A fault-tolerant one-way quantum computer
Robert Raussendorf1, Jim Harrington2 and Kovid Goyal1
1: California Institute of Technology, 2: Los Alamos National Laboratory
A fault-tolerant one-way quantum computer Robert Raussendorf 1 , Jim - - PowerPoint PPT Presentation
A fault-tolerant one-way quantum computer Robert Raussendorf 1 , Jim Harrington 2 and Kovid Goyal 1 1: California Institute of Technology, 2: Los Alamos National Laboratory QIP Paris, 20th February 2006 Main idea Universal quantum computation
1: California Institute of Technology, 2: Los Alamos National Laboratory
Universal quantum computation by local measurements:
Retain universality, Retain universality, add fault-tolerance add fault-tolerance 2D cluster state 3D cluster state
Make use of geometry!
tion of qubits on a three-dimensional lattice required.
Error sources: preparation, gates, storage and measurement.
Three cluster regions: V (Vacuum), D (Defect) and S (Singular qubits). Qubits q ∈ V : local X-measurements, Qubits q ∈ D: local Z-measurements, Qubits q ∈ S: local measurements of X±Y
√ 2 .
measurement of Z (⊙), X (↑), cos α X + sin α Y (ր)
read out by one-qubit measurements only.
A cluster state |φC on a cluster C is the single common eigenstate
Ka = Xa
Zb, ∀a ∈ C, (1) where b ∈ N(a) if a,b are spatial next neighbors in C. Z-Rule:
ZZ Z X X X X Z Z Z Z Z Z Z Z Z X X X X X X X X X Z = = plaquette stabilizer site stabilizer One qubit located on every edge Z X harmless error equivalent harmful errors syndrome at endpoint
equivalent chains correspond to physically equivalent errors.
rough edge smooth edge
Torus Plane segment 2 Qubits 1 Qubit
X Z hole
Plane with 2 holes 1 Qubit
the code surface.
pattern of Z- and X-measurements.
Cluster states in three spatial dimensions provide intrinsic topological error correction.
1 2 1 2 z x
elementary cell of L qubit location: (even, odd, odd)
qubit location: (odd, odd, even)
syndrome location: (odd, odd, odd)
syndrome location: (even, even, even)
Measurement pattern:
Errors:
(X-errors are absorbed into the X-measurement, I±X
2 X = ±I±X 2 .)
X Z Z Z Z face edge e f f X X X X X X bit syndrome
K(f) =
Xf
Zb. (2)
∂f = 0. (3)
What about the edge qubits?
Translation by vector (1, 1, 1)T:
→ L (dual). face of L − edge of L, edge of L − face of L, site of L − cube of L, cube of L − site of L, (4)
cluster harmful error elementary cell
[1] Dennis et al., quant-ph/0110143 (2001).
Map error correction to statistical mechanics:
Nishimori line
Error correction [1] Minimum weight chain matching [2]
[1] T. Ohno et al., quant-ph/0401101 (2004). [2] E. Dennis et al., quant-ph/0110143 (2001); J. Edmonds, Canadian J. Math. 17, 449 (1965).
Cluster region V Defects d ∈ D Singular qubits
Fault-tolerant quantum logic is realized via topologi- cally entangled engineered lattice defects.
sured in the Z-basis.
A quantum circuit is encoded in the way primal and dual defects are wound around another.
X Z hole
Plane with 2 holes 1 Qubit
Z X dual stabilizer element primal stabilizer element code surface
fault-tolerant measurement of X±Y
√ 2 .
Cluster region V Defects d ∈ D Singular qubits
Reed-Muller encoder
CNOTs, |0 , |+ -preps. qubit, encoded with surface code qubit, encoded with Reed-Muller code
√ 2
via local measurements of Xa±Ya
√ 2
and classical post-processing.
Error model:
from product state
a∈C |+a.
– |+-preparation:
Perfect preparation followed by 1-qubit par- tially depolarizing noise with probability pP.
– Λ(Z)-gates: Perfect gates followed by 2-qubit partially depolar-
izing noise with probability p2.
– Memory: 1-qubit partially depolarizing noise with probability pS
per time step.
– Measurement: Perfect measurement preceeded by 1-qubit par-
tially depolarizing noise with probability pM .
The fault-tolerance threshold is 1.1×10−3 for each
p2,c = 9.6 × 10−3, for pP = pS = pM = 0, pc = 5.8 × 10−3, for pP = pS = pM = p2 =: p. (5)
Error budget from Reed-Muller concatenation threshold: 76 15p2 + 2 3pP + 4 3pM + 4 3pS < 1 105. (6) Specific parameter choices: p2,c = 2.9 × 10−3, for pP = pS = pM = 0, pc = 1.1 × 10−3, for pP = pS = pM = p2 =: p. (7) The Reed-Muller code sets the overall threshold.
N: Number of non-Clifford operations in bare computation. Nft: Number of operations for fault-tolerant computation. Nft ≤ N2 (log N)10.8 . (8)
[quant-ph/0510135] Scenario:
Numbers:
gate-, storage- and measurement error (each source). Methods:
sic topological error correction related to the Random plaquette Z2-gauge model.
gineered lattice defects.