Chapter 1

Mathematical Tools of Quantum Mechanics

1.1Introduction

Quantum phenomena do not occur in a Hilbert space, they occur in a laboratory.Asher Peres, Quantum Theory: Concepts and Methods (1993)

The machinery in this chapter is the Hilbert space of that remark. It is worth keeping the second half of the sentence in view while learning the first: every object built here exists to predict the reading on an instrument, and the axioms of Chapter 1 say exactly which object predicts which reading.

In classical mechanics the state of a particle is specified by its position r → ( t ) and momentum p → ( t ) as functions of time. These quantities are real numbers that can, in principle, be measured simultaneously with arbitrary precision. Equations of motion, such as Newton's second law or Hamilton's equations, are ordinary differential equations for these real variables.

In quantum mechanics the situation is fundamentally different:

Because of this linear structure, the language of Hilbert spaces, Dirac bra–ket notation, and linear operators is the natural mathematical framework of quantum mechanics.

The goals of this chapter are:

  1. to formulate the basic axioms of quantum mechanics;

  2. to introduce the Hilbert–space structures that realize each axiom (states, operators, and representations);

  3. to organize the material so that different representations (matrix, position, momentum) are seen as the same physics in different bases.

Remark 1.1.1First-pass route

On a first reading, learn only the state and basis language of Sec. 1.2.4, the matrix language of Sec. 1.3.6, and the measurement rule of Sec. 1.5. Then work the two-state calculation near the end of the chapter. The sections on continuous bases, uncertainty, composite systems, and partial traces can be revisited when a later chapter calls for them.

RemarkPrerequisite recall: complex vectors

A complex column c → = ( c 0 , c 1 ) T has squared norm c → † c → = | c 0 | 2 + | c 1 | 2 . A matrix is Hermitian when H m n = H n m ∗ ; such a matrix has real eigenvalues and orthogonal eigenvectors. These two facts are all the linear algebra needed for the first-pass route. For a fuller review, see the companion Mathematical Methods chapters on vectors, matrices, and eigenvalue problems.

1.2Axiom I: States and Hilbert Space

1.2.1Statement of the Axiom

Axiom IStates

Every isolated physical system is associated with a complex Hilbert space ℋ . The physical state of the system at a given time is represented by a normalized ket | ψ ⟩ ∈ ℋ :

⟨ ψ | ψ ⟩ = 1. (1.1)

The ket | ψ ⟩ encodes all physically accessible information about the system. It is not directly observable; only probabilities and expectation values derived from it are measurable.

1.2.2Hilbert Space and Inner Product

DefinitionComplex vector space

A (complex) vector space V is a set whose elements are kets | v ⟩ , | w ⟩ , … , equipped with vector addition and multiplication by complex scalars α ∈ ℂ satisfying the usual linear algebra axioms (closure, associativity, commutativity of addition, distributivity, and existence of 0 and − | v ⟩ ).

DefinitionInner product

An inner product on V is a map

( | u ⟩ , | v ⟩ ) ⟼ ⟨ u | v ⟩ ∈ ℂ (1.2)

such that, for all kets | u ⟩ , | v ⟩ , | w ⟩ and all α , β ∈ ℂ :

  1. Conjugate symmetry: ⟨ u | v ⟩ = ⟨ v | u ⟩ ∗ .

  2. Linearity in the second argument: ⟨ u | α v + β w ⟩ = α ⟨ u | v ⟩ + β ⟨ u | w ⟩ .

  3. Positive definiteness: ⟨ v | v ⟩ ≥ 0 and ⟨ v | v ⟩ = 0 implies | v ⟩ = 0 .

The inner product defines the norm

∥ | v ⟩ ∥ = ⟨ v | v ⟩ , (1.3)

and two kets are orthogonal if ⟨ u | v ⟩ = 0 .

DefinitionHilbert space

A Hilbert space is a complex vector space H with an inner product ⟨ ⋅ | ⋅ ⟩ which is complete with respect to the norm ∥ | v ⟩ ∥ = ⟨ v | v ⟩ ; i.e., every Cauchy sequence of kets in H converges to a ket in H .

ExampleTypical Hilbert spaces
  1. ℂ n with ⟨ u → | v → ⟩ = ∑ j = 1 n u j ∗ v j is a finite-dimensional Hilbert space (spin systems, few-level atoms).

  2. L 2 ( ℝ 3 ) , the space of square-integrable functions ψ ( r → ) with

    ∫ ℝ 3 | ψ ( r → ) | 2 d 3 r < ∞ , (1.4)

    and inner product

    ⟨ ϕ | ψ ⟩ = ∫ ℝ 3 ϕ ∗ ( r → ) ψ ( r → ) d 3 r , (1.5)

    describes nonrelativistic wave functions in position space.

1.2.3Bases, Orthonormality, and Completeness

A ket is its list of projections and nothing else. The completeness relation nenen= 1 is the statement that laying the components back end to end returns the vector you started with.
Figure 1.1. A ket is its list of projections and nothing else. The completeness relation ∑ n | e n ⟩ ⟨ e n | = 1 ˆ is the statement that laying the components back end to end returns the vector you started with.
DefinitionBasis

A set of kets { | e n ⟩ } in H is a basis if:

  1. the kets are linearly independent;

  2. every | ψ ⟩ ∈ H can be written uniquely as

    | ψ ⟩ = ∑ n c n | e n ⟩ . (1.6)

If a basis has N elements, H is N –dimensional; otherwise H is infinite–dimensional.

DefinitionOrthonormal basis

A set of kets { | n ⟩ } is orthonormal if

⟨ m | n ⟩ = δ m n , (1.7)

and complete if any | ψ ⟩ ∈ H can be expanded as

| ψ ⟩ = ∑ n c n | n ⟩ , c n = ⟨ n | ψ ⟩ . (1.8)

A complete orthonormal set { | n ⟩ } is called an orthonormal basis.

Completeness can be written in operator form as

∑ n | n ⟩ ⟨ n | = I ˆ , (1.9)

where I ˆ is the identity operator on H .

For a continuum of labels α we instead have

⟨ α | α ′ ⟩ = δ ( α − α ′ ) , ∫ d α | α ⟩ ⟨ α | = I ˆ . (1.10)

1.2.4Dirac Notation and State Representations

A physical state is represented by a ket | ψ ⟩ ∈ ℋ . To each ket we associate a bra ⟨ ψ | via Hermitian conjugation:

| ψ ⟩ ⟷ ⟨ ψ | = ( | ψ ⟩ ) † . (1.11)

The inner product is written as

⟨ ϕ | ψ ⟩ = ⟨ ϕ | ψ ⟩ . (1.12)

Course notation convention. From this point onward (including atomic and molecular chapters), the abstract state vector is always written as a ket. Coordinate-space objects such as ψ ( r → ) , ϕ ( p → ) , or molecular forms like Ψ ( r → , R → ) are treated as representations of kets, never as a separate formalism:

ψ ( r → ) = ⟨ r → | ψ ⟩ , ϕ ( p → ) = ⟨ p → | ψ ⟩ , Ψ ( r → , R → ) = ⟨ r → , R → | Ψ ⟩ . (1.13)

We will switch representations when useful, but the state itself remains | ψ ⟩ or | Ψ ⟩ .

Matrix (abstract) representation

In a discrete orthonormal basis { | n ⟩ } , we define the components

c n = ⟨ n | ψ ⟩ . (1.14)

We can represent the state | ψ ⟩ as a column vector

| ψ ⟩ = ^ ( c 1 c 2 ⋮ ) , (1.15)

and bras as row vectors (complex conjugate transpose)

⟨ ψ | = ^ ( c 1 ∗ , c 2 ∗ , … ) . (1.16)

This is the usual matrix representation used in finite-dimensional quantum systems (two-level atoms, spin- 1 2 , etc.).

Position representation

The position basis { | r → ⟩ } satisfies

⟨ r → | r → ′ ⟩ = δ ( r → − r → ′ ) , ∫ d 3 r | r → ⟩ ⟨ r → | = I ˆ . (1.17)

The position-space wave function of | ψ ⟩ is

ψ ( r → ) = ⟨ r → | ψ ⟩ , (1.19)

and

| ψ ⟩ = ∫ d 3 r | r → ⟩ ψ ( r → ) . (1.20)

Normalization ⟨ ψ | ψ ⟩ = 1 becomes

∫ d 3 r | ψ ( r → ) | 2 = 1 , (1.21)

so | ψ ( r → ) | 2 d 3 r is the probability to find the particle near r → .

Momentum representation

Similarly, the momentum basis { | p → ⟩ } satisfies

⟨ p → | p → ′ ⟩ = δ ( p → − p → ′ ) , ∫ d 3 p | p → ⟩ ⟨ p → | = I ˆ . (1.22)

The momentum-space wave function is

ϕ ( p → ) = ⟨ p → | ψ ⟩ , | ψ ⟩ = ∫ d 3 p | p → ⟩ ϕ ( p → ) , (1.24)

with normalization

∫ d 3 p | ϕ ( p → ) | 2 = 1. (1.25)

Connection between representations

The position and momentum bases are related by

⟨ r → | p → ⟩ = 1 ( 2 π ℏ ) 3 / 2 exp ( i ℏ p → ⋅ r → ) , (1.26)

so that

ψ ( r → ) = ⟨ r → | ψ ⟩ = 1 ( 2 π ℏ ) 3 / 2 ∫ d 3 p e i ℏ p → ⋅ r → ϕ ( p → ) , ϕ ( p → ) = ⟨ p → | ψ ⟩ = 1 ( 2 π ℏ ) 3 / 2 ∫ d 3 r e − i ℏ p → ⋅ r → ψ ( r → ) . (1.27)

Thus:

They are simply different representations of the same abstract state | ψ ⟩ .

1.3Axiom II: Observables as Operators

1.3.1Statement of the Axiom

Axiom IIObservables

Every physical observable A is represented by a linear Hermitian operator A ˆ acting on the Hilbert space ℋ .

The mathematical objects we need here are:

1.3.2Linear Operators and the Adjoint

An operator A ˆ is a map that sends kets to kets:

| ψ ⟩ ↦ A ˆ | ψ ⟩ . (1.29)

It is linear if, for all complex numbers α , β and kets | ψ 1 ⟩ , | ψ 2 ⟩ ,

A ˆ ( α | ψ 1 ⟩ + β | ψ 2 ⟩ ) = α A ˆ | ψ 1 ⟩ + β A ˆ | ψ 2 ⟩ . (1.30)
DefinitionAdjoint operator

Given an operator A ˆ , its adjoint A ˆ † is defined by

⟨ ϕ | A ˆ | ψ ⟩ = ( ⟨ ψ | A ˆ † | ϕ ⟩ ) ∗ for all  | ϕ ⟩ , | ψ ⟩ ∈ ℋ . (1.31)

In a discrete orthonormal basis { | n ⟩ } with matrix elements A m n = ⟨ m | A ˆ | n ⟩ , the adjoint corresponds to the conjugate transpose:

( A ˆ † ) m n = A n m ∗ . (1.32)

1.3.3Hermitian Operators (Observables)

DefinitionHermitian operator

An operator A ˆ is Hermitian (or self–adjoint) if

A ˆ † = A ˆ . (1.33)

For Hermitian operators:

TheoremSpectral decomposition (discrete case)

If A ˆ has a discrete, nondegenerate spectrum { a n } with orthonormal eigenkets { | a n ⟩ } , then

A ˆ = ∑ n a n | a n ⟩ ⟨ a n | , ∑ n | a n ⟩ ⟨ a n | = I ˆ . (1.34)

In the presence of a continuous spectrum { a } we write instead

A ˆ = ∑ n a n | a n ⟩ ⟨ a n | + ∫ d a a | a ⟩ ⟨ a | , (1.35)

with the corresponding completeness relation

∑ n | a n ⟩ ⟨ a n | + ∫ d a | a ⟩ ⟨ a | = I ˆ . (1.36)

The Hamiltonian H ˆ , which governs time evolution, is one of the most important Hermitian operators.

1.3.4Identity and Projection Operators

DefinitionIdentity operator

The identity operator I ˆ is defined by

I ˆ | ψ ⟩ = | ψ ⟩ for all  | ψ ⟩ ∈ ℋ . (1.37)

In an orthonormal basis { | n ⟩ } we have

I ˆ = ∑ n | n ⟩ ⟨ n | , (1.38)

and in a continuous basis { | α ⟩ } ,

I ˆ = ∫ d α | α ⟩ ⟨ α | . (1.39)
DefinitionProjection operator

An operator P ˆ is a projection operator (projector) if

P ˆ 2 = P ˆ , P ˆ † = P ˆ . (1.40)

Projectors represent “yes/no’’ questions. For example, if P ˆ projects onto a subspace S ⊂ ℋ , then

For a nondegenerate eigenvalue a n of A ˆ , the projector onto the eigenspace spanned by | a n ⟩ is

P ˆ n = | a n ⟩ ⟨ a n | . (1.41)

The spectral decomposition can be written in terms of projectors:

A ˆ = ∑ n a n P ˆ n . (1.42)

Projectors will appear naturally in the measurement axiom (state reduction) and in decomposing dynamics into discrete and continuum contributions (Rydberg vs. ionization channels).

1.3.5Unitary Operators

DefinitionUnitary operator

An operator U ˆ is unitary if

U ˆ † U ˆ = U ˆ U ˆ † = I ˆ . (1.43)

Properties:

Important examples:

1.3.6Representations of Operators

Just like states, the same abstract operator A ˆ can be represented in different bases.

Matrix representation in a discrete basis

In an orthonormal basis { | n ⟩ } , the matrix elements of A ˆ are

A m n = ⟨ m | A ˆ | n ⟩ . (1.45)

If | ψ ⟩ = ∑ n c n | n ⟩ , then

( A ˆ | ψ ⟩ ) m = ⟨ m | A ˆ | ψ ⟩ = ∑ n A m n c n . (1.46)

In matrix form,

c → ′ = A c → , (1.47)

where c → and c → ′ are column vectors of components and A is the matrix representing A ˆ .

Kernels in continuous bases

In the position basis { | r → ⟩ } , the kernel of A ˆ is

A ( r → , r → ′ ) = ⟨ r → | A ˆ r → ′ ⟩ , (1.48)

and

( A ˆ ψ ) ( r → ) = ⟨ r → | A ˆ ψ ⟩ = ∫ d 3 r ′ A ( r → , r → ′ ) ψ ( r → ′ ) . (1.49)

Examples:

In the momentum basis { | p → ⟩ } we similarly define

A ( p → , p → ′ ) = ⟨ p → | A ˆ p → ′ ⟩ , ( A ˆ ϕ ) ( p → ) = ∫ d 3 p ′ A ( p → , p → ′ ) ϕ ( p → ′ ) . (1.52)

Thus:

1.3.7Functions of Operators

We often encounter functions of operators, such as f ( H ˆ ) or e − i H ˆ t / ℏ . For a Hermitian operator with spectral decomposition

A ˆ = ∑ n a n | a n ⟩ ⟨ a n | + ∫ d a a | a ⟩ ⟨ a | , (1.53)

we define

f ( A ˆ ) = ∑ n f ( a n ) | a n ⟩ ⟨ a n | + ∫ d a f ( a ) | a ⟩ ⟨ a | . (1.54)

Important special case:

e − i H ˆ t / ℏ = ∑ n e − i E n t / ℏ | E n ⟩ ⟨ E n | + ∫ d E e − i E t / ℏ | E ⟩ ⟨ E | , (1.55)

which is the time-evolution operator for a time-independent Hamiltonian.

This spectral calculus is crucial later when we work with mixed discrete (“Rydberg’’) and continuum (“ionization’’) states.

1.4Commutators and the Heisenberg Uncertainty Principle

1.4.1Commutators and Compatible Observables

Given two operators A ˆ and B ˆ , their commutator is

[ A ˆ , B ˆ ] = A ˆ B ˆ − B ˆ A ˆ . (1.56)

If [ A ˆ , B ˆ ] = 0 , we say that A ˆ and B ˆ are compatible: there exists a basis of common eigenkets { | a , b ⟩ } such that

A ˆ | a , b ⟩ = a | a , b ⟩ , B ˆ | a , b ⟩ = b | a , b ⟩ . (1.57)

Then A and B can be simultaneously measured with arbitrary precision.

If [ A ˆ , B ˆ ] ≠ 0 , no such complete common eigenbasis exists. The corresponding observables cannot, in general, be simultaneously sharp; this is quantified by the uncertainty principle.

Important example in one dimension:

[ x ˆ , p ˆ x ] = i ℏ I ˆ . (1.58)

1.4.2Uncertainties and Standard Deviations

For a normalized state | ψ ⟩ and Hermitian operator A ˆ , define

⟨ A ⟩ = ⟨ ψ | A ˆ | ψ ⟩ , ( Δ A ) 2 = ⟨ ψ | ( A ˆ − ⟨ A ⟩ ) 2 | ψ ⟩ = ⟨ A 2 ⟩ − ⟨ A ⟩ 2 .

Δ A is the standard deviation of measurement outcomes of A in state | ψ ⟩ .

Similarly for B ˆ :

( Δ B ) 2 = ⟨ ψ | ( B ˆ − ⟨ B ⟩ ) 2 | ψ ⟩ . (1.59)

1.4.3Heisenberg (Robertson) Uncertainty Relation

The relation is a statement about one state seen in two bases, not about a clumsy apparatus. Narrowing |(x)|^2 widens |(p)|^2 by the same factor, because the two are Fourier transforms of each other.
Figure 1.2. The relation is a statement about one state seen in two bases, not about a clumsy apparatus. Narrowing | ψ ( x ) | 2 widens | ψ ˜ ( p ) | 2 by the same factor, because the two are Fourier transforms of each other.
TheoremHeisenberg–Robertson uncertainty relation

For any Hermitian operators A ˆ and B ˆ and any normalized state | ψ ⟩ ,

Δ A Δ B ≥ 1 2 | ⟨ ψ | [ A ˆ , B ˆ ] | ψ ⟩ | . (1.60)

Thus, the commutator [ A ˆ , B ˆ ] sets a fundamental lower bound on the product of uncertainties in A and B .

For position and momentum in one dimension:

[ x ˆ , p ˆ x ] = i ℏ I ˆ ⇒ Δ x Δ p x ≥ ℏ 2 . (1.61)

This is the standard Heisenberg uncertainty relation. It is a direct consequence of:

Later, similar commutator structures appear in:

These will be built on top of the operator concepts introduced here.

1.5Axioms III and IV: Measurement and State Reduction

1.5.1Measurement Outcomes and Probabilities

Axiom IIIMeasurement outcomes and their probabilities

Let A ˆ be the Hermitian operator associated with an observable A , with spectral decomposition

A ˆ = ∑ n a n | a n ⟩ ⟨ a n | or A ˆ = ∫ d a a | a ⟩ ⟨ a | . (1.62)

If the system is in state | ψ ⟩ , then a measurement of A yields:

  • in the discrete case, outcome a n with probability P ( a n ) = | ⟨ a n | ψ ⟩ | 2 ;

  • in the continuous case, outcome in ( a , a + d a ) with probability P ( a ) d a = | ⟨ a | ψ ⟩ | 2 d a .

1.5.2State Reduction (Collapse)

Measurement is not the discovery of a value the system already had. The amplitudes that did not occur are gone from the state afterwards, which is what makes a second measurement of the same observable return the same answer with certainty.
Figure 1.3. Measurement is not the discovery of a value the system already had. The amplitudes that did not occur are gone from the state afterwards, which is what makes a second measurement of the same observable return the same answer with certainty.
Axiom IVState reduction

Immediately after an ideal measurement of A yielding the value a n , the state of the system becomes the corresponding normalized eigenstate | a n ⟩ (or, more generally, the projection onto the eigensubspace for a degenerate eigenvalue).

This axiom introduces intrinsically probabilistic, non-unitary evolution associated with measurement, in contrast to the unitary evolution governed by the Schrödinger equation.

1.6Axiom V: Time Evolution

1.6.1Schrödinger Equation and Unitary Evolution

Axiom VTime evolution

The time evolution of a closed quantum system is governed by the time-dependent Schrödinger equation

i ℏ ∂ ∂ t | ψ ( t ) ⟩ = H ˆ ( t ) | ψ ( t ) ⟩ , (1.63)

where H ˆ ( t ) is the Hermitian Hamiltonian of the system.

The formal solution can be written as

| ψ ( t ) ⟩ = U ˆ ( t , t 0 ) | ψ ( t 0 ) ⟩ , (1.64)

where U ˆ ( t , t 0 ) is a unitary operator satisfying

i ℏ ∂ ∂ t U ˆ ( t , t 0 ) = H ˆ ( t ) U ˆ ( t , t 0 ) , U ˆ ( t 0 , t 0 ) = I ˆ , (1.65)

and

U ˆ † ( t , t 0 ) U ˆ ( t , t 0 ) = I ˆ . (1.66)

For time-independent H ˆ ,

U ˆ ( t , t 0 ) = exp [ − i ℏ H ˆ ( t − t 0 ) ] . (1.67)

In the energy eigenbasis H ˆ | E n ⟩ = E n | E n ⟩ ,

| ψ ( t ) ⟩ = ∑ n c n e − i ℏ E n ( t − t 0 ) | E n ⟩ , (1.68)

so each energy component acquires only a phase factor.

Expectation values become time-dependent through

⟨ A ⟩ ( t ) = ⟨ ψ ( t ) | A ˆ | ψ ( t ) ⟩ . (1.69)

1.7Axiom VI: Composite Systems

Axiom VIComposite systems

For a composite system with subsystems 1 and 2 , whose state spaces are ℋ 1 and ℋ 2 , the state space of the total system is the tensor product

ℋ tot = ℋ 1 ⊗ ℋ 2 . (1.70)

A general state of the composite system is a ket | Ψ ⟩ in ℋ tot . Only a subset of such states can be written as simple products of subsystem states.

1.7.1Tensor Product and Product Basis

Let { | i ⟩ } be an orthonormal basis for ℋ 1 and { | α ⟩ } an orthonormal basis for ℋ 2 . The tensor product space

ℋ tot = ℋ 1 ⊗ ℋ 2 (1.71)

has a natural product basis

{ | i ⟩ ⊗ | α ⟩ } , (1.72)

which we often abbreviate as

| i ⟩ ⊗ | α ⟩ ≡ | i , α ⟩ ≡ | i ⟩ | α ⟩ . (1.73)

If dim ⁡ ℋ 1 = d 1 and dim ⁡ ℋ 2 = d 2 , then

dim ⁡ ℋ tot = d 1 d 2 . (1.74)

For example, two spin- 1 2 particles each have a two-dimensional state space; the composite system has dimension 2 × 2 = 4 .

The inner product on ℋ tot is defined by

⟨ i , α | j , β ⟩ = ⟨ i | j ⟩ ⟨ α | β ⟩ = δ i j δ α β , (1.75)

and extended by linearity to arbitrary superpositions.

Any state | Ψ ⟩ ∈ ℋ tot can be expanded as

| Ψ ⟩ = ∑ i , α C i α | i , α ⟩ , (1.76)

with complex coefficients C i α satisfying

∑ i , α | C i α | 2 = 1 (1.77)

for a normalized state.

1.7.2Product States vs. Entangled States

Entanglement is defined by a negative — a state that cannot be factorized — and the coefficient matrix turns that into something one can check. A product state has cij=aibj, so every row is a multiple of every other and the matrix has rank one. Any other rank is entanglement.
Figure 1.4. Entanglement is defined by a negative — a state that cannot be factorized — and the coefficient matrix turns that into something one can check. A product state has c i j = a i b j , so every row is a multiple of every other and the matrix has rank one. Any other rank is entanglement.

A state of the composite system is called a product state (separable state) if it can be written as

| Ψ ⟩ = | ψ 1 ⟩ ⊗ | ψ 2 ⟩ , (1.78)

where | ψ 1 ⟩ ∈ ℋ 1 and | ψ 2 ⟩ ∈ ℋ 2 are normalized subsystem kets.

Writing

| ψ 1 ⟩ = ∑ i a i | i ⟩ , | ψ 2 ⟩ = ∑ α b α | α ⟩ , (1.79)

we obtain

| Ψ ⟩ = ( ∑ i a i | i ⟩ ) ⊗ ( ∑ α b α | α ⟩ ) = ∑ i , α a i b α | i , α ⟩ . (1.80)

Thus, for a product state, the coefficient matrix C i α factorizes as

C i α = a i b α . (1.81)

A state that cannot be written in this form is called entangled. For such states the coefficients C i α do not factorize into a product of a purely i -dependent piece and a purely α -dependent piece.

ExampleTwo spin- 1 2 particles

Let subsystem 1 and 2 each have basis { | ↑ ⟩ , | ↓ ⟩ } . A product basis for the composite system is

{ | ↑↑ ⟩ , | ↑↓ ⟩ , | ↓↑ ⟩ , | ↓↓ ⟩ } , (1.82)

with the shorthand | ↑↓ ⟩ = | ↑ ⟩ ⊗ | ↓ ⟩ , etc.

  • The state

    | Ψ prod ⟩ = | ↑ ⟩ ⊗ ( 1 2 ( | ↑ ⟩ + | ↓ ⟩ ) ) = 1 2 | ↑↑ ⟩ + 1 2 | ↑↓ ⟩ (1.83)

    is a product state.

  • The state

    | Ψ ent ⟩ = 1 2 ( | ↑↓ ⟩ + | ↓↑ ⟩ ) (1.84)

    cannot be written as | ψ 1 ⟩ ⊗ | ψ 2 ⟩ : it is an entangled state.

In later chapters, tensor-product structures appear when we treat:

1.7.3Operators on Composite Systems

If A ˆ 1 is an operator on ℋ 1 and B ˆ 2 an operator on ℋ 2 , we can define an operator on ℋ tot = ℋ 1 ⊗ ℋ 2 by

A ˆ 1 ⊗ B ˆ 2 . (1.85)

Its action on a product state is

( A ˆ 1 ⊗ B ˆ 2 ) ( | ψ 1 ⟩ ⊗ | ψ 2 ⟩ ) = ( A ˆ 1 | ψ 1 ⟩ ) ⊗ ( B ˆ 2 | ψ 2 ⟩ ) , (1.86)

and extended by linearity to superpositions.

Important special cases:

In the product basis { | i , α ⟩ } ,

( A ˆ 1 ⊗ B ˆ 2 ) ( i , α ) , ( j , β ) = ⟨ i , α | A ˆ 1 ⊗ B ˆ 2 | j , β ⟩ = ⟨ i | A ˆ 1 | j ⟩ ⟨ α | B ˆ 2 | β ⟩ . (1.89)

At the matrix level, this corresponds to a Kronecker product of matrices.

Expectation values of composite observables are computed as

⟨ A 1 ⊗ B 2 ⟩ = ⟨ Ψ | A ˆ 1 ⊗ B ˆ 2 | Ψ ⟩ . (1.90)

1.7.4Reduced Descriptions and Partial Trace (Minimal Use)

In many situations we are interested only in one part of a composite system (e.g. electron vs. ionized core, molecule vs. radiation field). To describe subsystem 1 alone, we use a reduced state obtained by tracing over the unobserved subsystem 2 .

A convenient language for this is the density operator. For a normalized pure state | Ψ ⟩ of the composite system,

ρ ˆ tot = | Ψ ⟩ ⟨ Ψ | . (1.91)

The reduced density operator of subsystem 1 is defined as the partial trace over subsystem 2 :

ρ ˆ 1 = Tr 2 ( ρ ˆ tot ) . (1.92)

In a product basis { | i , α ⟩ } this can be written as

ρ ˆ 1 = ∑ α ⟨ α | ρ ˆ tot | α ⟩ , (1.93)

where the bra and ket refer only to subsystem 2 .

ExamplePartial trace for a two-qubit state

Let

| Ψ ⟩ = 1 2 ( | ↑↓ ⟩ + | ↓↑ ⟩ ) , (1.94)

and ρ ˆ tot = | Ψ ⟩ ⟨ Ψ | . Using the basis { | ↑ ⟩ , | ↓ ⟩ } for each subsystem, show that

ρ ˆ 1 = 1 2 | ↑ ⟩ ⟨ ↑ | + 1 2 | ↓ ⟩ ⟨ ↓ | . (1.95)

Thus, although the total state is pure, the reduced state of subsystem 1 is a statistical mixture.

For the present notes, we will use these concepts mainly in simple finite-dimensional examples (e.g. two-level systems, two spin- 1 2 particles) to:

A more systematic treatment of density operators and mixed states appears in advanced courses on quantum information or open quantum systems.

minimally in the present notes.

1.8First-Pass Calculation: One Two-State Hamiltonian

Use the ordered orthonormal basis { | 0 ⟩ , | 1 ⟩ } . The order is part of the representation: it tells us which coefficient and which matrix row belongs to which ket. Consider the teaching Hamiltonian

H ˆ = ^ H = ( 0 g g D ) , D > 0 , g ∈ ℝ . (1.96)

Here D is the uncoupled energy gap and g is a real coupling. The symbol D is used for an energy gap; frequency detuning is a different quantity introduced in Chapter 6.

Definition 1.8.1State, coefficients, operator, and matrix

The ket | ψ ⟩ is the physical state. Once the ordered basis is chosen, its coefficients form a column:

| ψ ⟩ = c 0 | 0 ⟩ + c 1 | 1 ⟩ = ^ ( c 0 c 1 ) . (1.97)

Likewise, H ˆ is the operator and H m n = ⟨ m | H ˆ | n ⟩ is its matrix in that basis. Changing basis changes the column and the matrix together; it does not change the state, operator, eigenvalues, or measurement probabilities.

Theorem 1.8.1Exact energies of the two-state model

The eigenvalues of Eq. (1.96) are

E ± = D ± D 2 + 4 g 2 2 . (1.98)

The eigenvalue equation has a nonzero solution only if

0 = det ⁡ ( H − E I ) = | − E g g D − E | = ( − E ) ( D − E ) − g 2 = E 2 − D E − g 2 .

The quadratic formula gives Eq. (1.98). At g = 0 the roots become 0 and D , which checks the uncoupled limit.

For either root, the first row of ( H − E I ) c → = 0 gives g c 1 = E c 0 . Choosing c 0 > 0 fixes the irrelevant overall sign, and normalization gives

| E ⟩ = | 0 ⟩ + ( E / g ) | 1 ⟩ 1 + ( E / g ) 2 ( g ≠ 0 ) . (1.99)

For g > 0 , E − < 0 , so the two ground-state coefficients have opposite signs. This relative sign is physical because it controls the negative interference contribution 2 g c 0 c 1 to the energy.

The three self-checks below, drawn. Left: the exact lower root of Eq. (1.98) against the second-order estimate -g^2/D. They are indistinguishable while g D and separate visibly once g approaches D/2 — the expansion parameter is g/D, and the figure shows where it stops being small. Right: the Rayleigh quotient H for the normalized trial state (,) at g/D=0.2. Its minimum over touches the exact lower eigenvalue and does not go below it, which is the variational principle in one picture. Both panels are computed by diagonalizing and by minimizing, not by plotting the formula.
Figure 1.5. The three self-checks below, drawn. Left: the exact lower root of Eq. (1.98) against the second-order estimate − g 2 / D . They are indistinguishable while g ≪ D and separate visibly once g approaches D / 2 — the expansion parameter is g / D , and the figure shows where it stops being small. Right: the Rayleigh quotient ⟨ H ⟩ for the normalized trial state ( cos ⁡ θ , sin ⁡ θ ) at g / D = 0.2 . Its minimum over θ touches the exact lower eigenvalue and does not go below it, which is the variational principle in one picture. Both panels are computed by diagonalizing and by minimizing, not by plotting the formula.
Example 1.8.1Self-check 1: exact diagonalization

Take D = 1 and g = 0.2 in a declared energy unit. Find the two energies, the normalized ground state, and the probabilities of the two basis outcomes.

Solution.

Equation (1.98) gives

E ± = 1 ± 1 + 4 ( 0.2 ) 2 2 = 1 ± 1.16 2 ,

so

E − = − 0.0385164807 , E + = 1.0385164807 .

The coefficient ratio is c 1 / c 0 = E − / g = − 0.192582404 . Normalizing,

| E − ⟩ = 0.981956 | 0 ⟩ − 0.189108 | 1 ⟩ .

Therefore a measurement in the ordered basis gives P ( 0 ) = 0.964238 and P ( 1 ) = 0.035762 ; their sum is one. The coupling lowers the ground energy below the smaller diagonal entry and mixes a small amount of | 1 ⟩ into the ground state.

The same answer can be obtained without solving a determinant. Every real, normalized state in this two-dimensional problem can be written

| ψ ( θ ) ⟩ = cos ⁡ θ | 0 ⟩ + sin ⁡ θ | 1 ⟩ . (1.100)

Its Rayleigh quotient is already normalized:

R ( θ ) = ⟨ ψ ( θ ) | H ˆ | ψ ( θ ) ⟩ = D sin 2 ⁡ θ + g sin ⁡ 2 θ . (1.101)
Example 1.8.2Self-check 2: the variational route

Minimize Eq. (1.101) and compare with the exact ground state for D = 1 , g = 0.2 .

Solution.

Differentiate rather than guess:

d R d θ = D sin ⁡ 2 θ + 2 g cos ⁡ 2 θ = 0 , tan ⁡ 2 θ = − 2 g D .

The minimum connected continuously to | 0 ⟩ lies in − π / 4 < θ < 0 , so

θ − = − 1 2 arctan ⁡ ( 0.4 ) = − 0.190253   rad .

Thus cos ⁡ θ − = 0.981956 and sin ⁡ θ − = − 0.189108 , exactly the coefficients found in Self-check 1. Substitution into R gives R ( θ − ) = − 0.0385164807 . The variational search is exact here because the one-parameter trial family contains every normalized real state in the two-dimensional space.

When | g | / D ≪ 1 , expand D 2 + 4 g 2 = D 1 + 4 g 2 / D 2 with 1 + x = 1 + x / 2 − x 2 / 8 + ⋯ . The lower root becomes

E − = − g 2 D + g 4 D 3 + O ( g 6 D 5 ) . (1.102)
Example 1.8.3Self-check 3: approximation and its failure

Use second order for D = 1 , g = 0.2 , then explain what happens as the uncoupled gap closes.

Solution.

Keeping only the first term of Eq. (1.102),

E − ( 2 ) = − ( 0.2 ) 2 1 = − 0.0400000 ,

which differs from the exact − 0.0385165 by 0.0014835 , or 3.85 % of the magnitude. Including g 4 / D 3 = 0.0016 gives − 0.0384 .

The expansion parameter is | g | / D . If D → 0 at fixed g , the second-order expression diverges, while the exact root remains finite:

E − ⟶ − | g | .

The approximation therefore fails at gap closure. The failure is not a defect in the exact two-state model; it says that coupling cannot be treated as a small correction when the uncoupled levels are degenerate.

1.9Summary

1.10Exercises

  1. For D = 2 and g = 0.1 in the same energy unit, compute E − exactly and from − g 2 / D . State the relative error.

  2. For the state ( 3 | 0 ⟩ + 4 i | 1 ⟩ ) / 5 , find P ( 0 ) and P ( 1 ) and verify normalization. Explain why the relative phase does not change these two probabilities.

  3. Show directly that replacing | 1 ⟩ by − | 1 ⟩ changes the matrix entry g to − g but leaves Eq. (1.98) unchanged. This is a basis change, not a new physical Hamiltonian.

1.11Optional Extension Problems

The former homework set is retained as an optional, more advanced extension. It ranges from formal proofs to composite systems. Unless stated otherwise, work in Dirac notation.

Hilbert Space, Inner Products, and Bases

  1. Let ℋ be a complex vector space with inner product ⟨ ⋅ | ⋅ ⟩ .

    1. Show that for any kets | ϕ ⟩ and | ψ ⟩ and any α ∈ ℂ ,

      ⟨ ϕ | α ψ ⟩ = α ⟨ ϕ | ψ ⟩ . (1.103)
    2. Prove the Cauchy–Schwarz inequality:

      | ⟨ ϕ | ψ ⟩ | ≤ ∥ | ϕ ⟩ ∥ ∥ | ψ ⟩ ∥ . (1.104)
    3. Use the Cauchy–Schwarz inequality to show the triangle inequality for the norm:

      ∥ | ϕ ⟩ + | ψ ⟩ ∥ ≤ ∥ | ϕ ⟩ ∥ + ∥ | ψ ⟩ ∥ . (1.105)
  2. Let { | n ⟩ } be an orthonormal basis of a Hilbert space ℋ .

    1. Show that any normalized state | ψ ⟩ can be written as

      | ψ ⟩ = ∑ n c n | n ⟩ , ∑ n | c n | 2 = 1 , (1.106)

      and identify c n in terms of | ψ ⟩ and | n ⟩ .

    2. Prove the completeness relation

      ∑ n | n ⟩ ⟨ n | = I ˆ (1.107)

      by acting on an arbitrary state | ψ ⟩ .

    3. Suppose | ϕ ⟩ = ∑ n d n | n ⟩ is orthogonal to | ψ ⟩ . Express the orthogonality condition ⟨ ϕ | ψ ⟩ = 0 in terms of the coefficients c n and d n .

  3. Consider a normalized state | ψ ⟩ in L 2 ( ℝ ) with wave functions

    ψ ( x ) = ⟨ x | ψ ⟩ , ϕ ( p ) = ⟨ p | ψ ⟩ . (1.108)
    1. Write explicitly the Fourier transforms relating ψ ( x ) and ϕ ( p ) using ⟨ x | p ⟩ .

    2. Show that normalization is preserved under the Fourier transform, i.e.,

      ∫ − ∞ ∞ | ψ ( x ) | 2 d x = ∫ − ∞ ∞ | ϕ ( p ) | 2 d p = 1. (1.109)
    3. For the Gaussian wave function

      ψ ( x ) = 1 ( π a 2 ) 1 / 4 exp ( − x 2 2 a 2 ) , (1.110)

      compute ϕ ( p ) up to an overall phase and check normalization.

  4. Let { | 1 ⟩ , | 2 ⟩ , | 3 ⟩ } be an orthonormal basis of a three-dimensional Hilbert space.

    1. Write the column-vector representation of a general state

      | ψ ⟩ = α | 1 ⟩ + β | 2 ⟩ + γ | 3 ⟩ , (1.111)

      in this basis.

    2. Suppose another orthonormal basis { | a ⟩ , | b ⟩ , | c ⟩ } is related by

      | a ⟩ = 1 2 ( | 1 ⟩ + | 2 ⟩ ) , | b ⟩ = 1 2 ( | 1 ⟩ − | 2 ⟩ ) , | c ⟩ = | 3 ⟩ . (1.112)

      Find the unitary matrix U that transforms components from the { | 1 ⟩ , | 2 ⟩ , | 3 ⟩ } basis to the { | a ⟩ , | b ⟩ , | c ⟩ } basis.

    3. Express the components of | ψ ⟩ in the { | a ⟩ , | b ⟩ , | c ⟩ } basis using U .

    Operators: Hermitian, Unitary, Projectors, and Functions

  5. Let A ˆ be a Hermitian operator and | ψ ⟩ a normalized state.

    1. Show that the expectation value ⟨ A ⟩ = ⟨ ψ | A ˆ | ψ ⟩ is always real.

    2. Suppose | ψ ⟩ is an eigenket of A ˆ with eigenvalue a . Show that

      Δ A = 0 , (1.113)

      where ( Δ A ) 2 = ⟨ A 2 ⟩ − ⟨ A ⟩ 2 .

    3. Conversely, show that if Δ A = 0 in a normalized state | ψ ⟩ , then | ψ ⟩ must be an eigenket of A ˆ .

  6. In the basis { | 1 ⟩ , | 2 ⟩ , | 3 ⟩ } , an operator A ˆ is represented by the matrix

    A = ( 1 i 0 − i 2 0 0 0 3 ) . (1.114)
    1. Check whether A ˆ is Hermitian.

    2. Find its eigenvalues.

    3. For each distinct eigenvalue, find a corresponding normalized eigenket in the given basis.

  7. Let | ϕ ⟩ be a normalized state in ℋ and define the operator

    P ˆ = | ϕ ⟩ ⟨ ϕ | . (1.115)
    1. Show that P ˆ is Hermitian and that P ˆ 2 = P ˆ .

    2. For an arbitrary normalized state | ψ ⟩ , interpret P ˆ | ψ ⟩ physically, and compute ∥ P ˆ | ψ ⟩ ∥ 2 .

    3. Show that I ˆ − P ˆ is also a projector and interpret the subspace onto which it projects.

  8. Let U ˆ be a unitary operator, U ˆ † U ˆ = I ˆ .

    1. Show that U ˆ preserves inner products, i.e.,

      ⟨ ϕ | ψ ⟩ = ⟨ ϕ ′ | ψ ′ ⟩ , | ϕ ′ ⟩ = U ˆ | ϕ ⟩ , | ψ ′ ⟩ = U ˆ | ψ ⟩ . (1.116)
    2. Prove that if | a ⟩ is an eigenket of a Hermitian operator A ˆ with eigenvalue a , then U ˆ | a ⟩ is an eigenket of A ˆ ′ = U ˆ A ˆ U ˆ † with the same eigenvalue a .

    3. Explain how this shows that unitary transformations correspond to changes of representation that preserve the spectrum of observables.

  9. Let A ˆ be a Hermitian operator with discrete spectrum A ˆ = ∑ n a n | a n ⟩ ⟨ a n | .

    1. Define f ( A ˆ ) for a real-valued function f ( x ) .

    2. Show that if f is real-valued on the spectrum { a n } , then f ( A ˆ ) is Hermitian.

    3. For the Hamiltonian H ˆ with eigenvalues E n , write the spectral decomposition of the time-evolution operator U ˆ ( t ) = e − i H ˆ t / ℏ .

    Commutators and the Heisenberg Uncertainty Principle

  10. For operators A ˆ , B ˆ and scalar λ ∈ ℂ , show:

    1. [ A ˆ , B ˆ ] = − [ B ˆ , A ˆ ] .

    2. [ A ˆ , λ B ˆ ] = λ [ A ˆ , B ˆ ] .

    3. [ A ˆ , B ˆ + C ˆ ] = [ A ˆ , B ˆ ] + [ A ˆ , C ˆ ] .

    4. [ A ˆ , B ˆ C ˆ ] = [ A ˆ , B ˆ ] C ˆ + B ˆ [ A ˆ , C ˆ ] .

  11. In one dimension, the position and momentum operators act in the position representation as

    ( x ˆ ψ ) ( x ) = x ψ ( x ) , ( p ˆ ψ ) ( x ) = − i ℏ d d x ψ ( x ) . (1.117)
    1. Compute ( x ˆ p ˆ ψ ) ( x ) and ( p ˆ x ˆ ψ ) ( x ) explicitly.

    2. Use the results to show that

      [ x ˆ , p ˆ ] = i ℏ I ˆ . (1.118)
  12. Let A ˆ and B ˆ be Hermitian operators and | ψ ⟩ a normalized state.

    1. Define Δ A and Δ B and show that Δ A ≥ 0 , Δ B ≥ 0 .

    2. Consider the kets

      | α ⟩ = ( A ˆ − ⟨ A ⟩ ) | ψ ⟩ , | β ⟩ = ( B ˆ − ⟨ B ⟩ ) | ψ ⟩ . (1.119)

      Use the Cauchy–Schwarz inequality to derive the Robertson inequality

      Δ A Δ B ≥ 1 2 | ⟨ ψ | [ A ˆ , B ˆ ] | ψ ⟩ | . (1.120)
    3. Apply this to A ˆ = x ˆ and B ˆ = p ˆ x to obtain the standard Heisenberg uncertainty relation.

    Measurement, Time Evolution, and Composite Systems

  13. Let A ˆ be an observable with nondegenerate discrete spectrum A ˆ = ∑ n a n | a n ⟩ ⟨ a n | and normalized state | ψ ⟩ = ∑ n c n | a n ⟩ .

    1. Show that the probability to obtain the result a n in a measurement of A is P ( a n ) = | c n | 2 .

    2. Show that immediately after obtaining the result a n , the state becomes | a n ⟩ (up to an overall phase).

    3. Suppose instead that A ˆ has a twofold degenerate eigenvalue a with orthonormal eigenkets { | u 1 ⟩ , | u 2 ⟩ } . Write the projector onto the degenerate eigenspace and describe the post-measurement state after observing a .

  14. Assume a time-independent Hamiltonian H ˆ with discrete eigenvalues E n and eigenkets | E n ⟩ :

    H ˆ | E n ⟩ = E n | E n ⟩ . (1.121)

    A normalized initial state at t = 0 is

    | ψ ( 0 ) ⟩ = ∑ n c n | E n ⟩ . (1.122)
    1. Find | ψ ( t ) ⟩ for t > 0 using the time-evolution operator.

    2. Show that | c n | 2 is time-independent and interpret this physically.

    3. Write the expectation value ⟨ H ⟩ ( t ) and verify that it is constant in time.

  15. Consider a composite system made of two spin- 1 2 particles. The single-particle Hilbert space is spanned by { | ↑ ⟩ , | ↓ ⟩ } .

    1. Write an orthonormal product basis for the two-particle Hilbert space ℋ 1 ⊗ ℋ 2 .

    2. Consider the state

      | Ψ ⟩ = 1 2 ( | ↑ ⟩ ⊗ | ↓ ⟩ + | ↓ ⟩ ⊗ | ↑ ⟩ ) . (1.123)

      Show that | Ψ ⟩ cannot be written as a product | ψ 1 ⟩ ⊗ | ψ 2 ⟩ of single-particle states.

    3. Interpret the physical meaning of such a state in terms of quantum correlations (entanglement).