Why Quantum Machines Make Errors
A plain-language map of where quantum errors come from.
People often hear “quantum computers are noisy” without being told what that sentence actually contains. The useful answer is not “because qubits are fragile.” The useful answer is a list of failure modes.
Once you name the error sources, the machine stops looking mystical. It starts to look like a system with preparation problems, timing limits, control imperfections, and measurement uncertainty.
The High-Level Map
Most errors happen before, during, or after the intended computation.
Before the circuit, you can prepare the wrong state. During the circuit, the qubit can decohere, drift, couple to neighbors, or experience an imperfect pulse. After the circuit, the readout can misclassify what state was actually present.
That simple timeline helps: initialize, evolve, measure. Each phase introduces its own kind of error. Hardware papers describe them precisely, but the conceptual structure is easy enough to keep in your head.
The Main Sources
The list for non-specialists: state preparation, decoherence, gate error, cross-talk, leakage, readout, and drift.
- State preparation error: the qubit does not start exactly where the circuit assumes it starts.
- Decoherence: interaction with the environment destroys the phase or energy state over time.
- Gate error: the control pulse is imperfect, so the intended operation is only approximate.
- Cross-talk: acting on one qubit disturbs another nearby qubit.
- Leakage: the device wanders outside the computational subspace you thought you were using.
- Readout error: measurement reports the wrong classical bit value.
- Drift: the calibration changes over time, so yesterday’s good circuit is not today’s good circuit.
What This Means Operationally
Different errors call for different responses.
If readout is your main issue, mitigation and calibration matrices help. If depth is killing you, shallower circuits and better layout matter more. If drift dominates, run-time scheduling and repeated calibration awareness become central. There is no universal “error knob.”
That is why a useful experiment log should always record backend, calibration window, transpilation output, depth, shot count, and the comparison to a simulator. Without that context, “the quantum machine was noisy” says almost nothing.
Summary
Quantum noise is not one thing. It is a stack of failure modes.
For a beginner, the most important shift is from vague fragility to a named error budget. Once errors are broken into preparation, evolution, and measurement problems, the machine becomes debuggable in principle.
Continue the Quantum FAQ
The next story disentangles three terms that are often used as if they were interchangeable: correction, suppression, and mitigation.