Charge-noise induced dephasing in silicon hole-spin qubits

Ognjen Malkoc,1 Peter Stano,1 and Daniel Loss1, 2

1
RIKEN Center for Emergent Matter Science, Wako-shi, Saitama 351-0198, Japan
2
Department of Physics, University of Basel, Klingelbergstrasse 82, CH-4056 Basel, Switzerland
We investigate theoretically charge-noise induced spin dephasing of a hole confined in a quasi-two-
dimensional silicon quantum dot. Central to our treatment is accounting for higher-order corrections
to the Luttinger Hamiltonian. Using experimentally reported parameters, we find that the new terms
give rise to sweet-spots for the hole-spin dephasing, which are sensitive to device details: dot size and
arXiv:2201.06181v1 [cond-mat.mes-hall] 17 Jan 2022

asymmetry, growth direction, and applied magnetic and electric fields. Furthermore, we estimate
that the dephasing time at the sweet-spots is boosted by several orders of magnitude, up to order
of milliseconds.

Introduction. Silicon is promising for realizing scalable a) [001]

[010]
B b) HH LH

qubits using quantum-dot electrons to store and process [100]

quantum information [1–4]. The recent attention to silicon
stems from compatibility with industrial fabrication [5–8] E
and low noise from nuclear spins. The latter effect, so far
a major obstacle for spin qubits in GaAs, can be further
suppressed using holes instead of electrons [9–12]. Holes also
offer stronger spin-orbit coupling [13–16], essential for elec-
tric spin control without micromagnets or on-chip ESR lines
[17]. Taken together, the reduced susceptibility to nuclear FIG. 1. a) A schematic of a lateral quantum dot hosting a
noise [18], the absence of valley degeneracy, and fully-electric spin qubit. A hole from the two-dimensional-hole-gas (2DHG)
control make holes in silicon a very attractive platform for is trapped by the confinement potential, that has a major axis
making an angle δ with the crystallographic [100] axis. b) The
scalable spin qubits. basis states in the calculation, showing the heavy hole and light
In a quasi-two-dimensional (planar) quantum dot,with the hole subbands of the unperturbed Hamiltonian. The red area
strongest confinement along the growth direction, the con- shows the states used in the perturbation series of Eq. (6).
finement splits the fourfold degeneracy at the Γ point into
light and heavy holes, offering a resilient spin qubit residing
in the heavy hole subspace [9, 19, 20]. Spin blockade detec- While the conduction band non-parabolicity has been
tion [21–24], control over the charge state down to a single studied in zinc-blende crystals in detail [55, 56], including
hole [25, 26], fabrication of arrays [27–29], and demonstra- its effects on the g-factor [57], the valence band requires a
tion of single [30, 31] and two-qubit operations [32] are among separate treatment. To this end, we derive the corrections
recent experimental achievements with planar dots. In con- to the Luttinger Hamiltonian up to the fourth order in mo-
trast, the strong confinement-induced spin-orbital mixing in mentum and up to linear in the electric field. Even though
a nanowire geometry [33–35] gives large and tunable spin- one can generate these terms by symmetry analysis, for ex-
orbit interaction [36–39] and fast spin manipulation [40, 41]. ample using the tables in Ref. [58], we are not aware of their
The strong sensitivity to the electric field is a generic fea- prefactors being known. To evaluate the spin-orbit effects
ture of hole-spin qubits. Among its most direct manifesta- reliably, however, these prefactors are necessary. We calcu-
tions, the electrical response of the g-factor has been reported late them within the 14-band k · p model [59], using up to the
for various designs [34, 42–51]. While it offers increased elec- fifth order of the Löwdin perturbation theory. The resulting
trical tunability, it also implies a higher susceptibility to elec- effective model is valid for any materials with diamond crys-
trical noise. With nuclear noise negligible, charge noise be- tal structures, such as Si and Ge, but we will focus on the
comes the primary concern for qubit coherence [17, 52, 53]. former material. By focusing exclusively on Si devices, we
Aiming at long coherence time, the most favorable scenario use the Si band structure parameters and identify a minimal
seems to be a single hole in an isolated planar quantum dot. set, given in Eq. (2), which, together with Eq. (1), describes
Assessing this ultimate limit on the hole-spin coherence is the essential correction terms to the valence band in the bulk
our main objective. for describing spin dephasing.
We find that in planar dots the spin-electric coupling is We obtain the spin-qubit Hamiltonian by projecting the
dominated by higher-order (non-quadratic in momentum) valence band Hamiltonian onto the lowest orbital state, de-
terms, which are not contained in the often used and well- fined by the three-dimensional confinement potential. Be-
known Luttinger Hamiltonian [54]. This finding is among low, we will use second order perturbation theory to include
our main results. the effects of higher orbital states. From the qubit Hamilto-
 2

nian we evaluate two quantities of interest: the g-tensor and and the band-warping terms,
the dephasing rate. Our main result in this part is twofold.
First, the typical dephasing time is found to be on the order H12 =Γ12 ({kX , kY }2 + {kY , kZ }2 + {kZ , kX }2 ),
2 2
(2b)
of tens of microseconds. This value is then the ultimate up- H32 =Γ32 J32 · (2kX kY −(kY2 + kX
2 2
)kZ , (kY2 − kX
2 2
)kZ ).
per limit in any design with holes in silicon gated dots with
an in-plane magnetic field. With other than planar dots one In these equations, J53 = ({Jx , Jy2 − Jz2 }, {Jy , Jz2 −
would expect the dephasing time to be much smaller Jx2 }, {Jz , Jx2 − Jy2 }), J32 = (JZ2 − J · J/3, JX
2
− JY2 ), and c.p.
Second, we find pronounced sweet spots, where the de- means cyclic permutation. The values of prefactors are given
phasing time is boosted up to milliseconds. Their position in Tab. I of the Supplemental Material [61]. The figures in
in parameter space is sensitive to all system parameters. The the following sections are plotted using all 27 corrections,
suggestion to search for experimentally robust sweet spots is together denoted as δHL . In the Supplementary Material
the main practical implication of our work. (SM) [61], we show analogous figures produced with Eq. (2)
Valence band corrections. All symmetry-allowed terms in instead. One finds that the latter is an excellent approxima-
the valence band of silicon up to quadratic in the kinetic tion.
momentum h̄k are contained in the Luttinger Hamiltonian Effective hole-spin qubit Hamiltonian. We consider a hole
[54], confined in a device shown in Fig. 1(a). Since the spin-orbit
interaction is anisotropic, as can be seen already in HL , the
h̄2
  
5γ2 2 2 2 splitting gap depends on the growth direction: [62] changing
HL = − γ1+ k +2γ2 (kX JX+kY2 JY2 +kZ
2 2
JZ )
2m0 2 (1) it from [001] to [111], in our model the gap increases from
i 2.1 meV to 29.7 meV for a triangular vertical confinement
+4γ3 (kXY JXY +kY Z JY Z+kZX JZX) −2µB (κJ+qJ3 )·B.
with 10 V/µm electric field. A similar splitting arises from
Here, the components of the vectors J = (JX , JY , JZ ) and strain
J3 = (JX 3
, JY3 , JZ3 ), are the spin 3/2 operators, m0 is the free- [63–65]. For instance, a strain of 0.5% in Si would give a
electron mass, Aij = Ai Aj + Aj Ai is the anticommutator, splitting of order 10 meV 3 . The principal difference is which
and the coefficients γ1 , γ2 , γ3 , κ, and q are the Luttinger pa- spin is the ground state: with strain, depending on its type,
rameters 1 . Also, X, Y , and Z denote the [001], [010], and it can be either the heavy or the light hole
[001] crystallographic axis, respectively. Finally, due to the [66, 67]. On the other hand, in the planar geometry the
relation k × k = −ieB/h̄, the magnetic B-field components heavy hole ground state always results from the vertical con-
are counted as quadratic in momentum. finement alone [68]. We assume that the heavy hole subspace
Using symmetry analysis, we derive corrections to Eq. (1) is the ground state and the qubit is defined therein, as a con-
up to the fourth order in momentum and up to linear in figuration most resilient to charge noise.
the electric field. For these two groups we get fifteen and We adopt standard choices to describe the quantum dot
twelve terms, respectively. We evaluate their prefactors us- confinement: a triangular potential for the vertical part and
ing the 14-band k · p model, in the fourth and fifth order of an anisotropic harmonic for the in-plane part,
the Löwdin perturbation theory [60], respectively. With the 3/2
(
mxy 2 2 2 2 eEz z for z > 0
full list including formulas for prefactors given elsewhere, we Vxy = 2 (εx x + εy y ), Vz = . (3)
restrict here ourselves to an excerpt. Namely, after analyz- 2h̄ V0 for z ≤ 0
ing each of the 27 terms as outlined below, we identify the 3/2
terms contributing dominantly to the spin dephasing for de- Here, V0 is the heterostructure band offset, mxy is the in-
vices grown along [001] 2 . There are magnetic-field generated plane effective electron mass discussed below, εx and εy are
terms, the in-plane excitation energies, Ez is the electric field, and
x, y, and z are dot coordinates. In calculations, we take the
2
H41 =µB (κ41 J + q41 J3 ) · B(kX + kY2 + kZ
2
), limit V0 → ∞, so that the wave function is zero for z ≤ 0 4 .
2
H42 =µB (κ42 J + q42 J3 ) · (BX kX , BY kY2 , BZ kZ
2
), Having specified the confinement, we have
(2a)
H43 =µB (κ43 J + q43 J3 ) · (kX (kY BY + kZ BZ ), c.p.), H = HL + δHL + Vxy + Vz , (4)
H53 =µB Γ53 J53 · (BX (kY2 − 2
kZ ), c.p.),

3 The estimation is for bi-axial strain in the dot-plane, where ∆strain =

1 We use the experimentally measured values γ1 = 4.285, γ2 = |b|(1 + c11 /c12 )εk , where εk is the in-plane strain. The deformation
0.339, γ3 = 1.446, κ = −0.42, and q = 0, instead of calculating them potential b = 2.2 and elastic coefficients cij are taken from Ref. [58].
within the 14-band model, since the latter procedure underestimates 4 The limit is subtle, as it does not commute with evaluating matrix
these parameters. elements of differential operators [App. C in Ref. [69], App. B in
2 The set of terms depends on the growth orientation of the device. [70]]. For crystal momentum expectation values, for powers higher
An analogous set of terms which dominate devices grown along [111] than quadratic, it is necessary to account for artifacts of the hard-wall
in Eq. 9 of the Supplemental Material. potential.
 3

ε̄[meV] η
as the full—three-dimensional—Hamiltonian describing the −3.0 −2.5 −2.0 −1.5 −1.0 0.00 0.25 0.50 0.75 1.00
confined hole. We include the orbital effects of the in-plane 0.1
a) b) 0.1
magnetic field components via the vector potential, enter- 0.0 0.0
ing the momentum h̄k = −ih̄∇ + eA. We fix the gauge
−0.1
to A = (z − z0 )(By , −Bx ) + eBz (−y, x), where z0 is the −0.1

ground state expectation value of the z coordinate. Our next −0.2

−0.2
goal is to reduce this microscopic description into an effective 0.1
0.2 c) d)
Hamiltonian for the spin qubit, a two-level system. 0.0
0.1
We first define the unperturbed Hamiltonian by supple- −0.1
menting the confinement by terms quadratic in momentum 0.0 −0.2
and not coupling the heavy hole and light hole subspaces, −0.1 −0.3
−0.2 −0.4
h̄2 (∂x2 + ∂y2 ) h̄2 ∂z2 0 π/4 π/2 3π/4 π 1 4 8 12 16
H0J = + − Vxy − Vz . (5) δ Ez [mV/nm]
2mJxy 2mJz
FIG. 2. Heavy-hole qubit g-tensor components for a dot with
The effective masses depend on the Luttinger parameters,
zd -axis along [001]. The colored lines show ĝxx (red), ĝxy (blue),
the hole-spin subspace (J = |Jz | ∈ {1/2, 3/2} for the light ĝyx (green), and ĝyy (yellow). The x-axis represents (a) the average
and heavy hole, respectively), and the growth direction. We dot energy ε̄ = (εx + εy )/2, (b) dot asymmetry η = (εy − εx )/2ε̄,
give formulas for some cases of interest in Tab. II in the Sup- (c) dot orientation δ, and (d) vertical confinement electric field Ez .
plementary Material. The unperturbed Hamiltonian defines The dashed line, defined by εx = −1 meV, εy = −3 meV, δ =
the basis for the perturbation theory. Since it is separable in π/4 , and Ez = 10 mV/nm, denotes a common reference point.
in-plane coordinates x and y, the vertical coordinate z, and
the spin, the basis states |J, nx , ny , nz i can be indexed by four
Effective g-tensor. Evaluating Eq. (6) gives the Hamilto-
quantum numbers: the pair (nx , ny ) is the Fock-Darwin spec-
nian H describing the qubit as a two-level system. Due to
trum indexes, while nz labels eigenstates of the triangular √ the time-reversal symmetry, at zero magnetic field the two
potential, associated with energy scale Jz = (h̄eEz / mJz )2/3
states are degenerate. Neglecting terms higher than linear in
(see App. A1 in Ref. [71] for details on the z-confinement
the magnetic field, an approximation that we adopt in eval-
eigenstates). The splitting of heavy and light holes ∆HL dis-
uating Eq. (6), the latter becomes the Hamiltonian of a spin
cussed in the introduction is the energy difference of the two
one-half,
ground states of Eq. (5) for the two values of the spin index
J.
X
H= µB Bi ĝij τj . (7)
The qubit Hamiltonian follows by integrating out the or- i,j=x,y,z
bital degrees of freedom, with the excited states taken into
account within the second-order perturbation theory, Here, τ is a vector of Pauli matrices defined with up and
down spin one-half states corresponding to perturbed spin
h X δH|J 0 , nihJ 0 , n|δH i states of H, where the perturbation in Eq. (6) mixes to the
H =hJ, 0| H + |J, 0i. (6)
E |J,0i −E 0
|J ,ni HH (Jz = ±3/2) spin ground state the LH (Jz = ±1/2)
(J 0 ,n)6=(J,0)
spin states, and ĝ is a second-rank tensor, the g-tensor. We
Here, δH = H − H0J , the summation is over all excited or- have thus reduced the three-dimensional qubit-description
bital states, the vector n = (nx , ny , nz ), and E|J,ni is the of Eq. (4) to a much simpler effective two-level model. Nev-
unperturbed eigenstate energy. ertheless, this model reflects orbital effects through the g-
This derivation follows the procedure of Refs. [57, 71] with tensor dependence on confinement electric fields, which we
one difference. In those references, the reduction proceeded now examine.
in two steps: first integrating out the vertical coordinate Figure 2 shows the g-tensor for a quantum dot with the
z, then the in-plane coordinates x and y. Here, we do not z-axis along [001]. The g-tensor in-plane components are
neglect the in-plane excitation energies in the denominator of plotted as functions of the lateral dot size (panel a), its
Eq. (6), as they are comparable to ∆HL . On the other hand, asymmetry (panel b), its orientation (panel c), and the
we restrict the sum over nz in Eq.(6) to the lowest excited vertical-confinement strength (panel d). The components
state [see Fig. 1(b)]. The restriction is suitable for quasi- off-diagonal in the in-plane versus vertical coordinate groups,
two-dimensional dots, where out-of-plane excitation energies gxz , gzx , gyz , gzy , are zero. The out-of-plane component gzz
are larger than the in-plane ones. The resulting approximate is typically an order of magnitude larger than the in-plane
form of H is the basis for the two main quantities of our work, ones, and does not depend appreciably on any parameter ex-
the effective g-tensor, and the qubit energy. The dependence cept of the vertical electric field. We include gzz in Fig. 5
of the latter on the electric field is responsible for dephasing of the Supplementary Material. The g-tensor is strongly
of the hole-spin qubit. anisotropic, a consequence of the confinement breaking all
 4

symmetries of the crystal [72, 73]. In realistic samples which

π π
are neither perfectly symmetric nor aligned with any partic- 2 a) b) 2 T2∗ [µs]
π π
5000
ular direction with respect to the crystal axes one expects 4 4
0 0

β
large variations of the g-tensor components. Most impor- 1000
− π4 n̂ = [001] n̂ = [111] − π4
tantly, the g-tensor components clearly depend on the con- HL + δHL HL + δHL 500
π π
finement electric field. Through this dependence, the charge 2 c) Ez [mV/nm] d) Ez [mV/nm] 2 250
π π
noise influences the qubit energy and causes dephasing. We 4 4 100
0 0

β
are now in a position to estimate the resulting dephasing 50
− π4 n̂ = [001] n̂ = [111] − π4
time. HL HL
− π2 − π2 0
Coherence time. The charge noise in the sample and ex- 1 5 10 15 20 1 5 10 15 20
periment electronics leads to fluctuations of the electric field Ez [mV/nm] Ez [mV/nm]
at the dot location, and thereby to fluctuations of the qubit
energy. The connection can bep seen by writing the qubit FIG. 3. Dephasing time T2∗ for spin-qubit devices grown along
energy, using Eq. (7), as h̄ω = µ2B Bj Bk gji gki , where re- n̂ = [001] and n̂ = [111] when using the full Hamiltonian (top)
peated indices are summed over. The electric field enters and only HL (bottom). We consider a dot with δ = 0, an in-plane
the g-tensor in two ways: by defining the shape of the dot magnetic field B = B(cos β, sin β, 0), where B = 1 T, and Ez is
the vertical electric field. We consider dot confinement lengths
confinement and by inducing bandstructure terms. We find 3/2
lx = h̄(mxy εx )−1/2 = 20 nm, ly = 15 nm, and ωh = 50 kHz,
that the latter terms have negligible influence; this is the rea- −2
ωl = 10 Hz for the frequency cut-offs.
son why no electric-field generated term is listed in Eq. (2).
We also neglect fluctuating electric in-plane fields, since a
uniform field does not change the shape of a harmonic con- sweet spots is not specific to the [001] growth direction. To
finement. The noise in the z-component of the electric field illustrate this, Fig. 3b and Fig. 3d show the dephasing for
remains, changing the qubit energy through changes in the [111]. The situation is analogous, though there are quanti-
vertical confinement
R ∞ strength. The noise is described by its tative differences: First, the discrepancy from not including
0
spectrum, S(ω)= −∞dτ 0 eiωτ hEz (0)Ez (τ 0 )i, with the bracket the δHL terms is even more drastic. Second, the sweet spot
standing for a statistical average. We further assume that appears within a larger range of the electric fields and for
1/f noise is dominant and take S(ω) = A/|ω|. The levels any orientation of the magnetic field. Since we found similar
of noise measured in silicon samples were recently reviewed sweet spots for [011] growth direction (plots not shown), we
in Ref. [74]. However, the majority of reports give them as conjecture that sweet spots in lateral spin hole qubits are
fluctuations of the dot energy, not the electric field. We take generic [16, 41].
2
A = 2800 V2 /m as measured in Ref. [75] for the electric Before concluding, we review available experimental re-
field fluctuations, which would yield dot energy fluctuations sults on the confined hole spin dephasing times. There are
within the range reported in Ref. [74]. not many: An optical probe of a single hole in a III-V self-
Reference [76] finds that 1/f noise causes a Gaussian decay assembled dot gave T2∗ of 100 ns in Ref. [77] and 0.5 µs in
with a pure-dephasing rate Ref. [12], with the coherence time T2 estimated an order
of magnitude larger. The only numbers we are aware of in
1/T2∗ = |∂Ez ω| A ln[ωh /2πωl ],
p
(8) silicon is T2∗ = 60 ns from Ref. [17], which was prolonged
where ωl and ωh are the low and high frequency cut-offs, four-fold upon Hahn-echo, as expected for a 1/f noise 5 , and
respectively. T2∗ = 440 ns from Ref. [15] for a hole spin qubit in a Si
Our main result is the analysis of the dephasing time T2∗ FinFET device.
calculated from Eq. (8). We plot the dephasing time in Fig. 3, While all these numbers are lower than our T2∗ , the differ-
varying the electric field strength Ez and the in-plane mag- ence is not so drastic considering that the charge noise levels
netic field orientation for fixed strength B = 1 T. Figure 3a have large variations among different materials and samples
(3c) shows the dephasing time T2∗ for devices grown along [74, 78]. Coincidentally, the dephasing times that we ob-
[001] when including (excluding) the δHL terms. We note tained are comparable to nuclear-limited dephasing times of
that in both models, the dephasing time is maximized for electrons in silicon dots: T2∗ in natural silicon is a few mi-
the B-field along x-axis, i.e. β = 0, except at small electric croseconds, close to our values away from the sweet spot,
fields in Fig. 3a. The dephasing time ranges from tens to and a millisecond in purified silicon-28, comparable to our
hundreds of µs in most of the plot area. However, including sweet-spot values. We conclude that the intrinsic material
δHL reveals a line of sweet spots with the dephasing time
boosted up to milliseconds. For the parameter space con-
sidered in Fig. 3, the sweet spot line appears only within a 5 For 1/f noise, the dephasing time under a Hahn-echo equals [76] T2∗
limited range of the magnetic-field directions, |β| <
p
∼ π/8, and times ln(ωh /2πωl )/ ln 2, evaluating to 4.4 for the parameters given
electric field magnitudes, Ez < ∼ 5V /µm. The appearance of
in the caption of Fig. 3.
 5

charge noise might be limiting coherence in some of these [12] J. H. Prechtel, A. V. Kuhlmann, J. Houel, A. Ludwig, S. R.
experiments, while in some, such as Si FinFETs, the nuclear Valentin, A. D. Wieck, and R. J. Warburton, Nature Mate-
noise is still the limiting factor, see Refs.[15, 18]. Our results rials 15, 981 (2016).
suggest that holes in planar dots can reach coherence com- [13] A. Bogan, S. Studenikin, M. Korkusinski, L. Gaudreau,
P. Zawadzki, A. S. Sachrajda, L. Tracy, J. Reno, and T. Har-
parable to electrons 6 , and searching for the hole-qubit sweet gett, Phys. Rev. Lett. 120, 207701 (2018).
spots experimentally looks attractive. [14] B. Venitucci and Y.-M. Niquet, Physical Review B 99,
Conclusions. In this work, we have quantified charge-noise 115317 (2019).
induced dephasing rate of a silicon spin hole qubit, and the [15] L. C. Camenzind, S. Geyer, A. Fuhrer, R. J. Warbur-
effective g-factor. For typical dot dimensions and external ton, D. M. Zumbühl, and A. V. Kuhlmann, (2021),
magnetic and electric fields, we find it is necessary to go arXiv:2103.07369.
beyond the Luttinger model to assess the g-tensor and de- [16] S. Bosco, B. Hetényi, and D. Loss, PRX Quantum 2, 010348
(2021).
phasing reliably; the difference is qualitative. Our model, [17] R. Maurand, X. Jehl, D. Kotekar-Patil, A. Corna, H. Bo-
which can also be extended to devices using other diamond huslavskyi, R. Laviéville, L. Hutin, S. Barraud, M. Vinet,
crystal materials, e.g., germanium hole spin devices, predicts M. Sanquer, and S. De Franceschi, Nature Communications
a sweet spot for the dephasing time. We find that the sweet 7, 13575 (2016).
spots depend on the device growth direction, confinement [18] S. Bosco and D. Loss, Phys. Rev. Lett. 127, 190501 (2021).
potential, and in-plane magnetic field orientation. Our work [19] D. V. Bulaev and D. Loss, Phys. Rev. Lett. 98, 097202
(2007).
leaves space for interesting extensions. For example, the de-
[20] R. W. Martin, R. J. Nicholas, G. J. Rees, S. K. Haywood,
pendence on the device geometry prompts the question of N. J. Mason, and P. J. Walker, Physical Review B 42, 9237
how the additional spin-orbit interactions impact spin de- (1990).
phasing in other devices, such as nanowire-based hole-spin [21] R. Li, F. E. Hudson, A. S. Dzurak, and A. R. Hamilton,
qubits [37] or FinFETs [16]. Nano Letters 15, 7314 (2015).
This work was partially supported by the Swiss National [22] H. Bohuslavskyi, D. Kotekar-Patil, R. Maurand, A. Corna,
Science Foundation and NCCR SPIN. S. Barraud, L. Bourdet, L. Hutin, Y.-M. Niquet, X. Jehl,
S. De Franceschi, M. Vinet, and M. Sanquer, Applied Physics
Letters 109, 193101 (2016).
[23] Y. Yamaoka, K. Iwasaki, S. Oda, and T. Kodera, Japanese
Journal of Applied Physics 56, 04CK07 (2017).
[24] D. Q. Wang, O. Klochan, J.-T. Hung, D. Culcer, I. Farrer,
[1] D. Loss and D. P. DiVincenzo, Physical Review A 57, 120
D. A. Ritchie, and A. R. Hamilton, Nano Letters 16, 7685
(1998). (2016).
[2] A. Chatterjee, P. Stevenson, S. De Franceschi, A. Morello, [25] S. D. Liles, R. Li, C. H. Yang, F. E. Hudson, M. Veldhorst,
N. P. de Leon, and F. Kuemmeth, Nature Reviews Physics
A. S. Dzurak, and A. R. Hamilton, Nature Communications
3, 157 (2021).
9, 3255 (2018).
[3] P. Stano and D. Loss, (2021), arXiv:2107.06485. [26] A. J. Sousa de Almeida, A. M. Seco, T. van den Berg,
[4] G. Burkard, T. D. Ladd, J. M. Nichol, A. Pan, and J. R. B. van de Ven, F. Bruijnes, S. V. Amitonov, and F. A.
Petta, (2021), arXiv:2112.08863. Zwanenburg, Physical Review B 101, 201301 (2020).
[5] S. Thompson, Guangyu Sun, Youn Sung Choi, and
[27] N. W. Hendrickx, W. I. Lawrie, M. Russ, F. van Riggelen,
T. Nishida, IEEE Transactions on Electron Devices 53, 1010
S. L. de Snoo, R. N. Schouten, A. Sammak, G. Scappucci,
(2006). and M. Veldhorst, Nature 591, 580 (2021).
[6] M. Gonzalez-Zalba, S. de Franceschi, E. Charbon, T. Meu- [28] W. I. L. Lawrie, H. G. J. Eenink, N. W. Hendrickx, J. M.
nier, M. Vinet, and A. Dzurak, Nature Electronics , 1 (2021).
Boter, L. Petit, S. V. Amitonov, M. Lodari, B. Paque-
[7] A. Zwerver, T. Krähenmann, T. Watson, L. Lampert,
let Wuetz, C. Volk, S. G. J. Philips, G. Droulers, N. Kalhor,
H. George, R. Pillarisetty, S. Bojarski, P. Amin, S. Amitonov, F. van Riggelen, D. Brousse, A. Sammak, L. M. K. Van-
J. Boter, et al., (2021), arXiv:2101.12650. dersypen, G. Scappucci, and M. Veldhorst, Applied Physics
[8] M. Vinet, Nature nanotechnology , 1 (2021). Letters 116, 080501 (2020).
[9] D. V. Bulaev and D. Loss, Phys. Rev. Lett. 95, 076805
[29] F. van Riggelen, N. W. Hendrickx, W. I. L. Lawrie, M. Russ,
(2005).
A. Sammak, G. Scappucci, and M. Veldhorst, Applied
[10] D. Heiss, S. Schaeck, H. Huebl, M. Bichler, G. Abstreiter, Physics Letters 118, 044002 (2021).
J. J. Finley, D. V. Bulaev, and D. Loss, Phys. Rev. B 76, [30] A. Hofmann, D. Jirovec, M. Borovkov, I. Prieto, A. Ballabio,
241306 (2007).
J. Frigerio, D. Chrastina, G. Isella, and G. Katsaros, (2019),
[11] E. A. Chekhovich, A. B. Krysa, M. S. Skolnick, and A. I.
arXiv:1910.05841.
Tartakovskii, Phys. Rev. Lett. 106, 027402 (2011). [31] N. Hendrickx, W. Lawrie, L. Petit, A. Sammak, G. Scap-
pucci, and M. Veldhorst, Nature communications 11, 1
(2020).
[32] N. W. Hendrickx, D. P. Franke, A. Sammak, G. Scappucci,
6 Using the hyperfine coupling coefficient for silicon in Ref. [79] with and M. Veldhorst, Nature 577, 487 (2020).
Eq. 29 of Ref. [80], we obtain T2∗ = 6µs, while for purified silicon [33] Y. Hu, F. Kuemmeth, C. M. Lieber, and C. M. Marcus,
with 800ppm we get T2∗ = 0.3ms.
 6

Nature Nanotechnology 7, 47 (2011). becq, G. Allison, T. Honda, T. Kodera, S. Oda, Y. Hoshi,

[34] V. S. Pribiag, S. Nadj-Perge, S. M. Frolov, J. W. G. van den N. Usami, K. M. Itoh, and S. Tarucha, Nature Nanotechnol-
Berg, I. van Weperen, S. R. Plissard, E. P. A. M. Bakkers, ogy (2017), 10.1038/s41565-017-0014-x.
and L. P. Kouwenhoven, Nature Nanotechnology 8, 170 [54] J. M. Luttinger, Physical Review 102, 1030 (1956).
(2013). [55] N. R. Ogg, Proceedings of the Physical Society 89, 431
[35] F. Gao, J.-H. Wang, H. Watzinger, H. Hu, M. J. Rančić, (1966).
J.-Y. Zhang, T. Wang, Y. Yao, G.-L. Wang, J. Kukučka, [56] U. Rössler, Solid State Communications 49, 943 (1984).
L. Vukušić, C. Kloeffel, D. Loss, F. Liu, G. Katsaros, and [57] P. Stano, C.-H. Hsu, M. Serina, L. C. Camenzind, D. M.
J.-J. Zhang, Advanced Materials 32, 1906523 (2020). Zumbühl, and D. Loss, Physical Review B 98, 195314 (2018).
[36] C. Kloeffel, M. Trif, and D. Loss, Physical Review B 84, [58] R. Winkler, Spin-orbit coupling effects in two-dimensional
195314 (2011). electron and hole systems, Springer tracts in modern physics
[37] F. N. M. Froning, M. J. Rančić, B. Hetényi, S. Bosco, M. K. No. v. 191 (Springer, Berlin; New York, 2003).
Rehmann, A. Li, E. P. A. M. Bakkers, F. A. Zwanenburg, [59] P. Pfeffer and W. Zawadzki, Physical Review B 53, 12813
D. Loss, D. M. Zumbühl, and F. R. Braakman, Phys. Rev. (1996).
Research 3, 013081 (2021). [60] P.-O. Löwdin, The Journal of Chemical Physics 19, 1396
[38] C. Kloeffel, M. J. Rančić, and D. Loss, Phys. Rev. B 97, (1951).
235422 (2018). [61] See Supplemental Material at [url...] for more details on the
[39] S. Bosco, M. Benito, C. Adelsberger, and D. Loss, Phys. numerical values used in the calculation, and justification for
Rev. B 104, 115425 (2021). the minimal model.
[40] F. N. M. Froning, L. C. Camenzind, O. A. H. van der Molen, [62] T. Takahashi, T. Kodera, S. Oda, and K. Uchida, Journal
A. Li, E. P. A. M. Bakkers, D. M. Zumbühl, and F. R. of Applied Physics 109, 034505 (2011).
Braakman, Nature Nanotechnology (2021), 10.1038/s41565- [63] N. W. Hendrickx, D. P. Franke, A. Sammak, M. Kouwen-
020-00828-6. hoven, D. Sabbagh, L. Yeoh, R. Li, M. L. V. Tagliaferri,
[41] Z. Wang, E. Marcellina, A. R. Hamilton, J. H. Cullen, M. Virgilio, G. Capellini, G. Scappucci, and M. Veld-
S. Rogge, J. Salfi, and D. Culcer, npj Quantum Informa- horst, Nature Communications 9 (2018), 10.1038/s41467-
tion 7, 1 (2021). 018-05299-x.
[42] T. Andlauer and P. Vogl, Phys. Rev. B 79, 045307 (2009). [64] Y. Sun, S. E. Thompson, and T. Nishida, Journal of Applied
[43] G. Katsaros, P. Spathis, M. Stoffel, F. Fournel, M. Mongillo, Physics 101, 104503 (2007).
V. Bouchiat, F. Lefloch, A. Rastelli, O. G. Schmidt, and [65] W. J. Hardy, C. T. Harris, Y.-H. Su, Y. Chuang, J. Moussa,
S. De Franceschi, Nature Nanotechnology 5, 458 (2010). L. N. Maurer, J.-Y. Li, T.-M. Lu, and D. R. Luhman, Nan-
[44] F. Klotz, V. Jovanov, J. Kierig, E. C. Clark, D. Rudolph, otechnology 30, 215202 (2019).
D. Heiss, M. Bichler, G. Abstreiter, M. S. Brandt, and J. J. [66] Y. H. Huo, B. J. Witek, S. Kumar, J. R. Cardenas, J. X.
Finley, Applied Physics Letters 96, 053113 (2010). Zhang, N. Akopian, R. Singh, E. Zallo, R. Grifone, D. Krieg-
[45] N. Ares, V. N. Golovach, G. Katsaros, M. Stoffel, F. Fournel, ner, R. Trotta, F. Ding, J. Stangl, V. Zwiller, G. Bester,
L. I. Glazman, O. G. Schmidt, and S. De Franceschi, Phys. A. Rastelli, and O. G. Schmidt, Nature Physics 10, 46
Rev. Lett. 110, 046602 (2013). (2014).
[46] A. J. Bennett, M. A. Pooley, Y. Cao, N. Sköld, I. Farrer, [67] M. Lodari, A. Tosato, D. Sabbagh, M. A. Schubert,
D. A. Ritchie, and A. J. Shields, Nature Communications 4, G. Capellini, A. Sammak, M. Veldhorst, and G. Scappucci,
1522 (2013). Physical Review B 100, 041304 (2019).
[47] J. H. Prechtel, F. Maier, J. Houel, A. V. Kuhlmann, A. Lud- [68] M. V. Fischetti, Z. Ren, P. M. Solomon, M. Yang, and
wig, A. D. Wieck, D. Loss, and R. J. Warburton, Phys. Rev. K. Rim, Journal of Applied Physics 94, 1079 (2003).
B 91, 165304 (2015). [69] M. J. Carballido, C. Kloeffel, D. M. Zumbühl, and D. Loss,
[48] M. Brauns, J. Ridderbos, A. Li, E. P. A. M. Bakkers, and Phys. Rev. B 103, 195444 (2021).
F. A. Zwanenburg, Phys. Rev. B 93, 121408 (2016). [70] G. E. Simion and Y. B. Lyanda-Geller, Phys. Rev. B 90,
[49] B. Voisin, R. Maurand, S. Barraud, M. Vinet, X. Jehl, 195410 (2014).
M. Sanquer, J. Renard, and S. De Franceschi, Nano Let- [71] P. Stano, C.-H. Hsu, L. C. Camenzind, L. Yu, D. Zumbühl,
ters 16, 88 (2016). and D. Loss, Physical Review B 99, 085308 (2019).
[50] F. K. de Vries, J. Shen, R. J. Skolasinski, M. P. Nowak, [72] G. Scappucci, C. Kloeffel, F. A. Zwanenburg, D. Loss,
D. Varjas, L. Wang, M. Wimmer, J. Ridderbos, F. A. Zwa- M. Myronov, J.-J. Zhang, S. De Franceschi, G. Katsaros,
nenburg, A. Li, S. Koelling, M. A. Verheijen, E. P. A. M. and M. Veldhorst, Nature Reviews Materials , 1 (2020).
Bakkers, and L. P. Kouwenhoven, Nano Letters 18, 6483 [73] C. Gradl, R. Winkler, M. Kempf, J. Holler, D. Schuh,
(2018). D. Bougeard, A. Hernández-Mı́nguez, K. Biermann, P. V.
[51] A. Crippa, R. Maurand, L. Bourdet, D. Kotekar-Patil, Santos, C. Schüller, and T. Korn, Phys. Rev. X 8, 021068
A. Amisse, X. Jehl, M. Sanquer, R. Laviéville, H. Bo- (2018).
huslavskyi, L. Hutin, S. Barraud, M. Vinet, Y.-M. Niquet, [74] L. Kranz, S. K. Gorman, B. Thorgrimsson, Y. He, D. Keith,
and S. De Franceschi, Physical Review Letters 120, 137702 J. G. Keizer, and M. Y. Simmons, Advanced Materials ,
(2018). 2003361 (2020).
[52] J. Houel, J. H. Prechtel, A. V. Kuhlmann, D. Brunner, C. E. [75] A. V. Kuhlmann, J. Houel, A. Ludwig, L. Greuter, D. Reuter,
Kuklewicz, B. D. Gerardot, N. G. Stoltz, P. M. Petroff, and A. D. Wieck, M. Poggio, and R. J. Warburton, Nature
R. J. Warburton, Phys. Rev. Lett. 112, 107401 (2014). Physics 9, 570 (2013).
[53] J. Yoneda, K. Takeda, T. Otsuka, T. Nakajima, M. R. Del- [76] G. Ithier, E. Collin, P. Joyez, P. J. Meeson, D. Vion, D. Es-
 7

teve, F. Chiarello, A. Shnirman, Y. Makhlin, J. Schriefl, [79] P. Philippopoulos, S. Chesi, and W. A. Coish, Phys. Rev. B
and G. Schön, Physical Review B 72 (2005), 10.1103/Phys- 101, 115302 (2020).
RevB.72.134519. [80] T. Struck, A. Hollmann, F. Schauer, O. Fedorets, A. Schmid-
[77] D. Brunner, B. D. Gerardot, P. A. Dalgarno, G. Wüst, bauer, K. Sawano, H. Riemann, N. V. Abrosimov,
K. Karrai, N. G. Stoltz, P. M. Petroff, and R. J. Warburton, L. Cywiński, D. Bougeard, et al., npj Quantum Information
Science 325, 70 (2009). 6, 1 (2020).
[78] B. M. Freeman, J. S. Schoenfield, and H. Jiang, Applied [81] S. Richard, F. Aniel, and G. Fishman, Physical Review B
Physics Letters 108, 253108 (2016). 70, 235204 (2004).

Supplemental Material: “Charge-noise induced dephasing in silicon hole-spin qubits”

In the Supplemental Material we provide details used for our numerical calculations and justify our minimal model proposed
for the spin dephasing in the main text. In particular, we show in Table I the numerical coefficients that are used for the
minimal model. Moreover, we provide omitted results such as out-of-plane g-tensor components and dephasing times for
devices with a dot-normal along [111]. For the latter result, we introduce the minimal set of terms, quartic in momentum, in
Eq. (9), that are needed to accurately reproduce the spin dephasing time T2∗ shown in Fig. 3 of the main text. In addition,
we demonstrate the accuracy of our minimal model by comparing its results with the ones obtained from the more general
extended model, Eq. (4), used in the main text for both considered growth directions.

Coupling strengths Band parameters Band structure

κ41 -0.15 nm2 P 0.871 eVnm
q41 0.71 nm2 Q 0.750 eVnm
κ42 79.81 nm2 E0 4.185 eV
q42 -32.46 nm2 ∆0 0.044 eV
0
κ43 -78.46 nm2 E0 3.4 eV
q43 31.05 nm2 ∆00 0 eV
Γ53 -0.71 nm2
Γ32 456.99 nm4
Γ12 939.59 nm4

TABLE I. Prefactors (left box) as calculated in this work. The band parameters (middle box), defined by the schematic band structure
(right box), are taken from Tables II and III of Ref. [81]. The energy splittings and the matrix elements coupling different bands
around the high-symmetry point Γ are shown by the red and blue lines, respectively.

The device growth orientation directly impacts the effective masses for holes occupying the heavy and light hole bands.
Table II summarizes the properties of the states entering our perturbation terms. In the following we demonstrate the

growth direction hole character total spin mass in the plane mass along z
LH or HH Jz m0 /mxy m0 /mz
[001] LH ±1/2 γ1 − γ2 γ1 + 2γ2
[001] HH ±3/2 γ1 + γ2 γ1 − 2γ2
[111] LH ±1/2 γ1 − γ3 γ1 + 2γ3
[111] HH ±3/2 γ1 + γ3 γ1 − 2γ3

TABLE II. The effective masses for the HH and LH states in devices with [001] and [111] growth directions, which are needed in the
numerical calculation.

accuracy of the minimal model proposed in the main text, and show that an analogous model is available for devices with
a dot-normal along the [111] crystallographic axis. The analogous model comprises the following set of terms, and those
contained within the Luttinger Hamiltonian HL , and describes accurately the spin dephasing behaviour, as shown in Fig. 4:
 8

π π
2 2
a) b) 5000
π n̂ = [001] n̂ = [111] π
4 Hmin Hmin 4

1000
0 0

β
− π4 − π4 500

T2∗ [µs]
− π2 − π2
π π
2 Ez [mV/nm] Ez [mV/nm] 2 250
c) d)
π n̂ = [001] n̂ = [111] π
4 HL + δHL HL + δHL 4
100
0 0
β

β
50
− π4 − π4

− π2 − π2 0
1 5 10 15 20 1 5 10 15 20
Ez [mV/nm] Ez [mV/nm]

FIG. 4. Estimated dephasing time T2∗ (color coded) for a single hole spin qubit in a planar quantum dot grown along n̂ = [001] and
n̂ = [111] when using the minimal Hamiltonian Hmin (top) and full Hamiltonian HL + δHL (bottom). The minimal Hamiltonian Hmin
contains only the subset of terms described in the main text for [001]-devices and Eq. (9) for [111]-devices. The dot axes are aligned
with the crystallographic axes, the magnetic field is in-plane with the magnitude B = 1T and direction β, and the vertical electric field
is Ez . The same parameters for the dot shape and charge noise are used as in Fig. 3.

2
H42 =µB (κ42 J + q42 J3 ) · (BX kX , BY kY2 , BZ kZ
2
),
H43 =µB (κ43 J + q43 J3 ) · (kX (kY BY + kZ BZ ), c.p.),
2 2 (9)
H32 =Γ32 J32 · (2kX kY − (kY2 + kX
2 2
)kZ , (kY2 − kX
2 2
)kZ ).
2
H51 =Γ51 ({kX , kY , kZ }JY Z + {kY2 , kZ , kX }JZX + {kZ
2
, kX , kY }JXY ),

where {a, b, c} = (abc + acb + cba + bca + bac + bca)/6 is used and the prefactor Γ51 = −1804.6 nm4 is introduced. In the
main text we briefly mention the difference between the in-plane and out-of-plane components of the g-tensor. In Fig. 5 we
show that the out-of-plane components are in general much larger than the in-plane ones and do not change significantly
when varying the dot parameters.
 9

ε̄[meV] η
−3.0 −2.5 −2.0 −1.5 −1.0 0.0 0.5 1.0
4 4
a) b)
3 3
2 2
1 1
0 0
4 4
c) d)
3 3
2 2

1 1

0 0

0 π/4 π/2 3π/4 π 1 4 7 10 13 16 19

δ Ez [mV/nm]

FIG. 5. Components of the heavy-hole qubit g-tensor. The colored lines show ĝxx (red), ĝxy (blue), ĝyx (green), ĝyy (yellow) and ĝzz
(black). The x-axis represents (a) the dot size ε̄ = (εx + εy )/2, (b) the dot asymmetry η = (εy − εx )/2ε̄, (c) the dot orientation δ, and
(d) the vertical confinement electric field Ez . The dashed vertical line in each panel, defined by εx = −1 meV, εy = −3 meV, δ = π/4 ,
and Ez = 10 mV/nm, denotes a common reference point.