Ask a quantum engineer how good their qubit is and they will almost always answer with two numbers. The first is called T1. The second is called T2. They are measured in microseconds, occasionally milliseconds, and they represent the two ways a qubit forgets what you told it. Understanding them is the fastest route to understanding why quantum computing is so hard.
Why a qubit needs two numbers
A classical bit forgets in only one way: it flips. A qubit is richer. Its state is not just a zero or a one but a blend of the two, described by an amplitude and a phase. You can picture it as an arrow pointing somewhere on the surface of a sphere. The height of the arrow encodes the probability of measuring one versus zero. The direction it points around the equator encodes the phase, the delicate quantum information that makes interference and entanglement possible.
There are two ways for that arrow to drift. It can fall from the top of the sphere toward the bottom, losing energy to its surroundings. Or it can keep its height but smear out around the equator, losing track of which way it was pointing. The first process is timed by T1. The second is timed by T2. A qubit can be excellent at one and terrible at the other, which is exactly why one number is never enough.
T1: the energy leak
T1 is the relaxation time, sometimes called the amplitude damping time. Left alone in its excited state, a qubit will eventually dump its energy into whatever it is coupled to and slump back to the ground state. T1 is the characteristic time for that decay. After one T1, roughly 63 percent of the excited population has leaked away.
Engineers measure it with a simple recipe. Excite the qubit, wait a variable amount of time, then read it out. Repeat thousands of times for each wait interval and watch the excited-state probability decay along an exponential curve. The time constant of that curve is T1. Long T1 values come from clean materials, careful shielding, and designs that keep the qubit from acting like an antenna that radiates its energy away.
T2: the phase leak
T2 is the coherence time, and it is usually the one that hurts. Even if a qubit never loses a scrap of energy, small fluctuations in its environment can nudge its frequency up and down. Each nudge advances or retards the phase by a tiny random amount, and over many runs those tiny errors add up to a blur. When the phase information is gone, so is the quantum in quantum computing.
Physicists actually track two flavors here. T2-star is the raw dephasing time you see in a straightforward experiment, dominated by slow drifts and inhomogeneities. T2, sometimes written T2-echo, is the longer time you recover when you apply a clever refocusing pulse partway through, flipping the qubit so that slow drifts cancel themselves out. That refocusing trick is the seed of dynamical decoupling, a whole family of pulse sequences used to stretch coherence in idle qubits.
The rule that ties them together
There is a hard mathematical relationship between the two: T2 can never exceed twice T1. Energy loss is itself a form of phase loss, so the amplitude clock caps the phase clock. When you see a qubit whose T2 sits close to that 2-times-T1 ceiling, it means dephasing has been beaten down so far that energy relaxation is the only thing left limiting it. That is the sweet spot every hardware team chases.
Why the numbers set the budget
These times matter because gates take time too. A single- or two-qubit operation might run in tens or hundreds of nanoseconds. Divide a coherence time by a gate time and you get a rough estimate of how many operations you can perform before the qubit turns to noise. That ratio is the real currency of the machine. A qubit with a millisecond T2 and a hundred-nanosecond gate can attempt thousands of steps; one with a ten-microsecond T2 gets a fraction of that.
It also explains the split personalities of different hardware. Trapped-ion qubits boast enormous coherence times, seconds in some cases, but slow gates. Superconducting qubits have shorter coherence but blisteringly fast gates. Neither wins outright, because what matters is the ratio, not either clock alone.
Error correction eventually promises to hide these limits behind logical qubits built from many physical ones. But the physical clocks never disappear. Push T1 and T2 higher and the whole tower gets cheaper to build. That is why so much of the field's grind, new materials, better shielding, cleaner fabrication, is really a quiet fight to add a few more microseconds to two very small stopwatches.
