Frequency Up-Conversion Schemes for Controlling Superconducting Qubits

2025-04-22 0 0 1.64MB 9 页 10玖币
侵权投诉
Frequency Up-Conversion Schemes for Controlling Superconducting Qubits
Johannes Herrmann,
1, 2,
Christoph Hellings,
1
Stefania Lazar,
1
Fabian Pf¨affli,
2
Florian Haupt,
2
Tobias Thiele,
2
Dante Colao Zanuz,
1
Graham J. Norris,
1
Flavio Heer,
2
Christopher Eichler,
1
and Andreas Wallraff
1, 3
1
Department of Physics, ETH Zurich, CH-8093 Zurich, Switzerland
2
Zurich Instruments AG, CH-8005 Zurich, Switzerland
3
Quantum Center, ETH Zurich, CH-8093 Zurich, Switzerland
(Dated: October 7, 2022)
High-fidelity control of superconducting qubits requires the generation of microwave-frequency
pulses precisely tailored on nanosecond timescales. These pulses are most commonly synthesized
by up-converting and superimposing two narrow-band intermediate-frequency signals referred to
as the in-phase (I) and quadrature (Q) components. While the calibration of their DC-offsets,
relative amplitude and phase allows one to cancel unwanted sideband and carrier leakage, this IQ
mixing approach suffers from the presence of additional spurious frequency components. Here, we
experimentally study an alternative approach based on double frequency conversion, which overcomes
this challenge and circumvents the need for IQ-calibration. We find a spurious-free dynamic range of
more than 70 dB and compare the quality of pulse generation against a state-of-the-art IQ mixing
scheme by performing repeated single-qubit randomized benchmarking on a superconducting qubit.
I. INTRODUCTION
Superconducting qubits have emerged as one of the
most promising platforms for building error-corrected
quantum computers [
1
4
], which have the potential to
solve problems beyond the reach of classical computers [
5
].
Large-scale quantum computing crucially relies on clas-
sical electronics hardware to orchestrate the control and
measurement signals. Of particular importance are de-
vices for generating nanosecond-long microwave pulses,
which are used for control [
6
,
7
], measurement [
8
10
], and
entangling operations [
11
,
12
] of superconducting qubits
and which also play an important role for operating semi-
conductor spin qubits [13, 14] and trapped ions [15, 16].
Microwave pulse generation typically employs a radio
frequency mixer to up-convert pulses generated by an
arbitrary waveform generator (AWG) at an intermediate
megahertz frequency to the typical gigahertz transition
frequency range of the superconducting qubit [
17
]. Com-
pared to the direct digital synthesis (DDS) of control
pulses [
18
,
19
], frequency up-conversion allows for using
AWGs with lower sampling rate, which relaxes resource re-
quirements and thereby eases the scale-up to large channel
numbers [20].
Conventional frequency up-conversion schemes utilize
an IQ mixer, which up-converts an in-phase and a quadra-
ture component to destructively interfere unwanted side-
band and carrier leakages, which can ultimately limit the
achievable single-qubit gate fidelity. However, realistic IQ
mixers require extensive calibration of the IQ components
to achieve optimal performance, and the up-converted
output spectrum exhibits additional spurious frequency
components, which cannot be canceled by the interference
mechanism.
johannes.herrmann@phys.ethz.ch
Here, we control a superconducting qubit with an alter-
native microwave pulse generation scheme, which makes
use of two frequency conversion stages and uses analog
filters to remove the unwanted sideband and carrier leak-
ages from the up-converted control pulse. We compare
the signal quality of this double frequency conversion
scheme to a conventional IQ mixing scheme, and find the
output of the double-conversion stage to exhibit smaller
spurious frequency components and to be less affected by
variations in the ambient temperature. We also find a
slight improvement in the single-qubit gate fidelity when
using the double frequency conversion scheme to generate
the control pulses. For our study, we use a high-density
IQ converter (HDIQ) and a super-high-frequency signal
generator (SHFSG) to investigate the two schemes.
II. FREQUENCY UP-CONVERSION
High-fidelity single-qubit control is achieved with reso-
nant microwave pulses of the general form
Vd(t) = ˜vI(t) cos(ωt +φ) + ˜vQ(t) sin(ωt +φ),(1)
where
ω
is the transition frequency of the qubit,
φ
a
global phase, and
˜vI
(
t
) and
˜vQ
(
t
) are two independent
pulse envelope functions. A common choice for
˜vI
(
t
)
and
˜vQ
(
t
) is the Gaussian DRAG pulse parametrization,
which allows to avoid leakage into non-computational
states [6, 7, 21].
To generate the control pulse at frequency
ω
by up-
conversion, we multiply an intermediate frequency signal
generated by an AWG with a local oscillator continuously
running at frequency
ωLO
in the gigahertz range. The IF
signal consists of the two independent pulse envelope func-
tions
˜vI
(
t
) and
˜vQ
(
t
) defined in software and modulated
digitally at frequency
ωIF
, see Appendix A for details. We
multiply the IF signal with the LO using a radio frequency
mixer resulting in an output spectrum
S
(
ω
) featuring two
arXiv:2210.02513v1 [quant-ph] 5 Oct 2022
2
FIG. 1. Frequency up-conversion schemes. (a) Diagram and schematic output spectrum S(ω) of a radio frequency mixer
used for up-conversion, where both sidebands at frequencies
ωLO ±ωIF
are equally present. The frequency of the LO carrier is
indicated with a dashed arrow. (b) Sketch of an IQ mixer, for which the outputs of two parallel up-conversion paths, with an
LO signal shared through a 90
-hybrid splitter, are superimposed in a microwave combiner. The IF signals for the IQ mixer are
generated by two independent digital-to-analog converters (DAC). (c) Sketch of a double frequency conversion scheme with two
separate mixing stages fed by two different LO signals. The analog filtering after each stage is indicated by a dashed box in the
respective spectrum Si(f).
sidebands centered around the frequencies
ωLO ±ωIF
, see
Fig. 1(a). Throughout the paper, we use the convention
of driving the qubit with the upper sideband at frequency
ω
=
ωLO
+
ωIF
and consider the frequency component
ωLO ωIF
the undesired image of the control pulse, which,
when present in the spectrum of the qubit drive pulse,
induces a gate error.
III. IQ MIXING SCHEME
An established method to eliminate the image compo-
nent is IQ mixing [
17
,
22
], as shown in Fig. 1(b). Here, two
parallel up-conversion paths, for two IF signal waveforms
vI
(
t
) and
vQ
(
t
), are equipped with radio frequency (RF)
mixers which share a local oscillator signal through a
hybrid splitter which adds a 90
-phase shift to the sig-
nal driving the Q port mixer. The signals from the two
parallel mixing paths are superimposed in a microwave
combiner, where ideally one of the two up-converted side-
bands is canceled out by destructive interference, see
Appendix A. The spectra in Fig. 1(b) illustrate this mech-
anism for the case in which both
vI
(
t
) and
vQ
(
t
) have
a real-valued Fourier transform. In this particular case,
the destructive interference in the microwave combiner
is visible from the opposite signs of the lower-sideband
signals at frequency
ωLO ωIF
, see the spectra
SI
and
SQ
after the respective RF mixers in Fig. 1(b). To generate
the control pulse described by Eq.
(1)
with an IQ mixer,
we follow the procedure described in Appendix A and
compute in software two IF signals
vI
(
t
) and
vQ
(
t
), which,
when up-converted using an ideal IQ mixer, result in an
image-free spectrum.
Under realistic conditions, however, the implementa-
tion of an IQ mixer is subject to imperfections, such as
amplitude and phase imbalances between the two parallel
mixing paths [
17
,
23
]. These imperfections result in an
up-converted spectrum that exhibits unwanted sideband
and carrier leakage, which we compensate for by calibrat-
ing the relative phase, amplitude and DC offsets of the
IF signals, as described in the following.
First, the LO carrier signal leaks through each of the
two RF mixers with amplitude
LI
,
LQ
and phase
θI
,
θQ
,
respectively, such that the LO signal at the output of the
IQ mixer is [23]
yLO(t) = LIcos(ωLOt+θI) + LQsin(ωLOt+θQ).(2)
In many basic applications, this effect is not compensated
for, but an additional DC bias at the IQ mixer input
can help to decrease the amplitude of
yLO
(
t
). For this
purpose, we apply DC voltages
VI
and
VQ
to the IF-ports
of the IQ mixer, which create additional signals of the
form
VIcos
(
ωLOt
) +
VQsin
(
ωLOt
) at the output. These
signals interfere destructively with
yLO
(
t
) to result in a
total amplitude
YLO =|VI+LIeiθIiVQiLQeiθQ|(3)
of carrier leakage at the output of the mixer.
YLO
is
minimized for one specific combination of DC compensa-
tion voltages. To find this combination experimentally,
we measure the amplitude
YLO
with an FPGA-based ac-
quisition system, see Appendix B, while varying the DC
offsets
VI
and
VQ
with coarse resolution, see Fig. 2(a).
We fit Eq.
(3)
to the measured data to obtain optimal
compensation parameters, which are not limited to the
resolution between the discrete values of the underlying
measurement data and further enhance the calibration
accuracy, see Fig. 2(b).
Second, imbalances in the relative amplitude and phase
between the two parallel mixing paths modeled by the
scaling factor
˜α·
e
i˜
φ
can result in an imperfect destructive
interference of the image component at frequency
ωIM
=
摘要:

FrequencyUp-ConversionSchemesforControllingSuperconductingQubitsJohannesHerrmann,1,2,ChristophHellings,1StefaniaLazar,1FabianPfai,2FlorianHaupt,2TobiasThiele,2DanteColaoZanuz,1GrahamJ.Norris,1FlavioHeer,2ChristopherEichler,1andAndreasWallra 1,31DepartmentofPhysics,ETHZurich,CH-8093Zurich,Switzerl...

展开>> 收起<<
Frequency Up-Conversion Schemes for Controlling Superconducting Qubits.pdf

共9页,预览2页

还剩页未读, 继续阅读

声明:本站为文档C2C交易模式,即用户上传的文档直接被用户下载,本站只是中间服务平台,本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私,请立即通知玖贝云文库,我们立即给予删除!
分类:图书资源 价格:10玖币 属性:9 页 大小:1.64MB 格式:PDF 时间:2025-04-22

开通VIP享超值会员特权

  • 多端同步记录
  • 高速下载文档
  • 免费文档工具
  • 分享文档赚钱
  • 每日登录抽奖
  • 优质衍生服务
/ 9
客服
关注