Decoherence in flux qubits
Yasunobu Nakamura
NEC Nano Electronics Research Laboratories RIKEN Frontier Research System CREST-JST
• Measurement of T1, T2 – qubit as a spectrum analyzer •Longer T1 samples •Flux qubitin a cavity Hyunsik Im Farrukh Abdumalikov
Yuri Shen Khalil Tsuyoshi Kazuaki Yasu Pashkin Tsai Harrabi Yamamoto Matsuba Nakamura
Antti Oleg Michio Fumiki Tiefu Niskanen Astafiev Watanabe Yoshihara Li Study of decoherence
= Characterization of environment • Qubit as a tool for investigating its environment
environment
interaction
qubit
tunable tunable
Charge-flux: Cottet 2002, Ithier 2005 (Saclay) Charge: Astafiev 2004 (NEC), Duty 2004 (Chalmers) Flux: Bertet 2005 (Delft), Yoshihara 2006 (NEC), Kakuyanagi 2007 (NTT) Phase: Simmonds 2004, Cooper 2005, Martinis 2005 (NIST/UCSB) Steffen 2006, Bialczak 2007 (UCSB), Claudon 2006 (Grenoble) Possible decoherence sources
magnetic-field noise?
trapped vortices?
paramagnetic/nuclear spins?
charge fluctuations? environment circuit modes?
phonons? quasiparticle tunneling?
charge/Josephson-energy fluctuations? Flux qubit: Hamiltonian and energy levels
100 ) z GH ( 0 y
g r e En
-100 0.81.01.2 γ /π nqφ/nφ* nφ*=0.5
J.E. Mooij et al. Science 285, 1036 (1999) Sensitivity to noises
relaxation
transverse coupling
dephasing
longitudinal coupling
nφ-nφ* Estimation of decoherence time
4JJ flux qubit @ optimal bias point constraints:
•EJ-fluctuation can be the largest contribution cf. transmon and CPB
J. Koch et al. PRA 76, 042319 (2007) preliminary Flux qubit: experimental setup
Rabi oscillations resonant microwave pulse
visibility~79.5% Energy relaxation
relaxation and excitation
for weak perturbation: Fermi’s golden rule
•qubitenergy E01 variable • relaxation ∝ S(+E01 ) and excitation ∝ S(-E01 ) ⇒ quantum spectrum analyzer U. Gavish et al. R. Aguado and L. Kouwenhoven R. Schoelkopf et al. O. Astafiev et al. zero-point fluctuation of
environment ex. Johnson noise in ohmic resistor R spontaneous absorption emission
T1 vs f: flux bias dependence
80 π ~ 4ns 70
delay readout pulse 60
50
Switching probability (%) 0.0 0.4 0.8 1.2 1.6 Time ( µs)
initialization to ground state is better than 90%
E01 ⇒ relaxation dominant ⇒ classical noise is not important at qubit frequency ~ 5 GHz E 01 /h (G Hz)
nφ Γ1 vs E01: qubit energy dependence
• Data from both sides of spectroscopy coincide • Floor at high-frequency
• Random high-frequency peaks sample3 E /h (GHz) • Broad structure at low frequency 01 • Depends on SQUID bias point
sample5 E01/h (GHz) Gamma_1 at different flux bias Dephasing
free evolution of the qubit phase
dephasing
for Gaussian fluctuations tunable tunable sensitivity of qubit energy to the fluctuations of external parameter
information of S(ω) at low frequencies Dephasing: T2Ramsey, T2echo measurement Ramsey interference (free induction decay) correspond to detuning π/2~2ns π/2
readout pulse
t 1 0.8 t h g
i 0.6 e
w 0.4 0.2
0.01 0.1 1 10 100 freq. [1/t] spin echo π/2~2ns π ~ 4ns π/2
readout pulse
t/2 t/2 1 0.8 t h g
i 0.6 e
w 0.4 0.2
0.01 0.1 1 10 100 freq. [1/t] Optimal point to minimize dephasing
n • two bias parameters φ Ib – External flux: nφ=Φex/Φ0
– SQUID bias current Ib
E01 (GHz)
I b
nφ
G. Burkard et al. PRB 71, 134504 (2005); P. Bertet et al. PRL 95, 257002 (2005). T1 and T2echo at optimal point nφ=nφ*, Ib=Ib*
T1=545±16ns
Echo decay time is limited by relaxation
Pure dephasing due to high frequency noise (>MHz) is negligible Echo at nφ≠nφ*, Ib=Ib*
consistent with 1/f flux noise
do not fit
No high-frequency cut-off (soft nor hard) below ~1 MHz 1/f noise cut off dependence ΓϕRamsey, Γϕecho vs nφ : flux bias dependence
Red lines: fit
For
Flux noise, not charge noise nor critical current noise
F. Yoshihara et al. PRL 97, 167001 (2006) Dephasing due to photon number fluctuations in SQUID plasma mode
qubit + resonator Red: dephasing due to 1/f flux noise
thermal fluctuation of photon number in resonator ⇒ dephasing of qubit
exponential decay
κ: cavity decay
D.I. Schuster et al. PRL 94, 123602 (2005); P. Bertet et al. Phys. Rev. Lett. 95, 257002 (2005). 1/f flux noise: sample dependence
3 4.58 242.9 11.0 1.63 5 5.08 246.4 9.68 1.40 6 3.85 229.5 18.7 2.90 11 6.07 232.6 10.1 1.55
Loop area ~3 µm2 14 5.45 132 4.9 1.32
-6 2 7±3x10 [Φ0] SQUID~2500-160000 µm F.C.Wellstood et al. APL50, 772 (1987) -6 2 1.5x10 [Φ0] phase qubit ~10000 µm (gradiometer) R.C. Bialczak et al. PRL 99, 187006 (2007) -6 2 ~ 1x10 [Φ0] flux qubit ~1000 µm (Berkeley, unpublished) -6 2 ~ 1x10 [Φ0] flux qubit ~ 25 µm K. Kakuyanagi et al. PRL 98, 047004 (2007) Loop size independent? Optimal point to minimize dephasing
n • two bias parameters φ Ib – External flux: nφ=Φex/Φ0
– SQUID bias current Ib
E01 (GHz)
I b
nφ
G. Burkard et al. PRB 71, 134504 (2005); P. Bertet et al. PRL 95, 257002 (2005). Γ1, ΓϕRamsey, Γϕecho vs Ib : bias current dependence F. Yoshihara et al. PRL 97, 167001 (2006). * Increase |Ib-Ib | ⇒ selectively introduce Ib noise coupling
•exponential decay
•inefficient echo
relaxation: dephasing:
Yu. Makhlin and A. Shnirman, PRL92, 178301 (2004); G. Burkard et al. PRB71, 134504 (2005). Summary
•T1, T2 measurement in flux qubit, T1,T2echo~ several µs • dependence on flux bias and SQUID-current bias conditions ⇒ characterization of environment
Optimal point nφ=nφ*, Ib=Ib* nφ=nφ*, Ib≠Ib*
T1 limited echo decay Pure dephasing due to low freq. noise
‘white’ Ib noise dominant
nφ≠nφ*, Ib=Ib*
Open questions:
-T1 vs flux bias -residual dephasing at optimal point 1/f flux noise dominant -origin of 1/f flux noise Effect of environment circuit design
R-environment
with large Cshunt
LC-environment
with less Cshunt Single qubit 06-06-06 with on-chip LC filter sonance in the LC filter Re
SQUID plasma mode Rabi measurements at optimal point
(corrected for drift) T2Rabi ~6µs 5 5 strong driving
0 0
-5 -5 0 1 2 3 4 0 1 2 3 4 t(µ s) t(µ s) 5 5 %) ( 0 0 Psw -5 -5 0 1 2 3 4 0 1 2 3 4 t(µ s) t(µ s) 80 90 weak driving
75 85
70 80 0 1 2 3 4 0 1 2 3 4 t(µ s) t(µ s) Relaxation at optimal point last logs of T1T2 B=opt Ipre=0.311opt2.lvd
38
37
36
35 T1 =6.3µs
34 %) ( 33 Psw 32
31
30
29 0 5 10 15 20 25 30 time (µ s) Ramsey at optimal point last logs of T1T2 B=opt Ipre=0.311opt2.lvd
39
38 T2Ramsey =2.7µs
37
36 %) ( 35
Psw 34
33
32
31 0 0.5 1 1.5 2 2.5 3 time (µ s) Echo at optimal point last logs of T1T2 B=opt Ipre=0.311opt2.lvd
36
35.5
35
34.5
34 %) (
33.5 T2echo = 3.7 µs Psw 33
32.5
32
31.5 0 5 10 15 20 25 30 time (µ s) Sample with best coherence
On-chip resistors were replaced by superconducting leads by mistake
∆/h=2.557 GHz, IP=157 nA 1/2 2.3 µΦ0/Hz @ 1Hz A. Abdumalikov, Jr. Flux qubit in a cavity: sample design O. Astafiev
Qubit
Nb – main part of the resonator SOG – isolator, used only for test structures Resonator quality factor is defined by the form of Al – adjust coupling of the resonator and qubit this Al island chip size: 2.5 × 5 mm Spectroscopy
• Plot shows the change of the phase dip when peak when Vacuum Rabi splitting
fit: Relaxation time (continuous measurement)
• measured by sweeping the pulse period – Pulse width was 10 ns
•T1= 0.73 µs – Qubit frequency
E01/h = 8.4 GHz – Coupling to cavity gsinθ/h = 30 MHz
•E01 dependence Remarks
•Qubit is powerful tool to analyze its environment
What we have tried and we have been trying
• Flux bias point dependence (nφ*=0.5, 1.5, 2.5; up to 20 Gs) •T1 depends on bias point – coupling to the SQUID is different •T2echo, T2Ramsey are almost independent ~ 20% variation • Area dependence -1/2 • 4 times larger qubit – flux noise 3.6 µΦ0/Hz • Qubit with small ∆/h~50 MHz
•Long T1 ~ 5 ms • Consistent with flux noise contribution • However, does not necessarily mean flux noise contribution • Spin locking
• Temperature dependence of T1 •Qubitwith variable ∆ • Superconducting leads