rigorous fault tolerance thresholds
play

Rigorous fault-tolerance thresholds Ben Reichardt UC Berkeley N - PowerPoint PPT Presentation

Rigorous fault-tolerance thresholds Ben Reichardt UC Berkeley N gate circuit 0/1 N gate circuit Need error 1/N 1/0 Quantum fault-tolerance problem Classical fault-tolerance: Von Neumann (1956) 0/1 Fault-tolerant, larger C High


  1. Rigorous fault-tolerance thresholds Ben Reichardt UC Berkeley

  2. N gate circuit 0/1

  3. N gate circuit Need error 1/N 1/0

  4. Quantum fault-tolerance problem – Classical fault-tolerance: Von Neumann (1956) 0/1 Fault-tolerant, larger C High tolerable noise  Low overhead 

  5. Quantum fault-tolerance problem – Classical fault-tolerance: Von Neumann (1956) 0/1 C Important problem! Fault-tolerant, larger High tolerable noise  Low overhead 

  6. Work on encoded data  Intuition Correct errors to prevent spread  Concatenate procedure for arbitrary reliability  EC EC • Quantum fault-tolerance: Shor (1996) – Using poly(log N)-sized code, tolerate 1/poly(log N) error • Aharonov & Ben-Or (‘97), Kitaev (‘97), Knill-Laflamme-Zurek (‘97) – Using concatenated constant-sized code, tolerate constant error

  7. Work on encoded data  Intuition Correct errors to prevent spread  Concatenate procedure for arbitrary reliability  EC EC • Quantum fault-tolerance: Shor (1996) – Using poly(log N)-sized code, tolerate 1/poly(log N) error • Aharonov & Ben-Or (‘97), Kitaev (‘97), Knill-Laflamme-Zurek (‘97) – Using concatenated constant-sized code, tolerate constant error

  8. Concatenation 0.0050.010.0150.020.0250.0050.010.0150.020.025 • N gate circuit Want error 1/N c p t+1 • m-qubit, t-error correcting code p Logical gate error rate Logical gate error rate Probability Physical bits of error per logical bit p 1 1/c 1/t c p t+1 m ~p (t+1)2 m 2 p (t+1)3 m 3 O(log log N) concatenations poly(log N) physical bits / logical Physical gate error rate p

  9. Recent results Magic states distillation [Bravyi & Kitaev ‘04, Knill ‘04]  – Universality method, related to best current threshold upper bounds Universal Stabilizer op. – Reduction fault-tolerance fault-tolerance B

  10. Recent results Magic states distillation [Bravyi & Kitaev ‘04, Knill ‘04]  – Universality method, related to best current threshold upper bounds – Reduction from FT universality to FT stabilizer operations Optimized fault-tolerance schemes: [Knill ‘03]  – Erasure error threshold is 1/2 for Bell measurements – [Knill ‘05]: > 5% estimated threshold for depolarizing noise 1% with substantial but more reasonable overhead Fault-tolerance threshold myth: Threshold is all that counts. Maximize the threshold at all costs.

  11. Steane-type error correction X Z operations Physical X X data apply correction X mZ ancilla mZ mX operations Logical mZ

  12. Steane-type Knill-type error correction error correction X Z mX X X data operations mX Physical X X mZ data X apply correction mZ ancilla X mZ ancilla mZ Teleportation mX operations mX Logical mZ mZ

  13. Knill-type error correction mX Advantages  data operations mX – Efficient Physical – Technical advantage: Reduces blockwise mZ independence to encoded Bell state mZ ancilla Teleportation operations mX Logical mZ

  14. Knill-type correction + computation mX Advantages  data operations mX – Efficient Physical – Technical advantage: Reduces blockwise mZ independence to encoded Bell state mZ ancilla U L Teleportation operations mX Logical mZ U

  15. Knill-type correction + computation mX Advantages  data operations mX – Efficient Physical – Technical advantage: Reduces blockwise mZ independence to encoded Bell state mZ ancilla U L Teleportation operations mX Logical mZ U

  16. Knill-type correction + Distance-two code + computation + Postselection mX Advantages  data operations mX – Efficient Physical – Technical advantage: Reduces blockwise mZ independence to encoded Bell state – Allows for more checking mZ ancilla U L Disadvantages  – High overhead at high error rates with error detection Teleportation – Renormalization penalty requires operations stronger control over error distribution mX Logical – No threshold has been proved to exist mZ U

  17. Main issues Bounded dependencies  – Between different blocks – In time – Between bit errors and logical errors Example:  w/ prob. 1-q w/ prob. q … … 3% bit error rate 1% bit error rate q .99 n (1-q) .97 n accepted w/ prob. ⇒ Probability of logical error increases exponentially!

  18. w/ prob. 1-q Main issues … 3% bit error rate Bounded dependencies  w/ prob. q – Between bit & logical errors … 1% bit error rate Monotonicity? want encoded Bell pair: get: low bit high bit monotonicity ⇒ error rate error rate But!

  19. w/ prob. 1-q Main issues … 3% bit error rate Bounded dependencies  w/ prob. q – Between bit & logical errors … 1% bit error rate Monotonicity? (repetition code)

  20. Recent results (continued) Magic states distillation [Bravyi & Kitaev ‘04, Knill ‘04]  – Universality method, related to best current threshold upper bounds – Reduction from FT universality to FT stabilizer operations Optimized fault-tolerance schemes: [Knill ‘03]  – Erasure error threshold is 1/2 for Bell measurements – [Knill ‘05]: > 5% estimated threshold for depolarizing noise 1% with substantial but more reasonable overhead Improved threshold proofs  more efficient – Aliferis/Gottesman/Preskill ‘05: 2.7 x 10 -5 distance three – R. ‘05: < 1.4 x 10 -5 – Ouyang, R. (unpublished): 10 -4

  21. Distance-3 code thresholds Basic estimates  – Aharonov & Ben-Or (1997) – Knill-Laflamme-Zurek (1998) – Preskill (1998) – Gottesman (1997) Optimized estimates  – Zalka (1997) – R. (2004) – Svore-Cross-Chuang-Aho (2005) 2-dimensional locality constraint  – Szkopek et al (2004) – Svore-Terhal-DiVincenzo (2005) But no constant threshold was even proven to exist for distance-3 codes!  – Aharonov & Ben-Or proof only works for codes of distance at least 5 Today: Threshold for distance-3 codes 

  22. Dist-2 code threshold & threshold gap Knill (2005) has highest threshold estimate ~5%  – … Albeit with large constant overhead (more reasonable at 1%) – Again, no threshold has been proved to exist Gaps between proven and estimated thresholds  – Estimates are as high as ~5% – Aliferis-Gottesman-Preskill (2005): 2.6 x 10 -5 Caveat: Small codes aren’t necessarily the most efficient  – Steane (‘03) found 23-qubit Golay code had higher threshold (based on simulations), particularly with slow measurements – 23-qubit Golay code proven: 10 -4

  23. Distance-three code threshold proof intuition  Idea: Maintain inductive invariant of wellness. (A block is well “if it has at most one unwell subblock, and that only rarely.”) EC well well EC well well What’s new: Control probability distribution of errors, not just error states.

  24. Aharonov/Ben-Or-style proof intuition  Idea: Maintain inductive invariant of goodness. (A block is good “if it has at most one bad subblock.”) EC good good X X X X X EC good good X X (assuming one level k-1 error, m ≥ 7)

  25. Aharonov/Ben-Or-style proof intuition  Idea: Maintain inductive invariant of goodness. (A block is good “if it has at most one bad subblock.”) X EC good good X X X X EC good good X (assuming one level k-1 error, m ≥ 7)

  26. Aharonov/Ben-Or-style proof intuition  Idea: Maintain inductive invariant of goodness. (A block is good “if it has at most one bad subblock.”) EC X good good X X X X X X X X EC X X good bad X X X X X (two level k-1 errors, m=7)

  27. Aharonov/Ben-Or-style proof intuition  Idea: Maintain inductive invariant of goodness. (A block is good “if it has at most one bad subblock.”) X X EC good bad X X X X EC good good X (two level k-1 errors) # CNOT locations level-k CNOT level-(k-1) failure rate failure rate

  28. Aharonov/Ben-Or-style proof intuition  Idea: Maintain inductive invariant of goodness. (A block is good “if it has at most one bad subblock.”) For distance-5 code: X X EC EC good good X X X EC EC good good

  29. Aharonov/Ben-Or-style proof intuition  Idea: Maintain inductive invariant of goodness. (A block is good “if it has at most one bad subblock.”) For distance-5 code: X X EC EC good good X X X EC EC good good  Inefficient: 1. 2. not (distance = 5) 3. No threshold for concatenated distance-three codes.

  30. Aharonov/Ben-Or-style proof intuition  Idea: Maintain inductive invariant of goodness. (A block is good “if it has at most one bad subblock.”)  Why not for distance-three codes? X EC X good bad X (one level k-1 error is already too many)  New idea: Most blocks should have no bad subblocks. Maintain inductive invariant of a controlled probability distribution of errors: “wellness.” (A block is well “if it only rarely has a bad subblock.”)

  31. Proof overview  Def: Error states (resolve ambiguity)  Def: Relative error states (encoded CNOT must work even on erroneous input)  Def: good block  Def: “well” block  Distance-3 code threshold setup  Def: Logical success and failure  Distance-3 code threshold proof

  32. Def: Error states  Problem: Different errors are equivalent, so it is ambiguous which bit is in error  Solution: Track errors from their introduction

Recommend


More recommend