Article

Coherent electrical control of a single

high-spin nucleus in silicon

https://doi.org/10.1038/s41586-020-2057-7 Serwan Asaad1,6, Vincent Mourik1,6, Benjamin Joecker1, Mark A. I. Johnson1,

Andrew D. Baczewski2, Hannes R. Firgau1, Mateusz T. Mądzik1, Vivien Schmitt1,
Received: 10 June 2019
Jarryd J. Pla3, Fay E. Hudson1, Kohei M. Itoh4, Jeffrey C. McCallum5, Andrew S. Dzurak1,
Accepted: 30 January 2020 Arne Laucht1 & Andrea Morello1 ✉

Published online: 11 March 2020

Check for updates Nuclear spins are highly coherent quantum objects. In large ensembles, their control
and detection via magnetic resonance is widely exploited, for example, in chemistry,
medicine, materials science and mining. Nuclear spins also featured in early proposals
for solid-state quantum computers1 and demonstrations of quantum search2 and
factoring3 algorithms. Scaling up such concepts requires controlling individual
nuclei, which can be detected when coupled to an electron4–6. However, the need to
address the nuclei via oscillating magnetic fields complicates their integration in
multi-spin nanoscale devices, because the field cannot be localized or screened.
Control via electric fields would resolve this problem, but previous methods7–9 relied
on transducing electric signals into magnetic fields via the electron–nuclear hyperfine
interaction, which severely affects nuclear coherence. Here we demonstrate the
coherent quantum control of a single 123Sb (spin-7/2) nucleus using localized electric
fields produced within a silicon nanoelectronic device. The method exploits an idea
proposed in 196110 but not previously realized experimentally with a single nucleus.
Our results are quantitatively supported by a microscopic theoretical model that
reveals how the purely electrical modulation of the nuclear electric quadrupole
interaction results in coherent nuclear spin transitions that are uniquely addressable
owing to lattice strain. The spin dephasing time, 0.1 seconds, is orders of magnitude
longer than those obtained by methods that require a coupled electron spin to
achieve electrical driving. These results show that high-spin quadrupolar nuclei could
be deployed as chaotic models, strain sensors and hybrid spin-mechanical quantum
systems using all-electrical controls. Integrating electrically controllable nuclei with
quantum dots11,12 could pave the way to scalable, nuclear- and electron-spin-based
quantum computers in silicon that operate without the need for oscillating magnetic
fields.

Nuclear magnetic resonance (NMR) relies on the presence of a static A theoretical idea crucial to this strategy was proposed by Bloember-
magnetic field, B0, that separates the energy levels of the nuclear gen as early as 196110: for nuclei with spin I > 1/2 and non-zero electric
spins, and a radio-frequency (RF) oscillating magnetic field, B1, that quadrupole moment qn, a resonant electric field induces nuclear spin
induces transitions between such levels. Magnetic fields cannot be transitions by modulating the nuclear quadrupole interaction, if the
easily confined or screened at the nanoscale. Therefore, identical nuclei are placed in solids that lack point-inversion symmetry at the
nuclear spins within large regions would all respond to the same lattice site. In bulk ensembles, the static shift of the NMR frequency by
signal, preventing the spins from being individually addressed. Elec- a d.c. electric field, named linear quadrupole Stark effect (LQSE), was
tric fields, instead, can be efficiently routed and confined within observed in the 1960s13. The resonant version of LQSE, called nuclear
highly complex nanoscale devices, with a prime example being the electric resonance (NER) was demonstrated only recently14 in a bulk
sophisticated interconnects found in modern silicon computer chips. gallium arsenide (GaAs) crystal.
These observations suggest that an ideal route to scale up nuclear- We report here the demonstration of NER and coherent electrical
spin-based quantum devices would involve the use of RF electric control of a single antimony (123Sb) nucleus in silicon (Si). The discovery
fields for spin control. that this nucleus could be electrically controlled was fortuitous. The 123Sb
1
Centre for Quantum Computation and Communication Technology, School of Electrical Engineering and Telecommunications, UNSW Sydney, Sydney, New South Wales, Australia. 2Center for
Computing Research, Sandia National Laboratories, Albuquerque, NM, USA. 3School of Electrical Engineering and Telecommunications, UNSW Sydney, Sydney, New South Wales, Australia.
School of Fundamental Science and Technology, Keio University, Yokohama, Japan. 5Centre for Quantum Computation and Communication Technology, School of Physics, University of
4

Melbourne, Melbourne, Victoria, Australia. 6These authors contributed equally: Serwan Asaad, Vincent Mourik. ✉e-mail: [email protected]

Nature | Vol 579 | 12 March 2020 | 205

Article
a b Zeeman Quadrupole c
ISET JnB0 + 500 kHz
J n B 0 Iz QxxIx2 f5/2 ↔ 7/2
I/V
f3/2 ↔ 5/2
D

frequency
|−7/2〉 f1/2 ↔ 3/2

Transition
Do f–1/2 ↔ 1/2
no JnB0
rg f–3/2 ↔ −1/2
at |−5/2〉
es f−7/2 ↔ −5/2 f–5/2 ↔ −3/2
Energy JnB0 − 500 kHz f–7/2 ↔ −5/2
SE |−3/2〉 f−5/2 ↔ −3/2
T S 0 66 100
VSD f−3/2 ↔ −1/2 Qxx (kHz)
Mic |−1/2〉
r Donor gates d
ant owav
enn e f−1/2 ↔ 1/2
a |1/2〉 Shear strain
SET
B0 Gap Hxx – Hzz
f1/2 ↔ 3/2
V gate
RF |3/2〉 SiO2
0.2%
+ –
200 nm –5 123Sb
– +
|5/2〉 f3/2 ↔ 5/2 –10

y (nm)
0%
antenna –15
VRF f3/2 ↔ 7/2 –20
|7/2〉
f5/2 ↔ 7/2 –25 –0.2%
–50 0 50
z (nm)

Fig. 1 | 123Sb nuclear spin in a silicon device. a, False-colour scanning electron transition frequencies as a function of Qxx. A non-zero Qxx  results in seven
micrograph of the silicon metal–oxide–semiconductor device used in the individually addressable nuclear resonances. The m I = −1/2 ↔ +1/2 transition
experiment. Note the gaps in the nominally short-circuited antenna (dashed blue line in c) is forbidden in NER. The measured quadrupole splitting
terminations. S, source; D, drain; SET, single-electron transistor. b, Energy-level fQ = 66 kHz is indicated by a dashed purple line. d, Shear strain in the silicon
diagram of the spin-7/2 nucleus of an ionized 123Sb donor. The magnetic field B0 substrate, calculated on a vertical cross-section under the orange dashed line
introduces a Zeeman splitting (green dashes), and the electric quadrupole in a.
interaction Qxx causes a further energy shift (blue dashes). c, Nuclear spin

donor atom has a nuclear spin of I = 7/2 with electric quadrupole moment constant and e is the electron charge), resulting in a quadrupole split-
qn = −0.69 b. Depending on its electrochemical potential relative to a ting  fQ of the nuclear resonance frequencies (Fig. 1d), making all transi-
nearby electron reservoir, an electron (with spin S = 1/2) may be bound to tions individually addressable.
the nucleus. The atom was implanted in a metal–oxide–semiconductor The application of an RF electric field of amplitude E1 modulates the
nanostructure (Fig. 1a) fabricated on isotopically enriched 28Si (for fabri- nuclear quadrupole energies by δQxz and δQyz, and induces transitions
cation details, see Supplementary Information section 3), similar to those between nuclear states at a rate of f Rabi,NER mI −1↔mI
∝ |δQxz mI − 1 IˆxIˆz + IˆzIˆx mI |,
developed for phosphorus (31P) spin qubits5,15,16. The structure contains where mI is the secondary spin quantum number, ranging from −I to I
a single-electron transistor (SET) for single-shot electron spin readout, in steps of 1, and Iˆx, Iˆy, Iˆz are the eight-dimensional operators describing
which is based on energy-selective electron tunnelling into a cold charge the x, y, z projections of the I = 7/2 spin. Notably, the transition rate is
reservoir17. Four electrostatic gates control the electrochemical potential predicted to be zero for the mI = −1/2 ↔ +1/2 transition (see equation
of the donor, and a broadband on-chip microwave antenna18 delivers (15) in Supplementary Information section 2C), a consequence of the
coherent control signals to the donor spins. Single-shot, quantum non- selection rules of electric quadrupole transitions. Because the quad-
demolition nuclear spin readout5 is obtained by combining single-shot rupole interaction is quadratic in the nuclear spin operators, first-order
electron readout with selective excitation at a specific electron spin transitions between spin states that differ by ΔmI = ±2 are allowed.
2
resonance frequency, which depends on the nuclear state because of the These occur at a rate of f Rabi,NER m I −2↔mI
∝ |δQxx mI − 2 Iˆx mI | (see equation
strong hyperfine interaction (see Methods). The antenna is nominally (19) in Supplementary Information section 2C) and, importantly, all
terminated by a short circuit, in order to obtain maximum current at its ΔmI = ±2 transitions have a non-zero rate.
tip and produce strong oscillating magnetic fields to control both the Figure 2a shows the experimental NER spectrum for ΔmI = ±1 transi-
electron (at about 40 GHz) and the nuclear (at about 10 MHz) spins of tions, which contains six sharp resonances separated by fQ = 66 kHz.
the donor. In this device, however, an electrostatic discharge damaged The mI = −1/2 ↔ +1/2 transition is absent, as expected from NER. All
the short-circuit termination (Fig. 1a). Although the small gap in the six predicted ΔmI = ±2 transitions are observed (Fig. 2b). The ability
termination had a low enough impedance at 40 GHz, to allow current to excite the mI = −1/2 ↔ +3/2 transition was used to ‘jump over’ the
flow for electron spin resonance, at about 10 MHz it produced solely forbidden mI = −1/2 ↔ +1/2 transition and observe the ΔmI = ±1 transi-
an RF electric field. Once we realized that NER was possible, we began tions at negative mI, which would otherwise be inaccessible if starting
to use the electric gates fabricated exactly above the donor, which had from a positive mI. Similarly, the NER spectrum for ΔmI = ±2 transitions
an even stronger effect. (Fig. 2b) could be completed only by employing a ΔmI = ±1 transition.
We focus here on the 123Sb donor in its ionized state; the removal of Figure 2c, d presents the observed transition rates between each
the donor-bound electron precludes any interpretation of the data pair of states, in excellent agreement with the predicted trends from
involving modulation of hyperfine fields7,9. The electron is introduced NER theory. Using NMR, the Rabi frequencies for the ΔmI = ±1 transi-
only for the final readout phase. tions would be f Rabi,NMR
mI −1↔mI
∝ |γnB1⟨mI − 1|I^x|mI ⟩| (γn = 5.55 MHz T−1 is the
In nanoscale Si devices, the aluminium (Al) gates can cause consid- nuclear gyromagnetic ratio), which is notably maximal for the
erable lattice strain at low temperatures, owing to the different ther- mI = −1/2 ↔ +1/2 transition. The ΔmI = ±2 NMR transitions are forbidden
mal contraction of Al and Si (ref. 18). Lattice strain creates an electric to first-order. These results prove decisively that our experiments do
field gradient (EFG) of Vαβ = ∂2V/∂α∂β (V is the electric potential and not constitute a form of magnetic resonance.
α, β ∈ {x, y, z}) at the nuclear site20,21 (Fig. 1b), which produces a static As observed in earlier experiments on 31P (refs. 16,22), the nuclear spins
nuclear quadrupole interaction Qαβ = eqnVαβ/[2I(2I − 1)h] (h is the Planck of ionized donors in 28Si have exceptional quantum coherence

206 | Nature | Vol 579 | 12 March 2020

 a ΔmI = ±1 b ΔmI = ±2 Fig. 2 | Nuclear electric resonance.
1 1 a, b, NER spectrum for the Δm I = ±1 (a) and
Δm I = ±2 (b) transitions, obtained by
gate
applying voltage V RF to a donor gate (see
Fig. 1a). The m I = −1/2 ↔ +1/2 transition
Pflip

Pflip
(a, red) was not observed, as expected in
NER. To acquire the complete Δm I = ±1
spectrum, the m I = −1/2 ↔ +3/2 transition
0 0 was used to bridge the positive and
8.0 8.1 8.2 8.3 8.4 8.5 16.2 16.4 16.6 16.8
Electric drive frequency (MHz) Electric drive frequency (MHz) negative m I values. P flip represents the
c d probability of flipping the nuclear spin
200
between two states. c, d, Rabi frequencies
1,000
of the Δm I = ±1 (c) and Δm I = ±2 (d)
150 transitions, each measured at a constant
800
NER drive amplitude (see Extended Data
fRabi (Hz)

fRabi (Hz)
600 100 Fig. 2 for the corresponding Rabi
400 oscillations). Measured values (circles in
Data
50 c, squares in d) are compared to the
200 NMR Data gate
gate theoretical predictions for NER (stars)
NER V RF = 20 mV V RF = 40 mV
NER
0 0 and NMR (triangles in c), using the drive
|−7/2〉 |−5/2〉 |−3/2〉 |−1/2〉 |1/2〉 |3/2〉 |5/2〉 |−7/2〉 |−5/2〉 |−3/2〉 |−1/2〉 |1/2〉 |3/2〉 amplitude as the single free-scaling
|−5/2〉 |−3/2〉 |−1/2〉 |1/2〉 |3/2〉 |5/2〉 |7/2〉 |−3/2〉 |−1/2〉 |1/2〉 |3/2〉 |5/2〉 |7/2〉
parameter to match the experimental
Transition Transition values. All Rabi frequencies closely follow
e |5/2〉 ↔ |7/2〉 f |3/2〉 ↔ |7/2〉 the NER prediction, including the absence
1 1 of the m I = −1/2 ↔ +1/2 transition (red
circle in c, red dots in a), and are
incompatible with NMR. e, f, Nuclear Rabi
Pflip

Pflip

oscillations on the m I = +5/2 ↔ +7/2 (e) and

m I = +3/2 ↔ +7/2 (f) transitions. A sinusoid
with no decay is used to fit the data.  tNER,
0 0
0 5 10 15 0 20 40 NER pulse duration. g, h, Nuclear Ramsey
tNER (ms) tNER (ms) fringes used to extract the pure
g h
1 1 dephasing time T 2n+ * on the m I = +5/2 ↔
T2* = 92(8) ms T2* = 28(1) ms +7/2 (g) and m I = +3/2 ↔ +7/2 (h)
transitions. The fits are sinusoids with
Pflip

Pflip

2
envelopes decaying as exp [− (τ /T *2n+) ],
where τ is the free precession time. Error
bars and uncertainties denote the 68%
0 0
0 25 50 75 100 0 20 40 60 confidence level.
W(ms) W(ms)

properties. We performed a Ramsey experiment (Fig. 2g) on the additional d.c. voltage ΔV DC gate
on the gates (Fig. 3c, d). All NER frequen-
mI = +5/2 ↔ +7/2 (ΔmI = ±1) transition to extract the pure dephasing gate gate
cies shifted according to Δf = (∂fQ /∂V DC ) ΔmI [mI − (ΔmI /2)]ΔV DC ,
time T *2n+(+5/2 ↔ + 7/2) = 92(8) ms (68% confidence level), which cor- gate −1
where ∂fQ /∂V DC = 9.9(3) Hz mV , and ΔmI [mI − (ΔmI /2)] is a factor
responds to an NER broadening (full-width at half-maximum) of of order unity that represents the matrix element of the electric quad-
Γn = ln2/(πT *2n+) = 2.4(2) Hz . The mI = +3/2 ↔ +7/2 (ΔmI = ±2) transition rupole interaction between the initial and final state of each transition
has shorter dephasing time, T *2n+(+3/2 ↔ + 7/2) = 28(1) ms  (Fig. 2h). (see Supplementary Information section 2C for details).
Both values, although extremely long in absolute terms, are noticeably The results reported here constitute the first, to our knowledge,
shorter than the timeT *2n+ = 250–600 ms measured on the 31P nucleus in observation of coherent, purely electrical control of a single nuclear
two other similar devices16 fabricated on the same 28Si wafer. Given that spin. Achieving this in silicon is, at first sight, remarkable: no effect
the 31P nucleus has zero quadrupole moment, this suggests that the 123Sb of electric fields on nuclear spins has ever been observed in a non-polar,
coherence may be affected by electrical noise23, in a way that the 31P coher- non-piezoelectric material in the absence of a hyperfine-coupled electron.
ence is not. Nonetheless, our dephasing time remains two orders of mag- To gain a microscopic understanding of this phenomenon, we conjec-
nitude longer than that observed in 31P when adding a hyperfine coupled tured that our results are a form of LQSE13. Resonant transitions between
electron, T ∗2n0 ≈ 430–570 μs (ref. 16) and three orders of magnitude nuclear levels induced by electric fields (NER) require that the crystal does
longer than the observedT ∗2 = 64 μsof a terbium nucleus in a single-atom not possess point-inversion symmetry at the atomic site10, as is indeed the
magnet7. This observation highlights the benefit of a purely electrical case for silicon. The observation of individual NER transitions, separated
control mechanism that does not rely on hyperfine interactions. by the nuclear quadrupole splitting fQ, implies that a static EFG must exist
We measured the Rabi frequencies of the ΔmI = ±1 and ΔmI = ±2 NER at the nuclear site. This requires breaking the Td (tetrahedral) symmetry
transitions as a function of the amplitude of the RF voltage applied to of the silicon crystal, as it would otherwise have zero net EFG. For instance,
the gate, finding transition rates gE,1 = 34.21(3) Hz mV−1 (Fig. 3a) and uniaxial strain (for example, ϵzz) lowers the symmetry to D2d (tetrago-
gE,2 = 1.995(4) Hz mV−1 (Fig. 3b). These transition rates show that NER nal scalenohedral), whereas shear strain (for example, ϵxx–ϵzz) lowers it
is a weak effect but, owing to the long nuclear spin coherence in 28Si, to C2v (rhombic pyramidal). The Td symmetry can also be broken by an
we were able to perform high-fidelity Rabi flops persisting for tens of electric field that polarizes the atomic bonds. This latter effect explains
milliseconds (Fig. 2e, f). both the observation of NER and the static shift of the nuclear spin
In addition to driving nuclear spin transitions with an RF voltage, we resonance lines (Fig. 3c, d) due to LQSE on application of a static gate
were able to apply Stark shifts to the resonance frequencies using an voltage.

Nature | Vol 579 | 12 March 2020 | 207

 Article
a |5/2〉 ↔ |7/2〉 b |3/2〉 ↔ |7/2〉 Fig. 3 | Linear quadrupole Stark effect.
2,000 a, b, Rabi frequencies f Rabi for varying
200 electric drive peak amplitude V RF gate,
1,500 measured on the Δm I = ±1 transition
5/2 ↔ 7/2 (a) and the Δm I = ±2 transition
fRabi (Hz)

fRabi (Hz)
1,000 3/2 ↔ 7/2  (b). The linear relationship
100 gate
between V RF and f Rabi is consistent with
500 gate gate
a first-order transition induced by
fRabi /VRF = 34.21(3) Hz mV–1 fRabi /VRF = 1.995(4) Hz mV–1 the LQSE. c, d, Quadrupole shift
0
0 10 20 30 40 50
0
0 20 40 60 80 100
(
ΔfQ = ∂fQ /∂V DC )
gate gate
ΔV DC measured while
applying an additional d.c. voltage
gate gate
V RF (mV) V RF (mV) gate
ΔV DC on a donor gate. The application
gate
c d of ΔV DC causes each transition
ΔmI = ±1 ΔmI = ±2
0 0 frequency fm −Δm ↔ m to shift by
( )
I I I
|5/2〉 ↔ |7/2〉 |3/2〉 ↔ |7/2〉 Δf = ∂fQ /∂V DC gate gate
ΔmI [mI − (ΔmI /2)]ΔV DC  
|3/2〉 ↔ |5/2〉 |1/2〉 ↔ |5/2〉
–100 |1/2〉 ↔ |3/2〉 –100 |−1/2〉 ↔ |3/2〉 (inset; see Extended Data Fig. 3 for nuclear
|−3/2〉 ↔ |–1/2〉 |−3/2〉 ↔ |1/2〉 spectra). A combined fit through all
|−5/2〉 ↔ |–3/2〉 |−5/2〉 ↔ |–1/2〉 Δm I = ±1 (c) and Δm I = ±2 (d) frequency
–200 |−7/2〉 ↔ |–5/2〉 –200 |−7/2〉 ↔ |–3/2〉
Fit Fit
shifts results in an LQSE coefficient of
gate
= 9.9(3) Hz mV −1 .
ΔfQ (Hz)

ΔfQ (Hz)
∂fQ /∂V DC
–300 –300 2
1
Δf (kHz)

Δf (kHz)
-400 0 –400 0

–1
–2
-500 –40 –20 0 –500 –40 –20 0
gate gate
ΔV DC (mV) ΔV DC (mV)
-600 –600
–50 –40 –30 –20 –10 0 –50 –40 –30 –20 –10 0
gate gate
ΔV DC (mV) ΔV DC (mV)

The larger and charged donor atom introduces a local lattice distor- have observed the manifestation of LQSE and NER in a single nuclear
tion, displacing its four coordinating Si atoms by 0.2 Å, and polarizes spin in silicon.
the charge density along the bonds (Fig. 4b, d). This, however, does not Our results have substantial consequences for the development of
yet break the Td symmetry. An EFG is obtained by further introducing nuclear-spin-based quantum computers and the design of nanoscale
strain. The S tensor that links EFG to strain has two unique components, quantum devices. The Hilbert space of the I = 7/2 123Sb nucleus has eight
S11 (uniaxial) and S44 (shear). We conducted a first-principles, density dimensions. It can encode the equivalent of three quantum bits of
functional theory calculation and extracted S11 = 2.4 × 1022 V m−2 and information, allowing simple quantum algorithms25 or quantum error
S44 = 6.1 × 1022 V m−2 (see Supplementary Information section 7C2 for correction codes26, all using solely electric fields. The donor electron
details). Using a finite-element numerical model we computed the and nuclear spins combined form a ‘flip-flop’ qubit11, controllable by
strain profile in our device, as caused by the different thermal expan- electric-dipole spin resonance. This scheme normally requires a mag-
sions of Si and Al on cooling to cryogenic temperatures19,21 (Fig. 1b). netic antenna to reset the nuclear state in the appropriate qubit sub-
Finally, we triangulated the most likely location of the 123Sb donor by space. This need could be removed completely by using an electrically
combining the implantation depth profile with a model of the relative drivable high-spin nucleus such as 123Sb. A recent result showed that
capacitive coupling between the donor and different pairs of control lithographic quantum dots in silicon can be entangled with nuclear
gates, extracted from the experimental charge stability diagrams (see spins and that the nuclear coherence can be preserved while shuttling
Extended Data Fig. 4 and Supplementary Information section 7C3 for the electron between different dots12. Electron spin qubits in silicon can
details). By combining these three pieces of information, we arrived be coherently controlled by electric fields with high speed and high
at a spatial map of quadrupole splittings fQ (Fig. 4c), which shows good fidelity27. Adding the ability to electrically control quadrupolar nuclei
agreement between the models and the experiment around the pre- paves the way to quantum computer architectures that integrate fast
dicted location of the donor under study. electron spin qubits with long-lived nuclear quantum memories while
The effect of electric fields on the quadrupole interaction, both fully exploiting the controllability and scalability of silicon metal–
static (LQSE) and dynamic (NER), can be understood as arising from oxide–semiconductor devices, without the complication of routing
the single unique component of the R tensor, R14 (see Supplementary RF magnetic fields within the device.
Information section 7D1 for details). By combining a finite-element The experimental validation of a microscopic model of the relation
model of the electric field in the device, the estimated 123Sb+ donor between strain and quadrupole splitting, obtained in a functional
position and the experimental values of LQSE and NER Rabi frequen- silicon electronic device, suggests the use of quadrupolar nuclei as
cies, we extracted R14 = 1.7 × 1012 m−1 (see Supplementary Information single-atom probes of local strain, which has a key role in enhancing
section 7D2 for details). The strength of this coupling is comparable the performance of ultra-scaled transistors28.
to prior bulk measurements of LQSE on arsenic (75As) in GaAs (ref. 24). The NER methods and microscopic models presented here could
This can be understood by observing that, although the Sb+–Si bond be extended to the study of quadrupolar nuclei in materials such as
has a weaker ionic character than the Ga–As bond, R14 scales with atomic diamond and silicon carbide, where electrical and strain tuning of opti-
number, leading to a similar overall value. Given that our model agrees cally addressable electronic spins has been demonstrated29,30.
with the experiment within a factor of order unity and no alternative The observation of a large quadrupole splitting of fQ = 66 kHz in a
explanation comes within orders of magnitude of the results (see Sup- high-spin nucleus creates a platform in which to study quantum chaotic
plementary Information section 7E for details), we conclude that we dynamics in a single particle31. This has further applications in quantum

208 | Nature | Vol 579 | 12 March 2020

 a toolbox of hybrid quantum systems for quantum information process-
Electron ing and precision sensing34.

28Si
Online content
Any methods, additional references, Nature Research reporting sum-
123Sb+
[001] maries, source data, extended data, supplementary information,
acknowledgements, peer review information; details of author con-
tributions and competing interests; and statements of data and code
[010] availability are available at https://doi.org/10.1038/s41586-020-2057-7.
[100] B0 1. Kane, B. E. A silicon-based nuclear spin quantum computer. Nature 393, 133–137 (1998).
2. Jones, J. A., Mosca, M. & Hansen, R. H. Implementation of quantum search algorithm on a
quantum computer. Nature 393, 344–346 (1998).
3. Vandersypen, L. M. K. et al. Experimental realization of Shor’s factoring algorithm using
b c Al nuclear magnetic resonance. Nature 414, 883–887 (2001).
Shear strain strain
SiO2 f Q (kHz) 4. Jelezko, F. et al. Observation of coherent oscillation of a single nuclear spin and
0 realization of a two-qubit conditional quantum gate. Phys. Rev. Lett. 93, 130501 (2004).
150 5. Pla, J. J. et al. High-fidelity readout and control of a nuclear spin qubit in silicon. Nature
–10 496, 334–338 (2013).
y (nm)

100 6. Willke, P. et al. Hyperfine interaction of individual atoms on a surface. Science 362,
–20
336–339 (2018).
–30 50 7. Thiele, S. et al. Electrically driven nuclear spin resonance in single-molecule magnets.
Science 344, 1135–1138 (2014).
–40 0 8. Laucht, A. et al. Electrically controlling single-spin qubits in a continuous microwave
–40 –20 0 20 40
field. Sci. Adv. 1, e1500022 (2015).
z (nm)
9. Sigillito, A. J., Tyryshkin, A. M., Schenkel, T., Houck, A. A. & Lyon, S. A. All-electric control
d e of donor nuclear spin qubits in silicon. Nat. Nanotechnol. 12, 958–962 (2017).
Electric field |5/2〉 ↔ |7/2〉 10. Bloembergen, N. Linear Stark effect in magnetic resonance spectra. Science 133,
1,500 |3/2〉 ↔ |7/2〉 1363–1364 (1961).
Model 11. Tosi, G. et al. Silicon quantum processor with robust long-distance qubit couplings.
fRabi (Hz)

1,000 Nat. Commun. 8, 450 (2017).

12. Hensen, B. et al. A silicon quantum-dot-coupled nuclear spin qubit. Nat. Nanotechnol. 15,
E 13–17 (2020).
500 13. Dixon, R. & Bloembergen, N. Electrically induced perturbations of halogen nuclear
quadrupole interactions in polycrystalline compounds. ii. Microscopic theory. J. Chem.
0 Phys. 41, 1739–1747 (1964).
0 20 40 60 80 100 14. Ono, M., Ishihara, J., Sato, G., Ohno, Y. & Ohno, H. Coherent manipulation of nuclear spins
gate
V RF (mV) in semiconductors with an electric field. Appl. Phys. Express 6, 033002 (2013).
15. Pla, J. J. et al. A single-atom electron spin qubit in silicon. Nature 489, 541–545 (2012).
16. Muhonen, J. T. et al. Storing quantum information for 30 seconds in a nanoelectronic
Fig. 4 | Microscopic origins of the quadrupole interaction. a, Valence charge
device. Nat. Nanotechnol. 9, 986–991 (2014).
density near the Sb+ atom (gold) and its 16 closest Si atoms (black) with a charge 17. Morello, A. et al. Single-shot readout of an electron spin in silicon. Nature 467, 687–691
density isosurface (red). The positive charge of the donor causes an (2010).
asymmetric charge density along the Sb+–Si bond but, in the absence of strain 18. Dehollain, J. et al. Nanoscale broadband transmission lines for spin qubit control.
Nanotechnology 24, 015202 (2013).
or external electric fields, the EFG at the 123Sb site vanishes by symmetry.
19. Thorbeck, T. & Zimmerman, N. M. Formation of strain-induced quantum dots in gated
b, Shear strain displaces the Si atoms and covalent bonds neighbouring the semiconductor nanostructures. AIP Adv. 5, 087107 (2015).
123
Sb nucleus, creating an EFG that results in a quadrupole shift. c, Quadrupole 20. Franke, D. P. et al. Interaction of strain and nuclear spins in silicon: quadrupolar effects on
splitting fQ, predicted by combining density functional theory calculations and ionized donors. Phys. Rev. Lett. 115, 057601 (2015).
21. Pla, J. J. et al. Strain-induced spin-resonance shifts in silicon devices. Phys. Rev. Appl. 9,
finite-element simulations (see Supplementary Information section 7C for
044014 (2018).
details). Black contours enclose the 68% and 95% confidence regions for the 22. Saeedi, K. et al. Room-temperature quantum bit storage exceeding 39 minutes using
location of the donor, as obtained from capacitance triangulation and the ionized donors in silicon-28. Science 342, 830–833 (2013).
donor implantation profile (see Supplementary Information section 7A for 23. Franke, D. P., Pflüger, M. P. D., Itoh, K. M. & Brandt, M. S. Multiple-quantum transitions and
charge-induced decoherence of donor nuclear spins in silicon. Phys. Rev. Lett. 118,
details). d, Electric fields applied via the gate voltage distort the charge
246401 (2017).
distribution, resulting in both linear frequency shifts (LQSE) and coherent spin 24. Gill, D. & Bloembergen, N. Linear Stark splitting of nuclear spin levels in GaAs. Phys. Rev.
transitions (NER). e, Calculation of the NER Rabi frequencies caused by 129, 2398–2403 (1963).
electrical EFG modulation (green lines), compared to experimental results for a 25. Godfrin, C. et al. Operating quantum states in single magnetic molecules:
implementation of Grover’s quantum algorithm. Phys. Rev. Lett. 119, 187702 (2017).
Δm I = ±1 (dots) and a Δm I = ±2 (squares) transition. All f Rabi values are determined
26. Waldherr, G. et al. Quantum error correction in a solid-state hybrid spin register. Nature
using a single parameter, R 14, calculated via finite-element modelling and 506, 204–207 (2014).
electronic structure theory. No free fitting parameters were used. 27. Yoneda, J. et al. A quantum-dot spin qubit with coherence limited by charge noise and
fidelity higher than 99.9%. Nat. Nanotechnol. 13, 102–106 (2018).
28. Thompson, S. E., Sun, G., Choi, Y. S. & Nishida, T. Uniaxial-process-induced strained-Si:
extending the CMOS roadmap. IEEE Trans. Electron Dev. 53, 1010–1020 (2006).
29. Dolde, F. et al. Electric-field sensing using single diamond spins. Nat. Phys. 7, 459–463 (2011).
information science, for example, because of the remarkable analo- 30. Falk, A. L. et al. Electrically and mechanically tunable electron spins in silicon carbide
gies between chaotic spin models and digital quantum simulations32. color centers. Phys. Rev. Lett. 112, 187601 (2014).
31. Mourik, V. et al. Exploring quantum chaos with a single nuclear spin. Phys. Rev. E 98,
Although the strain in the present device is static, our work allows 042206 (2018).
us to predict the nuclear Rabi frequencies that would arise from time- 32. Sieberer, L. M. et al. Digital quantum simulation, trotter errors, and quantum chaos of the
dependent strain (see Supplementary Information section 8 for details). kicked top. npj Quantum Inf. 5, 78 (2019).
33. Ghaffari, S. et al. Quantum limit of quality factor in silicon micro and nano mechanical
A dynamical strain of about 5 × 10−8 would cause a Rabi frequency of resonators. Sci. Rep. 3, 3244 (2013); corrigendum 4, 4331 (2013).
10 Hz, comparable to both the inhomogeneous nuclear linewidth 34. Kurizki, G. et al. Quantum technologies with hybrid systems. Proc. Natl Acad. Sci. USA 112,
3866–3873 (2015).
Γn ≈ 2.4 Hz and to the linewidth Γn of high-quality silicon mechanical
resonators in the megahertz range33. Therefore, it is conceivable that Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
the strong-coupling limit of cavity quantum electrodynamics might be published maps and institutional affiliations.

achieved between a single nuclear spin and a macroscopic mechanical © This is a U.S. government work and not under copyright protection in the U.S.; foreign
oscillator, adding a novel spin–mechanical coupling pathway to the copyright protection may apply 2020

Nature | Vol 579 | 12 March 2020 | 209

Article
Methods
H^
Fabrication = γnB0I^z + ∑ Q αβ I^αI^β (1)
h α , β ∈{x , y , z }
The device was fabricated on a  100 p-type silicon wafer, with a
900-nm-thick epitaxial layer of isotopically purified 28Si on top (con-
centration of residual 29Si, 730 ppm). Metallic leads for the SET were where h = 6.626 × 10−34 J Hz−1 is the Planck constant, γn = −5.553 MHz T−1
formed using optical lithography and phosphorus diffusion. The sub- is the nuclear gyromagnetic ratio and B0 = 1.496 T. In the presence of
strate was subsequently covered with a 200-nm-thick field oxide, with an RF electric field of amplitude E1, the ΔmI = ±1 transitions are driven
a small central window (10 × 20 μm2) containing a high-quality, ther- by an additional Hamiltonian term of the form:
mally grown layer of SiO2 with a thickness of 8 nm. Using a combination
NER
of standard optical and electron-beam lithography techniques, the H^ mI −1↔mI (t)/h = cos(2πft)[δQxz (I^xI^z + I^zI^x) + δQ yz (I^yI^z + I^zI^y)] (2)
device was fabricated on this thin oxide window. First, a small
(90 × 100 nm2) window was defined, through which 123Sb ions were The ΔmI = ±2 transitions are driven by a term of the form:
implanted at an energy of 8 keV and a fluence of 2 × 1011 cm−2, corre-
NER 2 2
sponding to an average of 14 donors in the implantation window. H^ mI −2↔mI (t)/h = cos(2πft)[δQxx I^ x + δQ yy I^ y + δQxy (I^xI^y + I^yI^x)] (3)
Donors were activated using a rapid thermal anneal at 1,000 °C for 5 s.
Next, in two electron-beam lithography steps, the gates forming the A detailed derivation of the matrix elements responsible for driving
SET, the donor gates and the microwave antenna were created using the ΔmI = ±1 and ΔmI = ±2 NER transitions is given in Supplementary
thermally evaporated aluminium and lift-off, with native aluminium Information section 2C. A finite-element model is used to compute
oxide as the gate dielectric. Ohmic contacts to the n-doped SET leads the strain and electric fields in the silicon layer near the donor posi-
were formed using optical lithography, evaporated Al and lift-off, fol- tion using the COMSOL multiphysics software. The donor position is
lowed by a forming gas anneal. A detailed step-by-step process flow is triangulated by comparing simulated gate-to-donor coupling strengths
given in Supplementary Information section 3. with the experimentally observed strength, combined with the donor
implantation profile (see Extended Data Fig. 4 and Supplementary
Experimental setup Information section 7A). Kohn–Sham density functional theory is
The sample was cooled to a temperature of 20 mK in a dilution refrig- employed to calculate the components of the S tensor that describe
erator (Bluefors BF-LD400) fitted with a superconducting magnet. the impact of strain on the EFG. To this end, 64- and 512-atom super-
During the measurements, arbitrary waveform generators (Signadyne cells were strained using the PAW (projector augmented-wave) for-
M3201A and M3300A) were used to tune the donor electrochemical malism35 with a plane-wave basis, as implemented in VASP (Vienna ab
potential, generate NER pulses and IQ-modulate the microwave sig- initio simulation package)36–38. The electric-field response tensor is
nals generated by a vector microwave source (Keysight E8267D). The estimated by comparing the data points from the d.c. LQSE (Fig. 3c,
SET current was amplified with a transimpedance amplifier (FEMTO d) and ΔmI = ±1 (Fig. 2c) and ΔmI = ±2 (Fig. 2d) Rabi frequencies with
DLPCA-200 in combination with Stanford Instruments SIM911) and the simulated electric fields at the triangulated donor position. The
subsequently measured with a digitizer (Signadyne M3300A). Full final R14 is found by minimizing the normalized residuals of the three
details of the experimental setup, including a wiring schematic, can separate R14 estimates. Full theoretical modelling details can be found
be found in Supplementary Information section 4. in Supplementary Information section 7.

Nuclear spin readout

The nuclear spin state is measured via electron spin readout. For Data availability
nuclear spin readout, an electron is introduced to the donor by tun- All data necessary to evaluate the claims of this paper are provided in
ing its electrochemical potential about the Fermi level of the SET such the main manuscript and Supplementary Information. Raw data files,
that a spin-down electron tunnels onto the donor. The electron spin data analysis code and simulation code are available at https://doi.
resonance (ESR) spectrum (Extended Data Fig. 1) shows eight distinct org/10.26190/5de9c295a8821.
resonance lines, each corresponding to a single nuclear spin eigenstate.
The electron spin can be flipped conditionally on the nuclear spin state, 35. Blöchl, P. E. Projector augmented-wave method. Phys. Rev. B 50, 17953–17979 (1994).
36. Kresse, G. & Furthmüller, J. Efficient iterative schemes for ab initio total-energy
resulting in single-shot nuclear spin readout. Electron spin readout is calculations using a plane-wave basis set. Phys. Rev. B 54, 11169–11186 (1996).
achieved by spin-to-charge conversion through spin-dependent tun- 37. Kresse, G. & Furthmüller, J. Efficiency of ab initio total energy calculations for metals and
nelling onto an SET and subsequent detection of the change in charge semiconductors using a plane-wave basis set. Comput. Mater. Sci. 6, 15–50 (1996).
38. Kresse, G. & Joubert, D. From ultrasoft pseudopotentials to the projector augmented-
occupation of the donor (see Supplementary Information section 5 wave method. Phys. Rev. B 59, 1758–1775 (1999).
for details). As each of these electron spin measurements project the 39. Mansir, J. et al. Linear hyperfine tuning of donor spins in silicon using hydrostatic strain.
Phys. Rev. Lett. 120, 167701 (2018).
nucleus into a single spin eigenstate, this is a quantum non-demolition
measurement. Therefore, each single-shot nuclear spin readout can Acknowledgements We thank T. Botzem and J. T. Muhonen for discussions. The research was
be repeated to increase the nuclear spin readout fidelity while retain- funded by the Australian Research Council Discovery Projects (grants DP150101863 and
ing the single-shot nature of the nuclear spin measurement. An NER DP180100969) and the Australian Department of Industry, Innovation and Science (grant
AUSMURI00002). V.M. acknowledges support from a Niels Stensen Fellowship. M.A.I.J. and
pulse has a probability Pflip of flipping the nuclear spin between two H.R.F. acknowledge the support of Australian Government Research Training Program
states. To measure Pflip, an NER pulse followed by nuclear spin readout Scholarships. J.J.P. is supported by an Australian Research Council Discovery Early Career
is performed Niterations times. The first record of the nuclear spin state Research Award (DE190101397). A.M. was supported by a Weston Visiting Professorship at the
Weizmann Institute of Science during part of the writing of this manuscript. We acknowledge
is used as a reference, and each subsequent record is compared to the support from the Australian National Fabrication Facility (ANFF), and from the laboratory of
one before it. This reveals the number of times that the nucleus flips, R. Elliman at the Australian National University for the ion implantation facilities. A.D.B. was
Nflips, between the two spin states. Therefore, the flip probability is supported by the Laboratory Directed Research and Development programme at Sandia
National Laboratories, Project 213048. Sandia National Laboratories is a multi-missions
simply the number of flips per number of recorded attempts, that is, laboratory managed and operated by National Technology and Engineering Solutions of
Pflip = Nflips/(Niterations − 1). Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the National
Nuclear Security Administration of the US Department of Energy under contract DE-
NA0003525. The views expressed in this manuscript do not necessarily represent the views
Theoretical modelling of the US Department of Energy or the US Government. K.M.I. acknowledges support from
The spin Hamiltonian of the 123Sb nucleus takes the form: Grant-in-Aid for Scientific Research by MEXT.
Author contributions S.A. and M.A.I.J. performed the measurements under the supervision of Competing interests S.A., V.M. and A.M. have submitted a patent application that describes
V.M., A.L. and A.M., with the assistance of V.S., M.T.M. and H.R.F.; S.A. and M.A.I.J. analysed the the use of electrically controlled high-spin nuclei for quantum information processing
data under the supervision of V.M. and A.M., with the assistance of H.R.F., V.S., J.J.P. and A.L.; (AU2018900665A).
A.D.B., S.A., V.M., B.J. and A.M. developed a microscopic theory supported by finite-element
modelling by B.J. and electronic structure calculations by A.D.B.; F.E.H. partially fabricated the Additional information
device under the supervision of A.S.D., on isotopically enriched material supplied by K.M.I. and Supplementary information is available for this paper at https://doi.org/10.1038/s41586-020-
M.T.M. subsequently fabricated the aluminium gate structures under the supervision of V.M. 2057-7.
and A.M.; J.C.M. designed and performed the 123Sb ion implantation; S.A., V.M., B.J., M.A.I.J., Correspondence and requests for materials should be addressed to A.M.
A.D.B., H.R.F. and A.M. wrote the manuscript and Supplementary Information, with input from Reprints and permissions information is available at http://www.nature.com/reprints.
all co-authors; A.M. initiated and supervised the research programme.
Article

fraction

Extended Data Fig. 1 | ESR spectrum in a magnetic field of B0 = 1.496 T. a, ESR possible cause for this deviation is strain, which is known to modify the
frequencies for eight nuclear states. The average difference between hyperfine interaction39. b, ESR spectral lines. For each nuclear state, the
successive ESR transition frequencies (black line) gives a hyperfine interaction nucleus was initialized at the start of each microwave sweep, and adiabatic ESR
of A = 96.5 MHz, substantially lower than the bulk value of 101.52 MHz. One pulses with 1 MHz frequency deviation were applied to excite the electron.
Extended Data Fig. 2 | NER Rabi oscillations on resonance. a, b, Nuclear Rabi for Δm I = ±2 transitions (b). Black lines are non-decaying sinusoidal fits to the
oscillations measured with varying NER pulse duration, tNER, while the pulse data, and error bars show the 68% confidence level.
amplitude was fixed atV RF RF
gate = 20 mV for Δm I = ±1 transitions (a) and V gate = 40 mV
Article

Extended Data Fig. 3 | NER spectral line shifts for varying d.c. gate voltage. resulting in shifts of the resonance peaks. A single fit to the resonance
a, b, The spectral lines of all Δm I = ±1 transitions (a) and Δm I = ±2 (b) transitions frequency shifts of all Δm I = ±1 and Δm I = ±2 transitions (solid lines) gives an
are measured while the d.c. gate voltage bias ΔV DC gate
is varied (columns) during estimate of the gate-dependent quadrupole shift of ∂fQ /∂V DC gate
= 9.9(3) Hz mV −1.
gate gate
the NER pulse. We note that this change in V DC is applied on top of large gate From the top to the bottom transition, the drive strengths V RF are 20 mV,
voltages, of the order of 0.5 V, which are necessary to electrostatically tune the 20 mV, 25 mV, 25 mV, 20 mV and 25 mV for Δm I = ±1 (a) and 30 mV, 30 mV, 40 mV,
gate
device to enable its operation. The varying ΔV DC modifies the quadrupole 40 mV, 40 mV and 40 mV for Δm I = ±2 (b). Error bars show the 68% confidence
interaction via the LQSE (see Supplementary Information section 7 for details), level.
Extended Data Fig. 4 | Position triangulation of the 123Sb donor. The colour Information section 7A for details). This has little effect laterally, but greatly
map shows the probability of finding the donor in a certain location. confines the likely depth range of the donor within the range expected on the
a, b, Probability density function found using a least-squares estimation basis of the donor implantation parameters. The most likely donor position,
comparing simulated gate-to-donor coupling strengths with the indicated by a cross, is at the lateral location (x, z) = (13 nm, 8 nm) at a depth of
experimentally observed strengths (see Supplementary Information y = −5 nm. The probability density functions are normalized over the model
section 7A for details, including the locations of the donor gates DFR, DFL and volume and are integrated over the out-of-plane axis in both panels,
DBR). To improve on the low resolving power of the triangulation method in the specifically, P(x, z) = ∫P(r)dy and P(y, z) = ∫P(r)dx. The contour lines mark the
y direction, the triangulation probability density function is multiplied with 68% and 95% confidence regions.
the donor implantation probability density function (see Supplementary