Purification-based quantum error mitigation of pair-correlated electron simulations

Thomas E. O’Brien, Gian-Luca Anselmetti, Fotios Gkritsis, Vincent E. Elfving, Stefano Polla, William J. Huggins, Oumarou Oumarou, Kostyantyn Kechedzhi, Dmitry A. Abanin, Rajeev Acharya, I. L. Aleǐner, R. M. Allen, T. I. Andersen, K. R. Anderson, M. Ansmann, Frank Arute, Kunal Arya, Abraham Asfaw, Juan Atalaya, Dave Bacon, Joseph C. Bardin, Andreas Bengtsson, Sergio Boixo, Gina Bortoli, Alexandre Bourassa, Jenna Bovaird, L. Brill, Michael Broughton, Bob B. Buckley, David A. Buell, Tim Burger, Brian Burkett, Nicholas Bushnell, Juan Campero, Yu Chen, Zijun Chen, B. Chiaro, Desmond Chik, Josh Cogan, Roberto Collins, Paul Conner, William Courtney, Alexander L. Crook, Ben Curtin, Dripto M. Debroy, Alexander Del Toro Barba, Sean Demura, Ilya Drozdov, A. Dunsworth, Daniel Eppens, Catherine Erickson, Lara Faoro, Edward Farhi, Reza Fatemi, Vinicius Ferreira, Leslie Flores Burgos, Ebrahim Forati, Austin G. Fowler, Brooks Foxen, William Giang, Craig Gidney, D. Gilboa, Marissa Giustina, Raja Gosula, Alejandro Grajales Dau, Jonathan A. Gross, Steve Habegger, Michael C. Hamilton, Monica Hansen, Matthew P. Harrigan, Sean D. Harrington, Paula Heu, Jeremy Hilton, M. R. Hoffmann, Sabrina Hong, Trent Huang, Ashley Huff, L. B. Ioffe, Sergei V. Isakov, Justin Iveland, E. Jeffrey, Jiang Zhang, Cody Jones, Pavol Juhás, Dvir Kafri, J. Kelly, Tanuj Khattar, Mostafa Khezri, Marika Kieferova, Seon Kim, Paul V. Klimov, Andrey Klots, Alexander N. Korotkov, Fedor Kostritsa, John Mark Kreikebaum, David Landhuis, Pavel Laptev, Kim-Ming Lau, Lily Laws, Joonho LeeResearch Square 2022 · DOI

Abstract

An important measure of the development of quantum computing platforms has been the simulation of increasingly complex physical systems. Prior to fault-tolerant quantum computing, robust error mitigation strategies are necessary to continue this growth. Here, we study physical simulation within the seniority-zero electron pairing subspace, which affords both a computational stepping stone to a fully correlated model, and an opportunity to validate recently introduced “purification-based” error-mitigation strategies. We compare the performance of error mitigation based on doubling quantum resources in time (echo verification) or in space (virtual distillation), on up to 20 qubits of a superconducting qubit quantum processor. We observe a reduction of error by one to two orders of magnitude below less sophisticated techniques (e.g. post-selection); the gain from error mitigation is seen to increase with the system size. Employing these error mitigation strategies enables the implementation of the largest variational algorithm for a correlated chemistry system to-date. Extrapolating performance from these results allows us to estimate minimum requirements for a beyond-classical simulation of electronic structure. We find that, despite the impressive gains from purification-based error mitigation, significant hardware improvements will be required for classically intractable variational chemistry simulations.

Introduction: Bridging the Gap in the NISQ Era

Simulating quantum chemistry is a primary objective for quantum computing. However, we currently operate in the Noisy Intermediate-Scale Quantum (NISQ) era, where hardware is susceptible to errors from decoherence and gate imperfections. While fault-tolerance is the ultimate goal, current research focuses on error mitigation to extract useful results from noisy devices. This study explores "purification-based" strategies to achieve the precision required for chemical accuracy in complex, correlated electron simulations.

Methodology: Advanced Error Mitigation Strategies

The research focuses on the seniority-zero subspace, where electrons are paired, using a Unitary pair Coupled Cluster Doubles (UpCCD) ansatz. The team compared three distinct mitigation levels:

  1. Post-selection (PS): A baseline method that discards results violating physical constraints, such as incorrect particle counts.
  2. Echo Verification (EV): This technique doubles resources in time by running the circuit forward and then "echoing" it backward. Deviations from the initial state are used to correct errors.
  3. Virtual Distillation (VD): This method doubles resources in space by preparing two simultaneous copies of the quantum state and performing a collective measurement to "distill" the pure state from the noisy mixture.

Experimental Results: Scaling and Accuracy

The methods were tested on two systems: the Richardson-Gaudin (RG) model (scaled up to 20 qubits) and the cyclobutene ring-opening reaction (6 and 10 qubits).

Key Findings:

  • Massive Error Reduction: Echo Verification achieved an average 85-fold reduction in error compared to raw data, peaking at 460-fold in the RG model.
  • Scaling Advantage: The benefits of purification-based mitigation were found to increase with system size, suggesting viability for larger molecules.
  • Chemical Accuracy: In the 6-qubit cyclobutene simulation, the team achieved an error of less than 0.011 Hartree, meeting the threshold for chemical accuracy.
  • Capturing Physical Nuance: The error-mitigated simulation successfully reproduced a physical "cusp" in the RG model's order parameter that classical approximations like BCS theory fail to capture.

Discussion: The Road to Beyond-Classical Simulation

Despite these successes, challenges remain. "Device drift"—small fluctuations in hardware performance over time—remains a significant error source. Furthermore, purification techniques involve a high sampling cost; a 50-qubit experiment could require up to $10^9$ shots.

Extrapolating from these results, the authors conclude that a 25-fold reduction in hardware error rates is necessary to reach a "beyond-classical" regime for variational chemistry simulations. Future efforts will focus on optimizing circuit depth, improving hardware stability, and integrating classical pre-optimization.

Summary updated

Abstract

<title>Abstract</title> An important measure of the development of quantum computing platforms has been the simulation of increasingly complex physical systems [1–3]. Prior to fault-tolerant quantum computing, robust error mitigation strategies are necessary to continue this growth [4–11]. Here, we study physical simulation within the seniority-zero electron pairing subspace, which affords both a computational stepping stone to a fully correlated model [12–17], and an opportunity to validate recently introduced “purification-based” error-mitigation strategies [8–10]. We compare the performance of error mitigation based on doubling quantum resources in time (echo verification [10]) or in space (virtual distillation [8, 9]), on up to 20 qubits of a superconducting qubit quantum processor. We observe a reduction of error by one to two orders of magnitude below less sophisticated techniques (e.g. post-selection); the gain from error mitigation is seen to increase with the system size. Employing these error mitigation strategies enables the implementation of the largest variational algorithm for a correlated chemistry system to-date. Extrapolating performance from these results allows us to estimate minimum requirements for a beyond-classical simulation of electronic structure. We find that, despite the impressive gains from purification-based error mitigation, significant hardware improvements will be required for classically intractable variational chemistry simulations.
BibTeX
@inproceedings{obrien2022purification,
  author = {Thomas E. O’Brien and Gian-Luca Anselmetti and Fotios Gkritsis and Vincent E. Elfving and Stefano Polla and William J. Huggins and Oumarou Oumarou and Kostyantyn Kechedzhi and Dmitry A. Abanin and Rajeev Acharya and I. L. Aleǐner and R. M. Allen and T. I. Andersen and K. R. Anderson and M. Ansmann and Frank Arute and Kunal Arya and Abraham Asfaw and Juan Atalaya and Dave Bacon and Joseph C. Bardin and Andreas Bengtsson and Sergio Boixo and Gina Bortoli and Alexandre Bourassa and Jenna Bovaird and L. Brill and Michael Broughton and Bob B. Buckley and David A. Buell and Tim Burger and Brian Burkett and Nicholas Bushnell and Juan Campero and Yu Chen and Zijun Chen and B. Chiaro and Desmond Chik and Josh Cogan and Roberto Collins and Paul Conner and William Courtney and Alexander L. Crook and Ben Curtin and Dripto M. Debroy and Alexander Del Toro Barba and Sean Demura and Ilya Drozdov and A. Dunsworth and Daniel Eppens and Catherine Erickson and Lara Faoro and Edward Farhi and Reza Fatemi and Vinicius Ferreira and Leslie Flores Burgos and Ebrahim Forati and Austin G. Fowler and Brooks Foxen and William Giang and Craig Gidney and D. Gilboa and Marissa Giustina and Raja Gosula and Alejandro Grajales Dau and Jonathan A. Gross and Steve Habegger and Michael C. Hamilton and Monica Hansen and Matthew P. Harrigan and Sean D. Harrington and Paula Heu and Jeremy Hilton and M. R. Hoffmann and Sabrina Hong and Trent Huang and Ashley Huff and L. B. Ioffe and Sergei V. Isakov and Justin Iveland and E. Jeffrey and Jiang Zhang and Cody Jones and Pavol Juhás and Dvir Kafri and J. Kelly and Tanuj Khattar and Mostafa Khezri and Marika Kieferova and Seon Kim and Paul V. Klimov and Andrey Klots and Alexander N. Korotkov and Fedor Kostritsa and John Mark Kreikebaum and David Landhuis and Pavel Laptev and Kim-Ming Lau and Lily Laws and Joonho Lee},
  title = {{Purification-based quantum error mitigation of pair-correlated electron simulations}},
  booktitle = {Research Square},
  year = {2022},
  doi = {10.21203/rs.3.rs-2185180/v1},
}
Figure