betaFeedback

WorksPhysics and Astronomy

Towards few-body QCD on a quantum computer

Abstract

Quantum computers are promising tools for the simulation of many-body systems, and among those, QCD stands out by its rich phenomenology. Every simulation starts with a codification, and here we succently review a newly developed compact encoding based on the identification between registers and particles; the quantum memory is divided into registers, and to each we associate a Hilbert space of dimension the number of degrees of freedom of the codified particles. In this way we gain an exponential compression over direct encodings for a low number of particles with many degrees of freedom. As an example we apply this encoding on a two-register memory and implement antisymmetrization and exponentiation algorithms.

Introduction

Numerical simulations of QCD for the description of its phase space [1] or the evaluation of real-time observables are thwarted by the so-called sign problem of Monte Carlo sampling [1, 2], which makes it challenging, for example, to derive an EoS for QCD matter in neutron stars’ cores; where only loose bounds by pQCD and chiral EFT [3, 4, 5] can be derived.

Quantum computers (QC), by their very definition, can realize Hamiltonian evolution on quantum memories, usually based on two level systems called qubits [6, 7]. They are thus envisioned as powerful tools to reach there where classical computers cannot [8]. Their first applications as simulators were centered on fermion systems [2, 9] and soon reached quantum chemistry [10, 11] and HEP [12], with implementations based on lattice formulations [13, 14, 15] and works that explore other implementations [16, 17, 18, 19, 20, 21]. Among their most recent applications we find studies for the simulation of nuclear systems [22, 23], heavy quarks [24, 25], light mesons [18], in medium jets [26, 27] and thermalization and phase diagrams [28, 29, 30], among others.

Simulating a system requires encoding its degrees of freedom (quantum numbers). The “direct” encodings, based on the identification of the occupation number basis and the computational basis are the most common, and they generally need a qubit for each fermionic mode to simulate; the Jordan-Wigner [31] and the Bravyi-Kitaev [32] encodings are two prominent examples. Here we present an implementation based on a “compact encoding” scheme [21] in which the memory is divided into registers with enough size to accommodate the basis states of each particle. This codification is then used for the implementation of time evolution under a Hamiltonian with kinetic and exchange terms. This encoding is expected to be efficient for studying few-body QCD problems, given the large number of internal degrees of freedom associated with quarks and gluons.

Codification

We now review the particle-register codification discussed in [21]. Consider a quantum memory of qubits divided into registers, each representing a particle with spin and momentum,

∣Ω⟩≡∣0⟩P/A⊗∣0⟩spin⊗∣0…0⟩momentum,(1)|\Omega\rangle\equiv|0\rangle_{\mathrm{P/A}} \otimes|0\rangle_{\mathrm{spin}} \otimes|0\ldots0\rangle_{\mathrm{momentum}}, \tag*{(1)}

each set of m>0m > 0 qubits is able to store 2m2^m different values of the corresponding degree of freedom as binary numbers. The presence/absence qubit indicates whether the register is empty or occupied and it is used as control, one qubit is enough for spin, and the amount of momentum qubits depends on discretization. The aa and a†a^\dagger are written as tensor product of set, scrap and control operators Eij∣k⟩=δjk∣i⟩\mathcal{E}_{ij}|k\rangle= \delta_{jk}|i\rangle:

ap,η†≡E10⊗sη†⊗sp†,ap,η≡E01⊗sη⊗sp,(2)a^\dagger_{p,\eta} \equiv\mathcal{E}_{10} \otimes\mathsf{s}^\dagger_{\eta} \otimes\mathsf{s}^\dagger_{p}, \quad a_{p,\eta} \equiv\mathcal{E}_{01} \otimes\mathsf{s}_{\eta} \otimes\mathsf{s}_{p}, \tag*{(2)}

which initiate the corresponding qubits to the binary number that represents the value of the degree of freedom along with the presence qubit:

ap,λ†∣Ω⟩=∣1⟩⊗∣η⟩⊗∣p⟩=∣1ηp⟩.(3)a^\dagger_{p,\lambda}|\Omega\rangle= |1\rangle\otimes|\eta\rangle\otimes|p\rangle= |1\eta p\rangle. \tag*{(3)}

We use the abbreviated notation ar†≡E10⊗sr†a^\dagger_r \equiv\mathcal{E}_{10} \otimes\mathsf{s}^\dagger_r to collectively denote all quantum numbers.

Some problems arise with states of several particles, the vacuum for up to nn particles is the tensor product of nn single-particle vacua

∣Ω⟩≡∣Ω⟩n⊗…⊗∣Ω⟩1,(4)|\Omega\rangle\equiv|\Omega\rangle_n \otimes\ldots\otimes|\Omega\rangle_1, \tag*{(4)}

and the memory is considered to be filled from right to left, the first unoccupied register being found using projectors:

Pj(n)∣Ω⟩n⊗…∣Ω⟩k+1⊗∣1η′p′⟩k⊗…⊗∣1ηp⟩1=δj,k∣Ω⟩n⊗…∣Ω⟩k+1⊗∣1η′p′⟩k⊗…⊗∣1ηp⟩1,(5)\mathbb{P}^{(n)}_j |\Omega\rangle_n \otimes\ldots|\Omega\rangle_{k+1} \otimes|1\eta' p'\rangle_k \otimes\ldots\otimes|1\eta p\rangle_1 = \delta_{j,k}|\Omega\rangle_n \otimes\ldots|\Omega\rangle_{k+1} \otimes|1\eta' p'\rangle_k \otimes\ldots\otimes|1\eta p\rangle_1, \tag*{(5)}

which can be written in terms of the set, scrap and control operators described earlier. To implement the appropriate particle statistics we introduce step-antisymmetrizers Aj←(j−1)A_{j\leftarrow(j-1)}, which antisymmetrize register jj with all the others, already assumed antisymmetric. In this way we can write the creation and annihilation operators

ar(n)†=∑j=1nAj←(j−1)⋅P0(n−j)⊗(E10⊗sr†)j⊗Pj−1(j−1),(6)a_r^{(n)\dagger} = \sum_{j=1}^{n} A_{j\leftarrow(j-1)} \cdot\mathbb{P}^{(n-j)}_0 \otimes(\mathcal{E}_{10} \otimes\mathsf{s}^\dagger_r)_j \otimes\mathbb{P}^{(j-1)}_{j-1}, \tag*{(6)}

and

ar(n)=∑j=1nP0(n−j)⊗(E01⊗sr)j⊗Pj−1(j−1)⋅Aj←(j−1),(7)a_r^{(n)} = \sum_{j=1}^{n} \mathbb{P}^{(n-j)}_0 \otimes(\mathcal{E}_{01} \otimes\mathsf{s}_r)_j \otimes\mathbb{P}^{(j-1)}_{j-1} \cdot A_{j\leftarrow(j-1)}, \tag*{(7)}

with a ‘⋅\cdot’ to separate operators over the entire memory and each register acts on the number of registers indicated by its superscript (n)(n), (n−j)(n-j), etc.

{ar(n),as(n)†}=δr,s(E00⊗i)n⊗∑j=0n−1Pj(n−1)+An←n−1⋅(E11⊗sr†ss)n⊗Pn−1(n−1)⋅An←n−1,(8)\left\{a_r^{(n)},a_s^{(n)\dagger}\right\} = \delta_{r,s}(\mathcal{E}_{00}\otimes i)_n \otimes\sum_{j=0}^{n-1}\mathbb{P}^{(n-1)}_j + A_{n\leftarrow n-1}\cdot(\mathcal{E}_{11}\otimes\mathsf{s}^\dagger_r\mathsf{s}_s)_n\otimes\mathbb{P}^{(n-1)}_{n-1}\cdot A_{n\leftarrow n-1}, \tag*{(8)}

with a non-canonical boundary term that vanishes as long as there is at least one empty register, and is equivalent to defining the boundary action of the creation operator as

ar(n)†∣1η′p′⟩n⊗…⊗∣1ηp⟩1=0.(9)a_r^{(n)\dagger}|1\eta' p'\rangle_n\otimes\ldots\otimes|1\eta p\rangle_1=0. \tag*{(9)}

In the examples below the number of particles is kept fixed, so the boundary condition has no effect. Thus we can use Eq. (6) and Eq. (7) to obtain expressions in terms of set and scrap operators that can be exponentiated and applied to the quantum memory. Details are to be found in [21]; we here just state the results for a quantum memory of two registers: the time evolution operator for a kinetic-energy term such as the exponential of the first term of Eq. (25) below can be written as

U11(Δt)=P1(2)+P1(1)⊗(E11⊗U11(Δt))1+(E11⊗L11(Δt))2⊗(E11⊗L11(Δt))1,(10)\mathcal{U}_{11}(\Delta t)=\mathbb{P}^{(2)}_1+\mathbb{P}^{(1)}_1\otimes(\mathcal{E}_{11}\otimes\mathcal{U}_{11}(\Delta t))_1 +(\mathcal{E}_{11}\otimes\mathcal{L}_{11}(\Delta t))_2\otimes(\mathcal{E}_{11}\otimes\mathcal{L}_{11}(\Delta t))_1, \tag*{(10)}

where

L11(Δt)≡exp⁡[−iΔt∑kEk⊗sk†sk].(11)\mathcal{L}_{11}(\Delta t)\equiv\exp\left[-i\Delta t\sum_k \mathcal{E}_k\otimes\mathsf{s}^\dagger_k\mathsf{s}_k\right]. \tag*{(11)}

adds to each register a phase corresponding to its kinetic energy eke_k and the spin operators have been summed to the identity ‘i’. (10) is the free-energy evolution operator (over the entire memory) and it defines how the operators over single registers, (11), should be applied. The implementation is completed once these exponentials are written in terms of basic gates, a decomposition that can be done using built-in functions of programming packages such as qiskit [33].

The exponential of the momentum exchange term of (25) simplifies to

U22(Δt)=P0(2)+P1(2)+{C11⊗C11}{U22,2.1(Δt)},(12)\mathcal{U}_{22}(\Delta t)=\mathbb{P}_{0}^{(2)}+\mathbb{P}_{1}^{(2)}+\{\mathcal{C}_{11}\otimes\mathcal{C}_{11}\}\{\mathcal{U}_{22,2.1}(\Delta t)\}, \tag*{(12)}

where the control operators act on the presence qubits of the first two registers. The register operator is now

U22,2.1(Δt)=e−iΔt∑q∑p,η1;k,η2λq[s~trssη1⊗s~trsuη2+s~trsuη2⊗s~trssη1].(13)\mathcal{U}_{22,2.1}(\Delta t)=e^{-i\Delta t\sum_q\sum_{p,\eta_1;k,\eta_2}\lambda_q\left[\tilde{s}_{t}^{r}s_{s}^{\eta_1}\otimes\tilde{s}_{t}^{r}s_{u}^{\eta_2}+\tilde{s}_{t}^{r}s_{u}^{\eta_2}\otimes\tilde{s}_{t}^{r}s_{s}^{\eta_1}\right]}. \tag*{(13)}

where the set and scrap operators separately act on spin and momentum qubits:

s~r†ss=s~η1†sη1⊗s~q†sp,\tilde{s}_{r}^{\dagger}s_s=\tilde{s}_{\eta_1}^{\dagger}s_{\eta_1}\otimes\tilde{s}_{q}^{\dagger}s_p,
s~t†su=s~η2†sη2⊗s~k†sq,(14)\tilde{s}_{t}^{\dagger}s_u=\tilde{s}_{\eta_2}^{\dagger}s_{\eta_2}\otimes\tilde{s}_{k}^{\dagger}s_q, \tag*{(14)}

so the spin operators can be again summed to the identity.

The total evolution operator is implemented using the Trotter formula

U(Δt)={U11(ΔtnTrotter)U22(ΔtnTrotter)}nTrotter+O(ΔtnTrotter)2,(15)\mathcal{U}(\Delta t)=\left\{\mathcal{U}_{11}\left(\frac{\Delta t}{n_{\mathrm{Trotter}}}\right)\mathcal{U}_{22}\left(\frac{\Delta t}{n_{\mathrm{Trotter}}}\right)\right\}^{n_{\mathrm{Trotter}}}+\mathcal{O}\left(\frac{\Delta t}{n_{\mathrm{Trotter}}}\right)^2, \tag*{(15)}

where we call the time step ΔtnTrotter\frac{\Delta t}{n_{\mathrm{Trotter}}} Trotter interval. The equation allows for the simulation of time evolution with a systematically improvable error.

Antisymmetrization

In the state preparation process or during the evolution of the quantum memory it is sometimes necessary to apply antisymmetrizers over arbitrary states, an operation that is not unitary. To circumvent the issue we impose an ordering to the memory registers using comparisons between binary numbers; we say that ∣1s0p0⟩>∣1s1p1⟩|1s_0p_0\rangle> |1s_1p_1\rangle if the binary representing the state ∣1s0p0⟩|1s_0p_0\rangle is greater than that representing the state ∣1s1p1⟩|1s_1p_1\rangle. This ordering can be used to disentangle the auxiliary qubits that make the operation feasible.

As an example, consider two scalar particles with two momentum qubits in which we codify three different momenta:

∣00⟩→None,∣01⟩→∣p0⟩,|00\rangle\rightarrow\mathrm{None},\quad|01\rangle\rightarrow|p_0\rangle,
∣10⟩→∣p1⟩,∣11⟩→∣p2⟩,(16)|10\rangle\rightarrow|p_1\rangle,\quad|11\rangle\rightarrow|p_2\rangle, \tag*{(16)}

in which the state ∣00⟩|00\rangle has been excluded for technical reasons. Adding the presence qubit in total gives three qubits per particle.

We first initiate the memory in the following superposition (note that p1p_1, whose binary number is larger than that of p0p_0, is written in the second register):

∣ϕ0⟩=12(∣Ω⟩2∣1p0⟩1+∣1p1⟩2∣1p0⟩1),(17)|\phi_0\rangle=\frac{1}{\sqrt{2}}\left(|\Omega\rangle_2|1p_0\rangle_1+|1p_1\rangle_2|1p_0\rangle_1\right), \tag*{(17)}

two auxiliary qubits are now added, the first one associated to the largest state (as defined earlier) if there are at least two active registers:

∣ϕ1⟩=12(∣00⟩ax∣Ω⟩2∣1p0⟩1+∣10⟩ax∣1p1⟩2∣1p0⟩1),(18)|\phi_1\rangle=\frac{1}{\sqrt{2}}\left(|00\rangle_{\mathrm{ax}}|\Omega\rangle_2|1p_0\rangle_1+|10\rangle_{\mathrm{ax}}|1p_1\rangle_2|1p_0\rangle_1\right), \tag*{(18)}

the auxiliary register is antisymmetrized, and SWAP operations and a phase flip are applied to the registers controlling on whether the second auxiliary qubit is 1:

∣ϕ2⟩=12∣00⟩ax∣Ω⟩2∣1p0⟩1+12(∣10⟩ax∣1p1⟩2∣1p0⟩1−∣01⟩ax∣1p0⟩2∣1p1⟩1),(19)|\phi_2\rangle=\frac{1}{\sqrt{2}}|00\rangle_{\mathrm{ax}}|\Omega\rangle_2|1p_0\rangle_1+\frac{1}{2}\left(|10\rangle_{\mathrm{ax}}|1p_1\rangle_2|1p_0\rangle_1-|01\rangle_{\mathrm{ax}}|1p_0\rangle_2|1p_1\rangle_1\right), \tag*{(19)}

the auxiliary qubit is now disentangled with an algorithm called Locate the Largest, which points, on an auxiliary register, to the position of the largest momenta in a quantum memory with at least two registers (this algorithm is described in [21]). Applying the algorithm we get

∣ϕ3⟩=∣00⟩ax[12∣Ω⟩2∣1p0⟩1+12(∣1p1⟩2∣1p0⟩1−∣1p0⟩2∣1p1⟩1)],(20)|\phi_3\rangle= |00\rangle_{ax}\left[\frac{1}{\sqrt{2}}|\Omega\rangle_2|1p_0\rangle_1+\frac{1}{2}\left(|1p_1\rangle_2|1p_0\rangle_1-|1p_0\rangle_2|1p_1\rangle_1\right)\right], \tag*{(20)}

so the memory is antisymmetrized and the auxiliary qubits unentangled. This procedure has been validated on a quantum computer, both via classical simulation with qiskit aer [33] and with the ibm brisbane chip, accessible by the IBM cloud.

Numerical demonstration

Figure 1-Figure 2 are the results of measuring the momentum qubits of the second register of Eq. (20), along with the auxiliary register, following the antisymmetrization process. In the figures, binary digits preceding the dash symbol correspond to auxiliary qubits. The histograms of Figure 1 are the outcomes of a quantum circuit simulation performed using IBM’s Qiskit aer simulator, both with and without the final unentangling step. A similar experiment was conducted on the IBM quantum chip ibm brisbane, with 5000 shots and an error mitigation procedure applied to obtain quasi-probabilities, see Figure 2. The circuits without the uncomputation step (that is, with result Eq. (19)), have a depth of approximately 300 layers (first histogram), and it shows a strong resemblance to the simulator results, Figure 1. However, incorporating the unentangling steps to obtain Eq. (20) increases the circuit depth to around 2000 layers, which exceeds the current capabilities of available quantum computers. The red vertical lines are the exact probabilities, momenta are measured to be in state None (corresponding to an empty register) with probability 1/21/2, while momenta p0p_0 or p1p_1 are measured with probabilities 1/41/4 each.

Measurement of the auxiliary and second registers of the antisymmetrization example

Figure 1. Measurement of the auxiliary and second registers of the antisymmetrization example § 3, binary digits before the dash correspond to auxiliary qubits. The circuits have been executed with the qiskit aer simulator.

Measurement of the auxiliary and second registers of the antisymmetrization example

Figure 2. Measurement of the auxiliary and second registers of the antisymmetrization example Sec. 3, binary digits before the dash correspond to auxiliary qubits. The histograms represent quasi-probabilities from the IBM quantum chip ibm brisbane, see the text for futher details.

Momentum exchange simulation

We now simulate the behaviour of two electrons under a positively charged background in two dimensions. We first state the Hamiltonian, closely following the discussion of Chap. 1, Sec. 3 of [34]. The operator is divided as follows

H^=H^el+H^b+H^el−b,(21)\hat{H} = \hat{H}_{el} + \hat{H}_{b} + \hat{H}_{el-b}, \tag*{(21)}

where H^el\hat{H}_{el} and H^b\hat{H}_{b} are the Hamiltonians for the electrons and the background and H^el−b\hat{H}_{el-b} that of the interaction among the two. Assuming an innert background and a spatially uniform system, H^b\hat{H}_{b} and H^el−b\hat{H}_{el-b} can be demonstrated to be pure c-numbers. Thus we focus on the electron part, which in first quantization reads

H^el=−∑iNeℏ2∇i22m+12e2∑i≠jNee−μ∣ri−rj∣∣ri−rj∣(22)\hat{H}_{el}=-\sum_{i}^{N_e}\frac{\hbar^2\nabla_i^2}{2m}+\frac{1}{2}e^2\sum_{i\ne j}^{N_e}\frac{e^{-\mu|r_i-r_j|}}{|r_i-r_j|} \tag*{(22)}

where the first term is the kinetic energy TT and the second the potential energy VV; NeN_e is the number of electrons, pip_i their momentum, mm their mass, ee their charge and rir_i their positions. μ\mu is a regulator needed to make the potential matrix elements finite, the pure c-numbers mentioned earlier also depend on μ\mu.

The passage to second-quantization is realized through the following equation

H^el=∑rs⟨r∣T∣s⟩ar†as+12∑rstu⟨rs∣V∣tu⟩ar†as†auat(23)\hat{H}_{el}=\sum_{rs}\langle r|T|s\rangle a_r^\dagger a_s+\frac{1}{2}\sum_{rstu}\langle rs|V|tu\rangle a_r^\dagger a_s^\dagger a_u a_t \tag*{(23)}

where rr, ss, ... are to be understood as generic indices encompassing both the momentum and spin degrees of freedom, i.e. ar≡ak1,η1a_r\equiv a_{k_1,\eta_1}, ∣r⟩≡∣k1,η1⟩|r\rangle\equiv|k_1,\eta_1\rangle, etc. The operators fulfill the anticommutation relations

{ak1,η1†,ak2,η2}=δk1,k2δη1,η2,(24)\left\{a_{k_1,\eta_1}^\dagger,a_{k_2,\eta_2}\right\}=\delta_{k_1,k_2}\delta_{\eta_1,\eta_2}, \tag*{(24)}

as we consider a discrete momentum values. To evaluate the matrix elements we approximate the discrete sums over position by integrals, normalized with a total surface SS, ∑x→1S∫d2x\sum_x\to\frac{1}{S}\int d^2x and take as basis functions plane waves with periodic boundary conditions ki=2πniSk_i=\frac{2\pi n_i}{\sqrt{S}}. A characteristic length r0=SNπr_0=\sqrt{\frac{S}{N\pi}} is also introduced to define dimensionless quantities, denoted with bars. The final Hamiltonian reads,

H^=∑kˉ,ηϵkˉakˉη†akˉ,η+∑kˉ1,kˉ2∑η1,η2∑qˉ≠(0,0)λqˉakˉ1+qˉ,η1†akˉ2−qˉ,η2†akˉ2,η2akˉ1,η1,(25)\hat{H}=\sum_{\bar{k},\eta}\epsilon_{\bar{k}}a_{\bar{k}\eta}^\dagger a_{\bar{k},\eta} +\sum_{\bar{k}_1,\bar{k}_2}\sum_{\eta_1,\eta_2}\sum_{\bar{q}\ne(0,0)}\lambda_{\bar{q}}a_{\bar{k}_1+\bar{q},\eta_1}^\dagger a_{\bar{k}_2-\bar{q},\eta_2}^\dagger a_{\bar{k}_2,\eta_2}a_{\bar{k}_1,\eta_1}, \tag*{(25)}

where the condition q≠(0,0)q\ne(0,0) is a consequence of regularization and

ϵk=∣kˉ∣2E02rs2,λq=E0rsNe∣qˉ∣,(26)\epsilon_k=\frac{|\bar{k}|^2E_0}{2r_s^2},\qquad\lambda_q=\frac{E_0}{r_sN_e|\bar{q}|}, \tag*{(26)}

with E0=e2/a0=340E_0=e^2/a_0=340 eV, a0a_0 the Bohr radius and rs=r0/a0r_s=r_0/a_0 a free parameter.

Implementation

The codification is now applied to the Hamiltonian Eq. (25). We consider two particles and two different discretizations with different qubit counts:

a) 1 presence qubit + 1 spin qubit + 3 qubits to represent 5 values of momenta: 5 qubits per particle

Momenta are arranged in the 2D symmetric lattice of Figure 3, where discretization a)a) corresponds to points {C,F,G,H,K}\{C,F,G,H,K\} and discretization b)b) includes all. Once physical states have been mapped to qubit states, the set and scrap operators which enter Eq. (11) and Eq. (13) can be written in terms of Pauli

Representation of the momentum lattice

Figure 3. Representation of the momentum lattice: Points lying on the same circumference have equal kinetic energy. Discretization a)a) corresponds to points {C,F,G,H,K}\{C,F,G,H,K\}, while discretization b)b) includes all. b)b) 1 presence qubit + 1 spin qubit + 4 qubits to represent 13 values of momenta: 6 qubits per particle

Qubit statesLattice labelPhysical state
∣000⟩|000\rangleNoneNone
∣001⟩|001\rangleC∣(0,1)⟩|(0,1)\rangle
∣010⟩|010\rangleF∣(−1,0)⟩|(-1,0)\rangle
∣011⟩|011\rangleG∣(0,0)⟩|(0,0)\rangle
∣100⟩|100\rangleH∣(0,1)⟩|(0,1)\rangle
∣101⟩|101\rangleK∣(0,−1)⟩|(0,-1)\rangle

Table 1. Table of codification a)a) with correspondence between qubit states and momentum states kˉ=4πNe(i,j)\bar{k}=\sqrt{\frac{4\pi}{N_{\mathrm{e}}}}(i,j). The qubit state ∣000⟩|000\rangle does not represent any physical state.

matrices II, XX, YY, ZZ, for example

s^(0,1)†=∣001⟩⟨000∣=I+Z2⊗I+Z2⊗X−iY2,(27)\hat{s}^{\dagger}_{(0,1)}=|001\rangle\langle000|=\frac{I+Z}{2}\otimes\frac{I+Z}{2}\otimes\frac{X-iY}{2}, \tag*{(27)}

the exponentials of the sum of these tensor products can then be decomposed in terms of basic gates and applied to a quantum circuit using the built-in qiskit’s function HamiltonianGate; finally, the Trotter formula Eq. (15) is used to control the errors comming from separate exponentiation of the free and exchange terms of the Hamiltonian Eq. (25). The operator thus obtained is used to evolve, as an example, the two electron state

∣Ψ⟩init=12(∣10⟩aux∣↑p0⟩2∣↓p0⟩1−∣01⟩aux∣↓p0⟩2∣↑p0⟩1),(28)|\Psi\rangle_{\mathrm{init}}=\frac{1}{\sqrt{2}}\left(|10\rangle_{\mathrm{aux}}|\uparrow p_0\rangle_2|\downarrow p_0\rangle_1-|01\rangle_{\mathrm{aux}}|\downarrow p_0\rangle_2|\uparrow p_0\rangle_1\right), \tag*{(28)}

which is obviously not an eigenstate of the Hamiltonian Eq. (25); the probabilities thus change with time, and we keep track of that of measuring p0p_0 on the first register. The auxiliary qubits are not uncomputed because the number of particles does not change and no further antisymmetrization is needed.

Results

To carry out the simulations we used the qiskit’s aer simulator with Ne=2N_{\mathrm{e}}=2 for two electrons and rs=30r_{\mathrm{s}}=30 in Eq. (26) for the kinetic and momentum exchange terms to be comparable in size.

The results are shown in Figure 4, with 100 time points and a Trotter interval of 0.055 for a)a) and 151 time points and a Trotter interval of 0.03 for b)b); the number of circuit shots per time point is always 1000. The first plot confirms our expectations for a)a), where we see a simple oscillatory behaviour, but the composite evolution of b)b) we see in the second plot is not as easy to interpret.

Time evolution and frequency plots for discretizations a) and b)

Figure 4. Probability of measuring momentum p0p_0 in the first register as a function of time for discretizations a)a) (up) and b)b) (down). Orange star-like points are obtained by the qiskit aer simulator of quantum circuits, while the blue line is the result of direct numerical exponentiation of the Hamiltonian matrix by conventional means.

Figure 5 shows the frequency domain results, with 400 frequency points and a Trotter interval of 0.067 for a)a) and 250 frequency points1 and a Trotter interval of again 0.067 for b)b), note that only half of the points really contributes to what we see in the figures, as the amplitudes are symmetric with respect to the ordinate axis. Numerical results for frequencies and periods are shown in Table 2. There we see a single frequency for discretization a)a), as expected for the simple oscillatory behaviour of the time evolution plot. In contrast, there are 5 clear peaks for discretization b)b), but the exact solution shows another at approximately 2.4 eV which is not resolved with only 250 frequency points; thus there are 6 peaks, corresponding to an oscillation around 3 states: the initial state, associated to point G on Figure 3 and two

Fourier transforms for discretizations a) and b)

Figure 5. Fourier transforms of the time-depending probability of measuring the state p0p_0 in the first register for discretizations a)a) (up) and b)b) (down). Orange star-like points are simulated, while the blue line is the result of direct numerical exponentiation.

Frequency (eV)Period (eV−1^{-1})
a1a_11.79±0.041.79 \pm0.040.56±0.010.56 \pm0.01
b1b_10.54±0.060.54 \pm0.061.9±0.21.9 \pm0.2
b2b_20.84±0.060.84 \pm0.061.2±0.11.2 \pm0.1
b3b_31.37±0.061.37 \pm0.060.73±0.030.73 \pm0.03
b4b_43.17±0.063.17 \pm0.060.32±0.010.32 \pm0.01
b5b_53.71±0.063.71 \pm0.060.269±0.0040.269 \pm0.004

Table 2. Frequencies and periods characterizing the evolution of the state Eq. (28) for discretizations a)a) and b)b) discussed in subsection 4.1. Errors correspond to the finite spacing between frequency values.

other states, superpositions of points {C,F,H,K}\{C,F,H,K\} and {A,E,I,M}\{A,E,I,M\}, the system does not oscillate to a superposition of points {B,D,J,L}\{B,D,J,L\}.

Conclusions

We have presented here a register-particle codification for a quantum computer, by which the unitary evolution operators are decomposed in terms acting on a small number of registers each time, and that can be implemented using standard quantum computation protocols. This has been applied to the antisymmetrization and time evolution of a tworegister quantum memory and we have seen that unentangling circuits are at present out of reach of this technology. Classical simulators can still be used to test and debug algorithms, but their applicability is limited to a low number of qubits, and therefore, to a very low number of particles. Finally, we would like to note that with a few more qubits per register this example can be easily extended to incorporate other quantum numbers such as colour, flavour, etc.; with Eq. (23) then promoting to a standard quark model, a step closer to QCD.

Acknowledgements

We would like to thank Felipe J. Llanes Estrada and María Gómez Rocha for a critical reading of the manuscript. Supported by grant PID2022-137003NB-I00 of the Spanish MCIN/AEI /10.13039/501100011033/; EU’s 824093 (STRONG2020); grant FPU21/04180 of the Spanish Ministry of Universities; Universidad Complutense de Madrid under research group 910309 and the IPARCOS institute.

References

  1. [1]P. De Forcrand, Simulating QCD at finite density, Sissa Medialab. doi:10.22323/1.091.0010.
  2. [2]G. Ortiz, J. E. Gubernatis, E. Knill, R. Laflamme, Phys. Rev. A 64 (2). doi:10.1103/PhysRevA.64.022319.DOI
  3. [3]H.-W. Hammer, S. König, U. Van Kolck, Rev. Mod. Phys. 92 (2). doi:10.1103/RevModPhys.92.025004.DOI
  4. [4]F. J. Llanes-Estrada, E. Lope-Oter, Prog. Part. Nucl. Phys. 109. doi:10.1016/j.ppnp.2019.103715.DOI
  5. [5]E. Lope-Oter, EPJ Web Conf. 258. doi:10.1051/epjconf/202225807007.DOI
  6. [6]S. Lloyd, Universal quantum simulators, Science 273 (5278). doi:10.1126/science.273.5278.1073.DOI
  7. [7]M. A. Nielsen, I. L. Chuang, Quantum Computation and Quantum Information, 1st Edition, Cambridge University Press, 2012. doi:10.1017/CBO9780511976667.DOI
  8. [8]R. P. Feynman, Int. J Theor. Phys. 21 (6-7). doi:10.1007/BF02650179.DOI
  9. [9]D. S. Abrams, S. Lloyd, Phys. Rev. Lett. 79 (13). doi:10.1103/PhysRevLett.79.2586.DOI
  10. [10]J. D. Whitfield, J. Biamonte, A. Aspuru-Guzik, Mol. Phys. 109 (5). doi:10.1080/00268976.2011.552441.DOI
  11. [11]S. McArdle, S. Endo, A. Aspuru-Guzik, S. C. Benjamin, et al., Rev. Mod. Phys. 92 (1). doi:10.1103/RevModPhys.92.015003.DOI
  12. [12]C. W. Bauer, Z. Davoudi, A. B. Balantekin, et al., PRX Quantum 4. doi:10.1103/PRXQuantum.4.027001.DOI
  13. [13]S. P. Jordan, K. S. M. Lee, J. Preskill, Science 336 (6085). doi:10.1126/science.1217069.DOI
  14. [14]M. C. Banuls, R. Blatt, J. Catani, A. Celi, et al., EPJ D 74 (8). doi:10.1140/epjd/e2020-100571-8.
  15. [15]R. C. Farrell, I. A. Chernyshev, S. J. M. Powell, et al., Phys. Rev. D 107 (5). doi:10.1103/PhysRevD.107.054512.
  16. [16]J. Barata, N. Mueller, A. Tarasov, R. Venugopalan, Phys. Rev. A 103 (4). doi:10.1103/PhysRevA.103.042410.
  17. [17]W. M. Kirby, S. Hadi, M. Kreshchuk, P. J. Love, Phys. Rev. A 104 (4). doi:10.1103/PhysRevA.104.042607.
  18. [18]W. Qian, R. Basili, S. Pal, G. Luecke, et al., Phys. Rev. Res. 4 (4). doi:10.1103/PhysRevResearch.4.043193.
  19. [19]M. Kreshchuk, W. M. Kirby, G. Goldstein, H. Beauchemin, et al., Phys. Rev. A 105 (3). doi:10.1103/PhysRevA.105.032418.
  20. [20]H. Chawdhry, M. Pellen, SciPost Phys. 15 (5). doi:10.21468/SciPostPhys.15.5.205.
  21. [21]J. J. Gálvez-Viruet, F. J. Llanes-Estrada (2024). doi:10.48550/arxiv.2406.03147.
  22. [22]A. Pérez-Obiol, A. M. Romero, J. Menéndez, A. Rios, et al., Sci. Rep. 13 (1). doi:10.1038/s41598-023-39263-7.
  23. [23]P. Wang, W. Du, W. Zuo, J. P. Vary, Phys. Rev. C 109 (6). doi:10.1103/PhysRevC.109.064623.
  24. [24]D. Gallimore, J. Liao, Phys. Rev. D 107 (7). doi:10.1103/PhysRevD.107.074012.
  25. [25]N. M. de Arenaza, J. J. Gálvez-Viruet, F. J. Llanes-Estrada (2024). doi:10.48550/arxiv.2407.07232.
  26. [26]J. Barata, X. Du, M. Li, W. Qian, et al., Phys. Rev. D 106 (7). doi:10.1103/PhysRevD.106.074013.
  27. [27]J. Barata, X. Du, M. Li, W. Qian, et al., Phys. Rev. D 108 (5). doi:10.1103/PhysRevD.108.056023.
  28. [28]M. Motta, C. Sun, A. T. K. Tan, M. J. O’Rourke, et al., Nat. Phys. 16 (2). doi:10.1038/s41567-019-0704-4.
  29. [29]Z.-Y. Zhou, G.-X. Su, J. C. Halimeh, R. Ott, et al., Science 377 (6603). doi:10.1126/science.abl6277.
  30. [30]Z. Davoudi, N. Mueller, C. Powers, Phys. Rev. Lett. 131 (8). doi:10.1103/PhysRevLett.131.081901.
  31. [31]P. Jordan, E. Wigner, Z. Physik 47 (9-10). doi:10.1007/BF01331938.
  32. [32]S. B. Bravyi, A. Y. Kitaev, Ann. Phys. 298 (1). doi:10.1006/aphy.2002.6254.
  33. [33]A. Javadi-Abhari, M. Treinish, K. Kruslich, C. J. Wood, et al., Quantum computing with Qiskit (2024). doi:10.48550/arXiv.2405.08810.
  34. [34]A. L. Fetter, J. D. Walecka, Quantum theory of many-particle systems, Dover Publications, 2003.

Paper details

vlnc:W_fmn6y8hr
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 5

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents