In short: Decoherence is the process by which a qubit loses its fragile quantum superposition because it cannot be perfectly isolated from its surroundings. The environment "measures" the qubit through stray fields, heat and vibration, scrambling phase and relaxing energy on characteristic timescales called T1 and T2. Fighting decoherence is the central engineering challenge of building a useful quantum computer.
Decoherence is the loss of the delicate quantum information a qubit holds when it interacts with an uncontrolled environment. A classical bit is either 0 or 1, but a qubit can sit in a superposition — a weighted blend of 0 and 1 with a well-defined relative phase. That phase is where quantum algorithms get their power, and it is also the first thing the outside world destroys.
No qubit is ever truly alone. It is coupled, however weakly, to photons, phonons, nearby spins, fluctuating electric fields and the readout wiring itself. Each of these interactions effectively leaks a tiny amount of "which-state" information into the environment. Once that information escapes, the superposition collapses into an ordinary statistical mixture — the qubit still has a 0 and 1 population, but the phase relationship between them is gone. That irreversible leak is decoherence, and it is why qubits are fragile.
A single qubit's state can be drawn as an arrow pointing somewhere on the surface of a sphere — the Bloch sphere. The north and south poles are the definite states |0⟩ and |1⟩; any point on the equator is an equal superposition whose longitude encodes the phase. Decoherence shrinks and drags this arrow inward over time.
Animating the arrow makes the two decay mechanisms intuitive: the vector precesses around the vertical axis (its phase evolving), while slowly spiralling toward the north pole as energy drains away.
Engineers quantify how long a qubit survives with two time constants, both measured in microseconds for today's best superconducting and trapped-ion devices.
T1 is the characteristic time for an excited qubit to lose its energy and decay from |1⟩ back to |0⟩, exactly as an atom emits a photon and drops to its ground state. It is set by energy exchange with the environment — lossy dielectrics, two-level defects in the materials, or coupling to the readout line. The excited-state population decays as e−t/T1.
T2 is the time over which the phase relationship between |0⟩ and |1⟩ stays well defined. Even if no energy is lost, slow fluctuations in the qubit's transition frequency — caused by magnetic-field noise, charge noise or temperature drift — cause the phase to wander unpredictably. Because any energy-loss event also destroys phase, T2 can never exceed 2×T1; in practice pure dephasing usually makes it shorter.
The cruel trade-off of quantum computing is that a good qubit must be both well isolated and easily controllable. Any port you open to write a gate or read a result is also a port through which the environment can reach in. Three factors make the fragility acute:
Decoherence cannot be eliminated, only outrun. The field attacks it on several fronts:
The race is simple to state: perform and verify your computation — or your error-correction cycle — faster than coherence decays. Longer T1 and T2 mean more gate operations per coherence window, which is exactly why those two numbers headline every hardware announcement.
Real projects and industry tool flows, with certification.
Explore CourseTron Courses →They are closely related but not identical. Decoherence is the gradual, physical leakage of phase information into the environment, which turns a pure superposition into a classical-looking mixture. It explains why quantum effects vanish at macroscopic scales, but the final "pick one outcome" of a measurement is a separate interpretational question.
In principle the full qubit-plus-environment system stays coherent, but recovering that information is practically impossible because it has dispersed into countless environmental degrees of freedom. Techniques like dynamical decoupling undo predictable drift, but true random decoherence is effectively irreversible.
Leading superconducting qubits reach roughly 100–300 microseconds for T1 and T2, while trapped-ion and neutral-atom qubits can stay coherent for seconds. The right figure of merit is coherence time divided by gate time — how many operations fit inside the window.
Per qubit, not necessarily, but larger chips add crosstalk, wiring and packaging that introduce fresh noise channels. Keeping per-qubit coherence high while scaling up is one of the hardest engineering problems in the field.
Related: Quantum error correction · Glossary · all courses.