Neutral-atom quantum computing · Pasqal Pulser · a real run, explained

Five atoms solve a puzzle by relaxing into their lowest-energy state

Below is an actual experiment (run on a quantum emulator): five atoms, held in place by laser tweezers, work out the largest set of “antenna sites” that don’t interfere with each other. No algorithm searches for the answer — physics settles into it.

problem Maximum Independent Set atoms 5 pulse 4 µs shots 1000 blockade 9.1 µm
The problem

Pick as many sites as possible — but neighbors clash

Imagine five candidate spots for radio antennas. Antennas placed too close together interfere, so from any two clashing spots you may only pick one. What is the largest set of spots with no clashes? Computer scientists call this the Maximum Independent Set problem, and for large maps it is one of the notoriously hard ones — the number of combinations explodes.

A neutral-atom quantum computer has a party trick: this exact problem is built into its physics. One atom is placed at each site, at the real geometric position, spaced in micrometers. When a laser excites an atom into a high-energy Rydberg state, it physically prevents any atom within 9.1 µm — the blockade radius — from being excited too. “Too close to both be picked” isn’t a rule the machine checks. It is a law the atoms cannot break.

drag to orbit
t 0 ns Ω 0.00 δ −12.57
The whole experiment as a sequence diagram. During a run, the amber number tracks the live step. Only step 4 is quantum — here the “quantum device” is a physics-faithful emulator (QuTiP); the same program runs unchanged on a real Pasqal machine.
atom in ground state (“not picked”) atom in Rydberg state (“picked”) blockade range — overlapping spheres = too close conflict link
The recipe

Every step of the experiment, in order

The sequence diagram beside the 3D experiment above traces the whole run — who does what, and when. Here is each of its six steps in detail. Only step 4 is quantum: everything before it is ordinary programming; everything after it is ordinary checking.

1 Describe the problem

You list the five sites and their coordinates. That’s the whole input — a tiny map, measured in micrometers.

2 Encode it as geometry

Each site becomes one atom, held at that exact position by an optical tweezer. Distance does the encoding: any two atoms closer than 9.1 µm are automatically “in conflict.” Seven pairs qualify here.

3 Program the laser schedule

Instead of code, you write a 4-microsecond laser schedule: how strongly to shine (amplitude Ω) and how far off-resonance to tune (detuning δ) at every moment. This schedule is the entire “program.”

4 Let physics search

The sweep slowly makes “being picked” energetically attractive while blockade keeps clashing picks impossible. The atoms drift toward their lowest-energy arrangement — which is precisely the biggest conflict-free set. This is repeated 1000 times.

5 Measure

At the end of each run a camera snapshots the atoms. Each atom reads out as one bit — 1 picked, 0 not — so every shot returns a five-character bitstring like 10100.

6 Verify classically

Quantum answers are probabilistic, so a plain classical program checks every returned set against the conflict list, discards the rare invalid ones, and keeps the largest certified answer. Trust comes from this step, not from the quantum box.

Step 3, up close

The “program” is a laser schedule

Both knobs are plotted below in the same unit (rad/µs). The amplitude Ω rises and falls — turning quantum mixing on, then gently off. The detuning δ sweeps from negative (“being picked costs energy”) through zero to positive (“being picked saves energy”). Done slowly enough, the atoms stay in the lowest-energy state the whole way — so they finish in the lowest-energy answer. Physicists call this an adiabatic sweep: change the question slowly, and the system keeps giving its best answer.

The actual 4 µs schedule from the run: Ω interpolates through 0 → 9.42 → 0 rad/µs and δ through −12.57 → 0 → +12.57 rad/µs. The playhead follows the 3D scene above — scrub it and watch the atoms respond. Hover for exact values.
Steps 5 & 6 — the data

1000 shots, one histogram, one certified answer

A quantum computer doesn’t print “the answer.” It produces a distribution — run it many times and the good answers show up most often. Here are the five most frequent bitstrings from the real 1000-shot run (reading left to right: atoms n0 n1 n2 n3 n4):

valid — no two picked atoms conflict invalid — breaks a conflict (caught & discarded in step 6)
Measured counts out of 1000 shots on the QuTiP emulator. The winner 10100 = pick n0 and n2 — appeared in 59% of shots. The hatched rows are physics noise: 18 shots in 1000 landed on a clashing pair, which is exactly why the classical verification step exists.
{ n0, n2 } certified independent set, size 2 seen in 588 / 1000 shots brute-force check agrees: optimum is 2

On five atoms, a classical brute-force check (32 combinations) confirms the quantum answer instantly — that’s the point of a demo. The reason to care is scaling: the same recipe on hundreds of atoms explores a solution space no brute force could touch, while step 6 stays cheap — verifying an answer is easy even when finding it is hard.

If you remember three things

The layout is the program

The problem is encoded by physically placing atoms. Geometry — who is within 9.1 µm of whom — is the conflict graph. No translation layer, no circuit.

Constraints are enforced by nature

Rydberg blockade makes invalid answers energetically forbidden rather than merely discouraged. The machine doesn’t check the rule; it embodies it.

Trust the verifier, not the oracle

Every quantum output is checked by a boring classical program before it is believed. Probabilistic engine, deterministic gatekeeper.