quantum control of superconducting...
TRANSCRIPT
Quantum Control of Superconducting Circuits
Liang Jiang Yale University, Applied Physics
Victor Albert, Stefan Krastanov, Chao Shen, Changling Zou
Brian Vlastakis, Matt Reagor, Andrei Petrenko, Steven Touzard, Zaki Leghtas, Reinier Heeres, Wolfgang Pfaff
Mazyar Mirrahimi, Michel Devoret, Rob Schoelkopf
City Tech Physics Seminar 2014.11.13
Supported by: DARPA, ARO, AFOSR, Sloan, Packard.
45mm
Superconduc*ng Cavity-‐Qubit System
storage cavity
transmon qubit
coupler port
Effec*ve Hamiltonian:
−χ sa†a e eH =ω sa
†a +ω q e e
readout cavity
χ s ≫κ s ,γ q
Interac9on strength dominates photon loss & qubit decoherence
κ s : photon decay rate γ q : qubit decoh. rate
• QND readout of the qubit has been demonstrated in many groups (e.g., F>99.5% in 300 ns)
• Long lived SC cavity (T>10 ms >> Tqubit~100us)
storage cavity Manipulate the cavity
H = ω s − χ s e e( )a†a +ω q e e
, gβ0,gD gβ
Weak Drive: Condi9onal cavity displacement
0,e0,gD eβ
β,g
0,e
in-‐phase
quadrature
Phase-‐space diagram
Quantum circuit
cavity
qubit
gDβ
Strong Drive: Uncondi9onal cavity displacement
βDβ 0indep. of qubit state
in-‐phase
quadrature
Phase-‐space diagram
θ
β
Quantum circuit
cavity Dβ
Create superposi4on of cavity state depending on the qubit.
+ε t( )a† eiω st + h.c.
storage cavity Manipulate the qubit
H =ω sa†a + ω q − χ sa
†a( ) e e
Query the qubit: ‘Are there m photons in the cavity?’
Kirchmair G. et al. Nature 495 205-‐209 2013
Johnson B.R. et al. Nature Phys. 6, 663-‐667 2010
Integrated Signal
Spectroscopy frequency (GHz)
Qubit Spectroscopy
χ s
cavity
qubit Yπm
m,eYπm m,g
Yπm n,g n,gn ≠ m
Weak Drive: Condi9onal qubit rota9on:
Yπ n,g n,efor all n
Strong Drive: Uncondi9onal qubit rota9on:
qubit Yπ
Quantum circuit
+Ωm t( ) e g ei ωq−mχs( )t + h.c.
Determinis)c qubit-‐cavity mapping
Leghtas et. al PRA 87, 042315 (2013) Theory:
Transfer arbitrary state from qubit to cavity
Storage
0
Yπ /2 Yπ0
Dβ Cπ Dβ†Dβ
qubit a g b e+
a bβ β+ −
g
Vlastakis et. al Science 342, 6158 (2013) Experiment:
0.0
-‐0.6
0.6
Measure Wigner func*on (based on QND parity meas)
-‐ Can we prepare the cavity in arbitrary superposiTon of photon number states?
( ) ( ) ( ) ( )† †1 a aW Tr D Dα α ρ α⎡ ⎤= ⎢ ⎥⎣ ⎦
−
-‐ Can we robustly encode quantum informaTon in cavity?
Theory of Cat Codes
Mo*va*ons: 1. How to construct robust quantum memory?
a) Overcome cavity dephasing errors b) Overcome cavity loss errors
2. How to control such robust quantum memory? a) Gates over single logical qubits b) Gates between two logical qubits
† † †1 1 1
1 12 2loss a a a a a aκ ρ κ ρ ρ ρ−
⎛ ⎞= − −⎜ ⎟⎝ ⎠D
Dominant Errors in Cavity Quantum Memory
2 21 12 2dephase n n n nφ φκ ρ κ ρ ρ ρ⎛ ⎞= − −⎜ ⎟⎝ ⎠
D
Dephasing Error Photon Loss Error
No Change in photon number! Reduce photon number by one!
(Env. is probing photon number) (Env. is stealing our photons one by one)
Strategy: quickly shuffle photon number … Strategy: monitor photon parity …
generator
Dissipation to Transmission line
LC oscillator
I
Q
D a⎡⎣ ⎤⎦ρ = aρa† − 1
2a†aρ − 1
2ρa†a
Steady state:
ρ∞ = α∞ α∞ ,α∞ = 2ε /κ
Driven damped harmonic oscillator
Driving + DissipaTon à Pure Steady State à Suppress dephasing noise!
ddt
ρ = [εa† − ε * a,ρ]+κ1D a⎡⎣ ⎤⎦ρ
ddt
ρ =κ1D a −α∞⎡⎣ ⎤⎦ρ
† *d d[ ( ) , ]d d d
dd a a adt
ρ ε ε ρ κ ρ⎡ ⎤= − + ⎣ ⎦D
Case d=2 : 2-dim steady state subspace!
{ }2 2, 2 /cρ α α ε κ∞ ± ∞ ∞∈ ± =∑
Case d=4 : 4-dim steady state subspace!
{ }1/4( ) 4 4( ) , (2 / )ic iρ α α ε κ∞ ± ∞ ∞∈ ± =∑
Mul*-‐photon Driven & Mul*-‐photon Damped Oscillator
I
Q
I
Q
I
Q
I
Q
I
Q
I
Q
I
QI
Q
I
Q
I
Q
I
Q
45mm
Pump & Readout Cavity
Storage Cavity
Vertical Transmon Qubit
Coupler Port
Mirrahimi, Leghtas, Albert, et al., NJP 16, 045014 (2014)
ddt
ρ =κ d D ad −α∞d⎡⎣ ⎤⎦ρ
α∞ = 2εd /κ d( )1/d
Ongoing experiment in Devoret’s group
𝑄
𝐼
𝑄
𝐼
Steady states dependent on ini*al states Case d=2 : Two dimensional steady state subspace!
{ }2 2, 2 /cρ α α ε κ∞ ± ∞ ∞∈ ± =∑
2( ) 2nN c nα α∞ ∞+ − =∑0n =ie δφα∞ α∞
0
1Z
Z
n
n
+ = =
− = =
Choice of memory basis (|n=0> and |n=1>)
|+X⟩ ≈ |α⟩|−X⟩ ≈ |−α⟩
|+Z⟩ = |C+α⟩
|−Z⟩ = |C−
α⟩
Y
|+Z⟩ = |C(0mod4)α
⟩
|−Z⟩ = |C(2mod4)α
⟩
|+X⟩ ≈ |C+α⟩|−X⟩ ≈ |C+
iα⟩
Y
(b)
1
|+𝑧⟩= |0⟩
|−𝑧⟩= |1⟩
|+𝑥⟩= |0⟩+ |1⟩ |−𝑥⟩= |0⟩− |1⟩
Vulnerable to both -‐ Dephasing Errors -‐ Photon Loss Errors
|+X⟩ ≈ |α⟩|−X⟩ ≈ |−α⟩
|+Z⟩ = |C+α⟩
|−Z⟩ = |C−
α⟩
Y
|+Z⟩ = |C(0mod4)α
⟩
|−Z⟩ = |C(2mod4)α
⟩
|+X⟩ ≈ |C+α⟩|−X⟩ ≈ |C+
iα⟩
Y
(b)
1
2
2 1
( ) 2
( ) 2 1
Z n
Z n
C N c n
C N c n
α
α
α α
α α +
++ = = + − =−− = = − − = +
∑∑
Choice of qubit basis (2-‐photon process) |+𝒛⟩= |𝜶⟩+ |−𝜶⟩
|−𝒛⟩= |𝜶⟩− |−𝜶⟩
|−𝒙⟩= |−𝜶⟩ |+𝒙⟩= |+𝜶⟩
2| |2
!
n
nc en
α α−=
+1
-1
Par
ity
|+X⟩ ≈ |α⟩|−X⟩ ≈ |−α⟩
|+Z⟩ = |C+α⟩
|−Z⟩ = |C−
α⟩
Y
|+Z⟩ = |C(0mod4)α
⟩
|−Z⟩ = |C(2mod4)α
⟩
|+X⟩ ≈ |C+α⟩|−X⟩ ≈ |C+
iα⟩
Y
(b)
1
Choice of qubit basis (2-‐photon process)
( )0
0
| |
| 1 |
nn
nn
n
x c n
x c n
∞
=
∞
=
+ 〉 = 〉
− 〉 = − 〉
∑
∑with
|−𝒙⟩= |−𝜶⟩ |+𝒙⟩= |+𝜶⟩
2| |2
!
n
nc en
α α−=
|+X⟩ ≈ |α⟩|−X⟩ ≈ |−α⟩
|+Z⟩ = |C+α⟩
|−Z⟩ = |C−
α⟩
Y
|+Z⟩ = |C(0mod4)α
⟩
|−Z⟩ = |C(2mod4)α
⟩
|+X⟩ ≈ |C+α⟩|−X⟩ ≈ |C+
iα⟩
Y
(b)
1
Choice of qubit basis (2-‐photon process)
|−𝒙⟩= |−𝜶⟩ |+𝒙⟩= |+𝜶⟩
Key property: Difference in average photon number <n> is exponen-ally small, for any superposiTon state of
e−α2
Cα+ and Cα
−
Choice of qubit basis (4-‐photon process)
+Z = Cα (0mod 4) = N α + −α( ) + iα + −iα( )( ) = c4n∑ 4n
−Z = Cα (2mod 4) = N α + −α( )− iα + −iα( )( ) = c4n+2∑ 4n+ 2
|+X⟩ ≈ |α⟩|−X⟩ ≈ |−α⟩
|+Z⟩ = |C+α⟩
|−Z⟩ = |C−
α⟩
Y
(a) |+Z⟩ = |C(0mod4)α
⟩
|−Z⟩ = |C(2mod4)α
⟩
|+X⟩ ≈ |C+α⟩|−X⟩ ≈ |C+
iα⟩
Y
1
+1
-1
Par
ity
Key property: Difference in average photon number <n> is exponen-ally small, for any superposiTon state of
e−α2 /2
Cα0mod 4( ) and Cα
2mod 4( )
Zaki Leghtas
Leghtas, et al., PRL 111, 120501 (2012).
No change in photon number (modulo 2 or 4) è
Constant populaTon of è
No logical bit-‐flip errors
±Z
Effect of photon dephasing (in presence of driven dissipa*ve process)
[ ] 2 21 12 2
n n n n nφ φκ ρ κ ρ ρ ρ⎛ ⎞= − −⎜ ⎟⎝ ⎠D
|𝛼|
(a) Two-photon ƉƌŽĐĞƐƐ (b) Four-photon ƉƌŽĐĞƐƐ
𝛾 �����
/2𝜅
2𝛾�����
/𝜅
|𝛼|
Suppression of cavity dephasing error -‐-‐ by mul*-‐photon driven dissipa*ve process
[ ]† *d d[ ( ) , ]d d d
dd a a a ndt φρ ε ε ρ κ ρ κ ρ⎡ ⎤= − + +⎣ ⎦D D
MulT-‐photon driven dissipaTon Dephasing
/ dφκ κ
( )1/d2 / ddα ε κ∞ =
Two Dimensionless Para:
Induced phase-‐flip rate: ExponenTally suppressed with the cat size . α∞
2
Out[133]=
0 5 10 1510-4
0.001
0.01
0.1
1
»a• 2
-2g
bit-flipêk f
Num: kfêk2 ph=1ê200Num: kfêk2 ph=1ê20
Out[133]=
0 5 10 1510-4
0.001
0.01
0.1
1
»a• 2
-2g
bit-flipêk f
Num: kfêk2 ph=1ê200Num: kfêk2 ph=1ê20
0 2 4 6 810-610-510-40.0010.010.11
»a• 2
-2g
bit-flipêk f
Pert. Thry.
Num: kfêk2 ph=1ê200Num: kfêk2 ph=1ê200 2 4 6 8
10-610-510-40.0010.010.11
»a• 2
-2g
bit-flipêk f
Pert. Thry.
Num: kfêk2 ph=1ê200Num: kfêk2 ph=1ê20
Victor Albert
† † †1 1 1
1 12 2loss a a a a a aκ ρ κ ρ ρ ρ−
⎛ ⎞= − −⎜ ⎟⎝ ⎠D
Dominant Errors in Cavity Quantum Memory
2 21 12 2dephase n n n nφ φκ ρ κ ρ ρ ρ⎛ ⎞= − −⎜ ⎟⎝ ⎠
D
Dephasing Error Photon Loss Error
No Change in photon number! Reduce photon number by one!
(Env. is probing photon number) (Env. is stealing our photons one by one)
Strategy: quickly shuffle photon number … Strategy: monitor photon parity …
QND Measurement of Qubit!
0
5
-5
0 50 100 150 Time (ms)
Results from Devoret group, Yale: Hatridge et al., Science 2013*
dispersive circuit QED readout + JJ paramp
Readout fidelity > 99.5% in ~ 300 nsec
Many groups now working with JJ paramps & feedback, including: Berkeley, Delft, JILA, ENS/Paris, IBM, Wisc., Saclay, UCSB, …
*First jumps: R. Vijay et al., (UCB)
QND Measurement of Photon Number Parity
H = −χqsa†a e e U T = π
χqs
⎛
⎝⎜⎞
⎠⎟=
g g + e e for even n
g g − e e for odd n
⎧⎨⎪
⎩⎪
Sun, Petrenko, et al, Nature 511, 444 (2014)
Effect of photon loss (in presence of driven dissipa*ve process)
a Cα
+ → Cα − and a Cα
− → Cα +
a Cα (0mod 4) → Cα
(3mod 4) = N α − −α( ) + i iα − −iα( )( )a Cα
(2mod 4) → Cα (1mod 4) = N α − −α( )− i iα − −iα( )( )
For two-‐photon process: photon loss à logical bit-‐flip error
For four-‐photon process: photon loss à tractable by parity measurement
Summary on Quantum Memory
Two-‐photon process: 1. Logical qubit basis of 2. Photon dephasing induces phase-‐flip errors whose rate
is exp. suppressed by the cat size. 3. Photon loss induces bit-‐flip errors.
Four-‐photon process: 1. Logical qubit basis of 2. Photon dephasing induces phase-‐flip errors whose rate is
exp. suppressed by the cat size. 3. Photon loss induces errors that are tractable by photon-‐
number parity measurements.
How to achieve universal gates on encoded quantum memory?
( )†2 2 22- 2-
†[ [ , ] [, ]] phX phid a aat
a ad
ρ ε ρ κε ρρ= + − +− + D
Quantum Zeno dynamics for 1. Resonant drive è Small displacement
2. Two-‐photon process è ProjecTon on to logical space
D(iε ) Cα
+ = N e− iεα −α + iε + eiεα α + iε( )Cα
+ , Cα −{ }
N e−iεα −α + iε + eiεα α + iε( )→ cos(εα) Cα + + isin(εα) Cα
−
ε XΠ Cα
+ , Cα − (a + a† )Π
Cα + , Cα
− = (α +α∗)ε X Cα + Cα
− + Cα − Cα
+( )∝ X L
2-X phε κ<<Steady State Subspace
gates (2-‐ph process) Xθ
Two-photon process:
Four-photon process:
Decay operator κ2-‐pha2 Driving Hamiltonian iε2-‐ph (a*2-‐a2) Arbitrary rot. around X εX(a*+a) π/2-‐rotaTon around Z -‐χKerr(a*a)2 Two-‐qubit entanglement εXX(a1*a2+a2*a1)
Decay operator κ4-‐pha4 Driving Hamiltonian iε4-‐ph (a*4-‐a4) Arbitrary rot. around X εX(a*2+a2) π/2-‐rotaTon around Z -‐χKerr(a*a)2 Two-‐qubit entanglement εXX(a1*2a22+a2*2a12)
Universal Gates on Quantum Memory
Mirrahimi, Leghtas, Albert, et al., NJP 16, 045014 (2014)
Mazyar Mirrahimi
Summary of Cat Codes
1. How to construct robust quantum memory? a) Overcome cavity dephasing errors b) Overcome cavity loss errors
2. How to control such robust quantum memory? a) Single qubit gates b) Two qubit gates
Autonomous QEC of cat code
Quantum Zeno Dynamics Steady State Subspace
I
Q
I
Q
Symmetry & Conserved Quan**es in Lindblad Systems
Q: What informaTon from iniTal state is preserved as infinite Tme?
Albert, L J, PRA 89,022118 (2014); Mirrahimi, et al., NJP 16, 045014 (2014)
For open system ddt
ρ = Lρ =κ 2D a2 −α∞2⎡⎣ ⎤⎦ρ, initial state ρinit involves into ρ∞ = eLtρinit t→∞
,
which is c++ c+−
c−+ c−−
⎡
⎣⎢⎢
⎤
⎦⎥⎥
in the steady state basis Cα+ , Cα
−{ }.
Each degree cjk is associated with a conserved quantity J jk , such that cjk = Tr J jk† ρinit
⎡⎣ ⎤⎦.
The corresponding quantities can be calculated:
which satisfies d
dtJ jk = L†J jk ≡ 0 with no evolution (conserved).
Key result: #(conserved quanTTes) = #(degrees in steady state density matrix) Conserved quantities: (1) efficient tool to compute steady state, (2) extract stored quantum information, (3) perturbative calculation for other decoherences,
Victor Albert
Outlook: Two-‐Qubit Quantum Modules
Design of Two-‐Qubit Quantum Modules
1. Memory Qubit (m) • Long coherence Tme
2. CommunicaTon Qubit (c)
• IniTalizaTon • Measurement • Entanglement generaTon
3. Local Two-‐Qubit Unitary Gates
𝑚 𝑐
Outlook: Scalable QC
DiVincenzo’s Criteria • A scalable physical system with well characterized qubits • The ability to ini*alize the state of the qubits • The ability to measure specific single qubits • A universal set of quantum gates • Long relevant decoherence *me, much longer than the gate opera*on *mes
…
Use cavity mode for long-‐lived quantum
memory!
Outlook: Scalable QC
DiVincenzo’s Criteria • A scalable physical system with well characterized qubits • The ability to ini*alize the state of the qubits • The ability to measure specific single qubits • A universal set of quantum gates (esp., focusing on the remote gates) • Long relevant decoherence *me, much longer than the gate opera*on *mes
…
… Switchable
Router