Download presentation
Presentation is loading. Please wait.
Published byFlora Day Modified over 9 years ago
1
Debbie Leung (IQC, UW) w/ Panos Aliferis, Yingkai Ouyang, Man-Hong Yung (IBM) (IQC, UW) (UIUC) $$: NSERC, CFI, ORF, CIFAR, MITACS, QWorks, ARO, Mike Lazaridis Fellowship(s) Two measurement problems in Fault-tolerant Quantum-Computation Dec 17, 2007. QEC07, USC
2
Measurement problem 1: * In measurement based quantum computation will 1 error in a measurement outcome cascade to many, overwhelming the ability to perform QEC? Measurement problem 2: * NonMarkovian noise is modeled as a system-bath Hamiltonian evolution. Its sup-norm should be sub-threshold for FT method to work. Will a sharp measurement, with an obligatorily large amplitude for the system-bath Hamiltonian, spoil the result? * * * * * * Plan: interpret the problem away, thus, the problem and result are inevitably trivial Disclaimer -- this is not on Foundations of QM
3
Measurement problem 1: * In measurement based quantum computation will 1 error in a measurement outcome cascade to many, overwhelming the ability to perform QECC? * * * * * * Will focus on graph state quantum computation
4
Graph state Quantum Computation GSQC: RBB 0301052 Ã 1WQC: RB01 Linear optics app: N 0402005 Circuit interpretation CLN04, AL04 w/ gate-teleportation (GC99, ZLC00) (1) Produce a graph state (2) Apply single measurements that can be controlled by prior measurement outcomes Simulate a circuit element-by-elment -- can identify "regions" responsible for each
5
Graph state simulation of a circuit element O 0/1 x1x1 x2x2 xnxn |+ i... e in P ein ( in ) e out P eout (O( in )) |+ i = |0 i + |1 i = controlled- z P i = Pauli rotation indexed by i S O (e in P ein ( in )) = (e out P eout O( in )) (will see this is composable)
6
Composable simulation for a universal gate set £ £ |+ i MxMx abab b©dab©da XaZb|iXaZb|i X b © d Z a H| i d S(H): |+ i MxMx abab ab©cab©c XaZb|iXaZb|i X a Z b © c Z | i c S(Z ): Z (-1) a H a2b2a2b2 X a 1 Z b 1 X a 2 Z b 2 | i S( CP ): a1b1a1b1 a 1 b 1 © a 2 a 2 b 2 © a 1 X a 1 Z b 1 © a 2 X a 2 Z b 2 © a 1 (CP| i )
7
Goal: perform Z rotation e iZ
8
H c |0 i |i|i Zc|iZc|i Goal: perform Z rotation e iZ Z-Telep (ZT)
9
H c |0 i |i|i Zc|iZc|i Goal: perform Z rotation e iZ c Z c e i(-1) a Z X a Z b | i XaZb|iXaZb|i H |0 i e i(-1) a Z Z-Telep (ZT) Input state = e i(-1) a Z X a Z b | i
10
H c |0 i |i|i Zc|iZc|i Goal: perform Z rotation e iZ c Z c e i(-1) a Z X a Z b | i XaZb|iXaZb|i H |0 i e i(-1) a Z = X a Z c+b e iZ | i Z-Telep (ZT) Xa eiZXa eiZ = § X a Z c+b e iZ | i
11
H c |0 i |i|i Zc|iZc|i Goal: perform Z rotation e iZ c Z c e i(-1) a Z X a Z b | i XaZb|iXaZb|i H |0 i e i(-1) a Z = X a Z c+b e iZ | i Z-Telep (ZT) Xa eiZXa eiZ = § X a Z c+b e iZ | i |+ i MxMx abab ab©cab©c XaZb|iXaZb|i X a Z b © c Z | i c S(Z ): Z (-1) a H
12
Composable simulation for state prep & meas S(|0 i ): MzMz a 0 X a Z 0 |0 i a0a0 0 S(|+ i ): MxMx b X 0 Z b |+ i 0b0b MzMz abab a©ca©c XaZb|iXaZb|i c S( M z ): MxMx abab a©ca©c XaZb|iXaZb|i c S( M x ): |+ i given
13
General quantum circuit C |0 i U1U1 U2U2 U3U3 U4U4 U5U5 0/1 y1y1 y2y2 y3y3
14
Composing simulations simulates the circuit S |0 i 0/1 y1y1 y2y2 y3y3 S O k... S O 2 S O 1 (P init (|0 ih 0|)) = P final O k... O 2 O 1 (|0 ih 0|) S SU1SU1 SU2SU2 SU3SU3 SU4SU4 SU5SU5 S S S O (e in P ein ( in )) = (e out P eout O( in ))
15
FT Qn 1: Can we achieve FT simply by simulating a FT circuit ? Physical noise ! effective noise in simulated circuit ? FT Qn 2: Will an error propagate via feedforward ? Say, 1 error corrupts a meas outcome, which may induce further errors as it is used to condition subsequent op... ? AL05: offer an interpretation in which measurement outcomes are error free... yes to FT R03, ND04 -- with much work in designing schemes and full-blown nonMarkovian treatment of meas outcome errors
16
S( X a’ Z b’ ) (or S(I)): abab a © a’ b © b’ XaZb|iXaZb|i a’ b’ X a © a’ Z b © b’ | i Composable simulation for Pauli’s Note: Known Pauli operations : shifts in the classical parts. Unknown Pauli errors : errors in the classical parts.
17
0. Add each |+ i to the graph state only slightly before it’s being measured ~ 1. Model noisy elementary operations O storage, gate, or state prep meas env UFUF O ~O~O » UFUF O ~O~O » where U F = I A 0 + i P i A i on sys env P i = nontrivial Pauli’s Now working out how physical noise transformed to dealing the noise
18
: where U F s act 2. Noisy simulation: interperse U F s between ideal operations e.g. |+ i MxMx abab ab©cab©c XaZb|iXaZb|i X a Z b © c Z | i c S(Z ): Z (-1) a H Sub U F = I sys A 0 env + i P i sys A i env in above, expand For @ summand, commute P i 's towards end of simulation - “Joint I term” : ideal simulation - Else:(1) classical part may flip (2) quantum part may suffer unknown Pauli error But they’re equivalent !! So classical part is always correct & faults do not propagate but are localized How physical errors affect each simulation?
19
Thus, in GSQC: No error in classical part (they're unknown Pauli errors) Each elementary faults is also localized within a simulation Faults of all elementary operations in 1 simulation give a combined fault in that simulated op, and no where else, with effective noise strength · max # locations in a sim £ max strength per elementary op All FT results for circuits thus apply (including the code design, threshold analysis, and size+depth bounds) In fact, the composable graph state simulation is like a lowest level encoding in FTQC Qn: can GS/other teleportation based implementation decohere different fault paths? Need to understand meas better.
20
* * * * * * Measurement problem 2: * NonMarkovian noise is modeled as a system-bath Hamiltonian evolution. Its sup-norm should be sub-threshold for FT method to work. Will a sharp measurement, with an obligatorily large amplitude for the system-bath Hamiltonian, spoil the result? Remember fast measurement & classical components should help FT but how to model these?
21
Modelling sharp measurements: Q C B M |x i Q ! |x i Q |x i C |x i M need to give a copy of the measurement outcome x to the environment to ensure its classicality g g H = H Q + H C + H M + H B + H QC + H CM + H BM + H QM + H QB + H BC || H QC || 1 = || H CM || 1 = g signifies how fast the meas is || H BM || 1 = quantities how fast adversarial noise can use the meas outcome || H QB || 1 = || H QM || 1 = the usual QC noise amplitude || H CB || 1 = ¿ Classical Comp noise amplitude g large if meas fast
22
Modelling sharp measurements: Q C B M g g H = H Q + H C + H M + H B + H QC + H CM + H BM + H QM + H QB + H BC || H QC || 1 = || H CM || 1 = g signifies how fast the meas is || H BM || 1 = quantities how fast adversarial noise can use the meas outcome || H QB || 1 = || H QM || 1 = the usual QC noise amplitude || H CB || 1 = ¿ Classical Comp noise amplitude || e -iHt - V || 1 · t ( + 2 + ) Let t be the meas time (tg ¼ 1), V be ideal meas Let t O be the time for a gate, noise strength ¼ t O (t O À t) Putting in the separation of scale, as long as · t O /t the threshold is unspoiled. Qns: do we need to include in the noise amplitude for gate alone? What to expect for in various systems? Decohering fault paths by (un)intentional "random" Pauli's? esp in FT designs heavily relying on gate teleportation...
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.