WorksPhysics and Astronomy
Exponential speedup in quantum simulation of Kogut-Susskind Hamiltonian via orbifold lattice
Abstract
We demonstrate that the orbifold lattice Hamiltonian – an approach known for its efficiency in simulating SU() Yang-Mills theory and QCD on digital quantum computers – can reproduce the Kogut-Susskind Hamiltonian in a controlled limit. We show that it emerges naturally as the infinite scalar mass limit of the orbifold lattice formulation, even at finite lattice spacing. Our analysis provides both a general analytical framework applicable to SU() gauge theories in arbitrary dimensions and specific numerical evidence for -dimensional SU() Yang-Mills theories (). Using Euclidean path integral methods, we quantify the approach to this limit by comparing the standard Wilson action with the orbifold lattice action, matching lattice parameters, and systematically extrapolating results as the bare scalar mass approaches infinity. This reformulation resolves longstanding technical obstacles and offers a straightforward implementation protocol for digital quantum simulation of the Kogut-Susskind Hamiltonian with exponential speedup compared to classical methods and previously known quantum methods, modulo standard assumptions made also for the original Kogut-Susskind approach. We emphasize that the present work establishes this equivalence using classical Euclidean Monte Carlo; the exponential speedup itself is a property of quantum simulation of the orbifold Hamiltonian established in earlier work, which the equivalence shown here lets one inherit for the Kogut-Susskind Hamiltonian.
Introduction
The Kogut-Susskind Hamiltonian [1] has emerged as a canonical framework for quantum simulation of Yang-Mills theory and QCD on quantum computing architectures. However, this approach presents a fundamental incongruity: the formalism, conceived prior to the conceptualization of quantum computation [2], was inherently not designed with quantum algorithmic efficiency in mind. Its reliance on compact variables — specifically unitary link variables — engenders a Hilbert space whose structural complexity becomes increasingly intractable for when extended beyond one spatial dimension, particularly when the continuum limit is considered. This intrinsic limitation has manifested as a persistent obstacle; despite two decades of scholarly pursuit since the inaugural quantum simulation protocol [3], the field has yet to produce a concrete implementation capable of demonstrating genuine quantum advantage in the continuum limit [4]. Although simulations on quantum hardware (digital or analog), including refs. [5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22], have led to promising results for certain limited cases, they leveraged specific features of systems (e.g., Abelian, low dimensions, or truncation to small number of levels) which cannot be extended to in higher spatial dimensions. The fundamental constraints are twofold: existing protocols necessitate classical preprocessing with computational demands that scale exponentially with qubit count, while the quantum circuit depth likewise exhibits exponential scaling relative to the number of qubits allocated to each link variable. Because of this, there is no known quantum simulation protocol that allows us to take the continuum limit of the Kogut-Susskind Hamiltonian [4]; the readers are encouraged to check the state-of-the-art effort on the explicit circuit construction of the theory on three spatial dimensions [23] that observed significant resource requirement. Strictly speaking, even asymptotic gate counting is difficult for the Kogut-Susskind approach in its original form. For example, while ref. [24] offered a quantum simulation protocol whose gate counting increases exponentially with the truncation level, ref. [25] provided only a counting that ignored the cost hidden in sparse oracles. Needless to say, generic methods for constructing such oracles typically lead to exponential growth in circuit depth and compilation cost [4]. There is no explicit circuit construction for the theory in dimensions based on the Kogut-Susskind Hamiltonian which avoids exponential scaling of the resource requirement.1 In fact, in most cases, even a resource estimate is difficult; see ref. [4, 27] for details.
To provide a viable way to realize quantum simulations of Yang-Mills theory and QCD, it is crucial to identify the reason for the technical complications in naive Kogut-Susskind-based approaches and provide alternative lattice formulations. In fact, it is not difficult to understand what makes the Kogut-Susskind Hamiltonian so complicated; a key observation is that theories based on non-compact variables, such as scalar quantum field theory [28] and Hermitian matrix models [29], can be simulated straightforwardly. This simple fact leads to a simple and profound guiding principle: use non-compact variables, and avoid compact variables, because the complicated structure in the Kogut-Susskind Hamiltonian comes from the use of compact variables. Once having this guiding principle, it is natural to use the orbifold lattice Hamiltonian [30, 31, 32], which is formulated using the non-compact variables. Indeed, the orbifold lattice Hamiltonian leads to exponential improvement in the quantum simulations of Yang-Mills theory and QCD, for any and any spatial dimensions [33, 4, 34]. For example, a quantum circuit for Hamiltonian time evolution via Trotter decomposition can be written explicitly using only CNOT gates, one-qubit rotations and quantum Fourier transform, leading to a proof of a polynomial scaling of the gate complexity and compiling cost with regard to the number of qubits assigned to each bosonic degree of freedom). This has been done for arbitrary truncation levels, using only analytic methods and hence without relying on a black box [33, 34]. Such strong results are obtained because of a simple, universal structure of the Hamiltonian that is agnostic to the details of the theories such as gauge group or spatial dimensions.
In this work, we present a synthesis of the two approaches: the Kogut-Susskind Hamiltonian emerges as a limiting case from the orbifold lattice Hamiltonian. From this point of view, the orbifold lattice formulation appears as a controlled relaxation of the constraints of the field-theoretical target space. The equivalence of the Kogut-Susskind and the orbifold lattice Hamiltonians in the limit of infinite scalar mass is trivial in the continuum and at weak coupling. This paper aims to test a smooth transition at finite lattice spacing and strong couplings as well.
That we can obtain the Kogut-Susskind Hamiltonian as a special limit of the orbifold lattice Hamiltonian means that the significant advantage of the latter — that *one can easily write the truncated Hamiltonian explicitly by hand, and efficient quantum circuits can be designed by hand — can be transferred to the former. Therefore, we can enjoy the advantages of the standard Kogut-Susskind formulation, in particular related to theoretical considerations — there is a direct correspondence with Wilson’s Euclidean path-integral formulation on the lattice [35], which is a framework that has underpinned countless investigations — resolving the technical disadvantages in its application to quantum simulations.
This paper demonstrates that the Kogut-Susskind Hamiltonian can be derived as a specific limiting case of the orbifold lattice Hamiltonian, precisely when the mass parameters of certain fields approach infinity, without introducing additional lattice artifacts. Crucially, this limit can be systematically approached without compromising the quantum simulation advantages inherent to the orbifold lattice formulation. This realization offers a methodological synthesis: the quantum simulation of the Kogut-Susskind Hamiltonian becomes achievable through the orbifold lattice Hamiltonian evaluated at various mass parameters, with subsequent extrapolation to the infinite-mass limit. Given that the orbifold lattice Hamiltonian facilitates exponentially faster quantum simulations than conventional Kogut-Susskind approaches, our method delivers an exponential computational acceleration. The elegance of this approach lies in its universal applicability across arbitrary gauge groups and spacetime dimensions.2
While this equivalence has been previously suggested [30, 31], comprehensive details were not provided, as the focus was primarily on establishing the equivalence in the continuum limit through radiative corrections rather than demonstrating the equivalence at finite lattice spacing and outside the weak coupling regime. This paper provides a rigorous analysis of the lattice-regularized equivalence. Following a general analytical treatment for Yang-Mills theory applicable to arbitrary and dimensions with straightforward extensions to QCD, we present numerical evidence for the -dimensional pure Yang-Mills theory. Our computational approach employs the path-integral formulation, which, while equivalent to the Hamiltonian formulation, offers complementary computational advantages. For quantitative investigation, we employ lattice Monte Carlo methods to compare -d Wilson action with -d orbifold action under identical lattice spacing and volume conditions for and . Here, we adopt the Euclidean signature, which is convenient for our purpose of seeing how heavy degrees of freedom affect physics at low energy. Within the orbifold-lattice framework, we examine multiple bare scalar mass values, confirming that Wilson action results emerge naturally through extrapolation to infinite bare mass. This methodology enables a quantitative assessment of convergence to the Kogut-Susskind Hamiltonian.
Note, however, that there is no reason that we have to take the Kogut-Susskind limit; the orbifold-lattice formulation leads to the same continuum limit as the Kogut-Susskind formulation, because the scalars acquire positive radiative mass corrections (generic for scalars coupled to gauge fields at constant bare mass) that grow relative to the dynamical scale as , so the scalars become heavy and decouple from the low-energy dynamics. What we demonstrate in this paper is that the Kogut-Susskind formulation can be reproduced from the orbifold lattice including unphysical lattice artifacts, if one prefers to do that; see Figure 1. Note also that no new energy scale beyond the cutoff scale has to be introduced to estimate the Kogut-Susskind limit. Still, there may be a nontrivial dynamics that causes complications for quantum simulations, for example the use of smaller Trotter steps; but that is also a potential problem of the original Kogut-Susskind approach.

Figure 1. Both orbifold lattice Hamiltonian and Kogut-Susskind Hamiltonian describe Yang-Mills theory in the continuum limit. Optionally, the Kogut-Susskind Hamiltonian can be obtained as a special limit of the orbifold lattice Hamiltonian — which is different from the continuum limit — including unphysical lattice artifacts.
The paper proceeds as follows: Section 2 introduces the orbifold lattice Hamiltonian for Yang-Mills theory. Section 3 presents relevant lattice actions and elucidates the relationship between orbifold lattice Hamiltonian/action and Kogut-Susskind Hamiltonian/Wilson action, with particular emphasis on the infinite-mass parameter limit. Section 4 provides numerical confirmations for -d and theories. Finally, Section 5 synthesizes our findings, comments on the extension to QCD, and explores promising future research directions.
Orbifold lattice Hamiltonian for SU(N) Yang-Mills theory
In this section, we introduce the orbifold Hamiltonian for an Yang-Mills theory in dimensions ( discrete spatial dimensions and continuous time). The spatial link variables are complex matrices , where and labels spatial points.
The complex link variable can be decomposed into a positive-definite Hermitian matrix and unitary link variable [37]:
The unitary variable describes the gauge field , with the well-known relation . Note that a priori the determinant of is not fixed to one. Instead, the Hamiltonian has an additional term that forces the U(1) part to be heavy and the determinant close to 1. The SU() part will be treated as the gauge field, but not the U(1) part. The Hamiltonian will also have a term that forces to be close to identity. is related to a scalar field by .
Let be the Hermitian conjugate of matrix , i.e., . We take and to be the conjugate momenta of and . For these operators, the commutation relations are
and
The Hamiltonian is
where
and
The meaning of the different contributions of the Hamiltonian are explained in ref. [31]. In this reference, we have included a term which would add a mass to the U(1) part of , but not to the U(1) part of . We have replaced this term by the second term in (6), which corresponds to a mass for both U(1) parts.
Using (R) and (I) to denote real and imaginary parts5 of as , we obtain
The Hilbert space is defined by using the coordinate eigenstates :
Specifically, we consider the states with the square-integrable wave function . Note that is the extended Hilbert space that contains the gauge nonsinglet states. A symmetry of the Hamiltonian can be gauged if we take only singlet states or avoid double-counting of the states related by the symmetry. The Hamiltonian is invariant under U() transformation, but only the SU() subgroup is gauged.
Following (1), the complex link variable is decomposed into and . Since
the first term of in (6) pushes close to the identity. In the continuum limit, at tree level, we obtain a mass term of scalar proportional to . The second term in (6) leads to the mass of the U(1) part, .
When the vacuum expectation value of the scalar field is nonzero, the lattice spacing shifts effectively from ; see e.g., ref. [38]. An optional term defined by (7) adds a linear term that can be used to set the expectation value to be zero and the ‘bare’ lattice spacing agrees precisely with the effective lattice spacing. This tuning of the effective spatial lattice spacing can be chosen such that it leads to an improved convergence towards the isotropic Kogut-Susskind limit. Therefore can be considered as an improvement term. The rest of the Hamiltonian describes Yang-Mills theory coupled to scalars.
When and are large (and, optionally, when is tuned to an appropriate value), and are well localized around 1 and the identity matrix , respectively. The scalars and the U(1) part of decouple,6 and only the SU(N) gauge field is left, leading to the Kogut-Susskind Hamiltonian for SU(N) pure Yang-Mills theory. In § 3.3, we will study this limit quantitatively by using the Euclidean path integral.7
The orbifold lattice Hamiltonian (4) belongs to a class of Hamiltonians of bosonic systems of the form
where and are the coordinate and momentum operators of the -th boson that satisfy the canonical commutation relations
We assume the potential part not to be complicated, e.g., a polynomial or analytic function that can be well approximated by a lower-order truncated Taylor series. In the case of the orbifold lattice Hamiltonian studied in this paper, is a polynomial of degree . This class of theories is simple enough to admit efficient quantum simulation algorithms [33].
Kogut-Susskind Hamiltonian from the orbifold lattice
We can see the relationship between the Kogut-Susskind Hamiltonian and the infinite-mass limit of the orbifold lattice in the following way. Let us focus on the scalars, because the same argument applies to the U(1) part. When the mass is large, it is reasonable to assume that the scalars are approximated well by keeping only the free part (momentum term and the mass term); indeed, we can justify this assumption shortly. If the mass is larger than the energy scale under consideration, then the scalars are frozen to the ground state of the harmonic oscillators. In the coordinate basis, using the analytic formula for the ground state, we can see that the fluctuation of the harmonic oscillator decays as . Because of this suppression, the interaction can be neglected, indeed. Therefore, we conclude that the theory reduces to Kogut-Susskind plus decoupled harmonic oscillators.
An easy way to see the connection between the Kogut-Susskind Hamiltonian and the orbifold lattice Hamiltonian quantitatively, even at finite mass, is to switch to the path-integral formalism. Because we want to see how heavy degrees of freedom affect physics at low energy, it is convenient to consider the Euclidean signature. With the Euclidean signature, we can introduce a spacetime lattice so that the path integral reduces to a usual integral with a finite number of variables, which brings the proof of the equivalence into a form accessible by numerical methods. We can use Monte Carlo simulations to obtain quantitative results.
In this section, we provide the two spacetime lattice actions to demonstrate the emergence of the Kogut-Susskind Hamiltonian from the orbifold lattice Hamiltonian. The first is the standard Wilson action [35] (§ 3.1) that corresponds to the Kogut-Susskind Hamiltonian when the continuum limit is taken along the time direction. The second is the orbifold lattice action [31] (§ 3.2). This is slightly different from those in the original papers [39, 40, 41] reflecting a difference of motivation (the target of the original papers was exact supersymmetry on the lattice). Specifically, we use the complex link variables only for spatial links, and, as a consequence, we can take the gauge group of these links to be rather than . As already explained, the spatial links are complex matrices, which get close to once the masses get sufficiently large.
In § 3.3, we show the equivalence of the two theories directly at the regularized level, without taking the continuum limit in either temporal or spatial dimensions. This equivalence guarantees the equivalence of the Hamiltonian formulations when the continuum limit is taken along the time direction.
Wilson action
The Wilson action for Yang-Mills theory in -dimensional theory is given by
Here, and are temporal and spatial lattice spacings, respectively. We introduced different spacings (anisotropic lattice) such that we can take a limit that corresponds to the Kogut-Susskind Hamiltonian. As a path-integral measure, we use the Haar measure for both and .
Orbifold lattice action
The orbifold lattice action for Yang-Mills theory in dimensions that follows from (4), see also [31], is given as
We take the unitary temporal link variable as elements of the gauge group and not . The complex link variable can be decomposed as in (1). The additional term forces and to fluctuate around the identity. The tree level continuum limit of the action in terms of adjoint scalar and gauge field is derived using and . This assumes that fluctuates around the identity matrix ; a small departure from the identity can be absorbed into a redefinition of lattice spacing, as demonstrated in detail in ref. [38]. Optionally, the last term in (15) can be used to eliminate such a departure by tuning , without taking large [33, 38]. Either way, the corresponding continuum action at tree level is
Note that, as the path-integral measure, we use the Haar measure for and the flat measure for .
Relationship between the two theories
To remove the scalars and the part of , we send and to infinity. In this limit,8 , , and . The orbifold lattice action (14) becomes
The second line vanishes since . The third line is
which is the standard plaquette terms. Likewise, the first line is also written in terms of plaquette. Therefore, up to an additive constant,
This is the same as the Wilson action. The measure of the integration of the unitary part arising from the flat measure is the Haar measure, which is the same as the measure used for the path integral with the Wilson action. Therefore, we obtain exactly the same path-integral weight and path-integral measure from the Wilson action and from the infinite-mass limit of the orbifold lattice action.
To see if such a limit is practically under control, numerical Monte Carlo simulations are required. In the next section, we study SU(2) and SU(3) pure Yang-Mills theory in 2 + 1 dimensions and confirm that it is straightforward to take this limit.
Lattice simulations for (2 + 1)-d SU(2) and SU(3) theory
We have seen that the SU() Wilson action and the Haar measure of the path integral are obtained from the orbifold lattice action and flat measure by sending and $m^2_{\mathrm{U}(1)} to infinity. In practice, for quantum simulations, we should take several values of and and then extrapolate the results to the infinite mass limit.
In this section, we demonstrate such an extrapolation for -d Yang-Mills theory, with gauge group SU(2) and SU(3). Specifically, we use the Hybrid Monte Carlo algorithm [42] (see ref. [43] for an introductory review) for the demonstration.9 We want to provide evidence for the equivalence even at the regularized level, considering small lattice sizes without the continuum extrapolation.
Note that the Yang-Mills coupling constant is dimensionful at , i.e., has mass dimension , and the lattice spacing should be measured in units of the coupling constant. To take the dependence into account, the ’t Hooft coupling provides a typical energy scale. Dimensionless combinations are and . If these dimensionless combinations are small, the system is close to the continuum limit. We set the coupling constant to . Therefore, dimensionful quantities like lattice spacing or temperature should be interpreted as the dimensionless combinations or . Furthermore, we take and study the limit of infinite mass. Whether these parameters are ‘large’ or ‘small’ should be considered in relation to the mass scale set by the coupling constant.
We will study two cases: with or without linear counter term .
Emergence of Kogut-Susskind as heavy mass limit of the orbifold theory
As a first step, we study the Kogut-Susskind (KS) limit without introducing the improvement term , i.e., setting zero. The aim of this investigation is to demonstrate that the Kogut-Susskind results are reproduced even in the ultraviolet regime including visible lattice artefacts. For simplicity, we consider the isotropic KS action at zero temperature. A more precise comparison in the spirit of the original orbifold approach would include a dynamically determined anisotropy, which we postpone to a later publication. At finite temperature we consider an anisotropic case, but only as an alternative method to adjust the temperature. We will use the following quantities for the confirmation of convergence to the KS limit:
To confirm agreement in the ultraviolet regime, we compute the spatial plaquettes. We use the one made of complex links, , and the one made of unitary links, . Note that can be obtained from using (1). As , these plaquette values should converge to the spatial plaquette obtained with the Wilson action at the same parameters up to an overall factor (4 for the former and 1 for the latter,
specific to the -dimensional case studied here; in higher dimensions this factor is -dependent).10
The temporal plaquette averaged over . The corresponding average value is denoted as . The comparison of spatial and temporal plaquettes indicates the residual anisotropy.
Another important observable is the Polyakov loop. It can be used to identify the signal for the deconfinement transition. Specifically, we consider the bare Polyakov loop without renormalization. Although it depends on the details of the ultraviolet regime through the renormalization factor, the general phase diagram detected by this quantity does not. Therefore, agreement of the bare Polyakov loop provides strong evidence for the agreement with the Kogut-Susskind action in the ultraviolet and infrared.
We also compute and . We take the average over , spacetime points , and samples with the corresponding averages and . These averages should converge to and , respectively in the KS limit.
In our measurements, we have used at least 400 independent configurations for each parameter set. The autocorrelation time is negligible except for the Polyakov loop. The Polyakov loop experiences, as expected, large autocorrelation times close to the phase transition. These are taken into account in the error estimates. Specifically, the errors of and of the Polyakov-loop susceptibility are obtained from a bootstrap over blocks of consecutive measurements, where is the integrated autocorrelation time determined on the same series, and the identical prescription is applied to every data set shown. In the transition region, several independent chains started from ordered and disordered configurations are combined for each parameter set.
Convergence of spatial and temporal plaquette expectation values
In Figure 2, spatial and temporal plaquettes are plotted against . For these plots, the lattice size is and the lattice spacing is for SU(2) and 0.2 for SU(3). There is a smooth convergence to the value obtained from the Wilson action as the mass is sent to infinity.
Figure 2. Spatial and temporal plaquette. [Left] SU(2), lattice, . [Right] SU(3), lattice, . Infinite-mass extrapolations by a quadratic function in from are shown at . The horizontal lines are the values obtained from the Wilson action. For and , the horizontal axis is slightly shifted so that the data points can be distinguished.
Throughout, the extrapolations to the infinite-mass limit are performed as expansions in powers of . This functional form is dictated by the heavy-scalar nature of the limit: the scalar fluctuations scale as , so that , and integrating out the heavy scalar generates corrections to the observables organized in powers of . The variable used for the analysis below is the same expansion parameter up to the fixed factor , so the two analyses employ the same expansion; it is truncated at quadratic order for the data, which lie very close to , and at cubic order for the data, which extend to substantially larger (i.e. ).
Figure 3 shows the finite-temperature behavior. As we will see later using the Polyakov loop, the confinement-deconfinement transition takes place in the range of shown in this plot. The gauge group is SU(3), and the lattice size is . The horizontal axis is , which is related to the temperature (in units of the coupling) by . The spatial lattice spacing is fixed to . We can see that and converge to the values in the Wilson action quickly at all values of , suggesting a decoupling of scalars and the U(1) part from the SU(N) gauge field.
Figure 3. [Left] and [Right] versus , for SU(3) on a lattice with fixed spatial spacing . Orbifold-lattice results at are compared with the Wilson action. Each point is based on at least 400 configurations, separated by 150 HMC trajectories; errors are autocorrelation-corrected.
Polyakov loop
The Polyakov loop is constructed by taking a trace of the product of temporal links at each spatial point as
We take the spatial average for each configuration,
For simplicity, we call this quantity Polyakov loop in the following.
A simple way to characterize the deconfinement transition in this theory is the breaking of center symmetry that acts on the Polyakov loop as . In the large-volume limit, a nonzero expectation value of indicates broken center symmetry. However, in a finite volume, tunnelings between different sectors take place and vanishes even in the broken regime. Therefore, we use instead.
In Figure 4, and the Polyakov-loop susceptibility , with spatial volume, are shown. The increase in and the peak of indicate the deconfinement transition. We observe convergence to the values of the Wilson action as the mass is sent to infinity, across a wide temperature region, including the phase transition. The same convergence holds for the distribution of ; see Figure 5. The residual deviation that remains at the smallest mass, —the transition is located at a visibly smaller than for the Wilson action—is revisited at the end of § 4, where the improvement term is switched on; see Figure 13.
Figure 4. [Left] versus . [Right] the Polyakov-loop susceptibility , with spatial volume , versus . SU(3), lattice, fixed as ; orbifold results at , 300, 1000 compared with the Wilson action. Based on the same configurations as Figure 3, supplemented near the transition by additional independent chains started from ordered and disordered configurations. The rise of and the peak of the susceptibility locate the deconfinement transition. Note that the transition of the orbifold theory moves towards the Wilson-action result as is increased. Rather large -dependence is due to a shift of effective lattice spacing, as we will see from Figure 13.
Figure 5. The distributions of from the Monte Carlo simulations. The horizontal and vertical axes are and , respectively. SU(3), lattice, fixed ; orbifold results at , , compared with the Wilson action. Based on the same configurations as Figure 4.
Figure 13. Effect of the improvement term on the deconfinement transition. [Left] versus ; [Right] the Polyakov-loop susceptibility versus . SU(3), lattice, fixed , at and at , compared with the Wilson action. The and Wilson data are those of Figure 4. The vertical lines mark the susceptibility peak positions obtained from weighted parabola fits, , and for , and the Wilson action, respectively. Each chain consists of about 400 measurements separated by 150 HMC trajectories; at the values of closest to the peak, four to five independent chains with ordered and disordered starts are combined. Errors are obtained as described in § 4.
The simplest investigation of the phase transition in Yang–Mills theory usually considers isotropic lattices without a continuum extrapolation. In order to test this scenario, we have set the same lattice spacing in both directions , which also determines the temperature. This is the same as a scan of the results as a function of the coupling constant. The results are shown in Figure 6 and illustrate the convergence to the Wilson action results. In this case we have also added some lower masses, which represent a stronger deviation from the Wilson data. The investigation of the lower mass regime is not the focus of the current paper, but will be investigated in a future publication.
Figure 6. A parameter scan with the orbifold action of SU(3) Yang-Mills theory on a lattice. The lattice spacing is isotropic, . For comparison, data from the Wilson action has been added; the Wilson action data were obtained with a heat bath algorithm, whereas all other data were computed with HMC.
Convergence of and
In Figure 7, and are plotted for SU(2) and SU(3), lattice, taking horizontal axis . We can see the convergence to 0 and 1 as the mass is sent to infinity.

Figure 7. (left column) and (right column). [Top] SU(2), lattice, . [Bottom] SU(3), lattice, . Infinite-mass extrapolations by a quadratic function of from are shown at .
In Figure 8, the behaviors of these quantities at finite temperature are shown for SU(3). The approach to the infinite-mass limit can be seen at any temperature in the plots.
Figure 8. [Left] versus . [Right] versus . SU(3), lattice, fixed ; orbifold results at . Based on the same configurations as Figure 3.
Improved extrapolation to the Kogut-Susskind limit
So far, we demonstrated the convergence to the Kogut-Susskind limit with fixed bare lattice spacing. The effective lattice spacing was different from the bare lattice spacing when the mass parameters are finite.
Now, we study the Kogut-Susskind limit at fixed effective lattice spacing. Specifically, we introduce . (As another option, we could also tune bare lattice spacing to keep the effective lattice spacing fixed [38].) At each value of , we find where becomes zero, and then, study the -dependence of several quantities at .
Determination of
To determine at each , we perform simulations at several values of , fit as a function of , simply using . An example is shown in Figure 9. Tables 1a and 1b summarize the results for SU(2) and SU(3) theories, respectively, on an lattice with for SU(2) and for SU(3).
Figure 9. Fit of values versus , lattice, SU(3), , .
-dependence at
We performed Monte Carlo simulations at the values of summarized in Tables 1a and 1b. Plaquette values are shown in Figure 10, taking the horizontal axis. We can see a significant improvement from . This shows that the most of the “deviation from the Kogut-Susskind limit” observed at merely came from a shift of lattice spacing from the input value.
Figure 10. Plaquette values versus at , lattice, [Left] SU(2), , and [Right] SU(3), .
Furthermore, the effectiveness of tuning can be better appreciated by examining how the plaquette values obtained at for different values of — shown in Table 1a for SU(2) and in Table 1b for SU(3) — can be used to extrapolate their values to the KS limit (). These extrapolations, based exclusively on plaquette values obtained from simulations at with , are shown in Figure 11 for SU(2) at , and in Figure 12 for SU(3) at . Therefore, as expected, the KS limit can be evaluated using , without introducing a new scale other than . We need only the same energy scales as the KS formulation, i.e., the cutoff scale , and actual physical scales of interest such as the glueball mass.
| 5 | 9.746680 | 0.670820 | 5.0 | 16.315680 | 0.447214 |
| 9 | 7.730493 | 0.900000 | 9.0 | 18.561845 | 0.6 |
| 11 | 7.671210 | 0.994987 | 16.0 | 17.933675 | 0.8 |
| 13 | 7.623557 | 1.081665 | 20.0 | 17.726980 | 0.894427 |
| 16 | 7.575180 | 1.200000 | 25.0 | 17.467360 | 1.0 |
| 22 | 7.502837 | 1.407125 | 30.0 | 17.283905 | 1.095445 |
| 25 | 7.476063 | 1.500000 | 36.0 | 17.060590 | 1.2 |
| 28 | 7.446263 | 1.587451 | 42.0 | 16.854545 | 1.296148 |
| 32 | 7.408923 | 1.697056 | 49.0 | 16.629630 | 1.4 |
| 36 | 7.386837 | 1.800000 | 50.0 | 16.586855 | 1.414214 |
| 40 | 7.354473 | 1.897367 | 56.0 | 16.438995 | 1.496663 |
| 44 | 7.346650 | 1.989975 | 64.0 | 16.222980 | 1.6 |
| 50 | 7.315410 | 2.121320 | 72.0 | 16.027750 | 1.697056 |
| 75 | 7.243050 | 2.598076 | 75.0 | 15.951550 | 1.732051 |
| 100 | 7.194417 | 3.000000 | 81.0 | 15.837810 | 1.8 |
| 125 | 7.161960 | 3.354102 | 90.0 | 15.643905 | 1.897367 |
| 150 | 7.140743 | 3.674235 | 100.0 | 15.459455 | 2.0 |
| 175 | 7.106953 | 3.968627 | 125.0 | 15.047240 | 2.236068 |
| 200 | 7.094080 | 4.242641 | 150.0 | 14.676130 | 2.449490 |
| 225 | 7.087750 | 4.500000 | 175.0 | 14.389375 | 2.645751 |
| 250 | 7.073507 | 4.743416 | 200.0 | 14.131770 | 2.828427 |
| 275 | 7.065953 | 4.974937 | 225.0 | 13.930280 | 3.000 |
| 300 | 7.046990 | 5.196152 | 250.0 | 13.718430 | 3.162 |
| 275.0 | 13.553465 | 3.317 | |||
| 300.0 | 13.397350 | 3.464 |
Table 1. (a) SU(2), . (b) SU(3), . Table 1: The values of used to demonstrate the improved extrapolation to the infinite-mass limit with fixed effective lattice spacing, and the corresponding and . The lattice size is . [left] SU(2) at , [right] SU(3) at .
Figure 11. Extrapolation of SU(2) plaquette values toward the KS limit using simulations at with on an lattice with . [Left] Data in the range , and [Right] Data in the range . The fit ansatz is a cubic polynomial of (equivalently, of ; see text).
Figure 12. Extrapolation of SU(3) plaquette values toward the KS limit using simulations at with on an lattice with . [Left] Data in the range , and [Right] Data in the range . The fit ansatz is a cubic polynomial of (equivalently, of ; see text).
The deconfinement transition at
So far, the improvement term has been tested with ultraviolet observables at zero temperature. It is natural to ask whether fixing the effective lattice spacing also improves an infrared quantity. The confinement-deconfinement transition provides such a test, and a demanding one: as noted in § 4, the location of the transition is the largest deviation from the Kogut-Susskind limit that we observe at .
We have therefore repeated the finite-temperature scan of Figure 4 at with the improvement term switched on. The coefficient is not re-tuned at each temperature. What is held fixed across the scan is the combination that multiplies in (15), at the value corresponding to of Table 1b at the isotropic point . The product is the natural quantity to hold fixed from a strong coupling perspective. For each link, the departure of from that this term compensates originates from the flat integration measure for , which carries no factor of (a detailed discussion will be given elsewhere). In this way one and the same term is added to the action at every temperature, and the comparison introduces no additional temperature-dependent parameter.
This prescription is justified a posteriori by the tuning condition itself, which remains well satisfied over the entire scan. At , where was determined, we find , i.e. the condition is met on the lattice with the value determined independently on the lattice. Away from that point the residual grows only slowly and monotonically, staying below 0.011 in absolute value over the whole range , to be compared with 0.10–0.17 at ; the linear term removes between 90% and 99.9% of the departure of from across the scan. The quantity of Figure 8 responds far less: it decreases by less than 3% over the same range (from 0.188 to 0.184 at ).
The effect on the transition is shown in Figure 13. At every the improved lies between the and the Wilson curve, with no crossings, and the susceptibility peak moves from
to be compared with ; the peak positions are obtained from weighted parabola fits to the five points closest to the maximum. This means that the improvement removes about 74% of the shift of the transition. Equivalently, in units of the coupling, the transition temperature moves from 1.4134(14) to 1.2502(62), against 1.2011(13) for the Wilson action.
Conclusions and outlook
A simple and universal framework [33] applicable to a wide class of theories, including the orbifold lattice, leads to a significant speedup compared to more complex formulations such as the Kogut-Susskind Hamiltonian [44]. Our work demonstrates that the Kogut-Susskind Hamiltonian can be obtained as a special limit of the orbifold lattice, allowing us to leverage this property to achieve exponential speedup in quantum simulations. Specifically, we can achieve exponential speedup in quantum simulation of the Kogut-Susskind Hamiltonian by performing quantum simulations of the orbifold lattice Hamiltonian at several values of mass parameters and extrapolating the results to the infinite-mass limit. Although the final result reproduces that of the Kogut-Susskind Hamiltonian, the simulations utilize the orbifold lattice Hamiltonian, thereby enabling us to implement the universal framework that delivers exponential speedup. Note that there is no need to introduce a new energy scale; as a simple dimensional analysis suggests, we can evaluate the Kogut-Susskind limit from bare scalar mass of order of the inverse lattice spacing, once the improvement term tuned to is included. Without this term, we used for the unimproved extrapolation. The same term also improves an infrared quantity: at it removes a large part of the shift of the deconfinement transition of the -d SU(3) theory relative
to the Wilson action, leaving a small residual that we have not attempted to remove.1112
We emphasize that the mechanism demonstrated here is not restricted to pure Yang-Mills theory. The orbifold lattice formulation of QCD – including quarks in the fundamental representation, in both the action and the Hamiltonian – was constructed in ref. [31]. There the noncompact link variable reduces to the ordinary unitary link as the scalar mass parameters and are sent to infinity (so that and ) through the same type of term (or ), and standard Wilson-fermion lattice QCD is recovered. Since this decoupling of the scalars is independent of the matter sector, the infinite-mass limit studied in the present work – and hence the route to the Kogut-Susskind Hamiltonian and the associated exponential speedup – carries over directly to full QCD. In particular, nothing in the limiting mechanism relies on the absence of matter, so the demonstration presented here should be read as evidence for the viability of the approach in QCD, not merely in pure gauge theory.
Note that several terms in the orbifold Hamiltonian (4) can be dropped without altering the infinite-mass limit. If we are interested only in this limit, it would be better to remove these terms and make the quantum circuits shorter. More generally, we might be able to use such flexibility in the choice of the Hamiltonian to connect the Kogut-Susskind limit and a simple model smoothly, allowing simpler extrapolations to the Kogut-Susskind limit. Furthermore, such an approach would enable adiabatic state preparation protocols with reduced computational overhead. We plan to explore this strategy using conventional lattice simulation methods on classical computers in forthcoming work.
Other embeddings of the group manifold to flat space can also be useful. For example, can be embedded into , interpreting the radial coordinate as a scalar. This approach can cut half of the bosons per link from the orbifold lattice. It would be nice if we could find the best embedding for each group, balancing the qubit requirement and gate complexity.
Our findings open numerous directions for future exploration. We can extend our numerical demonstration to -dimensional theory, incorporate quarks, increase to , and investigate the continuum limit. The constraint term that enforces may prove less critical than previously thought, at least for pure Yang-Mills theory, and verifying this numerically could substantially reduce simulation costs. Additionally, we can estimate the required qubit count for quantum simulations using the Monte Carlo technique introduced in ref. [45]. Developing optimized simulation algorithms — particularly state preparation methods using orbifold lattice techniques — represents another critical avenue. In this context, avoiding nontrivial oracles is essential to preserve the advantages offered by the orbifold lattice. Reexamining the Kogut-Susskind Hamiltonian literature through this new lens may reveal valuable insights, such as magnetic basis formulations that could facilitate more efficient state preparations.
The approach presented here represents the first viable method to make the Kogut-Susskind Hamiltonian programmable toward the continuum limit on digital quantum computers for arbitrary gauge groups and dimensions. Our framework uniquely enables seamless utilization of both coordinate (magnetic) and momentum (electric) bases. Furthermore, the orbifold Hamiltonian belongs to a broader class of Hamiltonians (11), creating opportunities for developing tailored simulation strategies that fully exploit these capabilities. Near-term goals should include testing simplified models of the same generic form on actual quantum devices as proof-of-principle demonstrations.
The fundamental simplicity of the orbifold lattice stems from emergent geometry, arising from matrix models with specific backgrounds [32, 46] – a concept that extends to noncommutative geometry in matrix models [29]. Theories on emergent spaces inherit elegant structures from their original formulations, suggesting that nature’s constructions may surpass human design, as dynamically generated spatial dimensions exhibit superior properties compared to artificially constructed lattices [29]. Gauge/gravity dualities [47] provide even deeper insights, connecting gravitational geometries with non-gravitational theories.13 Enhanced understanding of emergent geometry could lead to more efficient quantum simulation protocols. Importantly, the universal framework for Hamiltonians of the form (11) applies across diverse theories — from simple toy models to matrix models and quantum field theories dual to superstring/M-theory. This unifying perspective reveals that quantum simulations of QCD and superstring/M-theory are fundamentally interconnected; by advancing our understanding of both simultaneously, we stand to gain deeper insights and achieve more significant progress in quantum simulation capabilities.
Acknowledgment
The authors thank Graham Van Goffrier, Shunji Matsuura, Enrico Rinaldi, Andreas Schäfer, David Schaich, Jesse Stryker, and Carsten Urbach for discussions and comments. The authors thank Andreas Schäfer for carefully checking the manuscript and providing valuable feedback. G. B. is funded by the Deutsche Forschungsgemeinschaft (DFG) under Grant No. 432299911 and 431842497. M. H. thanks the STFC for the support through the consolidated grant ST/Z001072/1. E. M. was supported by UK Research and Innovation Future Leader Fellowship MR/X015157/1. Part of the computing time for this project has been provided by the compute cluster ARA of the University of Jena. Part of the numerical simulations were undertaken on Barkla High Performance Computing facilities at the University of Liverpool. The authors acknowledge the use of the AI-based assistant Claude (Anthropic) for editing assistance during the preparation of this manuscript. The authors take full responsibility for all content of the article.
References
- [1]J. B. Kogut and L. Susskind, “Hamiltonian Formulation of Wilson’s Lattice Gauge Theories,” Phys. Rev. D 11 (1975) 395–408.DOI
- [2]R. P. Feynman, “Simulating physics with computers,” Int. J. Theor. Phys. 21 (1982) 467–488.
- [3]T. Byrnes and Y. Yamamoto, “Simulating lattice gauge theories on a quantum computer,” Phys. Rev. A 73 (2006) 022328, arXiv:quant-ph/0510027.DOI
- [4]M. Hanada, S. Matsuura, E. Mendicelli, and E. Rinaldi, “Exponential improvement in quantum simulations of bosons,” arXiv:2505.02553 [quant-ph].arxiv.org/abs/2505.02553
- [5]E. A. Martinez et al., “Real-time dynamics of lattice gauge theories with a few-qubit quantum computer,” Nature 534 (2016) 516–519, arXiv:1605.04570 [quant-ph].DOI
- [6]B. Yang, H. Sun, R. Ott, H.-Y. Wang, T. V. Zache, J. C. Halimeh, Z.-S. Yuan, P. Hauke, and J.-W. Pan, “Observation of gauge invariance in a 71-site Bose–Hubbard quantum simulator,” Nature 587 no. 7834, (2020) 392–396, arXiv:2003.08945 [cond-mat.quant-gas].DOI
- [7]N. Klco, J. R. Stryker, and M. J. Savage, “SU(2) non-Abelian gauge field theory in one dimension on digital quantum computers,” Phys. Rev. D 101 no. 7, (2020) 074512, arXiv:1908.06935 [quant-ph].DOI
- [8]S. A Rahman, R. Lewis, E. Mendicelli, and S. Powell, “SU(2) lattice gauge theory on a quantum annealer,” Phys. Rev. D 104 no. 3, (2021) 034501, arXiv:2103.08661 [hep-lat].DOI
- [9]S. A Rahman, R. Lewis, E. Mendicelli, and S. Powell, “Self-mitigating Trotter circuits for SU(2) lattice gauge theory on a quantum computer,” Phys. Rev. D 106 no. 7, (2022) 074502, arXiv:2205.09247 [hep-lat].DOI
- [10]Y. Y. Atas, J. Zhang, R. Lewis, A. Jahanpour, J. F. Haase, and C. A. Muschik, “SU(2) hadrons on a quantum computer via a variational approach,” Nature Commun. 12 no. 1, (2021) 6499, arXiv:2102.08920 [quant-ph].DOI
- [11]A. N. Ciavarella, “String breaking in the heavy quark limit with scalable circuits,” Phys. Rev. D 111 no. 5, (2025) 054501, arXiv:2411.05915 [quant-ph].arxiv.org/abs/2411.05915
- [12]A. H. Z. Kavaki and R. Lewis, “From square plaquettes to triomand lattices for SU(2) gauge theory,” Commun. Phys. 7 no. 1, (2024) 208, arXiv:2401.14570 [hep-lat].DOI
- [13]A. T. Than et al., “The phase diagram of quantum chromodynamics in one dimension on a quantum computer,” arXiv:2501.00579 [quant-ph].arxiv.org/abs/2501.00579
- [14]A. Ciavarella, N. Klco, and M. J. Savage, “Trailhead for quantum simulation of SU(3) Yang-Mills lattice gauge theory in the local multiplet basis,” Phys. Rev. D 103 no. 9, (2021) 094501, arXiv:2101.10227 [quant-ph].DOI
- [15]M. Illa and M. J. Savage, “Basic elements for simulations of standard-model physics with quantum annealers: Multigrid and clock states,” Phys. Rev. A 106 no. 5, (2022) 052605, arXiv:2202.12340 [quant-ph].DOI
- [16]A. N. Ciavarella and I. A. Chernyshev, “Preparation of the SU(3) lattice Yang-Mills vacuum with variational quantum methods,” Phys. Rev. D 105 no. 7, (2022) 074504, arXiv:2112.09083 [quant-ph].DOI
- [17]Y. Y. Atas, J. F. Haase, J. Zhang, V. Wei, S. M. L. Pfaendler, R. Lewis, and C. A. Muschik, “Simulating one-dimensional quantum chromodynamics on a quantum computer: Real-time evolutions of tetra- and pentaquarks,” Phys. Rev. Res. 5 no. 3, (2023) 033184, arXiv:2207.03473 [quant-ph].arxiv.org/abs/2207.03473
- [18]R. C. Farrell, I. A. Chernyshev, S. J. M. Powell, N. A. Zemlevskiy, M. Illa, and M. J. Savage, “Preparations for quantum simulations of quantum chromodynamics in 1+1 dimensions. I. Axial gauge,” Phys. Rev. D 107 no. 5, (2023) 054512, arXiv:2207.01731 [quant-ph].DOI
- [19]R. C. Farrell, I. A. Chernyshev, S. J. M. Powell, N. A. Zemlevskiy, M. Illa, and M. J. Savage, “Preparations for quantum simulations of quantum chromodynamics in 1+1 dimensions. II. Single-baryon β-decay in real time,” Phys. Rev. D 107 no. 5, (2023) 054513, arXiv:2209.10781 [quant-ph].DOI
- [20]A. N. Ciavarella, “Quantum simulation of lattice QCD with improved Hamiltonians,” Phys. Rev. D 108 no. 9, (2023) 094513, arXiv:2307.05593 [hep-lat].arxiv.org/abs/2307.05593
- [21]A. N. Ciavarella and C. W. Bauer, “Quantum Simulation of SU(3) Lattice Yang-Mills Theory at Leading Order in Large-Nc Expansion,” Phys. Rev. Lett. 133 no. 11, (2024) 111901, arXiv:2402.10265 [hep-ph].arxiv.org/abs/2402.10265
- [22]T. Hayata and Y. Hidaka, “Floquet evolution of the q-deformed SU(3)1 Yang-Mills theory on a two-leg ladder,” Phys. Rev. D 111 no. 3, (2025) 034513, arXiv:2409.20263 [hep-lat].
- [23]P. Balaji, C. Conferey-Shinozaki, P. Draper, J. K. Elhaderi, D. Gupta, L. Hidalgo, A. Lytle, and E. Rinaldi, “Quantum Circuits for SU(3) Lattice Gauge Theory,” arXiv:2503.08866 [hep-lat].arxiv.org/abs/2503.08866
- [24]A. Kan and Y. Nam, “Lattice Quantum Chromodynamics and Electrodynamics on a Universal Quantum Computer,” arXiv:2107.12769 [quant-ph].arxiv.org/abs/2107.12769
- [25]M. L. Rhodes, M. Kreshchuk, and S. Pathak, “Exponential Improvements in the Simulation of Lattice Gauge Theories Using Near-Optimal Techniques,” PRX Quantum 5 no. 4, (2024) 040347, arXiv:2405.10416 [quant-ph].DOI
- [26]Z. Davoudi, A. F. Shaw, and J. R. Stryker, “General quantum algorithms for Hamiltonian simulation with applications to a non-Abelian lattice gauge theory,” Quantum 7 (2023) 1213, arXiv:2212.14030 [hep-lat].DOI
- [27]M. Hanada, “Comments on “Ether of Orbfolds”,” arXiv:2604.08622 [hep-lat].
- [28]S. P. Jordan, K. S. M. Lee, and J. Preskill, “Quantum Algorithms for Quantum Field Theories,” Science 336 (2012) 1130–1133, arXiv:1111.3633 [quant-ph].DOI
- [29]H. Gharibyan, M. Hanada, M. Honda, and J. Liu, “Toward simulating superstring/M-theory on a quantum computer,” JHEP 07 (2021) 140, arXiv:2011.06573 [hep-th].DOI
- [30]A. J. Buser, H. Gharibyan, M. Hanada, M. Honda, and J. Liu, “Quantum simulation of gauge theory via orbifold lattice,” JHEP 09 (2021) 034, arXiv:2011.06576 [hep-th].DOI
- [31]G. Bergner, M. Hanada, E. Rinaldi, and A. Schafer, “Toward QCD on quantum computer: orbifold lattice approach,” JHEP 05 (2024) 234, arXiv:2401.12045 [hep-th].DOI
- [32]D. B. Kaplan, E. Katz, and M. Unsal, “Supersymmetry on a spatial lattice,” JHEP 05 (2003) 037, arXiv:hep-lat/0206019.DOI
- [33]J. C. Halimeh, M. Hanada, S. Matsuura, F. Nori, E. Rinaldi, and A. Schäfer, “A universal framework for the quantum simulation of Yang-Mills theory,” arXiv:2411.13161 [quant-ph].arxiv.org/abs/2411.13161
- [34]J. C. Halimeh, M. Hanada, and S. Matsuura, “Universal framework with exponential speedup for the quantum simulation of quantum field theories including QCD,” arXiv:2506.18966 [quant-ph].arxiv.org/abs/2506.18966
- [35]K. G. Wilson, “Confinement of Quarks,” Phys. Rev. D 10 (1974) 2445–2459.DOI
- [36]T. Banks, W. Fischler, S. H. Shenker, and L. Susskind, “M theory as a matrix model: A conjecture,” Phys. Rev. D 55 (1997) 5112–5128, arXiv:hep-th/9610043.DOI
- [37]M. Unsal, “Compact gauge fields for supersymmetric lattices,” JHEP 11 (2005) 013, arXiv:hep-lat/0504016.DOI
- [38]G. Bergner, M. Hanada, and E. Mendicelli, “A minimal implementation of Yang-Mills theory on a digital quantum computer,” arXiv:2604.15132 [hep-lat].arxiv.org/abs/2604.15132
- [39]A. G. Cohen, D. B. Kaplan, E. Katz, and M. Unsal, “Supersymmetry on a Euclidean space-time lattice. 1. A Target theory with four supercharges,” JHEP 08 (2003) 024, arXiv:hep-lat/0302017.
- [40]A. G. Cohen, D. B. Kaplan, E. Katz, and M. Unsal, “Supersymmetry on a Euclidean space-time lattice. 2. Target theories with eight supercharges,” JHEP 12 (2003) 031, arXiv:hep-lat/0307012.DOI
- [41]D. B. Kaplan and M. Unsal, “A Euclidean lattice construction of supersymmetric Yang-Mills theories with sixteen supercharges,” JHEP 09 (2005) 042, arXiv:hep-lat/0503039.DOI
- [42]S. Duane, A. D. Kennedy, B. J. Pendleton, and D. Roweth, “Hybrid Monte Carlo,” Phys. Lett. B 195 (1987) 216–222.DOI
- [43]M. Hanada and S. Matsuura, MCMC from Scratch: A Practical Introduction to Markov Chain Monte Carlo. Springer Nature, Jan., 2022.DOI
- [44]M. Hanada, S. Matsuura, E. Mendicelli, and E. Rinaldi, “Exponential improvement in quantum simulations of bosons,” arXiv:2505.02553 [quant-ph].arxiv.org/abs/2505.02553
- [45]M. Hanada, J. Liu, E. Rinaldi, and M. Tezuka, “Estimating truncation effects of quantum bosonic systems using sampling algorithms,” Mach. Learn. Sci. Tech. 4 no. 4, (2023) 045021, arXiv:2212.08546 [quant-ph].DOI
- [46]N. Arkani-Hamed, A. G. Cohen, and H. Georgi, “(De)constructing dimensions,” Phys. Rev. Lett. 86 (2001) 4757–4761, arXiv:hep-th/0104005.
- [47]J. M. Maldacena, “The Large N limit of superconformal field theories and supergravity,” Adv. Theor. Math. Phys. 2 (1998) 231–252, arXiv:hep-th/9711200.DOI
- [48]D. B. Kaplan and M. Unsal. Private communication.










