• Abstract and Figures
• Public Full-text

Quantum Chemistry Simulations of Dominant Products in Lithium-Sulfur Batteries Julia E. Rice,1 Tanvi P. Gujarati,1 Tyler Y. Takeshita,2 Joe Latone,1 Mario Motta,1 Andreas Hintennach,3 and Jeannette M. Garcia1 1IBM Almaden Research Center, 650 Harry Road, San Jose, CA 95120, USA 2Mercedes Benz Research and Development North America, Sunnyvale, CA 94085 3Daimler AG (Mercedes-Benz Cars), Group Research, HPC G012-BB, 71034, Boeblingen, Germany Quantum chemistry simulations of four industrially relevant molecules are reported. Dissociation curves and dipole moments are reported for lithium hydride (LiH), hydrogen sulfide (H2S), lithium hydrogen sulfide (LiSH) and lithium sulfide (Li2S). Herein, we demonstrate the ability to calculate dipole moments using up to 21 qubits on a quantum simulator for a lithium sulfur salt molecule, and demonstrate the ability to calculate the dipole moment of the LiH molecule on the IBM Q Valencia device using four qubits. This is the first example to the best of our knowledge of dipole moment calculations being performed on quantum hardware. INTRODUCTION mechanism for production and identification of reactive inter- mediates must be validated. However, the reaction mechanism Lithium-sulfur batteries are a promising next-generation for sulfur reduction in the battery environment is highly com- battery technology with a high theoreti-cal capacity of up to plex and debated in the field. Intermediates are difficult (if not ∼1675 mAh/g and high theoretical energy density of ∼2600 impossible) to characterize during operation of the battery and Wh/kg (Li-ion theoretical energy density is ∼350-500 Wh/kg) competing chemical (vs electrochemical) pathways are also [1–3]. The reversible battery functioning is based on the elec- active, which confounds the mixture of products produced. trochemistry between lithium metal (Li0) and elemental sul- Generally, products are assumed to be produced though either one electron or two electron processes, which generate radi- fur (S8) to form lithium sulfide (Li2S) as the thermodynamic product, as illustrated in Figure 1. The reduction of sulfur is cals and Li-S salts, in which the sulfur anion consists of 1-8 known to proceed on discharge in the battery, through a se- sulfur atoms. Four electron processes have also been consid- ries of intermediate lithium polysulfide salts, some of which ered but are unlikely processes as there is little experimental are soluble in the electrolyte and can diffuse to the lithium evidence and for entropic reasons [4]. metal anode. The solubility of various intermediates provides The electron-cloud density distribution of molecules, and unique challenges for the realization of lithium-sulfur batter- particularly their dipole moment, are critical for understand- ies and limits its potential use as a practical alternative to state- ing a variety of phenomena occurring in batteries. In gen- of-the-art lithium-ion chemistries. eral, molecules with high polarity can easily attract or re- pel valence electrons from other compounds and generate re- actions through electron transfer. The dipole moment of a molecule also determines its response to an external electric field. Accurate computationof energetics and dipole moments of molecules is thus a problem with deep conceptual impor- tance, and significant applicability to the chemistry of LiS bat- teries. Achieving this goal requires solving the Schr¨odinger equation for the moleculesof interest, a problem that is known to be exponentially expensive for classical computers, unless approximation schemes are introduced. Quantum computing is a mode of attack of mathematical problems, that has an enormouspotential to provide advantage arXiv:2001.01120v1 [physics.chem-ph] 4 Jan 2020 over conventional computing in a number of areas, including quantum chemistry [5–7]. A number of heuristics to provide approximate but highly accurate solutions to the Schr¨odinger equation have been proposed, in particular the Variational Quantum Eigensolver (VQE) [8–10]. IBM researchers have FIG. 1. The chemistry of a lithium-sulfur battery. During discharge, demonstrated the use and accuracy of VQE in investigations elemental sulfur is reduced into sulfur salts containing chains of vary- ing lengths, ultimately resulting in lithium sulfide production. Dur- of a variety of molecules [11]. ing charge, the chemistry is reversed. The mechanism for lithium Motivated by these successes, and by the importance of sulfide production is proposed to occur through a two-electron pro- computing energies and electrostatic properties, in this work cess. we assess the performance of quantum algorithms in deter- mining ground state energies and dipole moments along bond In order to harness control over the battery technology, the stretching for LiH, H2S, LiSH and Li2S. 2 The calculations presented here are important for the devel- quantum processor, Valencia, where we employed readout- opment of quantum computing: to the best of our knowledge, error mitigation using Qiskit Ignis to correct measurement er- we present the first hard-ware simulation of dipole moments. rors. We also used a simple noise extrapolation scheme using Our results immediately generalize to expectation values of k- additional CNOT gates at the minimum energyVQE iterations body operators in materials, with significant implications for to account for errors introduced during the expensive 2-qubit the ability to study reactivity and electrostatics in batteries by entangling operations as was shown in Refs [24, 25]. quantum algorithms. RESULTS AND DISCUSSION METHODOLOGY Quantum computing simulations on lithium-sulfur molecules The overall strategy for the calculations performed in this work involved initial preprocessing by classical quantum chemistry codes on conventional computers, to generate op- RHF CCSD q-UCCSD FCI timized mean-field orbitals and matrix elements of the Hamil- -7.3 0.3 A B tonian prior to performing computations with quantum simu- -7.4 0.2 lators or devices. The restricted Hartree-Fock (RHF) singlet -7.5 0.1 Ha] m state has been chosen as the initial state for all of the calcula- [ [Ha] -7.6 0.0 FCI E tions described here, since experience has indicated this state E -7.7 -0.1 − E as a good choice for a variety of chemical problems [12]. Re- -7.8 -0.2 stricted coupled-cluster with singles and doubles (CCSD) cal- -7.9 -0.3 culations were performed using Psi4 [13], at STO-3G level C D 5.0 0.2 of theory, using the frozen-core approximation for correlated ] 4.0 0.1 D [ calculations. The choice of the minimal basis is motivated by ] FCI D [ 3.0 0.0 the fact that only a few molecular orbitals can be simulated on D D 2.0 -0.1 − contemporary quantum hardware, since the number of avail- D able qubits is still relatively small, devices are noisy, and full 1.0 -0.2 quantum error correction techniques are not yet available. Ad- 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 -0.3 ˚ ˚ ditional details for the studied molecules are listed in Table I RLiH A RLiH A in the Supplemental Information. Having selected a set of single-electron orbitals for each of FIG. 2. Fig. 2 Calculations for lithium hydride, LiH. (A) Disso- ciation curve (ground state energy in Hartrees, as a function of in- the studied species, VQE computations were performed with teratomic distance, in angstroms) from RHF, CCSD, q-UCCSD and quantum simulators and devices. We use IBM’s open-source FCI. (B) Deviation from FCI energy, for CCSD and q-UCCSD, in Python library for quantum computing, Qiskit [14]. Qiskit millihartrees. (C) Dipole moment (in Debye) as a function of inter- provides tools for various tasks such as creating quantum cir- atomic distance, in angstroms). (D) Deviation from FCI dipole mo- cuits, performing simulations, and computations on real hard- ment, for CCSD and q-UCCSD, in Debye. Ha and D denote Hartree ware. Qiskit Aqua contains an implementation of the VQE and Debye respectively. algorithm, a hybrid quantum-classical algorithm that uses both quantum and classical resources to solve the Schr¨odinger The lithium hydride (LiH) bond can be represented by equation. It also provides a classical exact eigensolver algo- molecular (spatial) orbitals (MOs) with 2s, 2pz character for rithm (FCI) to compare results. lithium and an MO with 1s character for hydrogen, for a total In the VQE algorithm, we take our wavefunction in the of 6 spin-orbitals. The ground state energy and dipole mo- form of a quantum circuit, which is the unitary coupled cluster ments were calculated on the simulator through the use of with singles and doubles (UCCSD) [15–17], its quantum vari- six qubits and the hardware with four qubits after tapering ant (q-UCCSD) as defined in Ref [18] and the Ry variational [26, 27] techniques were applied. form provided in Qiskit [14]. As seen in Figure 2, results from CCSD, q-UCCSD and We then minimize the expectation value of the Hamiltonian FCI coincide for both energy and dipole moment along bond with respect to the parameters of our circuit. The minimiza- stretching, due to the fact that CCSD and q-UCCSD are equiv- tion is carried out through the classical optimization method, alent to FCI for two-electron systems. In the case of q- L BFGS B[19–21] on the simulator, and Simultaneous Per- UCCSD some small deviations from FCI are observed. Of turbation Stochastic Approximation (SPSA) [22, 23] on the course, with molecules that contain more than two electrons, device. Once the VQE is complete, we get the optimized vari- different results are observed for energies and other proper- ational form and the upper bound for the ground state energy. ties. Despite its simplicity, LiH has an interesting evolution in We ran our experiments on both the statevector and qasm the dipole moment along the dissociation curve from a regime simulators in Qiskit.

Load more

  7

Copy Link