In short
A spin qubit encodes |0\rangle and |1\rangle in the two orientations — "up" and "down" — of the intrinsic angular momentum of a single particle, usually an electron, sometimes a nucleus. Two platforms dominate. Silicon quantum dots confine a single electron in a gate-defined well or on a phosphorus donor atom embedded in isotopically purified ^{28}Si; the electron spin is the qubit, its coherence time T_2 reaches seconds with dynamical decoupling because ^{28}Si has zero nuclear spin and therefore a silent magnetic environment. Diamond nitrogen-vacancy (NV) centres use a crystal defect — one nitrogen atom next to one missing carbon — that hosts an electron spin triplet with a spin-preserving optical transition, allowing initialisation, single-qubit gates, and readout all at room temperature using green laser light and a microwave loop. Single-qubit gates on both platforms are driven by microwave electron spin resonance (ESR) at 1–30 GHz. Two-qubit gates exploit the Heisenberg exchange interaction J\,\vec S_1 \cdot \vec S_2 between neighbouring electrons, tuned by gate voltages in silicon (Loss and DiVincenzo, 1998) or by dipolar coupling between two NV spins in diamond. Gate times: 10 ns – 1 μs. Fidelities: above 99.9% single-qubit and above 99% two-qubit on the best 2024 silicon devices. Players: Intel (Tunnel Falls, 12-qubit silicon chip, 2023), Diraq (UNSW spinout), Quantum Motion (UK), IMEC (Belgium foundry), Google Quantum AI / Seeqc (diamond NV networks), Element Six (diamond substrates). Advantages: silicon is CMOS-compatible — a fab that makes iPhone processors can in principle make spin-qubit chips; diamond runs at room temperature and doubles as a magnetometer. Challenges: each qubit has a slightly different resonance frequency (fabrication variability), measurement via spin-to-charge conversion is slower than superconducting readout, and individual-qubit addressing requires one control line per qubit, which does not scale to millions without on-chip multiplexing. In India, IIT Kanpur's Kasturi Saha group works on NV-centre sensors and qubits; TIFR Mumbai and IISc Bangalore host spin-qubit theory groups under the National Quantum Mission.
The superconducting qubit on an IBM Heron chip lives inside a dilution refrigerator the size of a bookshelf, cooled to twenty milliKelvin — colder than any point in the known universe except inside other dilution refrigerators. The trapped ion on a Quantinuum H2 chip lives in an ultra-high vacuum chamber pumped to less than 10^{-11} Torr, suspended by RF fields, imaged with precision optics. Neutral atoms demand cold-atom physics and laser tables that fill a small lab. Every one of these platforms is a serious piece of scientific apparatus.
Now look at your phone. Somewhere on its motherboard is a silicon chip with about twenty billion transistors on it, each one a switch on a two-centimetre-square die, each one fabricated by lithography at a few-nanometre resolution, each one costing a few picocents to make. The semiconductor industry has spent sixty years learning to build tiny structures on silicon very cheaply and very reliably. What if a qubit could ride on the back of that?
This is the silicon spin qubit bet. A single electron, trapped in a structure that looks remarkably like a transistor, with its spin up or down as |0\rangle and |1\rangle. No exotic superconductor. No ultra-high vacuum. A silicon chip — almost.
The other big spin-qubit bet is even more surprising. Take a diamond. Remove one carbon atom from its lattice, and replace a neighbouring carbon with a nitrogen. That crystallographic defect — a nitrogen-vacancy (NV) pair — hosts an electronic wavefunction with a ground-state electron spin that behaves, for quantum-computing purposes, almost exactly like a textbook spin-1. It can be prepared, addressed, and read out with nothing more than a green laser, a microwave loop, and a photodetector. And it all works at room temperature.
This chapter builds both platforms from scratch: the physics of an electron spin as a qubit, the silicon quantum dot, the NV centre, how you measure them, who is commercialising them, and where Indian research groups fit in. By the end you will understand why, despite trailing superconducting and trapped-ion platforms on current qubit count, spin qubits are the platform people keep betting on when they think about the path to a billion qubits.
What is a spin qubit, physically
Every electron — every one — has an intrinsic angular momentum called spin. The word is a historical accident; nothing actually spins. What is true is that the electron carries a small magnetic moment that behaves, in every equation you will ever write, like the angular momentum of a tiny gyroscope that can point in only two possible directions along any axis you choose to measure.
Call those two directions |\!\uparrow\rangle and |\!\downarrow\rangle along some chosen axis — usually the direction of an applied magnetic field \vec B_0, taken to be \hat z. In a magnetic field, the two spin orientations have different energies:
where g \approx 2 is the electron's gyromagnetic ratio (the dimensionless factor relating spin to magnetic moment) and \mu_B = 9.274 \times 10^{-24} J/T is the Bohr magneton. Plug in a laboratory field of B_0 = 0.5 T and you get a splitting of h \times 14 GHz — a microwave frequency.
Why the two spin orientations differ in energy: the electron's magnetic moment \vec \mu = -g\mu_B \vec S / \hbar in a magnetic field has energy U = -\vec \mu \cdot \vec B. Plug in \vec B = B_0 \hat z and the two S_z eigenvalues \pm \hbar/2 give energies \pm (1/2) g\mu_B B_0. Their difference is g\mu_B B_0 — the Zeeman splitting. For an NV centre the splitting is set by the crystal field even at zero external B_0; for a silicon electron it needs an applied field.
Identify these two Zeeman-split levels with |0\rangle and |1\rangle. You now have a qubit. The next question is how to drive transitions between them (single-qubit gates), how to make two of them interact (two-qubit gates), and how to measure them. The answers differ between silicon and diamond, but the physics is the same spin in both cases.
This is the bare-bones picture. The engineering problem, which differs between silicon and diamond, is: where does the single electron live, what magnetic environment does it see, and how do you read out its spin without shaking it to pieces?
Silicon quantum dots
Silicon is the most characterised material in human history. Decades of CMOS fabrication have tuned its lithography, doping, etching, and metallisation to nanometre precision, and the ecosystem — foundries, design software, test equipment — already exists at industrial scale. If you can build a qubit in silicon, you inherit all of that for free.
The trick is to trap one electron. A bulk silicon wafer is full of electrons; a nanoscale device is designed so that a carefully chosen region contains exactly one electron and nothing else.
Gate-defined quantum dots
Picture the cross-section of a standard MOS (metal-oxide-semiconductor) transistor. On top sits a stack of gate electrodes; below them, a thin oxide; below that, the silicon substrate. Apply a voltage to a gate electrode and it shapes the electrostatic potential in the silicon underneath, attracting electrons to the interface.
In a gate-defined quantum dot, the gate layout is designed so that positive voltages on some electrodes form a narrow channel where electrons can sit, and negative voltages on flanking electrodes pinch off the channel into a tiny bowl — a potential well a few tens of nanometres across. Tune the gate voltages carefully and only one electron fits in the bowl. Its wavefunction is confined to the 30-nm dot; its energy levels are quantised; its spin is the qubit.
Two neighbouring dots share a tunable tunnel barrier controlled by a middle "exchange gate." Raise the barrier high, the two electrons are isolated. Lower it, and their wavefunctions start to overlap — the Heisenberg exchange interaction J(V_{\text{exch}})\,\vec S_1 \cdot \vec S_2 turns on, and a two-qubit gate can be performed.
Phosphorus donors
An alternative, pushed by the Kane group at UNSW since 1998, uses a single phosphorus atom implanted into silicon. Phosphorus is a group-V element — silicon is group IV — so a P atom in the Si lattice has one more electron than the surrounding Si atoms. At low temperature that extra electron is weakly bound to its parent P nucleus in a hydrogen-like orbit of radius about 3 nm, with a binding energy of ~45 meV. The donor's spin is the electron spin qubit; the underlying P nucleus carries a nuclear spin of its own (I = 1/2 for ^{31}P) that can serve as a second qubit with coherence times of tens of seconds.
Fabricating single-atom devices is a tour de force: Michelle Simmons's group at UNSW uses a scanning-tunnelling-microscope tip to pattern hydrogen-passivated Si surfaces with atomic precision, dose PH_3 gas that chemisorbs at the exposed patches, and anneal so that one P atom ends up in each desired position. The resulting devices have demonstrated donor-based qubits with T_2 > 30 seconds — the longest coherence times of any electronic qubit on any platform.
Why ^{28}Si matters
Natural silicon is 92% ^{28}Si (nuclear spin I = 0, magnetically silent) plus 5% ^{29}Si (nuclear spin I = 1/2, magnetically noisy). The 5% of noisy nuclei create tiny fluctuating magnetic fields that dephase an electron spin sitting nearby — a random walk of the qubit phase due to nuclear-spin flipping. Isotopically purify the silicon to 99.99% ^{28}Si (leaving only 0.01% ^{29}Si) and the nuclear-spin bath shrinks by a factor of 500; the electron spin's dephasing time jumps from microseconds in natural Si to seconds in purified ^{28}Si. Every high-coherence silicon spin-qubit experiment since 2014 has used purified material.
Single-qubit gates by ESR
A resonant microwave magnetic field perpendicular to \vec B_0 drives Rabi oscillations between |0\rangle and |1\rangle — the standard electron spin resonance trick. The drive is delivered by a small microwave wire fabricated alongside the qubits, or (cleverly) by a local micromagnet that converts a microwave electric field into an effective microwave magnetic field at the dot, a technique called EDSR (electric-dipole spin resonance) pioneered by the Petta group at Princeton. Gate times are 10–200 ns; fidelities above 99.9% have been reported on single ^{28}Si qubits since 2022.
Two-qubit gates by exchange
Lower the middle gate voltage and two neighbouring electrons' wavefunctions overlap. In the overlap region, Pauli exclusion forces any spatially symmetric two-electron wavefunction to have an antisymmetric spin state — the singlet — and vice versa. The energetic difference between the singlet (|\!\uparrow\downarrow\rangle - |\!\downarrow\uparrow\rangle)/\sqrt 2 and the triplet |\!\uparrow\uparrow\rangle, |\!\downarrow\downarrow\rangle, (|\!\uparrow\downarrow\rangle + |\!\downarrow\uparrow\rangle)/\sqrt 2 is the exchange energy J. Turn J on for a controlled duration, and the two-qubit unitary evolves through a SWAP-like rotation — the \sqrt{\text{SWAP}} gate of Loss and DiVincenzo (1998), which combined with single-qubit gates gives universal quantum computing.
Example 1: The Loss–DiVincenzo two-qubit gate
Work through exactly what the Loss–DiVincenzo exchange gate does to two electron spins in neighbouring quantum dots.
Step 1. Write the Hamiltonian. Two electrons with spins \vec S_1 and \vec S_2 in dots 1 and 2 couple by the Heisenberg exchange:
where J(t) is a voltage-controlled coupling that is zero when the middle gate is high and turns on to a desired value when the gate is lowered. Why: Pauli exclusion forbids two electrons from occupying the same orbital state with the same spin. When two dots share a wavefunction, the energetic preference for the (antisymmetric) singlet over the (symmetric) triplet is exactly the exchange energy, which is well-approximated by J\,\vec S_1 \cdot \vec S_2 when the dots are weakly coupled.
Step 2. Express H_{\text{ex}} in the computational basis. Using \vec S_1 \cdot \vec S_2 = \tfrac{1}{2}(\vec S_1 + \vec S_2)^2 - \tfrac{3}{4}\hbar^2 and labelling |\!\uparrow\rangle = |0\rangle, |\!\downarrow\rangle = |1\rangle:
in the ordered basis \{|00\rangle, |01\rangle, |10\rangle, |11\rangle\} (with \hbar = 1 and an overall shift absorbed into the identity). Why: the two outer-diagonal states |00\rangle and |11\rangle are already triplets (eigenstates of total \vec S^2), so exchange only shifts them by a constant. The middle 2\times 2 block couples |01\rangle and |10\rangle — those two combine into the singlet (eigenvalue -3J/4) and triplet (eigenvalue J/4), which is precisely the Pauli-driven splitting.
Step 3. Evolve for a time t. The time-evolution operator is U(t) = \exp(-i\,H_{\text{ex}}\,t / \hbar). On the \{|01\rangle, |10\rangle\} subspace, this gives
while |00\rangle and |11\rangle each pick up the phase e^{-iJt/4}. Why: restrict H_{\text{ex}} to the two-dimensional \{|01\rangle, |10\rangle\} block; the relevant part is (J/2)\,\sigma_x (after subtracting the diagonal shift), whose exponential is the standard Rabi-like oscillation matrix.
Step 4. Choose the gate time. Set Jt = \pi/2, so \cos(Jt/2) = 1/\sqrt 2 and \sin(Jt/2) = 1/\sqrt 2. On |01\rangle: |01\rangle \to (|01\rangle - i|10\rangle)/\sqrt 2. This is an entangled state — the exchange pulse has converted a separable two-qubit input into a maximally entangled one. The resulting unitary is \sqrt{\text{SWAP}}, the square root of the full swap gate. Why: \text{SWAP} exchanges the two qubits; its square root is the gate that, applied twice, produces a full swap, but applied once produces an entangling operation. Combined with single-qubit rotations, \sqrt{\text{SWAP}} is universal — any two-qubit unitary can be decomposed as three \sqrt{\text{SWAP}}s plus single-qubit rotations.
Step 5. Make it a CNOT. To implement a CNOT from \sqrt{\text{SWAP}} you apply \sqrt{\text{SWAP}}, then single-qubit Z rotations with specific angles, then \sqrt{\text{SWAP}} again. The total two-qubit gate time on 2024 silicon hardware is ~100 ns; on phosphorus-donor devices about 1 μs. Why: the explicit decomposition is \text{CNOT} = (H \otimes I)\cdot (I \otimes R_z(-\pi/2)) \cdot \sqrt{\text{SWAP}}\cdot (R_z(\pi) \otimes I) \cdot \sqrt{\text{SWAP}} \cdot (H \otimes I) up to a global phase — a standard identity in Loss–DiVincenzo-style circuits.
Result. The exchange interaction, turned on for a calibrated duration by lowering the middle gate voltage, implements a \sqrt{\text{SWAP}} entangling gate. Two such pulses plus local rotations give a CNOT. On 2024 ^{28}Si spin-qubit chips this achieves fidelities of 99.2% (Xue et al., Nature) — competitive with superconducting transmons and approaching the surface-code threshold.
What this shows: a two-qubit gate in silicon is performed by momentarily shorting two potential wells together. The quantum physics — Pauli exclusion, spin singlet vs triplet — is exactly the physics of covalent bonding in chemistry; the engineering reuse is that you can turn it on with a voltage pulse on a lithographically defined gate, and the gate does not have to be any more exotic than the gate of a transistor.
Readout by spin-to-charge conversion
Here is the hardest part. You cannot "look at" a single electron's spin directly — no simple laser transition exists in the silicon system the way it does for an ion or an atom. The solution, called spin-to-charge conversion (or Elzerman readout, after its 2004 inventor), is to convert the question "is this spin up or down?" into the question "has this electron left the dot?", which you can answer by monitoring a nearby charge sensor.
Tune the dot's chemical potential so that the Fermi level of a nearby lead sits between E_\uparrow and E_\downarrow (achievable with gate voltages). If the electron is spin-up (higher energy), it can tunnel out to the lead; if spin-down (lower energy), it is energetically forbidden from leaving. Monitor the dot with a single-electron transistor (SET) or a quantum point contact (QPC) charge sensor — these devices have a current that depends sensitively on the number of electrons in the dot. If you see the current jump, the electron tunnelled out: it was spin-up. If the current stays put, it did not: spin-down. A few microseconds later, a fresh electron tunnels back in and reinitialises the dot.
Elzerman readout fidelities above 99% are routine; the readout time is 10–100 μs, which is slower than superconducting dispersive readout (~100 ns) but fast enough for error correction if the gate times are also slower in proportion. A newer variant — Pauli spin blockade (PSB) — uses a two-electron singlet-triplet readout that is faster, in some cases approaching the 1-μs mark.
Nitrogen-vacancy centres in diamond
Now the second platform, which looks like a completely different beast.
A natural diamond is (mostly) a perfect crystal of carbon atoms arranged in the tetrahedral diamond lattice. Occasionally — naturally or by engineered ion implantation — one carbon atom is missing, leaving a vacancy. If one of the neighbouring carbons happens also to be replaced by a nitrogen, you have a nitrogen-vacancy (NV) pair: a crystallographic defect about the size of a single atom that hosts a localised electronic state.
The NV centre comes in two charge states. The relevant one for quantum computing is NV⁻ (negatively charged, six electrons in the defect's molecular orbital). Its ground state is a spin triplet with total spin S=1: three sublevels labelled m_s = -1, 0, +1. The m_s = \pm 1 states are degenerate at zero magnetic field, while the m_s = 0 state sits about 2.87 GHz below them — a splitting called the zero-field splitting D, set by the internal crystal field of the defect itself. No external magnet required.
Identify |0\rangle \equiv |m_s = 0\rangle and |1\rangle \equiv |m_s = -1\rangle (the m_s = +1 state is parked by a small magnetic field) and you have a qubit with a built-in 2.87 GHz splitting.
Optical initialisation and readout
The NV centre has a wonderful property called the spin-preserving optical cycle. Shine green laser light at ~532 nm onto an NV⁻ centre and it drives a transition from the ground-state triplet (^3A_2) to an excited-state triplet (^3E) that preserves the m_s quantum number. The excited state relaxes back to the ground state via two channels: (1) a radiative fluorescence at ~637 nm, which preserves m_s; and (2) a non-radiative intersystem crossing through a spin singlet that selectively relaxes back into m_s = 0.
The net effect is a kind of optical pumping: repeated green-laser cycles progressively dump all the population into m_s = 0. After ~1 μs of green illumination, the spin is initialised to |0\rangle with fidelity above 95%.
Even better, the brightness of the red fluorescence differs between m_s states: m_s = 0 is bright, m_s = \pm 1 is about 30% dimmer (because the \pm 1 states leak more into the non-radiative intersystem crossing). Illuminate the NV with green, count red photons for 300 ns, and the photon count tells you whether the spin was |0\rangle (brighter) or |1\rangle (dimmer) — optical spin readout, entirely without tunnel junctions, single-electron transistors, or any electrical infrastructure. All at room temperature.
The NV centre's unique appeal
- Room-temperature operation. Unlike every other quantum-computing platform, an NV centre has T_2 up to a millisecond at 300 K. Dynamical decoupling extends this to ~1 s. No dilution refrigerator; no vacuum chamber; a benchtop setup suffices.
- All-optical single-shot readout. Photon counting gives a qubit measurement with no electrical infrastructure.
- Built-in nuclear-spin ancilla. The nitrogen nucleus (^{14}N: I=1, or ^{15}N: I=1/2) is weakly coupled to the electron spin via the hyperfine interaction; its coherence is tens of seconds. Nearby ^{13}C nuclei (1% natural abundance) in the surrounding lattice provide additional nuclear-spin ancillas. A single NV is effectively a small quantum register.
- Sensor applications. A handful of NV centres is the world's most sensitive magnetometer — used for imaging single neurons firing in a brain slice, mapping surface currents on single cells, and detecting single electron and nuclear spins.
The NV centre's weakness as a scalable computer
- No lithographic placement. Controlling where in the diamond an NV sits is hard — ion implantation gives a statistical distribution, not a precise array. "Aperture-aligned" implantation is improving, but nowhere near the lithographic precision of silicon.
- Weak inter-NV coupling. Two NV centres 10 nm apart have a dipole-dipole coupling of kHz — much weaker than the exchange in silicon (MHz). Two-qubit gates between distant NVs require photonic interconnects, a separate active research direction.
- Diamond crystal quality. Natural diamonds are full of other defects that shorten T_2; CVD-grown isotopically purified diamond is needed, supplied by Element Six and a small handful of other companies.
Example 2: Measuring an NV spin with photon counting
Walk through the optical readout of a single NV qubit.
Step 1. Prepare the qubit. Apply a 1 μs pulse of 532 nm green laser light focused onto the NV. The spin-preserving excitation cycle plus ISC pumping initialises the spin into m_s = 0 \equiv |0\rangle with ~95% fidelity. Why: the intersystem crossing from the ^3E excited state preferentially empties the m_s = \pm 1 manifold, funnelling population through a metastable singlet and depositing it into m_s = 0. Repeated cycles of green illumination therefore polarise the spin.
Step 2. Apply a microwave \theta pulse. Drive a microwave field at 2.87 GHz (the ground-state zero-field splitting) for a duration t, with Rabi frequency \Omega. After the pulse, the state is |\psi\rangle = \cos(\theta/2)|0\rangle - i\sin(\theta/2)|1\rangle where \theta = \Omega t. Why: a resonant microwave magnetic field rotates the spin on its Bloch sphere; the rotation angle is the pulse area. This is standard ESR, identical to the single-qubit gate on silicon spin qubits; the frequency differs (2.87 GHz for NV, 10–30 GHz for Si in external field) but the physics is the same.
Step 3. Illuminate for readout. Apply a 300 ns pulse of 532 nm green laser. Count the number of red (637 nm) photons emitted during a 300 ns collection window. Why: the excitation-plus-fluorescence cycle preserves spin during the brief readout window; the m_s = 0 state fluoresces at roughly 30% higher rate than m_s = \pm 1 because the \pm 1 populations leak into the ISC (which is dark) while the 0 populations keep cycling brightly.
Step 4. Interpret the photon count. A typical detection system collects ~0.1 photons per cycle from m_s = 0 and ~0.07 photons per cycle from m_s = \pm 1. A single readout window is too noisy to distinguish in one shot — the Poisson noise is comparable to the signal. Single-shot readout is recovered by repetitive measurement: map the electron spin state onto a nuclear ancilla (using a hyperfine-assisted CNOT), then read out the electron many times without disturbing the nucleus. A few hundred repetitions gives single-shot fidelity above 98%. Why: the nuclear spin has a longer coherence and does not respond to the green laser, so you can read the (correlated) electron spin hundreds of times while the nuclear ancilla holds the "true" answer. This is a form of quantum non-demolition (QND) measurement — textbook QND through ancilla repetition.
Step 5. The outcome. After integration, you have a classical bit: the original \theta-rotated qubit has been projected onto the \{|0\rangle, |1\rangle\} basis with probabilities \cos^2(\theta/2) and \sin^2(\theta/2), and the photon count has converted that choice into a classical readout. Total readout time: ~100 μs for single-shot. Why: this is the canonical projective measurement of a qubit, implemented by a specific physical protocol. The ancilla-assisted repetition converts a low-contrast single-shot into a high-fidelity one, a trick widely used on other platforms as well (e.g. trapped-ion repetitive readout).
Result. Initialise, drive, read — an NV centre runs the full qubit protocol at room temperature using nothing more than a green laser, a microwave loop, and a photon counter. The coherence time is ~1 ms (up to ~1 s with dynamical decoupling), single-qubit gate time ~10 ns, readout time ~100 μs, readout fidelity >98%.
What this shows: every step of the NV qubit — preparation, a single-qubit gate, readout — runs at room temperature using components a clever undergraduate lab group can build. This is unique among quantum-computing platforms. The consequence is that NV centres are the platform most likely to show up in classroom teaching, sensing, and QKD applications — even if they never dominate as a scalable quantum-computer chip.
Who is building spin qubits
- Intel is the largest silicon spin-qubit program. In 2023 Intel publicly shipped the Tunnel Falls chip to ~20 academic and national-lab collaborators: a 12-qubit device fabricated on a full-wafer 300 mm CMOS-compatible process at the Intel D1 foundry in Hillsboro, Oregon. The engineering thesis is explicit: use the same lithography, deposition, and test infrastructure that makes iPhone and server chips today to build qubits at wafer scale.
- Diraq (Sydney), the commercial spinout of Andrea Morello and Andrew Dzurak's UNSW group, has demonstrated ^{28}Si metal-oxide-semiconductor spin qubits with two-qubit gate fidelities above 99% and is collaborating with Global Foundries on a foundry-scale process.
- Quantum Motion (London) is developing silicon spin qubits using CMOS-compatible devices; they have published qubit demonstrations on chips fabricated in a commercial 28-nm CMOS process — the most aggressive scaling of a spin-qubit chip in an existing foundry to date.
- IMEC (Leuven) operates a European foundry that supplies CMOS-compatible silicon spin-qubit devices to dozens of academic groups.
- Element Six / Diamond Materials supply the isotopically purified CVD diamond substrates on which essentially every serious NV experiment in the world is performed.
- Google Quantum AI / Seeqc and a number of academic groups (Delft, Harvard, Stuttgart) work on NV-centre networks — using photonic interconnects between NV nodes to build a distributed quantum network rather than a monolithic chip.
- Academic powerhouses: UNSW Sydney (Dzurak, Morello), TU Delft (Vandersypen), QuTech Amsterdam, RIKEN Tokyo (Tarucha), Princeton (Petta), Sandia National Lab, Harvard (Lukin, again — his group also does NV work).
Indian context
- IIT Kanpur's Kasturi Saha group runs an NV-centre lab: quantum sensing and qubit-level experiments on diamond. The group has published on NV magnetometry and is actively recruiting for NQM-funded projects.
- TIFR Mumbai hosts theoretical and experimental spin-qubit work (condensed-matter theory of silicon qubits and donor-based systems).
- IISc Bangalore (Centre for Nano Science and Engineering) has NV-centre fabrication and NV-based sensing experiments.
- National Quantum Mission (2023, ₹6000 crore over 8 years) funds both silicon and NV programs, with specific milestones for a 10-qubit NV demonstrator by 2028 and a silicon spin-qubit demonstrator at IIT-Bombay/TIFR on a similar timeline. The budget allocated to the matter-qubit pillar is substantial — perhaps ₹800 crore across silicon and NV combined.
- IIT Bombay and IIT Delhi have smaller spin-qubit experimental and theoretical efforts, including magnetic-material qubit concepts.
The Indian spin-qubit community is growing quickly but remains small relative to the US and European scene — perhaps 50 active researchers across academic and industry groups in 2026.
Common confusions
- "Spin qubits are the same as spintronics." Different. Spintronics (hard drives, magnetic RAM) uses the spin of large collections of electrons to make classical memory and logic. Spin qubits use the spin of one electron, preserve quantum coherence, and support superposition and entanglement. The physics overlaps; the goal and the engineering do not.
- "A silicon quantum dot is just a transistor." Close but not quite. A transistor channels a current of many electrons; a quantum dot isolates exactly one. The lithography is similar (sub-50-nm gates, CMOS-compatible stacks), but the dot is operated in a few-electron regime, deliberately tuned so that adding one more electron costs a large charging energy and the device has a single controllable wavefunction. Transistors are designed to hide quantum effects; quantum dots are designed to display them.
- "The NV centre is a single electron." It is actually an electron triplet — total spin S = 1, three sublevels m_s = -1, 0, +1. The qubit is formed from two of those three sublevels. The third is usually parked by a small magnetic field that Zeeman-splits the degenerate m_s = \pm 1 pair.
- "Silicon needs to be purified so electrons don't scatter." Isotopic purification is not for mobility; it is for coherence. The 5% ^{29}Si in natural silicon has nuclear spin, and these nuclei fluctuate the magnetic field seen by the electron on microsecond timescales, dephasing the qubit. Electron mobility is fine in natural silicon; coherence is not.
- "NV centres work at room temperature, so they must be the best platform." They run at room temperature, but their inter-qubit coupling is weak (dipolar, kHz range, tiny for pairs more than 30 nm apart) so scaling beyond ~10 nearby NVs has not been demonstrated. Room-temperature operation is a huge convenience for sensing and single-qubit experiments but does not by itself make NV the winning computing platform.
- "Spin-to-charge conversion is the same as photon readout." It is the silicon analogue of photon readout for NV, but implemented by tunnelling electrons and charge sensors rather than fluorescence. Both convert "what is my spin state?" into a classical signal; the physical transducer is completely different.
Going deeper
If you understand that spin qubits are single-electron-spin (or single-nucleus-spin) qubits, that silicon quantum dots trap them in gate-defined potential wells or on phosphorus donor atoms in ^{28}Si, that NV centres use diamond crystal defects at room temperature, that two-qubit gates use Heisenberg exchange (silicon) or dipolar coupling + photonic links (NV), and that Intel's Tunnel Falls and Diraq are the silicon front-runners while IIT-Kanpur's Saha group leads Indian NV work — you have chapter 171. What follows is a closer look at the Loss–DiVincenzo and Kane proposals, the isotopic-purification requirement, the scaling challenge of individual-qubit control lines, and the NQM's spin-qubit roadmap.
The Loss–DiVincenzo 1998 proposal
Daniel Loss and David DiVincenzo published Quantum Computation with Quantum Dots in Phys. Rev. A in 1998, setting out the canonical spin-qubit architecture nearly a decade before the first experimental demonstration. The paper proposes: (a) electron spins as qubits, (b) gate-defined semiconductor quantum dots as trapping sites, (c) ESR for single-qubit gates, (d) voltage-controlled Heisenberg exchange for two-qubit gates, and (e) spin-to-charge readout via a nearby electrometer. Every silicon-spin-qubit experiment today follows this blueprint — the only major departure is the shift from GaAs (in the 1998 paper) to ^{28}Si (enabling the seconds-long T_2). The paper's arXiv preprint is freely available and remains the single best introduction to the theoretical foundations.
Kane's 1998 phosphorus-donor proposal
In the same year, Bruce Kane published a complementary scheme in Nature: instead of gate-defined dots, use individual phosphorus donor atoms in a silicon lattice. The nuclear spin of the ^{31}P, Kane argued, would have essentially unlimited coherence — seconds to tens of seconds — because the surrounding ^{28}Si lattice provides a magnetically silent environment. Kane's paper is visionary and, in many ways, describes the endpoint of the program: donor-based quantum computing with nuclear-spin memory qubits and electron-spin processing qubits. UNSW's Michelle Simmons has spent two decades building the STM-patterned devices that implement the Kane architecture, with the first two-qubit donor gates demonstrated in 2019 and donor-based coherence times of tens of seconds.
Why scaling to millions of qubits is hard
Every silicon spin qubit needs at least three control lines: a microwave line for ESR (single-qubit gates), gate voltages for plunger and exchange control, and a readout line (SET or RF reflectometry). At a thousand qubits, that is thousands of lines entering the dilution refrigerator, with thermal and wiring budgets that rapidly become dominant. The long-term solution is on-chip cryogenic CMOS control, pioneered by Intel's Horse Ridge cryoelectronics (a cryo-compatible chip that multiplexes microwave control of many qubits from a compact SoC at the cold stage), and by analogue demultiplexers at 4 K or 1 K that fan out from a few external wires to thousands of on-chip connections. This is a Heroic engineering problem; no one has demonstrated more than ~16 simultaneously controlled silicon spin qubits as of 2026, against the ~1000 achieved by transmons or neutral atoms.
NV-centre networks and the quantum internet
Individual NV centres have weak direct coupling to each other, but each NV can emit a photon entangled with its spin state. Send the photon to a beam splitter; interfere it with a photon from a second NV; detect coincidence events. Success heralds entanglement between the two distant NV spins — the Barrett–Kok protocol, first demonstrated in 2013 by the Hanson group at Delft. Iterated, this produces entanglement networks: the basis for proposed quantum internets. Delft's 2022 "3-node network" paper demonstrated deterministic entanglement among three NV nodes at separations of tens of metres. If this scales, NV networks become the backbone of a quantum internet — with qubits processed locally on each node and entanglement distributed between them via photonic channels.
NQM spin-qubit plans
The National Quantum Mission's 2023 roadmap includes a spin-qubit pillar with:
- 2024–2027: NV-centre demonstrators at IIT-Kanpur and IISc, targeting 3–5 coupled NV qubits and photonic interfaces.
- 2025–2028: silicon spin-qubit demonstrator at TIFR in collaboration with IMEC (Belgian foundry access) — targeting a 10-qubit gate-defined device.
- Coordination with ISRO on NV-based magnetometry for space applications (deep-space navigation, lunar surface magnetism).
- Budget allocation: approximately ₹400 crore across silicon and NV over the mission, modest compared with European and US programs but enough for 3–5 world-class demonstrators.
The bottleneck for India, as for neutral atoms, is the talent pipeline: perhaps 5–10 experimental groups of world-class quality as of 2026, with NQM funding expansion in the 2026–2030 window.
Platform comparison
| Platform | Best qubit count (2024) | T_2 | 2Q fidelity | Environment |
|---|---|---|---|---|
| Silicon (Intel Tunnel Falls) | 12 | 1–10 s | 99.2% | dilution fridge (~20 mK) |
| Silicon donors (UNSW) | 4 | >30 s | 99.3% | dilution fridge (~20 mK) |
| Diamond NV (coherent pairs) | 3–10 in network | 1 ms (up to 1 s with DD) | ~95% | room temperature |
| Superconducting (IBM Heron) | 133 | 200 μs | 99.7% | dilution fridge (~20 mK) |
| Trapped ions (Quantinuum H2) | 56 | 60 s | 99.9% | UHV, ~4 K |
| Neutral atoms (Atom Computing) | 1180 | 1 s | 99.4% | UHV, ~mK cooling |
Spin qubits currently lag on raw count but lead on coherence (silicon donors) and environmental accessibility (NV). The long-term bet is that CMOS-compatible silicon will scale faster than any other platform once the wiring and cryoelectronics problems are solved — but as of 2026, neutral atoms and superconductors remain substantially ahead on demonstrable qubit number.
Where this leads next
- Neutral Atom Qubits — another matter-qubit platform that solves the wiring problem differently, using global optical control instead of individual wires.
- Superconducting Transmon — the dominant current platform, contrasting sharply with spin qubits in environment, speed, and fabrication.
- DiVincenzo's Criteria — the hardware checklist; spin qubits score well on coherence, moderately on gate fidelity, and weakly on scalability.
- Topological Qubits — the condensed-matter alternative to spin qubits, with an even longer road to demonstration.
- Decoherence — T1 and T2 — the coherence framework in which silicon donors reach T_2 > 30 s.
References
- Daniel Loss and David P. DiVincenzo, Quantum Computation with Quantum Dots (1998), Phys. Rev. A 57, 120 — arXiv:cond-mat/9701055.
- Bruce E. Kane, A Silicon-Based Nuclear Spin Quantum Computer (1998), Nature 393, 133 — Nature (open archive).
- Marcus W. Doherty et al., The nitrogen-vacancy colour centre in diamond (2013), Physics Reports 528, 1 — arXiv:1302.3288.
- Wikipedia, Spin qubit quantum computer.
- Intel Newsroom, Tunnel Falls silicon spin-qubit research chip (2023) — intel.com/newsroom.
- John Preskill, Lecture Notes on Quantum Computation, Chapter 7 — theory.caltech.edu/~preskill/ph229.