###### tags: `papers` `Quantum Calculation` # Bravyi-Kitaev Superfast simulation of electronic structure on a quantum computer [arXiv:1712.00446v3](https://arxiv.org/pdf/1712.00446.pdf) This paper shows the results of hydrogen molevule simulations by BKSF and compare them to that of JW and BK transformations. ## Introduction * Brief explantions about quantum simulations * Mapping strategies used (e.g. JW, BK, BKSF) in quantum simulations * BKSF is differnt from spin-fermion mappings (JW and BK) in sence that qubits are representing interactions between fermionnic-modes. (In JW and BK, qubit is representing fermionnic-modes itself.) * The goal of this paper * explain the methodology of BKSF * consider the relation between BKSF and quantum error corrections * simulate hydrogen molecule by BKSF ## Background for quantum simulation ### Fermionic Systems and Second Quantization Explanation of second quantizations, that is seen in many textbooks * definitions of anihilation and creation operators * N body hamiltonian representations by anihilation and creation operators ### Implementing qubit Hamiltonians on quantum computers * How to implement $e^{iHt}$ to represent time evolutions. * If hamiltonian $H$ is devided into terms that are mutually comutable, this can be done easly. * In general the terms are not comutable, thus we use trotter approzimations. ### Phase Estimation Algorithms * Breif expalnations about PEA to calculate the ground energy of the hamiltonian. * In molecular simulations $U = e^{iHt}$ and $v$ is at ground state. $U$ is not trival so prepared by suzuki-trotter decompositions. * Measure the upper-resister to read the phase of $U$. ![](https://i.imgur.com/B09SE0c.png) ### State preparation State preparation $v$ is not trivial. The solutions to this problem will be: * instead prepare Hartree-Fock or post Hartree-Fock state (which has overlap between ground state) and use PEA for polynomial timses. * Methods using adiabatic state preparation. * VQE ### OpenFermion Explanations about openfermion library. * python based tool set, previously called fermilib * open source project providing the classical routines necessary for simulating fermionic models and quantum chemistry problems on quantum computers * support pis4 and projectQ * can comlile the hamiltonian to hardwares ## Transformations from fermionic operators to qubit operators ### Jordan-Wigner and Bravyi-Kitaev Transformation Explanation about JW and BK transformations. * Mapping from spin to fermion * JW and the alternative way (parity nethod) costs $O(n)$ * Bravyi-Kitaev striles the balance to cost only $O(\log n)$ where $n$ is number of qubits. ### Bravyi Kitaev Superfast (BKSF) Algorithm * draws imspiration from Kitaev's honeycomb methods (where majorana modes are used to diagonalize the hamiltonian) * pair the Majorana fermions which lead physically unpaired Majorana modes that cannot interact with environment * Summary of BKSF * map the modes to vertices of graphs * get the number of edges * find the representation of edge operator * find the representation of quubit operator * find independent loops in the graph and define stabilizers for each of them * find the relevant subspace for quantum simulation and the vacuum stat * stabilzer is a subspace where algebra of majorana modes satisfy ### Stabilizers and Vacuum State How to construct stabilizers. > Details will be summarized later. Need to learn about stabilaizers, majorana modes. ### BKSF representation of hamiltonian How to construct number operators, coulomb-exchange operators, excitation operator, number-excitation operator, double excitation operator. These operators are constructed using edge operators. > Details will be summarized later. ## Electronic hamiltonian of hydrogen molecule Construct hydrogen molecule using BKSF representation and compare to the JW and BK transformations. * calculate the one-body and two-body terms individually * construct the interaction hamiltonian as graph * calculate the creation and anihilation operators in terms of edge operators using the integrals in table II ![](https://i.imgur.com/hL0muEw.png) ## Results * The hamiltonian derived in the three differnt ways is presented at below. ![](https://i.imgur.com/JkiV8FC.png) ![](https://i.imgur.com/KnzoUE5.png) ![](https://i.imgur.com/SRTgm45.png) * Calculated the propagetor $e^{iHt}$ by trotter approximations. * Compared the ground energy wrt oredering of suzuki-trotter approximations. Simillar results were derived using BKSF, BK. ![](https://i.imgur.com/pP0Ojb8.png) * Compared the "magnitude ordering" of three mappings. The three mapping acheived the same errors. ![](https://i.imgur.com/CSvKg6Z.png) * The gate count were higher than the BK but lower than the JW. * BKSF might be sufficient comparaed to JW if the interaction in the hamiltonian is localized * Extend investigations to larger molecules * Explore the effects of noise of the BKSF, which have close connections to quantum error corrections ## Conclusions Compared the trotter errors on H$_2$ molecule using BKSF, JW and BK. ## My thoughts and impressions * In abstract it saids that BKSF had lower trootter errors than JW and BK, but I rather think, according to this paper, that they're the same. * I am also interested in the other molecule simulation. * What does physically mean by hamiltonian is locallized? * I haven't understand the details of BKSF and its physics so I want to try learning it someday. (Paper itself seems very helpful for understanding.) ## Applications [openfermion BKSF by google](https://quantumai.google/reference/python/openfermion/transforms/bravyi_kitaev_fast)