Chapter 6: The Formalism — Operators, Commutators, and the Generalized Uncertainty Principle

"If you want to learn about nature, to appreciate nature, it is necessary to understand the language that she speaks in." — Richard Feynman

6.1 The Mathematical Framework: Why Formalism Matters

We have spent the last four chapters doing quantum mechanics. We solved the infinite square well, tamed the harmonic oscillator, separated the hydrogen atom into tractable pieces, and computed expectation values for specific systems. Along the way, we used operators — $\hat{p} = -i\hbar \frac{d}{dx}$, $\hat{H} = -\frac{\hbar^2}{2m}\frac{d^2}{dx^2} + V(x)$ — without fully interrogating what these mathematical objects are, what makes them special, and why quantum mechanics demands them.

This chapter is where we step back and build the general theory.

Think of an analogy. A student of mechanics might first learn to solve specific problems: a ball rolling down an incline, a pendulum, a mass on a spring. But at some point, they encounter the Lagrangian and Hamiltonian formulations — frameworks powerful enough to unify all of classical mechanics. That transition feels abstract at first, but it multiplies their power enormously. We are at the analogous moment in quantum mechanics.

The formalism we develop here is not an optional layer of mathematical polish. It is the skeleton of quantum mechanics itself. Every result we have derived so far — the quantization of energy, the uncertainty principle, the orthogonality of stationary states — will emerge as consequences of a handful of postulates about operators and Hilbert spaces. And the formalism will predict new physics that we could never have guessed from solving differential equations alone.

What is a Hilbert space? We will formalize this concept fully in Chapter 8 when we introduce Dirac notation. For now, we need only the essentials. A Hilbert space $\mathcal{H}$ is a (possibly infinite-dimensional) vector space equipped with an inner product. In wave mechanics, our Hilbert space is the set of all square-integrable functions:

$$\mathcal{H} = L^2(-\infty, +\infty) = \left\{ \psi(x) \;\bigg|\; \int_{-\infty}^{+\infty} |\psi(x)|^2 \, dx < \infty \right\}$$

The inner product of two functions $f$ and $g$ is:

$$\langle f | g \rangle = \int_{-\infty}^{+\infty} f^*(x) \, g(x) \, dx$$

This inner product gives us everything we need: norms ($\langle f | f \rangle = \|f\|^2$), orthogonality ($\langle f | g \rangle = 0$), and completeness (every Cauchy sequence converges). If these words from linear algebra feel vaguely familiar, they should. Quantum mechanics is linear algebra — often in infinite dimensions.

🔗 Connection: In Chapter 8, we will rewrite everything in this chapter using Dirac notation ($|\psi\rangle$, $\langle\phi|$, $\langle\phi|\hat{A}|\psi\rangle$), which makes the linear algebra structure even more transparent. For now, we work in wave-function language, using integrals where Chapter 8 will use brackets.

The postulates we are heading toward. By the end of this chapter, you will understand — in precise mathematical terms — five claims that together constitute the measurement theory of quantum mechanics:

1. The state of a quantum system is described by a vector $|\psi\rangle$ in a Hilbert space.
2. Every observable corresponds to a Hermitian (self-adjoint) operator.
3. The only possible results of measuring an observable are the eigenvalues of its operator.
4. The probability of obtaining a particular eigenvalue is determined by the inner product of the state with the corresponding eigenstate.
5. After measurement, the state collapses to the eigenstate corresponding to the result.

Postulate 5 is the one that will haunt us for the rest of this book, and for good reason. But we must build the mathematical machinery before we can appreciate just how strange it is.

Why inner products matter. The inner product is the bridge between mathematics and measurement. When we write $\langle \psi_n | \psi \rangle$, we are computing the overlap between the state $\psi$ and the eigenstate $\psi_n$ — and the squared modulus $|\langle \psi_n | \psi \rangle|^2$ gives the probability of finding the system in state $\psi_n$ upon measurement. Without the inner product, quantum mechanics would be mathematics without prediction. With it, we have a complete probabilistic theory.

Completeness of eigenfunctions. For a Hermitian operator $\hat{A}$ with a discrete spectrum, the eigenfunctions $\{\psi_n\}$ form a complete set: any state $\psi$ in the Hilbert space can be expanded as

$$\psi = \sum_n c_n \psi_n, \qquad c_n = \langle \psi_n | \psi \rangle$$

This completeness relation — sometimes called the resolution of the identity — is one of the most useful results in quantum mechanics. It tells us that the eigenstates of any observable provide a "coordinate system" for the Hilbert space, analogous to the $\hat{x}$, $\hat{y}$, $\hat{z}$ unit vectors in ordinary 3D space. Every state has a unique expansion, and the coefficients $c_n$ contain all the physical information: $|c_n|^2$ is the probability of measuring the eigenvalue $a_n$.

For continuous spectra (like the position or momentum eigenstates), the sum becomes an integral and the Kronecker delta becomes a Dirac delta, but the essential idea is the same:

$$\psi(x) = \int_{-\infty}^{+\infty} c(k) \psi_k(x) \, dk$$

We will make this more precise in Chapters 8 and 9.

✅ Checkpoint: Before proceeding, make sure you can answer: What is the momentum operator in position space? What does it mean for two functions to be orthogonal in the $L^2$ inner product? If you are uncertain, revisit Sections 2.3 and 2.5 of Chapter 2.

6.2 Operators in Quantum Mechanics

What is an operator?

An operator $\hat{A}$ is a rule that takes one function and returns another:

$$\hat{A} : \psi(x) \mapsto \phi(x)$$

We write $\hat{A}\psi(x) = \phi(x)$. The hat notation ($\hat{A}$) distinguishes operators from ordinary numbers or functions.

Examples you already know:

Linear operators

An operator $\hat{A}$ is linear if, for any functions $\psi$ and $\phi$ and any complex constants $\alpha$ and $\beta$:

$$\hat{A}(\alpha\psi + \beta\phi) = \alpha\hat{A}\psi + \beta\hat{A}\phi$$

All the operators listed above are linear. The operator "take the absolute value" is not linear ($|a + b| \neq |a| + |b|$ in general). Quantum mechanics uses only linear operators — a restriction that reflects the linearity of the Schrodinger equation itself, and ultimately the superposition principle.

⚠️ Common Misconception: Students sometimes think operators "act on numbers." They do not. Operators act on functions (or, more generally, on vectors in a Hilbert space). Writing $\hat{p} \cdot 3$ is meaningless. Writing $\hat{p} \cdot (3e^{ikx})$ means $-i\hbar \frac{d}{dx}(3e^{ikx}) = 3\hbar k \, e^{ikx}$.

Operator algebra

Operators can be added and multiplied (composed):

Addition: $(\hat{A} + \hat{B})\psi = \hat{A}\psi + \hat{B}\psi$

Scalar multiplication: $(c\hat{A})\psi = c(\hat{A}\psi)$

Product (composition): $(\hat{A}\hat{B})\psi = \hat{A}(\hat{B}\psi)$

The product means: first apply $\hat{B}$, then apply $\hat{A}$ to the result. This is exactly like matrix multiplication — and it is not commutative in general:

$$\hat{A}\hat{B} \neq \hat{B}\hat{A} \quad \text{(in general)}$$

This seemingly innocuous mathematical fact — that the order of operators matters — has the most profound consequences in all of physics. It is the reason the uncertainty principle exists. We will make this precise in Section 6.4.

Powers and functions of operators. $\hat{A}^2 = \hat{A}\hat{A}$, and more generally $\hat{A}^n$ means "apply $\hat{A}$ $n$ times." For functions of operators defined by Taylor series:

$$e^{\hat{A}} = \hat{I} + \hat{A} + \frac{\hat{A}^2}{2!} + \frac{\hat{A}^3}{3!} + \cdots$$

This exponential will become central in Chapter 7 when we write the time-evolution operator as $\hat{U}(t) = e^{-i\hat{H}t/\hbar}$.

The identity operator

The identity operator $\hat{I}$ does nothing: $\hat{I}\psi = \psi$ for all $\psi$. It is the operator analogue of the number 1. We often omit it when the context is clear — writing $\hat{A} - a$ instead of $\hat{A} - a\hat{I}$ — but it is always there implicitly.

The projection operator

A concept we will use repeatedly is the projection operator onto a state $\psi_n$. In wave function language, it acts as:

$$\hat{P}_n \phi = \psi_n \langle \psi_n | \phi \rangle = \psi_n \int \psi_n^*(x) \phi(x) \, dx$$

It "projects" the state $\phi$ onto the direction $\psi_n$ — extracting the component of $\phi$ along $\psi_n$ and discarding everything else. Projection operators satisfy $\hat{P}_n^2 = \hat{P}_n$ (projecting twice is the same as projecting once) and $\hat{P}_n^\dagger = \hat{P}_n$ (they are Hermitian). They will play a starring role in the measurement postulates of Section 6.8.

The classical-to-quantum correspondence

How did physicists discover which operators correspond to which observables? The historical route was the canonical quantization prescription: take the classical expression for an observable (written in terms of position and momentum), and replace the classical variables with operators:

$$x \to \hat{x}, \qquad p \to \hat{p} = -i\hbar\frac{d}{dx}$$

For example, the classical kinetic energy $T = p^2/(2m)$ becomes $\hat{T} = \hat{p}^2/(2m) = -\frac{\hbar^2}{2m}\frac{d^2}{dx^2}$.

This recipe works beautifully for most observables. It breaks down when the classical expression involves products of $x$ and $p$ (because $\hat{x}\hat{p} \neq \hat{p}\hat{x}$, so the ordering matters), and it fails entirely for observables like spin that have no classical analogue. The general theory of operator construction is more subtle than a simple substitution rule, but canonical quantization remains the starting point for most applications.

Worked Example 6.1: Operator products and non-commutativity

Let us verify that $\hat{x}$ and $\hat{p}$ do not commute. Consider an arbitrary differentiable function $\psi(x)$.

Compute $\hat{x}\hat{p}\psi$:

$$\hat{x}\hat{p}\psi = \hat{x}\left(-i\hbar\frac{d\psi}{dx}\right) = -i\hbar \, x \frac{d\psi}{dx}$$

Compute $\hat{p}\hat{x}\psi$:

$$\hat{p}\hat{x}\psi = -i\hbar\frac{d}{dx}(x\psi) = -i\hbar\left(\psi + x\frac{d\psi}{dx}\right)$$

Difference:

$$(\hat{x}\hat{p} - \hat{p}\hat{x})\psi = -i\hbar \, x\frac{d\psi}{dx} + i\hbar\psi + i\hbar \, x\frac{d\psi}{dx} = i\hbar\psi$$

Since this holds for every $\psi$, we conclude:

$$\hat{x}\hat{p} - \hat{p}\hat{x} = i\hbar\hat{I}$$

where $\hat{I}$ is the identity operator. This is the canonical commutation relation, and we will study it exhaustively in Section 6.4.

🔄 Check Your Understanding: Verify that $[\hat{x}, \hat{x}^2] = 0$ by acting on an arbitrary function $\psi(x)$. Why is this result physically obvious?

Worked Example 6.1b: Verifying operator ordering matters for $\hat{x}\hat{p}$ vs $\hat{p}\hat{x}$

Students often wonder: which ordering should we choose when a classical expression involves both $x$ and $p$? Consider the classical quantity $xp$. There are two natural quantum versions: $\hat{x}\hat{p}$ and $\hat{p}\hat{x}$. Neither is Hermitian.

$(\hat{x}\hat{p})^\dagger = \hat{p}^\dagger\hat{x}^\dagger = \hat{p}\hat{x} \neq \hat{x}\hat{p}$

A Hermitian combination is the symmetrized product:

$$\frac{1}{2}(\hat{x}\hat{p} + \hat{p}\hat{x})$$

You can verify: $\left[\frac{1}{2}(\hat{x}\hat{p} + \hat{p}\hat{x})\right]^\dagger = \frac{1}{2}(\hat{p}\hat{x} + \hat{x}\hat{p})$, which is the same thing. This symmetrization (also known as Weyl ordering) is the standard prescription for resolving ordering ambiguities.

Using our result from Example 6.1, $\hat{x}\hat{p} = \hat{p}\hat{x} + i\hbar$, so:

$$\frac{1}{2}(\hat{x}\hat{p} + \hat{p}\hat{x}) = \hat{p}\hat{x} + \frac{i\hbar}{2} = \hat{x}\hat{p} - \frac{i\hbar}{2}$$

The two orderings differ by a constant — which is exactly the commutator $[\hat{x}, \hat{p}] = i\hbar$.

6.3 Hermitian Operators and Observables

The adjoint of an operator

The adjoint (or Hermitian conjugate) $\hat{A}^\dagger$ of an operator $\hat{A}$ is defined by the requirement:

$$\langle f | \hat{A} g \rangle = \langle \hat{A}^\dagger f | g \rangle \quad \text{for all } f, g \in \mathcal{H}$$

In integral notation:

$$\int_{-\infty}^{+\infty} f^*(x) \, (\hat{A} g)(x) \, dx = \int_{-\infty}^{+\infty} (\hat{A}^\dagger f)^*(x) \, g(x) \, dx$$

An operator is Hermitian (or self-adjoint) if $\hat{A}^\dagger = \hat{A}$, meaning:

$$\langle f | \hat{A} g \rangle = \langle \hat{A} f | g \rangle \quad \text{for all } f, g \in \mathcal{H}$$

🔴 Warning: Mathematicians distinguish between "Hermitian" (symmetric) and "self-adjoint" (symmetric with matching domains). For physically relevant operators on suitable domains, the distinction usually does not bite, and we will use the terms interchangeably. Readers pursuing mathematical physics should consult Reed and Simon, Methods of Modern Mathematical Physics, Vol. I.

Proof: Eigenvalues of Hermitian operators are real

This is one of the most important theorems in quantum mechanics.

Theorem 6.1. If $\hat{A}$ is Hermitian and $\hat{A}\psi = a\psi$ (with $\psi \neq 0$), then $a \in \mathbb{R}$.

Proof. Starting from the eigenvalue equation:

$$\langle \psi | \hat{A}\psi \rangle = \langle \psi | a\psi \rangle = a \langle \psi | \psi \rangle$$

Since $\hat{A}$ is Hermitian:

$$\langle \psi | \hat{A}\psi \rangle = \langle \hat{A}\psi | \psi \rangle = \langle a\psi | \psi \rangle = a^* \langle \psi | \psi \rangle$$

Setting these equal:

$$a \langle \psi | \psi \rangle = a^* \langle \psi | \psi \rangle$$

Since $\psi \neq 0$, we have $\langle \psi | \psi \rangle > 0$, so $a = a^*$, which means $a$ is real. $\blacksquare$

Why this matters physically: Measurement outcomes are real numbers. When you measure energy, you get 13.6 eV, not $(13.6 + 2i)$ eV. The fact that Hermitian operators have real eigenvalues is why quantum mechanics identifies observables with Hermitian operators. The math guarantees that the theory predicts sensible measurement results.

Proof: Eigenfunctions of different eigenvalues are orthogonal

Theorem 6.2. If $\hat{A}$ is Hermitian with $\hat{A}\psi_a = a\psi_a$ and $\hat{A}\psi_b = b\psi_b$, and $a \neq b$, then $\langle \psi_a | \psi_b \rangle = 0$.

Proof. Consider $\langle \psi_a | \hat{A}\psi_b \rangle$ and evaluate it two ways:

$$\langle \psi_a | \hat{A}\psi_b \rangle = b \langle \psi_a | \psi_b \rangle$$

Using the Hermitian property:

$$\langle \psi_a | \hat{A}\psi_b \rangle = \langle \hat{A}\psi_a | \psi_b \rangle = a^* \langle \psi_a | \psi_b \rangle = a \langle \psi_a | \psi_b \rangle$$

(where $a^* = a$ by Theorem 6.1). Setting these equal:

$$(a - b)\langle \psi_a | \psi_b \rangle = 0$$

Since $a \neq b$, we conclude $\langle \psi_a | \psi_b \rangle = 0$. $\blacksquare$

🔗 Connection: You have already seen this orthogonality in action. In Chapter 3, the infinite-well eigenfunctions $\psi_n(x) = \sqrt{2/L}\sin(n\pi x/L)$ are orthogonal — that is, $\int_0^L \psi_m^*(x)\psi_n(x) \, dx = \delta_{mn}$. Now you know why: they are eigenfunctions of the Hermitian Hamiltonian $\hat{H}$ with different eigenvalues $E_n = n^2\pi^2\hbar^2/(2mL^2)$.

Worked Example 6.2: Verifying that $\hat{p}$ is Hermitian

We need to show that $\langle f | \hat{p} g \rangle = \langle \hat{p} f | g \rangle$ for square-integrable functions that vanish at infinity.

$$\langle f | \hat{p} g \rangle = \int_{-\infty}^{+\infty} f^*(x)\left(-i\hbar\frac{dg}{dx}\right) dx$$

Integrate by parts, with $u = f^*$ and $dv = -i\hbar \frac{dg}{dx} dx$:

$$= \left[-i\hbar f^* g\right]_{-\infty}^{+\infty} + i\hbar \int_{-\infty}^{+\infty} \frac{df^*}{dx} g \, dx$$

The boundary term vanishes because $f, g \in L^2$ (they vanish at infinity). The remaining integral is:

$$= \int_{-\infty}^{+\infty} \left(i\hbar\frac{df^*}{dx}\right) g \, dx = \int_{-\infty}^{+\infty} \left(-i\hbar\frac{df}{dx}\right)^* g \, dx = \langle \hat{p} f | g \rangle$$

So $\hat{p}^\dagger = \hat{p}$: the momentum operator is Hermitian. $\checkmark$

The physical observable postulate

We can now state a central postulate of quantum mechanics:

Postulate (Observables): Every measurable physical quantity $A$ is associated with a Hermitian operator $\hat{A}$ acting on the Hilbert space. The possible results of measuring $A$ are the eigenvalues of $\hat{A}$.

This is not something we derive — it is an axiom, justified by its extraordinary empirical success. Every prediction of quantum mechanics, from atomic spectra to semiconductor band gaps to qubit gate fidelities, follows from this identification.

📊 By the Numbers: The magnetic moment of the electron, predicted by QED (which rests on this operator formalism), agrees with experiment to 12 decimal places: $g/2 = 1.001\,159\,652\,180\,73(28)$. This is the most precisely tested prediction in all of science.

Worked Example 6.3: Verifying that the position operator $\hat{x}$ is Hermitian

This is straightforward but worth doing explicitly:

$$\langle f | \hat{x} g \rangle = \int_{-\infty}^{+\infty} f^*(x) \cdot x \cdot g(x) \, dx$$

Since $x$ is a real number, $x^* = x$, so:

$$= \int_{-\infty}^{+\infty} (xf(x))^* g(x) \, dx = \langle \hat{x} f | g \rangle$$

Therefore $\hat{x}^\dagger = \hat{x}$: the position operator is Hermitian. $\checkmark$

Notice how the proof relied on the reality of $x$ — a complex-valued "position" would not give a Hermitian operator, which is physically sensible because position is a real quantity.

Non-Hermitian operators in quantum mechanics

Not all useful operators are Hermitian. The annihilation operator $\hat{a}$ from Chapter 4 is not Hermitian ($\hat{a}^\dagger = \hat{a}^\dagger \neq \hat{a}$), and neither is $\hat{p} + i\hat{x}$. Non-Hermitian operators do not correspond to observables, but they are indispensable computational tools. The ladder operators $\hat{a}$ and $\hat{a}^\dagger$ let us solve the QHO without ever solving a differential equation. Similarly, the time-evolution operator $\hat{U}(t) = e^{-i\hat{H}t/\hbar}$ is not Hermitian but unitary ($\hat{U}^\dagger\hat{U} = \hat{I}$), which guarantees probability conservation.

The distinction is: - Hermitian ($\hat{A}^\dagger = \hat{A}$): represents an observable. Real eigenvalues. - Unitary ($\hat{U}^\dagger\hat{U} = \hat{I}$): represents a transformation. Eigenvalues on the unit circle. - General: a computational tool. No guaranteed spectral properties.

✅ Checkpoint: State, in your own words, why quantum mechanics requires observables to be Hermitian operators rather than arbitrary operators. Your answer should reference two properties of Hermitian operators.

6.4 Commutators: The Heart of Quantum Mechanics

Definition and basic properties

The commutator of two operators $\hat{A}$ and $\hat{B}$ is:

$$[\hat{A}, \hat{B}] \equiv \hat{A}\hat{B} - \hat{B}\hat{A}$$

If $[\hat{A}, \hat{B}] = 0$, we say $\hat{A}$ and $\hat{B}$ commute. If $[\hat{A}, \hat{B}] \neq 0$, they are non-commuting.

We already proved the canonical commutation relation in Worked Example 6.1:

$$[\hat{x}, \hat{p}] = i\hbar$$

This single equation — five symbols plus a constant of nature — encodes the entire uncertainty principle and, in a real sense, the difference between classical and quantum physics.

💡 Key Insight: In classical mechanics, position and momentum are ordinary numbers. Numbers always commute: $xp = px$. The fact that quantum operators don't commute is not a mathematical inconvenience — it is the origin of quantum behavior. If $[\hat{x}, \hat{p}]$ were zero, there would be no uncertainty principle, no quantization, no atoms, no chemistry, no you.

Commutator algebra

Commutators obey a rich algebra. The following identities hold for any operators $\hat{A}$, $\hat{B}$, $\hat{C}$ and any scalar $\alpha$:

Identity 1 (Antisymmetry): $$[\hat{A}, \hat{B}] = -[\hat{B}, \hat{A}]$$

Identity 2 (Linearity): $$[\hat{A}, \hat{B} + \hat{C}] = [\hat{A}, \hat{B}] + [\hat{A}, \hat{C}]$$

Identity 3 (Scalar factor): $$[\hat{A}, \alpha\hat{B}] = \alpha[\hat{A}, \hat{B}]$$

Identity 4 (Product rule / Leibniz rule): $$[\hat{A}, \hat{B}\hat{C}] = [\hat{A}, \hat{B}]\hat{C} + \hat{B}[\hat{A}, \hat{C}]$$

Identity 5 (Jacobi identity): $$[\hat{A}, [\hat{B}, \hat{C}]] + [\hat{B}, [\hat{C}, \hat{A}]] + [\hat{C}, [\hat{A}, \hat{B}]] = 0$$

Identity 6 (Self-commutator): $$[\hat{A}, \hat{A}] = 0$$

Let us prove Identity 4 as an example, since it is the workhorse of commutator calculations.

Proof of Identity 4:

$$[\hat{A}, \hat{B}\hat{C}] = \hat{A}\hat{B}\hat{C} - \hat{B}\hat{C}\hat{A}$$

Now add and subtract $\hat{B}\hat{A}\hat{C}$:

$$= \hat{A}\hat{B}\hat{C} - \hat{B}\hat{A}\hat{C} + \hat{B}\hat{A}\hat{C} - \hat{B}\hat{C}\hat{A}$$

$$= (\hat{A}\hat{B} - \hat{B}\hat{A})\hat{C} + \hat{B}(\hat{A}\hat{C} - \hat{C}\hat{A})$$

$$= [\hat{A}, \hat{B}]\hat{C} + \hat{B}[\hat{A}, \hat{C}] \quad \blacksquare$$

Worked Example 6.3: A systematic commutator calculation

Problem: Calculate $[\hat{x}, \hat{p}^2]$.

Solution using Identity 4:

$$[\hat{x}, \hat{p}^2] = [\hat{x}, \hat{p}\hat{p}] = [\hat{x}, \hat{p}]\hat{p} + \hat{p}[\hat{x}, \hat{p}]$$

$$= (i\hbar)\hat{p} + \hat{p}(i\hbar) = 2i\hbar\hat{p}$$

This result has immediate physical relevance: since the kinetic energy is $\hat{T} = \hat{p}^2/(2m)$, we get $[\hat{x}, \hat{T}] = i\hbar\hat{p}/m$, which connects to the velocity operator.

Worked Example 6.4: Commutator of $\hat{x}$ and $\hat{H}$

Problem: For a particle in a potential $V(x)$, calculate $[\hat{x}, \hat{H}]$.

Solution:

$$[\hat{x}, \hat{H}] = \left[\hat{x}, \frac{\hat{p}^2}{2m} + V(\hat{x})\right] = \frac{1}{2m}[\hat{x}, \hat{p}^2] + [\hat{x}, V(\hat{x})]$$

From Example 6.3, $[\hat{x}, \hat{p}^2] = 2i\hbar\hat{p}$. And $[\hat{x}, V(\hat{x})] = 0$ because $\hat{x}$ commutes with any function of $\hat{x}$. Therefore:

$$[\hat{x}, \hat{H}] = \frac{i\hbar\hat{p}}{m}$$

This result, combined with Ehrenfest's theorem (Chapter 7), yields $\frac{d\langle x \rangle}{dt} = \frac{\langle p \rangle}{m}$ — the quantum version of $v = p/m$.

Worked Example 6.5: Commutator with the QHO ladder operators

🔗 Spaced Review from Chapter 4: Recall the creation and annihilation operators for the quantum harmonic oscillator:

$$\hat{a} = \sqrt{\frac{m\omega}{2\hbar}}\left(\hat{x} + \frac{i\hat{p}}{m\omega}\right), \qquad \hat{a}^\dagger = \sqrt{\frac{m\omega}{2\hbar}}\left(\hat{x} - \frac{i\hat{p}}{m\omega}\right)$$

Problem: Calculate $[\hat{a}, \hat{a}^\dagger]$.

Solution: We expand the product:

$$\hat{a}\hat{a}^\dagger = \frac{m\omega}{2\hbar}\left(\hat{x} + \frac{i\hat{p}}{m\omega}\right)\left(\hat{x} - \frac{i\hat{p}}{m\omega}\right)$$

$$= \frac{m\omega}{2\hbar}\left(\hat{x}^2 - \frac{i\hat{x}\hat{p}}{m\omega} + \frac{i\hat{p}\hat{x}}{m\omega} + \frac{\hat{p}^2}{m^2\omega^2}\right)$$

$$= \frac{m\omega}{2\hbar}\left(\hat{x}^2 + \frac{\hat{p}^2}{m^2\omega^2} + \frac{i}{m\omega}(\hat{p}\hat{x} - \hat{x}\hat{p})\right)$$

$$= \frac{m\omega}{2\hbar}\left(\hat{x}^2 + \frac{\hat{p}^2}{m^2\omega^2} - \frac{i}{m\omega}[\hat{x}, \hat{p}]\right)$$

$$= \frac{m\omega}{2\hbar}\left(\hat{x}^2 + \frac{\hat{p}^2}{m^2\omega^2} + \frac{\hbar}{m\omega}\right)$$

Similarly:

$$\hat{a}^\dagger\hat{a} = \frac{m\omega}{2\hbar}\left(\hat{x}^2 + \frac{\hat{p}^2}{m^2\omega^2} - \frac{\hbar}{m\omega}\right)$$

Subtracting:

$$[\hat{a}, \hat{a}^\dagger] = \hat{a}\hat{a}^\dagger - \hat{a}^\dagger\hat{a} = \frac{m\omega}{2\hbar} \cdot \frac{2\hbar}{m\omega} = 1$$

That is:

$$\boxed{[\hat{a}, \hat{a}^\dagger] = 1}$$

This deceptively simple result is the foundation of the entire algebraic approach to the harmonic oscillator, and later to quantum field theory (Chapter 34). It encodes the quantization of energy levels: $E_n = \hbar\omega(n + \frac{1}{2})$.

The physical meaning of commutators

Two observables commute if and only if they can be measured simultaneously with arbitrary precision. Here is why.

Suppose $[\hat{A}, \hat{B}] = 0$. Then, under mild conditions, $\hat{A}$ and $\hat{B}$ share a common set of eigenfunctions. That is, there exist states $\psi_{ab}$ such that:

$$\hat{A}\psi_{ab} = a\psi_{ab} \qquad \text{and} \qquad \hat{B}\psi_{ab} = b\psi_{ab}$$

If the system is in such a state, a measurement of $A$ gives $a$ with certainty, and a measurement of $B$ gives $b$ with certainty. Both observables have definite, simultaneous values.

If $[\hat{A}, \hat{B}] \neq 0$, no such states exist (in general). You cannot simultaneously know both observables with arbitrary precision. The commutator quantifies the unavoidable trade-off — this is the uncertainty principle, which we derive in the next section.

🔵 Historical Note: Heisenberg's original 1927 paper introduced the uncertainty principle using a thought experiment involving a gamma-ray microscope (see Case Study 1). The algebraic derivation from commutators came later, largely through the work of Robertson (1929) and Schrodinger (1930), and is both more general and more rigorous.

Worked Example 6.6: Commutator $[\hat{p}, V(\hat{x})]$

Problem: For a differentiable potential $V(x)$, compute $[\hat{p}, V(\hat{x})]$.

Solution: Act on a test function $\psi(x)$:

$$[\hat{p}, V(\hat{x})]\psi = \hat{p}(V\psi) - V(\hat{p}\psi)$$

$$= -i\hbar\frac{d}{dx}(V\psi) - V\left(-i\hbar\frac{d\psi}{dx}\right)$$

$$= -i\hbar\left(\frac{dV}{dx}\psi + V\frac{d\psi}{dx}\right) + i\hbar V\frac{d\psi}{dx}$$

$$= -i\hbar\frac{dV}{dx}\psi$$

Therefore:

$$[\hat{p}, V(\hat{x})] = -i\hbar\frac{dV}{dx}$$

Combined with the Ehrenfest relation (Chapter 7), this gives $\frac{d\langle p \rangle}{dt} = -\langle \frac{dV}{dx} \rangle$ — the quantum version of Newton's second law! The commutator encodes the dynamics.

The commutator-Poisson bracket correspondence

There is a deep structural parallel between quantum commutators and classical Poisson brackets. In classical Hamiltonian mechanics, the Poisson bracket of two functions $A(x,p)$ and $B(x,p)$ is:

$$\{A, B\}_{\text{PB}} = \frac{\partial A}{\partial x}\frac{\partial B}{\partial p} - \frac{\partial A}{\partial p}\frac{\partial B}{\partial x}$$

Dirac recognized that the quantum commutator is related to the classical Poisson bracket by:

$$[\hat{A}, \hat{B}] = i\hbar\{A, B\}_{\text{PB}} + O(\hbar^2)$$

The canonical Poisson bracket $\{x, p\}_{\text{PB}} = 1$ maps to $[\hat{x}, \hat{p}] = i\hbar \cdot 1 = i\hbar$. This is not a coincidence — it is the correspondence principle in its most precise algebraic form. Quantum mechanics reduces to classical mechanics when commutators (proportional to $\hbar$) become negligible.

🔄 Check Your Understanding: Without calculating, predict whether $[\hat{p}_x, \hat{p}_y]$ is zero or nonzero. Justify your prediction on physical grounds. (Hint: Can you know both components of momentum simultaneously?)

6.5 The Generalized Uncertainty Principle

This is the centerpiece of the chapter. We will derive a result that is simultaneously beautiful mathematics and deep physics.

Setup

Let $\hat{A}$ and $\hat{B}$ be two Hermitian operators, and let $|\psi\rangle$ be any normalized state. Define the uncertainties (standard deviations) as:

$$\sigma_A^2 = \langle (\hat{A} - \langle A \rangle)^2 \rangle = \langle \hat{A}^2 \rangle - \langle \hat{A} \rangle^2$$

$$\sigma_B^2 = \langle (\hat{B} - \langle B \rangle)^2 \rangle = \langle \hat{B}^2 \rangle - \langle \hat{B} \rangle^2$$

where $\langle A \rangle = \langle \psi | \hat{A} | \psi \rangle$. These are the statistical spreads: if you prepared many copies of $|\psi\rangle$ and measured $A$ on each, $\sigma_A$ would be the standard deviation of your results.

Define the shifted operators:

$$\hat{A}' = \hat{A} - \langle A \rangle, \qquad \hat{B}' = \hat{B} - \langle B \rangle$$

Note that $[\hat{A}', \hat{B}'] = [\hat{A}, \hat{B}]$ (the constant shifts cancel). Also:

$$\sigma_A^2 = \langle \hat{A}'^2 \rangle = \langle \hat{A}'\psi | \hat{A}'\psi \rangle \equiv \|f\|^2, \qquad \sigma_B^2 = \|g\|^2$$

where $f = \hat{A}'\psi$ and $g = \hat{B}'\psi$.

The derivation

Step 1: Cauchy-Schwarz inequality. For any two vectors $f$ and $g$ in a Hilbert space:

$$\|f\|^2 \|g\|^2 \geq |\langle f | g \rangle|^2$$

This is the infinite-dimensional generalization of the familiar fact that $|\vec{u} \cdot \vec{v}| \leq |\vec{u}||\vec{v}|$ — the dot product of two vectors cannot exceed the product of their magnitudes. Equality holds if and only if $f$ and $g$ are proportional ($f = \lambda g$ for some complex $\lambda$).

To prove the Cauchy-Schwarz inequality, consider the non-negative quantity $\|f + \lambda g\|^2 \geq 0$ for any complex $\lambda$. Expand: $\|f\|^2 + \lambda\langle f | g \rangle + \lambda^*\langle g | f \rangle + |\lambda|^2\|g\|^2 \geq 0$. Choose $\lambda = -\langle g | f \rangle / \|g\|^2$ to minimize. After simplification, you obtain $\|f\|^2\|g\|^2 \geq |\langle f | g \rangle|^2$, as claimed.

In our case:

$$\sigma_A^2 \sigma_B^2 \geq |\langle \hat{A}'\psi | \hat{B}'\psi \rangle|^2 = |\langle \psi | \hat{A}'\hat{B}' | \psi \rangle|^2$$

Step 2: Decompose the operator product. Any product of operators can be written as:

$$\hat{A}'\hat{B}' = \frac{1}{2}[\hat{A}', \hat{B}'] + \frac{1}{2}\{\hat{A}', \hat{B}'\}$$

where $\{\hat{A}', \hat{B}'\} = \hat{A}'\hat{B}' + \hat{B}'\hat{A}'$ is the anticommutator. The commutator part is anti-Hermitian (its expectation value is purely imaginary), and the anticommutator part is Hermitian (its expectation value is purely real). Therefore:

$$\langle \hat{A}'\hat{B}' \rangle = \underbrace{\frac{1}{2}\langle [\hat{A}', \hat{B}'] \rangle}_{\text{purely imaginary}} + \underbrace{\frac{1}{2}\langle \{\hat{A}', \hat{B}'\} \rangle}_{\text{purely real}}$$

Step 3: Apply the modulus squared. For a complex number $z = x + iy$:

$$|z|^2 = x^2 + y^2 \geq y^2$$

So:

$$|\langle \hat{A}'\hat{B}' \rangle|^2 \geq \frac{1}{4}|\langle [\hat{A}', \hat{B}'] \rangle|^2 = \frac{1}{4}|\langle [\hat{A}, \hat{B}] \rangle|^2$$

Step 4: Combine. Putting it all together:

$$\boxed{\sigma_A \sigma_B \geq \frac{1}{2}\left|\langle [\hat{A}, \hat{B}] \rangle\right|}$$

This is the Robertson uncertainty relation (1929), also called the generalized uncertainty principle.

💡 Key Insight: The uncertainty principle is not about the clumsiness of measurement instruments. It is not about disturbing the system. It is a mathematical theorem about the statistical spreads of any two non-commuting observables in any quantum state. It follows from the structure of Hilbert space itself.

🔴 Warning: The Schrodinger extension includes the anticommutator term:

$$\sigma_A^2 \sigma_B^2 \geq \frac{1}{4}|\langle [\hat{A}, \hat{B}] \rangle|^2 + \frac{1}{4}|\langle \{\hat{A}', \hat{B}'\} \rangle|^2$$

The Robertson relation is a weaker bound (it drops a non-negative term). The Schrodinger form is tighter but less commonly used because the anticommutator contribution depends on the state in a more complicated way.

Application 1: Position-momentum uncertainty

For $\hat{A} = \hat{x}$ and $\hat{B} = \hat{p}$:

$$[\hat{x}, \hat{p}] = i\hbar$$

The expectation value $\langle [\hat{x}, \hat{p}] \rangle = i\hbar$ (it is a constant, independent of the state). Therefore:

$$\sigma_x \sigma_p \geq \frac{1}{2}|i\hbar| = \frac{\hbar}{2}$$

$$\boxed{\Delta x \, \Delta p \geq \frac{\hbar}{2}}$$

This is the Heisenberg uncertainty principle in its precise mathematical form. The bound $\hbar/2$ is saturated (achieved as equality) by Gaussian wave packets — the coherent states of the harmonic oscillator.

📊 By the Numbers: For an electron confined to an atom ($\Delta x \sim 10^{-10}$ m), the minimum momentum uncertainty is $\Delta p \geq \frac{\hbar}{2\Delta x} \sim 5.3 \times 10^{-25}$ kg m/s, corresponding to a velocity uncertainty of $\sim 5.8 \times 10^{5}$ m/s — about 0.2% of the speed of light. This is why atoms are intrinsically quantum.

Application 2: Angular momentum components

Consider $\hat{L}_x$ and $\hat{L}_y$, two components of orbital angular momentum. Their commutator is (we will derive this from first principles in Chapter 12):

$$[\hat{L}_x, \hat{L}_y] = i\hbar\hat{L}_z$$

Applying the generalized uncertainty principle:

$$\sigma_{L_x}\sigma_{L_y} \geq \frac{\hbar}{2}|\langle \hat{L}_z \rangle|$$

Unlike the position-momentum case, the right side depends on the state. If $\langle \hat{L}_z \rangle = 0$ (say, for an $m = 0$ state), the bound becomes trivial and says nothing. If $\langle \hat{L}_z \rangle = m\hbar$, then:

$$\sigma_{L_x}\sigma_{L_y} \geq \frac{m\hbar^2}{2}$$

This state-dependence is characteristic of most uncertainty relations — the position-momentum case, where the commutator is a constant, is actually the exception rather than the rule.

Application 3: Number and phase in the QHO

For the quantum harmonic oscillator, we can ask about the uncertainty in the number of quanta $n$ (measured by $\hat{n} = \hat{a}^\dagger\hat{a}$) and in the "position-like" observable $\hat{X} = (\hat{a} + \hat{a}^\dagger)/\sqrt{2}$. Since $[\hat{n}, \hat{a}] = -\hat{a}$, there is a related uncertainty between number and phase — a result that will become central in quantum optics (Chapter 27). For now, we note that the number state $|n\rangle$ (definite number of quanta, $\sigma_n = 0$) has maximal phase uncertainty, while a coherent state $|\alpha\rangle$ (nearly definite phase for large $|\alpha|$) has $\sigma_n = |\alpha|$. The trade-off is an uncertainty relation in disguise.

What the uncertainty principle does NOT say

It is worth being explicit about common misinterpretations:

1. It does NOT say a single measurement is imprecise. Each individual measurement of $\hat{A}$ yields an exact eigenvalue. The spread $\sigma_A$ refers to the ensemble of results from many measurements on identically prepared states.

2. It does NOT say "everything is uncertain." A system in an eigenstate of $\hat{A}$ has $\sigma_A = 0$ — the observable $A$ is perfectly definite. The uncertainty principle constrains the product of uncertainties for incompatible observables.

3. It does NOT depend on the measurement apparatus. The bound $\sigma_A\sigma_B \geq \frac{1}{2}|\langle[\hat{A},\hat{B}]\rangle|$ is a property of the state and the operators, not of any experimental setup.

4. It does NOT imply a "disturbance" by measurement. While measurement does disturb the state (Section 6.8), the uncertainty principle is logically independent of this fact. It is a statement about preparation, not about disturbance. See Case Study 1 for a detailed discussion.

🧪 Experiment: The Stern-Gerlach experiment, which we discuss more carefully in Section 6.8, demonstrates the uncertainty principle vividly. A beam of spin-1/2 particles is first filtered through a $z$-oriented analyzer, producing a definite $S_z$ eigenstate. A subsequent $x$-oriented analyzer then reveals that $S_x$ is completely uncertain: the beam splits 50-50. You cannot simultaneously know $S_z$ and $S_x$ because $[\hat{S}_x, \hat{S}_z] \neq 0$.

🔄 Check Your Understanding: If two observables $\hat{A}$ and $\hat{B}$ satisfy $[\hat{A}, \hat{B}] = 0$, what does the generalized uncertainty principle tell you? Is this consistent with the claim in Section 6.4 that commuting observables can be simultaneously measured?

6.6 Compatible and Incompatible Observables

Compatible observables

Two observables $\hat{A}$ and $\hat{B}$ are compatible if $[\hat{A}, \hat{B}] = 0$. As we discussed in Section 6.4, compatible observables share a common eigenbasis — there exist states that are simultaneously eigenstates of both operators.

The theorem of simultaneous diagonalizability

Theorem 6.3. If $[\hat{A}, \hat{B}] = 0$, then $\hat{A}$ and $\hat{B}$ can be simultaneously diagonalized — they share a complete set of common eigenstates.

Proof sketch (non-degenerate case). Suppose $\hat{A}\psi_a = a\psi_a$. Apply $\hat{B}$ to both sides:

$$\hat{B}\hat{A}\psi_a = a\hat{B}\psi_a$$

Since $[\hat{A}, \hat{B}] = 0$, we have $\hat{A}\hat{B} = \hat{B}\hat{A}$, so:

$$\hat{A}(\hat{B}\psi_a) = a(\hat{B}\psi_a)$$

This says $\hat{B}\psi_a$ is also an eigenfunction of $\hat{A}$ with eigenvalue $a$. If $a$ is non-degenerate, there is only one linearly independent eigenfunction for that eigenvalue, so $\hat{B}\psi_a$ must be proportional to $\psi_a$:

$$\hat{B}\psi_a = b\psi_a$$

So $\psi_a$ is an eigenfunction of both operators. $\blacksquare$

The degenerate case requires more care (we choose the right linear combinations within each degenerate subspace), but the conclusion is the same: commuting operators share a complete set of simultaneous eigenstates.

Incompatible observables

If $[\hat{A}, \hat{B}] \neq 0$, then in general there is no state that is simultaneously an eigenstate of both. The generalized uncertainty principle quantifies the minimum unavoidable trade-off.

Complete Sets of Commuting Observables (CSCOs)

In many physical systems, a single observable is not sufficient to uniquely label every quantum state. For example, the hydrogen atom energy levels are degenerate: for a given $n$, there are $n^2$ states with the same energy (ignoring spin). The energy eigenvalue alone does not specify which state you are in.

A complete set of commuting observables (CSCO) is a set of observables $\{\hat{A}, \hat{B}, \hat{C}, \ldots\}$ such that:

1. All pairs commute: $[\hat{A}, \hat{B}] = [\hat{A}, \hat{C}] = [\hat{B}, \hat{C}] = \cdots = 0$
2. Specifying the eigenvalues of all operators in the set uniquely determines the state (up to an overall phase).

Example: The hydrogen atom. The CSCO for the hydrogen atom (without spin) is:

$$\{\hat{H}, \hat{L}^2, \hat{L}_z\}$$

• $\hat{H}$ determines $n$ (the energy level)
• $\hat{L}^2$ determines $l$ (the orbital angular momentum quantum number)
• $\hat{L}_z$ determines $m$ (the magnetic quantum number)

All three commute:

$$[\hat{H}, \hat{L}^2] = 0, \qquad [\hat{H}, \hat{L}_z] = 0, \qquad [\hat{L}^2, \hat{L}_z] = 0$$

🔗 Connection: In Chapter 5, you solved the hydrogen atom and found that the eigenstates $\psi_{nlm}(r, \theta, \phi)$ are labeled by three quantum numbers. Now you see the deeper reason: these quantum numbers are the eigenvalues of a CSCO. The separation of variables into radial and angular parts works precisely because these operators commute.

Together, the eigenvalues $(n, l, m)$ uniquely specify each state. If we include spin (Chapter 13), we extend the CSCO to $\{\hat{H}, \hat{L}^2, \hat{L}_z, \hat{S}^2, \hat{S}_z\}$, adding the quantum numbers $s$ and $m_s$.

Why "complete"? A set of commuting observables that does not uniquely label states is incomplete. For instance, $\{\hat{H}\}$ alone is an incomplete set for hydrogen because the energy is degenerate. The set $\{\hat{H}, \hat{L}^2\}$ is also incomplete (different $m$ values give different states with the same $n$ and $l$). Only when we add $\hat{L}_z$ do we get unique labels.

Why does this matter? CSCOs tell you what can be simultaneously known about a quantum system. They determine which quantum numbers are "good" quantum numbers — the ones that remain well-defined over time if the operators all commute with the Hamiltonian.

⚠️ Common Misconception: Students sometimes think there is a unique CSCO for each system. There is not. For example, $\{\hat{H}, \hat{L}^2, \hat{L}_x\}$ is an equally valid CSCO for hydrogen (we would just use different eigenstates). The choice of CSCO is one of physics and convenience, not uniqueness.

Good quantum numbers

A good quantum number is an eigenvalue of an operator that commutes with the Hamiltonian: $[\hat{A}, \hat{H}] = 0$. Its physical significance is that the corresponding observable is a conserved quantity — its expectation value does not change in time (this follows from the Ehrenfest relation, Eq. 6.7, which gives $d\langle \hat{A} \rangle / dt = 0$ when $[\hat{A}, \hat{H}] = 0$).

For hydrogen: - $n$ is a good quantum number because $[\hat{H}, \hat{H}] = 0$ (trivially). - $l$ is a good quantum number because $[\hat{L}^2, \hat{H}] = 0$ (the Coulomb potential is spherically symmetric). - $m$ is a good quantum number because $[\hat{L}_z, \hat{H}] = 0$.

If we add a magnetic field in the $z$-direction (the Zeeman effect, Chapter 18), $l$ and $m$ remain good quantum numbers, but the degeneracy in $m$ is lifted. If we add a magnetic field in the $x$-direction, $m$ is no longer a good quantum number — the appropriate CSCO changes.

The choice of CSCO is therefore physically meaningful: it reflects the symmetries of the problem. We will develop this connection between symmetries, commutators, and conservation laws fully in Chapter 10.

Worked Example 6.7: Building a CSCO for the 2D harmonic oscillator

Consider a particle in a 2D isotropic harmonic oscillator potential $V = \frac{1}{2}m\omega^2(x^2 + y^2)$. The Hamiltonian separates as $\hat{H} = \hat{H}_x + \hat{H}_y$ where $\hat{H}_x = \frac{\hat{p}_x^2}{2m} + \frac{1}{2}m\omega^2\hat{x}^2$ and similarly for $\hat{H}_y$.

The energy eigenvalues are $E_{n_x, n_y} = \hbar\omega(n_x + n_y + 1)$. Let $N = n_x + n_y$. The $N$-th level has degeneracy $N + 1$ (the pairs $(n_x, n_y)$ with $n_x + n_y = N$).

Is $\{\hat{H}\}$ a CSCO? No — the energy levels are degenerate.

Is $\{\hat{H}, \hat{L}_z\}$ a CSCO? The angular momentum $\hat{L}_z = \hat{x}\hat{p}_y - \hat{y}\hat{p}_x$ commutes with $\hat{H}$ (by the rotational symmetry of the potential). The eigenvalues of $\hat{L}_z$ are $m\hbar$ where $m = -N, -N+2, \ldots, N-2, N$. Together, $N$ and $m$ uniquely label each state, so $\{\hat{H}, \hat{L}_z\}$ is a valid CSCO.

Alternatively: $\{\hat{H}_x, \hat{H}_y\}$ is also a CSCO, using the Cartesian quantum numbers $(n_x, n_y)$ instead of $(N, m)$. The two CSCOs represent different physical choices — one adapted to rotational symmetry, the other to Cartesian separability.

✅ Checkpoint: Consider a particle in a 3D infinite cubical box with side length $L$. The energy eigenvalues are $E_{n_x, n_y, n_z} = \frac{\pi^2\hbar^2}{2mL^2}(n_x^2 + n_y^2 + n_z^2)$. Identify a CSCO for this system. Are the energy levels degenerate? If so, what additional operators resolve the degeneracy?

6.7 The Energy-Time Uncertainty Relation

Why energy-time is different

The position-momentum uncertainty relation $\Delta x \Delta p \geq \hbar/2$ arises from $[\hat{x}, \hat{p}] = i\hbar$. It is tempting to write an analogous $\Delta E \Delta t \geq \hbar/2$ — and indeed, such a relation is widely quoted. But there is a fundamental difference.

Time is not an observable in quantum mechanics. There is no Hermitian operator $\hat{t}$ whose eigenvalues are "the results of measuring time." Time is a parameter, not a dynamical variable. The system evolves in time; it does not have a "time eigenvalue."

⚠️ Common Misconception: The energy-time uncertainty relation $\Delta E \Delta t \geq \hbar/2$ is NOT directly analogous to $\Delta x \Delta p \geq \hbar/2$. It cannot be derived from a commutator $[\hat{H}, \hat{t}]$ because $\hat{t}$ does not exist as an operator.

So what does $\Delta E \Delta t$ actually mean?

The Mandelstam-Tamm derivation

The correct approach, due to Mandelstam and Tamm (1945), defines $\Delta t$ in terms of how fast a physical observable changes.

Let $\hat{Q}$ be any observable (position, spin, whatever). The time rate of change of its expectation value is given by the Ehrenfest relation (derived fully in Chapter 7):

$$\frac{d\langle \hat{Q} \rangle}{dt} = \frac{1}{i\hbar}\langle [\hat{Q}, \hat{H}] \rangle$$

(assuming $\hat{Q}$ has no explicit time dependence). Now apply the generalized uncertainty principle to $\hat{Q}$ and $\hat{H}$:

$$\sigma_Q \sigma_H \geq \frac{1}{2}|\langle [\hat{Q}, \hat{H}] \rangle| = \frac{\hbar}{2}\left|\frac{d\langle \hat{Q} \rangle}{dt}\right|$$

Define the characteristic time for the observable $Q$ to change appreciably:

$$\Delta t_Q \equiv \frac{\sigma_Q}{|d\langle \hat{Q} \rangle / dt|}$$

This is the time it takes for $\langle Q \rangle$ to shift by one standard deviation — a natural measure of "how quickly the system changes as seen through the lens of $Q$." Then:

$$\boxed{\sigma_H \, \Delta t_Q \geq \frac{\hbar}{2}}$$

or equivalently:

$$\Delta E \, \Delta t \geq \frac{\hbar}{2}$$

Interpretation: A system with a well-defined energy ($\sigma_H$ small) must have slowly varying expectation values ($\Delta t$ large) for any observable. Conversely, a system whose properties change rapidly cannot have a precisely defined energy. A stationary state ($\sigma_H = 0$) has $\Delta t = \infty$: nothing ever changes, as we know — the expectation values of all observables are time-independent.

The lifetime-linewidth relation

One of the most important applications of $\Delta E \Delta t \geq \hbar/2$ is to unstable states.

Consider an excited atomic state with lifetime $\tau$ (the average time before the atom decays by emitting a photon). The state exists for a time $\Delta t \sim \tau$, so:

$$\Delta E \geq \frac{\hbar}{2\tau}$$

The energy is not perfectly sharp — it has a natural linewidth $\Gamma = \Delta E / \hbar \sim 1/(2\tau)$ (the factor of 2 depends on precise definitions). In terms of frequency:

$$\Delta \nu \sim \frac{1}{2\pi\tau}$$

📊 By the Numbers: The hydrogen 2p → 1s Lyman-alpha transition has a lifetime $\tau \approx 1.6 \times 10^{-9}$ s, giving a natural linewidth:

$$\Delta\nu \sim \frac{1}{2\pi \times 1.6 \times 10^{-9}} \approx 100 \text{ MHz}$$

This corresponds to $\Delta\lambda \sim 1.2 \times 10^{-5}$ nm, or about one part in $10^7$ of the transition wavelength. The line is narrow, but not infinitely narrow — and the width is a direct consequence of the energy-time uncertainty principle.

🧪 Experiment: In nuclear and particle physics, unstable particles are observed with a measurable energy spread. The $Z^0$ boson has a decay width $\Gamma \approx 2.5$ GeV, corresponding to a lifetime $\tau = \hbar/\Gamma \approx 2.6 \times 10^{-25}$ s. The particle barely exists before it decays, and its mass is correspondingly uncertain.

Energy-time uncertainty and quantum tunneling

The energy-time relation also provides a heuristic understanding of quantum tunneling. A particle "borrowing" energy $\Delta E$ to pass through a barrier can only do so for a time $\Delta t \lesssim \hbar/(2\Delta E)$. This is not a rigorous derivation — the correct treatment uses the tunneling probability from Chapter 3 — but it provides useful physical intuition for estimates.

🔗 Spaced Review from Chapter 3: In Chapter 3, you calculated the tunneling probability through a rectangular barrier:

$$T \approx e^{-2\kappa d}$$

where $\kappa = \sqrt{2m(V_0 - E)}/\hbar$ and $d$ is the barrier width. The energy-time picture gives a consistent estimate: the particle "borrows" energy $\Delta E = V_0 - E$ for time $\Delta t \sim d/v \sim md/(\hbar\kappa)$, and the condition $\Delta E \Delta t \sim \hbar$ gives $\kappa d \sim 1$, consistent with significant tunneling when $\kappa d \lesssim 1$.

🔄 Check Your Understanding: A pion ($\pi^0$) has a lifetime of $\tau \approx 8.5 \times 10^{-17}$ s. Estimate the natural width $\Gamma$ of the pion's mass-energy peak. Express your answer in eV.

6.8 Measurement in Quantum Mechanics: A First Serious Look

Everything in this chapter has been building toward this section. We now have the mathematical machinery to state precisely what quantum mechanics says about measurement — and to appreciate how strange it is.

The measurement postulates

Let $\hat{A}$ be a Hermitian operator representing some observable, with eigenvalues $\{a_n\}$ and normalized eigenfunctions $\{\psi_n\}$. Let $\psi$ be the state of the system just before measurement.

Postulate 1 (Possible outcomes): The only possible results of measuring the observable $A$ are the eigenvalues $\{a_n\}$ of $\hat{A}$.

Postulate 2 (Born rule for probabilities): The probability of obtaining the result $a_n$ is:

$$P(a_n) = |\langle \psi_n | \psi \rangle|^2 = \left|\int_{-\infty}^{+\infty} \psi_n^*(x) \psi(x) \, dx\right|^2$$

This requires expanding $\psi$ in the eigenbasis of $\hat{A}$:

$$\psi = \sum_n c_n \psi_n, \qquad c_n = \langle \psi_n | \psi \rangle$$

Then $P(a_n) = |c_n|^2$, and $\sum_n P(a_n) = 1$ by normalization.

Postulate 3 (State collapse / projection postulate): Immediately after measuring $\hat{A}$ and obtaining the result $a_n$, the state of the system becomes $\psi_n$.

$$\psi \xrightarrow{\text{measure } A, \text{ get } a_n} \psi_n$$

This is the mathematically precise statement of what is colloquially called "wave function collapse."

The Stern-Gerlach experiment: Measurement made concrete

The Stern-Gerlach experiment (1922) provides the cleanest illustration of these postulates. Although spin is formalized in Chapter 13, we can describe the essential physics now.

A beam of silver atoms (each carrying a spin-1/2 magnetic moment) passes through an inhomogeneous magnetic field oriented along the $z$-axis. Classically, the beam should spread continuously. Quantum mechanically, it splits into exactly two discrete beams — corresponding to $S_z = +\hbar/2$ ("spin up") and $S_z = -\hbar/2$ ("spin down").

This is Postulate 1 in action: the only possible results are the eigenvalues of $\hat{S}_z$.

Now the crucial part. Take the "spin up" beam (which has definite $S_z = +\hbar/2$) and send it through a second Stern-Gerlach apparatus oriented along the $x$-axis. The result: the beam splits 50-50 into $S_x = +\hbar/2$ and $S_x = -\hbar/2$ components.

This is Postulate 2: The $S_z = +\hbar/2$ eigenstate is an equal superposition of $S_x$ eigenstates, so the probabilities are $|1/\sqrt{2}|^2 = 1/2$ each.

Now take either $S_x$ beam and send it through a third apparatus oriented along $z$. You might expect the result to be $S_z = +\hbar/2$ (since you started with that). But no — the beam again splits 50-50. The measurement of $S_x$ has destroyed the information about $S_z$.

This is Postulate 3 — state collapse. The $S_x$ measurement projected the state onto an $S_x$ eigenstate, which is not an eigenstate of $S_z$. The previous $S_z$ information was irrecoverably lost.

🚪 Threshold Concept: Measurement changes the state. Observation in quantum mechanics is not passive. It is not like looking at a ball and noting its position — the ball was there before you looked, and it's there after. In quantum mechanics, the act of measuring creates the definite value. Before measurement, the system was in a superposition. After measurement, it is in an eigenstate. This is not a statement about our ignorance. It is a statement about reality — or at the very least, about the mathematical structure of the most successful physical theory ever constructed.

Let this sink in. This is the single most important conceptual lesson in quantum mechanics.

Worked Example 6.8: Measurement probabilities for the QHO

🔗 Spaced Review from Chapter 4: Suppose a quantum harmonic oscillator is in the state:

$$\psi = \frac{1}{\sqrt{3}}\psi_0 + \sqrt{\frac{2}{3}}\psi_1$$

where $\psi_0$ and $\psi_1$ are the ground and first excited states with energies $E_0 = \frac{1}{2}\hbar\omega$ and $E_1 = \frac{3}{2}\hbar\omega$.

What are the possible results of an energy measurement, and with what probabilities?

The state is already expanded in energy eigenstates: $c_0 = 1/\sqrt{3}$, $c_1 = \sqrt{2/3}$.

• Possible results: $E_0 = \frac{1}{2}\hbar\omega$ or $E_1 = \frac{3}{2}\hbar\omega$. (Only eigenvalues of $\hat{H}$ — Postulate 1.)
• Probabilities: $P(E_0) = |c_0|^2 = 1/3$, $P(E_1) = |c_1|^2 = 2/3$. (Born rule — Postulate 2.)
• Check: $P(E_0) + P(E_1) = 1/3 + 2/3 = 1$. $\checkmark$

If you measure and get $E_0$, what is the state after measurement?

The state collapses to $\psi_0$ (Postulate 3). The $\psi_1$ component is gone. If you immediately measure energy again, you get $E_0$ with certainty.

What is $\langle E \rangle$ before measurement?

$$\langle E \rangle = \frac{1}{3}\left(\frac{1}{2}\hbar\omega\right) + \frac{2}{3}\left(\frac{3}{2}\hbar\omega\right) = \frac{1}{6}\hbar\omega + \hbar\omega = \frac{7}{6}\hbar\omega$$

Note: $\langle E \rangle = \frac{7}{6}\hbar\omega$ is not an eigenvalue of $\hat{H}$. You will never get $\frac{7}{6}\hbar\omega$ from a single measurement — you always get $\frac{1}{2}\hbar\omega$ or $\frac{3}{2}\hbar\omega$. The expectation value is the statistical average over many measurements.

The expectation value revisited

With the measurement postulates in hand, the expectation value formula from Chapter 2 acquires its full meaning:

$$\langle \hat{A} \rangle = \langle \psi | \hat{A} | \psi \rangle = \sum_n a_n |c_n|^2 = \sum_n a_n P(a_n)$$

This is not the value you get from a single measurement. It is the average over many identical measurements on identically prepared systems. Each individual measurement yields one eigenvalue $a_n$; the expectation value is the statistical mean.

Repeated measurement and the projection postulate

If you measure $\hat{A}$ and get $a_n$, the state collapses to $\psi_n$. If you immediately measure $\hat{A}$ again, you get $a_n$ with certainty:

$$P(a_n) = |\langle \psi_n | \psi_n \rangle|^2 = 1$$

The second measurement confirms the first. This is experimentally verified and is a consistency requirement of the theory.

But if you wait, and the Hamiltonian does not commute with $\hat{A}$, the state will evolve away from $\psi_n$ under time evolution (Chapter 7), and a later measurement of $\hat{A}$ may yield a different eigenvalue.

The measurement problem: An honest preview

We have just described the standard textbook account of measurement: collapse occurs, eigenvalues appear, probabilities follow the Born rule. This account is extraordinarily successful — it has never been contradicted by experiment.

But there is a deep problem. The Schrodinger equation is linear and deterministic. It says nothing about collapse. If we model the measurement apparatus as a quantum system (which it is, being made of atoms), the Schrodinger equation predicts that the apparatus+system ends up in a superposition — a "Schrodinger's cat" state — not a definite outcome.

This is the measurement problem, and it remains unsolved. Different interpretations of quantum mechanics (Copenhagen, Many Worlds, Bohmian, decoherence-based, etc.) offer different resolutions, but none has achieved consensus. We will return to this honestly and at length in Chapter 33.

⚖️ Interpretation: The measurement problem is not a failure of quantum mechanics as a predictive tool. The formalism works flawlessly. The problem is at the level of ontology: what is actually happening? Different interpretations agree on all experimental predictions but disagree on what is "real." This is one of the great open questions in physics.

For now, we use the projection postulate as a practical rule. It works. But the honest physicist remembers that why it works is an open question.

💡 Key Insight: The operator formalism we have developed in this chapter is interpretation-independent. Whether you believe in collapse, many worlds, or pilot waves, you use the same operators, the same commutators, the same uncertainty relations. The formalism is the bedrock; the interpretations are the contested superstructure.

Degeneracy and measurement

What happens if the eigenvalue $a_n$ is degenerate — that is, if multiple linearly independent eigenstates share the same eigenvalue? Suppose $\psi_{n,1}, \psi_{n,2}, \ldots, \psi_{n,g}$ are $g$ orthonormal eigenstates of $\hat{A}$ all with eigenvalue $a_n$.

The probability of getting $a_n$ is:

$$P(a_n) = \sum_{k=1}^g |c_{n,k}|^2 = \sum_{k=1}^g |\langle \psi_{n,k} | \psi \rangle|^2$$

After measurement, the state collapses to the projection of $\psi$ onto the degenerate subspace, normalized:

$$\psi \to \frac{\sum_{k=1}^g c_{n,k} \psi_{n,k}}{\sqrt{P(a_n)}}$$

This is the Luders rule (a generalization of the von Neumann projection postulate). It is important: the state does not collapse to a single eigenstate when there is degeneracy, but to the component of $\psi$ within the degenerate subspace. To determine the state completely, you need to measure additional commuting observables — which is precisely why CSCOs (Section 6.6) are essential.

Sequential measurements: A preview of quantum logic

The Stern-Gerlach chain experiment illustrates a remarkable feature of quantum logic. In classical physics, if you know that a particle has property $A$ (say, $S_z = +\hbar/2$) and then learn property $B$ ($S_x = +\hbar/2$), you still know $A$. Knowledge accumulates.

In quantum mechanics, learning $B$ can destroy your knowledge of $A$. This is not a failure of memory or record-keeping — it is a consequence of the projection postulate and the non-commutativity of $\hat{A}$ and $\hat{B}$. The logical structure of quantum propositions is fundamentally different from classical logic. This observation, first made by Birkhoff and von Neumann (1936), is the starting point of quantum logic — a topic we will touch upon in Chapter 33.

🔄 Check Your Understanding: An operator $\hat{A}$ has eigenvalues $a_1 = 1$ and $a_2 = 2$ with eigenstates $\psi_1$ and $\psi_2$. A system is in the state $\psi = \frac{3}{5}\psi_1 + \frac{4}{5}\psi_2$. (a) What are the possible results of measuring $A$? (b) What is the probability of each? (c) After measuring and getting $a_1 = 1$, what is the state? (d) What is the probability of getting $a_1 = 1$ on a second immediate measurement?

6.9 Summary and Project Checkpoint

What we have accomplished

This chapter has transformed quantum mechanics from a collection of specific solutions into a general framework. Let us summarize the key results:

The road ahead

In Chapter 7, we will use the operator formalism to construct the time-evolution operator $\hat{U}(t) = e^{-i\hat{H}t/\hbar}$ and derive Ehrenfest's theorem, connecting quantum expectation values to classical equations of motion. The commutators from this chapter will be essential.

In Chapter 8, we will rewrite everything in Dirac notation, which makes the linear algebra structure of quantum mechanics even more explicit. The Stern-Gerlach experiment will become a two-dimensional matrix problem, and operators will become matrices.

🔗 Connection: The spin-1/2 system introduced in Section 6.8 via Stern-Gerlach will become the paradigmatic quantum system in Chapter 8 onward. It lives in a two-dimensional Hilbert space where everything can be written as $2 \times 2$ matrices — a paradise for building intuition.

Project Checkpoint: Quantum Simulation Toolkit v0.6

In this chapter's code (see code/project-checkpoint.py), you will add to your growing quantum simulation toolkit:

1. Operator class — Represents quantum operators as matrices, supports addition, multiplication, and adjoint.
2. commutator(A, B) function — Computes $[\hat{A}, \hat{B}] = \hat{A}\hat{B} - \hat{B}\hat{A}$.
3. uncertainty_product(state, A, B) function — Computes $\sigma_A \sigma_B$ for a given state and verifies the generalized uncertainty principle.

These tools will be used in every subsequent chapter.

🔄 Spaced Review

Before moving on, revisit these concepts from earlier chapters:

• From Chapter 2: What are the position-space representations of $\hat{x}$ and $\hat{p}$? Write down the expectation value formula $\langle \hat{A} \rangle = \int \psi^* \hat{A} \psi \, dx$. (You now understand why this formula works: it is the statistical mean of eigenvalues weighted by Born-rule probabilities.)

• From Chapter 3: Why are the infinite-well eigenfunctions orthogonal? (Because they are eigenfunctions of a Hermitian operator with distinct eigenvalues — Theorem 6.2.)

• From Chapter 4: Show that $[\hat{a}, \hat{a}^\dagger] = 1$ for the QHO ladder operators. (We derived this in Worked Example 6.5, but try it again from scratch as practice.)

📜 Historical Note: Heisenberg's Matrix Mechanics

In 1925, Werner Heisenberg — then 23 years old and suffering from hay fever on the island of Helgoland — invented the first complete formulation of quantum mechanics. His approach represented physical observables not as differential operators (Schrodinger's approach) but as infinite matrices. Position and momentum became arrays of numbers, and the canonical commutation relation $[\hat{x}, \hat{p}] = i\hbar$ was the equation that replaced classical Poisson brackets.

Heisenberg initially did not recognize his arrays as matrices. It was Max Born, his doctoral advisor, who saw the connection and, together with Pascual Jordan, developed the full matrix mechanics. Schrodinger later proved that his wave mechanics and Heisenberg's matrix mechanics are mathematically equivalent — different languages for the same physics.

The operator formalism we have developed in this chapter is the common ground: it encompasses both Heisenberg's matrices and Schrodinger's differential operators as special representations of abstract operators on Hilbert space.

Key Equations Reference

For convenience, the essential equations from this chapter:

$$[\hat{x}, \hat{p}] = i\hbar \tag{6.1}$$

$$[\hat{A}, \hat{B}\hat{C}] = [\hat{A}, \hat{B}]\hat{C} + \hat{B}[\hat{A}, \hat{C}] \tag{6.2}$$

$$[\hat{a}, \hat{a}^\dagger] = 1 \tag{6.3}$$

$$\sigma_A \sigma_B \geq \frac{1}{2}|\langle [\hat{A}, \hat{B}] \rangle| \tag{6.4}$$

$$\Delta x \, \Delta p \geq \frac{\hbar}{2} \tag{6.5}$$

$$\Delta E \, \Delta t \geq \frac{\hbar}{2} \tag{6.6}$$

$$\frac{d\langle \hat{Q} \rangle}{dt} = \frac{1}{i\hbar}\langle [\hat{Q}, \hat{H}] \rangle \tag{6.7}$$