Skip to content

Bra-Ket Notation

Bra-ket notation is Dirac’s compact notation for vectors, dual vectors, inner products, outer products, operators, and matrix elements. It is notation, not a separate algebra. Every valid bra-ket expression has an underlying linear-algebra meaning.

The canonical translation is:

∣ψ⟩:vector,⟨ϕ∣:dual vector,⟨ϕ∣ψ⟩:scalar,∣ψ⟩⟨ϕ∣:operator,⟨ϕ∣A∣ψ⟩:scalar matrix element.\begin{array}{ccl} |\psi\rangle&:&\text{vector},\\ \langle\phi|&:&\text{dual vector},\\ \langle\phi|\psi\rangle&:&\text{scalar},\\ |\psi\rangle\langle\phi|&:&\text{operator},\\ \langle\phi|A|\psi\rangle&:&\text{scalar matrix element}. \end{array}

Use Dirac Notation as Linear Algebra for the mathematical construction and Representation Translation Table for the bra-ket, matrix, and wavefunction dictionary.

A ket ∣ψ⟩|\psi\rangle denotes a vector in a complex Hilbert space H\mathcal H. The symbol inside the ket is a label, not automatically a number. It may name:

  • a state, such as ∣ψ⟩|\psi\rangle;
  • a basis vector, such as ∣n⟩|n\rangle;
  • a set of quantum numbers, such as ∣nℓm⟩|n\ell m\rangle;
  • a subsystem value, such as ∣0⟩A|0\rangle_A;
  • a generalized eigenvalue, such as ∣x⟩|x\rangle or ∣p⟩|p\rangle.

Linear combinations are ordinary vector sums:

∣χ⟩=a∣ψ⟩+b∣ϕ⟩.|\chi\rangle =a|\psi\rangle+b|\phi\rangle.

Coefficients multiply kets from the left by convention. Writing ∣ψ⟩a|\psi\rangle a is avoided because operators also act from the left and the ordering would become ambiguous outside scalar algebra.

The bra associated with ∣ψ⟩|\psi\rangle is its adjoint:

⟨ψ∣=(∣ψ⟩)†.\langle\psi| =\left(|\psi\rangle\right)^\dagger.

If

∣ψ⟩=∑ncn∣n⟩,|\psi\rangle =\sum_n c_n|n\rangle,

then

⟨ψ∣=∑ncn∗⟨n∣.\langle\psi| =\sum_n c_n^*\langle n|.

Taking an adjoint reverses the order of products:

(A∣ψ⟩)†=⟨ψ∣A†.\left(A|\psi\rangle\right)^\dagger =\langle\psi|A^\dagger.

In a chosen orthonormal basis, kets may be represented by columns and bras by conjugate-transposed rows. That representation is useful, but the bra is not defined as a row of numbers independently of the inner product and basis.

Use the physics convention:

⟨ϕ∣ψ⟩.\langle\phi|\psi\rangle.

It is conjugate-linear in ϕ\phi and linear in ψ\psi. Thus

⟨aϕ1+bϕ2∣ψ⟩=a∗⟨ϕ1∣ψ⟩+b∗⟨ϕ2∣ψ⟩,⟨ϕ∣aψ1+bψ2⟩=a⟨ϕ∣ψ1⟩+b⟨ϕ∣ψ2⟩.\begin{aligned} \langle a\phi_1+b\phi_2|\psi\rangle &= a^*\langle\phi_1|\psi\rangle +b^*\langle\phi_2|\psi\rangle,\\ \langle\phi|a\psi_1+b\psi_2\rangle &= a\langle\phi|\psi_1\rangle +b\langle\phi|\psi_2\rangle. \end{aligned}

Conjugate symmetry gives

⟨ϕ∣ψ⟩=⟨ψ∣ϕ⟩∗.\langle\phi|\psi\rangle =\langle\psi|\phi\rangle^*.

The norm is

∥ψ∥=⟨ψ∣ψ⟩.\|\psi\| =\sqrt{\langle\psi|\psi\rangle}.

The detailed comparison with the mathematics convention belongs to Inner Product Conventions.

A normalized ket may represent a pure state, but the physical state is a ray:

∣ψ⟩∼eiα∣ψ⟩.|\psi\rangle \sim e^{i\alpha}|\psi\rangle.

An overall nonzero rescaling also leaves the ray unchanged before normalization. Once a normalized representative is chosen, only a phase freedom remains.

Relative phases are physical. The states

∣0⟩+∣1⟩2and∣0⟩−∣1⟩2\frac{|0\rangle+|1\rangle}{\sqrt2} \quad\text{and}\quad \frac{|0\rangle-|1\rangle}{\sqrt2}

are not related by one overall phase and generally give different interference probabilities.

See Rays and Global Phase for the canonical physical treatment.

The scalar ⟨ϕ∣ψ⟩\langle\phi|\psi\rangle is always an inner product when both vectors lie in the Hilbert space. It becomes a transition or measurement amplitude only after a physical question is specified.

If ∣ψ⟩|\psi\rangle is normalized and a projective measurement asks whether the state lies along normalized ∣ϕ⟩|\phi\rangle, then

p(ϕ∣ψ)=∣⟨ϕ∣ψ⟩∣2.p(\phi|\psi) =|\langle\phi|\psi\rangle|^2.

The vertical bar in p(ϕ∣ψ)p(\phi|\psi) denotes conditional notation; the vertical bars in ∣⟨ϕ∣ψ⟩∣|\langle\phi|\psi\rangle| denote absolute value; the vertical bars in ∣ψ⟩|\psi\rangle delimit a ket. Context and paired delimiters distinguish these uses.

For a degenerate outcome represented by projector PaP_a, the probability is

p(a∣ψ)=⟨ψ∣Pa∣ψ⟩,p(a|\psi) =\langle\psi|P_a|\psi\rangle,

not necessarily the squared overlap with one vector.

The outer product

∣ψ⟩⟨ϕ∣|\psi\rangle\langle\phi|

is an operator. It acts on ∣χ⟩|\chi\rangle as

(∣ψ⟩⟨ϕ∣)∣χ⟩=∣ψ⟩⟨ϕ∣χ⟩.\left( |\psi\rangle\langle\phi| \right)|\chi\rangle =|\psi\rangle \langle\phi|\chi\rangle.

The order matters. In general,

∣ψ⟩⟨ϕ∣≠∣ϕ⟩⟨ψ∣.|\psi\rangle\langle\phi| \ne |\phi\rangle\langle\psi|.

Their adjoints are related by

(∣ψ⟩⟨ϕ∣)†=∣ϕ⟩⟨ψ∣.\left( |\psi\rangle\langle\phi| \right)^\dagger =|\phi\rangle\langle\psi|.

An outer product is not an inner product:

⟨ϕ∣ψ⟩⏟scalar,∣ψ⟩⟨ϕ∣⏟operator.\underbrace{\langle\phi|\psi\rangle}_{\text{scalar}}, \qquad \underbrace{|\psi\rangle\langle\phi|}_{\text{operator}}.

If ∣ψ⟩|\psi\rangle is normalized, then

Pψ=∣ψ⟩⟨ψ∣P_\psi=|\psi\rangle\langle\psi|

is the orthogonal projector onto the one-dimensional subspace spanned by ∣ψ⟩|\psi\rangle. It satisfies

Pψ†=Pψ,Pψ2=Pψ.P_\psi^\dagger=P_\psi, \qquad P_\psi^2=P_\psi.

The projector is invariant under global phase:

(eiα∣ψ⟩)(eiα∣ψ⟩)†=∣ψ⟩⟨ψ∣.\left(e^{i\alpha}|\psi\rangle\right) \left(e^{i\alpha}|\psi\rangle\right)^\dagger =|\psi\rangle\langle\psi|.

For an orthonormal set spanning a subspace S\mathcal S,

PS=∑a∈S∣a⟩⟨a∣.P_{\mathcal S} =\sum_{a\in\mathcal S}|a\rangle\langle a|.

The result does not depend on which orthonormal basis is chosen inside S\mathcal S. See Projectors for the canonical operator treatment.

For an orthonormal basis {∣n⟩}\{|n\rangle\},

⟨m∣n⟩=δmn,∑n∣n⟩⟨n∣=I.\langle m|n\rangle=\delta_{mn}, \qquad \sum_n|n\rangle\langle n|=I.

Inserting the identity gives

∣ψ⟩=∑n∣n⟩⟨n∣ψ⟩=∑ncn∣n⟩,cn=⟨n∣ψ⟩.\begin{aligned} |\psi\rangle &=\sum_n|n\rangle\langle n|\psi\rangle\\ &=\sum_n c_n|n\rangle, \qquad c_n=\langle n|\psi\rangle. \end{aligned}

The bra expansion is

⟨ψ∣=∑ncn∗⟨n∣.\langle\psi| =\sum_n c_n^*\langle n|.

For a normalized state,

∑n∣cn∣2=1.\sum_n|c_n|^2=1.

Completeness must not be confused with normalization: the basis resolves the identity, while the state has unit norm.

Position and momentum “bases” are generalized spectral bases. Formally,

⟨x∣x′⟩=δ(x−x′),∫dx ∣x⟩⟨x∣=I.\langle x|x'\rangle=\delta(x-x'), \qquad \int dx\,|x\rangle\langle x|=I.

The position-space wavefunction is

ψ(x)=⟨x∣ψ⟩,\psi(x)=\langle x|\psi\rangle,

and the state is reconstructed formally as

∣ψ⟩=∫dx ∣x⟩ψ(x).|\psi\rangle =\int dx\,|x\rangle\psi(x).

The ket ∣x⟩|x\rangle is delta normalized and is not an ordinary normalizable Hilbert-space vector. These formulas are made precise through the spectral theorem or a rigged Hilbert space. Use Continuous Spectra and Rigged Hilbert Spaces: A First Look when that distinction matters.

Operators act on kets from the left:

A∣ψ⟩.A|\psi\rangle.

The product ABAB means first apply BB, then AA:

AB∣ψ⟩=A(B∣ψ⟩).AB|\psi\rangle=A(B|\psi\rangle).

In basis {∣n⟩}\{|n\rangle\}, the matrix elements are

Amn=⟨m∣A∣n⟩.A_{mn}=\langle m|A|n\rangle.

The operator can be expanded as

A=∑mn∣m⟩⟨m∣A∣n⟩⟨n∣A =\sum_{mn}|m\rangle \langle m|A|n\rangle \langle n|

for a finite-dimensional or suitably controlled discrete setting.

The expectation value in normalized ∣ψ⟩|\psi\rangle is

⟨A⟩ψ=⟨ψ∣A∣ψ⟩.\langle A\rangle_\psi =\langle\psi|A|\psi\rangle.

The subscript on ⟨A⟩ψ\langle A\rangle_\psi labels the state; the brackets here mean expectation value rather than a bra-ket pair.

Suppose ∣en⟩|e_n\rangle and ∣fα⟩|f_\alpha\rangle are orthonormal bases. Their overlap matrix is

Uαn=⟨fα∣en⟩.U_{\alpha n} =\langle f_\alpha|e_n\rangle.

Components transform as

⟨fα∣ψ⟩=∑n⟨fα∣en⟩⟨en∣ψ⟩.\langle f_\alpha|\psi\rangle =\sum_n \langle f_\alpha|e_n\rangle \langle e_n|\psi\rangle.

The abstract ket has not changed. Only its components in the chosen basis have changed. A physical unitary transformation of the state and a passive change of basis can use similar matrices; the surrounding statement must distinguish them.

Use subsystem labels when ambiguity is possible:

∣ψ⟩A⊗∣ϕ⟩B.|\psi\rangle_A\otimes|\phi\rangle_B.

The abbreviated form ∣ψ⟩A∣ϕ⟩B|\psi\rangle_A|\phi\rangle_B is allowed when the tensor product is unambiguous, but an ordinary product of kets is not defined within one Hilbert space.

An operator local to subsystem AA is

A⊗IB.A\otimes I_B.

Suppressing IBI_B is acceptable only after the subsystem action has been declared. Tensor-factor order follows Tensor Product Ordering.

For product bases,

∣i,j⟩≡∣i⟩A⊗∣j⟩B|i,j\rangle \equiv |i\rangle_A\otimes|j\rangle_B

only after that abbreviation and ordering have been declared.

  • ∣n⟩|n\rangle usually labels a discrete basis state, but the meaning of nn must be stated.
  • ∣x⟩|x\rangle and ∣p⟩|p\rangle denote generalized eigenkets.
  • ∣0⟩|0\rangle may mean a qubit basis state, oscillator ground state, or Fock vacuum.
  • ∣Ω⟩|\Omega\rangle often denotes a vacuum or distinguished ground state in many-body and field-theory contexts.
  • ∣ψ(t)⟩|\psi(t)\rangle is a time-dependent ket in the Schrödinger picture.
  • ∣a,λ⟩|a,\lambda\rangle uses multiple labels; commas do not imply tensor products.

Never infer the physical system from the typography alone.

  • Treating a bra as an unconjugated transpose.
  • Forgetting that adjoints reverse operator order.
  • Confusing the scalar ⟨ϕ∣ψ⟩\langle\phi|\psi\rangle with the operator ∣ψ⟩⟨ϕ∣|\psi\rangle\langle\phi|.
  • Reading every overlap as a probability amplitude without specifying a measurement.
  • Treating ∣ψ⟩|\psi\rangle and eiα∣ψ⟩e^{i\alpha}|\psi\rangle as different pure states while ignoring physically meaningful relative phases.
  • Inserting ∑n∣n⟩⟨n∣=I\sum_n|n\rangle\langle n|=I for an incomplete set.
  • Treating continuous kets as normalizable vectors.
  • Omitting tensor-product order or subsystem identity factors when ambiguity remains.
  • Reading AB∣ψ⟩AB|\psi\rangle as though AA acts first.
  • Using the same label ∣0⟩|0\rangle for different Hilbert spaces without a subsystem or context declaration.

Let

∣ψ⟩=16(∣0⟩+i∣1⟩−2∣2⟩).|\psi\rangle =\frac{1}{\sqrt6} \left( |0\rangle+i|1\rangle-2|2\rangle \right).

Write ⟨ψ∣\langle\psi| in the same orthonormal basis and verify normalization.

Solution

Taking the adjoint conjugates coefficients:

⟨ψ∣=16(⟨0∣−i⟨1∣−2⟨2∣).\langle\psi| =\frac{1}{\sqrt6} \left( \langle0|-i\langle1|-2\langle2| \right).

Orthonormality gives

⟨ψ∣ψ⟩=16(1+1+4)=1.\langle\psi|\psi\rangle =\frac{1}{6} \left(1+1+4\right) =1.

In the ordered basis {∣0⟩,∣1⟩}\{|0\rangle,|1\rangle\}, let

∣ψ⟩=∣0⟩+i∣1⟩2,∣ϕ⟩=∣0⟩.|\psi\rangle =\frac{|0\rangle+i|1\rangle}{\sqrt2}, \qquad |\phi\rangle=|0\rangle.

Find the matrix of ∣ψ⟩⟨ϕ∣|\psi\rangle\langle\phi| and its adjoint.

Solution

The column and row representations are

∣ψ⟩=12(1i),⟨ϕ∣=(10).|\psi\rangle =\frac{1}{\sqrt2} \begin{pmatrix} 1\\ i \end{pmatrix}, \qquad \langle\phi| = \begin{pmatrix} 1&0 \end{pmatrix}.

Therefore

∣ψ⟩⟨ϕ∣=12(10i0).|\psi\rangle\langle\phi| =\frac{1}{\sqrt2} \begin{pmatrix} 1&0\\ i&0 \end{pmatrix}.

Its adjoint is

∣ϕ⟩⟨ψ∣=12(1−i00).|\phi\rangle\langle\psi| =\frac{1}{\sqrt2} \begin{pmatrix} 1&-i\\ 0&0 \end{pmatrix}.

Exercise 3: Insert a resolution of identity

Section titled “Exercise 3: Insert a resolution of identity”

Show that

⟨ϕ∣A∣ψ⟩=∑mn⟨ϕ∣m⟩⟨m∣A∣n⟩⟨n∣ψ⟩\langle\phi|A|\psi\rangle =\sum_{mn} \langle\phi|m\rangle \langle m|A|n\rangle \langle n|\psi\rangle

for a complete orthonormal discrete basis.

Solution

Insert the identity on both sides of AA:

I=∑n∣n⟩⟨n∣.\displaystyle I=\sum_n|n\rangle\langle n|.

Then

⟨ϕ∣A∣ψ⟩=⟨ϕ∣IAI∣ψ⟩=∑mn⟨ϕ∣m⟩⟨m∣A∣n⟩⟨n∣ψ⟩.\begin{aligned} \langle\phi|A|\psi\rangle &=\langle\phi|IAI|\psi\rangle\\ &= \sum_{mn} \langle\phi|m\rangle \langle m|A|n\rangle \langle n|\psi\rangle. \end{aligned}

This is ordinary matrix multiplication written without choosing coordinate columns at the outset.

Let ∣ψ′⟩=eiα∣ψ⟩|\psi'\rangle=e^{i\alpha}|\psi\rangle. Show that the rank-one projector and all expectation values are unchanged.

Solution

The bra transforms as

⟨ψ′∣=e−iα⟨ψ∣.\langle\psi'| =e^{-i\alpha}\langle\psi|.

Hence

∣ψ′⟩⟨ψ′∣=eiαe−iα∣ψ⟩⟨ψ∣=∣ψ⟩⟨ψ∣.|\psi'\rangle\langle\psi'| =e^{i\alpha}e^{-i\alpha} |\psi\rangle\langle\psi| =|\psi\rangle\langle\psi|.

For any operator AA whose expectation is defined,

⟨ψ′∣A∣ψ′⟩=e−iαeiα⟨ψ∣A∣ψ⟩=⟨ψ∣A∣ψ⟩.\langle\psi'|A|\psi'\rangle =e^{-i\alpha}e^{i\alpha} \langle\psi|A|\psi\rangle =\langle\psi|A|\psi\rangle.
  • P. A. M. Dirac, The Principles of Quantum Mechanics, 4th ed., Oxford University Press, 1958.
  • J. von Neumann, Mathematical Foundations of Quantum Mechanics, Princeton University Press, 1955.
  • R. Shankar, Principles of Quantum Mechanics, 2nd ed., Springer, 1994.
  • J. J. Sakurai and J. Napolitano, Modern Quantum Mechanics, 3rd ed., Cambridge University Press, 2020.
  • B. C. Hall, Quantum Theory for Mathematicians, Springer, 2013.