Special Symbols and Notation

Special symbols specify how an expression is assembled: whether it is a scalar, vector, operator, distribution, or operation on a composite system. This page is a lookup guide to those distinctions. A symbol’s local definition takes precedence over its usual appearance; in particular, an asterisk, a pair of brackets, or an equality-like sign can change meaning between sources.

The Symbol Index covers typography and general ambiguity checks. Use Latin Symbols, Greek Symbols, and Operators for named quantities. The dedicated ħ entry gives the reduced Planck constant’s units, value, and conversion rules. Here, the emphasis is on reading operations correctly and finding their full explanations.

Helpful background. Bra–Ket Notation fixes the scalar-product convention; Operator Conventions fixes adjoints and matrix elements; Tensor Product Ordering fixes subsystem labels and basis order.

Brackets, adjoints, and identity symbols

In the finite-dimensional examples below, matrix representations use ordered orthonormal bases, and the inner product is conjugate-linear in the bra and linear in the ket. Statements about products of unbounded operators require a common domain on which those products exist; formal symbol manipulation does not supply one.

Symbol or expression	Read as; mathematical role	Distinction that matters
$\hbar=h/(2\pi)$	“h-bar”; a scalar constant with units of action	$E=\hbar\omega$ uses angular frequency. For ordinary frequency, $E=h\nu$. See ħ.
$I$, $I_A$	Identity operator, on the full or indicated space	$I_A$ alone is not an operator on a bipartite space until its action there is specified, usually as $I_A\otimes I_B$.
$\lvert\psi\rangle$, $\langle\psi\rvert$	Ket and bra; a vector and its adjoint functional	$\langle\psi\rvert$ conjugates the ket’s coefficients. A generalized position ket is not a normalizable Hilbert-space vector.
$\langle\phi\mid\psi\rangle$, $\lvert\psi\rangle\langle\phi\rvert$	Inner product, a scalar; outer product, an operator	An arbitrary outer product is not a projector; for normalized $\lvert\psi\rangle$, $\lvert\psi\rangle\langle\psi\rvert$ is a rank-one orthogonal projector.
$\langle A\rangle_\rho$	Expectation of $A$ in state $\rho$	In finite dimensions this is $\operatorname{Tr}(\rho A)$; it is neither an eigenvalue label nor an unweighted average over a basis.
$A^\dagger$	“A dagger”; adjoint	In an orthonormal matrix representation, conjugate and transpose. Adjoint reverses product order: $(AB)^\dagger=B^\dagger A^\dagger$.
$A^T$, $A^*$	Transpose; entrywise complex conjugate in the convention used here	Neither alone is the adjoint in general. Some mathematics texts use $A^*$ for the adjoint instead.
$\lvert z\rvert$, $\lVert\psi\rVert$	Absolute value of a scalar; norm of a vector	$\lVert\psi\rVert^2=\langle\psi\mid\psi\rangle$. An operator norm needs a declared choice, such as spectral, trace, or Hilbert–Schmidt norm.
$\lvert0\rangle$ versus $0$	A labeled state versus a zero vector or zero operator	A normalized vacuum, ground state, or qubit state has norm one. It is not the zero vector.

Bra–Ket Notation develops the inner/outer-product rules; Operator Conventions explains representation and adjoint conventions. In a nonorthonormal basis, taking a conjugate transpose of the coordinate matrix is not generally sufficient to obtain the adjoint’s coordinate matrix: the inner-product metric must also be included.

Watrous’s The Theory of Quantum Information, Chapter 1, uses $A^*$ for the adjoint and an overbar for entrywise conjugation. When comparing its formulas with the table above, translate the operations as well as the glyphs.

A dagger changes both coefficients and direction

Let $A=i\lvert0\rangle\langle1\rvert$ in an orthonormal qubit basis. Then

$$A^\dagger=-i\lvert1\rangle\langle0\rvert,\qquad A^T=i\lvert1\rangle\langle0\rvert,\qquad A^*=-i\lvert0\rangle\langle1\rvert.$$

The adjoint reverses the direction of the map and conjugates its coefficient. A useful check is $A^\dagger A=\lvert1\rangle\langle1\rvert\ge0$. By contrast, $A^TA=-\lvert1\rangle\langle1\rvert$, so substituting a transpose for an adjoint destroys this positivity check. These formulas are a notation diagnostic; the general theory belongs to the linked operator treatment.

Products, brackets, and subsystem operations

Symbol or expression	Read as; defining operation	Scope or warning
$AB$	Composition: $B$ acts first on a ket	Requires compatible spaces and, for unbounded operators, appropriate domains.
${[A,B]=AB-BA}$	Commutator	A classical Poisson bracket is a different operation. Square brackets can also indicate an argument or equivalence class when explicitly defined.
${\{A,B\}=AB+BA}$	Anticommutator	Braces may instead delimit a set. In classical mechanics, braces commonly denote a Poisson bracket.
$\mathcal H_A\otimes\mathcal H_B$	Tensor-product space for a composite system	For finite dimensions $d_A,d_B$, the dimension is $d_Ad_B$. Factor order must be specified.
$\mathcal H_1\oplus\mathcal H_2$	Direct sum of orthogonal sectors	Its finite dimension is $d_1+d_2$. Sector structure is not the same as a pair of subsystems.
$\operatorname{Tr}A$	Trace; sum of diagonal matrix elements in an orthonormal basis	The result is a scalar and is basis independent. In infinite dimensions, trace-class hypotheses give the standard basis-independent trace.
$\operatorname{Tr}_B\rho_{AB}$	Partial trace over subsystem $B$	The result is an operator on $A$, not a scalar and not a state of $B$.
${A\ge0}$, ${A\ge B}$	Positive semidefinite operator; positive semidefinite difference $A-B$	These are operator inequalities, not comparisons of individual matrix entries.
$\mathcal T$	Time ordering	Orders later-time factors to the left in the usual evolution convention. It does not denote matrix transpose.

For the corresponding definitions and assumptions, see Commutator Identities, Direct Sums versus Tensor Products, Positive Operators, Partial Trace, and Time Ordering. Even when two spaces have the same dimension, a chosen tensor decomposition contains extra physical information that cannot be inferred from dimension alone. A direct-sum decomposition does not by itself forbid coherent superpositions between its summands; that would require an additional physical restriction.

Read the subscript on a partial trace

In orthonormal product bases, the elementary rule is

$$\operatorname{Tr}_B\!\left(|i\rangle_A\langle j|\otimes|\mu\rangle_B\langle\nu|\right) =\delta_{\mu\nu}|i\rangle_A\langle j|.$$

The $B$ indices are contracted; the $A$ indices survive. For example, with $|01\rangle=|0\rangle_A\otimes|1\rangle_B$,

$$|\Psi\rangle=\frac{\sqrt3\,|01\rangle+i|10\rangle}{2},\qquad \rho_A=\operatorname{Tr}_B|\Psi\rangle\langle\Psi| =\frac34|0\rangle\langle0|+\frac14|1\rangle\langle1|.$$

The off-diagonal joint terms vanish under this trace because $\langle0|1\rangle_B=0$. The global state is pure, but $\operatorname{Tr}\rho_A^2=5/8<1$: the reduced state is mixed. This operation describes the state available locally without conditioning on a measurement outcome. It is not the selection of one branch. Partial Trace Exercises gives further calculations, and Watrous, Chapter 2, develops the operational meaning of reduced states.

Discrete deltas, distributions, and comparison signs

Symbol	Meaning in the stated context	Check
$\delta_{mn}$	Kronecker delta: one for equal discrete indices, zero otherwise	It is dimensionless; $\sum_n\delta_{mn}c_n=c_m$.
${\delta(x-x')}$	Dirac delta distribution with respect to the indicated integration measure	For a length coordinate and measure $dx$, it has units of inverse length. It is not a pointwise probability density with a finite value at the origin.
$\equiv$, $:=$	Identity or definition; definition	Read the accompanying declaration. Neither establishes a physical law by itself.
$\approx$, $\simeq$	Approximation or another locally specified relation	The sign alone states neither an error bound nor a controlled limit.
${f(\epsilon)\sim g(\epsilon)}$	Asymptotic equivalence, when defined by $f/g\to1$ in a specified limit	In informal scaling estimates, $\sim$ may instead mean order of magnitude. Do not interchange these readings.
$O(\epsilon^n)$	A remainder bounded in magnitude by $C\lvert\epsilon\rvert^n$ near the specified limit	State which variables are fixed and whether the bound is uniform in other parameters. An asymptotic expansion need not converge.

For a smooth compactly supported test function $f$ on the real line and a real nonzero constant $a$, the distributional rules are

$$\int_{\mathbb R}\delta(x-x_0)f(x)\,dx=f(x_0),\qquad \delta(ax)=\frac{\delta(x)}{|a|}.$$

The absolute Jacobian is essential. Continuous basis normalization, such as $\langle x|x'\rangle=\delta(x-x')$, must be used together with its measure and Fourier convention. The delta-normalization treatment explains the distributional interpretation; Fourier-Transform Conventions fixes factors of $2\pi$ and $\hbar$. Do not apply a discrete sum rule to a continuous label without changing both the measure and normalization.

Diagnostic exercises

A positive matrix can have negative entries

In an orthonormal qubit basis, let $P=\tfrac12\begin{pmatrix}1&-1\\-1&1\end{pmatrix}$. Does the notation $P\ge0$ apply despite its negative off-diagonal entries?

Solution

Yes. $P=|-\rangle\langle-|$, where $|-\rangle=(|0\rangle-|1\rangle)/\sqrt2$. For every vector $|v\rangle$, $\langle v|P|v\rangle=|\langle-|v\rangle|^2\ge0$. Equivalently, its eigenvalues are zero and one. Operator positivity constrains quadratic forms or eigenvalues, not the sign of each entry.

A frequency change also changes the delta distribution

Let $\omega=2\pi\nu$ and $\omega_0=2\pi\nu_0$, with ordinary frequency $\nu$ and angular frequency $\omega$. Express $\delta(\omega-\omega_0)$ in terms of $\nu$ and verify the measure in a sifting integral.

Solution

The scaling rule gives $\delta(\omega-\omega_0)=\delta(\nu-\nu_0)/(2\pi)$ when its argument is expressed through $\nu$. Since $d\omega=2\pi\,d\nu$,

$$\int_{\mathbb R}d\omega\,\delta(\omega-\omega_0)F(\omega) =\int_{\mathbb R}d\nu\,\delta(\nu-\nu_0)F(2\pi\nu) =F(\omega_0).$$

Changing the variable while keeping both the old delta factor and the new measure would introduce an erroneous factor of $2\pi$. This is independent of the separate energy conversion $h\nu=\hbar\omega$.

References

• Bureau International des Poids et Mesures, “The International System of Units (SI): Defining constants”, accessed 2026-10-19 — the SI status of the Planck constant; the ħ card records the reduced constant and its dimensions.
• R. Shankar, Principles of Quantum Mechanics, 2nd ed., Springer, 1994, Chapter 1 — Dirac notation, linear operators, inner products, and continuous bases.
• J. Watrous, The Theory of Quantum Information, Cambridge University Press, 2018, Chapters 1–2 — matrix operations, positive operators, tensor products, traces, and reduced states. Its asterisk denotes the adjoint, unlike the entrywise-conjugation convention used in this page’s tables.