Physics Demos

Physics you can poke

Small interactive demos of the ideas behind trapped-ion quantum computing. No equations required, but they’re there if you want them.

Build a Coulomb crystal

Ions repel each other but are held by the trap. Squeeze the trap hard sideways and they line up in a chain, which is what you want for a quantum computer. Loosen it and the chain buckles into a zigzag, then a 2D crystal. Click to drop in more ions.

10
6.00
—

Normal-mode explorer

The gates talk to each other through these modes, so knowing their frequencies and shapes is half the battle in designing a gate. This solves the real equilibrium positions and mode spectrum for N ions in a harmonic trap. Frequencies are in units of the axial center-of-mass frequency ωz.

7
Mode participation by ion

Modes · ω/ωz

Rotate a qubit on the Bloch sphere

Any state of a single qubit is a point on a sphere. Gates are rotations of that sphere. Start from a state, click a gate, and watch the arrow swing. A Raman pulse of a given length and phase is the rotation R(θ, φ), so you can also type those in directly. Open “Waiting and decoherence” to let the qubit sit and watch it lose phase (T2) and relax (T1). Drag the sphere to look at it from any side.

Waiting and decoherence
Bloch sphere (drag to rotate)

How it’s calculated: each gate rotates the Bloch vector by its angle about its axis (H is a π rotation about the x+z diagonal). Waiting multiplies the x and y components by e-t/T2, rotates them about z at the detuning, and relaxes z toward |0⟩ with T1.

Build a quantum circuit

Pick a gate, click a cell on a wire, and watch the output state update. Two-qubit gates take two clicks in the same column. When you are done, compile the circuit into the gates that a trapped-ion computer really has: single-qubit rotations R(θ, φ), free virtual Z rotations, and the Mølmer-Sørensen XX gate. The page checks the compiled circuit against the original as a whole, not just on one input.

Examples:
Output state: probability of each basis state (colour = phase)

Native trapped-ion gates
Press “Compile” to rewrite the circuit with only the gates the ions actually have: single-qubit rotations R(θ, φ), free virtual Z rotations, and the two-qubit Mølmer-Sørensen gate.

How it’s calculated: the compiler uses H = Z(π) then R(π/2, 90°), a CNOT built from one MS(π/4) gate and four single-qubit rotations, CZ = H·CNOT·H on the target and SWAP as three CNOTs. The check multiplies out both circuits on every basis state and compares them up to a global phase.

Laser-cool an ion

A trapped ion is sloshing back and forth. Each photon it absorbs from the laser pushes it to the right. Fire when it’s moving toward the laser (left) and you slow it down. Fire when it’s moving away and you speed it up. Get the energy into the green zone as fast as you can. Hold the button, press on the picture, or hold Space.

Energy — Time 0.0 s

How real labs do it: tune the laser slightly below the atom’s resonance (red detuning). The Doppler shift brings it into resonance only for an ion moving toward the beam, so the laser automatically “fires” at the right moments. Random recoil from re-emitted photons sets a floor on how cold you can get, called the Doppler limit.

Drive a Raman gate

Hyperfine qubits are flipped with a pair of laser beams whose frequency difference matches the qubit splitting. Here the two beams are counter-propagating, hitting the ion head-on from opposite sides. Neither beam is resonant with the atom: beam 1 drives the ion up to a far-detuned “virtual” level and beam 2 brings it back down to the other qubit state, a two-photon Raman transition. Pick an ion, set the laser’s wavelength, power, phase and pulse length, and watch the qubit rotate on the Bloch sphere.

More settings
Bloch sphere

How it’s calculated: Ω = 2√(I1I2)·|u3/2/2 − u1/2|/ħ, where uk is the vector light shift per unit intensity from the P1/2 and P3/2 lines, for two equal-power lin⊥lin beams.

Sideband Cooling

Doppler cooling still leaves the ion jiggling with about ten quanta of motion (phonons). To reach the ground state, two 532 nm beams hit the ion head-on from opposite sides. Together they drive a Raman transition between the two qubit states, which are hyperfine levels of the same 6S1/2 ground state (F = 1 and F = 2, 8 GHz apart). Their frequency difference is set to the red sideband: the qubit frequency minus the trap frequency. At that frequency the qubit can only flip if the ion gives up one phonon to pay the energy difference. The ion is then reset back to the lower qubit state, the phonon stays gone, and the cycle repeats. Set the trap frequency, how many phonons the ion starts with and how many cooling cycles to run, then press Cool and watch the ion pile up at zero phonons.

1.0 MHz
How likely the ion is to have n phonons: outline = start, filled = now

How it’s calculated: each cycle is a red-sideband Raman pulse (Rabi frequency 100 kHz) followed by an optical-pumping reset that scatters a few photons. It includes the recoil from those photons, off-resonant carrier and blue-sideband driving, and heating of 100 quanta/s. The sideband strengths use the full expressions from Leibfried et al., Rev. Mod. Phys. 75, 281 (2003). The starting state is a thermal distribution with the average phonon number you enter; the Doppler limit on the 493 nm line, Γ/2ω, is about 10 at 1 MHz.

Design a Paul trap

An ion trap can’t hold a charge with static fields alone, so it uses a fast RF field. Whether an ion stays trapped depends on two dimensionless numbers from the Mathieu equation, q (set by the RF voltage) and a (set by the DC voltages). Enter your drive frequency and voltages to see where your trap sits on the stability diagram and what secular frequencies it gives.

Trap geometry
First stability region of the Mathieu equation

How it’s calculated: with ±V on both rod pairs, q = 4eV/(mR2Ω2) and a = 8eU/(mR2Ω2); with RF on one pair and the other grounded, q = 2eV/(mR2Ω2) and a = 4eU/(mR2Ω2). The axial frequency is ωz2 = 2eκUend/(mz02) (Berkeland et al. - 1998). The radial frequency is βΩ/2, with β from the exact Floquet solution of the Mathieu equation.

Design a helical resonator

A helical resonator steps up the RF drive to the voltage an ion trap needs, and filters out everything but the resonant frequency. It’s a copper coil inside a copper can, fed by a small antenna coil at the top. Enter your drive frequency, your trap’s capacitance, the can size, and the wire. The calculator finds the coil diameter that gives the highest Q, sizes the antenna to match a 50 Ω feed, and draws the result to scale.

Coil and can
Antenna

How it’s calculated: the model from Siverns et al., Appl. Phys. B 107, 921 (2012). It assumes a 10 pF connecting wire, a 5 Ω trap, copper throughout, and a winding pitch of twice the wire diameter. The antenna inductance comes from matching 50 Ω with the antenna and coil treated as coupled inductors (k = 1); the turns come from the inductance of a circular loop, L = μ0N2a[ln(8a/r) − 2]. The coil sits D/4 above the floor of the can.

Pin an ion with an optical tweezer

A tightly focused laser beam pulls on an ion, adding extra confinement to just that one ion. Shine it on one ion in a chain and you change how the whole chain vibrates. Pick the ion, the tweezer’s wavelength, power and beam size, and compare each mode with and without the tweezer.

7
Ion participation in this mode, |bi|² (%): grey without the tweezer, colour with it (orange = tweezed ion)
Modes (MHz) · click to animate

How it’s calculated: the ion’s polarizability comes from its 493 nm and 455 nm transitions, which gives a tweezer spring constant k = 4αP/(πε0cw04). That is added to the tweezed ion in the chain’s normal-mode equations. w0 is the 1/e2 beam radius at the ion. It assumes a trap with 0.3 MHz axial and 2.5 MHz radial frequencies.

Read out a qubit with fluorescence

To read a qubit, shine light that makes one state glow and leave the other dark, then count the photons that arrive. Because photons arrive randomly, the bright and dark states give overlapping distributions of counts, and a threshold has to split them. Count longer and they separate, but the ion can change state while you look. Set the rates, the detection time and the threshold, take some shots, and see the error you get.

State lifetimes during detection
Photon counts in one detection window
Best achievable detection error vs detection time

How it’s calculated: counts are Poisson. A state that switches partway through the window gives counts at the mix of the two rates, integrated over the switching time. The error is the mean of the dark-read-as-bright and bright-read-as-dark probabilities, and the measured error comes with a Wilson 95% interval.

Focus a Gaussian beam

A laser beam can’t be focused to a point. Focus it as tightly as you like and it still has a smallest width, the waist, and spreads out again on either side. A tighter waist spreads faster. This calculator gives the waist from a lens setup (or the other way around), the distance over which the beam stays tight, the peak intensity and the field, which is what an optical tweezer cares about.

Beam intensity along the axis (light travels left → right)

How it’s calculated: w0 = M2λf/(πwin) for a collimated beam of radius win through a lens of focal length f, zR = πw02/(M2λ) and w(z) = w0√(1 + (z/zR)2). The peak intensity is 2P/(πw02). These are the paraxial formulas, so they get less accurate when the waist is smaller than a couple of wavelengths.

Match an impedance on the Smith chart

The RF drive for a trap only gets into the trap if the load looks like the source. The Smith chart shows an impedance as a point, and every element you add moves that point along a circle. Put in a load, add series and shunt inductors and capacitors, and walk the point to the centre of the chart, where everything is matched. Or press Auto-match and see what a two-element L network does.

Smith chart (normalised to Z₀)

How it’s calculated: Γ = (z − 1)/(z + 1) with z = Z/Z0. A series element adds jX to Z; a shunt element adds jB to the admittance. The auto-match cancels the load reactance with a series element and then uses the standard L-network: series X = √(R(Z0 − R)) and shunt B = X/(RZ0) when R < Z0, or the reverse order when R > Z0.

Teleport a qubit

Alice has a qubit in a state she doesn’t know, and she and Bob share an entangled pair. By measuring her two qubits and sending Bob two ordinary bits, she can make Bob’s half of the pair become exactly her state. Step through the circuit, watch each qubit’s Bloch sphere, and see the full state at every step. Pick the measurement outcome yourself or let it be random, then run it a thousand times.

Circuit: q0 holds the unknown state, q1 and q2 start as |0⟩

Each qubit on its own Bloch sphere (a short arrow means it is entangled with the others)
Full three-qubit state |q0 q1 q2⟩ (colour = phase)

More to play with