WorksPhysics and Astronomy
The importance of shear strain in narrow Si/SiGe quantum wells
Abstract
Several heterostructure profiles have been proposed for Si/SiGe quantum wells, to improve the operation of quantum-dot spin qubits. These structures share a common goal of generating either of the two wave vectors that govern the conduction-band valley coupling, dubbed and . Here, we focus on the long-wavelength term that couples valleys when paired with shear strain, and which governs the Dresselhaus spin-orbit coupling. We show that shear strain induced by metal top gates may be enough to activate the valley-coupling mechanism, mitigating the need for "ultrasharp" interfaces, as commonly assumed for large valley energy splittings. We propose a scheme featuring narrow quantum wells that does not require ultrasharp interfaces, but should still provide large deterministic valley splittings (when paired with shear strain), and large spin-orbit couplings and -factor corrections, which are of great interest for experimental implementations.
Recent years have witnessed the rapid expansion of quantum dots in large one- [2, 3] [1] and two-dimensional spin-qubit arrays [4, 5, 6], by leveraging advances in the semiconductor industry. Underlying this progress is the ability to operate devices at uniform voltage settings [7], easing the burden of tune-up and large-scale operation. Despite these successes, materials variability remains a key challenge across solid-state qubit platforms [8]. For quantum dots formed in silicon-germanium quantum wells, the random nature of the SiGe alloy is a persistent and fundamental source of variability [9, 10], as reflected in local fluctuations of energy splittings between the conduction-band valley eigenstates [11, 12, 13], their complex phases [14, 15], the spin-orbit coupling [16, 17, 18, 19], and the Landé -factor [20, 21].
Historically, the main focus of valley splitting experiments has been on extracting mean values [22]. However, recent theory work highlights the role of alloy disorder and the importance of characterizing valley splitting distributions, which are known to take a Rician form in the most general case [10, 23]. In recent years, systematic measurements of the valley splitting (e.g., across dot arrays or shuttlers [13], or in devices where dots are continuously positioned between plunger gates [24]) point to disorder-dominated behavior, which is well-described by the limiting Rayleigh distribution. Consequently, a significant fraction of dots will have low valley splittings, rendering such devices inoperable, as the spin states hybridize with excited valley states lying outside the logical subspace.
To overcome these challenges, several schemes have been proposed to deterministically enhance the valley splitting [10]. Some recent experiments also point to possible deterministic enhancement [25] [26], although the source of such behavior is unclear. On the other hand, all confinement schemes experience disorder-induced fluctuations of the valley splitting, due to the overlap of electron wavefunctions with disordered SiGe barriers. In fact, some schemes for enhancing the valley splitting require a significant Ge overlap. At the same time, the disorder also leads to spatially varying potential-energy landscapes that affect the shape and location of a dot, the tunnel coupling between dots, and even the formation of spurious dots. Optimal solutions to the valley splitting problem should therefore weigh the potential benefits from deterministic enhancement against complications arising from disorder.
The most straightforward scheme for coupling valleys requires engineering a quantum well confinement potential with a strong Fourier component at the special wave vector, , that directly connects the conduction valleys on opposite sides of the Brillouin zone [12], as illustrated in Figure 1a. Here, , where nm is the cubic lattice constant of Si, shown in Figure 1b. Heterostructures capable of providing such large wave vectors include the short-period Wiggle Well [27, 28] and quantum wells with ultrasharp interfaces [11, 29]. Unfortunately, the characteristic size of these features, nm, is so small that they are challenging to obtain by standard industrial techniques like chemical vapor deposition (CVD) [30]. (For recent progress based on molecular beam epitaxy, see [31].)

Figure 1. Using shear strain and small feature sizes to enhance the valley splitting. a Valley coupling requires a wave vector of length (red arrow), as illustrated in the band structure of the first Brillouin zone (BZ) for Si. Shear strain provides an alternative (second-order) scheme for coupling the valleys, by combining a lattice deformation of wave vector (orange arrow) with an engineered feature in the quantum well confinement potential of wave vector . Here, is the size of the Si cubic unit cell, and the required feature size is nm. b The Si cubic unit cell comprises a diamond lattice, with its two sublattices highlighted in red and blue. Shear strain can be achieved by stretching along and squeezing along (green arrows). As a result, the blue sublattice shifts toward (gray arrows), as shown on the right, yielding a lattice deformation of period , with a corresponding reciprocal wave vector of . c In this work, we consider a narrow quantum well defined by the sigmoid function of Eq. (6), with a quantum well width and/or an interface width comparable to . Here, we show an example wave function (red dashed line) computed for a quantum well defined by ML nm and ML nm, in a quantum well with a band offset of 180 meV and a vertical electric field of 2 mV/nm. Other types of engineered features in could also provide the required wave vector, including a Wiggle Well or a simple “bump.”
What then is the source of the persistently high valley splittings observed in some experiments? A recent proposal suggests that the strict requirement of ultrasharp () features can be relaxed in the presence of shear strain [32]. In this case, a much smaller wave vector is needed to couple the valleys, via the second-order process shown in Figure 1a (and explained below). The characteristic feature size for this coupling, nm, is fully compatible with current CVD growth techniques [27].
In this work, we show that narrow quantum wells can be used to deterministically enhance the valley splitting in the presence of shear strain. We also show that the quantum dot devices recently reported in [25] are well-suited for inducing shear strain at the levels needed for generating consistently high valley splittings, and we speculate that devices with micromagnets could further enhance this effect. In fact, we find that narrow quantum wells with shear strain can successfully compete with Wiggle Wells, which were specifically designed for this purpose [32]. We then show that the 3 nm quantum wells used in recent experiments [29, 33] could activate a combination of and -based enhancement mechanisms, and we conclude that, since such strains are present, they should certainly be taken into account. Finally, we note that the same conditions that enable high valley splittings can also enhance the Dresselhaus spin-orbit coupling and the Landé -factor.
Results
Shear strain
We recall that shear strain corresponds to the off-diagonal component of the strain tensor, . As an example, we consider a lattice distortion oriented along the crystallographic axis, whose effect on the Si diamond crystal lattice is illustrated in Figure 1b. Here, the shear strain is introduced by stretching along the axis while squeezing along , as indicated by the green arrows. The effect of these stresses is to displace the diamond crystal sublattices (indicated by blue and red) in opposite directions, causing a bilayer pairing of the atomic planes along [001] (gray arrows), as illustrated on the right-hand side of the panel. The resulting periodicity along is , with a corresponding wave vector of , which couples points across the Brillouin zone, as indicated in Figure 1a. A second-order process involving the wave vector then completes an effective coupling between the two valleys. (See [32] for additional details.)
An important consideration for the valley-coupling mechanism is the direction of the applied stress.
Although shear strain can be defined in any coordinate system, valley coupling requires shear strain defined with respect to the crystallographic axes of the cubic unit cell, as illustrated in Figure 1b. Such strain occurs naturally in many device settings, due to the presence of defects in the crystal lattice, including cross-hatch, whose features are naturally oriented along [110] and [10] [34]. Shear strain can also be engineered: for example, in [32], etched trenches were found to cause shear strain, while in [35], shear strain was generated by stressors fabricated on top. of the gates. In [36], very large shear strains of order were observed below a fabricated micromagnet.
In this work, we focus on strains arising from metal electrodes [37], noting that significant shear-strain components can be achieved when the devices are aligned with respect to [110] or [1\bar{1}0]. As an important example, we consider the device described in [25], which exhibits an unexpectedly high valley splitting. This industrially fabricated device incorporates gate structures in the form of vertical vias, composed predominantly of W. Such tall gate structures are expected to enhance the strain, as compared to the shorter, planar gates, commonly preferred in academic settings. Additionally, some of the gates have high aspect ratios, which we also expect to enhance the shear strain [35]. Since detailed specifications for the heterostructure and the gate stack in [25] have not been publicly disclosed, we adopt the details that have been provided, while filling in the missing information with parameters reported for older devices [29, 33].
A cross-sectional view of the device assumed in our simulations is shown in Figure 2a (not drawn to scale), with parameters described in the caption and Methods.

Figure 2. Shear strain simulations. a A cross-section view of the simulated device in the - plane (not drawn to scale), where is defined along the [110] direction. From bottom to top, the simulations include a SiGe virtual substrate (dark blue), a Si quantum well (light blue), a SiGe spacer layer (dark blue), an insulating AlO layer (medium gray), and W metal gates (dark gray), surrounded by SiO oxide (light gray), with corresponding thicknesses of 2 m, 3 nm, 30 nm, 5 nm, 300 nm, and 300 nm. b A simulated shear-strain color map for the gate geometry outlined in orange. (The three gates outlined in pink are referenced in other figures, below.) The gate structure is based on [25]. As in [25], the gate pitch is 70 nm, while the spacing between the gates is chosen to be 35 nm. The barrier gates outlined on the top and bottom of the figure are assumed to extend to the distant edges of the simulated device (see Methods and Supplementary Note S1). c The dashed pink curve shows a linecut of the shear strain in b, evaluated along the horizontal dashed line. The dashed blue curve shows a similar linecut obtained from Supplementary Fig. S1d. These two results are obtained for different depositional stress values for W, taken from the literature (see Methods). d Strain-induced conduction-band energy fluctuations for the same linecuts shown in c. For visual clarity, these curves are shifted vertically, so the regions between the gates have an average value of .
Typical results from the shear-strain simulations are presented in Figure 2b. Here, the direction is defined as the characteristic axis of the device geometry, which we assume to be along [110]. In this work, we consider two different values for the W depositional stress as reported in the literature (see Methods). Figure 2b shows results for the more moderate stress value of 1.12 GPa [38], while Supplementary Fig. S1b also shows results for a larger stress value of 4 GPa [39]. (Figure S1 also shows results for a device oriented along [100].) The gate structures, based on [25], are outlined in orange or pink in Figure 2b. In Figure 2c, we present a more detailed view of the simulation results, for both depositional stress values, evaluated along the linecut shown in Figure 2b (horizontal dashed line). A key result of these simulations is that the absolute value of the shear strain can range from 0.15-0.5% under every plunger gate, depending on the metal deposition method used. Comparing these results to the imposed stress conditions, we infer that depositional stress is the dominant source of shear strain for the blue curve, but that thermal contractions play a significant role for the pink curve. (See [36] for a related discussion.) We note that the strain approaches zero, as expected, far away from this central region of the device (Supplementary Fig. S1). We also note that the overall trends of the shear strain can be explained by arguments like those presented in [35], where the “outer field gates” at the top and bottom of Figure 2b play the same role as the stressors in [35].

Figure S1. Shear strain data for varying gate spacings, gate height, and depositional stress. a,b The average shear strain is studied for different gate geometries, as described in Note S1. Results are plotted as a function of a the gate spacing, and b the gate height. Two depositional stress values are used for the W gates: 1.12 GPa (pink) and 4 GPa (blue). a Here we assume a fixed gate height of 300 nm. Square markers indicate gate spacings of 35 nm. b Here we assume a fixed gate spacing of 35 nm. Square markers indicate gate heights of 300 nm. c,d Shear-strain maps obtained for a geometry with a 35 nm gate spacing and a 300 nm gate height. The W depositional stress values in these two simulations are given by c 1.12 GPa, and d 4 GPa. e The dashed pink and blue curves show linecuts obtained in c and d, under the pink and blue dashed lines. The dashed lines are obtained for devices oriented along the [110] crystallographic axis. The solid lines correspond to the same linecuts, again through the plunger gates, obtained for a device oriented along the [100] crystallographic axis.
In the discussion that follows, we show that the strain levels in Figure 2b and Supplementary Fig. S1 can provide consistently large valley splittings, depending on the details of the heterostructure. In Figure 2d, we also report the potential energy fluctuations caused by the diagonal elements of the strain tensor, which strongly modify the confinement potential over distances comparable to the gate pitch. These fluctuations can have deep minima below the plunger gates and large barriers between the plungers, suggesting that strong electrostatic potentials could be needed to overcome these effects. In contrast, the shear-strain variations shown in Figure 2c are relatively smooth below the plunger gates.
Deterministic valley splitting
We now compute the valley splitting using effective-mass methods. The theory, expressed in Eq. (7) of Methods, includes both and terms, but allows these effects to be explored independently. As discussed above, valley splitting is sensitive to the shape of the quantum well confinement potential , especially features containing or wave vectors. Here, we assume sigmoidal confinement potentials (see Methods), characterized by the quantum-well width and the interface width , as illustrated in Figure 1c. The sigmoidal shape is physically realistic [40] and, importantly, it avoids introducing sharp corners into the confinement potential, which artificially generate and wave vector components [10]. In the simulations below, we consider a range of shear-strain values, , as consistent with the strain simulations reported in Figure 2. We note that the results in this section do not include disorder, and are therefore deterministic. Disorder effects will be introduced in a later section. In Figure 3a-Figure 3d, we present valley splitting results as a function of , , and , while holding a different parameter constant in each plot. Throughout this work, we also assume a uniform, vertical electric field of mV/nm, which forces electrons against the top quantum-well interface. In Figure 3a, we set ML nm, where ML refers to the thickness of a single atomic monolayer (1 ML nm). On the far-left side of the map, we observe high valley splittings with no dependence on shear strain; this behavior arises from the direct valley coupling caused by ultrasharp interfaces with nm. In the rest of the map, we observe -dominated behavior, which depends approximately linearly on . The two regimes are separated by a knee-like feature at . In Supplementary Note S2, we confirm this crossover interpretation, from to -dominated behavior, by turning these terms on and off in the Hamiltonian. Next, in Figure 3b, we set ML nm, which does not qualify as an ultrasharp interface; this map, therefore, does not exhibit any substantial -type behavior. However, it does display interference fringes occurring when the wave function interacts with both the top and bottom quantum-well interfaces, analogous to interference fringes observed previously in the regime [11]. In Figure 3b, the fringes are separated by nm – the characteristic length scale associated with -type behavior. Finally, in Figure 3c and Figure 3d, we fix the values of . In Figure 3c we set , corresponding to the pink curve in Figure 2c, while in Figure 3d, we set , corresponding to the blue curve in Figure 2c. In both cases, we observe very strong enhancements of the valley splitting on the lower-left region of the map, induced by the mechanism, when either or approaches 1.7 nm. We also observe a series of interference fringes at the bottom of both maps, corresponding to interference effects; in this case, the distance between the fringes is closer to 0.16 nm, characteristic length scale for -type behavior. As a reference, in Figure 3c and Figure 3d, we highlight the quantum-well parameters nm, 2.04 nm) with stars, corresponding to deterministic valley splittings 0.33 and 0.88 meV, respectively. The same star locations are also shown in Figure 3a and Figure 3b.

Figure 3. Deterministic valley splitting results, based on effective-mass simulations, in the absence of random-alloy disorder. a Color map of the valley splitting , as a function of the interface width and shear strain , for a fixed quantum-well width of ML = 2.04 nm. ( is indicated in units of nm and ML.) In all panels, the stars identify the parameters for two cases: (white star) and (green star), corresponding to the average shear strains beneath the plunger gates, for the pink and blue curves in Figure 2c. b Color map of the valley splitting as a function of and , for the fixed value ML = 0.54 nm, revealing an interference pattern governed by the valley-coupling wave vector . ( is indicated in units of nm and ML.) c, d Color maps of the valley splitting as a function of and , for fixed values of c , or d . (Also see Supplementary Fig. S3.) e, f Spatial maps of the deterministic valley splitting for the fixed values ML and ML, using shear-strain maps from e Figure 2b, or f Supplementary Fig. S1d. Here, the colors of the gates are the same as the highlighted gates in those figures.
The stars in Figure 3a-Figure 3d represent desirable working points, owing to their large valley splitting values. Figure 3b reveals additional oscillation fringes that could serve as working points. For example, the second fringe is centered at a convenient quantum-well width of nm, which has already been achieved in recent experiments. The fringes are also observed in Figure 3, and particularly, in Figure 3d, where the and fringes work together to give higher valley splitting values. The simulations therefore reveal a potential interplay between and phenomena that is experimentally accessible. Figure 3e and Figure 3f show spatial maps of the deterministic valley splitting, computed with the strain profiles of Figure 2 and Supplementary Fig. S1, for the three gates highlighted in each of those figures. Since the simulations in this section do not include disorder, the variations observed here are attributed strictly to gate-induced strain variations. Summarizing these results: we have demonstrated the presence of consistently large valley splittings in the regime, which are accessible to current experiments. For sharper interfaces, behavior may also emerge, further enhancing the valley splitting.
L. As expected, we observe large enhancements for narrow quantum wells and sharp interfaces. For example, at the starred location in Figure 4a (the same starred location as Figure 3c), we obtain meV·nm. We can compare this to a conventional quantum well (e.g., with nm and nm), for which we find eV·nm, representing an enhancement factor of 65 for the narrow well. Due to their shared intervalley origins, the spin-orbit coupling shown in Figure 4a resembles the valley splitting distribution in Figure 3c, except for the absence of features at low , since the wave vector plays no (direct) role in the spin-orbit coupling. The results also do not depend on the shear strain, to leading order. (At higher order, small strain and -based corrections can occur, since these parameters enter the Hamiltonian and affect .) In Supplementary Note S3, we show that and in the -dominated regime, to a very good approximation.

Figure 4. Dependence of the spin-orbit coupling, -factor, and disorder parameters on the shape of the quantum well. a Color map of the Dresselhaus spin–orbit coupling strength , plotted as a function of quantum-well parameters and . b Color map of the -factor corrections (see Methods), as a function of and . c Color map of the standard deviation of the valley coupling , as a function of and . Dashed, dot-dashed, and dotted lines indicated the nominal crossover to deterministically enhanced valley splitting behavior (), for the cases , , and , respectively. d Color map of the deterministic-crossover ratio , as a function of and , for the fixed value . Dotted, dot-dashed, and dashed lines correspond to the contours , , and , respectively. (Also see Supplementary Fig. S5.)
Spin-orbit coupling and -factor
The dominant spin-orbit coupling in the Si/SiGe system takes the Dresselhaus form, proportional to , where , and are Pauli matrices acting on the electron spin. Since this coupling is an intervalley process [41], it shares some common features with the valley splitting. Unlike the valley splitting, however, the Dresselhaus coupling has a built-in wave vector of , and therefore does not require any additional shear strain [41]. In fact, the Dresselhaus coupling mechanism is described by the same second-order process illustrated in Figure 1a; so, it also needs a confinement potential with a wave vector. As discussed in Methods, the strength of the Dresselhaus coupling in the ground subband is defined as , where meV nm is the bare Dresselhaus coupling coefficient for Si. The ground subband envelope function, , is the same envelope used in valley splitting calculations; it contains spatial fluctuations that can induce spin-orbit coupling if they contain a wave vector, inherited from the quantum well confinement potential . fig:4a shows results for as a function of and
The spin-orbit coupling also contributes to the Landé -factor [20], resulting in a correction to the bare coupling , proportional to (see Methods). Here, is the magnitude of the correction, while is its complex phase and is the phase of the intervalley coupling (known as the “valley phase”). In Figure 4b, we plot as a function of and , also showing strong enhancement when these quantities approach nm. As in Figure 4a, we observed no or dependence here, to leading order. For more-conventional quantum wells, corresponding to larger and values, is of order
[21]. However, the enhancement mechanism studied here yields much larger values; for example, at the starred location in Figure 4b, we obtain .
Spatial variations due to the SiGe random alloy
Due to their small size, electrons in quantum dots are exposed to a limited number of Ge atoms within a Si/SiGe quantum well. In some structures, the Ge atoms are concentrated in the barrier regions, where they mainly contact the tails of the electron wave functions [9]. In other cases, Ge is intentionally added to the quantum well, where the electron density is highest, in the form of spikes [42], Wiggle Wells [27, 31], or a uniform doping layer [24]. Since the SiGe alloy is random, the Ge atoms are located at random sites throughout the SiGe crystal lattice, causing the electrons to sample random, local variations of the Ge density. Moreover, since Ge and Si atoms have different on-site potentials and bond lengths, this variability also leads to spatial fluctuations of the quantum dot confinement potential, valley splitting, and other local properties [10, 13, 21, 43]. The spatial variations of the intervalley coupling are known to follow a Gaussian distribution characterized by its standard deviation [10], which depends on the laterally-averaged Ge spatial profile of the quantum well, in the form of a weighted average:
Throughout this work, we set nm as the in-plane confinement length for a quantum dot, corresponding to a typical orbital confinement energy of meV. The spatial variations of the intravalley coupling (i.e., the potential energy landscape) also follow a Gaussian distribution with a standard deviation of [44]. In Figure 4c, we plot as a function of the quantum-well parameters and . This confirms our expectation that smaller and larger should both cause the electron wave function to overlap with more Ge, thereby increasing the alloy disorder felt by the electron, and . Indeed, we find that depends roughly linearly on both and .
Previously, in Figure 3, we studied the deterministic component of the valley splitting, obtained in the virtual crystal approximation, which does not include alloy disorder. Here, we explore this disorder and the emergence of deterministically enhanced valley splitting. We consider the same quantum-well profile and strain maps as Figure 3e and Figure 3f. However, we now include intervalley disorder potentials using the statistical methods of Refs. [10, 20] (see Supplementary Note S4). In Figure 5a, we first show a spatial map of the valley splitting for ML, including disorder but no shear strain – in this case, effects are not activated. (Compare to Figure 3e and Figure 3f, which include strain effects but no disorder.) We note that and are also not small enough to activate behavior; indeed, the contribution to the deterministic valley splitting is only 23.4 eV here (see Supplementary Figure S2). In this case, since the quantum well is narrow, the electron wave function overlaps with many Ge atoms, endowing the valley splitting with a large standard deviation ( eV) and a large disorder-enhanced mean value ( meV). However, many dot locations still have low valley splittings, as indicated by dark blue in Figure 5a. Figure 5b shows a second valley splitting map, with the same disorder profile as Figure 5a, but now including the shear strain from Figure 2b. While these results still appear disordered, the average valley coupling has been significantly enhanced by the coupling mechanism. In particular, the regions of very low valley splittings are suppressed. We also note that the gate structure is not clearly visible in the valley splitting map, since the shear strain is largely uniform across this window.

Figure 5. Comparison of disorder-dominated vs deterministically enhanced behaviors. a–f Color maps on the left are obtained with shear strain turned off, , while maps on the right are obtained using the strains computed in Figure 2b. All maps are computed using the same random-alloy disorder realization. The outlined gates refer to the same outlined gates as Figure 3e and Figure 3f: on the right-hand side, the pink gates generate the same shear strains as the pink gates in Figure 3e; however, on the left-hand side, the dashed green outlines are only shown for reference, since we set . We assume a sigmoid quantum-well profile, defined by ML and ML. a, b Valley-splitting maps showing a disorder-dominated behavior, or b deterministic enhancement. c, d Valley-phase maps showing c disorder-dominated behavior, or d deterministic enhancement. Note that valley vortices are observed in c but not in d. e, f Maps of the fluctuating -factor, , showing e disorder-dominated behavior, or f deterministic enhancement. g–i Sample-wide distributions of the valley splitting, h the valley phase, and i the -factor fluctuations, obtained from independently sampled disorder realizations, assuming spatially uniform values: green corresponds to (disorder-dominated), pink corresponds to (deterministically enhanced), blue corresponds to (strongly enhanced). Each distribution is overlaid with its theoretical probability distribution function (solid, dashed, and dotted lines; see Supplementary Note S5). In g, deterministic enhancement causes a suppression of low valley splitting values, while in h and i, it causes a narrowing of the distributions.

Figure S2. Individual contributions to the deterministic valley splitting from the and valley-coupling mechanisms. This figure extends our understanding of Figure 3a of the main text by separating the full results into their individual components. a The contribution to the valley splitting from the coupling mechanism of Eq. (11). This mechanism arises from ultrasharp features in the confinement potential, such as very small interface widths . b The contribution to the valley splitting from the coupling mechanism of Eq. (12). This mechanism is proportional to the shear strain. c The total valley splitting , which combines the and mechanisms. This panel is reproduced from Figure 3a.
The crossover to deterministically enhanced behavior depends on two key parameters [10]: the deterministic valley coupling, , representing the center of the valley-coupling distribution in the complex plane, and the standard deviation of this distribution, . We note that is also the valley coupling that could be observed in the absence of alloy disorder. In turn, these quantities are determined by the quantum-well confinement profile , which is averaged across the heterostructure, and by the vertical electric field and shear strain , which are locally varying. The term deterministically enhanced valley splitting refers to cases where the valley-coupling distribution is centered far enough from zero that dangerously small valley splittings are appreciably suppressed. The crossover to this regime occurs when [10] , although strong suppression of small valley-splitting values occurs when this ratio takes a larger value, in the range of -- (see Supplementary Figure S7).

Figure S7. Evolution of three different distribution functions, for several values of the disorder parameter . Since the nominal crossover to deterministic enhancement occurs when , we consider here the cases of , , , and (left to right), for three different distribution functions (top to bottom), where we set meV. In all plots, the colored regions correspond to histograms obtained from independently sampled disorder realizations, assuming meV. (Note that the colors used in this figure are not intended to match colors used elsewhere in this work.) a-d Valley-splitting distributions. Here, the solid-black lines correspond to the Rician distribution of Eq. (20). The vertical dashed lines correspond to the deterministic valley splitting, . Note that, while a has no deterministic valley coupling (), it is still difficult to distinguish from significant deterministic enhancement, corresponding to in b. Also note the exponential suppression of the lowest values in d. e-h Valley-phase distributions. Here, the solid-black lines correspond to Eq. (21). The vertical dashed lines correspond to the deterministic valley phase, . In this case, the deterministically enhanced distributions are much easier to verify, because they are strongly narrowed. i-l -factor distributions. Here, the solid-black lines correspond to Eq. (22).
In Figure 4d, we plot the crossover ratio as a function of the quantum-well parameters and , revealing interference patterns reminiscent of Figure 3c, with two distinct length scales arising from the different and valley-coupling mechanisms. The dashed black lines in Figure 4d highlight several crossover criteria for the deterministic enhancement, from “normal” enhancement to very strong enhancement; by comparing these curves to Figure 3c, we note that deterministic enhancement can occur even in cases where the valley splitting is not especially high ( eV). The white dashed lines in Figure 4c also indicate contours of deterministic enhancement – in this case, revealing that significant disorder can accompany high valley splittings. The valley-splitting maps in Figure 5a and Figure 5b exemplify both sides of the deterministic crossover, with (Figure 5a) corresponding to “randomly dominated” behavior and (Figure 5b) corresponding to “deterministically enhanced” behavior. These results confirm that the coupling mechanism can boost the valley splitting well into the deterministically enhanced regime. The sample-wide valley-splitting distributions for these difference cases are shown in Figure 5g (also see Supplementary Figs. S7a-d). In particular, for the shear-strain level corresponding to the blue line in Figure 2c, we observe a very strong suppression of low valley-splitting values. Since for -based valley coupling, this suggests a clear path for improving the valley splitting: by increasing the shear strain.
The valley phase also provides a key signature of deterministic behavior [14], although measuring it requires special experimental techniques [15, 21]. In Figure 5c and Figure 5d, we plot spatial maps of the valley phase, obtained for the same device parameters and disorder profiles as the valley-splitting maps in Figure 5a and Figure 5b. For the case of no shear strain (Figure 5c), we observe a full range of valley-phase values, , consistent with the center of the complex distribution lying near the origin. In Figure 5d, on the other hand, we observe a much narrower range of valley phases, consistent with the distribution shifting away from the origin. The sample-wide distribution functions for these valley-phase maps are shown in Figure 5h, for the same three parameter profiles as Figure 5g. (Note that the valley phases here and in Figure 5c and Figure 5d have been reported relative to .) It is clear that the narrowing of the distributions provides a definitive signature of deterministic enhancement [14]. For example, while the pink distribution in Figure 5g looks similar to the randomly dominated distribution in green, the pink distribution in Figure 5h shows clear deterministic enhancement.
Valley-phase spatial maps like Figure 5c and Figure 5d also display other interesting features. For example, the deterministically enhanced map (Figure 5d) contains no valley vortices, which can be identified as singularities with , around which the phase winds by . However, we contrast this with the disorder-dominated map (Figure 5c), which is obtained from the same disorder profile, but exhibits four vortices across the same window. The difference between these behaviors is also a consequence of deterministic enhancement, reflecting the suppression of valley-splitting zeros.
Finally, we note that Ge concentration fluctuations are atomistic by definition, giving rise to effects, which have a nearly atomistic length scale of 0.16 nm. However, such fluctuations are largely smoothed out over the much larger length scale of 1.7 nm. Consequently, quantities like the spin-orbit coupling () and , which are enhanced by features (not ), are largely robust to atomistic Ge fluctuations. In contrast, -factor fluctuations arise from the quantity , and are therefore susceptible to Ge fluctuations through the valley phase , resulting in spatial variations. In Figure 5e and Figure 5f, we plot -factor maps for the same parameter profiles as the valley-splitting and valley-phase maps in Figure 5a-Figure 5d. These results mirror the valley phase map. For example, we observe a significant narrowing of the -factor distribution for the deterministically enhanced case (Figure 5f), as compared to the disorder-dominated case (Figure 5e). Also analogous to Figure 5c and Figure 5d, we observe valley vortices in Figure 5e, but not Figure 5f. The narrowing of the distribution is more fully illustrated in Figure 5i, where we plot sample-wide distributions of the -factor, obtained for the same three parameter profiles as Figure 5g and Figure 5h.
Discussion
In summary, we have shown that a long-held assumption – that “ultrasharp” quantum-well interfaces (defined to be narrower than nm) are needed to obtain large, deterministic valley splittings – can be relaxed in the presence of shear strain. In this scenario, a much larger (and experimentally favorable) value of nm emerges as the characteristic length scale for valley coupling. Using a combination of simulation techniques, we have shown that gate geometries featured in recent experiments [25] could provide enough shear strain to generate large deterministic enhancements of the valley splitting, when coupled with sufficiently narrow quantum wells [29, 33], and that these same heterostructures should strongly enhance the Dresselhaus spin-orbit coupling and -factor fluctuations. We also speculate that wider quantum wells with internal features like steps, wiggles, or spikes could provide the required wave vector, and that micromagnets could alternatively provide the needed shear strain [36].
We conclude with some final remarks on -based coupling. First, we emphasize that the relatively large length scale is essentially immune to atomic-scale disorder at the quantum-well interface. This should be contrasted with the coupling mechanism, which is highly sensitive to single-atom steps [10]. Second, as strain engineering has only recently emerged as an active topic for quantum-dot qubits, further experiments are needed to characterize and exploit its effects. Indeed, our results indicate that shear strain could be present in many device settings, and may also arise from anomalous sources, such as cross-hatch phenomena. Here, we suggest that narrow quantum wells could provide a convenient platform for harnessing these effects. Conversely, we argue that experiments encountering shear strain should account for possible effects when analyzing energy dispersions. Finally, we emphasize that experiments hosting high valley splittings via the mechanism should also experience enhanced Dresselhaus spin-orbit couplings and -factor fluctuations, since these phenomena are governed by the same wave vector.
Acknowledgements
The authors are grateful for many enlightening discussions with E. S. Joseph. We also acknowledge discussions with L.K.M. Vandersypen and the S. Tarucha group, which motivated and clarified the importance of micromagnets for generating shear strain in these devices.
Funding
This research was sponsored in part by the Army Research Office under Award No. W911NF-23-1-0110. T.O. was sponsored by the Netherlands Ministry of Defence under Award No. QuBits R23/009. The views, conclusions, and recommendations contained in this document are those of the authors and are not necessarily endorsed by nor should they be interpreted as representing the official policies, either expressed or implied, of the Army Research Office, the U.S. Government, or the Netherlands Ministry of Defence. The U.S. Government is authorized to reproduce and distribute reprints for U.S. Government purposes notwithstanding any copyright notation herein. The Netherlands Ministry of Defence is also authorized to reproduce and distribute reprints for Government purposes notwithstanding any copyright notation herein.
Competing Interests
M.A.E., B.D.W., and M.F. have applied for a patent on strained Si/SiGe quantum wells. The remaining authors declare no competing interests.
Methods
Shear strain modeling in COMSOL
The strain modeling in this work closely follows the Methods described in [36], which we summarize as follows.
Simulations are performed using the Solid Mechanics module of COMSOL Multiphysics [45]. The simulations require input values for the Young’s modulus , the Poisson’s ratio , and the coefficients of thermal expansion (CTE) , for each material. While all parameters depend on temperature, the dependence of and is relatively weak, so we adopt room-temperature values for these quantities. We define temperature-averaged CTEs as
where and are the initial and final temperatures of the cool-down, respectively. These integrals are approximated using data provided in [46, 47, 48, 49, 50, 51]. We use Vegard’s law to evaluate the CTE of the SiGe alloy. The resulting temperature-averaged CTEs are given by . We use the full, orthotropic compliance tensors for Si and SiGe materials in our COMSOL simulations, as detailed in [36]. Biaxial strain for Si in the quantum well, arising from lattice mismatch with a strain-relaxed barrier, is given by and . This strain is implemented in COMSOL as an initial-strain parameter.
Depositional stress from the fabrication of metal gates is included as an initial-stress parameter that accounts for the effects of thermal contraction as the metal is cooled from depositional temperatures to 293 K. Consistent with the experiments described in [25], we adopt W metal gates for our simulations. We specifically consider two different values of the depositional stress for these W gates: 1.12 GPa [38] and 4 GPa [39], to account for possible variations in fabrication methods. No depositional stress is assumed for due to the wide range of reported values, including sign changes, and their sensitivity to deposition conditions and thermal post-processing [52]. We do not expect this approximation to qualitatively affect the results, but it may cause small quantitative errors. For the same reason we do not include any depositional stress for .
Free boundary conditions [45] are enforced on every surface, except for the bottom, where we enforce a contraction that matches the bulk thermal contraction of Si, which has the effect of emulating the presence of a much larger Si substrate at this boundary. The final results of our strain simulations are evaluated in the plane of the two-dimension electron gas, assumed to lie 1.5 nm below the top interface of the quantum well, where the low-energy electronic wave functions of the quantum well are concentrated. All device dimensions have been thoroughly checked, to avoid finite-size effects in the region of interest.
In the main text, we consider the gate structure whose outlines are shown in Figure 2b, oriented along the [110] crystallographic axis, while in Supplementary Note 1, we also consider a device oriented along [100]. To simplify the simulations in the [110] orientation, we consider a reference frame defined by , , and . In this frame, the device is oriented along . However, the orthotropic compliance tensors must also be rotated by , as described in [36]. As noted in the main text, the key input parameter for valley-coupling calculations is the shear strain evaluated in the crystallographic coordinate system. Thus, we must perform an inverse transformation between these two frames as the final step of the simulation procedure. The desired shear strain is then given by
In addition to valley-splitting energy variations arising from the local shear strain, we also observe variations of the conduction-band offsets arising from the diagonal strain components. (For example, see Figure 2d.) We compute these variations here using the standard relation [37]
where eV and eV are deformation-potential coefficients [53]. We note that Eq. (2) is independent of the coordinate system: . Unless otherwise stated, is shifted in each calculation, such that the asymptotic value of goes to zero, far away from the active region containing the plunger gates.
Effective-mass theory
In this work, we use a one-dimensional effective-mass model to calculate the valley splitting , the intrasubband (ground-state) Dresselhaus spin-orbit coupling coefficient , and the magnitude of the intervalley -factor renormalization parameter . This model is derived from a multiorbital tight-binding theory in Refs. [20, 32], where the two approaches were confirmed to give results in excellent agreement. In the valley basis, the effective-mass Hamiltonian takes the form
where is the intravalley Hamiltonian and contains the intervalley couplings. The intravalley Hamiltonian is given by
where is the longitudinal effective mass. The quantum-well confinement potential includes a uniform vertical electric field , and is given by
Here we adopt the value eV, as consistent with [54]. However, other sources suggest a slightly smaller value of eV [55], which would allow the wave function to penetrate further into the quantum-well barriers. The Ge concentration profile is modeled using smooth sigmoidal interfaces:
where is the quantum-well length, characterizes the interface width, and is the nominal Ge concentration in the barriers. Both and are varied throughout this work.
The intervalley Hamiltonian comprises a spin-conserving term and a spin-orbit contribution:
We adopt the values eV and , derived in [32] [56]. The first term in Eq. (7) describes conventional potential-induced intervalley coupling with momentum transfer , whereas the second describes the intervalley coupling enabled by shear strain. The final term in Eq. (7) describes the intervalley Dresselhaus spin-orbit interaction, where meV nm is the bare Dresselhaus coupling coefficient for Si obtained from a multiband tight-binding theory [20].
We numerically solve Eq. (4) to obtain the envelope functions and energies for each subband , with emphasis on the dominant ground subband, . The spin-conserving intervalley matrix element for (also known as the valley coupling) is given by , and the corresponding valley splitting is given by . We note that the effective-mass Hamiltonian inherits lateral spatial variations from and . The ground-state intrasubband Dresselhaus coefficient is obtained by projecting the third term in Eq. (7) onto the lowest subband:
The magnitude therefore quantifies the effective intrasubband spin-orbit coupling relevant for electrically driven spin rotations (EDSR) [43].
Following Ref. [20], we note that spin-orbit coupling also renormalizes the -factors of the two valley eigenstates. Except for certain anomalous cases, the renormalized -factors are given by
for the ground (excited) valley states, where is the valley phase, is a -factor phase, and is the in-plane angle of the magnetic field, relative to the crystallographic axis. The intervalley -factor renormalization parameter is given by
where is the in-plane orbital confinement energy, which we take to be meV throughout this work, and we define and . Based on these definitions, we note that describes the intrasubband spin-orbit coupling, while arises from virtual transitions involving higher subbands and consequently depends on the intersubband coefficients .
Supplementary information for “The importance of shear strain in narrow Si/SiGe quantum wells”
These Supplementary Notes and Figures provide details of the calculations reported in the main text.
GATE DESIGN SIMULATIONS AND GATE PARAMETER SWEEPS
The gate structure illustrated in Ref. [25], upon which our simulations are modeled, does not specify the full details of the gate geometry. Indeed, Fig. S24 of that work is only schematic, although the plunger gate pitch is specified as 70 nm. Taking this as a starting point, we consider the gate spacing (open area between the gates) to be a free parameter. Moreover, the gates are formed of W and they are described as “tall,” taking the form of vertical vias. We therefore also consider the total gate height as a free parameter. Although many other device parameters could be considered to be variable in this work, we limit our analysis to these two quantities. In this Supplementary Note, we then compute the shear strain as a function of these parameters. Since the depositional strain depends on many details of the fabrication procedure, we consider two values for the depositional stress of W reported in the literature: 1.12 GPa [38] and 4 GPa [39].
We first explore a variable gate spacing, with a fixed value 300 nm for the gate height. In this case, we continuously vary the dimensions of the inner gates while retaining the overall gate pitch. However, we do not vary the very top and very bottom gates, which are called “outer field gates” [25]. Although the detailed shapes of these gates are not described in [25], we assume they extend to the distant edges of the simulation box, since “field gates” refer to large-area gates used to deplete electrons, outside the active region [57]. Our approach to varying the plunger gate shapes can be intuited by comparing the gate structure shown in Figure 2b of the main text with Fig. S24 of Ref. [25]. The results of these simulations are shown in Figure S1, where the average shear strain is evaluated along the dashed lines shown in Figure S1 and Figure S1. The different colors represent the two depositional stress values assumed for W. Here, we observe a monotonic dependence of the shear strain on the gate spacing. We note that the maximum gate spacing for this sweep is equal to the gate pitch (70 nm). We also note that this limit is meaningful, even though it corresponds to a vanishing gate, since it reflects strain arising from just the outer field gates. Since no optimal gate spacing presents itself across the sweep range, we simply adopt the center value of 35 nm for our remaining simulations. Next, results for the gate-height simulation are shown in Figure S1. Here, a similar situation arises, in which no optimal value can be identified. Hence, we settle on the physically reasonable [58] gate height of 300 nm for our remaining simulations.
Figures S1c and d show shear-strain maps based on the gate parameters described in the previous paragraph. Here, Figure S1 corresponds to the smaller-magnitude depositional stress, while Figure S1 corresponds to the larger-magnitude depositional stress. Figure S1 extends the results shown in Figure 2b of the main text over a larger spatial range (and a different color scale), to demonstrate that the shear strain decays to zero as expected, far from the plunger gates. In Figure S1, we plot linecut results of the results obtained along the dashed lines in Figure S1 and Figure S1 (shown as pink and blue dashed lines). All the results discussed so far correspond to devices oriented along the [110] crystallographic direction. It is important to note that the shear strain depends strongly on the device orientation. To highlight this point, in Figure S1, we also report the same linecuts as before (solid lines), obtained along a line of plunger gates, but for a
device oriented along [100]. The resulting shear strain values are very close to zero, demonstrating the importance of device orientation.
Individual Contributions from the and Valley-Coupling Mechanisms
In this Supplementary Note, we study the individual contributions to the valley splitting from the conventional () and the shear-strain-induced () coupling mechanisms, to clarify the behavior observed in Figure 3a of the main text. That figure shows the deterministic valley splitting (i.e., without disorder), as a function of interface width and shear strain , for a quantum well of width ML nm. To distinguish the two mechanisms, we first solve for from Eq. (4) in Methods. We then use this to compute the conventional contribution to the valley coupling, given by
We also compute the shear-strain-induced contribution, given by
We note that the quantities and are both complex, so their relative phases determine whether they interfere constructively or destructively. We emphasize that only the sum is physically accessible; however, theoretically, we may consider and separately. In Figure S2 and Figure S2, we plot the quantities and , respectively, assuming the same simulation parameters as Figure 3a. The contribution (Figure S2) is independent of the shear strain within the present model. In contrast, the contribution (Figure S2) is proportional to , as anticipated from Eq. (12), and vanishes when . The total deterministic valley splitting, shown in Figure 3a, and reproduced in Figure S2, is given by . Again, since these two contributions are complex, this splitting is not generally equal to , since it contains an interference term:
To investigate the onset of -based valley coupling, Figure S3 shows the evolution of for experimentally relevant values of , as a function of quantum-well width and interface width . (Compare to Figure 3c and Figure 3d in the main text.) In the absence of shear strain (Figure S3), enhanced valley splittings are restricted to the ultrasharp-interface regime, defined by nm. Here we observe oscillations with periodicity, consistent with behavior reported by Boykin et al. [11]. With increasing shear strain (Figure S3-Figure S3), the portions of these diagrams displaying large valley splittings grow to include a much wider range of values, and a similarly wide range of values, thereby relaxing the sharp-feature requirement associated with the conventional mechanism. The shear-strain-induced features also show resonances with a characteristic periodicity, providing an additional signature of shear-strain-induced intervalley coupling.

Figure S3. Evolution of the deterministic valley splitting with shear strain. This figure extends the results of Figure 3c and Figure 3d in the main text to a range of shear-strain values, given by: a , b , c , d , e , and f . Dashed lines indicate the contour , which corresponds to the nominal onset of deterministically enhanced valley splitting.
In each panel of Figure S3, the dashed black line denotes the special contour , corresponding to the nominal threshold for deterministically enhanced behavior. However, these contours do not precisely follow the valley splitting contours, which can be understood as follows. Although the deterministic valley coupling is enhanced by small or values, the standard deviation of the valley coupling reflects the electron overlap with Ge atoms, and is therefore enhanced by large or small values. This leads to a competition between and , such that devices with large average valley splittings can still have dangerously low local valley splittings. The ratio is a much better indicator of deterministic behavior than [10]. Importantly, Figure S3 shows that regions above the deterministic threshold (i.e., lying below the dashed lines) are seen to grow with increasing shear strain.
Summarizing this Note: shear strain can drive devices into the deterministically enhanced regime while also redefining what it means to be a “narrow” quantum well or a “sharp” interface.
Approximate proportionality of the valley splitting and spin-orbit coupling in the regime
Here, we examine the approximate proportionality between the valley splitting and the Dresselhaus spin-orbit coupling strength in the regime. When the contribution [Eq. (11)] is negligible, the valley splitting is given by . Moreover, we find numerically that it is dominated by the term. The Dresselhaus coupling , defined in Eq. (8) of Methods, also depends on the Fourier component of the electron probability density. Thus, in the regime, which we define approximately as nm, we expect to observe a direct proportionality between and , given by
Figure S4 illustrates this relationship for the shear-strain values , , and (left to right). Figure S4-Figure S4 show and as functions of the interface width , for the fixed quantum-well width nm. The proportionality suggested by Eq. (14) becomes quite accurate as the relative contribution from increases, particularly for higher shear-strain values and when . Figure S4-Figure S4 show and as functions of , for a fixed interface width nm, again demonstrating the accuracy of Eq. (14) in the regime.

Figure S4. Approximate proportionality between the valley splitting and spin-orbit coupling in the regime. Valley splitting (red dashed curves) and Dresselhaus spin-orbit coupling strength (blue solid curves): a-c plotted as functions of the interface width , for a fixed quantum-well width nm, or d-f plotted as functions of the quantum-well width , for a fixed interface width of nm. The columns correspond to shear-strain values of , 0.2, and 0.4%, from left to right.
Alloy disorder and statistical distributions
The simulations reported in this work incorporate alloy disorder, which plays a key role in determining valley-coupling fluctuations and their resulting landscapes. The theory for these disorder models has been derived elsewhere [9, 10, 20]. For completeness, we summarize the pertinent results here.
Alloy-disorder variations of the valley coupling
We first characterize the variance of the valley coupling, and the crossover to deterministically enhanced behavior, using the effective-mass approach described in Refs. [9, 10, 20, 32]. For random-alloy disorder, every site in the crystal lattice has a probability of being occupied by a Ge atom, where is the layer-averaged Ge concentration and describes the average quantum-well Ge profile. As discussed in Methods, a single Ge atom contributes an additional onsite energy of eV, compared to Si. It is then easy to show that the fluctuations of the potential energy at a given site have zero mean and a local variance of . Expressing the valley coupling as a sum of its deterministic and fluctuating components, , we obtain the mean value and variance of the fluctuations:
Here, the ground-subband envelope function is computed using the average confinement potential, given by [see Eq. (5) of Methods], which also includes an applied, vertical electric field . The effective lattice constants nm and nm reproduce the atomic density of Si, where is the cubic lattice constant. For the transverse confinement, we consider a simple 2D harmonic potential, with a corresponding orbital energy splitting of and an effective radius of , where is the transverse effective mass. Throughout this work, we adopt the typical value meV, corresponding to nm. Many aspects of the valley coupling are determined by . As seen from this discussion, the main experimental knobs for tuning are , , and .
In Figure 4c and Figure 4d of the main text, we characterize the valley-coupling variations by plotting the quantities and as a function of the quantum-well parameters and . In Figure S5, we further explore the evolution of , for a sequence of shear-strain values. Here, we see that the parameter range yielding deterministically enhanced behavior (lying below the dashed black lines) grows quickly with , as expected.

Figure S5. Evolution of the disorder parameter with shear strain. This figure extends the results shown in Figure 4d in the main text, for a range of shear-strain values, given by: a , b , c , d , e , and f . For all panels: the dashed, dot-dashed, and dotted lines correspond to the contours , , and , respectively. (Only the contour is shown in panels a and b.)
Generating spatial maps of the valley splitting
Here, we summarize the methods used to generate spatial valley-splitting maps in this work, which were originally derived in Refs. [10, 20, 43]. We consider isotropic, harmonically confined quantum dots of radius , as described above. The dot centers are located at , and we consider a 2D grid of dot positions , where . Following [20], we decompose the fluctuating component of the valley coupling into its real and imaginary parts: . It can be shown that the covariance matrix describing and is given by
while .
To generate a realization of a spatial map , we first diagonalize the covariance tensor as
where are the eigenvalues of and the columns of are its orthonormal eigenvectors. Next, we generate two vectors and , whose entries are independent, and are drawn from a normalized Gaussian distribution. Spatial maps satisfying the covariance relation of Eq. (16) are then given by
In Figure 5a and Figure 5b of the main text, we plot two different versions of a disordered valley-splitting map, corresponding to cases with and without shear strain, respectively. These two plots differ in several ways, including the fact that they have very different deterministic valley-coupling components. Due to the absence of shear strain in Figure 5a, the deterministic component is rather small: µeV, arising from the confinement potential. In Figure 5b, the shear strain is large and spatially varying, yielding a distinct map. On the other hand, the valley-coupling fluctuations, , are determined by the alloy disorder, and are therefore the same for both cases. In Figure S6, we plot maps obtained for just the component of the valley coupling, for the same disorder realization used in Figure 5, with the deterministic component removed (). Here, we use the values nm and µeV. Figure S6 shows the resulting valley-splitting map, , while Figure S6 describes the corresponding valley phase, .

Figure S6. Spatial maps of the valley splitting and valley phase, obtained in the absence of deterministic valley coupling (). This figure extends the results of Figure 5a and Figure 5b in the main text, which shows two different versions of a disordered valley-coupling map, for cases that include a deterministic valley coupling ( eV), and without shear strain. a Here, the disordered valley splitting is given by . b The corresponding disordered valley phase, .
Statistics of the valley splitting, valley phase, and -factor
In this Note, we summarize or derive the probability distribution functions for three different quantities of interest. As discussed above, the real and imaginary parts of are uncorrelated Gaussian random variables, with zero mean, and a variance of . The conditional probability distribution for is then given by
where the deterministic valley phase is defined by . Equation (19) describes both the amplitude and phase of the valley coupling, which we can explore separately. For example, integrating over gives the well-known Rician distribution for the valley splitting [10, 14], defined as :
where is a modified Bessel function of the first kind. We note that Eq. (20) provides the expression used for the dashed lines in Figure 5g of the main text. In the disordered limit, , Eq. (20) reduces to a Rayleigh distribution.
An important criterion in many valley-splitting discussions is the crossover to deterministic behavior, which corresponds to the condition that the deterministic valley coupling exceeds the mean disordered contribution. This criterion can be derived from Eq. (20) by setting the deterministic component to zero () and computing the mean value , which leads to the desired crossover criterion . We note that this crossover is gradual, with deterministic behavior becoming more pronounced as the ratio increases. This can be seen in Figure S7-Figure S7, where Rician distributions are plotted for several multiples of the crossover criterion: , , , and . In all panels of Figure S7, we assume meV, which characterizes the disorder level for a quantum well with ML and ML, corresponding to the starred parameters in Figure 4c of the main text. In Figure S7 and Figure S7, we observe the emergence of strong deterministic enhancement, for which the distribution function is suppressed for the lowest values. On the other hand, we note the similarity of the distributions in Figure S7 and Figure S7, which makes it difficult to distinguish deterministic from random behavior, for low values.
Equation (20) also determines the distribution of the valley phase, (defined in ), by integrating over , giving
This provides the expression used for the dashed lines in Figure 5h of the main text. In the disordered limit, where , we see that the valley phase becomes uniformly distributed across . However, as the deterministic contribution grows, the distribution becomes progressively localized about . Thus, increasing suppresses not only fluctuations of the valley-splitting magnitude but also the valley phase. This can be seen in Figure S7-Figure S7, where we plot the valley-phase distributions corresponding to Figure S7-Figure S7. In [14], it was argued that, although deterministic valley coupling can be difficult to verify in distributions (e.g., in Figure S7 and Figure S7), the situation is much improved when valley phase distributions are considered (e.g., Figure S7 and Figure S7).
As noted in the main text, fluctuations of the valley phase are also reflected as fluctuations in the -factor. It is therefore important to derive the -factor distribution, since its statistics are relatively easy to gather in experiments.
Setting in Eq. (9), we write for the valley ground state. It is convenient to redefine the phase variables as and , such that . The distribution of has the same form as Eq. (21), with replaced by . We note that all valley phase quantities in this work, including spatial maps and statistical distributions, have been plotted using these primed variables, defined relative to . The probability distribution for the variable can then be solved using standard formulas for transformations of random variables [59]. Here we note that, since the cosine function is two-to-one over the interval , both phase branches must be considered. Defining the variable , we finally obtain
This provides the expression used for the dashed lines in Figure 5i of the main text. It is clear that the statistical properties of the -factor are inherited directly from those of the valley phase. This can be seen in Figure S7-Figure S7, where we plot the -factor distributions corresponding to Figure S7-Figure S7, observing similar narrowing of the distributions.
References
- [2]A. M. J. Zwerver, T. Krähenmann, T. F. Watson, L. Lampert, H. C. George, R. Pillarisetty, S. A. Bojarski, P. Amin, S. V. Amitonov, J. M. Boter, R. Caudillo, D. Correas-Serrano, J. P. Dehollain, G. Droulers, E. M. Henry, R. Kotlyar, M. Lodari, F. Lüthi, D. J. Michalak, B. K. Mueller, S. Neyens, J. Roberts, N. Samkharadze, G. Zheng, O. K. Zietz, G. Scappucci, M. Veldhorst, L. M. K. Vandersypen, and J. S. Clarke, Qubits made by advanced semiconductor manufacturing, Nature Electronics 5, 184 (2022).
- [3]A. Nickl, N. Dumoulin Stuyck, P. Steinacker, J. D. Cifuentes, S. Serrano, M. Feng, E. Vahapoglu, F. E. Hudson, K. W. Chan, S. Kubicek, J. Jussot, Y. Canvel, S. Beyne, Y. Shimura, R. Loo, C. Godfrin, B. Raes, S. Baudot, D. Wan, A. Laucht, C.-H. Yang, W. H. Lim, A. Saraiva, C. C. Escott, K. De Greve, A. S. D Zurak, and T. Tanttu, Eight-qubit operation of a 300 mm SiMOS foundry-fabricated device, Nature Communications 17, 5878 (2026).DOI
- [4]F. K. Unseld, M. Meyer, M. T. Mądzik, F. Borsoi, S. L. de Snoo, S. V. Amitonov, A. Sammak, G. Scappucci, M. Veldhorst, and L. M. K. Vandersypen, A 2D quantum dot array in planar 28Si/SiGe, Applied Physics Letters 123, 084002 (2023).arxiv.org/abs/2305.19681
- [5]S. D. Ha, E. Acuna, K. Raach, Z. T. Bloom, T. L. Brecht, J. M. Chappell, M. D. Choi, J. E. Christensen, I. T. Counts, D. Daprano, J. Dodson, K. Eng, D. J. Fialkow, C. A. C. Garcia, W. Ha, T. R. B. Harris, n. holman, I. Khalaf, J. W. Matten, C. A. Peterson, C. E. Plesha, M. J. Ruiz, A. Smith, B. J. Thomas, S. J. Whiteley, T. D. Ladd, M. P. Jura, M. T. Rakher, and M. G. Borselli, Two-dimensional Si spin qubit arrays with multilevel interconnects, PRX Quantum 6, 030327 (2025).DOI
- [6]V. John, C. X. Yu, B. van Straaten, E. A. Rodríguez-Mena, M. Rodríguez, S. D. Oosterhout, L. E. A. Stehouwer, G. Scappucci, M. Rimbach-Russ, S. Bosco, F. Borsoi, Y.-M. Niquet, and M. Veldhorst, Robust and localised control of a 10-spin qubit array in germanium, Nature Communications 16, 10560 (2025).DOI
- [7]S. Neyens, O. K. Zietz, T. F. Watson, F. Luthi, A. Nethwewala, H. C. George, E. Henry, M. Islam, A. J. Wagner, F. Borjans, E. J. Connors, J. Corrigan, M. J. Curry, D. Keith, R. Kotlyar, L. F. Lampert, M. T. Mądzik, K. Millard, F. A. Mohiyaddin, S. Pellerano, R. Pillarisetty, M. Ramsey, R. Savyt sky, S. Schaal, G. Zheng, J. Ziegler, N. C. Bishop, S. Bojarski, J. Roberts, and J. S. Clarke, Probing single electrons across 300-mm spin qubit wafers, Nature 629, 80 (2024).DOI
- [8]N. P. de Leon, K. M. Itoh, D. Kim, K. K. Mehta, T. E. Northup, H. Paik, B. S. Palmer, N. Samarth, S. Sangtawesin, and D. W. Steuerman, Materials challenges and opportunities for quantum computing hardware, Science 372, eabb2823 (2021).
- [9]B. Paquet Wue tz, M. P. Losert, S. Koelling, L. E. A. Stehouwer, A.-M. J. Zwerver, S. G. J. Philips, M. T. Mądzik, X. Xue, G. Zheng, M. Lodari, S. V. Amitonov, N. Samkharadze, A. Sammak, L. M. K. Vandersypen, R. Rahman, S. N. Coppersmith, O. Moutanabbir, M. Friesen, and G. Scappucci, Anticommutation fluctuations lifting the energy degeneracy in Si/SiGe quantum dots, Nature Communications 13, 7730 (2022).
- [10]M. P. Losert, M. A. Eriksson, R. Joynt, R. Rahman, G. Scappucci, S. N. Coppersmith, and M. Friesen, Practical strategies for enhancing the valley splitting in Si/SiGeDOI
- [11]T. B. Boykin, G. Klimeck, M. A. Eriksson, M. Friesen, S. N. Coppersmith, P. von Allmen, F. Oyafuso, and S. Lee, Valley splitting in strained silicon quantum wells, Appl. Phys. Lett. 84, 115 (2004).DOI
- [12]M. Friesen, S. Chutia, C. Tahan, and S. N. Coppersmith, Valley splitting theory of SiGe/Si/SiGe quantum wells, Phys. Rev. B 75, 115318 (2007).DOI
- [13]M. Volmer, T. Struck, A. Sala, B. Chen, M. Oberländer, T. Offermann, R. Xue, L. Visser, J.-S. Tu, S. Trellenkamp, Ł. Cywiński, H. Bluhm, and L. R. Schreiber, Mapping of valley splitting by conveyor-mode spin-coherent electron shuttling, npj Quantum Information 10, 61 (2024).
- [14]B. D. Woods, M. P. Losert, N. R. Elston, H. Soomro, M. A. Eriksson, S. N. Coppersmith, R. Joynt, and M. Friesen, Statistical characterization of valley coupling in Si/SiGe quantum dots via g-factor measurements near a valley vortex (2026), arXiv:2507.05160.arxiv.org/abs/2507.05160
- [15]D. J. King, M. Kim, J. Reily, J. C. Marcks, M. Friesen, B. D. Woods, and M. A. Eriksson, Complete measurement of tunnel- and valley-coupling parameters in a silicon double quantum dot (2026), arXiv:2607.09638.arxiv.org/abs/2607.09638
- [16]M. O. Nestoklon, L. E. Golub, and E. L. Ivchenko, Spin and valley-orbit splittings in SiGe/Si heterostructures, Phys Rev B 73, 235334 (2006).DOI
- [17]M. Prada, G. Klimeck, and R. Joynt, Spin-orbit splittings in Si/SiGe quantum wells: from ideal Si membranes to realistic heterostructures, New Journal of Physics 13, 013009 (2011).DOI
- [18]R. Ferdous, E. Kawakami, P. Scarlino, M. Nowak, D. R. Ward, D. E. Savage, M. G. Lagally, S. N. Coppersmith, M. Friesen, M. A. Eriksson, L. M. K. Vandersypen, and R. Rahman, Valley dependent anisotropic spin splitting in silicon quantum dots, npj Quantum Information 4, 26 (2018).DOI
- [19]X. Cai, E. J. Connors, L. F. Edge, and J. M. Nichol, Coherent spin–valley oscillations in silicon, Nature Physics 19, 386 (2023).DOI
- [20]B. D. Woods, M. P. Losert, R. Joynt, and M. Friesen, g-factor theory of Si/SiGe quantum dots: Spin-valley and giant renormalization effects, Phys. Rev. Lett. 136, 206201 (2026).DOI
- [21]M. Volmer, T. Struck, A. Sala, J.-S. Tu, S. Trellenkamp, D. D. Esposti, G. Scappucci, H. Bluhm, Łukasz Cywiński, and L. R. Schreiber, Mapping g-factors and complex intervalley coupling in Si/SiGe by conveyor-mode shuttling (2026), arXiv:2603.01844.arxiv.org/abs/2603.01844
- [22]G. Burkard, T. D. Ladd, A. Pan, J. M. Nichol, and J. R. Petta, Semiconductor spin qubits, Rev. Mod. Phys. 95, 025003 (2023).DOI
- [23]A. Thayil, L. Ermonnei, and M. Kantner, Theory of valley splitting in Si/SiGe spin qubits: Interplay of strain, resonances, and random alloy disorder, Phys. Rev. B 112, 115303 (2025).
- [24]J. C. Marcks, E. Eagen, E. C. Brann, M. P. Losert, T. Oh, J. Reily, C. S. Wang, D. Keith, F. A. Mohiyaddin, F. Luthi, M. J. Curry, J. Zhang, F. J. Heremans, M. Friesen, and M. A. Eriksson, Valley splitting correlations across a silicon quantum well containing germanium, Nature Communications 16, 11381 (2025).DOI
- [25]M. Abraham and Members of the HRL Quantum Team and Collaborators, A digitally controlled silicon quantum processing unit, Nature 655, 1154 (2026). [26] D. Degli Esposti, E. C. Brann, A. Elsayed, D. Costa, M. Friesen, and G. Scappucci, Confinement drives valley splitting above 4K in buried silicon quantum wells, Nano Letters 26, 15131 (2026).
- [27]T. McJunkin, B. Harpt, Y. Feng, M. Losert, R. Rahman, J. P. Dodson, M. A. Wolfe, D. E. Savage, M. G. Lagally, S. N. Coppersmith, M. Friesen, R. Joynt, and M. A. Eriksson, SiGe quantum wells with oscillating Ge concentrations for quantum dot qubits, Nature Commun. 13, 7777 (2022).DOI
- [28]Y. Feng and R. Joynt, Enhanced valley splitting in Si layers with oscillatory Ge concentration, Phys. Rev. B 106, 085304 (2022).DOI
- [29]E. H. Chen, K. Raach, A. Pan, A. A. Kiselev, E. Acuna, J. Z. Blumoff, T. Brecht, M. D. Choi, W. Ha, D. R. Hulbert, M. P. Jura, T. E. Keating, R. Noah, B. Sun, B. J. Thomas, M. G. Borselli, C. Jackson, M. T. Rakhier, and R. S. Ross, Detuning axis pulsed spectroscopy of valley-orbital states in Si/Si-Ge quantum dots, Phys. Rev. Applied 15, 044303 (2021).DOI
- [30]T. Koch, C. Godfrin, V. Adam, J. Ferrero, D. Schroller, N. Glaeser, S. Kubicek, R. Li, R. Loo, S. Massar, G. Simion, D. Wan, K. De Greve, and W. Wernsdorfer, Industrial 300 mw wafer processed spin qubits in natural silicon/silicon-germanium, npj Quantum Information 11, 59 (2025).
- [31]I. Rahfl, C. Richter, M. Schmidbauer, M. Oezkent, T. Remmele, M. Hanke, L. R. Schreiber, D. Dütz, S. Umezawa, M. Albrecht, Y.-M. Niquet, T. Salamone, B. M. Diaz, T. Schroeder, J. Martin, and K.-P. Gradwohl, Sharp periodic Ge concentration modulations beyond the conduction band valley wavevector k₀ in nuclear spin-free Si quantum wells (2026), arXiv:2605.31358.
- [32]B. D. Woods, H. Soomro, E. S. Joseph, C. C. D. Frink, R. Joynt, M. A. Eriksson, and M. Friesen, Coupling conduction-band valleys in SiGe heterostructures via shear strain and Ge concentration oscillations, npj Quantum Information 10, 54 (2024).DOI
- [33]E. Acuna, J. D. Broz, K. Shyamsundar, A. B. Mei, C. P. Feeney, V. Smetanka, T. Davis, K. Lee, M. D. Choi, B. Boyd, J. Suh, W. Ha, C. Jennings, A. S. Pan, D. S. Sanchez, M. D. Reed, and J. R. Petta, Coherent control of a triangular exchange-only spin qubit, Phys. Rev. Appl. 22, 044057 (2024).DOI
- [34]F. Schäffer, H.-J. Herzog, H. Jorke, and E. Kasper, Influence of thermal annealing on the electron mobility in modulation doped Si/SiGe heterostructures, J. Vac. Sci. Technol. B 9, 2039 (1991).
- [35]E. S. Joseph, T. Oh, D. E. Savage, D. S. Stone, D. Campbell, B. D. Woods, M. Friesen, and M. A. Eriksson, Enhancing valley splitting in Si/SiGe Wiggle Well quantum dots through integrated stressor-gate architectures (2026), arXiv:2611.12850.arxiv.org/abs/2611.12850
- [36]C. C. D. Frink, T. Oh, E. S. Joseph, M. P. Losert, E. R. MacQuarrie, B. D. Woods, M. A. Eriksson, and M. Friesen, Reducing strain fluctuations in quantum dot devices by gate-layer stacking (2025), arXiv:2312.09235.arxiv.org/abs/2312.09235
- [37]T. Thorbeck and N. M. Zimmerman, Formation of strain-induced quantum dots in gated semiconductor nanostructures, AIP Advances 5, 087107 (2015).
- [38]Y.-S. Kim, J. K. Kim, J. W. Park, D.-H. Kim, W.-J. Lee, D.-L. Cho, and Y. S. Song, Properties of chemically vapor deposited blanket tungsten films on tin glue layers prepared by chemical vapor deposition, Metals and Materials 6, 577 (2000).DOI
- [39]S. W. Lee, D. J. Kim, J. Jeon, and T.-S. Kim, Effect of W nucleation process on the adhesion of W/TiN/SiO₂ structures for BEOL interconnects, Applied Surface Science 736, 166754 (2026).DOI
- [40]O. Dyck, D. N. Leonard, L. F. Edge, C. A. Jackson, E. J. Pritchett, P. W. Deelman, and J. D. Poplawsky, Accurate quantification of Si/SiGe interface profiles via atom probe tomography, Adv. Mater. Interfaces 4, 1700622 (2017).DOI
- [41]B. D. Woods, M. A. Eriksson, R. Joynt, and M. Friesen, Spin-orbit enhancement in Si/SiGe heterostructures with oscillating Ge concentration, Phys. Rev. B 107, 035418 (2023).DOI
- [42]T. McJunkin, E. R. MacQuarrie, L. Tom, S. F. Neyens, J. P. Dodson, B. Thorgrimsson, J. Corrigan, H. E. Ercan, D. E. Savage, M. G. Lagally, R. Joynt, S. N. Coppersmith, M. Friesen, and M. A. Eriksson, Valley splittings in Si/SiGe quantum dots with a germanium spike in the silicon well, Phys. Rev. B 104, 085406 (2021).DOI
- [43]H. Soomro, M. Kim, A. Vivrekar, M. A. Eriksson, B. D. Woods, and M. Friesen, High-fidelity EDSR in Si/SiGe Wiggle Wells (2026), arXiv:2605.24790.arxiv.org/abs/2605.24790
- [44]R. Németh, V. K. Bandaru, P. Alves, E. Brann, O. M. Eskandari, H. Soomro, A. Vivrekar, M. Eriks son, M. P. Losert, and M. Friesen, Omnidirectional shuttling to avoid valley excitations in Si/SiGe quantum wells, PRX Quantum 7, 010336 (2026).DOI
- [45]COMSOL Multiphysics® v.6.4. www.comsol.com. COMSOL AB, Stockholm, Sweden (2025).
- [46]P. Bradley and R. Radebaugh, Properties of Selected Materials at Cryogenic Temperatures (CRC Press, Boca Raton, FL, 2013).
- [47]N. J. Simon, Cryogenic Properties of Inorganic Insulation Materials for ITER Magnets: A Review (National Institute of Standards and Technology, Boulder, CO, 1994).DOI
- [48]H. J. McSkimin, Measurement of elastic constants at low temperatures by means of high frequency ultrasonic waves, The Journal of the Acoustical Society of America 25, 826 (1953).DOI
- [49]M. Levinshtein, S. Rumyantsev, and M. Shur, Properties of Advanced Semiconductor Materials: GaN, AlN, InN, BN, SiC, SiGe, A Wiley-Interscience publication (Wiley, 2001).
- [50]R. J. Corruccini and J. J. Gniewek, Thermal Expansion of Technical Solids at Low Temperatures: A Compilation From the Literature, NBS Monograph No. 29 (National Bureau of Standards, Washington, D.C., 1961).DOI
- [51]M. Burke, Thermal expansion of ceramic materials at −200° to 0°C, Journal of the American Ceramic Society 45, 305 (1962).
- [52]O. M. E. Ylivaara, A. Langner, S. Ek, J. Malm, J. Julin, M. Laitinen, S. Ali, S. Sintonen, H. Lipsanen, T. Sajavaara, and R. L. Puurunen, Thermomechanical properties of aluminum oxide thin films made by atomic layer deposition, Journal of Vacuum Science & Technology A 40, 62414 (2022).DOI
- [53]M. V. Fischetti and S. E. Laux, Band structure, deformation potentials, and carrier mobility in strained Si, Ge, and SiGe alloys, Journal of Applied Physics 80, 2234 (1996).DOI
- [54]F. Stern and S. E. Laux, Charge transfer and low-temperature electron mobility in a strained Si layer in relaxed Si₁−ₓGeₓ, Applied Physics Letters 61, 1110 (1992).DOI
- [55]F. Schäffer, High-mobility Si and Ge structures, Semi- conductor Science and Technology 12, 1515 (1997).
- [56]The value of C₂ reported here corrects a factor-of-2 discrepancy in Ref. [32]. However, we note that the C₂ term in Eq. (7) has almost no effect on the valley splitting, compared to the C₁ term.
- [57]A. Jones, E. Pritchett, E. Chen, T. Keating, R. Andrews, J. Blumoff, L. De Lorenzo, K. Eng, S. Ha, A. Kiselev, S. Meenahan, S. Merkel, J. Wright, L. Edge, R. Ross, M. Rakher, M. Borselli, and A. Hunter, Spin-blockade spectroscopy of Si/Si-Ge quantum dots, Phys. Rev. Appl. 12, 014026 (2019).DOI
- [58]S. D. Ha, E. Acuna, K. Raach, Z. T. Bloom, T. L. Brecht, J. M. Chappell, M. D. Choi, J. E. Christensen, I. T. Counts, D. Daprano, J. P. Dodson, K. Eng, D. J. Fialkow, C. A. C. Garcia, W. Ha, T. R. B. Harris, N. Holman, I. Khalaf, J. W. Matten, C. A. Peterson, C. E. Plesha, M. J. Ruiz, A. Smith, B. J. Thomas, S. J. Whiteley, T. D. Ladd, M. P. Jura, M. T. Rakher, and M. G. Borselli, Two-dimensional Si spin qubit arrays with multilevel interconnects, PRX Quantum 6, 030327 (2025).DOI
- [59]G. Casella and R. L. Berger, Statistical Inference, 2nd ed. (Duxbury Press, Pacific Grove, CA, 2002).