Arxiv:2001.08251V1 [Cond-Mat.Mes-Hall] 22 Jan 2020

Democratizing Spin Qubits∗

Charles Tahan† Laboratory for Physical Sciences, 8050 Greenmead Rd, College Park, MD 20740 (Dated: January 24, 2020) I’ve been building Powerpoint-based quantum computers with electron spins in silicon for 17 19 years. Unfortunately, real-life-based quantum dot quantum computers are harder to implement. Fabrication, control, and materials challenges abound. The way to accelerate discovery is to make and measure more qubits. Here I discuss separating the qubit realization and testing circuitry from the materials science and on-chip fabrication that will ultimately be necessary. This approach should allow us, in the shorter term, to characterize wafers non-invasively for their qubit-relevant properties, to make small qubit systems on various different materials with little extra cost, and even to test spin-qubit to superconducting cavity entanglement protocols where the best possible cavity quality is preserved. Such a testbed can advance the materials science of semiconductor quantum information devices and enable small quantum computers. This article may also be useful as a light and light-hearted introduction to spin qubits.

The two states of a qubit are realized in the spin of an cavity (Figure 2). Superconducting circuits can be float- electron, spin up and spin down. A single electron can ing, requiring no source or sink of carriers. In contrast, be trapped in a semiconductor box, called a quantum using what we know now as best principles for making a dot (Figure 1). In silicon—currently the most promising silicon quantum dot qubit (Figure 3), one needs to make material for spin-based quantum computing—the indi- very small dots (due to large effective mass), with mul- rect band-gap means that the electron has extra nearby tiple layers of overlapping or tightly-aligned metal gates energy levels, combinations of the conduction band min- to limit cross-capacitance between dot gates [16] and to ima or valleys where the electron exists. Temperature, increase yield due to disorder, all with O(10 nm) wire- noise, and gate operations can cause unwanted excita- widths at 50-100 nm wire pitch. Worse, the materials tion into these states. This so-called “valley splitting” stack (causing current leakage between the top gates and problem, especially in silicon-germanium quantum dots, electrons, disorder, etc) can be critically important to impacts yield, initialization/readout, and quantum oper- whether the quantum dots you want can even be formed ations, and originally motivated this work. Although ob- where you want them, let alone will they work or have scure, this materials science issue is a roadblock to quan- desirable qubit properties. Spin quantum dot qubits tum information processors in silicon, and valley splitting also need multiple physical wires per dot, in addition is only representative of a greater challenge. to nearby charge sensors (for spin-to-charge conversion- Like for other qubit parameters such as coherence time based readout). This level of complexity in fabrica- and operation fidelity, to measure the valley splitting one tion—which must be coupled with good materials science must fabricate a quantum dot and test it, typically at properties of the wafer and the gate stack—retards both dilution refrigerator temperatures [1]. The general diffi- new qubit exploration and characterizing many, individ- culty in making qubits in semiconductors has hampered ual quantum dots to optimize materials parameters. progress in the field. Exciting recent success—functional We can separate the materials science challenge from dot qubits and compelling quantum gate demonstrations the “making the dots and measure them” challenge. [2–10]—has taught us a lot about how to make good If the dots and associated readout circuitry can be qubits. Yet there is a high barrier to entry for new experi- made on another circuit chip or board, then the actual mental groups compared to, say, superconducting qubits. electron(hole)-hosting wafer can be optimized separately Do superconducting qubit experimentalists have it (and even made of different crystals such as germanium). easy? Yeah, they kind of do. The transmon qubit [11– The idea of “flip-chip” engineering has already been ap- 13], which itself is robust by design [14], can be fabricated plied in the superconducting context [17, 18], while the with essentially a single layer of metal, at huge length concept of a “probe” trap has been used in ion trap quan- scales if wanted [15], and can be characterized and con- tum computing [19, 20] to search for heating mechanisms trolled with a single microwave cavity/generator/line or on relevant surfaces. We hope to motivate a new kind of arXiv:2001.08251v1 [cond-mat.mes-hall] 22 Jan 2020 even wirelessly, via a properly designed superconducting testbed where dots are induced on a separate chip to characterize materials [21] and to see if the protocols we design for qubit operation work. In doing so, an acceler- ation of progress similar to that driven by the “wireless” ∗ or, How to make semiconductor-based quantum computers with- 3D transmon (Figure 2bc) could be replicated in the spin out fabricating quantum dot qubits; Based on talks I gave at the community. ARO/LPS Quantum Computing Program Review, August 2018, and 4th School and Conference on Spin-Based Quantum Infor- mation Processing, September 2018. † [email protected] seApni XBfrmr eal) iue7epiil hw h alyrlxto aei h w ae ffixed of turns cases It two splitting. the valley in and rate field relaxation electric valley with the coe accurately shows changes Bloch that explicitly the height that 7 function out Figure wave Umklapp and details). the height more at function for wave peaked B be IX will Appendix (see rate relaxation decreasing phonon at function; the phonons wave that the predicts of prefactor extent This (assuming 10). energy phonon (Eq. case occurs orbital transition same exact this the relaxation have valley states for M final that and remember initial this (both elements understand manifold To lowest functions). the wave within gaussian to a due given to opposed as dependence phonon, Umklapp emitted B. IX the of roughness/steps, length interface wave to due here presented where results 1D the of versus layer meV, reduced donor 150 the be of from barriers likely separation nm, for will 10 space-charge splitting dots of to well valley etc. quantum due quantum and silicon a field states with electric realistic valley heterostructure direction in SiGe excited growth typical and large a for ground a (line) for and interpolated functions been wave has which for (points) model tight-binding 1D a10nmquantumwell.The from Results 8: Figure e scmaeE.1 opr ria eaain q 0 tfis lne h alyrlxto aehsa has rate relaxation valley the glance, first At 10. Eq. relaxation, orbital pure to 16 Eq. compare us Let _ x 0 2 v _ M stevle splitting, valley the is q u ⇠ h alyrlxto xrsinas nldspoietya prominently includes also expression relaxation valley The . = 1 nteobtlcs,w utcluaemti lmnsfrom elements matrix calculate must we case, orbital the In . v /

~ 2 v ffi l,t inst h ers alyat valley nearest the to cients hc onc aly nnihoigBilunznsaetems e most the are zones Brillouin neighboring in valleys connect which z z Valley Splitting, E (eV) cmoeto h lcrndtwv ucini h upto -adtgtbnigcalculation tight-binding 2-band a of output the is function wave dot electron the of -component 0 v 0.5 1.5 2.5 scntn with constant is 0 1 2 3 0 4

x 10 nteobtlcs asmn aaoi o oetasfrbt n arxelements matrix and both for potentials dot parabolic (assuming case orbital the in z 0 − 3 0.2 QW=10 nm (N=74) a b in gaussian) (assumed function wave the of extent the is c d z 0.4 0 “p-like” e These rate. relaxation the increases 0.6 q u ~130 meV v hc tint.Euto 6as hw h motneo the of importance the shows also 16 Equation isn’t). it which , Electric field, E kz =0 0.8 . 0 3 . k

3 excited max confinement k max

“valley state” 1

splitting Appendix in given are calculation this of details The . r ote most are z (V/m)

~0.1 – 8 meV 1.2

Ev ~ 0-3 meV

k1.4 z z s lk neoefnto)sc httematrix the that such function) envelope -like ffi 1e7 meV 200 meV 150 meV 100 meV 50 meV

in n hnn r hneitdi the in emitted then are phonons and cient spin “s-like” 1.6

exp( qubit 1.8 ff 2 csaerltd nta Umklapp that in related, are ects p lk to -like x 10 z 0 2 q 7

2

u 2 / 6 4) ⇥ ffi 1 10 in eaainchannel relaxation cient rfco beti the in absent prefactor s z ˆ lk tts uhthat such states, -like 6 and /.Vle splitting Valley V/m.

Spin qubit in silicon. q Figure 1. a) A trapping potential due to au heterostructure box or donor hosts our electron. b) The energy levels can be labeled like an atom if we assume an effective phonon the is mass theory [22] (note that the real wave function of the electron also has Kohn-Luttinger oscillations [23] due to the conduction band valleys at 0.8kF and atomistic oscillations 20 z z from the crystal underneath this envelope). The “valley-splitting”0 0 between3 the states can range from 0 to several meV. c) The conduction and valence band symmetry governs possible additional levels. For electrons in strained-silicon SiGe quantum wells or at inversion layers, there are two conduction band minima and thus double the number of states. d) Idealized electron wave functions of the lowest two valley states in the growth direction of a SiGe quantum well tshowing the Kohn-Luttinger oscillations [24].

We can make a quantum dot qubit without the excited states), the spontaneous decay time (called fabricating a quantum dot the T1 time, almost always due to emission of a phonon at these dilution refrigerator temperatures, but more com- A spin qubit is formed from the Zeeman split sub-levels plicated above ~1K), and the decay of the coherence of (see Figure 1c) of the ground state of an electron trapped the qubit as a function of time (or T2∗ for the specific 1/e in some potential inside a semiconductor. That electro- time assuming an exponential fall-off, which is actually static potential can be artificial, formed from the combi- not usually the case). All these parameters can change nation of a heterostructure [25] and an external voltage with the number of particles per dot and if multiple dots [26], or natural, the pull of an implanted donor [27]. One are coupled together, either via the Pauli exchange inter- or more electrons or holes can be placed in a dot to form a action or capacitively (more on this below). single qubit or a qubit can be “encoded” [28] into the large Let us now design the simplest quantum dot, using Hilbert space of multiple separate quantum dots [29–31], best practices from recent progress (Figure 3). Our op- and each approach has different potentially useful proper- tions include semiconductors such as GaAs, silicon, ger- ties for quantum computing. Typically, we focus on 1 or manium. GaAs has spinful nuclei, leading to poor T2∗ 3 electrons per dot [32] in a roughly parabolic potential in times (~ns). Germanium is interesting. But the pull the plane of the two-dimensional electron gas (2DEG) or of silicon is strong due to the CMOS industry (ultra- inversion layer, where the potential must be deep enough chemically pure and perfect crystals, precision lithogra- such that the excited state orbital levels (“p-like” envelope phy, dielectrics) and the fact that isotopically enriched 29 functions [33]) are well above the energy of thermal exci- silicon exists and is available (where the spin-1/2 Si tations ( kT ). For indirect band-gap conduction bands nuclei have been removed leaving only spin-0 silicon-28), ∼ where electrons live (or valence bands where holes live, leading to extremely long coherence times [36]. which can be considered spin-3/2 particles), there can First, create an electrostatic trap for the electron in z be other nearer levels than the orbital levels which one by making either a “quantum well” or an inversion layer. may also need to worry about. Somewhat surprisingly, For the former, use a strain-engineered SiGe-Si-SiGe the spin splitting (linearly proportional to the magnetic sandwich [37]. For the latter, use the well-known oxide- field) can be smaller than the effective temperature, and silicon (MOS) interface of silicon and silicon-dioxide. An typically is in many experiments. Kramers degeneracy accumulation gate (positive voltage) creates a triangular- (where a “forbidden” very tiny matrix element connects like potential, pulling electrons (if available) up the top phonons between the two spin states) results in extremely surface of the buried quantum well (Figure 1a). In ei- long lifetimes for spins in some solids, a fact that has been ther case there are no electrons until we put them there known at least since the 1950s [34, 35]. (assuming no doping). The effective mass of electrons in In designing a quantum computer based on quan- silicon is relatively high, meaning that to get an orbital tum dots we care about a number of intrinsic param- splitting of ~1meV = 10K, we need an effective “box” eters derived from the quantum dot potential, mate- of ~30 nm laterally. Finally, a combination of negative rial and material stack, and the proximity of other (depletion) and positive (accumulation) gates can form qubits/defects/oxides/gates, and still other parameters the trap in the (x, y) plane creating our dot potential. that are more relevant for two qubit operations. These Because of the large effective mass of electrons in sili- include the level structure of the dot (or spectroscopy of con compared to GaAs [38], the community has learned 3

Consider three options for loading electrons into the a Transmon qubit induced quantum dot (Figure 4). One, from the bulk Josephon junction by shining light with energy above the band-gap on the capacitors wafer, Figure 4a, (often done to ensure that the well

Coplanar was populated or to increase density in the 2DEG and cavity is also common practice in STM experiments), through b c light background doping (not recommended), or a doped back gate reservoir. Generate enough carriers and our dot will trap one (or more, depending on the depth of

antenna it’s trapping potential). Two, Figure 4b, we can move the gate tip to an implanted region on the wafer, and sapphire substrate physically move the loaded dot away from it to isolate

Figure 7.4: Images showing the fluxonium qubit and its coupling to the antennathe and dot. Three, we add another lead to the gate-chip, a 3D cavity modes. The top picture in an SEM image of the junction array shuntingmuch the wider and fatter “bath” gate (Figure 4c) that can phase slip junction. The coupling SQUID junctions join the antenna mode resonator Figure 2. From the transmonwith the fluxonium qubit qubit to loop. the The antenna wireless pads couple rev- the antennabring mode and electrons into the channel from the implanted re- fluxonium to the 3D cavity resonator, as shown in the bottom two images. olution. (a) The transmon superconducting qubit is a single gion to the dot much like is already done. Implanting is a layer of metal and a shadow-evaporated junction. It is ex- standard procedure and requires a mask but can be out- tremely robust in that nearly all fabricated qubits work. In a sourced, and of course leaves the wafer, well, implanted. co-planar geometry a superconducting cavity can be used to (So we’ve broken the non-invasive pledge for options two readout the qubit (circuit QED). (c) A crucial step to advance and three.) the field was applying these same ideas to a 3D163 geometry. By going larger the participation of loss mechanisms was propor- tionally lowered, and allowed systematic study of qubits with Characterizing an induced quantum dot longer life while the more scalable 2D qubits were catching up in performance. But perhaps more interesting the wireless approach can allow all different types of qubits to be tested We must confirm that the electron is present in the more quickly, even as in (c) fluxonium, where there isn’t a induced quantum dot to have a qubit. If the dot is truly natural coupling to an electric field but an antenna can be isolated with only one wire to both maintain the poten- used to couple to the 3D field. (Figures 2b and 2c courtesy tial and probe the dot, that is really hard to do. Best M. Devoret, Yale.) practice for qubit measurement or readout is to create a nearby quantum dot charge sensor (see the top-down SEM image in Figure 3b for an example), tuned to the that unwanted “dots” (think a disordered eggshell con- edge of a Coulomb blockade peak for maximum sensitiv- tainer) can form in the quantum well via the presence of ity (small charge redistributions in the qubit dots affect donors, or by possibly tiny strain due to metallic gates the current flowing through the dot sensor). By puls- [39] on the surface. Therefore, don’t dope the quantum ing the electrons in the dots one can figure out how to well; implant donors in source regions creating a bath of convert the spin information to different charge distribu- electrons that can be brought electrostatically near the tions, detectable with the readout dot with high sensitiv- active quantum dots with gates. And make the gates as ity. However, since we only have a single wire, a better uniform (“total coverage”) and as far away as possible. option is to use so-called dispersive readout, which should This is how the best arrays of lateral quantum dots for be called quantum capacitive or curvature readout. Here, qubits are fabricated today (Figure 3). a tank circuit is attached to a nearby dot gate, and small All we really need to induce a dot potential, however, is changes in the quantum capacitance (or more generally a single wire with a positive potential with the appropri- the curvature coupling [44]) of the system are detected ate geometry and distance from the target active layer. by measuring the phase shift of a reflected rf-pulse (dis- This can be formed from a metal gate on the dot wafer persive shift of the resonator) at the resonator frequency surface, or by a wire on a different chip or probe tip. In- (where the frequency is relatively low, 100 Mhz to 500 deed, Scanning Tunneling Microscope (STM) tips have MHz) [45–52]. But note that this technique as applied been used in the past to induce dots on the surface of so far has only worked in two ways: one, detecting a sig- materials [40, 41]. This is the core of our proposal: make nal when electrons tunnel in and out of a nearby bath the dot gates on a separate chip. Thus, the simplest dot (so-called “tunneling” capacitance) or two, at a charge we can imagine actually doesn’t require fabricating a dot transition (eg 11 to 20, see Figure 8 for a preview) to a on the “dot chip” at all. We can fabricate the “dot in- nearby dot (where the quantum capacitance is detected, ducing chip” or gate chip on a separate substrate, with a the curvature of the two-dot, one-electron system is max- single layer of metal. We can then place that perpendicu- imal at the degeneracy point, the symmetric state of a lar to the “wafer chip” (touching or not) to where the dot charge qubit). Thus if this condition is met, standard is actually intended to be created. All the circuitry we techniques can be applied to map out 0, 1, ... electrons need to measure the dot can be off the wafer chip under in the dot. A higher-frequency resonator can also be used test. as in [53, 54] (in these cases for cavity-QED not quantum APPLIED PHYSICS LETTERS 106, 223507 (2015)

A reconfigurable gate architecture for Si/SiGe quantum dots D. M. Zajac,1 T. M. Hazard,1 X. Mi,1 K. Wang,1,a) and J. R. Petta1,2 1Department of Physics, Princeton University, Princeton, New Jersey 08544, USA 2Department of Physics, University of California, Santa Barbara, California 93106, USA (Received 5 February 2015; accepted 27 May 2015; published online 5 June 2015) We demonstrate a reconfigurable quantum dot gate architecture that incorporates two interchangeable transport channels. One channel is used to form quantum dots, and the other is used for charge sensing. The quantum dot transport channel can support either a single or a double quantum dot. We demonstrate few-electron occupation in a single quantum dot and extract charging energies as large as 6.6 meV. Magnetospectroscopy is used to measure valley splittings in the range of 35–70 leV. By energizing two additional gates, we form a few-electron double quantum dot and demonstrate tunable tunnel coupling at the (1,0) to (0,1) interdot charge transition. VC 2015 AIP Publishing LLC.[http://dx.doi.org/10.1063/1.4922249]

1220 Hanson et al.: Spins in few-electron quantum dots Quantum dots have considerable potential for the realiza- by the Si/Si0.7Ge0.3 lattice mismatch, breaking the six-fold tion of spin-based quantum devices.1,2 Extremely long spin valley degeneracy.14,15 The degeneracy of the two lowest coherence times3–5 and the ability to utilize existing fabrica- lying valleys is further lifted by quantum confinement in the tion processes make silicon an attractive host material for growth direction.16 Accumulation mode Hall bar samples fab- quantum dot qubits.6–8 Existing depletion mode designs use ricated on this wafer yield a two-dimensional electron gas gate electrode patterns that are much larger than the spatial (2DEG) carrier mobility l 1:6 105 cm2/Vs at an electron extent of the resulting electron wavefunctions.9 As a result, it density n 2:2 1011/cm2¼and temperature T 350 mK. A ¼  ¼ is difficult to precisely control the electronic confinement valley splitting Dv 150 leV is measured at a magnetic field potential. Successful scaling to a larger number of quantum B 1.8 T. The 2DEG¼ also undergoes a metal-to-insulator ¼ 11 2 dots will require fine control of the confinement potential on transition (MIT) at a critical density nc 0:46 10 /cm . FIG. 1. Schematic picture20 nm ofa length quantum scales. dot Accumulation in ͑a͒ a lateral mode designs,10,11 where ¼  geometry and ͑b͒ in aelectrons vertical geometry. are accumulated The quantum under small dot positively biased gates 4 ͑represented by a disk͒ is connected to source and drain reser- 12 (instead of depleted using large “stadium” gate designs ) 2 voirs via tunnel barriers,allow allowing control the ofcurrent the confinement through the potential de- on a much smaller (a) B1 B2 B3 INCREASING INTER-DOT TUNNEL COUPLING vice I to be measured in response to a bias voltage VSD and a P1 P2 length scale and merit further development. (a) (b) (c) dI/dVsd (nS) gate voltage VG. 3.0 L2 0.8 In this letter, we present a reconfigurable accumulation L1

(V)

P1 0.4 mode device architecture that utilizes three overlapping alu- 50 nm V 2.7 0 Because a quantum dot is such a general kind of sys- (b) 2.5 VP2 (V) 2.6 2.5 VP2 (V) 2.6 2.5 VP2 (V) 2.6 minum gate layers. The device architecture has two parallel P1 P2 (d) V = 1.20 V (e) (f) V = 1.40 V tem, there exist quantum dots of many different sizes L1 B1 B2 B3 L2 B2 VB2 = 1.32 V B2 VB1 VB3 V VB3 V V (and interchangeable) transport channels. One of the channels S D B1 B1 B3 EF Electron Reservoir Dot 1 Dot 2 and materials: for instance, single molecules trapped be- S D S D S D E VP1 VP2 VP1 VP2 V V is used to create single and double quantum dots, while the Near Intrinsic Si n+ Ohmic Regions Silicon Dioxide C P1 P2 tween electrodes ͑Park et al., 2002͒, normal metal ͑Petta Aluminium Gates Oxidised Aluminium other channel is used to define a charge sensor quantum FIG. 2. Characteristics at di↵erent inter-dot tunnel and Ralph, 2001͒, superconducting13 ͑Ralph et al., 1995; -V -V FIG. 1. SEM and schematic+V view of the device.(a) coupling. Measured stability diagrams and energy landscape dot. The natural length scale of this gate architecture is Scanning electron micrograph of a device identical to that of the double dot system ranging from weak to strong inter- measured. (b) (Not to scale) Schematic cross-section view of dot tunnel coupling (a) (c) and (d) (f) respectively, for VL1 von Delft and Ralph, 2001͒, or ferromagnetic nanopar- comparable to the resulting dot size, allowing a higher degree -V -V +V the Si MOS- doubleV quantum dot. The architecture-V is defined = VL2 =3.0V,VB1 =0.76V,VB3 =1.0VandVSD =0. ticles ͑Guéron et al., 1999͒, self-assembled quantum dots FIG. 2.12 Lateral quantum dot device defined by metal surface by B1, B2 and B3 (barrier gates), L1 and L2 (lead gates), From lower to higher VB2, the tunnel+V barrier height decreases of control compared to depletion mode devices. Direct local Si cap Si cap and P1 and P2 (plunger gates). The gates are separated by resulting in stronger inter-dot tunnel coupling. (a) A checker an Al2O3 layer (light gray). Positive voltages applied to the box pattern, (b) honeycomb pattern and (c) diagonal parallel ͑Klein et al., 1996͒, semiconductor lateral ͑Kouwen- electrodes. ͑a͒ Schematic view. Negative voltages applied to lead and plunger gates induce an electron layer (black dashes) lines indicate that the two dots merge into a single dot as the accumulation also reduces capacitive cross-coupling, simpli- SiGe spacer SiGe spacer underneath the SiO2. By tuning the barrier gates, Dot 1 and coupling is increased [23]. hoven et al., 1997͒ or vertical dots ͑Kouwenhoven et al., metal gate electrodes ͑dark gray͒ lead to depleted regions Dot 2 are formed. The coupling of the dots is adjusted using fying the formation of double quantum dots and tuning of the the middle barrier (B2). The regions coloured with red are + 2001͒, and also semiconducting nanowires or carbon ͑white͒ in the 2DEG ͑lightModulation gray͒. Ohmic doping contacts ͑light gray the n source (S) and drain (D) contacts formed via di↵used V , with all other gate voltages held constant, together phosphorus.SiO2 P2 relevant tunnel rates. The architecture demonstratedcolumns͒ enable here pro- bonding wires ͑not shown͒ to make electrical with sketches of the energy landscape of the double dot. nanotubes ͑Dekker, 1999; McEuen, 2000; Björk et al., SiGe spacer The charge-stability maps moving from Fig. 2(a) to 2(c) vides a straightforward method for scaling to a larger series clearly show the e↵ects of an increasing inter-dot coupling contact to the 2DEG reservoirs. ͑b͒, ͑c͒ Scanning electron mi- I. RESULTS 2004͒. as the middle barrier-gate voltage VB2 is increased, lower- array of N quantum dots, with the requiredcrographs number of of͑b gate͒ a few-electron single-dot device and ͑c͒ a ing the tunnel barrier between the dots. Fig. 2(b) shows The electronic properties of quantum dots are domi- Device architecture. Figure 1 shows a scanning elec- the characteristic honeycomb-shaped stability map rep- electrodes in each channel growing linearlydouble as 2N dot1. device, showing theStrained gate Si electrodes ͑light gray͒ onStrained Si tron micrograph (SEM) and cross-sectional schematic of resenting intermediate inter-dot coupling [23], obtained nated by two effects. First, the Coulomb repulsion be- at V = 1.32 V. At lower middle barrier-gate voltage, þ the device, which incorporates 7 independently controlled B2 The device is fabricated on an undopedtop of Si/SiGe the surface hetero-͑dark gray͒. White dots indicate the location aluminium gates. When a positive bias is applied to the VB2 = 1.20 V, we observe a checker-box shaped map tween electrons on the dot leads to an energy cost for lead gates (L1 and L2) an accumulation layer of electrons [Fig. 2(a)], since the middle barrier is opaque enough of the quantum dots. Ohmic contacts are shown in the corners. is induced under the thin SiO2, to form the source and to almost completely decouple the two dots. In contrast, adding an extra electronstructure to the with dot. the Due growth to this profile charging shown in Fig. 1(a). A SiGe drain reservoirs for the double dot system. A positive the stability map in Fig. 2(c) shows the formation of di- White arrows outline the path of current IDOT from one reser- voltage on the plunger gate P1 (P2) causes electrons to agonal parallel lines at VB2 = 1.40 V. Here the two dots energy tunneling of electronsrelaxed buffer to or substrate from the is reservoirs grown on a Si wafer by linearly accumulate in Dot 1 (Dot 2). Independent biasing of P1 e↵ectively merge into a single dot due to the lowering of voir through the dot͑s͒ toFIG. the 1. other (a) Schematic reservoir. and For cross-sectional the devicetransmission in electron micrograph and P2 provides direct control of the double-dot electron the middle barrier [Fig. 2(f)]. The transport measure- can be suppressed atvarying low temperatures; the Ge concentration this phenom- from 0% to 30% over 3 lm. The occupancy (m, n). The tunnel barriers between the two ments shown here do not allow a precise determination ͑c͒, the two gates on theshowing side can the be growth used profile to create of the two heterostructure. quan- An 8 nm Si quantum well dots and the reservoirs are controlled using the barrier of the electron occupancy (m, n) in the dots, since it enon is called Coulombsurface blockade of this virtualvan Houten substrateet isal. then, polished before grow- gates: B1, B2 and B3. The middle barrier gate B2 deter- is possible that electrons remain in the dots even when ͑ is grown on top of a 225 nm Si0.7Ge0.3 layer, followed by a 50 nm Si0.7Ge0.3 tum point contacts, which can serve as electrometers by pass- mines the inter-dot tunnel coupling. The electrochemical ISD is immeasurably small. For the regime plotted in 1992͒. Second, the confinementing an additional in all 225 three nm thick directions Si0.7Ge0.3 layer, followed by spacerSiGe andsubstrate a 2 nm protective Si cap. (b) False-colorSiGe substrate scanning electron micro- potentialsSilicon of the coupled dots can also be easily tuned to Fig. 2 there were at least 10 electrons in each dot, based ing a current IQPC. Note that this device can also be used to be in resonance with those of the source and drain reser- on our measurement of Coulombd peaks as we further leads to quantum effectsan 8 that nm Siinfluence quantum the well electron (QW), dy- a 50 nm Si Ge spacer, graph of a device identical to the one measured. The first layer (blue) serves voirs. As shown in Fig. 1(b), gates L1 and L2 extend depleted the system. An absolute measurement of dot define0.7 a single0.3 dot. Image in b courtesy of A. Sachrajda. + as a screening͑ ͒ layera for layers 2 (red) and 3 (green), which form the plunger over the source and drain nc contacts, and also overlap occupancy would require integration of a charge sensor namics. Due to the resultingand a 2 nm discrete protective energy Si cap. spectrum, The Si QW is uniaxially strained b gates B1 and B3. The upper-layer gates (P1 and P2) into the system [7]. These results nevertheless demon- and barrier gates, respectively. (c) COMSOL simulation of the electron den- are patterned on top of the lead and barrier gates. The strate that the multi-gated structure provides excellent quantum dots behave in many ways as artificial atoms lithographic size of the dots is defined by the distance tunability of coupling while maintaining charge stability sity in the QW with the gate voltages tuned to form a single dot under L1 in between adjacent barrier gates ( 30 nm) and the width over a wide range of electron occupancy. ⇠ of the plunger gates ( 50 nm), as shown in Fig. 1(a). Capacitances and charging energies. Application ͑Kouwenhoven et al., 2001͒. Quantum dot, simplified. ⇠ beam epitaxy ͑see Fig.Figurethe2͒ upper. 3. By channel doping and the (d) AlGaAs with a double layer dot undera) LEarly1 and R depletion1 in the upper modeInter-dot quantum tunnel coupling dots tunability (modeled. Figure 2 onof a GaAs DC source-drain quantum bias V causes experiments the triple-points before a) SD The physics of dotsPresent containing address: more Department than of two Physics, elec- Harvardwith University, Si, free Cambridge, electronsthem) arechannel. utilizedintroduced. In (c) and doping (d), These the layers lower accumulate channel (modulation is configured doping) as a charge and sensor, negativelyshows the measured charged di↵erential metal conductance top of the gates de- in theto weakly-coupled push away regime [Fig. all 2(a)] electrons to extend to but one vice as a function of the plunger gate voltages, VP1 and form triangular shaped conducting regions [Fig. 3(a)] trons has been previouslyMA 02138, reviewed USA. ͑Kouwenhoven et at the GaAs/AlGaAs[42]; interface,with these a many-electron suffered typically singlefrom 50–100 dot severe formed nm disorder beneath be- L2 when. implemented in silicon. b) Accumulation mode devices use overlapping gates to al., 1997; Reimann and Manninen, 2002͒. Therefore we low the surface, formingform dot a two-dimensional and tunnel gates electron and represent gas the current state of the art [3, 4, 43]; often a nearby charge sensor (also a dot) is used focus on single and coupled0003-6951/2015/106(22)/223507/4/$30.00 quantum dots containing ͑2DEG͒—a thin106,͑ϳ 223507-1for10 spin-charge nm͒ sheet ofreadout. electrons Here, that the can dopingVC 2015 layer AIP canPublishing be removed LLC if implanted electron “sources” are nearby. c) A single quantum only one or two electrons. These systems are particularly Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions.only move along thedot interface. potential The in 2DEG it’s simplest can haveDownload form. high d)to IP: A quantum129.2.109.8 dot On: createdWed, 22 Jun with 2016 a single gate wire whose electrostatic potential has already important as they constitute the building blocks of pro- trapped an electron in the quantum well. (Top row device images are courtesy of Charlie Marcus, Jason Petta and Andrew mobility and relatively21:06:11 low electron density ͓typically posed electron spin-based quantum information proces- 105 −107 cm2 /V s andDzurak.)ϳ͑1−5͒ϫ1015 m−2, respectively͔. sors ͑Loss and DiVincenzo, 1998; DiVincenzo et al., The low electron-density results in a large Fermi wave- 2000; Byrd and Lidar, 2002; Levy, 2002; Wu and Lidar, length ͑ϳ40 nm͒ and aa large screening length, which al- b c 2002a, 2002b; Meier et al., 2003; Kyriakidis and Penney, lows us to locally deplete the 2DEG with an electric Ohmic regions Ohmic regions 2005; Taylor et al., 2005; Hanson and Burkard, 2007͒. (source of (source of field. This electric field isLight created by applying negative electrons for electrons for voltages to metal gategenerates electrodes on top of the hetero- bath gates) bath gates) electron- B. Fabrication of gated quantum dots structure ͓see Fig. 2͑a͔͒hole pairs. Electron-beam lithography enables fabrication of gate The bulk of the experiments discussed in this review structures with dimensions down to a few tens of na- n++ n++ was performed on electrostatically defined quantum nometers ͑Fig. 2͒, yielding local control over the deple- Electron loaded from ohmic region Induced quantum dot dots in GaAs. These devices are sometimes referred to tion of the 2DEG with roughlyCaptured electron the same spatial resolu- and dragged away and electron lead as lateral dots because of the lateral gate geometry. tion. Small islands of electrons can be isolated from the Lateral GaAs quantum dots are fabricated from het- rest of the 2DEG by choosing a suitable design of the erostructures of GaAs and AlGaAs grown by molecular- gate structure, thusFigure creating 4. quantumLoading dots. an Finally,electron low- in to an induced quantum dot from a perpendicular gate chip. (a) Light above the band-gap can put electrons into the conduction band. (or background dopants can create a reservoir to load from. (b) Rev. Mod. Phys., Vol. 79, No. 4, October–December 2007 Implanted regions on the chip can be used as loading zones if the tip/gate-chip has the ability to move. (c) By adding a bath gate to the gate-chip, one can load electrons even from a long distance away.

capacitive readout) putting the cavity into the quantum lead without tunneling transitions to a reservoir or dot regime. We calculated in Ref [21] that it may be possible (which has never been seen before). It could also be that to measure the small energy band curvature (Figure 5a) if the measurement apparatus is fast enough then the due to a single, stationary electron as compared to no sig- appearance of an electron from the bulk (due to light- nal (no electron trapped). Assuming the use of a super- created carriers for example) may be observable. If these conducting resonator in series with the dot-inducing gate, don’t work, then using the bath gate or a nearby dot (a quantum capacitances as low as 0.01 attoFarad should be 2 dot probe) will be necessary, slightly complicating our observable if a Q 105 can be achieved (see [21] for de- approach but making all the known techniques for qubit ∼ tails), allowing for the detection of electrons with a single characterization available. 5

a b c 0 10 20 30 40

50 Figure 6. 2 quantum dot simulation. (a) Gate-chip is 60 at 10 nm above the heterostructure, each wire has 10 nm by 10 nm cross-section. Wire-wire distance is 50 nm. Applied 70 voltages on the 3 wires are 0.02 V, -0.02 V, 0.02 V. b) Elec- Energy [meV] 80 trostatic potential of the induced double dot system in the xy plane of the wafer two-dimensional electron gas (2DEG) in 90 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 a silicon-germanium quantum well. c) Electrostatic potential along the green line. The solid blue line is the ground state Vg [V] energy and the red dashed line is the first excited state energy. (Figure courtesy Yun-Pil Shim.)

DC to induce AC for measurement 𝐶𝐶𝑝𝑝 𝐿𝐿 quantum dot (quantum capacitance) option is to introduce another rf field which drives tran- sitions in the dot. Unfortunately, given the large possi- ble range of valley splittings, and the possibility of vary Energy levels of 2e in a QD | , large orbital splittings (up to 8 meV), tunability of the ↓𝑣𝑣푣 ↓𝑣𝑣2⟩ 𝐸𝐸𝑇𝑇 microwave field would have to be over a vast range for a

𝐸𝐸| 𝑉𝑉𝑉𝑉, 𝑆𝑆 single dot. With two dots a new energy scale emerges, 𝐸𝐸 𝑣𝑣푣 𝑣𝑣푣 0 ↑ ↓ ⟩ = Magnetic field the detuning between the two dots, and there are many Quantum 𝑍𝑍 𝑉𝑉𝑉𝑉 Capacitance𝐸𝐸 𝐸𝐸 more options; valley spectroscopy can be achieved with- Captured electrons in an induced quantum dot out a magnetic field [55]. The reflectometry approach Magnetic field allows one to measure the critical parameters of the sys- tem even if the excited states are much larger than the Figure 5. Using quantum capacitance for dot charac- cavity frequency and for arbitrarily high valley splittings terization. (a) Simulated energy bands of lowest two states in a two dot system. versus gate voltage of induced single electron quantum dot To summarize, with a single wire we can detect if there of Figure 3d. Although the curvature is very small, it exists is an electron in the dot and measure the valley split- and may be measurable [21]. (b) Energy curvature as a func- ting given some rather stringent requirements. There are tion of magnetic field near the valley splitting and Zeeman other ways to measure valley splitting: already demon- splitting anti-crossing (when they are equal) in a quantum dot with two electrons, where a large quantum capacitance strated techniques like photon assisted tunneling, that signal will be observed. The curvature vs. probe gate volt- involve multiple dots or tunneling to leads. Those can age and the curvature vs. magnetic field are proportional to also be realized in multi-lead systems. each other, so a large quantum capacitance signal will also be seen through the gate voltage in the reflected signal. (Figures courtesy Rusko Ruskov and Yun-Pil Shim.) Characterizing an induced quantum dot quantum computer

With the electron number in the dot known, let’s fo- It’s straight-forward to imagine a one-dimensional ar- cus on spectroscopy, or charting of the excited states of ray of dots induced by a gate-chip on a chip wafer. Il- the dot, as doing so solves our original problem of mea- lustrations of such systems are shown in Figures 6 and 7. suring the valley state across a chip non-invasively. Ref While it is possible to make two dots with just two wires, [21] proposes solutions for a single wire. At the mag- and for many years the community relied on detuning of netic field that equals the valley splitting energy, there is energy levels between dots to produce two-electron inter- an anti-crossing which results in a quantum capacitance actions, now it is understood that using the barrier gate change. With two electrons in the dot, the curvature is to control electron wave function overlap is better. It even larger and occurs in the ground state (Figure 5b). allows for operation at a sweet spot (symmetric operat- Detecting this curvature allows the valley splitting to be ing point) [56, 57] and it is less sensitive to charge noise measured by sweeping the magnetic field (or valley split- than the plunger gates above the dots [58] (the exchange ting via Ez over a smaller range). Doing precise spec- interaction is less sensitive to the tunnel barrier than to troscopy requires a relative energy scale. The magnetic the detuning, so charge noise is minimized by using only field provides that if it can be well calibrated. Another high speed lines on the barrier if possible). Doing so will 6 allow us to measure dot qubit quantum properties such be strong [64]. as coherence times and even do quantum operations, that Another option for qubit readout, as advocated by us is, make a small quantum computer. recently [44], is to attempt quantum curvature readout Let’s quickly go through a sequence of experiments deep in the (1,1) regime, where the curvature of the sin- that would characterize the wafer in question (Figure 8). glet and triplet states is detected within the S-T relax- With a multi-wire device it is natural and preferable to ation time [30, 44]. This approach has the benefit of being use a bath gate to load electrons. In this context our first quantum non-demolition (the qubit is preserved) and the experiment is to chart out the charge stability diagram symmetry of the dot decreases sensitivity to charge noise for loading electrons from zero. To do this we vary VP 1 and increases S-T relaxation time (because the transi- versus VT 1, where VP 1 is the dot gate and VT 1 is a tunnel tion dipole matrix elements between S and T vanishes). gate between the bath and the dot. One should see lines It should also be noted (although this is the first place in this plot indicating the transitions between n 1 and n we have noted it), that this approach has some immunity − electrons (Figure 8a). (The straighter the line the better, to temperature - so may be the most compatible readout it indicates small cross capacitance of the gates. Curvi- approach for high temperature qubits in relatively small ness means changing cross capacitance which is very bad magnetic fields. because the electron is moving. It also makes it harder Exchange is the fundamental interaction between elec- to dynamically compensate for such gates [59, 60].) Once trons related to the Pauli exclusion principle that allows we can reliably load single electrons, then charge stability for fast two-qubit gates with large ON/OFF ratios. As diagrams (P vs. P, see Figure 8b) would be performed changing the gate potentials results in the electron wave to map out the parameter space of the two dot system. functions in the two dots overlapping, the spin state Then we can do quantum measurements. of the combined system evolves. If this interaction is To do quantum coherent measurements we need to ac- timed just right a given two-qubit spin operation can tually measure the spin qubits. Incorporating readout al- be achieved resulting in an entangling gate between the lows one to experimentally determine the coherence time qubits [65]. Primitive quantum operations can be per- (T2) and lifetime (T1) of the qubits as well as the error formed by measuring Rabi oscillations when the tunnel rates of one and two-qubit operations with the right pulse barrier is lowered to turn on the exchange interaction. sequence. Combining quantum capacitive readout [61] The latter, if done at the symmetric operating point, al- with Pauli-blockade gives a proven means of differentiat- lows one to characterize the charge noise of the device ing the singlet versus triplet states of the two dots [62]. as once you turn on exchange, the spins are no longer If you detune the dots into the Pauli-blockade regime “protected” to noise on their wave functions. More os- (where the (1,1) state equals the (0,2) state of the dou- cillations are better. A useful variant of this latter ex- ble dot system), one can distinguish singlet from triplet: perience is to perform a fingerprint plot (see Figure 8c). the singlet state will tend to allow two electrons to go into the lower dot, when the combined the dots are in the triplet state, they will each stay in their respective dots as the transition won’t be allowed [63]. This is sen- Caveat Emptor sitive to a readout window given by the temperature and the singlet-triplet relaxation time, but the signal can still Ideally, not only would this proposal enable easier ex- periments, the separation of material optimization from qubit formation and control will allow us to push the 13 lines supporting 6 dots limits of possible fidelity, or yield, or valley splitting, or with barrier whatever is limited by the material properties. In prac- gates. tice, there are concerns in translating knowledge gained Fan out to contact pads from our gate-chip induced dots as compared to the “real Bath Bath and/or gate DC gate DC thing.” Even assuming the wafer has been optimized bias/readout/con bias bias trol circuitry fully, there could be drastically different results when the dots are fabricated in a more scalable manner. Materials science still matters. Fabricating the gates with the associated interfaces, dielectrics, processing steps (e.g., anneals) will affect critical parameters like Bath gate Bath gate valley splitting and charge noise. Complex surface physics due to the passivation of the silicon/silicon- Figure 7. 6 quantum dot “testbed”. (a) On a single chip germanium surface can create unwanted potentials on the wires to support 6 dots with barrier gates in between, with quantum dot plane in the quantum well case or a charge bath gates, can be placed in a single layer of metal, along with noise environment different than or worse than a full gate associate electronics. With six dots you can create two 3-dot stack. Our hope is that the surface can be treated in such encoded qubits and test out many one and two-qubit encoded a way to minimize negative impact. gate protocols. The gate-chip and dot wafer may not be intimately 7

is valid when interfaces proximal to gates are the dom- (a) 20 a inant sourceb of noise [15]. In this context, noise sensi- c tivity may be reduced by biasing the device to a “sweet 10

(mV) D

! spot” where small changes(1,0) in gate voltages only(1,1) weakly / alter the strength of the exchange interaction. Previ- ! 0 ously explored methods include using a triple quantum 3 dot with balanced exchange interactions [21, 22] (see the -10 P2 time

Detuning V Supplementary Material for a comparison), operating far -20- P1 2 P2 time from the (1,1) charge regime where excited states flatten P1 2 2 -300.00 0.25 0.50 0.75 1.00 V mentary Material,V for large detuning dJ/d J /tc . (b) Evolution time (µs) theThus, exchange as J is profile increased [22–24], by detuning, using double-dots it| becomes populated| ⇡ quadrat- 1.00-20 1 withically more more than susceptible two electrons to charge [25],or noise. tailoring When exchange =0, =V

however, the dominant derivative is now dJ/dVX1 = -10D derivatives via magnetic field gradients [26]. The strategy (mV)

!

(@J/@tc)(dtc/dVX1), proportional(0,0) only to J.(0,1) Increasing / 0 ! 0.75 weJ pursuewith tc inthen this only Letter linearly has the increases advantage susceptibility of employing to 0 charge noise. (This scaling is valid when J is exponen- only pairwise exchange without requiring high or inho- 10 tially dependent on VX1; we later show that it can be mogeneous magnetic fields and maintains tunability of Detuning dI/dVP1 sub-exponential anddI thus/dV evenP1 more favorable.) 0.5020 theThe exchange shape coupling of the Rabi rate oscillations from being shown negligibly in Fig. small 2(b) 30 can be modeled with a two-channel decay process. One Singlet probability toprocess many GHz. is due to the hyperfine interaction between the 0 20 40 60 80 100 0.25 Exchange gate voltage V X1 (mV) VT1 electronSymmetric spin operation and that device’sis diagrammed naturalVP1 in abundance Fig. 1. The of dif-29Si 0.0 0.5 V1.0T2 1.5 2.0 ferencenuclei in and chemical is described potential by Eq. between 12 of two Ref. dots 30. is The denoted other FIG. 3. “Fingerprint” plotEvolution demonstrating time (µs) the dependence of process is due to charge noise, which, for the 1/f spec- exchange on and VX1. We plot the average singlet prob- trumand of is noisepredominantly seen in these controlled devices by [13], two imposes gates labeleda Gaus- ability after evolving for 500 ns at a potential specified by P1sian and envelope. P2 in Fig. For this 1(a). illustrative For an ideal example, double the quantum relatively FIG.the axes. 2. Rabi The oscillationsz-scale is at the a SOP. same (a) as We Fig. observe 2. A exchange faint set low value of J and the natural isotopic abundance of this oscillationsof additional by fringes biasing is to present a detuning in this (y-axis) data (prominent for a given neartime dot, = ↵(V V )where↵ is the “lever arm” that (,VX1)=( 20, 100)), likely due to excited state population sample rendersP1 theP2 charge decoherence time comparable (x-axis). The middle of the plot, where =0andJ is min- to the magnetic dephasing time. In the discussion that (Supplementary Material). The device used here and in all converts voltage to chemical potential. A third gate, la- imized,subsequent corresponds figures di to↵ers the from SOP. the The device grayscale used measures in Fig. 2 sin- by follows, however, because we focus on higher values of J the addition of a screening gate [29] and the use of enriched beledin isotopically X1, controls purified the potential silicon samples, barrier that charge sets noise the rate will glet28Si probability (800 ppm 29 andSi)[13]. ranges from 100% (white) to 25% (black). atdominate which an the electron decay can envelope. tunnel, tc/h. Figure 1(b) shows (b) Evolving at = 0 reveals a time-domain Rabi oscillation For gate-referred 1/f charge noise, this envelope can eigenenergies for a double2 dot as calculated2 2 2 with a Hub-2 showing a double Gaussian decay with a 1/e time of 1.0 µs be expressed as exp( dJ/dVj t /~ ). Here, evolve for a fixed time at a point that is swept in both V j | | V VB1 VT1 VP1 VT2 VP2 VT3 bardisVB2 the model. variance Crucially, ofV theB1 noisealthoughVT1 (e.g.V theP1 theV detuning noiseT2 VP2 spectralVT3is often den-VB2 duet (parametrized to hyperfineVB1 interactions byVT1V V)P1 and andVT2 1..V5 ThisµP2s dueV “fingerprint”T3 to chargeVB2 noise. plot P c X1 usedsity to integrated control J over, the relevant tunnel coupling timescales)t can and alsoj indexes mod- shows fringes due to varying J, the curvature of which all gates; see Ref. 23 and the Supplementaryc Material. indicates where dJ/d = 0. This locus of points, which ifyIncreasing the energy the di Rabi↵erence decay between time for the fixed singletJ can and then triplet be on this plot is approximately parallel to the x-axis due to 2 accomplished by reducing dJ/dVj [27]. We define our orthogonalization scheme, is known as the symmetric energy eigenstates, J(,tc). Inj | particular,| J( =0,tc) measurement (see Fig. 2(a) of Ref. 13). In the exper- Figure 8. Multi-dot experimentsthe that insensitivity characterizeas the system (a) Loadingaxis electrons and indicates into the a location single of dot the by SOP changing for a given theJ. I P is a “sweet spot” where the e↵ects of charge noise on imentsSymmetric described operation below, maximizes we prepare the. To qubit demonstrate state by plunger gate above the dot versus the tunnel gate from the electron2 bath, also called the charge stability diagramI of a dot (P = J/ dJ/dVj , (1) biasingthis, we near choose the various (1,0,1)-(1,0,2) combinations charge oftransition and t wherec where a exchange are reducedI because j@|J/@ =| 0 [27, 28], as ev- J/h = 160 MHz, shown in the inset of Fig. 4(a). At each vs. T plot). (b) Charge stability diagram of a doubler dotX system for a given tunneltwo-electron barrier singlet choice state (P is preferentially vs. P plot). loaded (c) into Upper dot identwhich from has Fig. units 1(b). Weof refervoltage. to J( =0 With,tc)this as a sym- met- evolution point, we explicitly measure by determining I plot shows coherent exchange oscillationsmetricric, the operating for expected a given point envelope tunnel(SOP) because ofbarrier Rabi thechoice oscillations double quan- asa is function3.how One the of ofRabi the detuning electrons oscillation is between frequency then transferred changes the dots. into due dot to At small 2 bythe 2 2 tumexp dot(Jt/ is biased~) (V/ to) the, centerso that of the the number (1,1) charge of oscilla- cell rampingperturbations P2 and in P3. each We of thedefine seven this relevant singlet stategate voltages. between symmetric operating point (SOP) ortions zero that detuning occur beforeI of the the amplitude two dot decays potentials by 1/e is the numberWe plot the of resulting coherent insensitivity oscillations in Fig. is4(a) enhanced and find ⇥ ⇤ and/(2 equidistant⇡ ). As from longthe as (2,0)is andnot too(0,2) strong anti-crossings. a function dotsthat 2 it and is maximized 3 as the north at = pole 0 with of a aqubit value Bloch of 30 sphere mV [6, 57]. The lower plot shows a so-calledI “fingerprint”V plotV demonstrating the dependence of exchange on ∆ and VT 2.⇠ In this ofAlthough control parameters any system (e.g. of exchange-coupled noise not varying quantum from one [11].and rapidly Exchange decreases between to less dots than 1 and 5 mV 2 occurs for large at a. fre- gate to the next), we can optimize device performance by To test the validity of as a metric for the e↵ects of plot the average singlet probability is shown after evolving for 500 ns at a potential specified by theI axes [6]. (Both c plots dotmaximizing pairs couldwith potentially respect benefitto V. In from particular, symmetric only op- the quencycharge noise,J(,t wec)/h measure= J(V the)/h,where numberV ofdenotes Rabi oscillations the gate I courtesy HRL.) charge noise variance and not the detailed structure of its voltages.NRabi J This⌧/h interactionthat occur inrotates a 1/e the decay qubit time state⌧.Ifthe about eration, we use Si-based triple-quantum-dot devices for ⌘ 2 spectral density is relevant to this calculation, enabling gate-referred charge-noise variance V were equal and un- ourpredictions demonstration. of bias-dependent A SEM image charge of a device noise performance is shown in ancorrelated axis which on allis tipped gates, 120then from we would the north expect pole [11–13].N . I / Rabi Fig.based 1(c). on Metallic device electrostatics. plunger gates P1-P3 are deposited on AfterTo determine some evolution,NRabi, we we measure measure⌧ thealong qubit the voltagestate using arc In order to demonstrate the advantage of symmetric Pauliwhere blockadeJ(,tc)/h by biasing= 160 MHz.near the The (1,0,2)-(1,1,1) resulting NRabi chargeis anoperation, undoped we Si/SiGe must heterostructure.independently control When theand plungerst .The plotted in Fig. 4(b). Though it qualitatively follows connected (although they could be mounted that way). nect arraysc in a 1.5 dimensional geometry. To keep thingsI areplunger forward and biased, exchange individual gates a↵ electronsect both are parameters drawn from due transition.and is maximum Sweeping near the evolution= 0, the time quantities yields Rabi are not os- Therefore, shaking of the gate-chipato bathwill capacitive andresult accumulate cross-talk. in the beneath mov- Using thea routine plungerssimple described [13, we 16]. have in cillationsstrictly discussed proportional, which one havelayer a indicating maximum of that metal contrast our assumptions on of 75% the due gate are to the Supplementary Material, we orthogonalize these con- not fully supported. In particular, as discussed in the the tilted rotation axis. ing of the electrons (with some symmetry).Thetrol X axes. and ModulationT gates We are have of depositedtc notis accomplished on anchip insulating by but changing thelayer conceptSupplementary can get Information, more complicated by including some (wafer knowl- to thatVX1 overlapsalong with the smallplungers compensating and control voltages tunnel barriers on plunger be- edgeWe of demonstrate the relative geometriessinglet-triplet of the Rabi P and oscillations X gates in seriously considered the possible implicationsgates, while is of modified this vi-by biasingwafer P1 and integration P2 with this instead device, we of can perpendicular more accurately model chips;NRabi multiplewith a tweenapproximately the dots and equal to the and electron opposite bath. voltages. Some devices We show in Fig.generalized 2(a) by definitionsweeping the of exchange(blue crosses duration in Fig. and 4(b)). while I bration on dot and quantum operationthe e↵ectparameters, of these parameters al- on J inlayers Fig. 3, whereon the we perpCharge chip, noise etc) is if not the the basic only source idea of pans degradation out in our study di↵er from Ref. 13 by the addition of a metal holding tc constant. The Rabi frequency is given by though any length scales of movementscreening gate are which likely prevents much chargeexperiments. accumulation un- J(,tc)/h and is large even with = 0 because VX1 c 2015 HRL Laboratories, LLC. All rights reserved. 3 larger than the dots, and a much lowerder gate relevant leads [29]. frequency A proximal dot charge sensor formed is forward-biased during evolution, increasing tc. J in- by the M and Z gates enables single-shot readout of the creases with , producing a chevron pattern. The compared to gate speed (nanoseconds).qubit state [13]. There P and can X gate also control lines are capable of number of resolvable| | oscillations is greatest at the SOP be a vacuum penalty: if the wirenanosecond is too far pulse from rise the times sur- and amplitudes of 140 mV. ( =What 0), giving would preliminary I do? indication that using a face than the potential of the dotThe candevices be are washedoperated in out. a dilution refrigerator, giving SOP can enhance the quality of the exchange interac- T 100 mK. tion. This improvement can be interpreted in the context Our point design simulations indicatee ⇠ that 10 nm sep- The third dots in our devices enable initializationPhysics. and Usingof gate-referred the gate-chip charge noise. approach As discussed canin the Supple-enable aration still allows for sufficient dot confinement (while many relevant and interesting experiments. STM tunneling typically occurs ~1c 2015 nm HRL from Laboratories, the surface LLC. All rights reserved.1) Let’s characterize quantum-relevant wafer proper-2 distances, ~10 nm can be gauged with a field emission ties, especially valley splitting, charge noise, and disor- current). der, across enough wafers and “devices” to be statistically The approach still requires dilution fridge tempera- conclusive. tures unless the qubits can operate and be operated on 2) Investigate proposed qubit approaches, encodings, at higher temperature (more below). Qubit operations and operation protocols. There are too many unexplored (e.g., an encoded CNOT gate made up of 20+ pulses) are proposals to enumerate, but we still don’t understand in just as difficult control-wise. Cross capacitance may be practice which qubit encodings offer the best trade-offs better or worse as compared to the “metal everywhere” for qubit quality and classical overhead (number of dots, current preference. Metal shields above and below the pulses, etc). In particular, I’d love to see our proposals dot gates may be needed to decrease cross capacitance for encoded qubit interconversion and noise-insensitive or improve screening. always-on, exchange-only qubits validated or dismissed. Only a linear array of qubits is possible which limits A 6-8 dot device would be sufficient to implement the scalability. Introducing longer distance couplers can con- vast majority of qubit and gate proposals. 8

3) Explore alternative readout and coupling ap- proaches: transverse versus longitudinal coupling, cavity- QED versus quantum capacitance. Already discussed Superconducting cavity control/readout electronics above. Bath Bath gate DC gate DC 4) Investigate different materials for their relevance to bias bias quantum computing: Optimizing valley splitting: various proposals have been made to increase valley splitting. Because all of them depend on the microscopic details of the heterostructure stack, many devices will need to Figure 9. Superconducting cavity coupling of 2 en- Holes be measured to have confidence in a solution. : coded dot qubits. It’s just as easy to put a superconducting Holes exist at the Gamma-point of the valence band, so metal on the gate-chip. there are no valley splitting issues (although there may be spin-orbit bands nearby). Germanium: germanium has a lower effective mass for holes (as compared to electrons ample, tend to have Q’s <‌< 100,000 as compared to mil- in silicon) which would relax the gate wire pitch require- lions achieved on clean sapphire or silicon wafers. A sim- ments (and sensitivity to disorder); using holes may offer ilar approach can be made with a flip-chip resonator and larger spin-orbit coupling as well as no complicating val- a traditional dot chip, but our approach should make fab- ley splitting physics. III-Vs: Although III-Vs suffer from rication much simpler. It should also allow networking of spinful nuclei, they offer a benefit of a direct band gap. small qubit registers enabling a 1.5D quantum geometry. The gate-chip approach would allow continued research in III-Vs for optical conversion or for other materials, such as GaN. II-VIs: II-VIs offer the potential for quan- End Speech tum well dots with spin-0 nuclei and a direct band-gap, they are notoriously difficult to fabricate. ZnO and other The introduction to virtually every silicon qubit paper oxide-based 2DEGs have shown inklings of relevance to goes something like this: silicon quantum dot spin qubits quantum devices. 2D materials such as graphene and provide a promising platform for large-scale quantum Van der Walls heterostructures (layers of 2D materials) computation because of their compatibility with con- offer a very large phase space of possible dot implemen- ventional CMOS manufacturing and the long coherence tations (with such materials, loss may be minimized), times due to enriched 28Si material and low spin-orbit the approach here would greatly accelerate exploration coupling, along with the possibility of high-density due of such materials. Topological materials: Many topolog- to small quantum dots. The future of silicon quantum ical materials are fragile to lithographic and gate pro- computing is strong, more-so given recent progress. Our cessing. In proximitized superconducting-semiconductor difficulty has been that we must immediately go to the stacks, there are opportunities to explore different ap- final dot dimensions just for the dots to work, we can’t proaches to qubit formation via the split-chip approach push it off. Qubits need to be small, materials need to be presented here. right, and microscopic effects matter immediately. Many 5) Study high temperature qubits: spin qubits continue of these problems will eventually need to be addressed in to have long coherence times even at elevated tempera- superconducting qubits, but now you can avoid them to tures relative to the Zeeman splitting [24, 35, 66, 67]. make progress. An open question is how robust a 2-qubit gate can be at It’s become all the rage to be building “quantum elevated temperatures (350 mK or 1-4K). testbeds”. These testbeds put as many of the best qubits 6) Search for non-QC applications of these small quan- we have today together in order to run small algorithms, tum systems, such as quantum dot-based SETs for volt- to achieve quantum supremacy [68]! This will be excit- age standards, this approach may allow for far easier fab- ing, and inconclusive, for some time. Going forward on rication and potentially better charge noise characteris- this path is obviously necessary, and also toward the first tics. true quantum error corrected logical qubit. Here I have in mind a different form of quantum testbed. My testbed can be used to improve or assess new materials stacks for One more thing: coupling spins to a superconducting cavity qubits. It can be used to build few qubit system to test new designs. It is optimal for a materials-design-test cy- cle. Because it separates qubit design (gate structure) We’ve already discussed using superconducting res- from wafer growth, it is easy to make both better simul- onators for readout, we can go further by exploring qubit taneously. It’s also easy to shift to a completely different entangling protocols via superconducting cavity or trans- type of material (e.g. holes in germanium versus elec- mission lines. By putting the cavity on the gate-chip, trons in silicon) with at most a gate pitch change. There’s see Figure 9, we can optimize for high Q. Certain en- also no reason one can’t use this approach to make small tangling protocols may benefit from high-Q resonators. quantum computers. Resonators deposited on typical SiGe dot wafers, for ex- In summary, please find a way to make and measure 9 more semiconductor qubits. about this idea with me on her summer visit to LPS in 2016 as part of her National Physical Sciences Con- sortium (NPSC) graduate fellowship. Thanks to Bob ACKNOWLEDGMENTS Butera, Matt Borselli, Gil Herrera, Bruce Kane, Will Oliver, Ben Palmer, Rusko Ruskov, and Yun-Pil Shim My highest thanks to my close collaborators Yun-Pil for critical reading and comments and special thanks to Shim and Rusko Ruskov whose foundational work is re- Tim Sweeney for useful early conversations and motiva- alized in Ref [21]. Thanks to Hilary Hurst who thought tion.

[1] 50 milli-Kelvin or about 60 times colder than deep space. [14] Meaning if you make them, they typically work if the [2] R. Hanson, L. P. Kouwenhoven, J. R. Petta, S. Tarucha, right handling precautions are taken. and L. M. K. Vandersypen, Reviews of Modern Physics [15] H. Paik, D. I. Schuster, L. S. Bishop, G. Kirch- 79, 1217 (2007). mair, G. Catelani, A. P. Sears, B. R. Johnson, M. J. [3] M. G. Borselli, K. Eng, R. S. Ross, T. M. Hazard, K. S. Reagor, L. Frunzio, L. Glazman, S. M. Girvin, M. H. Holabird, B. Huang, A. A. Kiselev, P. W. Deelman, L. D. Devoret, and R. J. Schoelkopf, Physical Review Warren, I. Milosavljevic, A. E. Schmitz, M. Sokolich, Letters 107 (2011), 10.1103/PhysRevLett.107.240501, M. F. Gyure, and A. T. Hunter, Nanotechnology 26 arXiv:1105.4652. (2015), arXiv:1408.0600 [cond-mat]. [16] Note that “gates” is used in two ways. First, there are [4] D. M. Zajac, T. M. Hazard, X. Mi, K. Wang, and J. R. quantum gates or operations, these are operations that Petta, Applied Physics Letters 106, 223507 (2015). change the state of the qubit. Second, in reference to [5] M. Veldhorst, C. H. Yang, J. C. C. Hwang, W. Huang, physical metal wires, or gates, with some voltage poten- J. P. Dehollain, J. T. Muhonen, S. Simmons, A. Laucht, tial, a terminology that comes from the semiconductor F. E. Hudson, K. M. Itoh, A. Morello, and A. S. Dzurak, transistor community. Nature 526, 410 (2015), arXiv:1411.5760. [17] D. Rosenberg, D. Kim, R. Das, D. Yost, S. Gustavs- [6] M. D. Reed, B. M. Maune, R. W. Andrews, M. G. son, D. Hover, P. Krantz, A. Melville, L. Racz, G. O. Borselli, K. Eng, M. P. Jura, A. A. Kiselev, T. D. Ladd, Samach, S. J. Weber, F. Yan, J. Yoder, A. J. Kerman, S. T. Merkel, I. Milosavljevic, E. J. Pritchett, M. T. and W. D. Oliver, npj Quantum Informationvolume 3 Rakher, R. S. Ross, A. E. Schmitz, A. Smith, J. A. (2017), arXiv:1706.04116 [quant-ph]. Wright, M. F. Gyure, and A. T. Hunter, Physical Review [18] B. Foxen, J. Y. Mutus, E. Lucero, R. Graff, A. Megrant, Letters 116, 110402 (2016). Y. Chen, C. Quintana, B. Burkett, J. Kelly, E. Jeffrey, [7] M. Rudolph, P. Harvey-Collard, R. Jock, N. T. Jacob- Y. Yang, A. Yu, K. Arya, R. Barends, Z. Chen, B. Chiaro, son, J. Wendt, T. Pluym, J. Dominguez, G. Ten-Eyck, A. Dunsworth, A. Fowler, C. Gidney, M. Giustina, R. Manginell, M. P. Lilly, and M. S. Carroll, IEEE In- T. Huang, P. Klimov, M. Neeley, C. Neill, P. Roushan, ternational Electron Devices Meeting (IEDM) , 34.1.1 D. Sank, A. Vainsencher, J. Wenner, T. C. White, and (2016), arXiv:1705.05887 [cond-mat]. J. M. Martinis, arXiv:1708.04270 [physics, physics:quant- [8] F. Martins, F. K. Malinowski, P. D. Nissen, E. Barnes, ph] (2017), arXiv:1708.04270 [physics, physics:quant- S. Fallahi, G. C. Gardner, M. J. Manfra, C. M. Marcus, ph]. and F. Kuemmeth, Physical Review Letters 116 (2016), [19] C. L. Arrington, K. S. McKay, E. D. Baca, J. J. Cole- 10.1103/PhysRevLett.116.116801, arXiv:1511.07336. man, Y. Colombe, P. Finnegan, D. A. Hite, A. E. Hol- [9] D. M. Zajac, A. J. Sigillito, M. Russ, F. Borjans, J. M. lowell, R. Jördens, J. D. Jost, D. Leibfried, A. M. Rowen, Taylor, G. Burkard, and J. R. Petta, Science 359, 439 U. Warring, M. Weides, A. C. Wilson, D. J. Wineland, (2018), arXiv:1708.03530 [cond-mat, physics:quant-ph]. and D. P. Pappas, Review of Scientific Instruments 84, [10] T. F. Watson, S. G. J. Philips, E. Kawakami, D. R. 085001 (2013). Ward, P. Scarlino, M. Veldhorst, D. E. Savage, M. G. [20] D. A. Hite, K. S. McKay, S. Kotler, D. Leibfried, D. J. Lagally, M. Friesen, S. N. Coppersmith, M. A. Eriks- Wineland, and D. P. Pappas, MRS Advances 2, 2189 son, and L. M. K. Vandersypen, Nature 555, 633 (2018), (2017), arXiv:1701.04814. arXiv:1708.04214 [cond-mat, physics:quant-ph]. [21] Y.-P. Shim, H. Hurst, R. Ruskov, and C. Tahan, Appl. [11] J. Koch, T. M. Yu, J. Gambetta, A. A. Houck, D. I. Phys. Lett. 114 (2019). Schuster, J. Majer, A. Blais, M. H. Devoret, S. M. Girvin, [22] P. Y. Yu and M. Cardona, Fundamentals of Semiconduc- and R. J. Schoelkopf, Physical Review A 76 (2007), tors: Physics and Materials Properties (Springer, 2001). 10.1103/PhysRevA.76.042319, arXiv:cond-mat/0703002. [23] W. Kohn and J. M. Luttinger, Phys. Rev. 98, 915 (1955). [12] J. A. Schreier, A. A. Houck, J. Koch, D. I. Schuster, [24] C. Tahan and R. Joynt, Phys. Rev. B 89, 075302 (2014). B. R. Johnson, J. M. Chow, J. M. Gambetta, J. Ma- [25] A stack of differing crystals or crystal alloys. jer, L. Frunzio, M. H. Devoret, S. M. Girvin, and R. J. [26] D. Loss and D. P. DiVincenzo, Physical Review A 57, Schoelkopf, Physical Review B 77 (2008), 10.1103/Phys- 120 (1998). RevB.77.180502, arXiv:0712.3581. [27] B. E. Kane, Nature 393, 133 (1998). [13] A. A. Houck, J. Koch, M. H. Devoret, S. M. Girvin, and [28] The notion of encoded or “decoherence-free subspace or R. J. Schoelkopf, Quantum Information Processing 8, 105 subsystem” (DFS) qubits is not exclusive to spins. DFS (2009), arXiv:0812.1865. encoding provides some immunity to global fields or er- 10

rors, although there is significant experimental evidence Dzurak, Scientific Reports 1 (2011), 10.1038/srep00110, that the dominant noise sources affecting dot qubits cur- arXiv:1012.1410. rently are local to each dot. The main benefits for spins [44] R. Ruskov and C. Tahan, Phys. Rev. B 99, 245306 are: 1) spins are small, so microwave gates can result (2019), arXiv:1704.05876 [cond-mat]. in a lot of cross talk, 2) microwave gates (driving sin- [45] D. Reilly, C. M. Marcus, M. P. Hanson, and A. C. Gos- gle dot single spin rotations with a magnetic field) are sard, Appl. Phys. Lett. 91, 162101 (2007). slow, 3) two qubit gates via the exchange interaction are [46] K. D. Petersson, C. G. Smith, D. Anderson, P. Atkinson, fast. Terminology-wise I prefer 2-DFS, 3-DFS, etc. as de- G. Jones, and D. A. Ritchie, Nano. Lett. 10, 2789 (2010). noting encoded qubits where the dots are isolated from [47] J. I. Colless, A. C. Mahoney, J. M. Hornibrook, A. C. each other when not undergoing (pair-wise only) encoded Doherty, H. Lu, A. C. Gossard, and D. J. Reilly, Physical qubit operations. Other terminology is often used such Review Letters 110, 046805 (2013). as singlet-triplet qubits (for double dots) and exchange- [48] M. F. Gonzalez-Zalba, S. Barraud, A. J. Ferguson, and only (for triple dots). More recently the community has A. C. Betz, Nature Communications 6, ncomms7084 been revisiting using small clusters of spins which form (2015). molecules with always-on exchange couplings between [49] M. F. Gonzalez-Zalba, S. N. Shevchenko, S. Barraud, dots starting with the resonant exchange (3 dot) qubit J. R. Johansson, A. J. Ferguson, F. Nori, and A. C. (single sweet spot) and derivatives like the AEON qubit Betz, Nano Letters 16, 1614 (2016). (double sweet spot). Because the exchange interaction is [50] A. Rossi, R. Zhao, A. S. Dzurak, and M. F. always on, these qubits are more sensitive to charge noise Gonzalez-Zalba, Appl. Phys. Lett. 110, 212101 (2017), in their off state than an isolated spin, always, but have arXiv:1610.00767. other potential advantages. [51] R. Mizuta, R. M. Otxoa, A. C. Betz, and M. F. Gonzalez- [29] B. H. Fong and S. M. Wandzura, Quantum.Inf.Comput. Zalba, Phys. Rev. B 95, 045414 (2017). 11, 1009 (2011), arXiv:1102.2909 [quant-ph]. [52] S. Schaal, S. Barraud, J. J. L. Morton, and M. F. [30] Y.-P. Shim and C. Tahan, Physical Review B 93 (2016), Gonzalez-Zalba, Phys. Rev. Applied 9, 054016 (2018), 10.1103/PhysRevB.93.121410, arXiv:1602.00320. arXiv:1708.04159 [cond-mat, physics:quant-ph]. [31] J. M. Taylor, V. Srinivasa, and J. Medford, [53] X. Mi, J. V. Cady, D. M. Zajac, P. W. Deelman, and Physical Review Letters 111 (2013), 10.1103/Phys- J. R. Petta, Science 355, 156 (2017), arXiv:1703.03047. RevLett.111.050502, arXiv:1304.3407. [54] N. Samkharadze, G. Zheng, N. Kalhor, D. Brousse, [32] Odd numbers of electrons keep the overall spin 1/2 but A. Sammak, U. C. Mendes, A. Blais, G. Scappucci, and also lower the excited energy levels which can cause com- L. M. K. Vandersypen, Science 359, 1123 (2018). plications in gates. [55] G. Burkard and J. R. Petta, Physical Review B 94 [33] Effective mass theory lets us throw many of the complex- (2016), 10.1103/PhysRevB.94.195305, arXiv:1607.08801. ities of actually being in a crystal under the rug and bun- [56] M. D. Reed, B. M. Maune, R. W. Andrews, M. G. dle that up in a new mass with associated hydrogenic-like Borselli, K. Eng, M. P. Jura, A. A. Kiselev, T. D. Ladd, orbital levels. S. T. Merkel, I. Milosavljevic, E. J. Pritchett, M. T. [34] G. Feher and E. A. Gere, Physical Review 114, 1245 Rakher, R. S. Ross, A. E. Schmitz, A. Smith, J. A. (1959). Wright, M. F. Gyure, and A. T. Hunter, Phys. Rev. [35] C. Tahan, Silicon in the quantum limit: Quan- Lett. 116, 110402 (2016). tum computing and decoherence in silicon architec- [57] F. Martins, F. K. Malinowski, P. D. Nissen, E. Barnes, tures, Ph.D. thesis, University of Wisconsin-Madison, S. Fallahi, G. C. Gardner, M. J. Manfra, C. M. Marcus, arxiv.org/abs/0710.4263 (2005). and F. Kuemmeth, Phys. Rev. Lett. 116, 116801 (2016). [36] A. M. Tyryshkin, S. Tojo, J. J. L. Morton, H. Riemann, [58] Y.-P. Shim and C. Tahan, Phys. Rev. B 97, 155402 N. V. Abrosimov, P. Becker, H.-J. Pohl, T. Schenkel, (2018). M. L. W. Thewalt, K. M. Itoh, and S. A. Lyon, Nature [59] T. A. Baart, M. Shafiei, T. Fujita, C. Reichl, Materials 11, 143 (2012). W. Wegscheider, and L. M. K. Vandersypen, Nature [37] F. Schäffler, Semiconductor Science and Technology 12, Nanotechnology 11, 330 (2016). 1515 (1997). [60] A. R. Mills, D. M. Zajac, M. J. Gullans, F. J. Schupp, [38] GaAs was the physicists semiconductor and III-V het- T. M. Hazard, and J. R. Petta, Nature Communications erostructures the host for the quantum hall effect, the 10, 1063 (2019). fractional quantum Hall effect (e.g., Noble prizes) and [61] The terminology of cavity-based readout of supercon- the first artificial atoms or quantum dots. ducting and semiconducting qubits is confusing. The lit- [39] T. Thorbeck and N. M. Zimmerman, 5, 087107. erature in both communities uses the term “dispersive” [40] R. Dombrowski, C. Steinebach, C. Wittneven, M. Mor- readout (because the cavity is pulled in one direction or genstern, and R. Wiesendanger, Phys. Rev. B 59, 8043 another by the state of the qubit) but the physical cou- (1999). pling between the qubit and the cavity may be completely [41] J. Salfi, B. Voisin, A. Tankasala, J. Bocquel, M. Us- different. man, M. Y. Simmons, L. C. L. Hollenberg, R. Rah- [62] Many qubit options from single spins to double and triple man, and S. Rogge, Phys. Rev. X 8, 031049 (2018), dot encoded qubits can be read out by conversion to a arXiv:1706.09261 [cond-mat]. singlets or triplets of two dots. [42] M. Friesen, P. Rugheimer, D. E. Savage, M. G. Lagally, [63] J. R. Petta, A. C. Johnson, J. M. Taylor, E. A. Laird, D. W. van der Weide, R. Joynt, and M. A. Eriksson, A. Yacoby, M. D. Lukin, C. M. Marcus, M. P. Hanson, Physical Review B 67, 121301 (2003). and A. C. Gossard, Science 309, 2180 (2005). [43] N. S. Lai, W. H. Lim, C. H. Yang, F. A. Zwanen- [64] G. Zheng, N. Samkharadze, M. L. Noordam, N. Kalhor, burg, W. A. Coish, F. Qassemi, A. Morello, and A. S. D. Brousse, A. Sammak, G. Scappucci, and L. M. K. 11

Vandersypen, Nature Nanotechnology 14, 742 (2019). [67] L. Petit, H. G. J. Eenink, M. Russ, W. I. L. Lawrie, [65] D. Loss and D. P. DiVincenzo, Physical Review A 57, N. W. Hendrickx, J. S. Clarke, L. M. K. Vander- 120 (1998). sypen, and M. Veldhorst, arXiv:1910.05289 [cond-mat, [66] L. Petit, J. M. Boter, H. G. J. Eenink, G. Droulers, physics:quant-ph]. M. L. V. Tagliaferri, R. Li, D. P. Franke, K. J. Singh, [68] Yes, I wrote the bulk of this essay back in the winter J. S. Clarke, R. N. Schouten, V. V. Dobrovitski, L. M. K. break of 2017-2018, pre-supremacy. Vandersypen, and M. Veldhorst, Phys. Rev. Lett. 121, 076801 (2018).