Chapter 8: Linear Algebra for Quantum Mechanics — Vector Spaces, Operators, and Dirac Notation

The Bridge Chapter

"The notation isn't just a shorthand — it is the physics."

This is the most important chapter in this book. That is not hyperbole. Every chapter from this point forward — angular momentum, spin, perturbation theory, entanglement, quantum computing — uses the language you are about to learn. If you leave this chapter comfortable reading and writing Dirac notation, the rest of the book will open up to you. If you leave confused, every subsequent chapter will feel harder than it should.

So we will take our time. We will be careful. And we will constantly connect the new notation back to the wave mechanics you already know, until the bridge between the two is so solid you can cross it without thinking.

Here is the promise of this chapter: by the end, you will understand that $\psi(x)$ — the wave function you have been computing with since Chapter 2 — is not the quantum state itself. It is one particular representation of the quantum state, in one particular basis. Dirac notation lets you talk about the state directly, before choosing any representation at all. This is not a philosophical nicety. It is the key to understanding spin (which has no position representation), composite systems (where the Hilbert space is a tensor product), and modern quantum mechanics in general.

Let us begin.

8.1 Why We Need a Better Notation

In Chapters 2 through 7, we developed a powerful framework for quantum mechanics. We learned to write down the Schrodinger equation, solve it for various potentials, compute expectation values, and track time evolution. All of this was done in what we will now call wave mechanics notation: the quantum state is a function $\psi(x)$ or $\Psi(x, t)$, operators are differential expressions like $\hat{p} = -i\hbar \frac{d}{dx}$, and inner products are integrals $\int_{-\infty}^{\infty} \psi^*(x) \phi(x) \, dx$.

This notation served us well. But it has limitations that become apparent precisely when we try to do the things that make quantum mechanics most interesting.

The three problems with wave mechanics notation

Problem 1: Not every quantum state is a wave function. Consider the spin of an electron. In Chapter 6, we encountered the Stern-Gerlach experiment, where a silver atom is deflected either up or down in a magnetic field. The spin state is not a function of position. There is no "$\psi(x)$" for spin-up. The spin state lives in a two-dimensional vector space, not in the infinite-dimensional space of square-integrable functions. Wave mechanics notation has no natural way to describe this.

You might object: can we not simply attach a two-component label to the wave function, writing $\psi_\uparrow(x)$ and $\psi_\downarrow(x)$? We can, and sometimes we do. But this sidesteps the real issue. The spin degree of freedom exists independently of position. We need a formalism that handles it directly — one that does not begin by assuming the state is a function of spatial coordinates.

Problem 2: The notation privileges position. When we write $\psi(x)$, we have already chosen to represent the state in the position basis. But we could equally well work in momentum space, writing $\phi(p) = \frac{1}{\sqrt{2\pi\hbar}} \int \psi(x) e^{-ipx/\hbar} dx$. Or in the energy basis, writing $c_n = \int \psi_n^*(x) \psi(x) \, dx$. These are all representations of the same physical state. Why should one be privileged? In many problems — scattering theory, perturbation theory, quantum computing — position space is not even the most natural choice. We need a notation that treats all representations on equal footing.

Problem 3: Composite systems are awkward. When two particles interact, the state is a function of both positions: $\Psi(x_1, x_2)$. For three particles: $\Psi(x_1, x_2, x_3)$. For $N$ particles: $\Psi(x_1, x_2, \ldots, x_N)$. This quickly becomes unwieldy, and it obscures the mathematical structure (tensor products) that makes multi-particle quantum mechanics tractable. When we study entanglement in Chapter 24, we will need to talk about states like the singlet $\frac{1}{\sqrt{2}}(|\uparrow\downarrow\rangle - |\downarrow\uparrow\rangle)$. There is no natural way to write this as a wave function.

What we need is a notation that: 1. Works for any quantum system — wave functions, spin, photon polarization, qubits, anything. 2. Treats all bases equally — position, momentum, energy, spin projection — without privileging any one. 3. Makes the algebraic structure of quantum mechanics transparent and manipulable.

Paul Dirac invented exactly this notation in 1939, and it bears his name.

🔵 Historical Note — Paul Adrien Maurice Dirac (1902--1984) was one of the most original thinkers in the history of physics. His Principles of Quantum Mechanics (1930) introduced the bra-ket notation that is now universal in the field. Dirac was famously laconic — his colleagues at Cambridge defined the unit "one Dirac" as the minimum number of words spoken per hour. But his notation speaks volumes. It is a masterpiece of mathematical design: compact, general, and physically transparent. When you learn it, you are learning to think the way Dirac thought. As Dirac himself wrote: "A good notation should suggest the right ideas."

🔗 Connection — In Chapter 6, we defined operators, commutators, and the generalized uncertainty principle using wave mechanics notation. In this chapter, we will re-derive all of those results in Dirac notation — and you will see how much cleaner and more general they become. In Chapter 7, we wrote the time-evolution operator as $\hat{U}(t) = e^{-i\hat{H}t/\hbar}$. That expression already hints at Dirac notation: it treats the operator abstractly, without choosing a basis. This chapter completes that transition.

8.2 Kets, Bras, and Inner Products

Hilbert space: the arena of quantum mechanics

Before introducing the notation, let us name the stage on which quantum mechanics is played. A Hilbert space $\mathcal{H}$ is a (possibly infinite-dimensional) vector space over the complex numbers, equipped with an inner product, and complete in the mathematical sense (every Cauchy sequence converges). For our purposes, the important features are:

• Vectors can be added and multiplied by complex scalars (it is a vector space).
• There is an inner product that lets you compute overlaps, norms, and angles between vectors.
• The scalars are complex numbers, not real numbers. This is essential — quantum mechanics is irreducibly complex. Real-valued wavefunctions exist for special cases (bound states with time-reversal symmetry), but the general theory requires complex amplitudes. The relative phases between components carry physical information: they determine interference patterns, transition probabilities, and measurement outcomes.

For a particle moving in one dimension, $\mathcal{H} = L^2(\mathbb{R})$, the space of square-integrable functions — this is the Hilbert space you have been working in since Chapter 2. For a spin-1/2 particle, $\mathcal{H} = \mathbb{C}^2$, just two complex dimensions. For a particle with both spatial and spin degrees of freedom, $\mathcal{H} = L^2(\mathbb{R}) \otimes \mathbb{C}^2$ (tensor product — see Chapter 11). For two particles, $\mathcal{H} = L^2(\mathbb{R}^3) \otimes L^2(\mathbb{R}^3)$. The dimension of the Hilbert space grows with the complexity of the system, but Dirac notation works for all of these spaces with no modification. This is its power: whether the Hilbert space is 2-dimensional or infinite-dimensional, the rules are the same. You write kets, bras, and brackets in exactly the same way.

The ket: an abstract state vector

In wave mechanics, you write a quantum state as $\psi(x)$. In Dirac notation, you write it as:

$$|\psi\rangle$$

This is called a ket (the right half of the word "bracket"). The ket $|\psi\rangle$ is an abstract vector in a Hilbert space $\mathcal{H}$. It represents the physical state of the system without reference to any particular basis or representation.

What does this mean concretely? It means that $|\psi\rangle$ contains all the information about the quantum state, but does not commit to expressing that information in any particular way. The wave function $\psi(x)$ is what you get when you project the ket onto the position basis — we will make this precise shortly.

Think of it this way: a vector in three-dimensional space, say $\mathbf{v}$, exists independently of any coordinate system. You can express it in Cartesian coordinates as $(v_x, v_y, v_z)$, or in spherical coordinates as $(v_r, v_\theta, v_\phi)$, but the vector itself is the same object regardless of how you describe it. The numbers change when you rotate the axes; the vector does not. The ket $|\psi\rangle$ is the quantum analogue: it is the state itself, and $\psi(x)$, $\phi(p)$, $c_n$ are its components in different coordinate systems (bases).

Kets satisfy the same algebraic rules as vectors:

Superposition (linearity): If $|\psi\rangle$ and $|\phi\rangle$ are kets, then so is

$$\alpha|\psi\rangle + \beta|\phi\rangle$$

for any complex numbers $\alpha, \beta$. This is the mathematical expression of the superposition principle — the most fundamental feature of quantum mechanics.

Zero vector: There exists a ket $|0\rangle$ (not to be confused with the QHO ground state or the vacuum state — context will always make clear which is meant) such that $|\psi\rangle + |0\rangle = |\psi\rangle$.

Scalar multiplication: $\alpha|\psi\rangle$ is a ket for any complex $\alpha$. Note that $|\psi\rangle$ and $\alpha|\psi\rangle$ (for $\alpha \neq 0$) represent the same physical state — the overall phase and magnitude of the ket do not affect measurable quantities. What matters is the ray in Hilbert space, not the vector itself. But for computation, we usually work with normalized kets satisfying $\langle\psi|\psi\rangle = 1$.

The label inside the ket can be anything that identifies the state: $|n\rangle$ for an energy eigenstate, $|\uparrow\rangle$ or $|+\rangle$ for spin-up, $|x\rangle$ for a position eigenstate, $|1, 0, 0\rangle$ for the hydrogen ground state quantum numbers, or even $|\text{alive}\rangle$ for Schrodinger's cat. The label is just a name — the mathematics does not care what you call the state, as long as you are consistent.

The bra: the dual vector

Every ket has a corresponding bra (the left half of "bracket"):

$$|\psi\rangle \longleftrightarrow \langle\psi|$$

The bra $\langle\psi|$ lives in the dual space $\mathcal{H}^*$, which is the space of linear functionals on $\mathcal{H}$. For our purposes, the key operational fact is:

The bra is the Hermitian conjugate of the ket. If you think of $|\psi\rangle$ as a column vector, then $\langle\psi|$ is the corresponding row vector with complex conjugated entries. In a two-dimensional space (like spin-1/2), if $|\psi\rangle = \begin{pmatrix} \alpha \\ \beta \end{pmatrix}$, then $\langle\psi| = \begin{pmatrix} \alpha^* & \beta^* \end{pmatrix}$. In wave mechanics language:

$$|\psi\rangle \leftrightarrow \psi(x) \qquad \Longrightarrow \qquad \langle\psi| \leftrightarrow \psi^*(x)$$

The rules for constructing bras from kets follow from Hermitian conjugation:

1. If $|\psi\rangle \to \langle\psi|$, then $\alpha|\psi\rangle \to \alpha^*\langle\psi|$ (complex conjugate the scalar)
2. If $|\chi\rangle = \alpha|\psi\rangle + \beta|\phi\rangle$, then $\langle\chi| = \alpha^*\langle\psi| + \beta^*\langle\phi|$ (conjugate all scalars and distribute)

These rules are not arbitrary — they follow from the requirement that the inner product (defined next) be positive-definite.

⚠️ Common Pitfall — A very frequent error: students write $\langle\alpha\psi| = \alpha\langle\psi|$. This is wrong! The correct rule is $\langle\alpha\psi| = \alpha^*\langle\psi|$. The scalar gets complex conjugated when you go from ket to bra. This matters whenever $\alpha$ is complex. For example, if $|\psi\rangle = i|1\rangle$, then $\langle\psi| = -i\langle 1|$, not $i\langle 1|$. Forgetting the conjugation will produce wrong inner products, wrong probabilities, and wrong physics. Get this right now and you will avoid countless errors later.

The inner product: bra meets ket

The inner product of two states is written by combining a bra and a ket into a bracket (hence the ingenious name "bra-c-ket"):

$$\langle\phi|\psi\rangle$$

This is a complex number. In wave mechanics, you know this as:

$$\langle\phi|\psi\rangle = \int_{-\infty}^{\infty} \phi^*(x) \psi(x) \, dx$$

The inner product satisfies four fundamental properties:

1. Conjugate symmetry: $\langle\phi|\psi\rangle = \langle\psi|\phi\rangle^*$. If you swap the bra and ket, the result gets complex conjugated. In particular, $\langle\psi|\psi\rangle = \langle\psi|\psi\rangle^*$, so the norm is always real.

2. Linearity in the second argument: $\langle\phi|\alpha\psi_1 + \beta\psi_2\rangle = \alpha\langle\phi|\psi_1\rangle + \beta\langle\phi|\psi_2\rangle$. The inner product distributes linearly over the ket (the second slot).

3. Anti-linearity in the first argument: $\langle\alpha\phi_1 + \beta\phi_2|\psi\rangle = \alpha^*\langle\phi_1|\psi\rangle + \beta^*\langle\phi_2|\psi\rangle$. The bra slot (first argument) is anti-linear — scalars get conjugated. This is a direct consequence of the bra conjugation rule.

4. Positive-definiteness: $\langle\psi|\psi\rangle \geq 0$, with equality if and only if $|\psi\rangle = 0$.

The quantity $\langle\psi|\psi\rangle$ is the squared norm of the state. Normalization means $\langle\psi|\psi\rangle = 1$. Two states are orthogonal if $\langle\phi|\psi\rangle = 0$.

🔗 Connection — Recall from Chapter 2 that the normalization condition for wave functions is $\int |\psi(x)|^2 \, dx = 1$. In Dirac notation, this is simply $\langle\psi|\psi\rangle = 1$. The Born rule tells us that $|\langle\phi|\psi\rangle|^2$ is the probability of finding the system in state $|\phi\rangle$ given that it is in state $|\psi\rangle$. Everything you learned about probability in Chapter 2 carries over unchanged — but the notation is now more general.

The crucial connection: wave function as representation

Here is the key insight that makes everything click. The wave function $\psi(x)$ is the inner product of the position eigenstate $|x\rangle$ with the abstract state $|\psi\rangle$:

$$\boxed{\psi(x) = \langle x|\psi\rangle}$$

Read this carefully. The wave function is not the state — it is the overlap of the state with a particular basis vector, namely the position eigenstate at point $x$. It is the component of $|\psi\rangle$ along the direction $|x\rangle$ in Hilbert space, exactly as $v_x = \hat{x} \cdot \mathbf{v}$ is the component of an ordinary vector along the $x$-axis.

Similarly, the momentum-space wave function is:

$$\phi(p) = \langle p|\psi\rangle$$

And the expansion coefficients in the energy basis are:

$$c_n = \langle n|\psi\rangle$$

The state $|\psi\rangle$ is the same in all three cases. What changes is the basis you project onto. This is exactly like describing the same displacement vector as $(3, 4)$ in one coordinate system and $(5, 0)$ in another. The vector has not changed — only its description has.

🚪 Threshold Concept — This is the central conceptual leap of this chapter: the wave function is a representation, not the state itself. The quantum state is the ket $|\psi\rangle$. The wave function $\psi(x) = \langle x|\psi\rangle$ is what you get when you ask: "What are the components of this state in the position basis?" This is analogous to saying that the components $(v_x, v_y, v_z)$ are not the vector $\mathbf{v}$ — they are the numbers you get when you project $\mathbf{v}$ onto the $\hat{x}$, $\hat{y}$, $\hat{z}$ axes. Different basis, different numbers, same vector. Different representation, different wave function, same quantum state.

Why does this matter? Because spin-1/2 particles have no position-basis representation at all. Their Hilbert space is two-dimensional. There is no wave function $\psi(x)$ for spin — yet they have perfectly well-defined quantum states described by kets. Dirac notation handles these systems without any special treatment. It also makes multi-particle systems natural: the state of two particles is a ket $|\psi\rangle$ in a tensor product space (Chapter 11), rather than a function of two sets of coordinates.

🔄 Check Your Understanding — The state $|\psi\rangle$ of a particle in an infinite square well (Chapter 3) can be expanded as $|\psi\rangle = \sum_n c_n |n\rangle$, where $|n\rangle$ are the energy eigenstates. (a) What is $c_n$ in Dirac notation? (b) What is $c_n$ in wave mechanics notation? (c) How do you verify that $\sum_n |c_n|^2 = 1$? (Answers: (a) $c_n = \langle n|\psi\rangle$; (b) $c_n = \int \psi_n^*(x) \psi(x) \, dx$; (c) From normalization: $\langle\psi|\psi\rangle = 1$, expand using $|\psi\rangle = \sum c_n |n\rangle$ and orthonormality.)

Worked Example 8.1: Inner product both ways

Let us compute the inner product of two infinite-square-well states and verify that the Dirac and wave-mechanics calculations agree.

Problem: Compute $\langle m | n \rangle$ for the infinite square well of width $a$ (Chapter 3).

Wave mechanics:

The energy eigenstates are $\psi_n(x) = \sqrt{2/a} \sin(n\pi x / a)$, so:

$$\langle m | n \rangle = \int_0^a \psi_m^*(x) \psi_n(x) \, dx = \frac{2}{a} \int_0^a \sin\!\left(\frac{m\pi x}{a}\right) \sin\!\left(\frac{n\pi x}{a}\right) dx = \delta_{mn}$$

The last equality uses the orthogonality of sine functions on $[0, a]$, which requires evaluating the integral using the product-to-sum formula.

Dirac notation: The states $|n\rangle$ form an orthonormal basis by construction, so $\langle m | n \rangle = \delta_{mn}$ follows directly from the definition of an orthonormal basis.

The wave mechanics calculation requires evaluating a trigonometric integral. The Dirac calculation requires knowing the basis is orthonormal. Both give the same answer, but the Dirac version makes the underlying structure transparent: orthonormality is a property of the basis, not of any particular integral. The integral is how orthonormality manifests in the position representation; it is not the fundamental statement.

Worked Example 8.2: A spin-1/2 inner product

Problem: A spin-1/2 particle is in state $|\psi\rangle = \frac{3}{5}|\uparrow\rangle + \frac{4i}{5}|\downarrow\rangle$. Compute $\langle\psi|\psi\rangle$ and $|\langle\uparrow|\psi\rangle|^2$.

Solution: First, form the bra: $\langle\psi| = \frac{3}{5}\langle\uparrow| - \frac{4i}{5}\langle\downarrow|$ (note the conjugation of $4i$ to $-4i$).

$$\langle\psi|\psi\rangle = \left(\frac{3}{5}\langle\uparrow| - \frac{4i}{5}\langle\downarrow|\right)\left(\frac{3}{5}|\uparrow\rangle + \frac{4i}{5}|\downarrow\rangle\right)$$

$$= \frac{9}{25}\langle\uparrow|\uparrow\rangle + \frac{12i}{25}\langle\uparrow|\downarrow\rangle - \frac{12i}{25}\langle\downarrow|\uparrow\rangle + \frac{16}{25}\langle\downarrow|\downarrow\rangle$$

Using orthonormality ($\langle\uparrow|\uparrow\rangle = \langle\downarrow|\downarrow\rangle = 1$, $\langle\uparrow|\downarrow\rangle = 0$):

$$= \frac{9}{25} + \frac{16}{25} = 1 \quad \checkmark$$

For the measurement probability: $\langle\uparrow|\psi\rangle = \frac{3}{5}$, so $|\langle\uparrow|\psi\rangle|^2 = \frac{9}{25}$. This is the probability of measuring spin-up along $z$.

Notice that this problem has no wave function at all. There is no $\psi(x)$. The entire calculation is algebraic. This is a purely quantum system that Dirac notation handles effortlessly.

8.3 Bases, Completeness, and Resolution of Identity

Orthonormal bases

An orthonormal basis for the Hilbert space is a set of kets $\{|n\rangle\}$ satisfying two conditions:

Orthonormality: $\langle m | n \rangle = \delta_{mn}$

Completeness: Every ket $|\psi\rangle$ in the Hilbert space can be expanded as a linear combination of the basis kets.

The expansion of an arbitrary state in the basis $\{|n\rangle\}$ is:

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

This should look familiar: it is the Fourier expansion of the wave function, generalized to any basis. The coefficients $c_n = \langle n|\psi\rangle$ are the "components" of $|\psi\rangle$ along the "axes" $|n\rangle$. In matrix language, the ket is a column vector with entries $c_n$:

$$|\psi\rangle \leftrightarrow \begin{pmatrix} c_1 \\ c_2 \\ c_3 \\ \vdots \end{pmatrix}$$

Different bases give different column vectors for the same ket, just as different coordinate systems give different components for the same physical vector. The coefficients depend on the basis; the state does not.

The completeness relation (resolution of the identity)

The completeness of the basis can be expressed as an operator equation:

$$\boxed{\sum_n |n\rangle\langle n| = \hat{I}}$$

where $\hat{I}$ is the identity operator. This is called the completeness relation or resolution of the identity. It is one of the most important equations in all of quantum mechanics. You will use it hundreds of times before this book is over.

To see why it works, apply it to an arbitrary ket:

$$\hat{I}|\psi\rangle = \left(\sum_n |n\rangle\langle n|\right)|\psi\rangle = \sum_n |n\rangle \langle n|\psi\rangle = \sum_n c_n |n\rangle = |\psi\rangle$$

Each term $|n\rangle\langle n|$ is a projection operator (more on these in Section 8.4). The completeness relation says: if you project onto every basis vector and add up the results, you recover the original state. Nothing is lost. The basis is "complete" in the sense that no direction in Hilbert space is missing.

💡 Key Insight — The completeness relation is the single most useful computational tool in Dirac notation. The technique is called inserting a complete set of states or colloquially "inserting a 1." Whenever you need to change basis, evaluate a matrix element, or connect two different representations, you insert $\hat{I} = \sum_n |n\rangle\langle n|$ at the appropriate point in the expression. Since $\hat{I}$ is the identity, this changes nothing mathematically — but it can transform an opaque expression into a computable one. Master this technique and you can do almost anything.

Continuous bases

For continuous variables like position and momentum, the sums become integrals and the Kronecker delta becomes a Dirac delta:

Position basis:

$$\langle x | x' \rangle = \delta(x - x'), \qquad \int_{-\infty}^{\infty} |x\rangle\langle x| \, dx = \hat{I}$$

Momentum basis:

$$\langle p | p' \rangle = \delta(p - p'), \qquad \int_{-\infty}^{\infty} |p\rangle\langle p| \, dp = \hat{I}$$

The position-momentum overlap (which connects the two continuous bases) is:

$$\langle x|p\rangle = \frac{1}{\sqrt{2\pi\hbar}} e^{ipx/\hbar}$$

This single quantity is the kernel of the Fourier transform. We will use it repeatedly.

🔴 Warning — Position eigenkets $|x\rangle$ are not normalizable in the usual sense: $\langle x | x \rangle = \delta(0)$, which is infinite. They are not physical states — no particle can be exactly at a single point with zero uncertainty in position. They are mathematical tools that form a complete basis. This is the domain of rigged Hilbert spaces, which we will discuss in Chapter 9. For now, treat them as limiting cases of narrow wave packets and trust the formalism: it gives correct, finite answers for all physical quantities.

🔄 Spaced Review (Chapter 2) — Recall that $|\psi(x)|^2$ is the probability density for finding the particle at position $x$ (Born rule, Chapter 2). In Dirac notation, this becomes $|\langle x|\psi\rangle|^2$. The Born rule in Dirac notation is: the probability of measuring the eigenvalue $a_n$ of observable $\hat{A}$ is $|\langle a_n | \psi \rangle|^2$. This is more general than the position-space Born rule — it works for any observable, in any basis.

Worked Example 8.3: Recovering the wave function using completeness

Problem: Starting from the abstract state $|\psi\rangle$, derive the position-space expansion $\psi(x) = \sum_n c_n \psi_n(x)$ using the completeness relation.

Solution: We want $\psi(x) = \langle x|\psi\rangle$. Insert the energy-basis completeness relation:

$$\psi(x) = \langle x|\psi\rangle = \langle x| \hat{I} |\psi\rangle = \langle x| \left(\sum_n |n\rangle\langle n|\right) |\psi\rangle = \sum_n \langle x|n\rangle \langle n|\psi\rangle = \sum_n \psi_n(x) \, c_n$$

where we identified $\langle x|n\rangle = \psi_n(x)$ (the energy eigenfunction in position representation) and $\langle n|\psi\rangle = c_n$ (the expansion coefficient). This is the Fourier expansion of the wave function in the energy eigenbasis — exactly what you learned in Chapter 3, but now derived in one line using completeness.

The power of this technique cannot be overstated. We took an abstract equation ($\psi(x) = \langle x|\psi\rangle$) and, by inserting a single completeness relation, derived the concrete expansion that took several pages of motivation in Chapter 3.

Worked Example 8.4: The Fourier transform from completeness

Problem: Derive $\phi(p) = \frac{1}{\sqrt{2\pi\hbar}} \int \psi(x) e^{-ipx/\hbar} dx$ using Dirac notation.

Solution:

$$\phi(p) = \langle p|\psi\rangle = \int \langle p|x\rangle \langle x|\psi\rangle \, dx = \int \langle x|p\rangle^* \, \psi(x) \, dx$$

Using $\langle x|p\rangle = \frac{1}{\sqrt{2\pi\hbar}} e^{ipx/\hbar}$, we have $\langle p|x\rangle = \langle x|p\rangle^* = \frac{1}{\sqrt{2\pi\hbar}} e^{-ipx/\hbar}$:

$$\phi(p) = \frac{1}{\sqrt{2\pi\hbar}} \int \psi(x) \, e^{-ipx/\hbar} \, dx$$

This is the Fourier transform connecting position and momentum representations. In Dirac notation, it is a one-line derivation: insert a complete set of position states and evaluate the overlap $\langle p|x\rangle$. Compare this to the multi-step derivation you likely saw in Chapter 2 or Chapter 5. The Dirac method is not just shorter — it makes the structure visible. The Fourier transform is nothing but a change of basis, mediated by the overlap $\langle x|p\rangle$.

🔄 Check Your Understanding — By inserting the momentum completeness relation $\int |p\rangle\langle p| \, dp = \hat{I}$ into $\psi(x) = \langle x|\psi\rangle$, derive the inverse Fourier transform. You should get $\psi(x) = \frac{1}{\sqrt{2\pi\hbar}} \int \phi(p) e^{ipx/\hbar} dp$.

8.4 Operators in Dirac Notation

Operators acting on kets and bras

An operator $\hat{A}$ maps kets to kets:

$$\hat{A}|\psi\rangle = |\phi\rangle$$

This is the Dirac notation version of "$\hat{A}$ acting on $\psi$ gives $\phi$." In wave mechanics, this might look like $-\frac{\hbar^2}{2m}\frac{d^2\psi}{dx^2} + V(x)\psi(x) = E\psi(x)$. In Dirac notation: $\hat{H}|\psi\rangle = E|\psi\rangle$. Same physics, vastly cleaner notation.

Operators can also act to the left on bras:

$$\langle\phi|\hat{A}$$

This is itself a bra (an element of the dual space). The expression $\langle\phi|\hat{A}|\psi\rangle$ is unambiguous — you can evaluate it either as $(\langle\phi|\hat{A})|\psi\rangle$ or as $\langle\phi|(\hat{A}|\psi\rangle)$, and both give the same complex number. This associativity is built into the notation.

Matrix elements

The matrix element of $\hat{A}$ between states $|\phi\rangle$ and $|\psi\rangle$ is:

$$\langle\phi|\hat{A}|\psi\rangle$$

This is a complex number. It generalizes the wave mechanics integral $\int \phi^*(x) \hat{A}\psi(x) \, dx$ to arbitrary states and operators.

When $|\phi\rangle$ and $|\psi\rangle$ are basis states $|m\rangle$ and $|n\rangle$, the matrix element

$$A_{mn} = \langle m|\hat{A}|n\rangle$$

is literally the $(m,n)$ entry of the matrix representing $\hat{A}$ in the basis $\{|n\rangle\}$. This is the bridge between abstract operators and concrete matrices.

🔄 Spaced Review (Chapter 6) — In Chapter 6, you learned that Hermitian operators ($\hat{A}^\dagger = \hat{A}$) have real eigenvalues and orthogonal eigenstates. In Dirac notation, the eigenvalue equation is $\hat{A}|a_n\rangle = a_n|a_n\rangle$, and Hermiticity means $\langle\phi|\hat{A}|\psi\rangle = \langle\psi|\hat{A}|\phi\rangle^*$ for all states. Compare this with the wave mechanics definition: $\int \phi^* (\hat{A}\psi) \, dx = \left[\int \psi^* (\hat{A}\phi) \, dx\right]^*$. Same content, cleaner notation. The commutator $[\hat{A}, \hat{B}] = \hat{A}\hat{B} - \hat{B}\hat{A}$ has the same form in both notations, but in Dirac notation you can evaluate $\langle m|[\hat{A}, \hat{B}]|n\rangle$ using matrix elements rather than integrals.

Outer products and projection operators

The expression $|n\rangle\langle m|$ is an outer product — an operator, not a number. It takes a ket $|\psi\rangle$ and produces:

$$(|n\rangle\langle m|)|\psi\rangle = |n\rangle \langle m|\psi\rangle = \langle m|\psi\rangle \, |n\rangle$$

It extracts the component of $|\psi\rangle$ along $|m\rangle$ (the scalar $\langle m|\psi\rangle$) and returns a ket proportional to $|n\rangle$. When $n = m$, we get the projection operator:

$$\hat{P}_n = |n\rangle\langle n|$$

This operator projects any state onto $|n\rangle$:

$$\hat{P}_n |\psi\rangle = |n\rangle\langle n|\psi\rangle = c_n |n\rangle$$

Projection operators have two key properties: - Idempotency: $\hat{P}_n^2 = \hat{P}_n$ (projecting twice is the same as projecting once). Proof: $\hat{P}_n^2 = |n\rangle\langle n|n\rangle\langle n| = |n\rangle \cdot 1 \cdot \langle n| = \hat{P}_n$. - Hermiticity: $\hat{P}_n^\dagger = \hat{P}_n$ (projection operators are Hermitian, so they are observables).

The completeness relation is $\sum_n \hat{P}_n = \hat{I}$ — the projections onto all basis states add up to the identity.

💡 Key Insight — Any Hermitian operator can be written using outer products. If $\hat{A}$ has eigenvalues $a_n$ and eigenstates $|a_n\rangle$, then:

$$\hat{A} = \sum_n a_n |a_n\rangle\langle a_n|$$

This is the spectral decomposition of $\hat{A}$. It says: to apply $\hat{A}$ to a state, project onto each eigenstate, multiply each component by the corresponding eigenvalue, and add up the results. Every Hermitian operator is completely determined by its eigenvalues and eigenstates. This is profoundly useful and will be explored further in Chapter 9.

The Hermitian conjugate (adjoint)

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

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

for all $|\phi\rangle$ and $|\psi\rangle$. This is the operator analogue of complex conjugation for numbers.

Rules for Hermitian conjugation (memorize these — you will use them constantly): - $(|\psi\rangle)^\dagger = \langle\psi|$ and $(\langle\psi|)^\dagger = |\psi\rangle$ - $(\hat{A}\hat{B})^\dagger = \hat{B}^\dagger\hat{A}^\dagger$ (reverse the order) - $(\hat{A}|\psi\rangle)^\dagger = \langle\psi|\hat{A}^\dagger$ - $(\alpha\hat{A})^\dagger = \alpha^*\hat{A}^\dagger$ - $(|n\rangle\langle m|)^\dagger = |m\rangle\langle n|$

The general rule is: to take the Hermitian conjugate of any expression, reverse the order of everything and conjugate all scalars and operators.

An operator is Hermitian (or self-adjoint) if $\hat{A}^\dagger = \hat{A}$. Observables in quantum mechanics are represented by Hermitian operators. An operator is unitary if $\hat{U}^\dagger\hat{U} = \hat{I}$. Time evolution and basis changes are represented by unitary operators.

⚠️ Common Pitfall — The order-reversal rule $(\hat{A}\hat{B})^\dagger = \hat{B}^\dagger \hat{A}^\dagger$ catches many students off guard. Think of it this way: if you put on socks then shoes in the morning, you take off shoes then socks in the evening. Hermitian conjugation reverses the order of operations. So $(\alpha \hat{A}|\psi\rangle \langle\phi|\hat{B})^\dagger = \hat{B}^\dagger |\phi\rangle \langle\psi| \hat{A}^\dagger \alpha^*$.

Expectation values

The expectation value of $\hat{A}$ in state $|\psi\rangle$ is:

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

In wave mechanics: $\langle\hat{A}\rangle = \int \psi^*(x) \hat{A} \psi(x) \, dx$. The Dirac form is cleaner and basis-independent. If $\hat{A}$ is Hermitian, $\langle\hat{A}\rangle$ is guaranteed to be real — a necessary condition for a measurable quantity.

🔄 Spaced Review (Chapter 4) — Recall the harmonic oscillator ladder operators from Chapter 4: $\hat{a}|n\rangle = \sqrt{n}|n-1\rangle$ and $\hat{a}^\dagger|n\rangle = \sqrt{n+1}|n+1\rangle$. Using Dirac notation, we can instantly compute matrix elements: $\langle m|\hat{a}|n\rangle = \sqrt{n}\,\delta_{m,n-1}$ and $\langle m|\hat{a}^\dagger|n\rangle = \sqrt{n+1}\,\delta_{m,n+1}$. These are the matrix elements of the ladder operators in the energy basis — they give matrices with entries only on one off-diagonal. The algebraic approach you learned in Chapter 4 is Dirac notation. You were using it all along; now you have the full framework.

🔄 Check Your Understanding — Write the Hamiltonian of the QHO, $\hat{H} = \hbar\omega(\hat{a}^\dagger\hat{a} + \frac{1}{2})$, in outer-product form using the energy eigenstates. (Answer: $\hat{H} = \sum_{n=0}^{\infty} \hbar\omega(n + \frac{1}{2})|n\rangle\langle n|$.)

8.5 Matrix Representations

From operators to matrices

Given an orthonormal basis $\{|n\rangle\}$, every operator $\hat{A}$ can be represented as a matrix with entries:

$$A_{mn} = \langle m|\hat{A}|n\rangle$$

To see why this works, apply $\hat{A}$ to an arbitrary state and take components in the basis:

$$\hat{A}|\psi\rangle = \hat{A}\left(\sum_n c_n |n\rangle\right) = \sum_n c_n \hat{A}|n\rangle$$

The component of the result along $|m\rangle$ is:

$$\langle m|\hat{A}|\psi\rangle = \sum_n \langle m|\hat{A}|n\rangle \, c_n = \sum_n A_{mn} \, c_n$$

This is matrix multiplication: $b_m = \sum_n A_{mn} c_n$, where $b_m$ are the components of $\hat{A}|\psi\rangle$ and $c_n$ are the components of $|\psi\rangle$. The abstract operator equation $\hat{A}|\psi\rangle = |\phi\rangle$ becomes the matrix equation $\mathbf{A}\mathbf{c} = \mathbf{b}$. Dirac notation unifies operator algebra and matrix algebra.

Example: Spin-1/2 matrices

The spin-1/2 system has a two-dimensional Hilbert space spanned by $\{|\uparrow\rangle, |\downarrow\rangle\}$. We choose the basis where $\hat{S}_z$ is diagonal:

$$\hat{S}_z|\uparrow\rangle = +\frac{\hbar}{2}|\uparrow\rangle, \qquad \hat{S}_z|\downarrow\rangle = -\frac{\hbar}{2}|\downarrow\rangle$$

The matrix representation of $\hat{S}_z$ is:

$$[S_z] = \begin{pmatrix} \langle\uparrow|\hat{S}_z|\uparrow\rangle & \langle\uparrow|\hat{S}_z|\downarrow\rangle \\ \langle\downarrow|\hat{S}_z|\uparrow\rangle & \langle\downarrow|\hat{S}_z|\downarrow\rangle \end{pmatrix} = \frac{\hbar}{2}\begin{pmatrix} 1 & 0 \\ 0 & -1 \end{pmatrix} = \frac{\hbar}{2}\sigma_z$$

where $\sigma_z$ is the Pauli $z$-matrix. Similarly:

$$[S_x] = \frac{\hbar}{2}\begin{pmatrix} 0 & 1 \\ 1 & 0 \end{pmatrix} = \frac{\hbar}{2}\sigma_x, \qquad [S_y] = \frac{\hbar}{2}\begin{pmatrix} 0 & -i \\ i & 0 \end{pmatrix} = \frac{\hbar}{2}\sigma_y$$

The spin-1/2 system is the paradigmatic example of a quantum system with no wave-function representation. There is no $\psi(x)$ for spin. The state is a two-component vector (spinor):

$$|\psi\rangle = \alpha|\uparrow\rangle + \beta|\downarrow\rangle \longleftrightarrow \begin{pmatrix} \alpha \\ \beta \end{pmatrix}$$

where $|\alpha|^2 + |\beta|^2 = 1$. This system will be studied in full detail in Chapter 13, but you can already see why Dirac notation is essential: it gives us a framework that handles both infinite-dimensional systems (particle in a well) and finite-dimensional ones (spin) with the same tools.

Example: QHO ladder operators as matrices

For the quantum harmonic oscillator in the energy basis $\{|0\rangle, |1\rangle, |2\rangle, \ldots\}$, the matrix elements of the lowering operator are $\langle m|\hat{a}|n\rangle = \sqrt{n}\,\delta_{m,n-1}$. The matrix is:

$$[\hat{a}] = \begin{pmatrix} 0 & 1 & 0 & 0 & \cdots \\ 0 & 0 & \sqrt{2} & 0 & \cdots \\ 0 & 0 & 0 & \sqrt{3} & \cdots \\ 0 & 0 & 0 & 0 & \cdots \\ \vdots & & & & \ddots \end{pmatrix}, \quad [\hat{a}^\dagger] = \begin{pmatrix} 0 & 0 & 0 & 0 & \cdots \\ 1 & 0 & 0 & 0 & \cdots \\ 0 & \sqrt{2} & 0 & 0 & \cdots \\ 0 & 0 & \sqrt{3} & 0 & \cdots \\ \vdots & & & & \ddots \end{pmatrix}$$

The number operator $\hat{n} = \hat{a}^\dagger\hat{a}$ has matrix $[\hat{n}] = \text{diag}(0, 1, 2, 3, \ldots)$, and the Hamiltonian $\hat{H} = \hbar\omega(\hat{n} + \frac{1}{2})$ is diagonal with entries $E_n = \hbar\omega(n + \frac{1}{2})$. The matrix representation makes the algebraic structure from Chapter 4 completely explicit.

🔄 Check Your Understanding — Using the matrix representations above, verify that $[\hat{a}, \hat{a}^\dagger] = \hat{I}$ by computing $[\hat{a}][\hat{a}^\dagger] - [\hat{a}^\dagger][\hat{a}]$ for the upper-left $3\times 3$ block. Do you get the identity matrix? (You should. The bottom-right corner of a truncated matrix will show artifacts of the truncation, but the interior is exact.)

Building a matrix from scratch: the position operator in the QHO energy basis

To solidify the procedure, let us build one more matrix representation step by step. We want the position operator $\hat{x}$ in the QHO energy basis. The matrix elements are $x_{mn} = \langle m|\hat{x}|n\rangle$.

We could try to evaluate these as integrals $\int \psi_m^*(x)\, x\, \psi_n(x)\, dx$, which involves products of Hermite polynomials — a significant computation. Instead, we use the algebraic identity $\hat{x} = \sqrt{\frac{\hbar}{2m\omega}}(\hat{a} + \hat{a}^\dagger)$ and compute:

$$x_{mn} = \sqrt{\frac{\hbar}{2m\omega}}\left(\langle m|\hat{a}|n\rangle + \langle m|\hat{a}^\dagger|n\rangle\right) = \sqrt{\frac{\hbar}{2m\omega}}\left(\sqrt{n}\,\delta_{m,n-1} + \sqrt{n+1}\,\delta_{m,n+1}\right)$$

The matrix is tridiagonal, with nonzero entries only on the super-diagonal and sub-diagonal:

$$[x] = \sqrt{\frac{\hbar}{2m\omega}}\begin{pmatrix} 0 & 1 & 0 & 0 & \cdots \\ 1 & 0 & \sqrt{2} & 0 & \cdots \\ 0 & \sqrt{2} & 0 & \sqrt{3} & \cdots \\ 0 & 0 & \sqrt{3} & 0 & \cdots \\ \vdots & & & & \ddots \end{pmatrix}$$

This matrix is real and symmetric (Hermitian), as it must be for an observable. Its eigenvalues are the possible outcomes of a position measurement — but in the truncated $N\times N$ version, these are only approximations. The exact position eigenvalues form a continuous spectrum (all of $\mathbb{R}$), which requires an infinite-dimensional matrix. This is one of the subtleties of continuous spectra that we will address properly in Chapter 9.

The key lesson: the ladder-operator method (pure Dirac notation algebra) avoids all integrals and produces exact matrix elements. This is a pattern you will see again and again in this book: Dirac notation turns integration problems into algebraic problems.

8.6 Change of Basis and Unitary Transformations

Why change basis?

Different bases are useful for different purposes. The energy basis $\{|n\rangle\}$ is natural for time evolution (since energy eigenstates evolve as $e^{-iE_nt/\hbar}|n\rangle$). The position basis $\{|x\rangle\}$ is natural for spatial probability distributions. The momentum basis $\{|p\rangle\}$ is natural for scattering problems. The ability to switch between bases fluently is essential for any working physicist.

Unitary operators

A unitary operator $\hat{U}$ satisfies:

$$\hat{U}^\dagger \hat{U} = \hat{U}\hat{U}^\dagger = \hat{I}$$

This means $\hat{U}^{-1} = \hat{U}^\dagger$, so inversion is trivial. Unitary operators preserve inner products:

$$\langle\phi'|\psi'\rangle = \langle\phi|\hat{U}^\dagger\hat{U}|\psi\rangle = \langle\phi|\psi\rangle$$

This means unitary transformations preserve probabilities, norms, and all physically meaningful quantities. They are the "rotations" of Hilbert space — they change the description of the state without changing any measurable property.

Basis change as a unitary transformation

Suppose we have two orthonormal bases, $\{|n\rangle\}$ and $\{|n'\rangle\}$. The unitary operator connecting them is:

$$\hat{U} = \sum_n |n'\rangle\langle n|$$

It maps $|n\rangle \to |n'\rangle$ for each $n$. The matrix elements of $\hat{U}$ in the old basis are $U_{mn} = \langle m'|n\rangle$, the overlaps between the old and new basis vectors.

Under a change of basis, an operator $\hat{A}$ transforms as:

$$\hat{A}' = \hat{U}^\dagger \hat{A} \hat{U}$$

and a state transforms as $|\psi\rangle' = \hat{U}^\dagger|\psi\rangle$. The matrix representation changes, but the operator and state themselves do not.

Worked Example 8.5: Position to momentum (Fourier transform as basis change)

Problem: Show that the Fourier transform connecting $\psi(x) = \langle x|\psi\rangle$ and $\phi(p) = \langle p|\psi\rangle$ is a unitary change of basis.

Solution: The "transformation matrix" element is:

$$\langle p|x\rangle = \frac{1}{\sqrt{2\pi\hbar}} e^{-ipx/\hbar}$$

The transformation from position to momentum representation is:

$$\phi(p) = \langle p|\psi\rangle = \int \langle p|x\rangle \langle x|\psi\rangle \, dx = \frac{1}{\sqrt{2\pi\hbar}} \int e^{-ipx/\hbar} \psi(x) \, dx$$

This is the Fourier transform — and it is a unitary transformation. That is, $\int |\phi(p)|^2 dp = \int |\psi(x)|^2 dx = 1$ (Parseval's theorem). The deep lesson: the Fourier transform is not just a mathematical trick — it is a change of basis in Hilbert space.

The importance of unitarity

Why do we insist that basis changes be unitary? Because physics does not depend on how we choose to describe it. If I compute the probability of a measurement outcome in one basis and you compute it in another, we must get the same answer. Unitary transformations guarantee this: they preserve $|\langle\phi|\psi\rangle|^2$, which is the fundamental physical quantity (Born rule). A transformation that changed inner products would change probabilities, and different observers would predict different experimental outcomes. Unitarity is not a mathematical nicety — it is the consistency condition that makes quantum mechanics a physical theory.

The two most important classes of unitary operators in quantum mechanics are: - Basis changes (this section): rotating the "coordinate axes" in Hilbert space. - Time evolution (Chapter 7): $\hat{U}(t) = e^{-i\hat{H}t/\hbar}$, which is unitary because the Hamiltonian is Hermitian.

Both preserve the physics. Basis changes correspond to looking at the same state from different perspectives. Time evolution corresponds to the state physically evolving. Dirac notation treats both with the same mathematical machinery.

Worked Example 8.6: Spin-1/2 basis change

Problem: A spin-1/2 particle is in the state $|\uparrow\rangle$ (spin-up along $z$). What is its representation in the $S_x$ eigenbasis?

Solution: The eigenstates of $\hat{S}_x$ are:

$$|+x\rangle = \frac{1}{\sqrt{2}}(|\uparrow\rangle + |\downarrow\rangle), \qquad |-x\rangle = \frac{1}{\sqrt{2}}(|\uparrow\rangle - |\downarrow\rangle)$$

We can invert these to express the $S_z$ eigenstates in the $S_x$ basis. Adding: $|+x\rangle + |-x\rangle = \sqrt{2}|\uparrow\rangle$, so $|\uparrow\rangle = \frac{1}{\sqrt{2}}(|+x\rangle + |-x\rangle)$. The components of $|\uparrow\rangle$ in the $S_x$ basis are:

$$\langle +x|\uparrow\rangle = \frac{1}{\sqrt{2}}, \qquad \langle -x|\uparrow\rangle = \frac{1}{\sqrt{2}}$$

So $|\uparrow\rangle = \frac{1}{\sqrt{2}}|+x\rangle + \frac{1}{\sqrt{2}}|-x\rangle$. In the $S_x$ basis, the state vector is $\begin{pmatrix} 1/\sqrt{2} \\ 1/\sqrt{2} \end{pmatrix}$.

The unitary matrix connecting the $S_z$ and $S_x$ bases is:

$$[U] = \begin{pmatrix} \langle +x|\uparrow\rangle & \langle +x|\downarrow\rangle \\ \langle -x|\uparrow\rangle & \langle -x|\downarrow\rangle \end{pmatrix} = \frac{1}{\sqrt{2}}\begin{pmatrix} 1 & 1 \\ 1 & -1 \end{pmatrix}$$

You can verify: $[U][U]^\dagger = [I]$, confirming unitarity. This matrix is its own inverse (it is a Hadamard matrix — you will encounter it again in Chapter 25 as the Hadamard gate, one of the fundamental operations in quantum computing).

💡 Key Insight — The spin-up state along $z$ is an equal superposition of spin-up and spin-down along $x$. This means that measuring $S_x$ on a particle known to be spin-up along $z$ gives $+\hbar/2$ or $-\hbar/2$ with equal probability (50/50). Dirac notation makes this completely transparent: $P(+x) = |\langle +x|\uparrow\rangle|^2 = 1/2$. This is the uncertainty principle at work — complementary observables ($S_z$ and $S_x$) cannot both be definite simultaneously. A state that is sharp in $S_z$ must be maximally uncertain in $S_x$.

🔄 Check Your Understanding — Compute the representation of $\hat{S}_z$ in the $S_x$ eigenbasis using the transformation $[S_z]' = [U]^\dagger [S_z] [U]$. What do you get? Is the matrix still diagonal? (It should not be — $\hat{S}_z$ is not diagonal in the $S_x$ basis because the $S_x$ eigenstates are not eigenstates of $\hat{S}_z$. You should find that $[S_z]'$ looks like the matrix for $[S_x]$ in the $S_z$ basis. Can you explain why?)

8.7 Functions of Operators and the Trace

Functions of operators

Given a Hermitian operator $\hat{A}$ with spectral decomposition $\hat{A} = \sum_n a_n |a_n\rangle\langle a_n|$, we define a function of $\hat{A}$ by:

$$f(\hat{A}) = \sum_n f(a_n) |a_n\rangle\langle a_n|$$

This is the natural and unique definition: $f(\hat{A})$ acts on each eigenvector by replacing the eigenvalue $a_n$ with $f(a_n)$, and leaves the eigenvectors unchanged.

Example: The time-evolution operator. From Chapter 7, $\hat{U}(t) = e^{-i\hat{H}t/\hbar}$. Using the spectral decomposition of $\hat{H} = \sum_n E_n |n\rangle\langle n|$:

$$\hat{U}(t) = \sum_n e^{-iE_nt/\hbar} |n\rangle\langle n|$$

This is manifestly unitary: $\hat{U}^\dagger(t)\hat{U}(t) = \sum_n e^{+iE_nt/\hbar}e^{-iE_nt/\hbar} |n\rangle\langle n| = \sum_n |n\rangle\langle n| = \hat{I}$.

Applied to a state: $\hat{U}(t)|\psi(0)\rangle = \sum_n e^{-iE_nt/\hbar} c_n |n\rangle$, which is exactly the time-evolution formula from Chapter 7. Now you see where it comes from: it is the exponential function applied to the Hamiltonian via spectral decomposition.

Example: The inverse of a non-singular operator is $\hat{A}^{-1} = \sum_n a_n^{-1} |a_n\rangle\langle a_n|$ (provided no $a_n = 0$).

Example: The square root: $\hat{A}^{1/2} = \sum_n \sqrt{a_n} |a_n\rangle\langle a_n|$ (provided all $a_n \geq 0$).

🔗 Connection — In Chapter 7, we introduced the time-evolution operator $\hat{U}(t) = e^{-i\hat{H}t/\hbar}$ and used it to evolve quantum states. Now you see the full machinery: it is the exponential function applied to the Hamiltonian via spectral decomposition. Every energy eigenstate picks up a phase $e^{-iE_nt/\hbar}$, and the time evolution of an arbitrary state is a superposition of these rotating phases. This is precisely what we derived in Chapter 7 using wave mechanics — but Dirac notation makes the structure crystalline. The spectral decomposition is the reason we care about finding energy eigenstates: they are the "natural modes" of the time-evolution operator.

The trace

The trace of an operator is the sum of its diagonal matrix elements in any orthonormal basis:

$$\text{Tr}(\hat{A}) = \sum_n \langle n|\hat{A}|n\rangle$$

The trace has several remarkable and useful properties:

1. Basis-independent: $\sum_n \langle n|\hat{A}|n\rangle = \sum_m \langle m'|\hat{A}|m'\rangle$ for any two orthonormal bases. (We prove this below.)

2. Cyclic property: $\text{Tr}(\hat{A}\hat{B}\hat{C}) = \text{Tr}(\hat{B}\hat{C}\hat{A}) = \text{Tr}(\hat{C}\hat{A}\hat{B})$. You can cycle the operators under the trace. But you CANNOT arbitrarily permute them: $\text{Tr}(\hat{A}\hat{B}\hat{C}) \neq \text{Tr}(\hat{A}\hat{C}\hat{B})$ in general.

3. Invariance under unitary transformations: $\text{Tr}(\hat{U}^\dagger \hat{A} \hat{U}) = \text{Tr}(\hat{A})$.

4. Sum of eigenvalues: $\text{Tr}(\hat{A}) = \sum_n a_n$.

5. Linearity: $\text{Tr}(\alpha\hat{A} + \beta\hat{B}) = \alpha\,\text{Tr}(\hat{A}) + \beta\,\text{Tr}(\hat{B})$.

Proof of basis-independence (an elegant use of completeness):

$$\sum_n \langle n|\hat{A}|n\rangle = \sum_n \sum_m \langle n|m'\rangle\langle m'|\hat{A}|n\rangle$$

where we inserted $\hat{I} = \sum_m |m'\rangle\langle m'|$. Rearranging the sums and using $\sum_n |n\rangle\langle n| = \hat{I}$:

$$= \sum_m \langle m'|\hat{A}\left(\sum_n |n\rangle\langle n|\right)|m'\rangle = \sum_m \langle m'|\hat{A}|m'\rangle$$

This proof is a masterclass in completeness insertion. It uses the technique twice — once to introduce the new basis, and once to collapse the sum over the old basis. Read it again slowly and make sure you follow each step.

Why does the trace matter? The trace will become essential in Chapter 23 (density matrix formalism), where the expectation value of an observable in a mixed state is $\langle \hat{A}\rangle = \text{Tr}(\hat{\rho}\hat{A})$. It is also central to quantum information theory (Chapter 25), where the partial trace defines reduced density matrices for subsystems, and to decoherence theory (Chapter 33). Think of the trace as the quantum generalization of "averaging over all states."

🔄 Check Your Understanding — Compute $\text{Tr}(\hat{S}_z)$ for spin-1/2 using the matrix representation. Then compute it using the sum-of-eigenvalues formula. Do they agree? (Answer: $\text{Tr}(\hat{S}_z) = \frac{\hbar}{2} + (-\frac{\hbar}{2}) = 0$, consistent with the matrix trace.)

🔄 Check Your Understanding — Show that $\text{Tr}(|n\rangle\langle m|) = \delta_{mn}$. (Hint: $\text{Tr}(|n\rangle\langle m|) = \sum_k \langle k|n\rangle\langle m|k\rangle = \sum_k \delta_{kn}\delta_{mk} = \delta_{mn}$.)

Worked Example: Expectation value via the trace

A useful identity connects expectation values to the trace. For a pure state $|\psi\rangle$, define the density operator (or density matrix, fully treated in Chapter 23):

$$\hat{\rho} = |\psi\rangle\langle\psi|$$

Then the expectation value of any operator $\hat{A}$ can be written:

$$\langle\hat{A}\rangle = \langle\psi|\hat{A}|\psi\rangle = \text{Tr}(\hat{\rho}\hat{A})$$

Proof: $\text{Tr}(\hat{\rho}\hat{A}) = \text{Tr}(|\psi\rangle\langle\psi|\hat{A}) = \sum_n \langle n|\psi\rangle\langle\psi|\hat{A}|n\rangle = \langle\psi|\hat{A}\left(\sum_n |n\rangle\langle n|\right)|\psi\rangle^{\!\!\!*}$

Wait — let us be more careful. Using the cyclic property:

$$\text{Tr}(|\psi\rangle\langle\psi|\hat{A}) = \text{Tr}(\hat{A}|\psi\rangle\langle\psi|) = \sum_n \langle n|\hat{A}|\psi\rangle\langle\psi|n\rangle = \sum_n \langle\psi|n\rangle^*\langle n|\hat{A}|\psi\rangle$$

Insert completeness: $= \langle\psi|\hat{A}|\psi\rangle$. $\checkmark$

This identity will become the foundation of the density matrix formalism in Chapter 23, where it generalizes naturally to mixed states — statistical mixtures that cannot be described by any single ket.

8.8 Translating Between Wave Mechanics and Dirac: The Rosetta Stone

This section is the culmination of the chapter. We present a complete translation between the wave mechanics notation you learned in Chapters 2--7 and the Dirac notation that will be used for the remainder of this book. Study this table until you can read it in both directions without hesitation.

The Complete Rosetta Stone

Note how the Dirac column is systematically shorter and more general. The wave mechanics expressions all implicitly assume a position-space representation. The Dirac expressions are representation-independent and apply to any quantum system.

Study this table until you can translate in both directions without thinking. Here are some practice translations to get you started:

Quick Translation Drill:

• "The state is normalized" $\to$ $\langle\psi|\psi\rangle = 1$
• $\langle\psi|\hat{H}|\psi\rangle$ $\to$ "the expectation value of energy" $\to$ $\int \psi^* \hat{H}\psi \, dx$ (in position space)
• "The probability of measuring $E_n$" $\to$ $|\langle n|\psi\rangle|^2$ $\to$ $\left|\int \psi_n^*\psi \, dx\right|^2$
• "Expand $|\psi\rangle$ in the energy basis" $\to$ $|\psi\rangle = \sum_n \langle n|\psi\rangle |n\rangle$ $\to$ $\psi(x) = \sum_n c_n \psi_n(x)$

The two most important lines in the table are arguably the completeness relation and the wave-function equation. The completeness relation $\sum_n |n\rangle\langle n| = \hat{I}$ is the computational engine. The wave-function equation $\psi(x) = \langle x|\psi\rangle$ is the conceptual engine — it tells you that the wave function is not the state, but a representation of the state. Every other entry in the table follows from these two, combined with the basic rules of bra-ket algebra.

🔄 Check Your Understanding — Translate the following Dirac expression into wave mechanics: $\langle\phi|\hat{x}^2|\psi\rangle$. (Answer: Insert position completeness to get $\int \phi^*(x) \, x^2 \, \psi(x) \, dx$. The $\hat{x}^2$ becomes multiplication by $x^2$ in the position representation.)

Worked Example 8.7: Translating an expectation value step by step

Problem: Translate the wave mechanics calculation $\langle\hat{p}\rangle = \int \psi^*(x) \left(-i\hbar \frac{d}{dx}\right) \psi(x) \, dx$ into Dirac notation and back.

Forward (wave mechanics to Dirac):

The wave mechanics expression is $\int \psi^*(x) \hat{p} \psi(x) \, dx$. Identifying $\psi^*(x) = \langle\psi|x\rangle$ and $\psi(x) = \langle x|\psi\rangle$, and recognizing the integral as the position-space evaluation of $\langle\psi|\hat{p}|\psi\rangle$:

$$\int \psi^*(x) \hat{p} \psi(x) \, dx = \langle\psi|\hat{p}|\psi\rangle$$

That is the Dirac form: $\langle\hat{p}\rangle = \langle\psi|\hat{p}|\psi\rangle$.

Backward (Dirac to wave mechanics):

Start with $\langle\hat{p}\rangle = \langle\psi|\hat{p}|\psi\rangle$. Insert the position completeness relation twice:

$$\langle\psi|\hat{p}|\psi\rangle = \int\!\!\int \langle\psi|x\rangle \langle x|\hat{p}|x'\rangle \langle x'|\psi\rangle \, dx \, dx'$$

Now, $\langle\psi|x\rangle = \psi^*(x)$, $\langle x'|\psi\rangle = \psi(x')$, and $\langle x|\hat{p}|x'\rangle = -i\hbar \frac{\partial}{\partial x}\delta(x - x')$.

Substituting and integrating over $x'$ (the delta function collapses the integral):

$$= \int \psi^*(x) \left(-i\hbar \frac{d}{dx}\right) \psi(x) \, dx$$

The round trip is complete. The Dirac expression $\langle\psi|\hat{p}|\psi\rangle$ and the wave mechanics integral $\int \psi^* \hat{p} \psi \, dx$ are the same quantity expressed in different languages.

Worked Example 8.8: QHO expectation value both ways

Problem: Compute $\langle n|\hat{x}|n\rangle$ for the quantum harmonic oscillator.

Method 1 — Wave mechanics:

$$\langle n|\hat{x}|n\rangle = \int_{-\infty}^{\infty} \psi_n^*(x) \, x \, \psi_n(x) \, dx = \int_{-\infty}^{\infty} x\,|\psi_n(x)|^2 \, dx$$

The integrand is $x|\psi_n(x)|^2$. Since $|\psi_n(x)|^2$ is an even function of $x$ (for any $n$, the QHO probability density is symmetric about the origin), and $x$ is odd, the integrand is odd. The integral of an odd function over a symmetric interval vanishes:

$$\langle n|\hat{x}|n\rangle = 0$$

Method 2 — Dirac notation with ladder operators:

From Chapter 4, $\hat{x} = \sqrt{\frac{\hbar}{2m\omega}}(\hat{a} + \hat{a}^\dagger)$. Therefore:

$$\langle n|\hat{x}|n\rangle = \sqrt{\frac{\hbar}{2m\omega}} \langle n|(\hat{a} + \hat{a}^\dagger)|n\rangle = \sqrt{\frac{\hbar}{2m\omega}} \left(\langle n|\hat{a}|n\rangle + \langle n|\hat{a}^\dagger|n\rangle\right)$$

Now, $\hat{a}|n\rangle = \sqrt{n}|n-1\rangle$ and $\hat{a}^\dagger|n\rangle = \sqrt{n+1}|n+1\rangle$, so:

$$\langle n|\hat{a}|n\rangle = \sqrt{n}\langle n|n-1\rangle = 0, \qquad \langle n|\hat{a}^\dagger|n\rangle = \sqrt{n+1}\langle n|n+1\rangle = 0$$

Both terms vanish by orthogonality. Result: $\langle n|\hat{x}|n\rangle = 0$.

Both methods give the same answer. The Dirac method avoids integration entirely — it uses the algebraic properties of the ladder operators and the orthonormality of the basis. For more complicated calculations (e.g., $\langle m|\hat{x}^2|n\rangle$ or transition matrix elements $\langle m|\hat{x}|n\rangle$ with $m \neq n$), the algebraic method is far more efficient.

Worked Example 8.9: Transition matrix element

Problem: Compute $\langle n+1|\hat{x}|n\rangle$ for the QHO.

Dirac method:

$$\langle n+1|\hat{x}|n\rangle = \sqrt{\frac{\hbar}{2m\omega}} \langle n+1|(\hat{a} + \hat{a}^\dagger)|n\rangle$$

$$= \sqrt{\frac{\hbar}{2m\omega}} \left(\sqrt{n}\langle n+1|n-1\rangle + \sqrt{n+1}\langle n+1|n+1\rangle\right)$$

$$= \sqrt{\frac{\hbar}{2m\omega}} \cdot \sqrt{n+1}$$

This transition matrix element determines the strength of electric dipole transitions between adjacent QHO levels — a result central to molecular spectroscopy and quantum optics (Chapter 27). The selection rule $\Delta n = \pm 1$ is immediately visible: $\langle m|\hat{x}|n\rangle$ vanishes unless $m = n \pm 1$, because the position operator (being a sum of $\hat{a}$ and $\hat{a}^\dagger$) can only raise or lower the quantum number by one. In wave mechanics, deriving this requires evaluating the integral $\int \psi_m^*(x) \, x \, \psi_n(x) \, dx$ using Hermite polynomial recurrence relations — a far more tedious route to the same result.

🔄 Check Your Understanding — Use the ladder-operator technique to compute $\langle n|\hat{x}^2|n\rangle$. (Hint: expand $\hat{x}^2 = \frac{\hbar}{2m\omega}(\hat{a} + \hat{a}^\dagger)^2 = \frac{\hbar}{2m\omega}(\hat{a}^2 + \hat{a}\hat{a}^\dagger + \hat{a}^\dagger\hat{a} + (\hat{a}^\dagger)^2)$ and use the ladder-operator rules. Only terms that return you to $|n\rangle$ survive. You should get $\langle n|\hat{x}^2|n\rangle = \frac{\hbar}{2m\omega}(2n+1)$, which connects to the zero-point energy and the uncertainty relation.)

Worked Example 8.10: A complete two-level system calculation

Problem: A two-level atom has states $|g\rangle$ (ground) and $|e\rangle$ (excited) with energies $E_g = 0$ and $E_e = \hbar\omega_0$. An electric field couples the levels with matrix element $V = \langle g|\hat{V}|e\rangle = \langle e|\hat{V}|g\rangle^* = \hbar\Omega/2$ (here $\Omega$ is the Rabi frequency, and we assume $V$ is real). Find the eigenstates and eigenvalues of the total Hamiltonian $\hat{H} = \hat{H}_0 + \hat{V}$.

Solution: In the $\{|g\rangle, |e\rangle\}$ basis, the unperturbed Hamiltonian is:

$$[H_0] = \begin{pmatrix} 0 & 0 \\ 0 & \hbar\omega_0 \end{pmatrix}$$

The coupling operator is:

$$[V] = \frac{\hbar\Omega}{2}\begin{pmatrix} 0 & 1 \\ 1 & 0 \end{pmatrix}$$

The total Hamiltonian is:

$$[H] = \begin{pmatrix} 0 & \hbar\Omega/2 \\ \hbar\Omega/2 & \hbar\omega_0 \end{pmatrix}$$

To find the eigenvalues, solve $\det([H] - E[I]) = 0$:

$$E^2 - \hbar\omega_0 E - (\hbar\Omega/2)^2 = 0$$

$$E_\pm = \frac{\hbar\omega_0}{2} \pm \frac{\hbar}{2}\sqrt{\omega_0^2 + \Omega^2}$$

The two eigenvalues are split by $\Delta E = \hbar\sqrt{\omega_0^2 + \Omega^2}$, which is always larger than the unperturbed splitting $\hbar\omega_0$. This is the avoided crossing or level repulsion — a fundamental phenomenon in quantum mechanics that we will study systematically in Chapter 18 (degenerate perturbation theory).

The eigenstates (called "dressed states" in quantum optics) are:

$$|+\rangle = \cos\theta|g\rangle + \sin\theta|e\rangle, \qquad |-\rangle = -\sin\theta|g\rangle + \cos\theta|e\rangle$$

where $\tan(2\theta) = \Omega/\omega_0$. On resonance ($\omega_0 = 0$), $\theta = \pi/4$ and the dressed states are equal superpositions of $|g\rangle$ and $|e\rangle$.

In Dirac notation, the spectral decomposition of $\hat{H}$ is:

$$\hat{H} = E_+|+\rangle\langle +| + E_-|-\rangle\langle -|$$

The time-evolution operator is:

$$\hat{U}(t) = e^{-iE_+t/\hbar}|+\rangle\langle +| + e^{-iE_-t/\hbar}|-\rangle\langle -|$$

This is a complete, solvable quantum system: two levels, one coupling, exact eigenstates. The same structure appears in NMR (Chapter 13), the ammonia maser (Chapter 21), and the two-state quantum computer (Chapter 25). The calculation we just did — diagonalizing a $2\times 2$ Hamiltonian in Dirac notation — is one of the most frequently used techniques in all of quantum physics.

🔗 Connection — The two-level system above is the simplest model of Rabi oscillations, which you previewed in Chapter 7. If a particle starts in $|g\rangle$, the probability of finding it in $|e\rangle$ at time $t$ oscillates as $P_{g \to e}(t) = \frac{\Omega^2}{\omega_0^2 + \Omega^2}\sin^2\left(\frac{\sqrt{\omega_0^2 + \Omega^2}}{2}\,t\right)$. This result will be derived properly in Chapter 21 using time-dependent perturbation theory, but the exact solution above via Dirac notation is more elegant and more general.

8.9 Summary and Project Checkpoint

What you have learned

This chapter has introduced the language that the rest of this book speaks. Let us summarize the key ideas:

1. The ket $|\psi\rangle$ is the abstract quantum state, independent of any representation. The wave function $\psi(x) = \langle x|\psi\rangle$ is the state's representation in the position basis — one choice among many.

2. The bra $\langle\psi|$ is the Hermitian conjugate of the ket. Inner products are brackets: $\langle\phi|\psi\rangle$. The bra introduces a complex conjugation that ensures the inner product is positive-definite.

3. Completeness relations ($\sum_n |n\rangle\langle n| = \hat{I}$) are the master tool. "Insert a 1" to change basis, evaluate matrix elements, or connect representations. If you are stuck, try inserting a completeness relation.

4. Outer products $|n\rangle\langle m|$ are operators. Projection operators $\hat{P}_n = |n\rangle\langle n|$ project onto basis states. Every Hermitian operator has a spectral decomposition $\hat{A} = \sum_n a_n |a_n\rangle\langle a_n|$.

5. Matrix elements $A_{mn} = \langle m|\hat{A}|n\rangle$ connect abstract operators to concrete matrices. Choosing a basis turns the abstract formalism into linear algebra.

6. Unitary transformations ($\hat{U}^\dagger\hat{U} = \hat{I}$) change bases while preserving all physical content. The Fourier transform is a unitary basis change from position to momentum.

7. The trace $\text{Tr}(\hat{A}) = \sum_n \langle n|\hat{A}|n\rangle$ is basis-independent and cyclic. It will be essential for density matrices and quantum information.

8. Everything in wave mechanics translates to Dirac notation via the Rosetta Stone. From now on, we use Dirac notation as our primary language, with wave mechanics translations shown in parentheses during the transition period (Chapters 8--10).

🚪 Threshold Concept — Final Statement — Dirac notation unifies all representations of quantum mechanics. The wave function $\psi(x)$ is not the quantum state — it is the position-basis representation of the ket $|\psi\rangle$. The momentum-space wave function $\phi(p)$, the energy-basis coefficients $c_n$, and the spinor components $(\alpha, \beta)$ are all representations of the same abstract state in different bases. Dirac notation lets you work with the state itself, choosing a basis only when you need numbers. This is not just notational convenience — it is the correct way to think about quantum mechanics, and it is the framework within which spin, entanglement, and quantum computing all become natural.

💡 Key Insight — If you remember one technique from this chapter, make it this: insert a complete set of states. The identity $\sum_n |n\rangle\langle n| = \hat{I}$ (or its continuous version $\int |x\rangle\langle x| \, dx = \hat{I}$) is the Swiss Army knife of quantum mechanics. Whenever you are stuck — cannot simplify an expression, cannot connect two representations, cannot evaluate a matrix element — ask: "Can I insert a 1 here?" Nine times out of ten, the answer is yes, and the problem opens up.

What this means for your problem-solving

The practical impact of this chapter is immediate. Starting now, you have two methods for every quantum mechanics calculation:

1. The wave mechanics method: Write everything in position space, set up integrals, evaluate them. This works, but it is labor-intensive and limited to systems with position-space representations.

2. The Dirac method: Work abstractly with kets, operators, and inner products. Use completeness to change basis. Use ladder operators and matrix elements to avoid integrals. Convert to a specific representation only when you need a number.

For most problems in the rest of this book, the Dirac method will be faster, more general, and more insightful. It will not always be shorter in raw steps — sometimes an integral is quicker than setting up a matrix — but it will always be more transparent. You will see the physics more clearly because the notation does not clutter the view with representation-specific details.

The mark of a mature quantum mechanic is the ability to switch between these methods fluently, choosing whichever is more efficient for the problem at hand. The exercises at the end of this chapter are designed to build that fluency.

Looking ahead

In Chapter 9, we will use Dirac notation to study eigenvalue problems and spectral theory systematically — including the subtleties of continuous spectra and the Dirac delta function. In Chapter 10, we will discover how symmetry operators naturally live in this framework, and how Noether's theorem connects symmetries to conservation laws through commutators with the Hamiltonian. In Chapter 11, we will see how tensor products describe composite systems — something that would be extremely awkward in wave mechanics but is elegant in Dirac notation. And in Chapter 13, the full theory of spin will emerge naturally from the abstract vector-space formalism you have just learned.

The door is open. Let us walk through.

📐 Project Checkpoint — Toolkit v0.8: Dirac Notation Layer

In this chapter's code, you will build: - A Ket class representing abstract quantum states as column vectors - A Bra class representing dual vectors as row vectors - inner_product() and outer_product() functions - to_matrix() for constructing operator matrices in any basis - A basis-change utility using unitary transformations - trace(), spectral_decompose(), and function_of_operator() utilities

These classes will serve as the foundation for all subsequent computational work. From Chapter 9 onward, every code example imports from this module. See code/project-checkpoint.py for the complete implementation and self-test suite.

Key Equations of Chapter 8

For quick reference, the essential equations introduced or reformulated in this chapter:

$$\psi(x) = \langle x|\psi\rangle \qquad \text{(wave function as representation)}$$

$$\langle\phi|\psi\rangle = \int \phi^*(x)\psi(x) \, dx \qquad \text{(inner product in position basis)}$$

$$\sum_n |n\rangle\langle n| = \hat{I} \qquad \text{(completeness, discrete)}$$

$$\int |x\rangle\langle x| \, dx = \hat{I} \qquad \text{(completeness, continuous)}$$

$$A_{mn} = \langle m|\hat{A}|n\rangle \qquad \text{(matrix elements)}$$

$$\hat{A} = \sum_n a_n |a_n\rangle\langle a_n| \qquad \text{(spectral decomposition)}$$

$$f(\hat{A}) = \sum_n f(a_n)|a_n\rangle\langle a_n| \qquad \text{(function of an operator)}$$

$$\hat{U}^\dagger\hat{U} = \hat{I} \qquad \text{(unitary operator)}$$

$$\text{Tr}(\hat{A}) = \sum_n \langle n|\hat{A}|n\rangle \qquad \text{(trace, any basis)}$$

$$\langle\hat{A}\rangle = \langle\psi|\hat{A}|\psi\rangle \qquad \text{(expectation value)}$$

$$\langle x|p\rangle = \frac{1}{\sqrt{2\pi\hbar}}e^{ipx/\hbar} \qquad \text{(position-momentum overlap)}$$

A Final Word

You have just completed the most conceptually dense chapter in this book. If some of it feels like drinking from a fire hose, that is entirely normal. Dirac notation is not just a new set of symbols — it is a new way of thinking about quantum mechanics. The old way (wave functions, integrals, differential operators) is not wrong, but it is incomplete. The new way (kets, bras, operators, completeness) is the language of modern physics.

The good news: you do not need to master every detail in one reading. What you need right now is comfort with the basic objects ($|\psi\rangle$, $\langle\psi|$, $\langle\phi|\psi\rangle$, $\langle m|\hat{A}|n\rangle$, $\sum_n |n\rangle\langle n| = \hat{I}$) and the core technique (inserting completeness). The rest will solidify through practice — in the exercises that follow, in the case studies, and in every subsequent chapter of this book.

Return to this chapter often. When Chapter 12 uses $\langle j, m|\hat{J}_+|j, m'\rangle$ and it looks unfamiliar, come back here and remind yourself: it is a matrix element of an operator between two basis states. When Chapter 24 discusses the singlet state $\frac{1}{\sqrt{2}}(|\uparrow\downarrow\rangle - |\downarrow\uparrow\rangle)$ and you wonder how to compute its properties, the tools are all in this chapter. This is the foundation. Build on it with confidence.

From this chapter forward, Dirac notation is our primary language. Wave mechanics equivalents will be shown in parentheses during the transition period (Chapters 8--10), then assumed known.