English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Where Quantum Errors Come From: T1, T2 and the "Lifespan Table"

Forum topic · 小凯 · 2026-11-07

Summary

This article, episode 051 of a Chinese quantum computing column, explains the two core metrics of qubit health: T1 (relaxation time, the energy lifetime over which a qubit decays from |1⟩ back to |0⟩, producing amplitude errors) and T2 (decoherence time, the lifetime over which phase information is lost, producing phase errors), with the built-in constraint T2 ≤ 2·T1. It identifies four major noise sources: thermal/environmental radiation (hence dilution refrigerators operating near 0.01 K), control-line crosstalk and gate imprecision, atomic-scale material defects such as two-level systems, and measurement readout errors. Any single-qubit error decomposes into combinations of X (bit flip), Z (phase flip), and Y (both) errors, discretizing continuous noise—the foundation of quantum error correction. Typical T1/T2 values of microseconds to milliseconds allow only about a thousand gate operations, far below the depth required by algorithms like Shor's, motivating two paths: improving hardware coherence times and quantum error correction. The next episode covers why quantum error correction cannot simply recompute, due to the no-cloning theorem.

> Prerequisite navigation: You need 006 (coherence and decoherence), 012 (the state sphere), 017 (axis swaps), 016 (measurement). > Next episode preview: 052, "Why You Can't Just Recompute: The Innate Difficulty of Error Correction."

Where Quantum Errors Come From: T1, T2 and the "Lifespan Table"

A hospital checkup report always has a "reference value" column—how far your indicator is from the reference determines whether the doctor frowns.

A qubit also has a checkup report, and its two core metrics are T1 and T2—jokingly called the "lifespan table" in the industry. Batch six (errors and error correction, the hardest batch in this column) starts from this checkup report: what errors actually are, where they come from, and how fast things die. Without understanding the enemy, none of the defenses later in batch six (codes, syndrome checks, surface codes, thresholds) can even be discussed.

The Two Lines on the Checkup Report

Let's read the table directly (the state sphere from episode 012 is the frame of reference):

T1—relaxation time (the "falling over" lifespan). On the qubit's state sphere, the north pole is |0⟩ and the south pole is |1⟩. In real devices, the arrow slides down on its own—energy leaks into the environment (lattice vibrations, stray radiation), and the high-energy state |1⟩ slowly decays back to the low-energy state |0⟩, like a spinning top that gradually topples over. T1 is "how long, on average, before it slides down"—it is the "energy lifetime."

T2—coherence time (the "forgetting" lifespan). What happens even earlier than falling is twisting: the arrow is still on the sphere's surface, but its longitude angle begins to drift (the phase scrambling of episode 006—the environment's electromagnetic noise keeps giving the arrow "random kicks"). Once the relative phase relationships are scrambled, interference is ruined—T2 is "how long, on average, before the phase is completely forgotten." Mathematically, T2 can never exceed about half of T1 (falling necessarily comes with losing phase—a dying top can't mind its posture). This inequality is built into the logic of the checkup table.

One sentence to remember: T1 governs falling (amplitude errors), T2 governs drift (phase errors)—the two axes of episode 012, "latitude governs ratios, longitude governs phase," map exactly onto the two diseases.

The Patient List: Where Errors Come From

The checkup report only gives conclusions; the causes must be investigated in the ward. A qubit's list of "noise sources" is short, but every one is nasty:

One: heat and environmental radiation. Superconducting qubits live at 0.01 kelvin (the dilution refrigerator from episode 007)—precisely to suppress this: any stray thermal photon entering the circuit is an "energy injection" or a "phase kick." Refrigeration isn't for coldness—it's for clearing the field.

Two: crosstalk on control lines. You want to rotate qubit 3, but neighboring qubit 4 gets dragged along a little too—gate errors (intending a 90-degree rotation but getting 90.5 degrees) and crosstalk are the same family: control-precision problems.

Three: material defects. Atomic-scale "traps" on the chip surface and interfaces (jargon: two-level systems/defects) act like tiny noise machines that randomly absorb and release energy—one of the top killers of T1.

Four: errors in measurement itself. Readout also has a misjudgment rate (reading a 0 as a 1)—in the error-correction system this is a first-class citizen and cannot be ignored.

The common feature of these four diseases deserves its own line: none of them "negotiate"—they occur at any time and stack randomly. Noise is continuous, while the consequences of errors are discrete (the next sentence is the key to batch six, expanded in episode 052).

The "Three Faces" of Errors: X, Y, Z Return

The trio from batch three (episode 020) returns here in a villain role. Mathematically, any single-qubit error can be decomposed into a combination of the trio:

  • X error: the arrow flips half a turn around the X axis—|0⟩↔|1⟩, a bit flip (the discrete incarnation of T1-type disease);
  • Z error: the arrow rotates half a turn along a meridian—the phase is inverted, a phase flip (the discrete incarnation of T2-type disease);
  • Y error: both at once.
  • This "trio decomposition" is not mathematical decoration—it is the foundation of the entire error-correction empire: as long as all errors can be viewed as "combinations of a finite set of discrete errors," "check and repair" has a foothold (from episode 053 onward). Continuous noise is discretized by the geometry of the state sphere—the luckiest thing about this batch.

    Reading the Lifespan Table and "How Fast Errors Happen"

    In engineering reports, T1 and T2 are measured in microseconds to milliseconds—doesn't sound short? Convert it into "how many gates can be executed" and your heart sinks immediately: a single gate operation takes on the order of hundreds of nanoseconds, so the coherence time supports about a thousand gates—this is the ledger behind episode 030's claim that "depth is the hard currency." Once circuit depth exceeds a thousand layers, the signal sinks to the bottom.

    Meanwhile, Shor's algorithm needs n³ layers (episode 038), and chemistry simulations need thousands upon thousands of frames—a gap of four to five orders of magnitude. How to fill this gap? Two legs: ① make T1/T2 themselves longer (materials and fabrication, a slow variable, the hardware race of batch seven); ② error correction (compounding a thousand-layer lifespan into arbitrary length—the centerpiece around episode 068). The next episode covers something counterintuitive first: why error correction cannot rely on "recomputing."

    Dismantling the Analogy

    What the checkup-report analogy borrowed—and what it didn't:

  • Dissimilarity #1: Human checkup indicators have a "normal range"; only exceeding it counts as disease. A qubit's T1/T2 are exponential-decay time constants—there is no "critical point"; every microsecond things degrade at a fixed rate, differing only in speed.
  • Dissimilarity #2: Diseases can be treated with medicine or heal on their own; quantum errors, if untreated, only accumulate—decoherence is one-way entropy increase (episode 006). "Rest a while and recover" does not exist in the quantum world. The analogy borrows only "two metrics govern the whole body," not "wait for it to heal."

Luggage to Take Away

1. T1 = the lifespan of falling (amplitude/energy errors), T2 = the lifespan of drifting (phase errors); the built-in logic of T2 ≤ 2·T1—episode 012's two state-sphere axes correspond to the two diseases. 2. Four major noise sources: thermal radiation (refrigeration clears the field), control crosstalk (same family as gate errors), material defects (T1's top killer), readout misjudgment—noise is continuous, consequences are discrete; this mismatch is the starting point of error correction. 3. The trio returns as villains: any single-qubit error = a combination of X/Z/Y—continuous noise is discretized by geometry, giving "check and repair" its foothold; coherence time ≈ a thousand-gate depth, orders of magnitude short of Shor's n³—walk on two legs: fabrication extends lifetimes, error correction compounds them.

Hook for the next stop: In the classical world, error correction is a matter of course—"copy three copies, majority rules." In the quantum world, this method violates rule number one—no-cloning (episode 028). If you can't copy, how do you correct? The clever detour behind this "impossibility" is the most elegant beginning of all quantum error correction. Episode 052: "Why You Can't Just Recompute: The Innate Difficulty of Error Correction."

Tags: quantum-computing, from-coins-to-error-correction, coherence-time, noise-model

Tags

#quantum-computing#qubit-errors#t1-relaxation#t2-decoherence#noise-model#error-correction#coherence-time

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178636885