Quantum decoherence describes how a quantum system stops behaving quantum mechanically once it interacts with its environment. Every stray photon, air molecule, or vibration that touches a particle carries away a tiny bit of information about its state, and that leakage destroys the clean superposition the particle was in.
This is why superpositions are so hard to see in everyday life. A single atom can be kept in superposition in a carefully isolated lab, but the moment it bumps into the outside world, decoherence sets in almost instantly, and it starts behaving like an ordinary, classical object.
Decoherence is also the biggest engineering obstacle in building quantum computers. Qubits have to be shielded from heat, vibration, and stray fields, because any leak of information to the environment ruins the calculation before it finishes.
Decoherence offers a fairly practical resolution to the famous Schrodinger’s cat thought experiment, in which a cat is imagined as being simultaneously alive and dead until observed. In reality, a cat sized object is constantly bombarded by air molecules, photons, and its own internal heat, all of which interact with its surroundings and destroy any quantum superposition almost instantaneously, long before a human observer ever needs to open the box. This means decoherence does not require a conscious observer or any special act of measurement to resolve superpositions, ordinary physical interaction with the environment does the job on its own. It is part of why quantum weirdness, so evident with single particles in a lab, essentially disappears by the time you scale up to anything visible to the naked eye.
Quantum computer engineers fight decoherence with techniques like quantum error correction, which spreads a single unit of quantum information, a qubit, redundantly across several physical qubits so that a small amount of environmental noise on any one of them does not immediately destroy the calculation. Building a large, reliably error corrected quantum computer, capable of outperforming classical computers on useful problems, remains one of the field’s central engineering challenges.