Deterministic State Evolution and Probabilistic Measurement Outcomes

Welcome back. In the previous lesson, we established a precise and limited claim: for a closed system with a specified Hamiltonian, the Schrödinger equation maps each initial quantum state to one unique later quantum state. The map is unitary, reversible, and deterministic.

This lesson identifies why that result does not yet tell us that a measurement has one uniquely determined result. The key move is to include the measuring apparatus in the quantum description. Unitary dynamics then determines a unique entangled system-apparatus state—but, when the measured system began in a superposition, that state generally does not correspond to one definite pointer reading.

By the end, you should be able to separate two assertions that are often conflated:

The total quantum state evolves uniquely.
One particular measurement outcome is uniquely realized.

Two different targets for “determinism”

Let a system have state $∣ ψ ⟩$ , and suppose a measurement has possible recorded outcomes $i$ . There are two logically different maps one might seek.

State determinism is a map on quantum states:

∣ ψ (t_{0})⟩ ⟼ ∣ ψ (t)⟩ = U (t, t_{0}) ∣ ψ (t_{0})⟩ .

For closed-system Schrödinger dynamics, this map is unique.

Outcome determinism would instead require that the complete physical state before an experiment fixes one particular outcome:

complete physical state before measurement ⟼ i .

These maps have different codomains. The first produces another vector (or ray) in Hilbert space. The second produces one element of an outcome set, such as “up” or “down.”

The fact that the first map is deterministic does not logically entail that the second is. A deterministic law can take a state into a superposition or entangled state rather than into one of several mutually exclusive classical-looking alternatives.

This is not a technicality about incomplete calculations. It is a consequence of linearity.

A small but important clarification: applying an observable operator to a state is not the same physical act as measuring that observable. An operator represents an observable in the formalism; an actual measurement is a dynamical interaction involving a system, an apparatus, and eventually many environmental degrees of freedom.

The unitary model of an ideal measurement

Take a two-outcome observable of a system $S$ , with orthonormal eigenstates

∣0 ⟩_{S}, ∣1 ⟩_{S} .

Let the apparatus $A$ initially be in its ready state,

∣ R ⟩_{A} .

A functioning ideal measurement correlates each system eigenstate with a distinguishable apparatus record:

U_{S A} (∣0 ⟩_{S} ∣ R ⟩_{A}) = ∣0 ⟩_{S} ∣ A_{0} ⟩_{A},

U_{S A} (∣1 ⟩_{S} ∣ R ⟩_{A}) = ∣1 ⟩_{S} ∣ A_{1} ⟩_{A} .

Here $∣ A_{0} ⟩_{A}$ means “the apparatus displays outcome $0$ ,” while $∣ A_{1} ⟩_{A}$ means “the apparatus displays outcome $1$ .” In an idealized macroscopic measurement, these record states are effectively orthogonal:

⟨ A_{0} ∣ A_{1} ⟩ \approx 0.

This is the desirable calibration condition. If the system enters already in an eigenstate, the apparatus produces the matching record.

Read Preskill’s account of the physical pointer coupling before applying the same structure to a superposed input.

Chapter 3 Foundations II: Measurement and Evolution

In John Preskill’s lecture notes, the opening discussion of von Neumann measurement shows how an ordinary Hamiltonian coupling correlates eigenstates of an observable with shifted, distinguishable pointer states. Read it to keep the apparatus model physical rather than treating “measurement” as a primitive black box.

In Chapter 3, Section 3.1.1, “Orthogonal Measurements,” read from the discussion beginning with the system-pointer correlation. Then continue through the pointer Hamiltonian and its action on a superposed input, ending at the explanation that the pointer becomes correlated with the observable value. Focus on the role of the unitary coupling U: it creates correlations but does not, at this stage, select one correlated term.

The model becomes decisive when the incoming system is not an eigenstate. Suppose instead that

∣ ψ ⟩_{S} = α ∣0 ⟩_{S} + β ∣1 ⟩_{S}, ∣ α ∣^{2} + ∣ β ∣^{2} = 1,

with both $α$ and $β$ nonzero. The initial joint state is

∣ Ψ_{in} ⟩ = (α ∣0 ⟩_{S} + β ∣1 ⟩_{S}) ∣ R ⟩_{A} .

Because $U_{S A}$ is linear,

U_{S A} ∣ Ψ_{in} ⟩ = α U_{S A} (∣0 ⟩_{S} ∣ R ⟩_{A}) + β U_{S A} (∣1 ⟩_{S} ∣ R ⟩_{A}) = α ∣0 ⟩_{S} ∣ A_{0} ⟩_{A} + β ∣1 ⟩_{S} ∣ A_{1} ⟩_{A} .

Call this final state $∣ Ψ_{pre} ⟩$ :

∣ Ψ_{pre} ⟩ = α ∣0 ⟩_{S} ∣ A_{0} ⟩_{A} + β ∣1 ⟩_{S} ∣ A_{1} ⟩_{A} .

This is a unique final state. Given the initial state, apparatus ready state, and interaction Hamiltonian, there is no ambiguity about which joint quantum state unitary evolution produces.

But it is an entangled superposition of distinct record correlations, not one of the two record states.

A system initially in an up-down superposition undergoes a unitary interaction with a ready measuring device. The deterministic output is an entangled state correlating “up” with an UP pointer record and “down” with a DOWN pointer record, rather than a uniquely selected pointer value.

Why the entangled state is not one unique record

To state the issue sharply, represent the apparatus pointer observable by an operator $Z_{A}$ satisfying

Z_{A} ∣ A_{0} ⟩_{A} = z_{0} ∣ A_{0} ⟩_{A},

Z_{A} ∣ A_{1} ⟩_{A} = z_{1} ∣ A_{1} ⟩_{A}, z_{0} \neq = z_{1} .

For the joint system, the relevant pointer operator is $I_{S} \otimes Z_{A}$ . Acting on the post-interaction state gives

(I_{S} \otimes Z_{A}) ∣ Ψ_{pre} ⟩ = α z_{0} ∣0 ⟩_{S} ∣ A_{0} ⟩_{A} + β z_{1} ∣1 ⟩_{S} ∣ A_{1} ⟩_{A} .

When both coefficients are nonzero and $z_{0} \neq = z_{1}$ , this is not proportional to $∣ Ψ_{pre} ⟩$ . Therefore, $∣ Ψ_{pre} ⟩$ is not an eigenstate of the pointer observable.

Under the usual eigenstate-eigenvalue criterion for a definite record, the unitary final state does not assign the pointer one unique value. It assigns a coherent superposition of correlations:

system outcome $0$ correlated with apparatus record $0$ ;
system outcome $1$ correlated with apparatus record $1$ .

Notice what the problem is not:

It is not that the apparatus has failed to correlate with the system.
It is not that unitary evolution gives multiple possible final quantum states.
It is not ordinary uncertainty about which dynamics occurred.

The unitary dynamics does exactly what it should: it creates a reliable correlation. The unresolved question is why an individual experiment appears to yield one definite record rather than merely the entangled state predicted by the universal unitary model.

A schematic version of this argument is worth reading in the Stanford Encyclopedia of Philosophy discussion of the measurement problem.

Philosophical Issues in Quantum Theory

This Stanford Encyclopedia of Philosophy section gives a compact formulation of the measurement problem: an apparatus that works correctly for eigenstate inputs, combined with linear Schrödinger evolution, produces entanglement for superposed inputs rather than a single pointer eigenstate.

In Section 4.1, “The measurement problem formulated,” begin with the sentence on universal application of quantum theory. Then follow the schematic experiment involving system states \ket{0}_S, \ket{1}_S, the ready apparatus state, and the two apparatus records. Pay particular attention to the transition from calibrated eigenstate inputs to a superposed input: linearity forces the latter to evolve into an entangled state, not one definite instrument reading.

Born probabilities do not select an actual result

For the state

∣ ψ ⟩_{S} = α ∣0 ⟩_{S} + β ∣1 ⟩_{S},

the textbook Born rule gives

p (0) = ∣ α ∣^{2}, p (1) = ∣ β ∣^{2} .

These probabilities are indispensable for predicting frequencies across many similarly prepared trials. But a probability assignment is not, by itself, a dynamical account of which outcome occurs on one particular trial.

In standard textbook measurement rules, one adds a selective state-update rule. Conditional on observing outcome $i$ , the state is updated to the corresponding branch. In projector notation, with

Π_{i} = I_{S} \otimes ∣ A_{i} ⟩_{A} ⟨ A_{i} ∣,

the conditional update is

∣ Ψ_{pre} ⟩ ⟼ \frac{Π _{i} ∣ Ψ _{pre} ⟩}{⟨ Ψ _{pre} ∣ Π _{i} ∣ Ψ _{pre} ⟩},

with probability

p (i) = ⟨ Ψ_{pre} ∣ Π_{i} ∣ Ψ_{pre} ⟩ .

For example, if outcome $0$ is registered, then

∣ Ψ_{pre} ⟩ ⟼ ∣0 ⟩_{S} ∣ A_{0} ⟩_{A} .

The point is not that this rule is empirically unsuccessful. It is extraordinarily successful. The point is that it is not the same kind of evolution as the preceding unitary measurement interaction:

Unitary premeasurement	Selective outcome update
Maps one initial total state to one final total state	Selects one result-conditioned state
Linear	Not a linear map on state vectors
Reversible in principle	Not reversible as a stand-alone update
Produces entangled correlations	Produces a definite recorded result
Deterministic for the total state	Stochastic in the textbook formulation

So the often-heard sentence “quantum evolution is deterministic” is true only when its subject is carefully specified: the quantum state of a closed system evolves deterministically under unitary dynamics. It does not follow that the state alone fixes a unique outcome in the standard single-outcome measurement account.

The exact logical pressure point

The issue can be summarized as a tension among three claims:

Completeness: the quantum state gives the complete physical description.
Universal unitarity: every physical interaction, including measurement, always evolves according to the linear Schrödinger equation.
Single outcomes: each measurement has exactly one actual, definite outcome.

The ideal measurement calculation shows why accepting the first two yields the entangled state

α ∣0 ⟩_{S} ∣ A_{0} ⟩_{A} + β ∣1 ⟩_{S} ∣ A_{1} ⟩_{A},

rather than one unique record. Accepting all three therefore requires further interpretive work; simply repeating that $U$ is deterministic does not supply it.

This point should be handled with care. The equation above does not straightforwardly mean that an observer literally experiences a fuzzy or half-formed record. Its terms contain robust correlations: in the first term, the system and apparatus agree on $0$ ; in the second, they agree on $1$ . The question is instead how this formal structure is related to the single definite outcome reported in an experiment.

Different approaches respond by rejecting or reinterpreting different elements of the triad:

Textbook collapse formulations retain single outcomes by adding a stochastic outcome update.
Hidden-variable approaches retain deterministic dynamics by supplementing the quantum state with further physical variables.
Everettian approaches retain universal unitarity and treat the correlated terms as branches containing different relative records.
Operational or epistemic approaches may interpret the state update primarily as an update of an agent’s description after obtaining a record.

At this stage, these are not conclusions to endorse. They are signposts showing why a bare assertion of unitary determinism is insufficient to settle the question of quantum determinism more broadly.

A diagnostic to retain

Whenever someone argues either that quantum mechanics is deterministic or that it is indeterministic, identify three things:

What state is being evolved?
The microscopic system alone, system plus apparatus, or a larger closed system?
What law is being invoked?
Unitary Schrödinger evolution, a collapse rule, or some proposed extended dynamics?
What is claimed to be determined?
A later quantum state, an outcome distribution, or one actual outcome?

This diagnostic blocks two opposite mistakes:

“The Schrödinger equation is deterministic, therefore every measurement result was uniquely fixed.”
This overlooks the entangled superposition produced by an ideal measurement interaction.
“Measurement outcomes are probabilistic, therefore quantum theory has no deterministic dynamics.”
This overlooks the unique unitary evolution of the closed quantum state.

Key takeaways

An ideal measurement is a unitary interaction that correlates system eigenstates with distinct apparatus records.
For eigenstate inputs, calibration gives definite correlations:
$∣ i ⟩_{S} ∣ R ⟩_{A} ⟼ ∣ i ⟩_{S} ∣ A_{i} ⟩_{A} .$
For a superposed input, linearity requires
$(α ∣0 ⟩_{S} + β ∣1 ⟩_{S}) ∣ R ⟩_{A} ⟼ α ∣0 ⟩_{S} ∣ A_{0} ⟩_{A} + β ∣1 ⟩_{S} ∣ A_{1} ⟩_{A} .$
This final entangled state is uniquely determined by unitary dynamics, yet it is not generally a definite eigenstate of the apparatus pointer.
The Born rule supplies probabilities for outcomes, while the textbook collapse rule supplies a stochastic update conditional on an observed outcome. Neither follows merely from the claim that unitary state evolution is deterministic.
The resulting tension between completeness of the quantum state, universal unitary dynamics, and unique outcomes is the core motivation for the measurement problem.

Next, we will sharpen a distinction needed throughout the course: the difference between operational claims about the probabilities and records observed in experiments and ontological claims about what physical reality consists of between and during those observations.

Recommended resources