañuls and Krzysztof Cichy, “Review on Novel Methods for Lattice Gauge Theories,”
Rept. Prog. Phys. 83, 024401 (2020), arXiv:1910.00257 [hep-lat].
[37] Yannick Meurice, Ryo Sakai, and Judah Unmuth-Yockey, “Tensor lattice field theory for
renormalization and quantum computing,” Rev. Mod. Phys. 94, 025005 (2022), arXiv:2010.06539
[hep-lat].
[38] Mari Carmen Bañuls, “Tensor Network Algorithms: A Route Map,” Ann. Rev. Condensed Matter
Phys. 14, 173–191 (2023), arXiv:2205.10345 [quant-ph].
 43

[39] J. Ignacio Cirac, David Perez-Garcia, Norbert Schuch, and Frank Verstraete, “Matrix product states
and projected entangled pair states: Concepts, symmetries, theorems,” Rev. Mod. Phys. 93, 045003
(2021), arXiv:2011.12127 [quant-ph].
[40] Giuseppe Magnifico, Timo Felser, Pietro Silvi, and Simone Montangero, “Lattice quantum
electrodynamics in (3+1)-dimensions at finite density with tensor networks,” Nature Commun. 12,
3600 (2021), arXiv:2011.10658 [hep-lat].
[41] Pietro Silvi, Enrique Rico, Tommaso Calarco, and Simone Montangero, “Lattice Gauge Tensor
Networks,” New J. Phys. 16, 103015 (2014), arXiv:1404.7439 [quant-ph].
[42] Luca Tagliacozzo, Alessio Celi, and Maciej Lewenstein, “Tensor Networks for Lattice Gauge Theories
with continuous groups,” Phys. Rev. X 4, 041024 (2014), arXiv:1405.4811 [cond-mat.str-el].
[43] E. Rico, T. Pichler, M. Dalmonte, P. Zoller, and S. Montangero, “Tensor networks for Lattice Gauge
Theories and Atomic Quantum Simulation,” Phys. Rev. Lett. 112, 201601 (2014), arXiv:1312.3127
[cond-mat.quant-gas].
[44] Boye Buyens, Jutho Haegeman, Henri Verschelde, Frank Verstraete, and Karel Van Acoleyen,
“Confinement and string breaking for QED2 in the Hamiltonian picture,” Phys. Rev. X 6, 041040
(2016), arXiv:1509.00246 [hep-lat].
[45] T. Pichler, M. Dalmonte, E. Rico, P. Zoller, and S. Montangero, “Real-time Dynamics in U(1)
Lattice Gauge Theories with Tensor Networks,” Phys. Rev. X 6, 011023 (2016), arXiv:1505.04440
[cond-mat.quant-gas].
[46] Boye Buyens, Jutho Haegeman, Florian Hebenstreit, Frank Verstraete, and Karel Van Acoleyen,
“Real-time simulation of the Schwinger effect with Matrix Product States,” Phys. Rev. D 96, 114501
(2017), arXiv:1612.00739 [hep-lat].
[47] Stefan Kühn, Erez Zohar, J. Ignacio Cirac, and Mari Carmen Bañuls, “Non-Abelian string breaking
phenomena with Matrix Product States,” JHEP 07, 130 (2015), arXiv:1505.04441 [hep-lat].
[48] Emil Mathew, Navya Gupta, Saurabh V. Kadam, Aniruddha Bapat, Jesse Stryker, Zohreh Davoudi,
and Indrakshi Raychowdhury, “Tensor-network toolbox for probing dynamics of non-Abelian gauge
theories,” PoS LATTICE2024, 472 (2025), arXiv:2501.18301 [hep-lat].
[49] Mari Carmen Bañuls, Krzysztof Cichy, C. J. David Lin, and Manuel Schneider, “Parton Distribution
Functions in the Schwinger model from Tensor Network States,” (2025), arXiv:2504.07508 [hep-lat].
[50] Marco Rigobello, Simone Notarnicola, Giuseppe Magnifico, and Simone Montangero, “Entanglement
generation in (1+1)D QED scattering processes,” Phys. Rev. D 104, 114501 (2021), arXiv:2105.03445
[hep-lat].
[51] Ron Belyansky, Seth Whitsitt, Niklas Mueller, Ali Fahimniya, Elizabeth R. Bennewitz, Zohreh
Davoudi, and Alexey 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].
[52] Irene Papaefstathiou, Johannes Knolle, and Mari Carmen Bañuls, “Real-time scattering in the lattice
Schwinger model,” (2024), arXiv:2402.18429 [hep-lat].
[53] Raghav G. Jha, Ashley Milsted, Dominik Neuenfeld, John Preskill, and Pedro Vieira, “Real-Time
Scattering in Ising Field Theory using Matrix Product States,” (2024), arXiv:2411.13645 [hep-th].
[54] Giuseppe Magnifico, Giovanni Cataldi, Marco Rigobello, Peter Majcen, Daniel Jaschke, Pietro Silvi,
and Simone Montangero, “Tensor Networks for Lattice Gauge Theories beyond one dimension: a
Roadmap,” (2024), arXiv:2407.03058 [hep-lat].
[55] Ariel Kelman, Umberto Borla, Itay Gomelski, Jonathan Elyovich, Gertian Roose, Patrick Emonts, and
Erez Zohar, “Gauged Gaussian projected entangled pair states: A high dimensional tensor network
formulation for lattice gauge theories,” Phys. Rev. D 110, 054511 (2024), arXiv:2404.13123 [hep-lat].
[56] Atis Yosprakob, “Reduced Tensor Network Formulation for Non-Abelian Gauge Theories in Arbitrary
Dimensions,” PTEP 2024, 073B05 (2024), arXiv:2311.02541 [hep-th].
[57] Richard P. Feynman, “Simulating physics with computers,” Int. J. Theor. Phys. 21, 467–488 (1982).
[58] Daniel R. Simon, “On the Power of Quantum Computation,” SIAM J. Comput. 26, 1474–1483 (2006).
[59] Seth Lloyd, “Universal Quantum Simulators,” Science 273, 1073 (1996).
[60] John Preskill, “Quantum computing 40 years later,” (2021), arXiv:2106.10522 [quant-ph].
[61] Michael A. Nielsen and Isaac L. Chuang, Quantum Computation and Quantum Information
(Cambridge University Press, 2012).
 44

[62] Yulin Chi et al., “A Programmable Qudit-based Quantum Processor,” in Conference on Lasers and
Electro-Optics (2023).
[63] Anastasiia S. Nikolaeva, Evgeniy O. Kiktenko, and Aleksey K. Fedorov, “Universal quantum
computing with qubits embedded in trapped-ion qudits,” Phys. Rev. A 109, 022615 (2024),
arXiv:2302.02966 [quant-ph].
[64] Sergey B. Bravyi and Alexei Yu. Kitaev, “Fermionic Quantum Computation,” Annals Phys. 298,
210–226 (2002), arXiv:quant-ph/0003137.
[65] Ulysse Chabaud and Mattia Walschaers, “Resources for Bosonic Quantum Computational Advantage,”
Phys. Rev. Lett. 130, 090602 (2023), arXiv:2207.11781 [quant-ph].
[66] M. Dalmonte and S. Montangero, “Lattice gauge theory simulations in the quantum information era,”
Contemp. Phys. 57, 388–412 (2016), arXiv:1602.03776 [cond-mat.quant-gas].
[67] M. C. Bañuls et al., “Simulating Lattice Gauge Theories within Quantum Technologies,” Eur. Phys.
J. D 74, 165 (2020), arXiv:1911.00003 [quant-ph].
[68] Natalie Klco, Alessandro Roggero, and Martin J. Savage, “Standard model physics and the
digital quantum revolution: thoughts about the interface,” Rept. Prog. Phys. 85, 064301 (2022),
arXiv:2107.04769 [quant-ph].
[69] Christian W Bauer, Zohreh Davoudi, A Baha Balantekin, Tanmoy Bhattacharya, Marcela Carena,
Wibe A De Jong, Patrick Draper, Aida El-Khadra, Nate Gemelke, Masanori Hanada, et al., “Quantum
Simulation for High-Energy Physics,” PRX Quantum 4, 027001 (2023), arXiv:2204.03381 [quant-ph].
[70] Christian W. Bauer, Zohreh Davoudi, Natalie Klco, and Martin J. Savage, “Quantum simulation of
fundamental particles and forces,” Nature Rev. Phys. 5, 420–432 (2023), arXiv:2404.06298 [hep-ph].
[71] Jad C. Halimeh, Monika Aidelsburger, Fabian Grusdt, Philipp Hauke, and Bing Yang, “Cold-atom
quantum simulators of gauge theories,” (2023), arXiv:2310.12201 [cond-mat.quant-gas].
[72] Alberto Di Meglio et al., “Quantum Computing for High-Energy Physics: State of the Art and
Challenges,” PRX Quantum 5, 037001 (2024), arXiv:2307.03236 [quant-ph].
[73] Douglas Beck et al., “Quantum Information Science and Technology for Nuclear Physics. Input into
U.S. Long-Range Planning, 2023,” (2023) arXiv:2303.00113 [nucl-ex].
[74] Stephen P. Jordan, Keith S. M. Lee, and John Preskill, “Quantum Algorithms for Quantum Field
Theories,” Science 336, 1130–1133 (2012), arXiv:1111.3633 [quant-ph].
[75] Stephen P. Jordan, Keith S. M. Lee, and John Preskill, “Quantum Computation of Scattering in Scalar
Quantum Field Theories,” Quant. Inf. Comput. 14, 1014–1080 (2014), arXiv:1112.4833 [hep-th].
[76] Edward Farhi, Jeffrey Goldstone, Sam Gutmann, and Michael Sipser, “Quantum Computation by
Adiabatic Evolution,” (2000), arXiv:quant-ph/0001106.
[77] Federica Maria Surace et al., “String-Breaking Dynamics in Quantum Adiabatic and Diabatic
Processes,” (2024), arXiv:2411.10652 [quant-ph].
[78] João Barata, Niklas Mueller, Andrey Tarasov, and Raju Venugopalan, “Single-particle digitization
strategy for quantum computation of a ϕ4 scalar field theory,” Phys. Rev. A 103, 042410 (2021),
arXiv:2012.00020 [hep-th].
[79] Thomas D. Cohen and Hyunwoo Oh, “Efficient vacuum-state preparation for quantum simulation
of strongly interacting local quantum field theories,” Phys. Rev. A 109, L020402 (2024),
arXiv:2310.19229 [hep-lat].
[80] N. Klco, E. F. Dumitrescu, A. J. McCaskey, T. D. Morris, R. C. Pooser, M. Sanz, E. Solano,
P. Lougovski, and M. J. Savage, “Quantum-classical computation of Schwinger model dynamics
using quantum computers,” Phys. Rev. A 98, 032331 (2018), arXiv:1803.03326 [quant-ph].
[81] Yasar Y. Atas, Jinglei Zhang, Randy Lewis, Amin Jahanpour, Jan F. Haase, and Christine A.
Muschik, “SU(2) hadrons on a quantum computer via a variational approach,” Nature Commun. 12,
6499 (2021), arXiv:2102.08920 [quant-ph].
[82] Michael Fromm, Owe Philipsen, Michael Spannowsky, and Christopher Winterowd, “Simulating Z2
lattice gauge theory with the variational quantum thermalizer,” EPJ Quant. Technol. 11, 20 (2024),
arXiv:2306.06057 [hep-lat].
[83] Arianna Crippa, Karl Jansen, and Enrico Rinaldi, “Analysis of the confinement string in
(2+1)-dimensional Quantum Electrodynamics with a trapped-ion quantum computer,” (2024),
arXiv:2411.05628 [hep-lat].
[84] Bipasha Chakraborty, Masazumi Honda, Taku Izubuchi, Yuta Kikuchi, and Akio Tomiya, “Classically
emulated digital quantum simulation of the Schwinger model with a topological term via adiabatic
 45

state preparation,” Phys. Rev. D 105, 094503 (2022), arXiv:2001.00485 [hep-lat].

[85] Matteo D’Anna, Marina Krstic Marinkovic, and Joao C. Pinto Barros, “Adiabatic state preparation
for digital quantum simulations of QED in 1+1D,” (2024), arXiv:2411.01079 [hep-lat].
[86] Roland C. Farrell, Marc Illa, Anthony N. Ciavarella, and Martin 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].
[87] Roland C. Farrell, Marc Illa, Anthony N. Ciavarella, and Martin 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].
[88] Zhiyao Li, Dorota M. Grabowska, and Martin J. Savage, “Sequency Hierarchy Truncation (SeqHT) for
Adiabatic State Preparation and Time Evolution in Quantum Simulations,” (2024), arXiv:2407.13835
[quant-ph].
[89] James Ingoldby, Michael Spannowsky, Timur Sypchenko, Simon Williams, and Matthew
Wingate, “Real-Time Scattering on Quantum Computers via Hamiltonian Truncation,” (2025),
arXiv:2505.03878 [quant-ph].
[90] Nikita A. Zemlevskiy, “Scalable Quantum Simulations of Scattering in Scalar Field Theory on 120
Qubits,” (2024), arXiv:2411.02486 [quant-ph].
[91] Roland C. Farrell, Nikita A. Zemlevskiy, Marc Illa, and John Preskill, “Digital quantum simulations
of scattering in quantum field theories using W states,” (2025), arXiv:2505.03111 [quant-ph].
[92] Junyu Liu, Zimu Li, Han Zheng, Xiao Yuan, and Jinzhao Sun, “Towards a variational
Jordan–Lee–Preskill quantum algorithm,” Mach. Learn. Sci. Tech. 3, 045030 (2022), arXiv:2109.05547
[quant-ph].
[93] Matteo Turco, Gonçalo M. Quinta, João Seixas, and Yasser Omar, “Quantum Simulation of Bound
State Scattering,” PRX Quantum 5, 020311 (2024), arXiv:2305.07692 [quant-ph].
[94] Matteo Turco, Gonçalo M. Quinta, João Seixas, and Yasser Omar, “Creation of Wavepackets for
Quantum Chromodynamics on Quantum Computers,” (2025), arXiv:2501.13069 [quant-ph].
[95] Michael Kreshchuk, James P. Vary, and Peter J. Love, “Simulating Scattering of Composite Particles,”
(2023), arXiv:2310.13742 [quant-ph].
[96] Michael Hite, “Simplified Fermionic Scattering State Preparation for the NISQ Era,” (2025),
arXiv:2505.00476 [quant-ph].
[97] Tianyin Li, Wai Kin Lai, Enke Wang, and Hongxi Xing (QuNu), “Scattering amplitude from quantum
computing with reduction formula,” Phys. Rev. D 109, 036025 (2024), arXiv:2301.04179 [hep-ph].
[98] Raúl A. Briceño, Robert G. Edwards, Miller Eaton, Carlos González-Arciniegas, Olivier Pfister, and
George Siopsis, “Toward coherent quantum computation of scattering amplitudes with a measurement-
based photonic quantum processor,” (2023), arXiv:2312.12613 [quant-ph].
[99] Raúl A. Briceño, Juan V. Guerrero, Maxwell T. Hansen, and Alexandru M. Sturzu, “Role of boundary
conditions in quantum computations of scattering observables,” Phys. Rev. D 103, 014506 (2021),
arXiv:2007.01155 [hep-lat].
[100] Erik Gustafson, Yingyue Zhu, Patrick Dreher, Norbert M. Linke, and Yannick Meurice, “Real-time
quantum calculations of phase shifts using wave packet time delays,” Phys. Rev. D 104, 054507 (2021),
arXiv:2103.06848 [hep-lat].
[101] Alberto Peruzzo, Jarrod McClean, Peter Shadbolt, Man-Hong Yung, Xiao-Qi Zhou, Peter J. Love,
Alán Aspuru-Guzik, and Jeremy L. O’Brien, “A variational eigenvalue solver on a photonic quantum
processor,” Nature Commun. 5, 4213 (2014), arXiv:1304.3061 [quant-ph].
[102] Jarrod R. McClean, Jonathan Romero, Ryan Babbush, and Alán Aspuru-Guzik, “The theory of
variational hybrid quantum-classical algorithms,” New J. Phys. 18, 023023 (2016), arXiv:1509.04279
[quant-ph].
[103] Jules Tilly et al., “The Variational Quantum Eigensolver: A review of methods and best practices,”
Phys. Rept. 986, 1–128 (2022), arXiv:2111.05176 [quant-ph].
[104] Sanket Sharma, T. Papenbrock, and L. Platter, “Scattering phase shifts from a quantum computer,”
Phys. Rev. C 109, L061001 (2024), arXiv:2311.09298 [nucl-th].
[105] Anthony Ciavarella, “Algorithm for quantum computation of particle decays,” Phys. Rev. D 102,
094505 (2020), arXiv:2007.04447 [hep-th].
[106] Yahui Chai, Arianna Crippa, Karl Jansen, Stefan Kühn, Vincent R. Pascuzzi, Francesco Tacchino,
and Ivano Tavernelli, “Entanglement production from scattering of fermionic wave packets: a quantum
 46

computing approach,” (2023), arXiv:2312.02272 [quant-ph].

[107] Zohreh Davoudi, Chung-Chun Hsieh, and Saurabh V. Kadam, “Scattering wave packets of hadrons
in gauge theories: Preparation on a quantum computer,” Quantum 8, 1520 (2024), arXiv:2402.00840
[quant-ph].
[108] G. Magnifico, D. Vodola, E. Ercolessi, S. P. Kumar, M. Müller, and A. Bermudez, “ZN gauge theories
coupled to topological fermions: QED2 with a quantum-mechanical θ angle,” Phys. Rev. B 100, 115152
(2019), arXiv:1906.07005 [cond-mat.quant-gas].
[109] Clement Charles, Erik J. Gustafson, Elizabeth Hardt, Florian Herren, Norman Hogan, Henry Lamm,
Sara Starecheski, Ruth S. Van de Water, and Michael L. Wagman, “Simulating Z2 lattice gauge
theory on a quantum computer,” Phys. Rev. E 109, 015307 (2024), arXiv:2305.02361 [hep-lat].
[110] Julius Mildenberger, Wojciech Mruczkiewicz, Jad C. Halimeh, Zhang Jiang, and Philipp Hauke,
“Probing confinement in a Z2 lattice gauge theory on a quantum computer,” (2022), arXiv:2203.08905
[quant-ph].
[111] Reinis Irmejs, Mari Carmen Banuls, and J. Ignacio Cirac, “Quantum simulation of Z2 lattice gauge
theory with minimal resources,” Phys. Rev. D 108, 074503 (2023), arXiv:2206.08909 [quant-ph].
[112] Jernej Frank, Emilie Huffman, and Shailesh Chandrasekharan, “Emergence of Gauss’ law in a Z2
lattice gauge theory in 1+1 dimensions,” Phys. Lett. B 806, 135484 (2020), arXiv:1904.05414 [cond-
mat.str-el].
[113] Umberto Borla, Ruben Verresen, Fabian Grusdt, and Sergej Moroz, “Confined Phases of One-
Dimensional Spinless Fermions Coupled to Z2 Gauge Theory,” Phys. Rev. Lett. 124, 120503 (2020),
arXiv:1909.07399 [cond-mat.str-el].
[114] Umberto Borla, Bhilahari Jeevanesan, Frank Pollmann, and Sergej Moroz, “Quantum phases of two-
dimensional Z2 gauge theory coupled to single-component fermion matter,” Phys. Rev. B 105, 075132
(2022), arXiv:2012.08543 [cond-mat.str-el].
[115] Aritra Das, Umberto Borla, and Sergej Moroz, “Fractionalized holes in one-dimensional Z2 gauge
theory coupled to fermion matter: Deconfined dynamics and emergent integrability,” Phys. Rev. B
107, 064302 (2023), arXiv:2111.13205 [cond-mat.str-el].
[116] Rhine Samajdar, Darshan G. Joshi, Yanting Teng, and Subir Sachdev, “Emergent Z2 Gauge
Theories and Topological Excitations in Rydberg Atom Arrays,” Phys. Rev. Lett. 130, 043601 (2023),
arXiv:2204.00632 [cond-mat.quant-gas].
[117] Elio J. König, Piers Coleman, and Alexei M. Tsvelik, “Soluble limit and criticality of fermions in Z2
gauge theories,” Phys. Rev. B 102, 155143 (2020), arXiv:1912.11106 [cond-mat.str-el].
[118] H. Yarloo, M. Mohseni-Rajaee, and A. Langari, “Emergent statistical bubble localization in a Z2
lattice gauge theory,” Phys. Rev. B 99, 054403 (2019), arXiv:1810.08434 [cond-mat.str-el].
[119] Jwo-Sy Chen et al., “Benchmarking a trapped-ion quantum computer with 30 qubits,” Quantum 8,
1516 (2024), arXiv:2308.05071 [quant-ph].
[120] Guang Hao Low and Isaac L. Chuang, “Hamiltonian Simulation by Qubitization,” Quantum 3, 163
(2019), arXiv:1610.06546 [quant-ph].
[121] William M. Kirby, Sultana Hadi, Michael Kreshchuk, and Peter J. Love, “Quantum simulation
of second-quantized Hamiltonians in compact encoding,” Phys. Rev. A 104, 042607 (2021),
arXiv:2105.10941 [quant-ph].
[122] Yi-Hsiang Chen, Amir Kalev, and Itay Hen, “Quantum Algorithm for Time-Dependent Hamiltonian
Simulation by Permutation Expansion,” PRX Quantum 2, 030342 (2021).
[123] Amir Kalev and Itay Hen, “Quantum Algorithm for Simulating Hamiltonian Dynamics with an Off-
diagonal Series Expansion,” Quantum 5, 426 (2021), arXiv:2006.02539 [quant-ph].
[124] Guang Hao Low and Nathan Wiebe, “Hamiltonian Simulation in the Interaction Picture,” (2018),
arXiv:1805.00675 [quant-ph].
[125] Wren A. Yamada, Osamu Morimatsu, Toru Sato, and Koichi Yazaki, “Survival probability of unstable
states in coupled-channels: Nonexponential decay of threshold cusp,” Phys. Rev. D 108, L071502
(2023), arXiv:2305.11695 [hep-ph].
[126] F.M. Izrailev and A. Castañeda-Mendoza, “Return probability: Exponential versus gaussian decay,”
Physics Letters A 350, 355–362 (2006).
[127] Tom Banks, Leonard Susskind, and John B. Kogut, “Strong Coupling Calculations of Lattice Gauge
Theories: (1+1)-Dimensional Exercises,” Phys. Rev. D 13, 1043 (1976).
 47

[128] Chris Nagele, J. Eduardo Cejudo, Tim Byrnes, and Matthew Kleban, “Flux unwinding in the lattice
Schwinger model,” Phys. Rev. D 99, 094501 (2019), arXiv:1811.03096 [hep-th].
[129] Torsten Victor Zache, Quantum simulation of high-energy physics with ultracold atoms., Ph.D. thesis,
U. Heidelberg (main) (2020).
[130] T. V. Zache, N. Mueller, J. T. Schneider, F. Jendrzejewski, J. Berges, and P. Hauke, “Dynamical
Topological Transitions in the Massive Schwinger Model with a θ Term,” Phys. Rev. Lett. 122, 050403
(2019), arXiv:1808.07885 [quant-ph].
[131] Luca Lumia, Pietro Torta, Glen B. Mbeng, Giuseppe E. Santoro, Elisa Ercolessi, Michele Burrello,
and Matteo M. Wauters, “Two-Dimensional Z2 Lattice Gauge Theory on a Near-Term Quantum
Simulator: Variational Quantum Optimization, Confinement, and Topological Order,” PRX Quantum
3, 020320 (2022), arXiv:2112.11787 [quant-ph].
[132] Steven R. White, “Density matrix formulation for quantum renormalization groups,” Phys. Rev. Lett.
69, 2863–2866 (1992).
[133] David Perez-Garcia, Frank Verstraete, Michael M. Wolf, and J. Ignacio Cirac, “Matrix product state
representations,” Quant. Inf. Comput. 7, 401–430 (2007), arXiv:quant-ph/0608197.
[134] Rahul Trivedi, Adrian Franco Rubio, and J. Ignacio Cirac, “Quantum advantage and stability to errors
in analogue quantum simulators,” Nature Commun. 15, 6507 (2024), arXiv:2212.04924 [quant-ph].
[135] Vikram Kashyap, Georgios Styliaris, Sara Mouradian, Juan Ignacio Cirac, and Rahul Trivedi,
“Accuracy Guarantees and Quantum Advantage in Analog Open Quantum Simulation with and
without Noise,” Phys. Rev. X 15, 021017 (2025), arXiv:2404.11081 [quant-ph].
[136] Nhung H. Nguyen, Minh C. Tran, Yingyue Zhu, Alaina M. Green, C. Huerta Alderete, Zohreh Davoudi,
and Norbert M. Linke, “Digital Quantum Simulation of the Schwinger Model and Symmetry Protection
with Trapped Ions,” PRX Quantum 3, 020324 (2022), arXiv:2112.14262 [quant-ph].
[137] Joel J. Wallman and Joseph Emerson, “Noise tailoring for scalable quantum computation via
randomized compiling,” Phys. Rev. A 94, 052325 (2016), arXiv:1512.01098 [quant-ph].
[138] Miroslav Urbanek, Benjamin Nachman, Vincent R. Pascuzzi, Andre He, Christian W. Bauer, and
Wibe A. de Jong, “Mitigating depolarizing noise on quantum computers with noise-estimation
circuits,” Phys. Rev. Lett. 127, 270502 (2021), arXiv:2103.08591 [quant-ph].
[139] Sarmed A Rahman, Randy Lewis, Emanuele Mendicelli, and Sarah 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].
[140] Jutho Haegeman, J. Ignacio Cirac, Tobias J. Osborne, Iztok Pizorn, Henri Verschelde, and Frank
Verstraete, “Time-Dependent Variational Principle for Quantum Lattices,” Phys. Rev. Lett. 107,
070601 (2011), arXiv:1103.0936 [cond-mat.str-el].
[141] Jutho Haegeman, Christian Lubich, Ivan Oseledets, Bart Vandereycken, and Frank Verstraete,
“Unifying time evolution and optimization with matrix product states,” Phys. Rev. B 94, 165116
(2016).
[142] Mingru Yang and Steven R. White, “Time-dependent variational principle with ancillary krylov
subspace,” Physical Review B 102 (2020), 10.1103/physrevb.102.094315.
[143] Tim Byrnes and Yoshihisa Yamamoto, “Simulating lattice gauge theories on a quantum computer,”
Phys. Rev. A 73, 022328 (2006), arXiv:quant-ph/0510027.
[144] Alexander F. Shaw, Pavel Lougovski, Jesse R. Stryker, and Nathan Wiebe, “Quantum Algorithms
for Simulating the Lattice Schwinger Model,” Quantum 4, 306 (2020), arXiv:2002.11146 [quant-ph].
[145] Anthony Ciavarella, Natalie Klco, and Martin J. Savage, “Trailhead for quantum simulation of
SU(3) Yang-Mills lattice gauge theory in the local multiplet basis,” Phys. Rev. D 103, 094501 (2021),
arXiv:2101.10227 [quant-ph].
[146] Angus Kan and Yunseong Nam, “Lattice Quantum Chromodynamics and Electrodynamics on a
Universal Quantum Computer,” (2021), arXiv:2107.12769 [quant-ph].
[147] Henry Lamm, Scott Lawrence, and Yukari Yamauchi (NuQS), “General Methods for Digital Quantum
Simulation of Gauge Theories,” Phys. Rev. D 100, 034518 (2019), arXiv:1903.08807 [hep-lat].
[148] Jan F. Haase, Luca Dellantonio, Alessio Celi, Danny Paulson, Angus Kan, Karl Jansen, and
Christine A. Muschik, “A resource efficient approach for quantum and classical simulations of gauge
theories in particle physics,” Quantum 5, 393 (2021), arXiv:2006.14160 [quant-ph].
[149] Zohreh Davoudi, Alexander F. Shaw, and Jesse R. Stryker, “General quantum algorithms for
Hamiltonian simulation with applications to a non-Abelian lattice gauge theory,” Quantum 7, 1213
 48

(2023), arXiv:2212.14030 [hep-lat].

[150] Edison M. Murairi, Michael J. Cervia, Hersh Kumar, Paulo F. Bedaque, and Andrei Alexandru, “How
many quantum gates do gauge theories require?” Phys. Rev. D 106, 094504 (2022), arXiv:2208.11789
[hep-lat].
[151] Mason L. Rhodes, Michael Kreshchuk, and Shivesh Pathak, “Exponential Improvements in the
Simulation of Lattice Gauge Theories Using Near-Optimal Techniques,” PRX Quantum 5, 040347
(2024), arXiv:2405.10416 [quant-ph].
[152] Henry Lamm, Ying-Ying Li, Jing Shu, Yi-Lin Wang, and Bin Xu, “Block encodings of discrete
subgroups on a quantum computer,” Phys. Rev. D 110, 054505 (2024), arXiv:2405.12890 [hep-lat].
[153] Praveen Balaji, Cianán Conefrey-Shinozaki, Patrick Draper, Jason K. Elhaderi, Drishti Gupta, Luis
Hidalgo, Andrew Lytle, and Enrico Rinaldi, “Quantum Circuits for SU(3) Lattice Gauge Theory,”
(2025), arXiv:2503.08866 [hep-lat].
[154] Mari Carmen Bañuls, Krzysztof Cichy, J. Ignacio Cirac, Karl Jansen, and Stefan Kühn, “Efficient
basis formulation for 1+1 dimensional SU(2) lattice gauge theory: Spectral calculations with matrix
product states,” Phys. Rev. X 7, 041046 (2017), arXiv:1707.06434 [hep-lat].
[155] Indrakshi Raychowdhury and Jesse R. Stryker, “Loop, string, and hadron dynamics in SU(2)
Hamiltonian lattice gauge theories,” Phys. Rev. D 101, 114502 (2020), arXiv:1912.06133 [hep-lat].
[156] Saurabh V. Kadam, Indrakshi Raychowdhury, and Jesse R. Stryker, “Loop-string-hadron formulation
of an SU(3) gauge theory with dynamical quarks,” Phys. Rev. D 107, 094513 (2023), arXiv:2212.04490
[hep-lat].
[157] Saurabh V. Kadam, Aahiri Naskar, Indrakshi Raychowdhury, and Jesse R. Stryker, “Loop-string-
hadron approach to SU(3) lattice Yang-Mills theory: Gauge invariant Hilbert space of a trivalent
vertex,” (2024), arXiv:2407.19181 [hep-lat].
[158] Guy Pardo, Tomer Greenberg, Aryeh Fortinsky, Nadav Katz, and Erez Zohar, “Resource-efficient
quantum simulation of lattice gauge theories in arbitrary dimensions: Solving for Gauss’s law and
fermion elimination,” Phys. Rev. Res. 5, 023077 (2023), arXiv:2206.00685 [quant-ph].
[159] Zohreh Davoudi, Indrakshi Raychowdhury, and Andrew Shaw, “Search for efficient formulations for
Hamiltonian simulation of non-Abelian lattice gauge theories,” Phys. Rev. D 104, 074505 (2021),
arXiv:2009.11802 [hep-lat].
[160] S. Chandrasekharan and U. J. Wiese, “Quantum link models: A Discrete approach to gauge theories,”
Nucl. Phys. B 492, 455–474 (1997), arXiv:hep-lat/9609042.
[161] Uwe-Jens Wiese, “From quantum link models to D-theory: a resource efficient framework for the
quantum simulation and computation of gauge theories,” Phil. Trans. A. Math. Phys. Eng. Sci. 380,
20210068 (2021), arXiv:2107.09335 [hep-lat].
[162] Erez Zohar and Michele Burrello, “Formulation of lattice gauge theories for quantum simulations,”
Phys. Rev. D 91, 054506 (2015), arXiv:1409.3085 [quant-ph].
[163] Erez Zohar and J. Ignacio Cirac, “Eliminating fermionic matter fields in lattice gauge theories,” Phys.
Rev. B 98, 075119 (2018), arXiv:1805.05347 [quant-ph].
[164] Erez Zohar and J. Ignacio Cirac, “Removing Staggered Fermionic Matter in U (N ) and SU (N ) Lattice
Gauge Theories,” Phys. Rev. D 99, 114511 (2019), arXiv:1905.00652 [quant-ph].
[165] Andrei Alexandru, Paulo F. Bedaque, Andrea Carosso, Michael J. Cervia, Edison M. Murairi, and
Andy Sheng, “Fuzzy gauge theory for quantum computers,” Phys. Rev. D 109, 094502 (2024),
arXiv:2308.05253 [hep-lat].
[166] Andrei Alexandru, Paulo F. Bedaque, Siddhartha Harmalkar, Henry Lamm, Scott Lawrence, and
Neill C. Warrington (NuQS), “Gluon Field Digitization for Quantum Computers,” Phys. Rev. D 100,
114501 (2019), arXiv:1906.11213 [hep-lat].
[167] Yao Ji, Henry Lamm, and Shuchen Zhu (NuQS), “Gluon digitization via character expansion for
quantum computers,” Phys. Rev. D 107, 114503 (2023), arXiv:2203.02330 [hep-lat].
[168] Yao Ji, Henry Lamm, and Shuchen Zhu (NuQS), “Gluon Field Digitization via Group Space
Decimation for Quantum Computers,” Phys. Rev. D 102, 114513 (2020), arXiv:2005.14221 [hep-lat].
[169] Ali H. Z. Kavaki and Randy Lewis, “From square plaquettes to triamond lattices for SU(2) gauge
theory,” Commun. Phys. 7, 208 (2024), arXiv:2401.14570 [hep-lat].
[170] Marc Illa, Martin J. Savage, and Xiaojun Yao, “Improved Honeycomb and Hyper-Honeycomb Lattice
Hamiltonians for Quantum Simulations of Non-Abelian Gauge Theories,” (2025), arXiv:2503.09688
[hep-lat].
 49

[171] Irian D’Andrea, Christian W. Bauer, Dorota M. Grabowska, and Marat Freytsis, “New basis for
Hamiltonian SU(2) simulations,” Phys. Rev. D 109, 074501 (2024), arXiv:2307.11829 [hep-ph].
[172] Torsten V. Zache, Daniel González-Cuadra, and Peter Zoller, “Quantum and Classical Spin-Network
Algorithms for q-Deformed Kogut-Susskind Gauge Theories,” Phys. Rev. Lett. 131, 171902 (2023),
arXiv:2304.02527 [quant-ph].
[173] Tobias Hartung, Timo Jakobs, Karl Jansen, Johann Ostmeyer, and Carsten Urbach, “Digitising SU(2)
gauge fields and the freezing transition,” Eur. Phys. J. C 82, 237 (2022), arXiv:2201.09625 [hep-lat].
[174] Simone Romiti and Carsten Urbach, “Digitizing lattice gauge theories in the magnetic basis:
reducing the breaking of the fundamental commutation relations,” Eur. Phys. J. C 84, 708 (2024),
arXiv:2311.11928 [hep-lat].
[175] Pierpaolo Fontana, Marc Miranda Riaza, and Alessio Celi, “An efficient finite-resource formulation
of non-Abelian lattice gauge theories beyond one dimension,” (2024), arXiv:2409.04441 [quant-ph].
[176] Christian W. Bauer and Dorota 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].
[177] Dorota M. Grabowska, Christopher F. Kane, and Christian W. Bauer, “A Fully Gauge-Fixed SU(2)
Hamiltonian for Quantum Simulations,” (2024), arXiv:2409.10610 [quant-ph].
[178] I. M. Burbano and Christian W. Bauer, “Gauge Loop-String-Hadron Formulation on General Graphs
and Applications to Fully Gauge Fixed Hamiltonian Lattice Gauge Theory,” (2024), arXiv:2409.13812
[hep-lat].
[179] Anthony N. Ciavarella, I. M. Burbano, and Christian W. Bauer, “Efficient Truncations of SU(Nc )
Lattice Gauge Theory for Quantum Simulation,” (2025), arXiv:2503.11888 [hep-lat].
[180] Christian W. Bauer, “Efficient use of quantum computers for collider physics,” (2025),
arXiv:2503.16602 [hep-ph].
[181] Henry Lamm, Scott Lawrence, and Yukari Yamauchi (NuQS), “Parton physics on a quantum
computer,” Phys. Rev. Res. 2, 013272 (2020), arXiv:1908.10439 [hep-lat].
[182] M. G. Echevarria, I. L. Egusquiza, E. Rico, and G. Schnell, “Quantum simulation of light-front parton
correlators,” Phys. Rev. D 104, 014512 (2021), arXiv:2011.01275 [quant-ph].
[183] Tianyin Li, Xingyu Guo, Wai Kin Lai, Xiaohui Liu, Enke Wang, Hongxi Xing, Dan-Bo Zhang, and
Shi-Liang Zhu (QuNu), “Partonic collinear structure by quantum computing,” Phys. Rev. D 105,
L111502 (2022), arXiv:2106.03865 [hep-ph].
[184] Niklas Mueller, Andrey Tarasov, and Raju Venugopalan, “Deeply inelastic scattering structure
functions on a hybrid quantum computer,” Phys. Rev. D 102, 016007 (2020), arXiv:1908.07051 [hep-
th].
[185] Adrián Pérez-Salinas, Juan Cruz-Martinez, Abdulla A. Alhajri, and Stefano Carrazza, “Determining
the proton content with a quantum computer,” Phys. Rev. D 103, 034027 (2021), arXiv:2011.13934
[hep-ph].
[186] Wenyang Qian, Robert Basili, Soham Pal, Glenn Luecke, and James P. Vary, “Solving hadron
structures using the basis light-front quantization approach on quantum computers,” (2021),
arXiv:2112.01927 [quant-ph].
[187] J. S. Pedernales, R. Di Candia, I. L. Egusquiza, J. Casanova, and E. Solano, “Efficient
quantum algorithm for computing n-time correlation functions,” Physical Review Letters 113 (2014),
10.1103/physrevlett.113.020505, arXiv:1401.2430 [quant-ph].
[188] Carter M. Gustin and Gary Goldstein, “Generalized Parton Distribution Functions via Quantum
Simulation of Quantum Field Theory in Light-front Coordinates,” (2022), arXiv:2211.07826 [hep-th].
[189] Harry Buhrman, Richard Cleve, John Watrous, and Ronald de Wolf, “Quantum Fingerprinting,”
Phys. Rev. Lett. 87, 167902 (2001), arXiv:quant-ph/0102001.
[190] Tyler A. Cochran et al., “Visualizing Dynamics of Charges and Strings in (2+1)D Lattice Gauge
Theories,” (2024), arXiv:2409.17142 [quant-ph].
[191] Arinjoy De, Alessio Lerose, De Luo, Federica M Surace, Alexander Schuckert, Elizabeth R Bennewitz,
Brayden Ware, William Morong, Kate S Collins, Zohreh Davoudi, Alexey V Gorshkov, Or Katz, and
Christopher Monroe, “Observation of string-breaking dynamics in a quantum simulator,” (2024),
arXiv:2410.13815 [quant-ph].
[192] Daniel Gonzalez-Cuadra, Majd Hamdan, Torsten V Zache, Boris Braverman, Milan Kornjaca,
Alexander Lukin, Sergio H Cantu, Fangli Liu, Sheng-Tao Wang, Alexander Keesling, Mikhail D Lukin,
 50

Peter Zoller, and Alexei Bylinskii, “Observation of string breaking on a (2+1)D Rydberg quantum
simulator,” (2024), arXiv:2410.16558 [quant-ph].
[193] Anthony N. Ciavarella, “String Breaking in the Heavy Quark Limit with Scalable Circuits,” (2024),
arXiv:2411.05915 [quant-ph].
[194] Ying Liu, Wei-Yong Zhang, Zi-Hang Zhu, Ming-Gen He, Zhen-Sheng Yuan, and Jian-Wei Pan,
“String breaking mechanism in a lattice Schwinger model simulator,” (2024), arXiv:2411.15443 [cond-
mat.quant-gas].
[195] Constantia Alexandrou, Andreas Athenodorou, Kostas Blekos, Georgios Polykratis, and Stefan Kühn,
“Realizing string breaking dynamics in a Z2 lattice gauge theory on quantum hardware,” (2025),
arXiv:2504.13760 [hep-lat].
[196] De Luo, Federica Maria Surace, Arinjoy De, Alessio Lerose, Elizabeth R Bennewitz, Brayden
Ware, Alexander Schuckert, Zohreh Davoudi, Alexey V Gorshkov, Or Katz, and Christopher
Monroe, “Quantum simulation of bubble nucleation across a quantum phase transition,” (2025),
arXiv:2505.09607 [quant-ph].
[197] Zhao-Yu Zhou, Guo-Xian Su, Jad C. Halimeh, Robert Ott, Hui Sun, Philipp Hauke, Bing Yang,
Zhen-Sheng Yuan, Jürgen Berges, and Jian-Wei Pan, “Thermalization dynamics of a gauge theory
on a quantum simulator,” Science 377, abl6277 (2022), arXiv:2107.13563 [cond-mat.quant-gas].
[198] Niklas Mueller, Tianyi Wang, Or Katz, Zohreh Davoudi, and Marko Cetina, “Quantum Computing
Universal Thermalization Dynamics in a (2+1)D Lattice Gauge Theory,” (2024), arXiv:2408.00069
[quant-ph].
[199] Tianyin Li, Hongxi Xing, and Dan-Bo Zhang, “Simulating Parton Fragmentation on Quantum
Computers,” (2024), arXiv:2406.05683 [hep-ph].
[200] Roland C. Farrell, Marc Illa, and Martin J. Savage, “Steps toward quantum simulations of
hadronization and energy loss in dense matter,” Phys. Rev. C 111, 015202 (2025), arXiv:2405.06620
[quant-ph].
[201] Kyle Lee, Francesco Turro, and Xiaojun Yao, “Quantum Computing for Energy Correlators,” (2024),
arXiv:2409.13830 [hep-ph].
[202] Adrien Florio, David Frenklakh, Kazuki Ikeda, Dmitri E. Kharzeev, Vladimir Korepin, Shuzhe
Shi, and Kwangmin Yu, “Quantum real-time evolution of entanglement and hadronization in
jet production: Lessons from the massive Schwinger model,” Phys. Rev. D 110, 094029 (2024),
arXiv:2404.00087 [hep-ph].
[203] Zohreh Davoudi, Christopher Jarzynski, Niklas Mueller, Greeshma Oruganti, Connor Powers, and
Nicole Yunger Halpern, “Quantum Thermodynamics of Nonequilibrium Processes in Lattice Gauge
Theories,” Phys. Rev. Lett. 133, 250402 (2024), arXiv:2404.02965 [quant-ph].
[204] Qiskit contributors, “Qiskit: An open-source framework for quantum computing,” (2023).
[205] Matthew Fishman, Steven R. White, and E. Miles Stoudenmire, “The ITensor Software Library for
Tensor Network Calculations,” SciPost Phys. Codebases , 4 (2022).
[206] Jeff Bezanson, Alan Edelman, Stefan Karpinski, and Viral B Shah, “Julia: A fresh approach to
numerical computing,” SIAM Review 59, 65–98 (2017).
[207] Thomas Kluyver, Benjamin Ragan-Kelley, Fernando Pérez, Brian E Granger, Matthias Bussonnier,
Jonathan Frederic, Kyle Kelley, Jessica B Hamrick, Jason Grout, Sylvain Corlay, et al., “Jupyter
notebooks-a publishing format for reproducible computational workflows.” Elpub 2016, 87–90 (2016).
[208] Anaconda Software Distribution, “Anaconda,” .
[209] Jutho Haegeman, Spyridon Michalakis, Bruno Nachtergaele, Tobias J Osborne, Norbert Schuch, and
Frank Verstraete, “Elementary excitations in gapped quantum spin systems,” Physical review letters
111, 080401 (2013).
[210] Valentin Zauner-Stauber, Laurens Vanderstraeten, Matthew T Fishman, Frank Verstraete, and Jutho
Haegeman, “Variational optimization algorithms for uniform matrix product states,” Physical Review
B 97, 045145 (2018).
[211] Jutho Haegeman, Bogdan Pirvu, David J Weir, J Ignacio Cirac, Tobias J Osborne, Henri Verschelde,
and Frank Verstraete, “Variational matrix product ansatz for dispersion relations,” Physical Review
B—Condensed Matter and Materials Physics 85, 100408 (2012).