Quantum Computing Relatable — Part 3
The first two parts treated a qubit as a clean object that obeys perfect rules. Real hardware is much less cooperative.
The information in a qubit is not stored as an ordinary bit sitting in a box. It is a carefully controlled physical state. Heat, vibration, electromagnetic noise and unwanted interactions with the environment can change it. The longer and more complicated the calculation, the more chances there are for something to go wrong.
That is why a quantum computer is not just a normal computer with a more exotic processor. The engineering problem is to create a quantum system, control it, keep it isolated, and still measure it when the algorithm is finished.
Superconducting circuits
One approach uses superconducting circuits. These are manufactured circuits whose behaviour becomes quantum mechanical at very low temperatures.
The chip sits inside a dilution refrigerator, with several temperature stages leading down towards absolute zero. At those temperatures, the circuit can support carefully controlled quantum states. Microwave pulses act as instructions: they rotate the state of a qubit, create relationships between qubits, and eventually help us measure the result.
The refrigerator is not decorative. Thermal energy is noise, and noise destroys the delicate state the algorithm is trying to preserve. The control electronics also have to send signals into a cold environment and read signals back out without disturbing the computation.
Superconducting qubits can be operated quickly and manufactured using techniques related to conventional chip fabrication. They also lose their state relatively quickly, and the control wiring and calibration become harder as the system grows.
Trapped ions
A different approach holds individual charged atoms, such as ytterbium ions, in an electromagnetic trap. Lasers address the ions and control their internal states.
The ions are naturally very similar to one another, and they can remain coherent for a relatively long time. The trade-off is speed: laser operations can be slower, and arranging large numbers of ions and their connections is difficult.
This is not a contest in which one technology has already won. Superconducting circuits, trapped ions, neutral atoms, photons and other approaches make different engineering choices. Each one has to solve the same underlying problem: perform enough accurate operations before errors overwhelm the useful information.
A physical qubit is not a logical qubit
The qubits a company counts in a hardware announcement are usually physical qubits. They are the actual devices that interact with the control system, and they are noisy.
A logical qubit is an error-protected unit of quantum information encoded across multiple physical qubits. Additional measurements look for patterns that indicate an error without directly measuring and destroying the logical state.
This sounds like a simple backup system, but quantum information cannot be copied in the same way as an ordinary file. Error correction therefore uses carefully designed codes, repeated checks and a large amount of extra hardware.
The overhead can be substantial. A machine with many physical qubits may still have only a small number of useful logical qubits, especially if the physical error rates are not low enough. This is why a raw qubit count tells us much less than a headline suggests.
IBM's introduction to fault tolerance explains the relationship between physical and logical qubits.
Gates are control pulses
On paper, a quantum circuit is a tidy sequence of gates: Hadamards, controlled operations and measurements. In a laboratory, a gate is a calibrated physical interaction.
For a superconducting qubit, it may be a microwave pulse with a precisely chosen frequency, duration and phase. For a trapped ion, it may be a laser pulse. A two-qubit gate requires the devices to interact in a controlled way while the rest of the system remains quiet.
Every pulse is an opportunity for a small error. The algorithm assumes that the intended operation happened, but the hardware produces an approximation. Quantum error correction only helps when the error rate is low enough and the errors have the right structure for the code to detect them.
This is also why “just add more qubits” is not a complete strategy. More devices create more possible interactions, more calibration work and more routes for errors to spread.
What the machine actually does
The classical computer does not disappear when a quantum processor is connected. It prepares the input, compiles the circuit, schedules the job, controls the hardware, collects measurement results and often optimizes the next circuit.
The quantum processor executes a particular part of that workflow. The useful system is hybrid: classical and quantum processors working together.
That makes the question “Is the quantum computer faster?” incomplete. Faster at which subroutine, compared with which classical method, at what accuracy, and after including the cost of error correction and data movement?
The next part looks at the applications where those questions have concrete answers, and the claims where they do not: Part 4 — What Are Quantum Computers Actually Good For?.