Upload
others
View
8
Download
0
Embed Size (px)
Citation preview
Quantum error correction thresholds for the universalFibonacci Turaev-Viro code
arXiv:2012.04610
Alexis Schotte 1 Guanyu Zhu 2 Lander Burgelman 1 Frank Verstraete 1
1 Ghent University
2 IBM Quantum
Outline
Non-Abelian topological quantum error correction
The extended string-net code
Classical simulation
Summary & Outlook
Alexis Schotte arXiv:2012.04610 Non-Abelian topological quantum error correction 2 / 25
Topological quantum computationUse anyons to perform fault-tolerant quantum computation [Kitaev ’97]
Passive TQC Majorana wires, FQH states, . . .
I Information encoded in groundspace of physical systemI Protected by energy gap at T = 0I Need error correction for thermal noiseI non-Abelian anyons are possible
Active QEC surface code, color code, . . .
I Information encoded in codespace of system of qubitsI Actively measure stabilizersI Error correction when stabilizers are violated
Alexis Schotte arXiv:2012.04610 Non-Abelian topological quantum error correction 3 / 25
Active Non-Abelian quantum error correctionI Codespace of system of qubitsI Actively measure set of commuting projectorsI Noise → Non-abelian anyons
Our goals: 1. Construction of a non-Abelian code of qubits2. Classical simulation of this code
Necessary ingredients:
QEC
TQFT TN
You are here
Alexis Schotte arXiv:2012.04610 Non-Abelian topological quantum error correction 4 / 25
Abelian vs non-Abelian codes
Abelian Non-Abelian
Low-weight local stabilizer codesWell studiedHigh thresholds
Native universal gate set in 2DBeyond stabilizer codes
No native non-Clifford gate in 2DI Magic state distillationI Code-switching to higher
dimensional codes→ high overhead
ComplicatedHard to simulate classically
Alexis Schotte arXiv:2012.04610 Non-Abelian topological quantum error correction 5 / 25
Previous work
I König, Kuperberg, Reichardt: Turaev-Viro codesarXiv:1002.2816
I Bonesteel, DiVincenzo: Measurement circuits for Levin-Wen OperatorsarXiv:1206.6048
I Feng, Bonesteel: Fixing violated vertices in Fibonacci Levin-Wen codePhD Thesis W. Feng
I Burton, Brell, Flammia: Threshold for phenomenological model of Fib anyonsarXiv:1506.03815
Alexis Schotte arXiv:2012.04610 Non-Abelian topological quantum error correction 6 / 25
Outline
Non-Abelian topological quantum error correction
The extended string-net code
Classical simulation
Summary & Outlook
Alexis Schotte arXiv:2012.04610 The extended string-net code 7 / 25
The Levin-Wen model
HΛ = −∑
v
Qv −∑
p
Bp
Qv
∣∣i jk⟩
= δijk
∣∣i jk⟩
Bp = 1D2
∑s
dsOsp
v
p
Defined by a Unitary Fusion Category C:I string types {1, i2, . . . , iN} → qudit states |0〉 , |1〉 , . . . , |N〉I branching rules δijk → determine Qv
I numerical data {di, F} → determine Bp
Alexis Schotte arXiv:2012.04610 The extended string-net code 8 / 25
The Levin-Wen model
HΛ = −∑
v
Qv −∑
p
Bp
Qv
∣∣i jk⟩
= δijk
∣∣i jk⟩
Bp = 1D2
∑s
dsOsp
v
p
surface code: C = Z2{1, e} e× e = 1
C = FIB{1, τ} τ × τ = 1 + τ
Alexis Schotte arXiv:2012.04610 The extended string-net code 9 / 25
The Levin-Wen model
HΛ = −∑
v
Qv −∑
p
Bp
Qv
∣∣i jk⟩
= δijk
∣∣i jk⟩
Bp = 1D2
∑s
dsOsp
v
p
a
b
e
c
d=∑n
Fabecdf
a df
cb
=1φ
+1√φ
=1√φ
−1φ
Alexis Schotte arXiv:2012.04610 The extended string-net code 10 / 25
The extended Levin-Wen model
HΛ = −∑
v
Qv −∑
p
Bp
Qv
∣∣i jk⟩
= δijk
∣∣i jk⟩
Bp = 1D2
∑s
dsOsp
v
p
I Localized excitations follow DFIB anyonic statistics: {11, 1τ, τ1, ττ}
I Inside Hs.n. = {|ψ〉 |∏
v Qv |ψ〉 = |ψ〉}, the anyon charge of a plaquette is awell-defined quantum-number.
Alexis Schotte arXiv:2012.04610 The extended string-net code 11 / 25
The tube algebra
Oxαyβ
∣∣∣∣m1
m2
m3
jr
j1j2
mr
jr+1
mr−1
j3
x′⟩
= δx,x′
∣∣∣∣m1
m2
m3
jr
j1j2
mr
jr+1
mr−1
j3
x yα
β
⟩
= δx,x′vαvβ
vy
∑k1,...,kr+1
Fj1jr+1xαβkr+1
( r∏ν=1
Fmνjνjν+1αkν+1kν
)Fkr+1k1yαj1β
∣∣∣∣m1
m2
m3
kr
k1k2
mr
kr+1
mr−1
k3
y
⟩
Central idempotents project on anyonic sectors
Bp = P11 ∼1 1
1
1
+ φ1
τ
1τ
P1τ ∼τ
τ
τ1
+√φe−3πi/5
τ
τ
ττ+ e4πi/5
τ
1
ττPτ1 = P1τ
Pττ ∼ φ21 1
1
1
− φ1
τ
1τ+ φ
τ
τ
1τ+
1√φ τ
τ
ττ+ (1 +
1φ
)τ
1
ττ
Alexis Schotte arXiv:2012.04610 The extended string-net code 12 / 25
Measuring anyon charges
. . .
F-move(2-2 Pachner move)
0
CNOT0
0
(a) (b) (c)
(d) (e) (f)
(g) (h)
(i) (j)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
a
b
c
e
a
b
c
d
e a
b
c
d
e a
b
c
d
e a
b
c
de a
b
c
de
a
b
c
de a
b
c
d
e
a
b
c
de
12
11 109
8 7
6
543
21
13 14 15
1617
U Ua
b
c
e
a
b
c
d
ea
b
c
d
ea
b
c
d
ea
b
c
dea
b
c
de
a
b
c
dea
b
c
d
e
a
b
c
de
‘grow’ circuit ‘resolve’ circuit
abc
d
e
=
F-move
fe
Alexis Schotte arXiv:2012.04610 The extended string-net code 13 / 25
Vertex correctionSome errors σe can take the initial state out of the string-net subspace:
[σe,∏
v
Qv] 6= 0 ⇒ σe |Ψ0〉 /∈ Hs.n.
After measuring all Qv projectors withoutcomes V the state is returned toHs.n. by unitary UV conditioned on V
UV |ΨV 〉 ∈ Hs.n.
UV : 7→
Alexis Schotte arXiv:2012.04610 The extended string-net code 14 / 25
Error correction protocol
Apply noise
Measure vertex projectors + tails
Apply vertex correction
Measure anyon charge in all plaquetteswhile syndrome is nonempty1. Fuse pairs of anyons as dictated by decoding algorithm2. Measure anyon charges
Alexis Schotte arXiv:2012.04610 The extended string-net code 15 / 25
The clustering decoderDecoding graph:
A
B
C
D
E
F
(A)
(A)
(B)
(C)
(D)
(E)
(F)
(A)(G) (H) (I) (J) (K)
G H I J KA
B
C
D
E
F
(A)
(A)
(B)
(C)
(D)
(E)
(F)
(A)
G H I J K
(G) (H) (I) (J) (K)
A
B
C
D
E
F
(A)
(A)
(B)
(C)
(D)
(E)
(F)
(A)
G H I J K
(G) (H) (I) (J) (K)
A
B
C
D
E
F
(A)
(A)
(B)
(C)
(D)
(E)
(F)
(A)
G H I J K
(G) (H) (I) (J) (K)
A
B
C
D
E
F
(A)
(A)
(B)
(C)
(D)
(E)
(F)
(A)
G H I J K
(G) (H) (I) (J) (K)
Alexis Schotte arXiv:2012.04610 The extended string-net code 16 / 25
Outline
Non-Abelian topological quantum error correction
The extended string-net code
Classical simulation
Summary & Outlook
Alexis Schotte arXiv:2012.04610 Classical simulation 17 / 25
Translating qubit errors to anyonic processesI Non-stabilizer code → Pauli noise is not natural here!I Fusion states of anyons in plaquettes ↔ unique lattice qubit states:
|~a,~b〉 =
a2
a4a3
a1
b
I The states {|~a,~b〉} form an orthonormal basis of Hs.n.
I Express joint action of noisevertex measurementvertex correction
in fusion basis:
〈~a′,~b′| UV PV σe |~a,~b〉
〈~a′,~b′| σe PV σe |~a,~b〉
Alexis Schotte arXiv:2012.04610 Classical simulation 18 / 25
Computing the matrix elements
I UV PV σe |~a,~b〉 and σePV σ
e |~a,~b〉 onlydepend on local properties of |~a,~b〉
σx
a1 a2
a4a3b1
b3
b2
I Tensor network representation of anyonic fusion states:
|~a,~b〉 =
a1 a2
a3 a4
a5
a4a3 b
1 b2
a2 a
1 b1
a4 b
2 a5
Alexis Schotte arXiv:2012.04610 Classical simulation 19 / 25
Computing the matrix elements
|~a,~b〉 =
. . .
. . .
. . . . . .
a1
a2
a3
a4
b1 b2
a5
〈~a′,~b′|σexPV σ
ex|~a,~b〉 =
where =σex
σexPA
a2 and =σex
σexPB
a4
Alexis Schotte arXiv:2012.04610 Classical simulation 20 / 25
Assumptions for the classical simulation
I Noise = random Pauli errorsI Total number of errors T ∼ Poisson(|E| p)I Perfect measurementsI Perfect recovery operations
We simulate the dynamics of the system for T timesteps
Individual time step in the simulation1. An edge e is selected at random
2. A random Pauli operator σei is applied to the state
3. All vertex projectors {Qv} and tail qubits are measured
4. Depending on the outcome V , vertex correction UV is applied
5. The anyon charge of all plaquettes is measured.
Alexis Schotte arXiv:2012.04610 Classical simulation 21 / 25
Results
Clustering decoder, depolarizing noise
pc ≈ 0.0475
Clustering decoder, dephasing noise
pc ≈ 0.0725
Comparable to surface code thresholds under similar assumptions!
Alexis Schotte arXiv:2012.04610 Classical simulation 22 / 25
Outline
Non-Abelian topological quantum error correction
The extended string-net code
Classical simulation
Summary & Outlook
Alexis Schotte arXiv:2012.04610 Summary & Outlook 23 / 25
Summary
I Active Non-Abelian QEC → system of qubits
I Code space defined using modified Levin-Wen Hamiltonian
I 3 ingredients:I Quantum error correctionI Topological quantum field theory → tube algebra + anyonic fusion basisI Tensor networks → translating qubit errors to anyons
I First threshold for 2D code which natively supports a universal gate set
I Threshold comparable to surface code threshold under similar assumptions
Alexis Schotte arXiv:2012.04610 Summary & Outlook 24 / 25
Outlook
I More natural noise model?
I Simulating fully fault-tolerant setup
I Finding better decoders
I Applying our techniques to different non-Abelian codes (with lower-depthmeasurement circuits)
I Planar string-net codes
I Road to experimental implementation→ hardware-efficient multi-qubit gates
Alexis Schotte arXiv:2012.04610 Summary & Outlook 25 / 25