Position Operator

"Where is the particle?" was the first question this track asked of the wavefunction, and Born Statistical Interpretation answered it statistically: $|\Psi(x)|^2$ is the probability density for position. This article asks the same question in a new grammar — the grammar the entire rest of quantum mechanics is written in. Position becomes an operator: a machine $\hat x$ that acts on wavefunctions, and whose action is almost comically simple. It acts by multiplication: $\hat{x}\Psi(x)=x\,\Psi(x)$ . That is the complete definition of the position operator in the position representation. Why dignify multiplication with a hat and an article? Because this simplest of all observables is the cleanest place to install the operator ideas that every observable will use: how a mean value becomes an operator sandwich $\int\Psi^*\hat x\,\Psi\,dx$ ; what it means for an operator to be Hermitian and why that guarantees real measured values; and what happens when you ask the standard question "what are this operator's eigenfunctions?" — because for $\hat x$ the answer is gloriously pathological. The states of perfectly definite position turn out to be infinitely thin delta spikes: not normalizable, not physical, yet indispensable as limits. Position is your first meeting with a continuous spectrum, and it teaches the right reflexes cheaply. There is also a genuinely useful moral hiding in the simplicity: in the position language, all the quantum strangeness lives in the state, none in the operator. The operator itself is simple; the subtlety lies in interpreting the quantum state correctly. The partner operator, where the derivative does the work, is next — Momentum Operator — and the pair together will generate the uncertainty structure of the whole theory.

Key takeaways

In the position representation, the position operator is multiplication: $\hat x\,\Psi(x) = x\,\Psi(x)$ ; in 3D, $\hat{\mathbf r}\,\Psi = \mathbf r\,\Psi$ .
Its mean is the operator sandwich $\langle x\rangle = \int\Psi^*\,x\,\Psi\,dx = \int x\,|\Psi|^2dx$ — the Born-rule average, rewritten.
$\hat x$ is Hermitian because $x$ is real — the template credential every observable must present.
$\langle x\rangle$ is the expectation value of repeated measurements on identically prepared systems; it need not coincide with either the most probable position or any position obtained in a single measurement.
The eigenvalue problem $\hat x\,g = x_0\,g$ forces $g(x) = \delta(x - x_0)$ : perfectly located, zero-width spikes — a continuous spectrum of non-normalizable idealizations.
Real position measurements have finite resolution: a detector with pixels reads out coarse-grained probabilities $\int_{\text{pixel}}|\Psi|^2dx$ .
Key idea: position acts by multiplication — the operator is trivial, and all the physics lives in how the state is spread over $x$ and in the operator's refusal to commute with momentum.

The core insight

In the position representation the position operator is just multiplication by $x$ : it reads off the coordinate, while all the dynamics and quantum strangeness live in how the state is spread over $x$ and in its non-commutation with momentum.

Where this lives today

The pixel histogram of Worked Example 2 is essentially what modern position-sensitive detectors record experimentally. Quantum-gas microscopes image ultracold atoms in optical lattices with single-site resolution: each shot returns one position measurement per atom (which lattice site), and stacking thousands of shots builds $|\Psi|^2$ site by site — the coarse-grained position spectrum read at a pixel width of one lattice constant, about $500\ \mathrm{nm}$ . The 2020s generation of these machines watches correlated many-body states assemble in real space, one Born-rule sample at a time. At the opposite end of the precision scale, LIGO's interferometers are position measurements of mirror coordinates at the $10^{-19}\ \mathrm{m}$ level — so sharp that the apparatus collides with the operator structure of position itself: measuring $\hat x$ too well squeezes the mirror's momentum spread upward (the delta-spike trade-off of this article's second figure), which feeds back as future position noise. The "standard quantum limit" and the frequency-dependent squeezing installed to dodge it are engineering consequences of $[\hat x, \hat p] \neq 0$ — the commutator built in the next article.

The physical picture

How does the position operator convert the probability distribution into the expectation value of position? It acts as a weighting operator. At each position $x$ , the operator multiplies the wavefunction by the coordinate: $\hat{x}\Psi(x) = x\Psi(x)$ . Regions far from the origin are weighted in proportion to their distance, while regions near $x = 0$ contribute little. The resulting function $x\Psi(x)$ is not a new physical state but a bookkeeping object whose overlap with the original state, $\int \Psi^*(x)\,x\,\Psi(x)\,dx$ , is exactly the expectation value $\langle x\rangle = \int x|\Psi(x)|^2dx$ . The position operator therefore extracts the first spatial moment of the probability distribution by weighting each point according to its coordinate.

Figure

The position operator multiplies the wavefunction by the coordinate, weighting distant regions more strongly when computing ⟨x⟩

Now run the machine hunting for its eigenfunctions — waves it merely rescales. $x\,g(x) = x_0\,g(x)$ demands that $g$ vanish at every $x \neq x_0$ : the entire wave must be piled on one single point. The limit of ever-narrower spikes is the Dirac delta, $g(x) = \delta(x - x_0)$ — a "function" that is zero everywhere, infinite at one point, and integrates to 1. It is the caricature of perfect localization, and like all good caricatures it is not a portrait: no physical particle is ever in such a state. But as limits the spikes organize everything — every allowed $x_0$ on the line is an eigenvalue, which is exactly what "position has a continuous spectrum" means, and the statement "measure position, find the particle somewhere" becomes "the delta spikes are complete."

Figure

Sharpening toward the delta spike — and what a real detector does instead

Keep the two panels as poles of the same idea: the delta spike is what the mathematics of $\hat x$ wants for an eigenstate; the pixel histogram is what the laboratory delivers. Everything practical happens between them.

The operator and its credentials

The definition first, in the representation this chapter lives in:

\hat x\,\Psi(x,t) = x\,\Psi(x,t) \qquad (1)

where: $x$ — the coordinate itself, acting as a multiplier; the hat marks the role (an operator on states), not any complexity of action. In three dimensions each component multiplies: $\hat{\mathbf r}\Psi = \mathbf r\Psi$ . Note what kind of object equation (1) creates: $x\Psi$ is generally not a multiple of $\Psi$ — the machine reshapes the wave except in the delta-spike limit — and it may not even be normalizable if $\Psi$ has fat tails. The operator is simple; it is not tame. The first credential is that the operator reproduces the Born-rule mean. The average of many position measurements on identically prepared systems is, from Probability Density and Normalization,

\langle x\rangle = \int_{-\infty}^{\infty} x\,|\Psi(x,t)|^2\,dx = \int_{-\infty}^{\infty} \Psi^*\,\hat x\,\Psi\,dx \qquad (2)

where: the two expressions are identical because $\Psi^*(x)\,x\,\Psi(x)=x\,|\Psi(x)|^2$ — and that triviality is the point: the sandwich pattern (conjugate state, operator, state, integrate) is the universal grammar for every observable's mean, and position is the one case where you can watch it reduce to elementary probability with no sleight of hand. This operator-sandwich structure, with operators whose action is not mere multiplication, is the business of Expectation Values, two articles ahead. The second credential is Hermiticity — the property that will define "observable" in Operators and Observables. Here it is one line: for any two normalizable states,

\int f^*\,(\hat x\,g)\,dx = \int f^*x\,g\,dx = \int (x f)^*\,g\,dx = \int (\hat x f)^*\,g\,dx \qquad (3)

where: the middle equality holds because $x$ is real — $x^* = x$ — so it slides freely between the starred and unstarred slots. No integration by parts, no boundary terms: multiplication by a real-valued function is Hermitian on its natural domain. Contrast the momentum operator, whose Hermiticity is earned through integration by parts and a factor of $-i$ (the derivation card of Momentum Operator); position gets the credential for free. The payoff of Hermiticity — means guaranteed real, eigenvalues guaranteed real — is developed there and in Operators and Observables; here, note simply that a machine reporting positions had better output real numbers, and this one provably does.

Derivation AThe eigenvalue problem — perfect position means a delta spike

Every operator gets asked the standard question: which states do you merely rescale? For $\hat x$ the question is $\hat x\,g(x) = x_0\,g(x)$ with $x_0$ a candidate number, and the answer teaches three permanent lessons at once: the eigenfunctions are delta spikes, the spectrum is continuous, and the eigenstates are not physical states — the triple that defines how continuous observables work throughout quantum mechanics. The argument itself is two lines of pointwise logic; the care is all in reading the result.

Step 1. Write the condition pointwise. $x\,g(x) = x_0\,g(x)$ means $(x - x_0)\,g(x) = 0$ for every $x$ . Why: multiplication acts locally — the equation is not one condition but a continuum of them, one at each point.

Step 2. Solve pointwise. At any $x \neq x_0$ the factor $(x - x_0)$ is nonzero, so $g(x) = 0$ there. The equation $(x-x_0)g(x)=0$ implies that any ordinary function solution vanishes identically; a nontrivial solution therefore requires a distribution concentrated at $x=x_0$ . Why: an ordinary function that is zero except at one point integrates to zero and is useless; to carry any weight, the "function" must be a distribution — an infinitely tall, infinitely thin spike.

Step 3. Normalize as a distribution. The standard choice is the Dirac delta:

g_{x_0}(x) = \delta(x - x_0), \qquad \int_{-\infty}^{\infty}\delta(x - x_0)\,dx = 1 \qquad (4)

Why: the delta is defined by what it does under integrals — $\int f(x)\delta(x - x_0)dx = f(x_0)$ , the "sifting" property — and you have already computed with it as a potential shape in Delta-Function Potential Well; here it appears in its other costume, as an idealized state.

Step 4. Read off the spectrum. The construction works for every real $x_0$ : no boundary condition, no quantization mechanism, no discrete menu. The spectrum of $\hat x$ is the entire real line — continuous. Why: contrast the bound-state Hamiltonians of this chapter, where normalizability at infinity culled the menu to a ladder. Every real number $x_0$ labels a generalized eigenvalue of $\hat{x}$ ; the spectrum is therefore continuous.

Step 5. Disqualify the eigenstates as states — honestly. $\int|\delta(x - x_0)|^2dx$ diverges (products such as $\delta(x-x_0)^2$ are not well defined as ordinary distributions, so these eigenstates cannot be normalized in the usual Hilbert-space sense). Thus the position eigenstates are not normalizable elements of the Hilbert space; they are generalized eigenstates (distributions) belonging to the rigged Hilbert-space extension. The orthonormality that replaces the usual $\int\psi_m^*\psi_n = \delta_{mn}$ is delta normalization, $\int g_{x_0}^*(x)\,g_{x_0'}(x)\,dx = \delta(x_0 - x_0')$ — orthogonality with a distribution on the right. Why: this is the same disease, and the same cure, as the plane waves of Bound versus Scattering States — idealized basis elements, physical states as superpositions. The systematic calculus of delta-normalized bases is owned by the Formalism chapter's Continuous Spectra and Dirac Delta Normalization, and the mathematically clean home for the spikes (rigged Hilbert space) by its Rigged Hilbert Space; this article needs only the three lessons above.

Go deeper · Derivation B · 1 minMean versus mode — what the expectation value is not

Derivation BMean versus mode — what the expectation value is not

The single most common misreading of $\langle x\rangle$ is "the place the particle probably is." We build the counterexample culture early: a two-line proof that the mean can sit at a forbidden location, plus the general distinction between the mean (lever-arm average) and the mode (peak of the density). The key point is that expectation values describe ensemble averages rather than the outcome of a single measurement.

Step 1. Mean at a node. Take any odd-symmetric density with a hole at the centre — say the ISW first excited state $\psi_2$ , whose density $|\psi_2|^2 \propto \sin^2(2\pi x/L)$ vanishes at $x = L/2$ . By symmetry, $\langle x\rangle = L/2$ exactly. Why: the average of a symmetric distribution sits at the symmetry point regardless of whether any probability lives there — the mean is a balance point, not a location. A particle in $\psi_2$ is never found at $L/2$ , yet $L/2$ is its mean position.

Step 2. Mean versus mode in a lopsided state. For the triangular state of Worked Example 1, $\Psi \propto x$ on $[0, a]$ : the density $\propto x^2$ peaks at the wall $x = a$ (the mode), while the mean computes to $\langle x\rangle = 3a/4$ . Why: the long lever arms near the wall pull the average toward the peak, but the low- $x$ tail drags it back; mean and mode agree only for symmetric single-peaked densities.

Step 3. State the general reading. $\langle x\rangle$ is the number whose ensemble average the Born rule guarantees: measure position once each on $N$ identically prepared systems and the sample mean converges to $\langle x\rangle$ as $N$ grows. It is not the outcome of any single measurement, not the most likely outcome, and not "where the particle is between measurements." Why: the ensemble doctrine is the safe reading of every expectation value; its full development — including what fluctuates around the mean — is Expectation Values and Standard Deviation and Quantum Uncertainty, in that order.

Limiting cases

Case	Result	Meaning
eigenstate limit $\sigma \to 0$	$g \to \delta(x - x_0)$ , $\sigma_p \to \infty$	perfect localization accompanied by divergent momentum uncertainty — non-normalizable, a generalized (distributional) state
broad, symmetric, unimodal $\lvert\Psi\rvert^2$	$\langle x\rangle =$ mode	for symmetric single-peaked densities the expectation value coincides with the most probable position
symmetric density about $x_c$	$\langle x\rangle = x_c$	The expectation value equals the center of symmetry, even if the probability density vanishes there.
pixel width $w \rightarrow 0$	histogram $\rightarrow \lvert\Psi\rvert^2$	Measured histograms converge to the Born probability density as detector pixels become arbitrarily small.
3D position	$\hat{\mathbf r}=(\hat x,\hat y,\hat z)$ , $[\hat x,\hat y]=[\hat y,\hat z]=[\hat z,\hat x]=0$	The Cartesian position operators commute, so all three coordinates can be specified simultaneously and a joint position measurement exists.

Common trap“⟨x⟩ is where the particle actually is”

Three errors ride on that sentence, each fatal in its own way. First, $\langle x\rangle$ is an ensemble average: it emerges from many measurements on identically prepared copies, and no single measurement need ever return it — the ISW state $\psi_2$ has its mean at a point where the particle is never found. Second, the mean is not the most probable value: lopsided densities split the two (mode at the wall, mean at $3a/4$ , in this article's worked example), and confusing them corrupts every estimate you make about skewed distributions. Third — subtlest — the phrase presumes the particle has a position between measurements, which the statistical interpretation deliberately refuses to assert: $\hat x$ 's job is to organize measurement statistics, not to report a hidden coordinate. Say "the mean of position measurements is $\langle x\rangle$ " and all three errors die of precision.

Worked examples

1 · A lopsided wave: mean, mode, and the difference (confidence builder)

A particle is confined to $[0, a]$ with $a = 2.0\ \mathrm{nm}$ , in the triangular state $\Psi(x) = \sqrt{3/a^3}\;x$ (and $\Psi = 0$ outside). Verify normalization, then find the most probable position and the mean position.

Step 1. Normalization audit: $\int_0^a |\Psi|^2 dx = \dfrac{3}{a^3}\int_0^a x^2\,dx = \dfrac{3}{a^3}\cdot\dfrac{a^3}{3} = 1$ ✓ — the state arrives properly normalized (the craft of Probability Density and Normalization).

Step 2. Mode: the density $|\Psi|^2 = 3x^2/a^3$ rises monotonically, so the most probable position is the wall, $x_{\text{mode}} = a = \boxed{2.0\ \mathrm{nm}}$ .

Step 3. Mean, by the sandwich (2): $\langle x\rangle = \dfrac{3}{a^3}\displaystyle\int_0^a x\cdot x^2\,dx = \dfrac{3}{a^3}\cdot\dfrac{a^4}{4} = \dfrac{3a}{4} = \boxed{1.5\ \mathrm{nm}}$ .

Step 4. Read the gap: the expectation value and the mode differ by $0.5\,\mathrm{nm}$ , one quarter of the box. The expectation value is pulled below the peak because probability at smaller $x$ carries a shorter lever arm. Any statement about "the position" of this particle must specify whether it refers to the expectation value, the mode, or an individual measurement outcome; here the expectation value and the mode differ by $25\%$ .

2 · Reading the spectrum through pixels (exam level)

A quantum-dot electron sits in the ISW ground state $\psi_1 = \sqrt{2/L}\sin(\pi x/L)$ of a well of width $L$ . A position detector divides the well into four equal pixels. Predict the probability histogram — the odds that a single position measurement lands in each pixel.

Step 1. The pixel probability is the coarse-grained Born rule: $P_k = \int_{\text{pixel } k}|\psi_1|^2dx$ , with $|\psi_1|^2 = \frac{2}{L}\sin^2(\pi x/L)$ .

Step 2. Integrate over $[0, L/4]$ : using $\int\sin^2 u\,dx = \frac{x}{2} - \frac{L}{4\pi}\sin\frac{2\pi x}{L}$ , $P_1 = \left[\dfrac{x}{L} - \dfrac{1}{2\pi}\sin\dfrac{2\pi x}{L}\right]_0^{L/4} = \dfrac14 - \dfrac{1}{2\pi} = \boxed{0.0908}$ .

Step 3. By the symmetry of $|\psi_1|^2$ about $L/2$ : $P_4 = P_1 = 0.0908$ , and $P_2 = P_3 = \dfrac12 - P_1 = \dfrac14 + \dfrac{1}{2\pi} = \boxed{0.4092}$ . Audit: $2(0.0908) + 2(0.4092) = 1.0000$ ✓.

Step 4. Read the histogram: the two central pixels are about $4.5$ times more probable than the edge pixels, reflecting the underlying $\sin^2$ probability density. This is how position is measured in practice: a detector with pixel width $w=L/4$ partitions the continuous position spectrum into four finite bins. As the pixel width decreases, the measured histogram converges to the Born probability density $|\psi(x)|^2$ . No idealized delta-function eigenstate is required—the continuous spectrum is sampled through finite-resolution detector elements.

Origins · 1 minDavid Hilbert and how this was discovered

David Hilbert

1862–1943 · Königsberg and Göttingen

Quantum mechanics found its house already built, and Hilbert was the builder. Between 1904 and 1910, studying integral equations, he developed the theory of the space now bearing his name — infinite-dimensional vectors with an inner product — and coined the word spectrum for the set of values attached to an operator, a name he chose with no knowledge that physics would one day match it to actual atomic spectra. He reportedly savoured the coincidence later: "I developed my theory of infinitely many variables from purely mathematical interests, and even called it 'spectral analysis', without any presentiment that it would later find application to the actual spectrum of physics." The position operator is the theory's simplest tenant: multiplication by $x$ on the space $L^2$ of square-integrable functions — Hilbert's own example class — with the subtlety that its "eigenvectors" live just outside the space, a puzzle his student von Neumann and the rigged-space formalism would later make rigorous. When Heisenberg's matrices and Schrödinger's waves seemed to be rival theories in 1926, it was in Hilbert's Göttingen, with his assistants Nordheim and von Neumann, that both were recognized as one algebra represented on one space — the mathematical unification this chapter's operator grammar quietly assumes on every page.

History

1904

— Hilbert begins the theory of integral equations; "spectral analysis" of operators is born as pure mathematics.

1926

— Schrödinger's wave mechanics casts observables as operators on functions; position acts by multiplication.

1926

— Born's statistical interpretation makes ∫x|Ψ|²dx the measured mean of position.

1927

— Dirac's transformation theory introduces delta-normalized position eigenstates as working tools.

1932

— Von Neumann's spectral theorem puts continuous spectra — position first among them — on rigorous footing.

Going further

Texts. Griffiths, Introduction to Quantum Mechanics, 3rd ed., Ch. 1–3 · Shankar, Principles of Quantum Mechanics, 2nd ed., Ch. 1, 4, 7 · Sakurai & Napolitano, Modern Quantum Mechanics, 2nd ed., Ch. 1 · Cohen-Tannoudji, Diu & Laloë, Quantum Mechanics, Vol. I, Ch. II · Townsend, A Modern Approach to Quantum Mechanics, 2nd ed., Ch. 6. The key results in this article are: (1) $\langle\hat x\rangle = \int x|\psi|^2dx$ ; (2) the delta normalization $\langle x|x'\rangle = \delta(x - x')$ with eigenfunction $\delta(x - x_0)$ ; (3) the Hermiticity of multiplication by real $x$ ; (4) the worked numbers — triangular state: normalization, mode $a$ , mean $3a/4$ ( $a = 2\ \mathrm{nm}$ : $1.5\ \mathrm{nm}$ ); ISW ground-state pixel probabilities $\frac14 \mp \frac{1}{2\pi} = 0.0908 / 0.4092$ , summing to 1. The commutator $[\hat x, \hat p] = i\hbar$ and the Ehrenfest relation are treated in Momentum Operator and Ehrenfest's Theorem respectively and are cited, not derived, here.

Prerequisites. The Wavefunction · Born Statistical Interpretation · Probability Density and Normalization · Wave–Particle Duality. Leads to. Momentum Operator (the partner machine, where the derivative works) · Expectation Values (the sandwich doctrine in general) · Operators and Observables · the delta-normalization formalism of the Formalism chapter.

If x̂ is just multiplication, why does quantum position behave so differently from classical position?

Because the operator was never where the quantum content lived. Classical strangeness-freedom comes from positions being plain numbers that coexist with sharp momenta; quantum position differs in two structural ways. The state is a spread — $|\Psi|^2$ is irreducibly a distribution, not ignorance of a hidden number — and the partner refuses to cooperate: $\hat x$ and the momentum operator do not commute, so no state is sharp in both (the commutator is built in Momentum Operator). Multiplication is trivial; multiplication versus differentiation is the whole theory. This is also why the article insists all the physics is in the state: change representation (momentum space, later) and $\hat x$ becomes a derivative while $\hat p$ becomes multiplication — the operators trade costumes, the physics stays.

Is a delta spike ever actually prepared in the lab?

Never exactly, and the theory itself explains why the attempt self-destructs. A state of width $\sigma$ costs momentum spread $\sigma_p \gtrsim \hbar/2\sigma$ ; as $\sigma \to 0$ the kinetic energy $\langle p^2\rangle/2m$ diverges — infinitely sharp localization needs infinite energy. Real localizations stop far short: an atom imaged by a quantum-gas microscope is pinned to $\sim500\ \mathrm{nm}$ ; an electron microscope's focused probe reaches $\sim50\ \mathrm{pm}$ . Each is described by a narrow wavepacket — normalizable, finite energy — for which the delta spike is the convenient idealized limit, exactly as the plane wave idealizes a long monochromatic packet on the momentum side.

What does it mean to say the delta spikes are 'complete' if none of them is a state?

It means they work perfectly as an analysis basis: any physical state can be written as a superposition of spikes, $\Psi(x') = \int \Psi(x)\,\delta(x' - x)\,dx$ — which looks like a triviality (it is the sifting property read backwards) and is in fact the position-representation completeness relation. The expansion coefficients are the wavefunction's values: $\Psi(x)$ is literally "how much of the spike at $x$ " the state contains. That is why the position representation feels so natural — it is expansion over position eigenstates, done silently. The formal statement $\int|x\rangle\langle x|dx = \mathbb{1}$ , and the calculus for using it safely, is the Formalism chapter's Continuous Spectra and Dirac Delta Normalization.

Do x̂, ŷ, ẑ interfere with each other the way position and momentum do?

No — the three coordinate operators all commute with one another ( $x\cdot y\cdot\Psi = y\cdot x\cdot\Psi$ , multiplications in any order), which is why a single measurement can return all three coordinates at once and why "the particle was found at $\mathbf r$ " is a meaningful joint statement. The incompatibility that defines quantum mechanics is specifically between a coordinate and its own momentum: $\hat x$ with $\hat p_x$ , never $\hat x$ with $\hat p_y$ . The general theory of which observables can be jointly sharp — commuting versus non-commuting — is developed in the Formalism chapter (The Commutator and Compatible Observables).

Why do we bother with the operator at all, if ⟨x⟩ = ∫x|Ψ|²dx already follows from the Born rule?

For position alone, you would not need it — the density carries everything. The operator grammar pays off the moment you leave position: the momentum mean is not $\int p|\Psi(p...)|$ -anything computable from $|\Psi(x)|^2$ ; it requires the sandwich $\int\Psi^*(-i\hbar\partial_x)\Psi\,dx$ , an object with no density-only expression. Position is the training-wheels case where the sandwich visibly reduces to elementary probability, so you can trust the pattern when it stops reducing. There is also a structural payoff: writing all observables as operators exposes their algebra — commutators — which is where uncertainty relations and conservation laws live. One grammar, every observable.

A measurement found the particle at x₀. Is the wavefunction now δ(x − x₀)?

The textbook collapse rule would say so, and the delta's pathology is one clean way to see that rule needs refinement for continuous spectra. A physical detector has resolution $w$ : what the measurement certifies is "in the pixel around $x_0$ ," and the updated state is (idealizing) the old wavefunction truncated to that pixel and renormalized — a narrow packet, not a spike, with momentum spread $\sim\hbar/w$ , finite energy, and honest subsequent dynamics (it spreads fast). The infinitely sharp version is the $w \to 0$ limit that no apparatus reaches. Collapse itself — what exactly happens, and in what sense — is the Formalism chapter's Measurement and Wavefunction Collapse and the track's final chapter; the operational pixel version above is what laboratories actually use.

Does the particle have a position when nobody measures?

The statistical interpretation of this track declines to say yes, and experiments punish theories that insist too naively. If "has a position" means a pre-existing value that measurement merely reveals, you have a hidden-variable theory; such theories survive only by paying steep prices elsewhere (Bohmian mechanics keeps definite positions but makes them nonlocally choreographed; local hidden-variable theories are ruled out by the Bell-test experiments discussed in the Entanglement chapter). The safe operational stance: $\Psi$ encodes the statistics of possible position measurements, and between measurements the question "where is it?" has no theory-sanctioned answer. Uncomfortable, testable, and — so far — undefeated.