Chapter 03

Measurement & Collapse

Amplitudes are private. The only way information leaves a qubit is a measurement — which returns a single bit and destroys the blend. Here is how the Born rule plays out in practice, and the matrices that describe it.

↩ before you start · keep these handy
·From Ch. 2: the Born rule, P = |amplitude|², and |ψ⟩ = α|0⟩ + β|1⟩.
·From Ch. 2: |+⟩ = (|0⟩ + |1⟩)/√2, and that overlap measures how distinguishable two states are.
·From 0.3: probabilities are slices of one pie; over many tries the observed frequencies settle onto them.
🔑 symbol decoder · every new mark, in plain words
collapsethe jump a state makes at the instant of measurement: from a blend to a single definite outcome. P₀, P₁projector matrices, one per outcome. P₀ = |0⟩⟨0| “keeps only the |0⟩ part” of a state. ⟨ψ|P₀|ψ⟩a “sandwich” — bra, matrix, ket multiplied in a row. It returns one number: here, P(0). ⟨Z⟩the expectation (long-run average) of Z, where outcome 0 scores +1 and outcome 1 scores −1. Z, X basistwo different questions you can ask: Z asks “0 or 1?”, X asks “|+⟩ or |−⟩?”. diag(a, b)a matrix with a, b down the diagonal and zeros elsewhere.
§1

One look, one bit

Measure a qubit and you get back a single 0 or 1; the state collapses onto that answer and forgets the rest. One shot reveals almost nothing. But prepare the same state again and again, and the tally of outcomes converges on |α|² and |β|².

Set the qubit's bias, then run shots and watch the bars climb onto the dashed odds.

💡 everyday picture

A coin spinning on a tabletop is, for that moment, neither heads nor tails. Slap your hand down and it commits — instantly, and at random — to one face, and there it stays. Measurement does the same to a qubit: it forces a single answer and leaves it stuck. The one twist a coin lacks: with a qubit, which way you slap (which basis you measure) changes the odds you see.

▸ runmeasure(shots)
last shot
{{ lastShot }}
{{ t0pct }}%
0
{{ t1pct }}%
1
{{ total }} shots · {{ hz }} : {{ ho }}
law of large numbers:
frequencies → probabilities as shots → ∞
recapOne shot returns a single bit; the squared-amplitude odds only emerge as a tally over many fresh copies.
§2

Collapse is permanent

Here is the part that trips everyone up: a measurement doesn't just read the qubit, it changes it. The first look lands on a 0 or 1 at random — and from then on the qubit is stuck there. Ask again and you get the same answer, forever. The blend is gone.

Prepare a 70/30 qubit, then keep measuring the same copy — not a fresh one each time.

▸ livequbit.collapse
current state
{{ collLabel }}
{{ collSub }}
reads, in order
— not measured yet — {{ r.v }}
{{ collMsg }}

Compare with §1: there, every shot re-prepared a fresh qubit, so the tally spread across both outcomes. Here a single qubit is measured repeatedly — and it never wavers after the first look.

recapThe first look fixes the value for good — repeat reads of the same qubit never change.
§3 · the mathematics

Measurement as a projector

A measurement in the 0/1 basis is described by two projector matrices, P₀ = |0⟩⟨0| and P₁ = |1⟩⟨1|. Each picks out one slot:

P₀ =1000
P₁ =0001

The probability of outcome 0 is the "sandwich" ⟨ψ|P₀|ψ⟩ = |α|² — exactly the Born rule, now as matrix arithmetic.

recapA projector keeps one outcome; the sandwich ⟨ψ|P₀|ψ⟩ is just the Born rule written as matrix arithmetic.
§4 · the mathematics

Expectation value

Label outcome 0 as +1 and outcome 1 as −1 — that's the Z observable, the matrix diag(+1, −1). Its average over many shots is

✎ worked example · ⟨Z⟩ of a 70/30 qubit
1.the qubit gives P(0) = 0.7, P(1) = 0.3
2.score 0 as +1 and 1 as −1, then take the weighted average:  ⟨Z⟩ = (+1)(0.7) + (−1)(0.3)
3.= 0.7 − 0.3 = +0.4  — leaning toward 0, but not certain
⟨Z⟩ = P(0) − P(1) = {{ expZ }}

It runs from +1 (a sure 0) through 0 (an even superposition) to −1 (a sure 1) — and it's exactly the height of the Bloch arrow you'll meet next chapter.

recap⟨Z⟩ = P(0) − P(1) runs from +1 (sure 0) to −1 (sure 1) — and it is the height of the Bloch arrow.
§5

There is no single “true” value

So far we always asked the same question: “are you 0 or 1?” But that is a choice of measurement — the 0/1 axis, called the Z basis. You can just as well ask along the X basis, whose two answers are |+⟩ and |−⟩.

The punchline: a state that is perfectly certain in one basis is a fair coin in the other. Switch the state and the question and watch.

▸ livemeasure(basis)
prepared state
measure in basis
P({{ mbLbl0 }}){{ mbP0 }}%
P({{ mbLbl1 }}){{ mbP1 }}%
{{ mbVerdict }}

The reason is overlap (Chapter 2): |0⟩ = (|+⟩ + |−⟩)/√2, so a definite |0⟩ is an even split along X — and vice versa. The two bases are maximally incompatible; sharpness in one is total blur in the other. That trade-off is the seed of the uncertainty principle.

recapThere is no single “true” value — which basis you measure is a choice, and sharpness in one is a fair coin in the other.
⚠ common misconception

“The qubit secretly was 0 the whole time — measuring just revealed it.” Tempting, but false. There was no hidden 0 or 1 waiting to be found: a state on the equator is genuinely both, and §5 shows the answer you get even depends on which question you ask. Bell's theorem in Chapter 6 closes the door on every “it was decided all along” story — experimentally.

✓ you can now
predict measurement frequencies, and say why a single shot tells you almost nothing
explain why collapse is permanent — the blend is gone after the first read
compute a projector sandwich and ⟨Z⟩, and predict how the odds change with the measurement basis
← 02 The Qubit next · 04 The Bloch Sphere