What makes a good qubit
By now you know a qubit in the abstract: a two-level quantum system whose state is |psi> = alpha|0> + beta|1>, with |alpha|^2 + |beta|^2 = 1. That is the math. The hardware question is brutally concrete — what real physical thing can you point a wire or a laser at and have it behave like that? You need a system with two clean energy levels you can call |0> and |1>, that you can put into a superposition of them, and that you can read out at the end.
It turns out nature is full of two-level systems, but almost all of them are useless for computing because they fail one of four requirements. A good qubit has to be controllable (you can apply gates on demand), measurable (you can read 0 or 1 at the end via the Born rule), connectable (it can interact with other qubits to make entanglement and two-qubit gates), and — the hard one — isolatable. The same coupling that lets you control a qubit also lets the environment disturb it, and any stray nudge causes decoherence, quietly scrambling the state before your computation finishes.
Superconducting qubits
The first approach is to build an artificial atom out of an electrical circuit instead of borrowing a real one. A superconducting qubit is a tiny circuit etched onto a chip, much like an ordinary microchip, but chilled to a few thousandths of a degree above absolute zero (a few millikelvin). At that temperature the metal becomes superconducting — current flows with zero resistance — and the whole circuit can behave like one quantum object.
A plain electrical circuit, though, has *evenly spaced* energy levels, like a ladder with identical rungs. That is a problem: if your |0>-to-|1> step is the same size as the |1>-to-|2> step, any pulse meant to flip the qubit will also kick it up to level 2 and beyond, and it stops being a clean two-level system. The fix is a special component called a Josephson junction, which makes the rungs uneven. Now the lowest two levels are spaced differently from the rest, so you can single them out as |0> and |1> and leave the higher levels alone.
The transmon
Early superconducting qubits had a crippling weakness: they were painfully sensitive to charge noise. Stray electric charges drifting near the chip would jostle the energy levels and scramble the information almost immediately, giving you a coherence time far too short to compute with. The transmon is the design that fixed this, and it is now by far the most common kind of superconducting qubit — when people say 'IBM and Google qubits,' they almost always mean transmons.
The transmon's trick is simple to state: add a relatively large capacitor across the circuit. This flattens out how much the qubit's energy depends on stray charge, so much so that charge noise barely matters anymore. That is the main reason transmons hold their state long enough to be useful. The trade-off is honest and worth naming: the larger capacitor also makes the energy levels *more* evenly spaced again (less anharmonicity), so |0> and |1> sit closer in spacing to the unwanted higher levels. Engineers manage this with carefully shaped microwave pulses that nudge the qubit between |0> and |1> without leaking it up to level 2.
Trapped ions
The opposite philosophy is to stop building artificial atoms and just use a real one. A trapped-ion qubit stores the information inside a single charged atom (an ion), held almost perfectly still in empty space by carefully shaped electric and magnetic fields — like a marble settling into the bottom of an invisible bowl. The atom has internal energy levels; you pick two of them as |0> and |1>. You write and read with precisely tuned lasers: one laser flips the qubit between its two levels, and another makes the atom glow for one state and stay dark for the other, so a camera can tell which it is.
Because every atom of a given element is *naturally identical* and the ion floats far from any surface, trapped ions are extremely well isolated. That buys two real advantages over superconducting chips: much longer coherence times and very high gate fidelity — operations come out close to what you intended, with few errors. Ions in one trap can also interact with many partners through their shared vibration, which makes two-qubit gates flexible.
Photonic qubits
A third route encodes the qubit in a single particle of light — a photon. A photonic qubit uses a yes-or-no degree of freedom of one photon, such as its polarization (vertical versus horizontal) or which of two paths it travels. The big appeal: photons barely interact with their surroundings, so they hold their state well and need no dilution refrigerator — many photonic systems run at room temperature. And since light already travels through fibers and free space, photons are the natural carrier for sending quantum information between machines, which makes them a strong fit for communication, networking, and quantum key distribution.
But the same indifference that makes photons so isolatable is also their curse. Two photons normally pass straight through each other without noticing, so building a two-qubit gate — where one qubit must conditionally affect another — is genuinely hard. The common workarounds lean on measurement and extra 'helper' photons and only succeed part of the time, so they need clever schemes to become reliable. Great messenger, awkward processor.
# A 'qubit' in software is platform-agnostic: the same Qiskit # circuit can target a transmon, an ion, or a photonic backend. # The hardware differs; the logical description does not. from qiskit import QuantumCircuit qc = QuantumCircuit(2, 2) qc.h(0) # put qubit 0 into a superposition (a Hadamard gate) qc.cx(0, 1) # entangle qubit 1 with qubit 0 (a CNOT) qc.measure([0, 1], [0, 1]) # read out -> one outcome, by the Born rule print(qc) # ┌───┐ ┌─┐ # q_0: ─┤ H ├──■──┤M├─── # └───┘┌─┴─┐└╥┘┌─┐ # q_1: ──────┤ X ├─╫─┤M├ # └───┘ ║ └╥┘ # c: 2/════════════╩══╩═ # 0 1