Enhancing Dispersive Readout of Superconducting Qubits Through Dynamic Control

of the Dispersive Shift: Experiment and Theory

François Swiadek,1, 2, ∗ Ross Shillito,3 Paul Magnard,1 Ants Remm,1, 2
Christoph Hellings,1, 2 Nathan Lacroix,1, 2 Quentin Ficheux,1 Dante Colao Zanuz,1, 2
Graham J. Norris,1, 2 Alexandre Blais,3, 4 Sebastian Krinner,1, 2 and Andreas Wallraff1, 2, 5
1
Department of Physics, ETH Zurich, 8093 Zurich, Switzerland
2
Quantum Center, ETH Zurich, 8093 Zurich, Switzerland
3
Institut Quantique and Département de Physique,
Université de Sherbrooke, Sherbrooke J1K 2R1 QC, Canada
4
Canadian Institute for Advanced Research, Toronto, Ontario M5G 1M1, Canada
5
ETH Zurich - PSI Quantum Computing Hub, Paul Scherrer Institute, 5232 Villigen, Switzerland
(Dated: July 18, 2023)
arXiv:2307.07765v1 [quant-ph] 15 Jul 2023

The performance of a wide range of quantum computing algorithms and protocols depends crit-
ically on the fidelity and speed of the employed qubit readout. Examples include gate sequences
benefiting from mid-circuit, real-time, measurement-based feedback, such as qubit initialization,
entanglement generation, teleportation, and perhaps most importantly, quantum error correction.
A prominent and widely-used readout approach is based on the dispersive interaction of a super-
conducting qubit strongly coupled to a large-bandwidth readout resonator, frequently combined
with a dedicated or shared Purcell filter protecting qubits from decay. By dynamically reducing
the qubit-resonator detuning and thus increasing the dispersive shift, we demonstrate a beyond-
state-of-the-art two-state-readout error of only 0.25 % in 100 ns integration time. Maintaining low
readout-drive strength, we nearly quadruple the signal-to-noise ratio of the readout by doubling
the readout mode linewidth, which we quantify by considering the hybridization of the readout-
resonator and its dedicated Purcell-filter. We find excellent agreement between our experimental
data and our theoretical model. The presented results are expected to further boost the performance
of new and existing algorithms and protocols critically depending on high-fidelity, fast, mid-circuit
measurements.

Realizing high-fidelity and fast single-shot readout of a Faster readout protocols have been realized, with a
qubit [1–3] is essential for quantum error correction pro- 9 × 10−3 fidelity readout achieved in 40 ns by utilizing
tocols [4–8] in which qubit decoherence during readout the distributed-element, multimode nature of the read-
and reset contributes significantly to the logical error. It out resonator [33].
is also key for algorithms requiring real-time feedback,
such as teleportation [9–12], distillation [13, 14] and ini-
tialization [15–18]. One of the critical parameters governing dispersive
In superconducting circuits, the most commonly used qubit readout is the detuning between the qubit and the
readout architecture employs the state-dependent disper- readout resonator, which controls both the magnitude
sive shift of the resonance frequency of a resonator cou- of the dispersive shift and the nonlinearities induced in
pled to the qubit to infer the qubit state [19–21]. Whilst the resonator. Different detuning regimes have been ex-
the frequency of the resonator is typically fixed, flux- plored, including cases where the resonator frequency is
tunable transmons allow to control the qubit-resonator lower than the qubit [3, 34]. Notably, the measurement
detuning by modifying the transmon frequency [22], en- fidelity has been shown to improve for smaller detun-
able high-fidelity fast entangling gates [23–25] and avoid ings [7, 35], although these observations were not fully
frequency collisions. Additionally, each qubit is often explained.
coupled to a microwave transmission line via a dedicated
[6, 7, 26] or common Purcell filter [8, 27, 28] to pro-
tect the qubit from radiative decay [29–31]. Such mea-
surements are usually performed with weak measurement In this work, we demonstrate an increase in the signal-
tones to avoid nonlinearities and detrimental qubit state to-noise ratio (SNR) and assignment fidelity by bringing
transitions, although high-power readout has been stud- the qubit frequency closer to the readout resonator’s fre-
ied both theoretically [32] and experimentally [2]. quency using a flux pulse, see illustration in Fig. 1 (a,b),
In the past few years, significant improvements to the achieving a minimum two-level readout error of 2.5×10−3
single-shot readout have been realized, reaching a two- in 100 ns. We accredit this remarkable performance not
level readout assignment fidelity of 4 × 10−3 in 88 ns [3]. only to an increase in the dispersive shift χ imparted by
the qubit on the cavity, but also to an increase in the ef-
fective linewidth of the targeted normal mode response,
caused by bringing the Lamb-shifted readout resonator
∗ [email protected] closer to resonance with the Purcell filter, see Fig. 1 (c).
 2

(a) (b) flux pulses are Gaussian-filtered rectangular pulses with

short rising and falling edges minimizing coupling to two-
level systems [7], see Fig. 1 (b).
We repeat the experiment for five different qubit—
readout-resonator detunings ∆qr /2π, spanning −2.7 GHz
(c) to −1.3 GHz, where ∆qr = ωq − ωrg . We denote
g/e
ωr as the readout resonator frequency with the qubit
prepared in the ground/excited state. The measured
(light colored lines) and calculated (dark colored lines)
transmission response is shown in Fig. 2 (a), with
blue/red lines corresponding to the qubit prepared in
the ground/excited state. From a fit to a coupled qubit—
readout-resonator—Purcell-filter model (see Appendix A
and solid blue and red lines in Fig. 2 (a)), we extract
FIG. 1. (a) Schematic of a qubit coupled to a readout the relevant readout parameters for each value of ∆qr .
resonator-Purcell-filter system. The qubit of transition fre- The measured (dots) and calculated (lines) dressed read-
quency ωq is coupled capacitively at rate g to a readout res- out resonator frequencies ωrg and ωre = ωrg + 2χ are
onator of frequency ωr . The readout resonator in turn is
shown in Fig. 2 (b) (blue and red circles) as a func-
coupled at rate J to a Purcell filter of frequency ωp , which
is coupled to a feedline at rate κp . We probe the system by
tion of the detuning ∆qr , along with the Purcell filter
measuring the transmission of a readout pulse at frequency ωd frequency ωp /2π = 6.900 GHz, which remains constant.
g/e
through the feedline. The effective decay rate of the readout The variation in the resonator frequencies ωr is due to
resonator is indicated as κ. (b) Schematic of time-dependence 2
the Lamb shift g /∆qr caused by the qubit [22]. Fur-
of qubit frequency ωq relative to the readout resonator fre- thermore, we extract both a large intended Purcell filter
quency ωr . The qubit, initially idling at the lower flux sweet linewidth κp /2π = 34.5 MHz and a large intended cou-
spot, is pulsed to a smaller detuning from the readout res- pling strength between the readout resonator and the
onator using a fast Gaussian-filtered, rectangular flux pulse.
Purcell filter J/2π = 27.5 MHz.
(c) Illustration of the rise of the SNR of the qubit read-
out with integration time τ parameterized by the effective We consider a standard circuit-QED approach to
linewidth of the readout resonator κ at approximately con- model the transmon—resonator—Purcell-filter system,
stant dispersive shift χ. Increasing color saturation in panels see Appendix B for details. In the case of a weak drive
(b) and (c) indicate increasing κ at reduced detuning between E applied to the filter mode, the readout resonator and
ωq and ωr . the Purcell filter responses can be considered as linear.
As such, the dynamics can be effectively mapped to the
equations of motion [37]
I. READOUT PARAMETER
CHARACTERIZATION  g/e   g/e   g/e 
α̇ ωr J α
= − i
β̇ g/e J ωp − iκp /2 β g/e
We perform the experiment with a transmon qubit of   (1)
transition frequency ωq /2π = 4.14 GHz at the lower flux 0
+ ,
sweet spot [36] and anharmonicity α/2π = −181 MHz. Ee−iωd t
It has a lifetime T1 = 30.4 µs and is capacitively cou-
pled to a readout resonator with a coupling strength where α and β represent the coherent fields of the readout
g/2π = 224 MHz. The readout resonator is coupled to resonator and Purcell filter, respectively. In the regime
a feedline used for multiplexed readout [26] via a dedi- J ≈ κp , we observe two distinct hybridized readout-
cated Purcell filter of linewidth κp and with a coupling resonator—Purcell-filter modes, see Fig. 2 (a). We de-
strength J, see Fig. 1 (a). The qubit is located on a de- note these as the low and high readout modes, respec-
vice used to execute a distance-three surface code (see tively, the lowest and highest of the two modes in the
Fig. 7, Appendix A). Further information on the device transmission spectrum. The frequency and linewidth of
properties and its fabrication can be found in Ref. [7]. these modes can be determined, respectively, from the
To determine the readout parameters as a function of real and the imaginary part of the eigenvalues of the
the frequency detuning between the qubit and its read- equations of motion in Eq. (1) in the absence of a drive:
out resonator, we perform pulsed spectroscopy experi-
ments. We first prepare the qubit in the ground state g/e
s 2
|g⟩ or excited state |e⟩, pulse the qubit to a chosen read- g/e ωr + ωp 1 g/e iκp
ωl,h = ± Re ∆rp + + 4J 2 ,
out frequency ωq using a baseband flux pulse, and probe 2 2 2
the readout circuit using a 2.2 µs long microwave tone. s (2)
2
This duration corresponds to the maximum integration g/e κp g/e iκp
κl,h = ∓ Im ∆rp + + 4J 2 .
time of our readout electronics (see Appendix A). The 2 2
 3

(a) (b) Detuning, rp/2 (MHz) (c) Detuning, rp/2 (MHz) (d) Detuning, rp/2 (MHz)
|g |e d -24 -18 -4 6 29 -24 -18 -4 6 29 -24 -18 -4 6 29
5 6.94 20

(MHz)
Transmission Amp., |Sout, in|

Disp. shift, 2| l, h|/2 (MHz)

r /2
g
6.93 2| l + h|/2 25

Frequency, /2 (GHz)
r /2
e
4 2| h|/2
6.92 p/2 15 2| l|/2 20

l, h/2
3 6.91
10 15

Res. linewidth,
2 6.90
10
6.89 5
1 g
l /2
g
h /2
6.88 5
l /2 h /2
e e
0 6.87 0 0
6.8 6.9 7.0 2.5 2.0 1.5 2.5 2.0 1.5 2.5 2.0 1.5
Frequency, /2 (GHz) Detuning, qr/2 (GHz) Detuning, qr/2 (GHz) Detuning qr/2 (GHz)
FIG. 2. (a) Readout circuitry transmission spectra measured for five qubit—readout-resonator detunings (∆qr /2π ∈
[−2.7, −2.4, −1.9, −1.6, −1.3] GHz from bottom to top, vertically shifted by increments of one). Spectra are shown for both the
qubit prepared in the ground state |g⟩ (blue) and in the excited state |e⟩ (red). Solid lines are fits based on a coupled qubit—
readout-resonator—Purcell-filter model (see Appendix A). Dashed black lines indicate the selected readout frequency, chosen
|e⟩ |g⟩
such that the response in transmission |Sout,in − Sout,in | is maximum. (b) Resonator frequency ωrg (ωre ) conditioned on the qubit
being prepared in the ground (excited) state, and Purcell filter frequency ωp as a function of ∆qr . (c) Measured dispersive shift
χl (χh ) of the lower (higher) frequency hybridized readout mode (see green (purple) points), as a function of ∆qr . The two
contributions sum up to the bare readout resonator mode dispersive shift (solid black line) χ = χl + χh ≈ αg 2 /∆2qr . The solid
lines (purple and green) are fits based on a qubit—readout-resonator—Purcell-filter model (see Appendix A). (d) Measured
effective readout resonator linewidth κl (κh ) for the lower (higher) frequency hybridized readout mode for the qubit prepared in
the ground/excited (see blue/red), as a function of ∆qr . Solid lines are fits based on a qubit—readout-resonator—Purcell-filter
model (see Appendix A). In (b,c,d) the corresponding detunings ∆rp = ωrg − ωp between the Purcell filter and the readout
resonator, indicating the degree of hybridization of the two resonator modes, are shown on the top axis.

The readout-resonator—Purcell-filter hybridization resonator detuning ∆qr . As shown in Fig. 2 (d), while
leads to a distribution of the total qubit-induced dis- g/e g/e
|χl | > |χh | for ∆qr /2π ≤ −1.6 GHz, we have κl < κh .
persive shift χ on the readout-resonator—Purcell-filter This is expected as for ∆qr /2π ≤ −1.6 GHz, the low
system. Using the model in Eq. (2), we can extract mode has a larger weight in the readout resonator. The
the dispersive shifts of the low and high modes respec- difference between κgl and κel for the low mode derives
e g
tively, χl/h = (ωl/h − ωl/h )/2, see purple (green) cir- g/e g/e
from the frequency detuning ∆rp = ωr − ωp , between
cles for the low (high) mode in Fig. 2 (c). While the the readout-resonator frequency and the Purcell-filter fre-
total dispersive shift χ = χl + χh shows the expected quency. In particular, for the low mode, κgl > κel for all
αg 2 /∆2qr dependence in Fig. 2 (c) (solid black line), with detunings while for the high mode κgh < κeh , which can be
2χ/2π ∈ [−5.67, −19.49] MHz, the low mode dispersive seen from the analysis of the normal mode Hamiltonian
shift only shows small variations in that range, staying in Appendix B.
between 2χl /2π ∈ [−4.17, −6.69] MHz (see solid purple In the vicinity of the detuning leading to an equal
line in Fig. 2 (c)). In contrast, the high mode dispersive hybridization of the low and high modes ∆qr /2π ≈
shift shows a similar scaling with ∆qr as the total dis- −1.6 GHz, we further note that all κeh /2π ≈ κgl /2π ≈
persive shift, with 2χh /2π ∈ [−1.5, −13.18] MHz (solid 19 MHz and κgh /2π ≈ κel /2π ≈ 14 MHz. After this
green line in Fig. 2 (c)). crossing point, we observe that while |χl | > |χh | for
We observe that the dispersive shift of the low mode g/e g/e
∆qr /2π ≥ −1.6 GHz, we find κl > κh , which we ex-
is dominant for qubit—readout-resonators detunings be-
ploit in Sec. II. The detailed parameters are summarized
low −1.6 GHz, after which the dispersive shift of the high
in Table I.
mode becomes larger. The crossing point where χl = χh ,
in the vicinity of the qubit—readout-resonator detun-
ing ∆qr /2π = −1.6 GHz, corresponds to an equal hy-
bridization of the two readout modes. It coincides with II. SINGLE-SHOT READOUT
ωp = ωrg + χ = ωre − χ being equidistant to the ground
and excited state responses of the readout resonator, see We perform single-shot readout for the
Fig. 2 (b). qubit—readout-resonator detunings ∆qr /2π ∈
Our model also gives us valuable information about [−2.7, −2.4, −1.9, −1.6, −1.3] GHz (see Fig. 2) as a
the linewidth of the low and high modes, for the qubit function of the readout-pulse power and integration time
prepared in the ground |g⟩ or excited state |e⟩, namely τ ∈ [50, 100, 200, 300, 400] ns. Each experiment consists
κgl (κgh ) and κel (κeh ), as a function of the qubit—readout- of 104 single-shot measurements with the qubit prepared
 4

103

Int. Unit 2 (arb. units) Counts

in the ground or excited state. The detuning is varied
by tuning the qubit to a chosen frequency ωq , using a
0
flux pulse as described above. We use a rectangular
readout pulse with a duration of 450 ns convolved with 1
a Gaussian filter of width σ = 0.5 ns, and integrate the
readout signal for a time τ using mode-matched weights 0
[38] to discriminate the ground |g⟩ and excited |e⟩ qubit 1
state responses. The flux pulse lasts longer than the
readout pulse. In addition, we use a preselection readout 2
to reduce residual excited state population of the qubit
to below 0.1% [39]. 3
We express the readout power as a function of the 3 2 1 0 1 2 3 4 5 0 103
number of photons in the readout resonator ng when Int. Unit 1 (arb. units) Counts
the qubit is prepared in the ground state, relative to
the critical number of photons in the resonator ncrit = FIG. 3. Single-shot readout histogram for a qubit—readout-
∆2qr /4g 2 [19] at a given qubit—readout-resonator detun- resonator detuning of ∆qr /2π = −1.3 GHz, a readout integra-
ing ∆qr . By measuring the qubit-induced ac-Stark shift tion time of τ = 100 ns, and ng /ncrit = 0.93. We assign the
∆ac = 2g 2 /∆qr on the readout resonator at ∆qr /2π = measured state using a bimodal Gaussian mixture model. The
−2.7 GHz we can infer the number of photons ng in the marginal distributions of this model along each axis are plot-
ted along the corresponding axis. A solid black line indicates
resonator when the qubit is prepared in the ground state,
the distance between the means µg and µe of the Gaussian
ng = ∆ac /2(χl +χh ) (see Appendix D). The photon num- distributions of the ground and excited state responses. The
ber ng at other detunings and the photon number ne square root of the covariance matrix diagonal elements of the
when the qubit is prepared in the excited state for all Gaussian distributions σg and σe are used as the radii of the
detunings are inferred using semi-classical analysis, see black circles.
Appendix B.
The readout drive frequency ωd is chosen such that
the difference in the response in transmission when the We characterize the measurement by the average as-
|e⟩ signment error εa for two-state readout, limited by the
qubit is prepared in the ground or excited state |Sout,in −
|g⟩ overlap error between the Gaussian distributions and the
Sout,in |, is maximum for the low mode, see vertical black qubit lifetime T1 , defined as [38]
dashed lines in Fig. 2 (a). In the theoretical model
(see Appendix. C), this choice corresponds to selecting εa = 1 − Fg,e
the drive frequency which leads to the largest steady-
= [P (e|g) + P (g|e)] /2
state displacement between the coherent g-state and e- (4)
state Purcell-filter-mode responses. This assumes a fixed 1h p i τ
≳ 1 − erf SNR/8 + ,
weak drive power, such that the response is in the linear 2 2T1
regime. We found that this choice consistently leads to
a stronger resonator response for the qubit being in the where P (i|j) is the probability of measuring the state
excited state (ne > ng ). We accredit this to a smaller ef- |i⟩ when having prepared the state |j⟩, and where the
fective linewidth for the excited state for the lower mode, average two-state readout fidelity Fg,e characterizes the
κel < κgl , see Fig. 2 (d). We find this to be an appropri- quality of the readout. The factor two present in the T1
ate choice of drive frequency, as the Kerr nonlinearity limit term arises from the fact that only P (g|e) is affected
imparted on the resonator is weaker for the excited state by loss events.
than the ground state (see Appendix B). In Fig. 4 (a) we present the lowest measured av-
erage assignment errors εa as a function of the
We extract the signal-to-noise ratio (SNR) in terms of
qubit—readout-resonator detuning ∆qr /2π ∈ [−2.7,
power, of the acquired single-shot histograms (see Fig. 3)
−2.4, −1.9, −1.6, −1.3] GHz and as a function of the read-
from a bimodal Gaussian distribution as [38]
out integration times τ ∈ [50, 100, 200, 300, 400] ns.
2 We observe that εa < 1 × 10−2 for τ ≥ 100 ns. When
µg − µe τ ≥ 100 ns and for all qubit—readout-resonator detun-
SNR ≡ , (3)
(σg + σe )/2 ings, the variations in the average assignment error are
small and stay between 2.5 × 10−3 ≤ εa ≤ 1 × 10−2 , ex-
where µg/e and σg/e are, respectively, the mean and the cept for ∆qr /2π = −2.7 GHz and τ = 100 ns. The best
standard deviation of the Gaussian distributions of the assignment error εa = 2.5 × 10−3 is reached at τ = 100 ns
g/e-state responses. In Fig. 3 the solid black line indi- and ∆qr /2π = −1.3 GHz.
cates the distance between the means µg and µe , and This observation suggests that beyond this integration
radii of the black circles are given by the square root of time the assignment error is no longer limited by the
the diagonal covariance matrix elements of the bimodal SNR, which would continue to increase for longer readout
Gaussian distribution. times. This is further demonstrated in Fig. 4 (b), where
 5

(a) 70
= 100 ns

min
a
60
10 4

Assignement error limit,

50
40 10 3

SNR
30
qr/2 = 1.3 GHz
20 qr/2 = 1.6 GHz 10 2
qr/2 = 1.9 GHz
qr/2 = 2.4 GHz
10 qr/2 = 2.7 GHz 10 1

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
ng/ncrit
(b) qr/2 = 1.3 GHz = 400 ns
= 300 ns
= 200 ns
= 100 ns

a
10 1 = 50 ns 7

Min. required SNR

Assignment error,
10 2 22

10 3 38
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75
FIG. 4. (a) Minimum assignment error εa measured as a ng/ncrit
function of the qubit—readout-resonator detuning ∆qr /2π
∈ [−2.7, −2.4, −1.9, −1.6, −1.3] GHz and the readout integra-
tion time τ ∈ [50, 100, 200, 300, 400] ns. Annotated values FIG. 5. (a) SNR as a function of ng /ncrit for the indi-
are in per mille unit of probability: the lowest assignment er- cated qubit—readout-resonator detunings ∆qr and at a fixed
ror εa = 2.5 × 10−3 is reached at a qubit—readout-resonator readout-integration time of τ = 100 ns. The shaded regions
detuning of ∆qr /2π = −1.3 GHz, for an integration time of provide estimates from the analytical solution in the linear
τ = 100 ns. (b) Measured SNR corresponding to the mini- regime, detailed in Appendix C. (b) Average assignment er-
mum assignment error in (a). ror εa as a function of ng /ncrit for the indicated readout
integration times τ at a fixed qubit—readout-resonator de-
tuning ∆qr /2π = −1.3 GHz. Solid lines are plotted for ease
of visualization. Dashed lines correspond to the theoretical
we indicate the measured SNR corresponding to each
limit imposed by thep relaxation time of the qubit given by
lowest measured average assignment error in Fig. 4 (a) εmin = 0.5[1 − erfc( SNR/8)] + τ /2T1 .
a
as a function of the same qubit—readout-resonator de-
tuning and readout-integration-time range. We observe
that a SNR ≥ 30 leads to 2.5 × 10−3 ≤ εa ≤ 1 × 10−2
for τ ≥ 100 ns and for all detunings ∆qr . On the other in line with Eq. (3). We note that the SNR for the small-
hand, SNR ≤ 14 leads to a larger assignment error est detunings ∆qr ∈ [−1.6, −1.3] GHz is significantly
3.87 × 10−2 ≤ εa ≤ 2.87 × 10−1 . In particular, we higher, which we accredit to the increase in the linewidth
g/e
find that the best assignment error εa = 2.5 × 10−3 of the targeted lower mode κl , see Fig. 2 (d). This
is reached for SNR = 48.5. An SNR ≥ 200 leads to increased linewidth results in the pointer states βg/e (t)
assignment errors on the same order as an SNR ≈ 50 reaching the steady state faster, thus maximizing the
(see for example at ∆qr /2π = −1.6 GHz compared to at SNR rate. The shaded region contains the upper- and
∆qr /2π = −1.3 GHz, with τ ∈ [300, 400] ns). lower- bound estimates of the SNR based on uncertain-
We next compare the readout performance in terms ties in the model parameters, see Appendix C.
of SNR at different qubit—readout-resonator detunings In all instances we observe a saturation of the SNR at
as a function of the readout power ng /ncrit , for a fixed a readout power ng ≳ ncrit , where the dispersive approx-
integration time τ = 100 ns, see Fig. 5 (a). The shaded imation is known to break down [32, 41, 42]. This is in
regions indicate the theoretical SNR prediction from the part due to the broadening of the pointer states caused
linear response to the readout drive power [40] by the qubit-induced Kerr nonlinearity of the resonator
Z t (see Fig. 9), measurement-induced state transitions [43]
SNR(t) = 2ηκp |βe (t′ ) − βg (t′ )|2 dt′ , (5) and ionization [44].
0
Finally, we compare in Fig. 5 (b) the average assign-
where η is the measurement efficiency. This expression is ment error εa at different readout integration times τ ∈
 6

10 1
Johnson, '11 rable to the Purcell filter linewidth κp . We showed that
by probing the dispersive regime via flux pulses we can
Dassonneville, '20 increase the effective decay rate of the targeted readout
a

Touzard, '19
Assignment error,

Bultink, '16 mode, thus allowing us to reach larger SNR in a shorter

integration time.
Jeffrey, '14
10 2
Sunada, '22 Our findings open opportunities to study other regimes
Chen, '22 and help optimize the readout parameters in the design
Walter, '17 stage of quantum processors in order to adjust the ef-
Swiadek, '23 fective decay rate of the readout mode depending on
the applications. For instance, we expect this work to
10 3
help reduce the readout contribution to the quantum er-
0 100 200 300 400 500 600 700 800 ror correction cycle time on superconducting qubit plat-
Integration time, (ns) forms [7], without compromising on the readout fidelity
constraints. Such techniques combined with machine
FIG. 6. Two-level average assignment error reported in John-
learning methods as in Ref. [50] for the optimization of
son et al. [41], Jeffrey et al. [30], Bultink et al. [45], Walter
et al. [3], Dassonneville et al. [46], Touzard et al. [47] (blue pulse shapes, could continue to decrease readout times
circles) and in this work (red circle) as a function of the read- while maintaining low readout errors.
out integration time. Jurcevic et al. [48] reached a two-level
assignment error of 3.5 × 10−2 using the excited state promo-
tion technique [49] and is not plotted here.

[50, 100, 200, 300, 400] ns for a fixed qubit—readout- ACKNOWLEDGEMENTS

resonator detuning ∆qr /2π = −1.3 GHz as a function of
the readout power ng /ncrit . For ng /ncrit < 1, we find ex-
cellent agreement between the experimental data (dots) The team in Zurich thanks Johannes Herrmann and
and the approximate theoretical limit (dashed lines) in Stefania Laz̆ar for contributions to the experimental
Eq. (4). Here, we note that the 50 ns measurement is setup. The team in Sherbrooke thanks Cristóbal Lledó
clearly limited by the SNR, which consistently improves and Catherine Leroux for insightful discussions.
at higher drive powers. For 100 ns, the minimum assign- The team in Zurich acknowledges financial support
ment error εmin
a = 2.5 × 10−3 is limited by the intrinsic by the Office of the Director of National Intelligence
lifetime of the qubit T1 rather than the SNR, which can (ODNI), Intelligence Advanced Research Projects
be seen by comparison to the calculated assignment er- Activity (IARPA), through the U.S Army Research
ror (gray dashed line), which plateaus at higher readout Office grant W911NF-16-1-0071, by the EU Flagship
powers. We notice a distinct upturn in the 100 ns mea- on Quantum Technology H2020-FETFLAG-2018-03
surement at higher drive powers, which we attribute to project 820363 OpenSuperQ, by the National Center
non-linearities and measurement-induced transitions. of Competence in Research ’Quantum Science and
For the longer readout times τ ∈ [200, 300, 400] ns, Technology’ (NCCR QSIT), a reseach instrument of the
the minimum assignment error is obtained at lower drive Swiss National Science Foundation (SNSF, grant number
powers, since these measurements reach a larger SNR. 51NF40-185902), by the SNSF R’Equip grant 206021-
Given that the assignment fidelity is limited by the qubit 170731, by the EU-programme H2020-FETOPEN
lifetime, the increase in SNR by increasing drive power project 828826 Quromorphic and by ETH Zurich. S.K
has little impact on the final assignment error εa , as in- acknowledges financial support from Fondation Jean-
dicated by the plateaus (dashed lines). Jacques et Félicia Lopez-Loreta and the ETH Zurich
Foundation. The team in Sherbrooke acknowledges the
financial support by NSERC, the Canada First Research
III. CONCLUSION Excellence Fund, and the Ministère de l’Économie et de
l’Innovation du Québec. Support is also acknowledged
from the U.S. Department of Energy, Office of Science,
We have demonstrated beyond-state-of-the-art single National Quantum Information Science Research Cen-
shot readout reaching a minimum assignment error of ters, Quantum Systems Accelerator. The views and
2.5 × 10−3 in only 100 ns when reducing the qubit detun- conclusions contained herein are those of the authors
ing from the resonator by applying a flux pulse, see our and should not be interpreted as necessarily representing
work in perspective with other techniques in Fig. 6. We the official policies or endorsements, either expressed
provided new insights on dispersive readout for a qubit— or implied, of the ODNI, IARPA or the U.S Government.
readout-resonator—Purcell-filter system, in a strongly
hybridized regime where the coupling strength between
the readout resonator and the Purcell filter J is compa- The authors declare no competing interests.
 7

Flux Drive Readout

1x 1x 1x 1x
Flux DC Drive UHFQA
AWG Source AWG Out In

-23dB UC UC DC

-10dB (iii)

RT
-10dB -20dB -20dB (ii) -20dB
4K
-20dB -20dB -10dB
100 mK Ecco Ecco
Ecco -20dB -20dB 4-8 GHz
9 mK

(i)

1 mm
Quantum Device
Amplifier -x dB Attenuator 50Ω term.
Drive Flux Purcell Readout Coupling Feed (i) Lincoln Lab. TWPA
Qubits
Lines Lines Filters Resonators Elements Lines (ii) LNF HEMT 4-8GHz
(iii) Agile AMT-A0284
Ecco IR filter IQ mixer
Dir. coupler
FIG. 7. False-colour micrograph of the 17-qubit device used Transmon Qubit
Isolator BP filter
for the experiment, adapted from Ref. [7]; the scale bar de-
Qubit Readout
notes 1 mm. The experiment was realized using the qubit in Microwave Circuitry
LP filter
the feedline at the bottom left corner, as presented in Fig. 1. Generator

FIG. 8. Schematic of the wiring and control electronics.

AUTHOR CONTRIBUTIONS The qubit (yellow) on the quantum device is connected to
the room-temperature electronics via flux lines (green), drive
lines (pink) and readout lines (purple) through its readout
S.K., P.M. and F.S. planned the experiments with sup- resonator (red) and Purcell filter (blue). The background col-
port from all co-authors, and S.K., F.S. performed the ex- ors indicate to the temperature stages of the experimental
periments. F.S., R.S., and S.K. analyzed the data. R.S. setup.
and F.S. worked on the theory. F.S. and A.R. designed
the device, and A.R., S.K., D.C.Z. and G.J.N. fabricated
the device. C.H., N.L. and A.R. developed the experi- ground plane on the device and connect signal lines split
mental software framework. S.K., A.R., F.S., C.H. and by crossings. We fabricated aluminium-based Joseph-
N.L. contributed to the experimental setup and main- son junctions using shadow evaporation of aluminium
tained it. F.S., R.S. and S.K. prepared the figures for through a resist mask defined by electron-beam lithog-
the manuscript and S.K., A.W., A.B. and Q.F. provided raphy.
feedback. F.S. and R.S. wrote the manuscript with in- We characterized the properties of the qubit using
puts from all co-authors. A.W., S.K. and A.B. supervised spectroscopy and standard time-domain measurements.
the work. The qubit has an idling frequency ωq /2π = 4.144 GHz,
an anharmonicity α/2π = −181 MHz, a lifetime T1 =
30.4 µs, a Ramsey decay time T2∗ = 29.2 µs, and an echo
decay time T2e = 33.9 µs.
Appendix A: Experimental setup and device
characterization Following the method in Ref. [26], we fit the transmis-
sion amplitude of the readout signal through a feed-line
to the function
We used a qubit of a 17-qubit quantum device, shown
in Fig. 7, to perform the experiment. We fabricated |Sout,in |(ω) = (A + k(ω − ω0 ))×
the 17-qubit quantum processor by sputtering a nio- g/e
bium 150-nm-thin film onto a high-resistivity intrinsic κp (−2i∆r ) (A1)
cos(ϕ) − eiϕ g/e
,
silicon substrate. All coplanar waveguides, capacitors 4J 2 + (κp − 2i∆p )(−2i∆r )
and qubit islands were patterned using photolithogra-
phy and reactive-ion etching. The aluminium-titanium- where A is the amplitude, k describes a tilt in the spec-
aluminium trilayer airbridges establish a well-connected trum centered at ω0 , ϕ is a phase rotation induced by the
 8

qubit—readout-resonator detuning ∆qr /2π GHz -2.7 -2.4 -1.9 -1.6 -1.3
Qubit frequency during readout ωq /2π MHz 4144 4500 5000 5300 5600
Bare readout resonator frequency ωr,b /2π MHz 6854.63 6858.02 6857.98 6859.74 6864.86
Dressed readout resonator frequency ωrg /2π MHz 6876.27 6881.98 6896.09 6906.33 6928.43
Purcell filter frequency ωp /2π MHz 6899.86 6899.86 6899.86 6899.86 6899.86
Readout drive frequency ωd /2π MHz 6857.4 6861.2 6870.0 6874.0 6881.6
Qubit readout resonator coupling gb /2π MHz 224.32 205.61 211.49 204.2 205.53
Qubit charge-readout resonator coupling g/2π MHz 284.01 271.40 293.71 292.27 302.34
Readout resonator-Purcell filter coupling J/2π MHz 27.9 27.9 27.9 27.9 27.9
Low mode linewidth, qubit in |g⟩ state κgl /2π MHz 10.16 11.61 15.81 19.07 25.00
Low mode linewidth, qubit in |e⟩ state κel /2π MHz 8.88 10.03 12.66 14.84 19.87
High mode linewidth, qubit in |g⟩ state κgh /2π MHz 23.86 22.41 18.21 14.95 9.02
High mode linewidth, qubit in |e⟩ state κeh /2π MHz 25.14 23.99 21.36 19.18 14.15
Low mode dispersive shift 2χl /2π MHz -4.17 -4.35 -6.11 -6.69 -6.31
High mode dispersive shift 2χh /2π MHz -1.50 -1.90 -4.24 -6.64 -13.18
Critical readout resonator photon number ncrit 23.14 19.26 10.42 7.55 4.83

TABLE I. List of readout parameters extracted for the qubit—readout-resonator detunings ∆qr /2π spanning -2.7 GHz to -
1.3 GHz using pulsed-spectroscpy measurements.

capacitive couplings to other lines, κp is the external cou- Appendix B: Model

pling rate of the Purcell filter, ∆p = ω−ωp is the detuning
between the drive frequency ω and the Purcell-filter fre- To model the system, we use the Hamiltonian
g/e g/e
quency ωp , and ∆r = ω − ωr is the detuning between
the drive frequency and the resonator frequency condi- Ĥtrp = 4Ec n̂2t − EJ (Φ) cos φ̂t
tioned on the state of the qubit. The relevant parameters + ωr,b ↠â − ig(n̂t − ng )(â − ↠)
for the studied qubit at the indicated qubit-resonator de- (B1)
tunings are provided in Table I. + ωp fˆ† fˆ + J(fˆ† − fˆ)(↠− â)
+ 2iE sin(ωd t)(fˆ† − fˆ),
We installed the device in a magnetically-shielded sam- where n̂t is the charge operator of the transmon, â the
ple holder mounted at the base plate (9 mK) of a cryo- readout resonator mode creation operator and fˆ the Pur-
genic measurement setup [51] and connected it to the cell filter mode creation operator. Ec is the charging en-
control and measurement electronics as shown in Fig. 8. ergy of the transmon, EJ (Φ) the flux-tunable Josephson
We use a DC signal to generate a current inducing a energy of the transmon, ωr,b , ωp the bare resonator and
magnetic flux in the SQUID-loop of the transmon qubit, Purcell filer frequencies, and g, J the transmon-resonator
to control its idle frequency. We use arbitrary waveform and resonator-Purcell coupling rates respectively. E, ωd
generators to apply voltage pulses (2.4 GSa/s sampling are the the drive amplitude and drive frequency. Further,
rate) to the qubit to tune its frequency for readout. The we consider a master equation
DC and AWG signals are combined using a bias-tee. A
precompensation of distortions in the flux line is applied, ρ̂˙ = −i[Ĥtrp , ρ̂] + κp D[fˆ], (B2)
as in Ref. [7]. where κp is the coupling rate between the Purcell fil-
ter and the feedline. We first diagonalize the transmon-
We perform the single-shot readout experiments with resonator subsystem. We follow the notation of Ref. [21],
an ultra-high frequency quantum analyzer (UHFQA) and assume a Kerr-nonlinear oscillator model for the
by using an IQ-mixer to upconvert the frequency- transmon, valid in the low readout power regime. A
multiplexed readout pulses from an intermediate fre- Schrieffer-Wolff transformation yields an effective Hamil-
quency signal sampled at 1.8 GSa/s to the gigahertz tonian
frequency range of the readout circuitry. At the Ĥ = ω̄q b̂† b̂ + ωp fˆ† fˆ + ω g ↠â + 2χ↠âb̂† b̂ (B3)
r
output of the device feedline, the readout signal
α Ec 4 †2 2
passes through a wide-bandwidth near-quantum lim- − 2λ′ λ3 EC â†2 â2 b̂† b̂ − b̂†2 b̂2 − λ â â
ited traveling-wave parametric amplifier (TWPA) [52], a h i 2 2 
high-electron-mobility transistor (HEMT) amplifier, and + J 1 − 2λλ′ b̂† b̂ ↠fˆ + λb̂† fˆ + H.c. ,
room-temperature amplifiers. It is then down-converted
with an IQ-mixer and digitally demodulated and inte- where
grated in the UHFQA. Further details on the device fab- χ = −g 2 EC /(∆qr (∆qr − EC )), (B4)
rication, characterization, and the experimental setup,
λ′ = λEc / ∆qr + Ec (1 − 2λ2 )
 
can be found in Ref. [7]. (B5)
 9

(a) 100 (b) 100

Overlap error P(e|g)

Overlap error P(g|e)

10 1

10 1

10 2
qr/2 = 2.7 GHz
qr/2 = 2.4 GHz
qr/2 = 1.9 GHz
qr/2 = 1.6 GHz 10 2
10 3
qr/2 = 1.3 GHz

0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 1 2 3 4 5

ng/ncrit ne/ncrit
(c) 0.55 (d) 0.55
0.50 0.50
g

g
0.45 0.45
0.40 0.40
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 1 2 3 4 5
ng/ncrit ne/ncrit
(e) (f)
0.55 0.55
0.50 0.50
e

e
0.45 0.45

0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 1 2 3 4 5

ng/ncrit ne/ncrit
FIG. 9. (a), (b) Overlap error P (e|g) and P (g|e) for the indicated qubit—readout-resonator frequency detunings ∆qr at a fixed
integration time of τ = 100 ns. (c), (d) Gaussian width σg of the ground state single-shot histogram as a function of ng /ncrit
and ne /ncrit . (e), (f ) Gaussian width σe of the ground state single-shot histogram as a function of ng /ncrit and ne /ncrit .

with λ = g/∆qr and ∆qr = ωq − ωr,b . Further, the Appendix C: Linear response
qubit and resonator frequencies become Lamb-shifted,
with ω̄q ≈ ωq + g 2 /∆qr , ωrg ≈ ωr,b − g 2 /∆qr . The con- For sufficiently small drive amplitudes, we can assume
tribution −2λ′ λ3 EC â†2 â2 b̂† b̂ normalizes down the effec- negligible impact from the Kerr nonlinearity and take the
tive Kerr nonlinearity when the qubit is in the excited resonator and filter responses to be linear. As such, we
state, and we note that Ke /Kg ≈ 1 + 4λ′ /λ, where can use the relation
4λ′ /λ < 0 for ∆qr < 0. This nonlinearity leads to a  g/e   g/e   g/e 
significantly larger increase in the Gaussian width of the α̇ ωr − ωd J g/e α
g/e = − i g/e κp
β g/e
ground state response than the excited state response, as β̇ J ωp − ωd − i 2
(C1)
seen in Fig. 9 (c,d,e,f) [53, 54]. For this reason, we quote
 
0
the drive power in the main text as a function of ng /ncrit + ,
E
as opposed to ne /ncrit .
where α and β represent the coherent fields of the read-
We also note a correlation between the broadening of out resonator and Purcell filter respectively, E is the
the ground state response and an increase in the overlap drive amplitude and J g/e = J [1 − λλ′ (⟨σ̂z ⟩ + 1)] is the
errors P (g|e) and P (e|g) for ng /ncrit ≳ 1 where non- effective readout-resonator—Purcell-filter coupling, and
linear effects are expected to be more important, see ωre = ωrg + 2χ. Diagonalizing the equation of motion in
Fig. 9 (a,b). Frequency renormalizations from the ef- the absence of a drive (E = ωd = 0) yields eigenvalues
fective coupling of the filter to the qubit are on the order g/e
ωr + ωp − iκp /2
of J 2 λ2 /∆2qr and can be safely ignored in this regime, g/e
λl,h =
which was corroborated by numerical diagonalization of s 2 2 (C2)
Eq. (B1) and simulation of the master equation. More 1 g/e iκp
importantly, we note that the effective coupling strength ± ∆rp + + 4J 2,g/e .
2 2
between the resonator and Purcell filter, J[1 − 2λλ′ b̂† b̂],
only weakly depends on the qubit state. For 4J ≫ κ, the eigenvalues approximately correspond
 10

70 14

AC Stark shift, | ac|/2 (MHz)

Y Quadrature (arb. units) 250 60 12

Photon number, ng
200 50 10
40 8
150
g, ME 30 6
100 e, ME 20 4
50 g, SC. 10 2
e, SC.
0 0 0
125 100 75 50 25 0 25 0.0 0.2 0.4 0.6 0.8 1.0
X Quadrature (arb. units) Normalized readout power, P
FIG. 10. Example comparison of the master equation FIG. 11. Measured (dots) ac-Stark shift ∆ac of the qubit pre-
βg/e,M E , against the semi-classical trajectories βg/e,SC , plot- pared in the ground state at ωq /2π = 4.14 GHz as a function
ted in the phase space of the Purcell filter mode for ∆qr /2π = of the normalized readout power (real range spans 2.0 µV to
−1.3 GHz and E/2π = 10 MHz. Solid lines correspond to 12.5 µV). The corresponding inferred resonator photon num-
the master equation expectations of the resonator and fil- ber ng = ∆ac /2(χl + χh ) (solid line) is shown on the right
ter modes, ⟨a⟩, and ⟨f ⟩, respectively, where ⟨X⟩ = Re{a, f }, axis.
⟨Y ⟩ = Im{a, f }. Transparent lines correspond to the semi-
classical solution.
The relevant frequencies and linewidths extracted from
the normal-mode Hamiltonian in Eq. (C1) are plotted in
to a normal mode, where the indices l, h corresponds to Fig. 2.
the low and high mode respectively. In this fashion, the The steady state expressions are found to be
g/e
real and imaginary components of λl,h correspond to the " #
g/e  
frequency and linewidth of these low and high modes αss E −J
= . (C6)
g/e
βss ∆g/e (∆p − iκp /2) − J 2 ∆g/e
g/e g/e g/e g/e
ωl,h = Re[λl,h ], κl,h = −2Im[λl,h ]. (C3)
Using this expression, the full time-dependent response
To obtain a qualitative understanding of the eigenvalues, takes the form
g/e
we perform an expansion of the square root in (∆rp +
iκ/2). Assuming J g/e ≈ J, this yields λh − (ωr + χ⟨σz ⟩) e−i(λh −ωd )t
β g/e (t) = βss
g/e
−E
d (λh − ωd )
g/e (C7)
g/e ωr+ ωp − iκp /2 λl − (ωr + χ⟨σz ⟩) e−i(λl −ωd )t
λl,h = +E ,
2 d (λl − ωd )
g/e
! (C4)
∆rp − i∆rp κp − κ2 /4
2ge
± J+ .
p
8J where d = (−∆rp − iκp /2 − χ⟨σz ⟩)2 + 4J 2 .
We then use Eq. (C7) to express the SNR as [40]
g/e t
Consequently, we see that the frequency ωl,h of the two
Z
sets of normal modes are approximately separated by 2J, SNR(t) = 2ηκp |βe (t′ ) − βg (t′ )|2 dt′ , (C8)
0
with the relative dispersive shift of each mode χl,h be-
tween the low and high mode being where η is the measurement efficiency. We note that
this expression is the square of the often used expression
∆grp χ + χ2 but is in line with Eq. (3). We then plot these results
2χl = ωle − ωlg ≈ χ − ,
2J for Fig. 5 (a) allowing for a ±1 MHz deviation in the
∆grp χ + χ2 calculated values of g, J, ωr and κp to allow for uncer-
2χh = ωhe − ωhg ≈ χ + , tainties in the fitted parameters and nonidealities caused
2J
g/e (C5) by spurious couplings to two-level systems, alongside a
g/e κp ∆rp κp variation of up to 5% in the measurement efficiency at
κl ≈ + ,
2 4J different frequencies. The shaded region contains the up-
g/e
g/e κp ∆rp κp per and low bound estimates of the SNR based on these
κh ≈ − . uncertainties.
2 4J
Finally, we verify the validity of the semiclassical ap-
Noting that ∆erp < ∆grp , we see that κgh < κeh and κgl > κel proximation. Negligible difference was noted in the tra-
for ∆grp < 0, and vice versa for ∆grp > 0. jectories in phase space between the expected internal
 11

coherent fields ⟨â⟩, ⟨fˆ⟩, calculated by solving the mas- of variable power and a π-pulse of variable frequency.
ter equation (B2), and the corresponding semiclassical We measure the excited state population as a function
predictions α and β of the resonator and Purcell filter of the drive pulse frequency for each readout power for
respectively, confirming that the semiclassical model (in the ωq /2π = 4.14 GHz qubit frequency. The frequency
Appendix C) captures the state separation at low pow- at which the excited state population is maximum cor-
ers. Example trajectories at low power for the Purcell responds to the instantaneous and ac-Stark shifted qubit
filter mode are plotted for the ∆qr /2π = −1.3 GHz case frequency.
in Fig. 10. We determine this frequency using a Gaussian fit. We
infer and then calibrate the steady state readout res-
onator photon number ng with the qubit prepared in
Appendix D: Photon number and drive power the ground state from ng = ∆ac /2(χl + χh ) for the
calibration chosen drive powers [55]. Then, using the steady state
resonator response from Eq. (C1), this allows us to ex-
We measure the ac-Stark shift ∆ac caused on the qubit tract the effective drive amplitudes E. The steady state
prepared in the ground state by the readout resonator as resonator responses for the qubit-resonator detunings
a function of power, see Fig. 11. To this mean we simul- ∆qr /2π ∈ [−4.5, −5.0, −5.3, −5.6] GHz are then inferred
taneously apply a readout tone with a length of 0.6 µs from Eq. (C6) at the same drive powers.

[1] F. Mallet, F. R. Ong, A. Palacios-Laloy, F. Nguyen, rington, O. Higgott, J. Hilton, M. Hoffmann, S. Hong,
P. Bertet, D. Vion, and D. Esteve, Single-shot qubit read- T. Huang, A. Huff, W. J. Huggins, L. B. Ioffe, S. V.
out in circuit quantum electrodynamics, Nat. Phys. 5, Isakov, J. Iveland, E. Jeffrey, Z. Jiang, C. Jones, P. Juhas,
791 (2009). D. Kafri, K. Kechedzhi, J. Kelly, T. Khattar, M. Khezri,
[2] M. D. Reed, L. DiCarlo, B. R. Johnson, L. Sun, D. I. M. Kieferová, S. Kim, A. Kitaev, P. V. Klimov, A. R.
Schuster, L. Frunzio, and R. J. Schoelkopf, High-fidelity Klots, A. N. Korotkov, F. Kostritsa, J. M. Kreikebaum,
readout in circuit quantum electrodynamics using the D. Landhuis, P. Laptev, K.-M. Lau, L. Laws, J. Lee,
Jaynes-Cummings nonlinearity, Phys. Rev. Lett. 105, K. Lee, B. J. Lester, A. Lill, W. Liu, A. Locharla,
173601 (2010). E. Lucero, F. D. Malone, J. Marshall, O. Martin,
[3] T. Walter, P. Kurpiers, S. Gasparinetti, P. Mag- J. R. McClean, T. McCourt, M. McEwen, A. Megrant,
nard, A. Potočnik, Y. Salathé, M. Pechal, M. Mondal, B. Meurer Costa, X. Mi, K. C. Miao, M. Mohseni,
M. Oppliger, C. Eichler, and A. Wallraff, Rapid, high- S. Montazeri, A. Morvan, E. Mount, W. Mruczkiewicz,
fidelity, single-shot dispersive readout of superconducting O. Naaman, M. Neeley, C. Neill, A. Nersisyan, H. Neven,
qubits, Phys. Rev. Appl. 7, 054020 (2017). M. Newman, J. H. Ng, A. Nguyen, M. Nguyen, M. Y.
[4] A. Y. Kitaev, Fault-tolerant quantum computation by Niu, T. E. O’Brien, A. Opremcak, J. Platt, A. Petukhov,
anyons, Annals of Physics 303, 2 (2003). R. Potter, L. P. Pryadko, C. Quintana, P. Roushan,
[5] D. P. DiVincenzo, Fault-tolerant architectures for super- N. C. Rubin, N. Saei, D. Sank, K. Sankaragomathi, K. J.
conducting qubits, Phys. Scr. 2009, 014020 (2009). Satzinger, H. F. Schurkus, C. Schuster, M. J. Shearn,
[6] C. K. Andersen, A. Remm, S. Lazar, S. Krinner, A. Shorter, V. Shvarts, J. Skruzny, V. Smelyanskiy, W. C.
N. Lacroix, G. J. Norris, M. Gabureac, C. Eichler, and Smith, G. Sterling, D. Strain, M. Szalay, A. Torres, G. Vi-
A. Wallraff, Repeated quantum error detection in a sur- dal, B. Villalonga, C. Vollgraff Heidweiller, T. White,
face code, Nature Physics 16, 875 (2020). C. Xing, Z. J. Yao, P. Yeh, J. Yoo, G. Young, A. Zalcman,
[7] S. Krinner, N. Lacroix, A. Remm, A. D. Paolo, E. Genois, Y. Zhang, N. Zhu, and G. Q. AI, Suppressing quantum
C. Leroux, C. Hellings, S. Lazar, F. Swiadek, J. Her- errors by scaling a surface code logical qubit, Nature 614,
rmann, G. J. Norris, C. K. Andersen, M. Müller, A. Blais, 676 (2023).
C. Eichler, and A. Wallraff, Realizing repeated quantum [9] C. H. Bennett, G. Brassard, C. Crépeau, R. Jozsa,
error correction in a distance-three surface code, Nature A. Peres, and W. K. Wootters, Teleporting an unknown
605, 669 (2022). quantum state via dual classical and Einstein-Podolsky-
[8] R. Acharya, I. Aleiner, R. Allen, T. I. Andersen, M. Ans- Rosen channels, Phys. Rev. Lett. 70, 1895 (1993).
mann, F. Arute, K. Arya, A. Asfaw, J. Atalaya, R. Bab- [10] D. Gottesman and I. L. Chuang, Demonstrating the via-
bush, D. Bacon, J. C. Bardin, J. Basso, A. Bengtsson, bility of universal quantum computation using teleporta-
S. Boixo, G. Bortoli, A. Bourassa, J. Bovaird, L. Brill, tion and single-qubit operations, Nature 402, 390 (1999).
M. Broughton, B. B. Buckley, D. A. Buell, T. Burger, [11] L. Steffen, Y. Salathe, M. Oppliger, P. Kurpiers, M. Baur,
B. Burkett, N. Bushnell, Y. Chen, Z. Chen, B. Chiaro, C. Lang, C. Eichler, G. Puebla-Hellmann, A. Fedorov,
J. Cogan, R. Collins, P. Conner, W. Courtney, A. L. and A. Wallraff, Deterministic quantum teleportation
Crook, B. Curtin, D. M. Debroy, A. Del Toro Barba, with feed-forward in a solid state system, Nature 500,
S. Demura, A. Dunsworth, D. Eppens, C. Erickson, 319 (2013).
L. Faoro, E. Farhi, R. Fatemi, L. Flores Burgos, E. Fo- [12] J. Qiu, Y. Liu, J. Niu, L. Hu, Y. Wu, L. Zhang,
rati, A. G. Fowler, B. Foxen, W. Giang, C. Gidney, W. Huang, Y. Chen, J. Li, S. Liu, Y. Zhong, L. Duan,
D. Gilboa, M. Giustina, A. Grajales Dau, J. A. Gross, and D. Yu, Deterministic quantum teleportation be-
S. Habegger, M. C. Hamilton, M. P. Harrigan, S. D. Har- tween distant superconducting chips, arXiv:2302.08756
 12

10.48550/arXiv:2302.08756 (2023). fidelity multiplexed readout of superconducting qubits,

[13] C. H. Bennett, H. J. Bernstein, S. Popescu, and B. Schu- Phys. Rev. Appl. 10, 034040 (2018).
macher, Concentrating partial entanglement by local op- [27] F. Arute, K. Arya, R. Babbush, D. Bacon, J. C.
erations, Phys. Rev. A 53, 2046 (1996). Bardin, R. Barends, R. Biswas, S. Boixo, F. G. S. L.
[14] S. Bravyi and A. Kitaev, Universal quantum computa- Brandao, D. A. Buell, B. Burkett, Y. Chen, Z. Chen,
tion with ideal clifford gates and noisy ancillas, Phys. B. Chiaro, R. Collins, W. Courtney, A. Dunsworth,
Rev. A 71, 022316 (2005). E. Farhi, B. Foxen, A. Fowler, C. Gidney, M. Giustina,
[15] J. E. Johnson, C. Macklin, D. H. Slichter, R. Vijay, E. B. R. Graff, K. Guerin, S. Habegger, M. P. Harrigan,
Weingarten, J. Clarke, and I. Siddiqi, Heralded state M. J. Hartmann, A. Ho, M. Hoffmann, T. Huang,
preparation in a superconducting qubit, Phys. Rev. Lett. T. S. Humble, S. V. Isakov, E. Jeffrey, Z. Jiang,
109, 050506 (2012). D. Kafri, K. Kechedzhi, J. Kelly, P. V. Klimov, S. Knysh,
[16] D. Ristè, J. G. van Leeuwen, H.-S. Ku, K. W. Lehn- A. Korotkov, F. Kostritsa, D. Landhuis, M. Lind-
ert, and L. DiCarlo, Initialization by measurement of a mark, E. Lucero, D. Lyakh, S. Mandrà, J. R. Mc-
superconducting quantum bit circuit, Phys. Rev. Lett. Clean, M. McEwen, A. Megrant, X. Mi, K. Michielsen,
109, 050507 (2012). M. Mohseni, J. Mutus, O. Naaman, M. Neeley, C. Neill,
[17] Y. Salathé, M. Mondal, M. Oppliger, J. Heinsoo, M. Y. Niu, E. Ostby, A. Petukhov, J. C. Platt, C. Quin-
P. Kurpiers, A. Potočnik, A. Mezzacapo, U. Las Heras, tana, E. G. Rieffel, P. Roushan, N. C. Rubin, D. Sank,
L. Lamata, E. Solano, S. Filipp, and A. Wallraff, Digital K. J. Satzinger, V. Smelyanskiy, K. J. Sung, M. D. Tre-
quantum simulation of spin models with circuit quantum vithick, A. Vainsencher, B. Villalonga, T. White, Z. J.
electrodynamics, Phys. Rev. X 5, 021027 (2015). Yao, P. Yeh, A. Zalcman, H. Neven, and J. M. Marti-
[18] J. Herrmann, S. M. Llima, A. Remm, P. Zapletal, N. A. nis, Quantum supremacy using a programmable super-
McMahon, C. Scarato, F. Swiadek, C. K. Andersen, conducting processor, Nature 574, 505 (2019).
C. Hellings, S. Krinner, N. Lacroix, S. Lazar, M. Ker- [28] Z. Chen, K. J. Satzinger, J. Atalaya, A. N. Korotkov,
schbaum, D. C. Zanuz, G. J. Norris, M. J. Hartmann, A. Dunsworth, D. Sank, C. Quintana, M. McEwen,
A. Wallraff, and C. Eichler, Realizing quantum convo- R. Barends, P. V. Klimov, S. Hong, C. Jones,
lutional neural networks on a superconducting quantum A. Petukhov, D. Kafri, S. Demura, B. Burkett, C. Gid-
processor to recognize quantum phases, Nature Comm. ney, A. G. Fowler, A. Paler, H. Putterman, I. Aleiner,
13, 4144 (2022). F. Arute, K. Arya, R. Babbush, J. C. Bardin, A. Bengts-
[19] A. Blais, R.-S. Huang, A. Wallraff, S. M. Girvin, and son, A. Bourassa, M. Broughton, B. B. Buckley, D. A.
R. J. Schoelkopf, Cavity quantum electrodynamics for Buell, N. Bushnell, B. Chiaro, R. Collins, W. Courtney,
superconducting electrical circuits: An architecture for A. R. Derk, D. Eppens, C. Erickson, E. Farhi, B. Foxen,
quantum computation, Phys. Rev. A 69, 062320 (2004). M. Giustina, A. Greene, J. A. Gross, M. P. Harrigan,
[20] A. Wallraff, D. I. Schuster, A. Blais, L. Frunzio, J. Ma- S. D. Harrington, J. Hilton, A. Ho, T. Huang, W. J.
jer, M. H. Devoret, S. M. Girvin, and R. J. Schoelkopf, Huggins, L. B. Ioffe, S. V. Isakov, E. Jeffrey, Z. Jiang,
Approaching unit visibility for control of a superconduct- K. Kechedzhi, S. Kim, A. Kitaev, F. Kostritsa, D. Land-
ing qubit with dispersive readout, Phys. Rev. Lett. 95, huis, P. Laptev, E. Lucero, O. Martin, J. R. McClean,
060501 (2005). T. McCourt, X. Mi, K. C. Miao, M. Mohseni, S. Mon-
[21] A. Blais, A. L. Grimsmo, S. M. Girvin, and A. Wallraff, tazeri, W. Mruczkiewicz, J. Mutus, O. Naaman, M. Nee-
Circuit quantum electrodynamics, Rev. Mod. Phys. 93, ley, C. Neill, M. Newman, M. Y. Niu, T. E. O’Brien,
025005 (2021). A. Opremcak, E. Ostby, B. Pató, N. Redd, P. Roushan,
[22] J. Koch, T. M. Yu, J. Gambetta, A. A. Houck, D. I. N. C. Rubin, V. Shvarts, D. Strain, M. Szalay, M. D.
Schuster, J. Majer, A. Blais, M. H. Devoret, S. M. Girvin, Trevithick, B. Villalonga, T. White, Z. J. Yao, P. Yeh,
and R. J. Schoelkopf, Charge-insensitive qubit design de- J. Yoo, A. Zalcman, H. Neven, S. Boixo, V. Smelyanskiy,
rived from the Cooper pair box, Phys. Rev. A 76, 042319 Y. Chen, A. Megrant, J. Kelly, and A. I. Google Quan-
(2007). tum, Exponential suppression of bit or phase errors with
[23] V. Negirneac, H. Ali, N. Muthusubramanian, F. Battis- cyclic error correction, Nature 595, 383 (2021).
tel, R. Sagastizabal, M. S. Moreira, J. F. Marques, W. J. [29] M. D. Reed, B. R. Johnson, A. A. Houck, L. DiCarlo,
Vlothuizen, M. Beekman, C. Zachariadis, N. Haider, J. M. Chow, D. I. Schuster, L. Frunzio, and R. J.
A. Bruno, and L. DiCarlo, High-fidelity controlled-z Schoelkopf, Fast reset and suppressing spontaneous emis-
gate with maximal intermediate leakage operating at sion of a superconducting qubit, Appl. Phys. Lett. 96,
the speed limit in a superconducting quantum processor, 203110 (2010).
Phys. Rev. Lett. 126, 220502 (2021). [30] E. Jeffrey, D. Sank, J. Y. Mutus, T. C. White, J. Kelly,
[24] L. DiCarlo, M. D. Reed, L. Sun, B. R. Johnson, J. M. R. Barends, Y. Chen, Z. Chen, B. Chiaro, A. Dunsworth,
Chow, J. M. Gambetta, L. Frunzio, S. M. Girvin, M. H. A. Megrant, P. J. J. O’Malley, C. Neill, P. Roushan,
Devoret, and R. J. Schoelkopf, Preparation and measure- A. Vainsencher, J. Wenner, A. N. Cleland, and J. M.
ment of three-qubit entanglement in a superconducting Martinis, Fast accurate state measurement with super-
circuit, Nature 467, 574 (2010). conducting qubits, Phys. Rev. Lett. 112, 190504 (2014).
[25] F. W. Strauch, P. R. Johnson, A. J. Dragt, C. J. Lobb, [31] N. T. Bronn, Y. Liu, J. B. Hertzberg, A. D. Córcoles,
J. R. Anderson, and F. C. Wellstood, Quantum logic A. A. Houck, J. M. Gambetta, and J. M. Chow, Broad-
gates for coupled superconducting phase qubits, Phys. band filters for abatement of spontaneous emission in cir-
Rev. Lett. 91, 167005 (2003). cuit quantum electrodynamics, Applied Physics Letters
[26] J. Heinsoo, C. K. Andersen, A. Remm, S. Krinner, 107, 172601 (2015).
T. Walter, Y. Salathé, S. Gasparinetti, J.-C. Besse, [32] M. Boissonneault, J. M. Gambetta, and A. Blais, Im-
A. Potočnik, A. Wallraff, and C. Eichler, Rapid high- proved superconducting qubit readout by qubit-induced
 13

nonlinearities, Phys. Rev. Lett. 105, 100504 (2010). rahimi, H. J. Carmichael, and M. H. Devoret, To catch
[33] Y. Sunada, S. Kono, J. Ilves, S. Tamate, T. Sugiyama, and reverse a quantum jump mid-flight, Nature 570, 200
Y. Tabuchi, and Y. Nakamura, Fast readout and reset (2019).
of a superconducting qubit coupled to a resonator with [43] D. Sank, Z. Chen, M. Khezri, J. Kelly, R. Barends,
an intrinsic purcell filter, Phys. Rev. Applied 17, 044016 B. Campbell, Y. Chen, B. Chiaro, A. Dunsworth,
(2022). A. Fowler, and et al., Measurement-induced state tran-
[34] M. Khezri, A. Opremcak, Z. Chen, A. Bengtsson, sitions in a superconducting qubit: Beyond the rotat-
T. White, O. Naaman, R. Acharya, K. Anderson, ing wave approximation, Physical Review Letters 117,
M. Ansmann, F. Arute, K. Arya, A. Asfaw, J. C. Bardin, 190503 (2016).
A. Bourassa, J. Bovaird, L. Brill, B. B. Buckley, D. A. [44] R. Shillito, A. Petrescu, J. Cohen, J. Beall, M. Hauru,
Buell, T. Burger, B. Burkett, N. Bushnell, J. Campero, M. Ganahl, A. G. Lewis, G. Vidal, and A. Blais, Dynam-
B. Chiaro, R. Collins, A. L. Crook, B. Curtin, S. Demura, ics of transmon ionization, Phys. Rev. Appl. 18, 034031
A. Dunsworth, C. Erickson, R. Fatemi, V. S. Ferreira, (2022).
L. F. Burgos, E. Forati, B. Foxen, G. Garcia, W. Gi- [45] C. C. Bultink, M. A. Rol, T. E. O’Brien, X. Fu, B. C. S.
ang, M. Giustina, R. Gosula, A. G. Dau, M. C. Hamil- Dikken, C. Dickel, R. F. L. Vermeulen, J. C. de Sterke,
ton, S. D. Harrington, P. Heu, J. Hilton, M. R. Hoff- A. Bruno, R. N. Schouten, and L. DiCarlo, Active res-
mann, S. Hong, T. Huang, A. Huff, J. Iveland, E. Jef- onator reset in the nonlinear dispersive regime of circuit
frey, J. Kelly, S. Kim, P. V. Klimov, F. Kostritsa, J. M. QED, Phys. Rev. Applied 6, 034008 (2016).
Kreikebaum, D. Landhuis, P. Laptev, L. Laws, K. Lee, [46] R. Dassonneville, T. Ramos, V. Milchakov, L. Planat,
B. J. Lester, A. T. Lill, W. Liu, A. Locharla, E. Lucero, E. Dumur, F. Foroughi, J. Puertas, S. Leger, K. Bharad-
S. Martin, M. McEwen, A. Megrant, X. Mi, K. C. Miao, waj, J. Delaforce, C. Naud, W. Hasch-Guichard, J. J.
S. Montazeri, A. Morvan, M. Neeley, C. Neill, A. Ner- Garcı́a-Ripoll, N. Roch, and O. Buisson, Fast high-
sisyan, J. H. Ng, A. Nguyen, M. Nguyen, R. Potter, fidelity quantum nondemolition qubit readout via a
C. Quintana, C. Rocque, P. Roushan, K. Sankarago- nonperturbative cross-kerr coupling, Phys. Rev. X 10,
mathi, K. J. Satzinger, C. Schuster, M. J. Shearn, 011045 (2020).
A. Shorter, V. Shvarts, J. Skruzny, W. C. Smith, G. Ster- [47] S. Touzard, A. Kou, N. E. Frattini, V. V. Sivak, S. Puri,
ling, M. Szalay, D. Thor, A. Torres, B. W. K. Woo, A. Grimm, L. Frunzio, S. Shankar, and M. H. Devoret,
Z. J. Yao, P. Yeh, J. Yoo, G. Young, N. Zhu, N. Zo- Gated conditional displacement readout of superconduct-
brist, D. Sank, A. Korotkov, Y. Chen, and V. Smelyan- ing qubits, Phys. Rev. Lett. 122, 080502 (2019).
skiy, Measurement-induced state transitions in a super- [48] P. Jurcevic, A. Javadi-Abhari, L. S. Bishop, I. Lauer,
conducting qubit: Within the rotating wave approxima- D. F. Bogorin, M. Brink, L. Capelluto, O. Günlük,
tion, arXiv 10.48550/arXiv:2212.05097 (2022). T. Itoko, N. Kanazawa, A. Kandala, G. A. Keefe, K. Kr-
[35] M. Malekakhlagh, W. Shanks, H. Paik, M. Malekakhlagh, sulich, W. Landers, E. P. Lewandowski, D. T. McClure,
W. Shanks, and H. Paik, Optimization of the resonator- G. Nannicini, A. Narasgond, H. M. Nayfeh, E. Pritchett,
induced phase gate for superconducting qubits, Phys. M. B. Rothwell, S. Srinivasan, N. Sundaresan, C. Wang,
Rev. A 105, 022607 (2022). K. X. Wei, C. J. Wood, J.-B. Yau, E. J. Zhang, O. E.
[36] M. D. Hutchings, J. B. Hertzberg, Y. Liu, N. T. Bronn, Dial, J. M. Chow, and J. M. Gambetta, Demonstration
G. A. Keefe, M. Brink, J. M. Chow, and B. L. T. Plourde, of quantum volume 64 on a superconducting quantum
Tunable superconducting qubits with flux-independent computing system, Quantum Science and Technology 6,
coherence, Phys. Rev. Appl. 8, 044003 (2017). 025020 (2021).
[37] C. W. Gardiner and M. J. Collett, Input and output in [49] S. S. Elder, C. S. Wang, P. Reinhold, C. T. Hann,
damped quantum systems: Quantum stochastic differen- K. S. Chou, B. J. Lester, S. Rosenblum, L. Frunzio,
tial equations and the master equation, Phys. Rev. A 31, L. Jiang, and R. J. Schoelkopf, High-fidelity measurement
3761 (1985). of qubits encoded in multilevel superconducting circuits,
[38] J. Gambetta, W. A. Braff, A. Wallraff, S. M. Girvin, Phys. Rev. X 10, 011001 (2020).
and R. J. Schoelkopf, Protocols for optimal readout of [50] B. Lienhard, A. Vepsäläinen, L. C. G. Govia, C. R. Hof-
qubits using a continuous quantum nondemolition mea- fer, J. Y. Qiu, D. Ristè, M. Ware, D. Kim, R. Winik,
surement, Phys. Rev. A 76, 012325 (2007). A. Melville, B. Niedzielski, J. Yoder, G. J. Ribeill,
[39] P. Magnard, P. Kurpiers, B. Royer, T. Walter, J.-C. T. A. Ohki, H. K. Krovi, T. P. Orlando, S. Gustavsson,
Besse, S. Gasparinetti, M. Pechal, J. Heinsoo, S. Storz, and W. D. Oliver, Deep-neural-network discrimination
A. Blais, and A. Wallraff, Fast and unconditional all- of multiplexed superconducting-qubit states, Phys. Rev.
microwave reset of a superconducting qubit, Phys. Rev. Applied 17, 014024 (2022).
Lett. 121, 060502 (2018). [51] S. Krinner, S. Storz, P. Kurpiers, P. Magnard, J. Heinsoo,
[40] C. C. Bultink, B. Tarasinski, N. Haandbæk, S. Poletto, R. Keller, J. Lütolf, C. Eichler, and A. Wallraff, Engineer-
N. Haider, D. J. Michalak, A. Bruno, and L. DiCarlo, ing cryogenic setups for 100-qubit scale superconducting
General method for extracting the quantum efficiency of circuit systems, EPJ Quantum Technology 6, 2 (2019).
dispersive qubit readout in circuit qed, Appl. Phys. Lett. [52] C. Macklin, K. O’Brien, D. Hover, M. E. Schwartz,
112, 092601 (2018). V. Bolkhovsky, X. Zhang, W. D. Oliver, and I. Siddiqi,
[41] J. E. Johnson, E. M. Hoskinson, C. Macklin, D. H. A near-quantum-limited Josephson traveling-wave para-
Slichter, I. Siddiqi, and J. Clarke, Dispersive readout of metric amplifier, Science 350, 307 (2015).
a flux qubit at the single-photon level, Phys. Rev. B 84, [53] N. Bartolo, F. Minganti, W. Casteels, and C. Ciuti,
220503 (2011). Exact steady state of a kerr resonator with one- and
[42] Z. K. Minev, S. O. Mundhada, S. Shankar, P. Rein- two-photon driving and dissipation: Controllable wigner-
hold, R. Gutiérrez-Jáuregui, R. J. Schoelkopf, M. Mir- function multimodality and dissipative phase transitions,
 14

Phys. Rev. A 94, 033841 (2016). [55] D. I. Schuster, A. Wallraff, A. Blais, L. Frunzio, R.-S.
[54] D. Roberts and A. A. Clerk, Driven-dissipative quantum Huang, J. Majer, S. M. Girvin, and R. J. Schoelkopf,
kerr resonators: New exact solutions, photon blockade ac Stark shift and dephasing of a superconducting qubit
and quantum bistability, Phys. Rev. X 10, 021022 (2020). strongly coupled to a cavity field, Phys. Rev. Lett. 94,
123602 (2005).