1 / 32

Julia Kempe CNRS & LRI, Univ. of Paris-Sud Alexei Kitaev Caltech Oded Regev Tel-Aviv University

FSTTCS, Chennai, December 18 th , 2004. The Complexity of the Local Hamiltonian Problem. Julia Kempe CNRS & LRI, Univ. of Paris-Sud Alexei Kitaev Caltech Oded Regev Tel-Aviv University. Also implies:. 2-local adiabatic computation is equivalent to standard quantum computation. Results.

marged
Download Presentation

Julia Kempe CNRS & LRI, Univ. of Paris-Sud Alexei Kitaev Caltech Oded Regev Tel-Aviv University

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. FSTTCS, Chennai, December 18th, 2004 The Complexity of the Local Hamiltonian Problem Julia Kempe CNRS & LRI, Univ. of Paris-Sud Alexei Kitaev Caltech Oded Regev Tel-Aviv University

  2. Also implies: 2-local adiabatic computation is equivalent to standard quantum computation Results • Result: • 2-local Hamiltonian is QMA complete

  3. Outline • Introduction • Quantum computing • QMA • Local Hamiltonians • Previous Constructions • The 3-qubit Gadget • Implications • Adiabatic computation • Other applications of the technique

  4. U |’ = U | | Quantum Computation • A qubit is described by a unit vector in two-dimensional space: • | = c0|0 + c1|1 such that |c0|2+|c1|2=1 • |0 and |1 are simply two orthogonal vectors • An n-qubit system is described by a unit vector in a 2n dimensional space: • |C{0,1}nsuch that|| |||2=1 • An operation on an n-qubit system is described by a unitary matrix: • U C2n2nsuch that UU†=I(i.e., unitary)

  5. Quantum Computation • The model of quantum computation is as strong as classical computation • Moreover, there exists a small set of quantum gates that are universal [Deutsch ’95, Barenco et al. ’95, DiVincenzo’95] |0 |0 |1 U CNOT H “Hadamard” • Quantum complexity theory is born!

  6. V V 1 (accept) exists witness: y The class NP • NP – Nondeterministic Polynomial Time • Def: L  NP if there is a poly-time verifier V and a polynomial p s.t. “yes” instance: x  L “no” instance: x  L 0 (reject) for all y • Cook-Levin Theorem: SAT is NP-complete

  7. Def: L  QMA if there is a poly-time quantum • verifier U and a polynomial p s.t. U U exists witness | prob  0 (reject) for all | prob  1 (accept) The class QMA • QMA – Quantum Merlin Arthur “yes” instance: x  L 1 (accept) prob 1- “no” instance: x  L 0 (reject) prob 1-

  8. Local Hamiltonian Problem Kitaev’s quantum Cook-Levin Theorem (’99): Local Hamiltonian is QMA-complete. • Def. k-local Hamiltonian problem: • Input: k-local Hamiltonian , , Hi acts on k • qubits, a<b constants • Promise: • Smallest eigenvalue of H either  a or  b (b-a const.) • Output: • 1 if H has eigenvalue  a • 0 if all eigenvalues of H  b “witness = ground state”

  9. Intuition: Formula: , Hamiltonians: H2 local Hamiltonians H1 Satisfying assignment is groundstate of • Energy-penalty 1 for each unsatisfied constraint. • x1x2 … xn| H |x1x2 … xn  = #unsatisfied constraints Local Hamiltonian Problem Penalties for: x1x2x3 = 010 x3x4x5 = 100 …

  10. MAX2SAT is NP-complete • 2-local Hamiltonian is NP-hard Results Classical Quantum • log|x|-local Hamiltonian is QMA-compl. • [Kitaev’99] • 5-local Hamiltonian is QMA-complete • [Kitaev’99] • 3-local Hamiltonian is QMA-complete • [KempeRegev’02] • MAX3SAT is NP-complete 2-local Hamiltonian ?? New result: 2-local Hamiltonian is QMA-complete • 1-local Hamiltonian is in P

  11. Adiabatic Computation • Quantum computers can simulate adiabatic computation [Farhi et al. 00] • Adiabatic computation can simulate quantum computers [ADKLLR 04] • In fact, 3-local adiabatic computation is enough [ADKLLR 04] • New result: 2-local adiabatic computation can simulate quantum computers

  12. Outline • Introduction • Quantum Computation • QMA • Local Hamiltonians • Previous Constructions • The 3-qubit Gadget • Implications • Adiabatic computation • Other applications of the technique

  13. V Vx Classical Cook-Levin Theorem • Thm: SAT is NP-complete • Proof: First, given a verifier V, encode the input into V x1 x2 … y1 y2 … 0 0 … y1 y2 … 0 0 … input x 1 1 witness y witness y ancilla 0 ancilla 0

  14. propagationclauses output clause Classical Cook-Levin Theorem • Thm: SAT is NP-complete • Proof: Next, create a tableau of variables and 3 kinds of clauses. z01 ancilla clauses z02 z03 z04 ancilla z0N z1N z2N zTN time = 0 1 2 3 4 … T

  15. | |0 |0 … |1 Ux ancilla qubits Quantum Cook-Levin Theorem • Let us try to extend this to the quantum setting

  16. output clause Quantum Cook-Levin Theorem • Let us try to extend this to the quantum setting • The naïve attempt does not work • There is no local way to check local consistency z01 ancilla clauses z02 z03 propagationclauses   z04 ancilla qubits z0N zTN |0|1 |2 … |T

  17. output clause Quantum Cook-Levin Theorem • Instead of tensoring the columns, we put them in superposition • So the witness is a sum over history |0|1|2 … |T z01 ancilla clauses z02 z03 propagationclauses z04 ancilla qubits z0N zTN | |0|0+|1|1+…+|T|T

  18. input Time register {|0, |1,…, |T} Computation qubits • propagation • output Quantum Cook-Levin Theorem Thm [Kitaev]: Local Hamiltonian is QMA-complete Proof: Expect the witness described before. Construct the following Hamiltonians. H= Jin Hin + Jprop Hprop + Hout

  19. t T-t Reducing Locality Notice that we have log-local terms: Thm [Kitaev]: 5-local Hamiltonian is QMA-complete Proof idea: Use unary encoding |t | 11…100…0 Penalise illegal time states: Sclock - space of legal time-states is preserved (invariant) ▪ Thm [KempeRegev]: 3-local Hamiltonian is QMA-complt. |tt|  |1010|t,t+1 |tt-1|  |110100|t-1,t,t+1

  20. Outline • Introduction • Quantum Computation • QMA • Local Hamiltonians • Previous Constructions • The 3-qubit Gadget • Implications • Adiabatic computation • Other applications of the technique

  21. Spectrum: H H’ = H + V Energy gap: ||H||>>||V|| 0 groundspace S Three-qubit gadget Idea: use perturbation theory to obtain effective 3-local Hamiltonians from 2-local ones by restricting to subspaces What is the effective Hamiltonian in the lower part of the spectrum?

  22. Energy gap> ||V|| V--- restriction of V to S V++ - restriction of V to S S S Perturbation Theory Spectrum: H H’ = H + V S Energy gap: ||H||>>||V|| 0 groundspace S What is the effective Hamiltonian in the lower part of the spectrum?

  23. Energy gap> ||V|| V--- restriction of V to S V++ - restriction of V to S S S Theorem: Perturbation Theory Spectrum: H H’ = H + V S Energy gap: ||H||>>||V|| 0 groundspace S The lower spectrum of H’ is close to the spectrum of Heff (under certain conditions).

  24. Heff=P1P2P3 3-local H=P1P2P3 3-local 2 2 3 3 P2XB P3XC ZZ B C ZZ ZZ A P1XA 1 1 Terms in H’=H+V are 2-local Fine-tune the energy gap = -3 Three-qubit gadget

  25. ZZ B C ZZ ZZ A Three-qubit gadget S={|001,|010,|100, |110,|101,|011} =-3 Energy gap: S={|000, |111} 0

  26. P2XB |100  |110 Ex.: P1XA P3XC |000 |111 Three-qubit gadget 2 3 P2XB P3XC B C S={|001,|010,|100, |110,|101,|011} =-3 Energy gap: A P1XA S={|000, |111} 0 1 V++ S  S V-+ V+- Third order: S S Theorem:

  27. 2-local Hamiltonian is QMA-complete • Start with the QMA-complete 3-local Hamiltonian • Replace each 3-local term by a 3-qubit gadget

  28. Outline • Introduction • Quantum Computation • QMA • Local Hamiltonians • Previous Constructions • The 3-qubit Gadget • Implications • Adiabatic computation • Other applications of the technique

  29. Adiabatic theorem: g(t) gap between ground- and first excitedstate If then the final state arbitrarily close to groundstate of HP. Idea of Adiabatic Computation* • Start in the groundstate of a Hamiltonian H0 (easy to prepare) • Encode problem as a Hamiltonian HP (groundstate gives solution) • Adiabatically (slowly!) evolve from H0 to HP *E. Farhi, J. Goldstone, S. Gutmann, M. Sipser:“Quantum Computation by Adiabatic Evolution”, q-p/’00

  30. U Adiabatic simulation*: • Hfinal • groundstate = • Kitaev’s “history state” • Hinitial • groundstate • |0…0 |0 T’=poly(T): The gap between groundstate and first excited state is 1/poly(T) at all times. H(t) = (1-t/T’)Hinitial +t/T’ Hfinal Adiabatic Computation simulates quantum computation Standard quantum circuit: |0…0 |T T gates *D. Aharonov, W.  van Dam, J. Kempe, Z. Landau, S. Lloyd, O. Regev: "Adiabatic Quantum Computation is Equivalent to Standard Quantum Computation", FOCS’04, p.42-51

  31. Adiabatic simulation*: • Hfinal • groundstate = • Kitaev’s “history state” • Hinitial • groundstate • |0…0 |0 H(t) = (1-t/T’)Hinitial +t/T’ Hfinal Adiabatic Computation simulates quantum computation Result: 2-local adiabatic computation is equivalent to standard quantum computation Use the gadget to replace everything by 2-local terms.

  32. -2ZA -1P1XA -1P2XA H=P1P2 Heff=P1P2 “Proxy Interaction”: (with A. Landahl) -1Z1ZA -1X2XB -2YAYB Heff=Z1X2 Other applications of the gadget(work in progress) “Interaction at a distance”: H=Z1X2 only XX,YY,ZZ available Useful for Hamiltonian-based quantum architectures

More Related