Mathematical physics

Mathematical foundations of quantum mechanics

The functional analysis and operator theory that make quantum mechanics mathematically rigorous.

IntuitionQuantum states as vectors, observables as instruments

Imagine every possible state of a quantum system — an electron's spin, the position of a particle, the polarization of a photon — as an arrow (a vector) in some possibly infinite-dimensional space, and every physical quantity you could measure — position, momentum, energy — not as a plain number attached to the arrow, but as an instrument: a linear transformation that acts on the arrow and, when you measure, forces it to jump to one of the instrument's own special directions and reads out the corresponding number. This single picture — vectors, and instruments acting on them — is the entire mathematical skeleton of quantum mechanics; uncertainty, quantization and interference are all just what happens when you take that picture seriously and follow the mathematics wherever it leads.

A wave packet built by superposing many momentum eigenstates

e^{ikx}

; raising the harmonic count

n

narrows the packet in position but spreads it out in momentum — the Fourier-analytic face of the uncertainty principle.

SchoolFrom wavefunctions to Hilbert space

Definition: Hilbert space and quantum states

A Hilbert space $\mathcal H$ is a complex vector space equipped with an inner product $\langle\cdot,\cdot\rangle$ that is complete (every Cauchy sequence converges) with respect to the norm $\|\psi\|=\sqrt{\langle\psi,\psi\rangle}$ . For a single particle moving in $d$ dimensions, the relevant space is $\mathcal{H} = L^2(\mathbb{R}^d)$ , the square-integrable complex functions $\psi(x)$ with $\int_{\mathbb R^d}|\psi(x)|^2\,dx<\infty$ . A pure quantum state is a unit vector $\psi\in\mathcal H$ (strictly, a ray $\{e^{i\theta}\psi\}$ , since an overall phase is unobservable), and $|\psi(x)|^2$ is interpreted as a probability density for finding the particle at $x$ .

\langle \psi,\varphi\rangle = \int_{\mathbb{R}^d} \overline{\psi(x)}\,\varphi(x)\,dx, \qquad \|\psi\|^2=\langle\psi,\psi\rangle=\int_{\mathbb{R}^d}|\psi(x)|^2\,dx = 1

A physical observable — position, momentum, energy, spin — is represented not by a number but by a self-adjoint (Hermitian) operator $\hat A$ on $\mathcal H$ , meaning $\hat{A} = \hat{A}^*$ : for all $\psi,\varphi$ in its domain, $\langle\psi,\hat A\varphi\rangle=\langle\hat A\psi,\varphi\rangle$ . Self-adjointness is exactly the algebraic condition that forces the possible measurement outcomes (the spectrum of $\hat A$ ) to be real numbers, and that guarantees eigenvectors for different eigenvalues are orthogonal — the mathematical reason a measurement of a real physical quantity always returns a real number.

i\hbar\,\partial_t \psi(x,t) = \hat H\psi(x,t), \qquad \hat H = -\frac{\hbar^2}{2m}\nabla^2 + V(x), \qquad \hat U(t)=e^{-i t\hat H/\hbar}

Stone's theorem says these two pictures are the same fact seen two ways: a strongly continuous one-parameter group of unitary operators $\hat U(t)$ (time evolution that preserves total probability, $\hat U(t)^*\hat U(t)=I$ ) is always of the form $\hat{U}(t) = e^{-i t \hat{H}/\hbar}$ for a unique self-adjoint operator $\hat H$ , the Hamiltonian — and conversely every self-adjoint $\hat H$ generates such a group. Differentiating $\hat U(t)\psi$ at $t=0$ recovers the Schrödinger equation $i\hbar \partial_t \psi = \hat{H}\psi$ : the two are logically equivalent formulations of the same dynamics.

Classical vs. quantum description of a physical system
Concept	Classical mechanics	Quantum mechanics
State	A point $(x,p)$ in phase space	A unit vector $\psi\in$ $\mathcal{H} = L^2(\mathbb{R}^d)$
Observable	A function $f(x,p)$ on phase space	A self-adjoint operator $\hat{A} = \hat{A}^*$
Measurement outcome	The value $f(x,p)$ , exactly	An eigenvalue of $\hat A$ , with probability $\|\langle e\|\psi\rangle\|^2$
Time evolution	Hamilton's equations $\dot q=\partial H/\partial p$	Unitary flow $\hat{U}(t) = e^{-i t \hat{H}/\hbar}$

UndergraduateTheorems that make the picture rigorous

Theorem: Robertson–Heisenberg uncertainty principle

For any two self-adjoint operators $\hat A=\hat A^*$ , $\hat B=\hat B^*$ on $\mathcal H$ and any unit vector $\psi$ in the domain of $\hat A\hat B$ and $\hat B\hat A$ , the standard deviations $\sigma_A=\sqrt{\langle\psi|(\hat A-\langle\hat A\rangle)^2|\psi\rangle}$ and $\sigma_B$ (expectations taken in state $\psi$ ) satisfy $\sigma_A \sigma_B \ge \tfrac{1}{2}\left|\langle [\hat{A},\hat{B}]\rangle\right|$ , where $[\hat A,\hat B]=\hat A\hat B-\hat B\hat A$ .

Why is it true?

Two observables can only be measured with simultaneous perfect precision if their operators commute; the size of the commutator is a direct measure of how incompatible they are. Cauchy–Schwarz turns "these two vectors cannot both be short" into a hard numerical bound, and separating the inner product into its real and imaginary parts is exactly what isolates the commutator (the genuinely quantum part) from the anticommutator (a classical-looking correlation term).

Proof

Step 1 (centered operators). Let $\Delta\hat A=\hat A-\langle\hat A\rangle I$ and $\Delta\hat B=\hat B-\langle\hat B\rangle I$ , both still self-adjoint since $\langle\hat A\rangle,\langle\hat B\rangle$ are real scalars. By definition $\sigma_A^2=\langle\psi|\Delta\hat A^2|\psi\rangle=\|\Delta\hat A\,\psi\|^2$ and likewise $\sigma_B^2=\|\Delta\hat B\,\psi\|^2$ .

Step 3 (split into real and imaginary parts). Write $\langle\Delta\hat A\,\psi|\Delta\hat B\,\psi\rangle=\langle\psi|\Delta\hat A\Delta\hat B|\psi\rangle$ . Since $\Delta\hat A,\Delta\hat B$ are self-adjoint, complex-conjugating flips the operator order: $\langle\Delta\hat A\Delta\hat B\rangle^*=\langle\Delta\hat B\Delta\hat A\rangle$ . Hence the anticommutator expectation $\langle\{\Delta\hat A,\Delta\hat B\}\rangle=\langle\Delta\hat A\Delta\hat B\rangle+\langle\Delta\hat B\Delta\hat A\rangle=2\,\mathrm{Re}\langle\Delta\hat A\Delta\hat B\rangle$ is real, while the commutator expectation $\langle[\Delta\hat A,\Delta\hat B]\rangle=\langle\Delta\hat A\Delta\hat B\rangle-\langle\Delta\hat B\Delta\hat A\rangle=2i\,\mathrm{Im}\langle\Delta\hat A\Delta\hat B\rangle$ is purely imaginary. Also $[\Delta\hat A,\Delta\hat B]=[\hat A,\hat B]$ , since the constant shifts cancel in the commutator.

Step 4 (recombine). By Pythagoras applied to real and imaginary parts, $|\langle\Delta\hat A\Delta\hat B\rangle|^2=\big(\mathrm{Re}\langle\Delta\hat A\Delta\hat B\rangle\big)^2+\big(\mathrm{Im}\langle\Delta\hat A\Delta\hat B\rangle\big)^2=\tfrac14|\langle\{\Delta\hat A,\Delta\hat B\}\rangle|^2+\tfrac14|\langle[\hat A,\hat B]\rangle|^2 \ge \tfrac14|\langle[\hat A,\hat B]\rangle|^2,$ dropping the manifestly nonnegative anticommutator term. Combining with Step 2, $\sigma_A^2\sigma_B^2\ge\tfrac14|\langle[\hat A,\hat B]\rangle|^2$ , and taking square roots (both sides nonnegative) gives $\sigma_A \sigma_B \ge \tfrac{1}{2}\left|\langle [\hat{A},\hat{B}]\rangle\right|$ .

Theorem: Discrete spectrum of the quantum harmonic oscillator

The Hamiltonian $\hat H=\dfrac{\hat P^2}{2m}+\dfrac12 m\omega^2\hat X^2$ , with $[\hat X,\hat P]=i\hbar$ , has discrete, nondegenerate spectrum $E_n = \hbar\omega\left(n+\tfrac12\right)$ , $n=0,1,2,\dots$ .

Why is it true?

Rather than solving a differential equation directly, factor the Hamiltonian algebraically into 'raising' and 'lowering' operators, exactly as one factors a quadratic form. Positivity of the resulting number operator, plus the algebra of how the ladder operators shift its eigenvalues by one unit, forces the eigenvalues to form an evenly spaced ladder starting at a nonnegative bottom rung — energy quantization falls out of algebra alone.

Proof

Step 1 (ladder operators). Define $\hat a=\sqrt{\tfrac{m\omega}{2\hbar}}\left(\hat X+\tfrac{i}{m\omega}\hat P\right)$ and $\hat a^\dagger=\sqrt{\tfrac{m\omega}{2\hbar}}\left(\hat X-\tfrac{i}{m\omega}\hat P\right)$ . Using $[\hat X,\hat P]=i\hbar$ , a direct computation gives $[\hat a,\hat a^\dagger]=\tfrac{m\omega}{2\hbar}\left(\tfrac{-i}{m\omega}[\hat X,\hat P]+\tfrac{i}{m\omega}[\hat P,\hat X]\right)=\tfrac{1}{2\hbar}\left(\hbar+\hbar\right)=1$ .

Step 2 (rewrite the Hamiltonian). Inverting, $\hat X=\sqrt{\tfrac{\hbar}{2m\omega}}(\hat a+\hat a^\dagger)$ and $\hat P=i\sqrt{\tfrac{m\omega\hbar}{2}}(\hat a^\dagger-\hat a)$ ; substituting into $\hat H=\tfrac{\hat P^2}{2m}+\tfrac12 m\omega^2\hat X^2$ and using $[\hat a,\hat a^\dagger]=1$ to reorder terms gives $\hat H=\hbar\omega\left(\hat a^\dagger\hat a+\tfrac12\right)$ . Define the number operator $\hat N=\hat a^\dagger\hat a$ ; it is self-adjoint, and for any state $\phi$ , $\langle\phi|\hat N|\phi\rangle=\|\hat a\phi\|^2\ge0$ , so every eigenvalue $n$ of $\hat N$ satisfies $n\ge0$ .

Step 3 (ladder relations). From $[\hat a,\hat a^\dagger]=1$ one gets $[\hat N,\hat a]=-\hat a$ and $[\hat N,\hat a^\dagger]=\hat a^\dagger$ . So if $\hat N|n\rangle=n|n\rangle$ , then $\hat N\hat a=\hat a\hat N-\hat a=\hat a(\hat N-1)$ , giving $\hat N(\hat a|n\rangle)=(n-1)(\hat a|n\rangle)$ : $\hat a$ lowers the eigenvalue by $1$ (or annihilates the state), and $\hat a^\dagger$ raises it by $1$ .

Step 4 (termination and the spectrum). Because $\hat N\ge0$ , repeatedly applying $\hat a$ to any eigenstate cannot produce eigenvalues below $0$ ; the descending chain $n,n-1,n-2,\dots$ must terminate at $n=0$ (a non-integer starting value would force a negative eigenvalue after finitely many steps, a contradiction), so every eigenvalue of $\hat N$ is a nonnegative integer, and the ground state $|0\rangle$ satisfies $\hat a|0\rangle=0$ . Applying $\hat a^\dagger$ repeatedly to $|0\rangle$ produces normalizable eigenstates $|n\rangle\propto(\hat a^\dagger)^n|0\rangle$ for every $n=0,1,2,\dots$ , and no others exist. Substituting $\hat N|n\rangle=n|n\rangle$ into $\hat H=\hbar\omega(\hat N+\tfrac12)$ gives $E_n = \hbar\omega\left(n+\tfrac12\right)$ .

UndergraduateReal-World Applications and Worked Examples

These are not textbook abstractions. Quantum tunneling — a direct consequence of the Schrödinger equation being a wave equation, not a particle-trajectory equation — is what lets flash-memory cells store a bit by pushing electrons through an insulating barrier, and what lets a scanning tunneling microscope (STM) image individual atoms by measuring a tunneling current. Discrete energy levels of the kind guaranteed by the spectral theorem are what an MRI machine reads out as a resonance frequency, and what a superconducting qubit uses as its two lowest energy levels $|0\rangle,|1\rangle$ to store quantum information; the uncertainty principle sets the ultimate noise floor for how precisely such a qubit's state can be measured and controlled.

Example: Tunneling current in a scanning tunneling microscope

An electron approaches a rectangular potential barrier of height $V_0=4.0\,\text{eV}$ (the vacuum gap between an STM tip and a metal surface) and width $L=0.50\,\text{nm}$ , with kinetic energy $E=3.0\,\text{eV}<V_0$ . Using the transmission coefficient $T\approx 16\dfrac{E}{V_0}\left(1-\dfrac{E}{V_0}\right)e^{-2\kappa L}$ with $\kappa=\dfrac{\sqrt{2m(V_0-E)}}{\hbar}$ , estimate $T$ , and explain why moving the tip $0.1\,\text{nm}$ closer changes the tunneling current so dramatically.

Solution

Step 1: compute $\kappa$ . With $V_0-E=1.0\,\text{eV}=1.6\times10^{-19}\,\text{J}$ , $m=9.11\times10^{-31}\,\text{kg}$ , $\hbar=1.055\times10^{-34}\,\text{J·s}$ : $\kappa=\sqrt{2m(V_0-E)}/\hbar\approx 5.1\times10^{9}\,\text{m}^{-1}$ , i.e. about $5.1\,\text{nm}^{-1}$ .

Step 2: compute the exponent. $2\kappa L\approx 2\times5.1\,\text{nm}^{-1}\times0.50\,\text{nm}\approx5.1$ , so $e^{-2\kappa L}\approx e^{-5.1}\approx6.1\times10^{-3}$ .

Step 3: assemble $T$ . The prefactor $16\frac{E}{V_0}(1-\frac{E}{V_0})=16\times0.75\times0.25=3.0$ , so $T\approx3.0\times6.1\times10^{-3}\approx1.8\times10^{-2}$ .

Step 4: sensitivity to distance. Because $T\propto e^{-2\kappa L}$ , decreasing $L$ by $0.1\,\text{nm}$ multiplies $T$ by $e^{2\kappa\times0.1\,\text{nm}}=e^{1.02}\approx2.8$ — a roughly threefold jump in tunneling current from a sub-atomic change in distance. This exponential sensitivity is what lets an STM resolve individual atoms.

Example: Minimum uncertainty of a spin-1/2 qubit

A superconducting qubit is prepared in the state $|0\rangle$ (spin up along $z$ ), an eigenstate of $\hat\sigma_z$ with eigenvalue $+1$ . Using the Pauli operators $\hat\sigma_x,\hat\sigma_y,\hat\sigma_z$ (each self-adjoint, each with eigenvalues $\pm1$ , and $[\hat\sigma_x,\hat\sigma_y]=2i\hat\sigma_z$ ), compute $\sigma_{\sigma_x}$ and $\sigma_{\sigma_y}$ in this state and check the Robertson–Heisenberg bound for the pair $(\hat\sigma_x,\hat\sigma_y)$ .

Solution

Step 2: variances. Because $\hat\sigma_x^2=\hat\sigma_y^2=I$ , $\sigma_{\sigma_x}^2=\langle0|\hat\sigma_x^2|0\rangle-\langle0|\hat\sigma_x|0\rangle^2=1-0=1$ , and identically $\sigma_{\sigma_y}^2=1$ ; so $\sigma_{\sigma_x}=\sigma_{\sigma_y}=1$ and $\sigma_{\sigma_x}\sigma_{\sigma_y}=1$ .

Step 4: compare. The product $1$ equals the bound $1$ : $|0\rangle$ is a minimum-uncertainty state for $(\hat\sigma_x,\hat\sigma_y)$ , meaning a measurement of $\hat\sigma_x$ or $\hat\sigma_y$ on a qubit prepared along $z$ is maximally random (50/50).

Research frontier as of 2026

The Hilbert-space picture is the accepted foundation for nonrelativistic quantum mechanics, but several directions remain open. Algebraic quantum field theory (Haag–Kastler axioms) tries to extend the self-adjoint-operator framework to relativistic quantum fields rigorously; a full, mathematically rigorous construction of an interacting 4-dimensional quantum Yang–Mills theory with a mass gap remains unsolved (it is one of the Clay Millennium Problems). In quantum information, current research explores how the Hilbert-space formalism applies to designing and verifying quantum error-correcting codes on noisy processors, and to classifying topological phases of matter (anyons, symmetry-protected phases), where operator algebras meet category theory. These are active research programs as of 2026.

What is the commutator $[\hat X,\hat P] = i\hbar$ equal to?

By Stone's theorem, for $\hat{U}(t) = e^{-i t \hat{H}/\hbar}$ to be unitary for every $t$ , what must $\hat H$ be?

A quantum harmonic oscillator has $\hbar\omega=2\,\text{eV}$ . What is the energy of its $n=1$ level?

In a scanning tunneling microscope, the extreme sensitivity of the tunneling current to tip-sample distance comes from which feature of $T\approx 16\frac{E}{V_0}(1-\frac{E}{V_0})e^{-2\kappa L}$ ?

References

John von Neumann (1955). Mathematical Foundations of Quantum Mechanics
Howard P. Robertson (1929). The Uncertainty Principle · DOI:10.1103/PhysRev.34.163
Marshall H. Stone (1932). On One-Parameter Unitary Groups in Hilbert Space · DOI:10.2307/1968538