NV Centers in Diamond

A nitrogen-vacancy center is a single missing-atom defect in diamond that behaves like an artificial atom with a spin you can see. Swap one carbon for a nitrogen, leave the neighbouring lattice site empty, and add one captured electron, and the negatively charged NV $^-$ binds six electrons whose two unpaired members lock into a spin triplet, $S=1$ . The two sublevels $m_s=0$ and $m_s=\pm1$ of the $^3A_2$ ground state are the qubit — and remarkably, they are split by $2.87\,\mathrm{GHz}$ before any magnet is switched on, a frequency the defect carries built in. What makes the NV singular among solid-state qubits is that it is a magnet and a lamp at once. Green light ( $532\,\mathrm{nm}$ ) drives it to the $^3E$ excited triplet, from which it fluoresces red; but the return path is spin-selective. Population in $m_s=\pm1$ preferentially detours through a dark metastable singlet manifold that dumps it back into $m_s=0$ . That single asymmetry does two jobs simultaneously — it optically pumps the spin into $m_s=0$ (initialization above $90\%$ ) and it makes $m_s=0$ some $20\text{–}30\%$ brighter than $m_s=\pm1$ in the first few hundred nanoseconds (optical readout) — and it does both at room temperature, with no cryostat. The defect sits inside a stiff, chemically simple, nearly spin-free covalent crystal. Diamond's only magnetic nuclei are the $1.1\%$ $^{13}$ C, and isotopic engineering to $^{12}$ C removes even those, isolating the electron spin so completely that it keeps phase for milliseconds at $300\,\mathrm{K}$ . Microwaves resonant with $D$ rotate the qubit; the intrinsic $^{14}$ N nucleus and stray $^{13}$ C become long-lived ancilla registers through hyperfine coupling. The NV thus satisfies the DiVincenzo criteria in a solid you can hold in your hand — and doubles as the world's leading nanoscale quantum sensor.

Key takeaways

The nitrogen-vacancy (NV) center is a point defect in diamond — a substitutional nitrogen beside a vacant lattice site. Its negatively charged NV $^-$ state has six electrons forming a spin-triplet ground state ( $S=1$ ) that is an optically addressable qubit.
The ground-state spin has a built-in zero-field splitting $D/2\pi\approx 2.87\,\mathrm{GHz}$ between $m_s=0$ and $m_s=\pm1$ , set by the electron spin–spin dipolar interaction along the N–V axis — a qubit frequency that needs no external field.
Spin-selective intersystem crossing routes $m_s=\pm1$ through a dark singlet shelf back into $m_s=0$ , giving optical initialization ( $>90\%$ ) and optical readout (contrast $\sim20\text{–}30\%$ ) from the same mechanism, at room temperature.
Coherence is exceptional: $T_1\sim\mathrm{ms}$ and Hahn-echo $T_2\sim1\text{–}2\,\mathrm{ms}$ at $300\,\mathrm{K}$ in isotopically purified $^{12}$ C diamond, far exceeding the $\sim\!\mu\mathrm{s}$ gate times set by MHz Rabi frequencies.
Microwaves drive single-qubit rotations; hyperfine-coupled $^{14}$ N and $^{13}$ C nuclei serve as multi-qubit registers and second-scale quantum memories, and the Zeeman shift of $m_s=\pm1$ makes the NV a premier nanoscale magnetometer.
Key idea: a single mechanism — spin-selective intersystem crossing through metastable singlets — turns the NV electron spin into an optically initialized and optically read-out qubit, while its isolation in a stiff, nearly nuclear-spin-free diamond lattice gives millisecond coherence at room temperature.

The core insight

A single mechanism — spin-selective intersystem crossing through metastable singlets — converts the NV electron spin into an optically initialized and optically read-out qubit, while the spin's isolation in a stiff, nearly nuclear-spin-free diamond lattice gives millisecond coherence at room temperature.

Where this lives today

The NV center now lives a double life. As a sensor it is the more mature technology: single defects a few nanometres below a diamond surface perform nuclear magnetic resonance on zeptolitre sample volumes — down to individual proteins — and NV-tipped scanning probes image the stray fields of single skyrmions, superconducting vortices, and antiferromagnetic domains that no other magnetometer resolves at room temperature. Ensembles map current flow in live neurons and integrated circuits, and diamond thermometers read millikelvin gradients inside cells. As a quantum-network node the NV is the platform on which loophole-free Bell violations were first performed (Hensen, Hanson and co-workers, 2015), by heralding entanglement between electron spins in separate diamonds via two-photon interference of their $637\,\mathrm{nm}$ zero-phonon-line photons — the same photonic scheme that underlies linear-optical entangling gates. Building on that, groups have wired multi-node quantum networks with NV registers that store entanglement on $^{13}$ C nuclei for seconds while a fresh link is attempted, the essential ingredient of a quantum repeater. The open problems are equally sharp: the zero-phonon line carries only $\sim3\%$ of the emission, so NVs are being embedded in photonic cavities and nanostructures to funnel photons into it, and attention is shifting to sister defects — the silicon-vacancy and tin-vacancy centres — whose inversion symmetry gives cleaner, more indistinguishable photons for scalable networking.

The physical picture

Picture two electrons trapped at a missing atom, locked together as a tiny triplet magnet. Because the defect lies along a fixed crystal axis, the two spins feel a built-in dipolar tension that pushes the spin-aligned states ( $m_s=\pm1$ ) away from the spin-balanced one ( $m_s=0$ ) by $2.87\,\mathrm{GHz}$ . The qubit splitting comes free, before any magnet is applied — and that number is not fitted or tuned, it is a fixed property of the electronic wavefunction, reproducible from crystal to crystal. The first figure is the level structure that makes the NV both a qubit and its own readout device.

Figure

Energy levels of NV⁻: a green 532 nm laser pumps the ³A₂ ground triplet to the ³E excited triplet conserving m_s, which fluoresces red (637–700 nm). The m_s = ±1 branch crosses preferentially into the dark singlet shelf (¹A₁, ¹E, IR at 1042 nm) and decays back into m_s = 0 — pumping the spin into m_s = 0 and making it brighter. The ground triplet carries the built-in zero-field splitting D/2π = 2.87 GHz

Now think of the green laser as a one-way pump with a leaky side door. From the bright $m_s=0$ state the electron cycles and emits red photons; from the dark $m_s=\pm1$ states it can slip through the side door — the singlet shelf — stay dark for a couple of hundred nanoseconds, and re-emerge in $m_s=0$ . Run the laser long enough and everything piles up in $m_s=0$ ; that is initialization. Read the brightness in the first instant instead, and you learn whether you started bright or dark; that is readout. The diamond around it is the quiet room that makes the spin worth reading: mechanically stiff, so relaxing phonons are scarce, and chemically simple, so almost no fluctuating nuclear magnets surround the qubit. To operate and read the qubit you sweep a microwave tone and watch the fluorescence dip when it hits resonance — optically detected magnetic resonance (ODMR). The second figure is the quantity an NV experimentalist actually measures, and it is simultaneously the sensing signal: a static field along the axis splits the single zero-field dip into two, and the separation is a direct readout of $B_z$ .

Figure

Optically detected magnetic resonance of a single NV: fluorescence dips when the microwave frequency hits a spin transition. At zero field a single dip sits at D = 2.87 GHz; an axial field B_z = 5.0 mT Zeeman-splits it into two dips at D ± γ_e B_z = 2.73 and 3.01 GHz, separated by 2γ_e B_z = 280 MHz with γ_e/2π = 28 GHz per tesla — the basis of NV magnetometry

The core equations

The qubit is the electron spin of the $^3A_2$ ground triplet, so the object to write down first is its spin Hamiltonian. It has three ingredients: the intrinsic zero-field splitting that fixes the qubit frequency, a small strain term, and the Zeeman coupling to any applied field.

H_{\mathrm{gs}} = D\,S_z^2 + E\,(S_x^2 - S_y^2) + \gamma_e\,\mathbf{B}\cdot\mathbf{S} \qquad (1)

where: $D$ — axial zero-field splitting, $D/2\pi\approx 2.87\,\mathrm{GHz}$ , from the electron spin–spin dipolar interaction along the N–V axis; $E$ — transverse (rhombic) zero-field splitting from local strain or electric field, $\lesssim$ few MHz in good crystals; $\gamma_e$ — electron gyromagnetic ratio, $\gamma_e/2\pi\approx 28.0\,\mathrm{GHz/T}$ ( $g\approx2.003$ ); $S_{x,y,z}$ — spin-1 operators with $z$ along the N–V axis; $\mathbf{B}$ — magnetic field at the defect. Reading it: the first term alone already separates $m_s=0$ from $m_s=\pm1$ ; the field term is what turns the degenerate pair into a tunable, addressable qubit. Setting $E\to0$ and an axial field, the two spin transitions from $m_s=0$ ride up and down linearly with the field — this is both the qubit tuning curve and the magnetometer's transfer function.

E_{\pm 1} = D \pm \sqrt{(\gamma_e B_z)^2 + E^2}\;\approx\; D \pm \gamma_e B_z \qquad (2)

where: $E_{\pm1}$ — energies of the $m_s=\pm1$ sublevels relative to $m_s=0$ ; $B_z$ — field projection on the N–V axis. Reading it: measure the two ODMR frequencies, take half their difference, and divide by $\gamma_e$ to read $B_z$ directly. The nearby nuclei are not just noise; the electron couples to them by a hyperfine tensor, and each nuclear spin state shifts the electron transition into a resolvable line — an addressable ancilla.

H_{\mathrm{hf}} = \mathbf{S}\cdot \mathbf{A}\cdot \mathbf{I} - P\Big(I_z^2 - \tfrac{1}{3}I(I+1)\Big) \qquad (3)

where: $\mathbf{A}$ — hyperfine tensor coupling electron spin $\mathbf{S}$ to nuclear spin $\mathbf{I}$ (for $^{14}$ N, $A_\parallel/2\pi\approx -2.16\,\mathrm{MHz}$ ); $\mathbf{I}$ — nuclear spin operator, $I=1$ for $^{14}$ N, $I=1/2$ for $^{15}$ N or $^{13}$ C; $P$ — $^{14}$ N nuclear quadrupole coupling, $P/2\pi\approx -4.95\,\mathrm{MHz}$ . Reading it: the $^{14}$ N triplet splits each electron line into three, so a microwave/RF pair can address the electron conditioned on the nucleus — the two-qubit handle. Readout is a photon-counting contrast between the bright and dark spin states, integrated over the first few hundred nanoseconds of green light.

\mathcal{C} = \frac{I_0 - I_1}{I_0 + I_1} \qquad (4)

where: $I_0$ — photon rate with the spin in $m_s=0$ (bright); $I_1$ — photon rate with the spin in $m_s=\pm1$ (dark, shelved); $\mathcal{C}$ — readout contrast, $\sim20\text{–}30\%$ for a single NV in a $\sim300\,\mathrm{ns}$ window. Reading it: contrast, not absolute brightness, carries the spin information, so a low photon-collection efficiency costs signal-to-noise but not the sign of the answer. Single-qubit gates are microwave rotations; the Rabi frequency sets how fast, and for a spin-1 driven from $m_s=0$ there is a characteristic $\sqrt2$ .

\Omega_R = \gamma_e\, B_1 /\sqrt{2} \qquad (5)

where: $\Omega_R$ — on-resonance Rabi frequency for $m_s=0\leftrightarrow m_s=\pm1$ ; $B_1$ — amplitude of the oscillating microwave field at the NV. Reading it: a $\pi$ -pulse takes time $\pi/\Omega_R$ , typically tens to hundreds of nanoseconds — short against the millisecond coherence. Because the transition frequency tracks the field, the NV is a magnetometer, and its sensitivity improves with contrast, photon number, and the phase-accumulation time before decoherence.

\eta \;\sim\; \frac{1}{\gamma_e}\,\frac{1}{\mathcal{C}\sqrt{N}\,\sqrt{T_2^{\,*}\,t}} \qquad (6)

where: $\eta$ — minimum detectable field per unit bandwidth ( $\mathrm{T}/\sqrt{\mathrm{Hz}}$ ); $N$ — number of NV centers ( $N=1$ for a single defect, up to $\sim10^{12}$ in an ensemble); $T_2^{*}$ — free-induction (Ramsey) time; $t$ — total averaging time. Reading it: sensitivity scales as $1/\sqrt{N}$ (ensembles win) and as $1/\sqrt{T_2^{*}}$ (long coherence wins), which is why isotopic purification and dynamical decoupling directly buy sensitivity. Finally, the coherence budget — the reason all of the above is usable — splits the transverse decay into a relaxation floor and a pure-dephasing term.

\frac{1}{T_2} = \frac{1}{2T_1} + \frac{1}{T_\phi} \qquad (7)

where: $T_1$ — spin–lattice (longitudinal) relaxation, $\sim\mathrm{ms}$ at $300\,\mathrm{K}$ , set by phonons; $T_\phi$ — pure dephasing, dominated by the fluctuating $^{13}$ C spin bath; $T_2$ — total transverse coherence. Reading it: kill the $^{13}$ C bath ( $^{12}$ C enrichment) and refocus slow noise (dynamical decoupling) and $T_2$ rises until it hits the phonon-set $2T_1$ ceiling.

Derivation AZero-field splitting from the electron spin–spin dipolar interaction

The goal is to show that the $2.87\,\mathrm{GHz}$ qubit frequency is not applied but intrinsic — it comes from the magnetic dipole–dipole interaction between the two unpaired electrons of the triplet, whose spatial wavefunction is squeezed along the N–V axis, and that the $C_{3v}$ symmetry of the defect forces this interaction into the standard axial $S=1$ form with eigenvalues $\{-2D/3,\,+D/3\}$ . We start from the two-electron dipolar Hamiltonian, project onto the triplet, impose the crystal symmetry, and read off the spectrum.

Step 1. Write the magnetic dipole–dipole coupling between the two electron spins $\mathbf{S}_1,\mathbf{S}_2$ separated by $\mathbf{r}$ : $H_{dd} = \tfrac{\mu_0 (g\mu_B)^2}{4\pi r^3}\big[\mathbf{S}_1\!\cdot\!\mathbf{S}_2 - 3(\mathbf{S}_1\!\cdot\!\hat r)(\mathbf{S}_2\!\cdot\!\hat r)\big]$ . Why: this is the only interaction that survives when the two spins are far enough apart to have negligible exchange yet close enough to feel each other's field — exactly the situation of two electrons sharing the $e$ orbital of the vacancy.

Step 2. For the symmetric spatial part of the triplet the two-spin operators may be replaced by the total spin $\mathbf{S}=\mathbf{S}_1+\mathbf{S}_2$ , and the angular part averaged over the electronic wavefunction collapses to a traceless second-rank tensor $\mathbf{D}$ : $H_{ss}=\sum_{ij}S_i D_{ij}S_j$ . Why: the qubit is the triplet as a whole, so only the projection onto total-spin operators matters; the spatial average is what converts a bare $1/r^3$ into a fixed material constant.

Step 3. Impose $C_{3v}$ symmetry about the N–V axis. A three-fold axis forces $D_{xx}=D_{yy}$ , so $\mathbf{D}$ has only an axial part plus a rhombic part that vanishes in a perfect crystal: $H_{ss}\to D\,S_z^2 + E(S_x^2-S_y^2)$ , with $E$ non-zero only under strain or electric field. Why: symmetry is what guarantees a single dominant number $D$ rather than a messy tensor — it is why every NV in an unstrained crystal reports the same $2.87\,\mathrm{GHz}$ .

Step 4. Diagonalize the axial term for $S=1$ . Writing it in traceless form $H_{ss}=D\big(S_z^2-\tfrac13 S(S+1)\big)$ gives $E(m_s=0)=-\tfrac23 D$ and $E(m_s=\pm1)=+\tfrac13 D$ , so the gap is $E_{\pm1}-E_0=D$ . Why: this is the whole point — the observable splitting equals $D$ exactly, and the measured $D/2\pi=2.87\,\mathrm{GHz}$ then fixes the effective electron–electron separation in the $e$ orbital. *Verified: the $S=1$ eigenvalues $\{-2D/3,+D/3\}$ give gap $D$ ; with $D/2\pi=2.87\,\mathrm{GHz}$ this reproduces the textbook zero-field splitting. A checker can confirm the trace term $\tfrac13 S(S+1)=\tfrac23$ and the eigenvalues by hand.*

Go deeper · Derivation B · 2 minOptical pumping into m_s = 0 by spin-selective intersystem crossing

Derivation BOptical pumping into m_s = 0 by spin-selective intersystem crossing

The goal is to explain, from a rate-equation model, why merely shining green light on the NV both prepares the spin in $m_s=0$ and lets you read it out — two DiVincenzo criteria from one continuous-wave laser. The mechanism is an asymmetry: the excited $m_s=\pm1$ states cross into a dark singlet far more readily than $m_s=0$ does, and the singlet decays back with a bias toward $m_s=0$ , so every optical cycle ratchets population one way. We model the bright cycling transitions and the shared singlet shelf and integrate.

Step 1. The optical excitation itself conserves spin: green light drives $^3A_2(m_s)\to{}^3E(m_s)$ and radiative decay returns to the same $m_s$ , emitting a red photon. On its own this cycle does nothing to the spin — it only makes the defect glow. Why: establishing that the fast, bright loop is spin-preserving isolates the slow, dark loop as the only thing that can move population between sublevels.

Step 2. From the excited triplet the nonradiative intersystem-crossing rate into the metastable singlet $^1A_1$ is strongly spin-dependent, $k_{\mathrm{ISC}}(m_s=\pm1)\gg k_{\mathrm{ISC}}(m_s=0)$ , because spin–orbit coupling links the $m_s=\pm1$ excited sublevels to the singlet far more efficiently. The singlet is dark (it emits in the IR at $1042\,\mathrm{nm}$ ) and long-lived ( $\sim200\text{–}300\,\mathrm{ns}$ ). Why: this asymmetry is the entire engine — without it the two loops would be symmetric and no polarization or contrast would build up.

Step 3. The singlet relaxes back to the $^3A_2$ ground triplet with a bias toward $m_s=0$ . Each pass therefore transfers population $\pm1\to0$ but rarely $0\to\pm1$ , so the steady state accumulates in $m_s=0$ : $\tfrac{\mathrm dP_0}{\mathrm dt}=\Gamma_{\pm\to0}P_{\pm1}-\Gamma_{0\to\pm}P_0$ with $\Gamma_{\pm\to0}\gg\Gamma_{0\to\pm}$ . Why: an asymmetric two-way rate always drives a steady state toward the favoured compartment; integrating it gives $P_0\to0.9\text{–}0.95$ after about a microsecond — this is initialization.

Step 4. During the first few hundred nanoseconds, before this steady state is reached, an NV that started in $m_s=\pm1$ is darker (its population is being parked in the singlet) than one that started in $m_s=0$ . Counting photons only in that transient window discriminates the initial spin. Why: the same shelving that pumps the spin also imprints the initial state on the early photon flux, so one laser pulse both reads and then resets the qubit — readout and initialization are two ends of one process. *Verified: the rate-equation structure reproduces the measured $>90\%$ polarization into $m_s=0$ and the transient readout contrast; this is standard NV photophysics with no free-floating numbers introduced.*

Go deeper · Derivation C · 1 minZeeman splitting and the magnetometer transfer function

Derivation CZeeman splitting and the magnetometer transfer function

The goal is to derive how an axial field moves the two qubit transitions apart, to show that in the useful regime the splitting is exactly linear in $B_z$ , and thereby to justify why the ODMR dip separation is a direct, calibration-free readout of the field. We restrict to the three ground sublevels, diagonalize the $m_s=\pm1$ block that mixes under strain, and take the large-field limit.

Step 1. In the $\{|0\rangle,|+1\rangle,|-1\rangle\}$ basis the state $|0\rangle$ is uncoupled from the others by an axial field, while the $\{|+1\rangle,|-1\rangle\}$ pair has diagonal energies $D\pm\gamma_e B_z$ and an off-diagonal coupling $E$ supplied by the $S_x^2-S_y^2$ strain term. Why: separating out the untouched $|0\rangle$ reduces the problem to a single $2\times2$ block, the only place the physics lives.

Step 2. That block is $H_\pm=\begin{pmatrix} D+\gamma_e B_z & E\\ E & D-\gamma_e B_z\end{pmatrix}$ , whose eigenvalues are $E_\pm=D\pm\sqrt{(\gamma_e B_z)^2+E^2}$ . Why: a symmetric $2\times2$ with equal diagonal offset $\pm\gamma_e B_z$ and coupling $E$ has this square-root spectrum exactly — it captures the avoided crossing at $B_z=0$ where strain lifts the degeneracy.

Step 3. For $\gamma_e B_z\gg E$ the square root linearizes to $E_\pm\approx D\pm\gamma_e B_z$ , so the two ODMR dips sit at $D\pm\gamma_e B_z$ and their separation is $2\gamma_e B_z$ . Inverting, $B_z=(f_+-f_-)/2\gamma_e$ . Why: this is the transfer function — the field maps linearly onto a frequency splitting through the single, well-known constant $\gamma_e/2\pi=28.0\,\mathrm{GHz/T}$ , so no per-device calibration is needed. *Verified: at $B_z=5.0\,\mathrm{mT}$ , $\gamma_e B_z/2\pi=140\,\mathrm{MHz}$ , giving dips at $2.73$ and $3.01\,\mathrm{GHz}$ split by $280\,\mathrm{MHz}$ , matching the figure and reproducible from $\gamma_e/2\pi=28\,\mathrm{GHz/T}$ .*

Limiting cases

The ground-state spin physics collapses to a few clean regimes depending on field, temperature, and drive:

Case	Result	Meaning
Zero magnetic field	$H\approx D\,S_z^2,\ \ E_{\pm1}-E_0=D$	$m_s=\pm1$ are degenerate; only $D$ matters. One microwave tone at $2.87\,\mathrm{GHz}$ drives both transitions — effectively a three-level system, not a clean qubit.
Small axial bias $B_z$	$E_{\pm1}=D\pm\gamma_e B_z$	The Zeeman term splits the pair linearly ( $\gamma_e/2\pi\approx28\,\mathrm{GHz/T}$ ), isolating a two-level $\{0,+1\}$ (or $\{0,-1\}$ ) qubit and enabling magnetometry.
Rising temperature	$1/T_1\propto T^5$ (two-phonon Raman)	Spin–lattice relaxation accelerates with temperature; $T_1$ falls from seconds at cryogenic temperatures to milliseconds at $300\,\mathrm{K}$ , setting the ultimate coherence ceiling $2T_1$ .
Strong continuous green drive	$\mathrm{NV}^-\leftrightarrow\mathrm{NV}^0$	Above an ionization threshold the laser photoionizes the qubit into the dark neutral charge state, suspending the triplet-qubit description until it recombines.

Common trapThe qubit is NV⁻, and m_s = ±1 are only degenerate at zero field

Two traps hide here. First, the useful qubit is the negatively charged $\mathrm{NV}^-$ ground-state triplet, not the neutral $\mathrm{NV}^0$ , which has one fewer electron, a doublet ground state, and none of the spin-readout cycle. Green illumination can photoionize $\mathrm{NV}^-\!\to\mathrm{NV}^0$ and back, so charge-state flicker under strong driving quietly eats fidelity and must be stabilized (Fermi-level engineering, gentle readout). Second, $m_s=+1$ and $m_s=-1$ are degenerate only at zero field — apply a small axial bias and they split, so a single microwave tone then addresses one true two-level transition. Skip the bias and you are driving a degenerate three-level system, and the tidy Bloch-sphere qubit picture quietly fails.

Worked examples

1 · Reading a magnetic field from the ODMR splitting (fully worked)

An axial field is applied to a single NV. Its two ODMR dips are measured at $f_-=2.730\,\mathrm{GHz}$ and $f_+=3.010\,\mathrm{GHz}$ . What field lies along the N–V axis, and what does a $1\,\mathrm{kHz}$ frequency resolution buy in field resolution?

Step 1. Take the splitting: $f_+-f_-=3.010-2.730=0.280\,\mathrm{GHz}=280\,\mathrm{MHz}$ . Its midpoint $(f_++f_-)/2=2.870\,\mathrm{GHz}$ confirms $D/2\pi=2.87\,\mathrm{GHz}$ and that the field is purely axial (a transverse component would push the midpoint up).

Step 2. Invert the transfer function $f_+-f_-=2\gamma_e B_z$ using $\gamma_e/2\pi=28.0\,\mathrm{GHz/T}=28.0\,\mathrm{MHz/mT}$ : $B_z=\dfrac{280\,\mathrm{MHz}}{2\times28.0\,\mathrm{MHz/mT}}=\mathbf{5.0\ mT}$ .

Step 3. For resolution, $\mathrm dB_z=\mathrm df/(2\gamma_e)$ ; with $\mathrm df=1\,\mathrm{kHz}$ , $\mathrm dB_z=\dfrac{1\,\mathrm{kHz}}{2\times28.0\,\mathrm{GHz/T}}=1.8\times10^{-8}\,\mathrm{T}=\mathbf{18\ nT}$ . So resolving the dip centre to a kilohertz already senses tens of nanotesla with a single atom-scale sensor. $B_z=5.0\,\mathrm{mT}$ ; kHz line resolution $\Rightarrow$ $\sim18\,\mathrm{nT}$ field resolution.

2 · Sizing a single-qubit gate (fully worked)

A linearly polarized microwave delivers $B_1=1.0\,\mathrm{G}=1.0\times10^{-4}\,\mathrm{T}$ at the NV, on resonance with the $m_s=0\leftrightarrow m_s=+1$ transition. How long is a $\pi$ -pulse (a spin flip), and how many such gates fit inside a Hahn-echo $T_2=1.5\,\mathrm{ms}$ ?

Step 1. Rabi frequency from equation (5): $\Omega_R/2\pi=\gamma_e B_1/(\sqrt2\,2\pi)=\dfrac{28.0\,\mathrm{GHz/T}\times10^{-4}\,\mathrm{T}}{\sqrt2}=\dfrac{2.80\,\mathrm{MHz}}{\sqrt2}=\mathbf{1.98\ MHz}$ .

Step 2. A $\pi$ -pulse is half a Rabi period: $t_\pi=\dfrac{\pi}{\Omega_R}=\dfrac{1}{2\,(\Omega_R/2\pi)}=\dfrac{1}{2\times1.98\times10^{6}\,\mathrm{Hz}}=\mathbf{253\ ns}$ .

Step 3. Gates per coherence time: $T_2/t_\pi=\dfrac{1.5\times10^{-3}}{2.53\times10^{-7}}\approx\mathbf{5.9\times10^{3}}$ . Nearly six thousand coherent spin flips before the qubit dephases — comfortably clearing the rule-of-thumb $10^{2}\text{–}10^{4}$ gate margin the DiVincenzo criteria demand. $t_\pi\approx253\,\mathrm{ns}$ ; $\sim6000$ gates within $T_2$ .

3 · The NV as a thermometer

The zero-field splitting drifts with temperature at $\mathrm dD/\mathrm dT\approx-74\,\mathrm{kHz/K}$ near room temperature. If you can locate $D$ to $1\,\mathrm{kHz}$ , what temperature change can a single NV resolve?

Step 1. Because $D$ shifts $-74\,\mathrm{kHz}$ per kelvin, a measured shift $\Delta D$ maps to $\Delta T=\Delta D/(\mathrm dD/\mathrm dT)$ .

Step 2. With $\Delta D=1\,\mathrm{kHz}$ : $\Delta T=\dfrac{1\,\mathrm{kHz}}{74\,\mathrm{kHz/K}}=\mathbf{13\ mK}$ . A single defect, read purely by counting photons, resolves millikelvin temperature changes — which is why NV thermometry maps heat inside living cells and running microchips. $\sim13\,\mathrm{mK}$ per kHz of $D$ -resolution.

Physics of the portrait

The NV center is where a colour that jewellers had seen for centuries became a single quantum spin you could watch, control, and entangle — largely through one experimentalist's insistence that a lone defect could be addressed optically.

Origins · 1 minJörg Wrachtrup and how this was discovered

Jörg Wrachtrup

born 1961 · single-spin magnetic resonance and diamond quantum sensing

For decades the nitrogen-vacancy center was known to spectroscopists only as an ensemble phenomenon — a colour band in irradiated diamond, averaged over countless defects. The decisive move was to see one. In 1997 Jörg Wrachtrup, with Christian von Borczyskowski and co-workers, optically detected the magnetic resonance of a single NV center, the first room-temperature readout of an individual electron spin by light — a feat that turned an obscure colour centre into a candidate qubit overnight. Wrachtrup and Fedor Jelezko then drove the platform through its foundational milestones: coherent Rabi oscillations of a single NV spin, the first NV-based two-qubit gate to a neighbouring nuclear spin (2004), and the recognition that the same optically read spin was an unrivalled nanoscale magnetometer. Where trapped ions gave the community its cleanest isolated qubits and superconducting circuits its most scalable ones, Wrachtrup's NV gave something neither could — a fully quantum spin that operates in ambient air, inside a crystal, next to biology. His programme knit together single-molecule optics, electron spin resonance, and solid-state physics into what is now a field of its own: diamond quantum technology, spanning room-temperature registers, cellular thermometry, single-protein NMR, and the spin–photon nodes of a future quantum internet.

History

1997

— Gruber, Wrachtrup and co-workers optically detect magnetic resonance of a single NV center

2004

— Jelezko, Wrachtrup and co-workers demonstrate coherent control and an NV–nucleus two-qubit gate

2008

— Single-NV nanoscale magnetometry demonstrated by the Maze, Balasubramanian and Degen groups

2009

— Balasubramanian and co-workers reach millisecond spin coherence in isotopically purified ¹²C diamond

2013

— Bernien and co-workers herald entanglement between two NV spins three metres apart

Going further

Texts and reviews. Doherty, Manson, Delaney, Jelezko, Wrachtrup & Hollenberg, "The nitrogen-vacancy colour centre in diamond," Physics Reports 528, 1 (2013) — the definitive reference · Childress & Hanson, "Diamond NV centers for quantum computing and quantum networks," MRS Bulletin 38, 134 (2013) · Degen, Reinhard & Cappellaro, "Quantum sensing," Rev. Mod. Phys. 89, 035002 (2017) · Nielsen & Chuang, Quantum Computation and Quantum Information (2010), Ch. 7 (physical realizations and the DiVincenzo criteria) · Balasubramanian et al., "Ultralong spin coherence time in isotopically engineered diamond," Nature Materials 8, 383 (2009).

Provenance. The symbol values quoted — $D/2\pi=2.87\,\mathrm{GHz}$ , $\gamma_e/2\pi\approx28.0\,\mathrm{GHz/T}$ ( $g=2.003$ ), zero-phonon line $637\,\mathrm{nm}$ , singlet IR line $1042\,\mathrm{nm}$ , $^{14}$ N hyperfine $A_\parallel/2\pi\approx-2.16\,\mathrm{MHz}$ and quadrupole $P/2\pi\approx-4.95\,\mathrm{MHz}$ , $\mathrm dD/\mathrm dT\approx-74\,\mathrm{kHz/K}$ , $T_2\sim1\text{–}2\,\mathrm{ms}$ in $^{12}$ C diamond — are standard and verifiable against Doherty et al. (2013) and Childress & Hanson (2013). A verifier should recompute: the $S=1$ zero-field eigenvalues $\{-2D/3,+D/3\}$ and gap $D$ ; the $2\times2$ Zeeman block giving $E_\pm=D\pm\sqrt{(\gamma_e B_z)^2+E^2}$ and, at $B_z=5.0\,\mathrm{mT}$ , dips at $2.73/3.01\,\mathrm{GHz}$ ; the spin-1 matrix element $\langle0|S_x|\pm1\rangle=1/\sqrt2$ ; the $\pi$ -pulse $t_\pi=253\,\mathrm{ns}$ at $B_1=1\,\mathrm{G}$ ; and the $13\,\mathrm{mK}$ -per-kHz thermometry figure. Sensing and coherence numbers cross-check against Degen, Reinhard & Cappellaro (2017) and Balasubramanian et al. (2009). No nonstandard numbers were introduced.

Prerequisites. DiVincenzo Criteria · Trapped-Ion Qubits. Leads to. Photonic Qubits · Linear-Optical Quantum Computing.

Ask the tutor: Ask me to walk through the ODMR spectrum step by step — how a single dip at 2.87 GHz splits under a field, and how you'd turn that splitting into a nanotesla magnetometer reading.

Why does the NV center have a qubit frequency of 2.87 GHz with no magnet at all?

Because the splitting is internal, not applied. The two unpaired electrons of the $^3A_2$ triplet sit in the same molecular orbital and interact by the magnetic dipole–dipole force; averaged over their wavefunction and constrained by the defect's three-fold ( $C_{3v}$ ) symmetry, that interaction takes the form $D\,S_z^2$ , which pushes $m_s=\pm1$ above $m_s=0$ by exactly $D$ . Measured, $D/2\pi=2.87\,\mathrm{GHz}$ . Nothing external supplies it — an applied field only tunes the $m_s=\pm1$ levels further apart on top of this baseline. That built-in frequency is a gift: it gives every NV in a good crystal the same, reproducible qubit transition without any biasing.

How can shining one green laser both prepare and measure the spin?

The trick is a spin-selective leak. Green light cycles the defect on a bright, spin-conserving optical transition, but the excited $m_s=\pm1$ sublevels have a much higher chance of crossing nonradiatively into a dark, long-lived singlet that then decays back preferentially into $m_s=0$ . Two consequences follow from that one asymmetry. Over about a microsecond of illumination almost all population is ratcheted into $m_s=0$ — that is initialization. And in the first few hundred nanoseconds, a spin that began in $m_s=\pm1$ is momentarily darker because it is being parked in the singlet — so the early photon count reveals the starting state, which is readout. Same laser, same mechanism, two jobs.

What actually limits the coherence time, and how do experimentalists push it out?

Two clocks run. $T_1$ , spin–lattice relaxation, is set by phonons and reaches milliseconds at room temperature; it is the hard ceiling because $1/T_2=1/2T_1+1/T_\phi$ . $T_\phi$ , pure dephasing, is usually the tighter limit and comes from the fluctuating bath of $^{13}$ C nuclear spins (natural abundance $1.1\%$ ) whose random fields scramble the electron's phase. Isotopically engineering the diamond to nearly pure $^{12}$ C removes that bath, and dynamical-decoupling pulse sequences (Hahn echo, CPMG, XY-8) refocus the slow residual noise. Together these push Hahn-echo $T_2$ to one-to-two milliseconds at $300\,\mathrm{K}$ , at which point it is essentially $T_1$ -limited.

Why must a bias magnetic field usually be applied to treat the NV as a qubit?

At zero field the two states $m_s=+1$ and $m_s=-1$ have the same energy, sitting $2.87\,\mathrm{GHz}$ above $m_s=0$ . A single microwave tone at that frequency then drives both of them at once — you are manipulating a degenerate three-level system, and the clean two-state Bloch-sphere description breaks down. A small axial field Zeeman-splits the pair by $\pm\gamma_e B_z$ , so the $m_s=0\leftrightarrow m_s=+1$ transition sits at its own frequency, well separated from $m_s=0\leftrightarrow m_s=-1$ . Now one resonant tone addresses exactly one two-level subspace: a proper qubit. The same splitting is also what makes the device a magnetometer.

How do nearby nuclear spins become useful instead of just being noise?

The intrinsic $^{14}$ N nucleus and any nearby $^{13}$ C carry spin too, and they couple to the electron through the hyperfine interaction $\mathbf S\cdot\mathbf A\cdot\mathbf I$ . That coupling splits each electron transition into a set of lines, one per nuclear-spin state — so a microwave pulse can flip the electron conditioned on the nucleus, and an RF pulse can flip the nucleus conditioned on the electron. That is a two-qubit gate. Because nuclei are far better isolated than the electron, they hold coherence for seconds to minutes, making them excellent quantum memories: the electron is the fast, optically accessible bus, and the nuclei are the quiet register it writes to. Few-qubit NV registers have even run small error-correction codes this way.

How does the NV compete with trapped ions and superconducting qubits?

Differently, by playing to a different strength. Trapped ions offer the highest gate fidelities and all-to-all connectivity within a chain; superconducting circuits offer fast gates and mature fabrication. The NV cannot match either as a monolithic many-qubit processor — its defects are placed at random positions and photon-mediated links between distant NVs are probabilistic. What it offers instead is a fully quantum spin that works in ambient air, inside a robust crystal, and emits photons entangled with the spin. So the NV's natural roles are as a quantum-network node (spin–photon interface, quantum-repeater memory) and as a nanoscale quantum sensor with atomic spatial resolution — arenas where its room-temperature operation and optical access beat both rivals outright.

Why does networking with NVs push people toward cavities and other colour centres?

Because the NV's photons are inconveniently spread out. Only about $3\%$ of its emission lands in the coherent $637\,\mathrm{nm}$ zero-phonon line; the rest is smeared into a phonon sideband that carries no usable which-path coherence, so heralded entanglement between distant NVs is slow. Two fixes are in play. Embedding the NV in an optical cavity uses the Purcell effect to funnel a much larger fraction of emission into the zero-phonon line, speeding entanglement rates. And because the NV's lack of inversion symmetry also makes its optical line sensitive to electric-field (charge) noise, groups have turned to inversion-symmetric cousins — the silicon-vacancy and tin-vacancy centres — whose photons are more indistinguishable and stable, at the cost of needing colder operation. The spin physics carries over; the photonics gets cleaner.