Commutation Relations in the Weyl Algebra

Q: Why do ∂_i and x_j commute when i ≠ j?

Because differentiating with respect to x_i treats x_j as a constant. Concretely ∂_i(x_j f) = x_j,∂ f/∂ x_i with no extra term, since ∂ x_j/∂ x_i = 0 . Only the diagonal case produces the identity operator.

Q: Is the constant in [∂, x] = 1 a choice?

Its value is, its non-vanishing is not. Rescaling ∂ by λ ≠ 0 turns the relation into [λ∂, x] = λ , so any non-zero scalar can be arranged, and physics arranges iℏ . What no rescaling can do is make the commutator zero, and every theorem about A_n depends only on that.

Q: Do the relations determine the algebra completely?

Yes in characteristic zero: A_n is the free associative algebra on x_1, … ,x_n,∂_1, … ,∂_n modulo the two-sided ideal generated by (1.1). That is the presentation theorem . It is what allows an algebra map out of A_n to be defined by choosing 2n images and verifying (1.1).

Q: How do I compute a commutator quickly?

Use adP as a derivation. Expand [P, QR] = [P,Q]R + Q[P,R] until only generators remain, then apply (1.1). Equivalently, if only the leading term is wanted, take the Poisson bracket of the symbols — but remember that this discards everything below the top degree.

Q: Why is the centre only the scalars?

Because commuting with ∂_i forces the symbol of an element to have zero derivative in x_i , commuting with x_i forces zero derivative in ξ_i , and in characteristic zero a polynomial in 2n variables with all partial derivatives zero is a constant. The characteristic-zero hypothesis is essential; in characteristic p the elements x_i^p and ∂_i^p are central.

Q: Does [P,Q] = 0 mean P and Q are polynomials in a common element?

Not in general, though it is close to true in A_1 . The centraliser of a non-scalar element of A_1 is a commutative subalgebra and is known to be a free module of finite rank over K[P] for suitable P , but the statement that any two commuting elements are polynomials in a common one is a delicate result and not something to assume in higher n .

Overview

The definition of P0 as a ring of operators leaves one question outstanding: how do the generators interact? Composition of operators is not commutative, so the order in which $x_i$ and $∂_j$ are applied matters, and the whole arithmetic of the algebra is determined by the size of the discrepancy.

The answer is as small as it could be without being zero. The generators commute in every pair except one: $∂_i$ and $x_i$ , the derivative and the multiplication in the same variable, fail to commute by exactly the identity operator. Everything else — the $x$ 's among themselves, the $∂$ 's among themselves, and $∂_i$ against $x_j$ for $i ≠ j$ — commutes.

That single unit of non-commutativity is responsible for essentially every distinctive property of the Weyl algebra: the existence of a canonical form, the fact that the algebra is simple, the absence of finite-dimensional representations, and the appearance of $A_n$ as the algebra of quantum mechanics. This page derives the relations from the product rule, assembles the identities they generate, and draws out the consequences that can be reached from the relations alone.

One of those consequences deserves advertising: the centre of $A_n$ is just $K$ . The proof is three lines once the relations are in place, and it is the point at which characteristic zero first becomes indispensable.

Definition

Let $R$ be any ring. The commutator of $P, Q ∈ R$ is defined by

[P,Q] = PQ - QP,

so $[P,Q] = 0$ says precisely that $P$ and $Q$ commute. In $A_n$ the product is composition of operators, so $[P,Q]$ is the operator $f ↦ P(Q(f)) - Q(P(f))$ .

The commutation relationsCoutinho, Ch. 1 §1

Let $K$ have characteristic zero and let $A_n = A_n(K)$ act on $K[X] = K[x_1, … ,x_n]$ as usual. Then for all $1 ≤ i, j ≤ n$ ,

[∂_i, x_j] = δ_ij, [x_i, x_j] = 0, [∂_i, ∂_j] = 0,

(1.1)

where $δ_ij$ is the Kronecker delta and the right-hand side of the first relation means $δ_ij$ times the identity operator.

Proof

The second relation holds because $K[X]$ is commutative: $x_i(x_j f) = x_i x_j f = x_j x_i f = x_j(x_i f)$ .

The third is the equality of mixed partial derivatives, which holds for polynomials over any commutative ring by direct computation on monomials: both $∂_i ∂_j$ and $∂_j ∂_i$ send $x^α$ to $α_i α_j, x^α - e_i - e_j$ when $α_i, α_j ≥ 1$ and $i ≠ j$ , and to $0$ otherwise.

For the first, apply the product rule. For any $f ∈ K[X]$ ,

∂_i(x_j f) = (∂ x_j)/(∂ x_i), f + x_j (∂ f)/(∂ x_i) = δ_ij f + x_j (∂ f)/(∂ x_i),

and the second term is $(x_j ∂_i)(f)$ . Subtracting, $(∂_i x_j - x_j ∂_i)(f) = δ_ij f$ for every $f$ , which is the assertion.

Remark

The $1$ on the right-hand side of $[∂_i, x_i] = 1$ is the identity operator, and its appearance is not a normalisation that could be scaled away by rescaling the generators. Replacing $∂_i$ by $λ ∂_i$ gives $[λ∂_i, x_i] = λ$ , so the constant can be changed to any non-zero scalar — but never to $0$ . It is the non-vanishing, not the value, that carries the content.

Core Concepts

The commutator as a derivative

The useful way to read $[P, -]$ is as a differentiation operator on the algebra. For fixed $P$ define $adP : A_n → A_n$ by $(adP)(Q) = [P,Q]$ . This map is $K$ -linear and satisfies the Leibniz rule

(adP)(QR) = ((adP)Q),R + Q,((adP)R),

so it is a derivation of $A_n$ — an inner derivation. Once this is noticed, commutators need never be computed by expanding products: they are computed by differentiating, one generator at a time.

What the derivation differentiates

Write an element of $A_n$ in canonical form $D = ∑ c_αβ x^α ∂^β$ and let $σ(D) = ∑ c_αβ x^α ξ^β$ be the commutative polynomial obtained by replacing each $∂_i$ by a commuting variable $ξ_i$ . Then the two families of inner derivations act on $σ(D)$ as literal partial derivatives:

σ([∂_i, D]) = (∂ σ(D))/(∂ x_i), σ([x_i, D]) = -,(∂ σ(D))/(∂ ξ_i).

These are exact identities, not statements about leading terms. They convert questions about commutators in a non-commutative algebra into questions about partial derivatives of ordinary polynomials, and they make the computation of the centre immediate.

How far the non-commutativity reaches

Because $[∂_i, x_j]$ is a scalar, taking a commutator lowers complexity. In the Bernstein filtration, where every generator has degree $1$ , one has $[B_r, B_s] ⊂ eq B_r+s-2$ : a commutator is two degrees cheaper than a product. In the order filtration, where $x_i$ has order $0$ and $∂_i$ order $1$ , one has $[F_r, F_s] ⊂ eq F_r+s-1$ . This drop is the reason that the associated graded algebra of $A_n$ is commutative, and that the leading term of a commutator is a Poisson bracket rather than a product.

Where the physics enters

Written for position and momentum, with $\hat{q}$ multiplication by the coordinate and $p = -iℏ,∂$ , relation (1.1) becomes $[ q, p] = iℏ$ . The Weyl algebra is the coordinate-free algebraic content of that statement, and the impossibility of finite matrices satisfying it is the same fact as the absence of finite-dimensional $A_n$ -modules. See the quantum origins page.

Construction and Proof

Three results follow from (1.1) with no further input. Each is proved here in full; each is used constantly later.

Commuting a derivative past a polynomialCoutinho, Ch. 1 Ex. 4.1

For every $f ∈ K[X]$ , viewed as a multiplication operator, $[∂_i, f] = ∂ f / ∂ x_i$ .

Proof

Both sides are $K$ -linear in $f$ , so it is enough to treat monomials, and by (1.3) the map $f ↦ [∂_i, f]$ satisfies

[∂_i, fg] = [∂_i, f],g + f,[∂_i, g],

which is the Leibniz rule. So $f ↦ [∂_i,f]$ and $f ↦ ∂ f/∂ x_i$ are two $K$ -derivations of $K[X]$ with values in $A_n$ . Two derivations that agree on a generating set of the algebra agree everywhere, and on generators both send $x_j$ to $δ_ij$ by (1.1). Hence they are equal.

Inner derivations act as partial derivatives on symbols

With $σ$ the symbol map of the Core Concepts section, $σ([∂_i, D]) = ∂σ(D)/∂ x_i$ and $σ([x_i,D]) = -∂σ(D)/∂ξ_i$ for every $D ∈ A_n$ .

Proof

By linearity it suffices to check on a basis element $D = x^α ∂^β$ . Since $∂_i$ commutes with every $∂_j$ , (1.3) gives $[∂_i, x^α∂^β] = [∂_i, x^α]∂^β$ , and the previous lemma evaluates $[∂_i, x^α]$ as $α_i x^α - e_i$ . That is exactly $∂(x^α ξ^β)/∂ x_i$ read back through $σ$ .

Similarly $[x_i, x^α ∂^β] = x^α [x_i, ∂^β]$ , and iterating $[x_i,∂_i] = -1$ through (1.3) gives $[x_i,∂^β] = -β_i ∂^β - e_i$ , which corresponds to $-∂(x^αξ^β)/∂ξ_i$ . Note that both results are exact: no lower-order terms are discarded.

The centre of $A_n$ is $K$

If $K$ has characteristic zero then the centre of $A_n$ is $K$ : the only operators commuting with everything are the scalars.

Proof

Let $D$ be central. Then $[∂_i, D] = 0$ and $[x_i, D] = 0$ for every $i$ , so by the previous lemma all $2n$ partial derivatives of the polynomial $σ(D) ∈ K[x_1, … ,x_n,ξ_1, … ,ξ_n]$ vanish. Over a field of characteristic zero a polynomial with all partial derivatives zero is constant, so $σ(D) ∈ K$ and therefore $D ∈ K$ . Conversely scalars are obviously central.

The characteristic-zero hypothesis is used exactly once, and it cannot be dropped: in characteristic $p$ the polynomial $x_1^p$ has vanishing derivatives, and correspondingly $x_1^p$ and $∂_1^p$ are central in the Weyl algebra defined by generators and relations. This is the root of every failure described on the positive characteristic page.

Exponentiating the relationsCoutinho (3.2)

For every $i$ , the derivation $ad∂_i$ is locally nilpotent: applied enough times to any element of $A_n$ it gives zero, since each application lowers the total $x$ -degree of the symbol by one. Hence for $t ∈ K$ the formal exponential

(tad∂_i) = ∑ _k ≥ 0 (t^k)/(k!) (ad∂_i)^k

is a finite sum on each element and defines an algebra automorphism of $A_n$ with $x_i ↦ x_i + t$ and all other generators fixed. This is the translation automorphism, and it is the simplest case of Coutinho's construction of automorphisms of P2. Note that the operator $e^t∂_i$ itself does not lie in $A_n$ ; only the conjugation by it does.

Key Equations

The defining relations, and then the identities they generate. First, the relations themselves in the compact form used throughout:

[∂_i, x_j] = δ_ij, [x_i,x_j] = [∂_i,∂_j] = 0 (1 ≤ i,j ≤ n).

(1.1)

The commutator is bilinear and antisymmetric, and it obeys a Leibniz rule in each argument:

[P, λ Q + μ R] = λ[P,Q] + μ[P,R], [P,Q] = -[Q,P],

(1.2)

[P, QR] = [P,Q]R + Q[P,R], [PQ, R] = P[Q,R] + [P,R]Q.

(1.3)

Nested commutators satisfy the Jacobi identity, valid in any ring:

[P,[Q,R]] + [R,[P,Q]] + [Q,[R,P]] = 0.

(1.4)

Applying (1.3) repeatedly to (1.1) gives the two rules used in every hand computation:

[∂_i, f] = (∂ f)/(∂ x_i) (f ∈ K[X]), [x_i, ∂^β] = -β_i, ∂^β - e_i,

(1.5)

where $e_i$ is the $i$ -th standard basis vector of $N^n$ and the second expression is read as $0$ when $β_i = 0$ . The general reordering rule, which is what a computer algebra system implements, is the Leibniz expansion

∂^β f = ∑ _κ ≤ β βκ ( ∂^|κ| f∂ x^κ ) ∂^β - κ (f ∈ K[X]),

(1.6)

the sum running over multi-indices $κ$ with $κ_i ≤ β_i$ for all $i$ , and $βκ = ∏ _i β_iκ_i$ .

Finally, the Euler operators $θ_i = x_i ∂_i$ act on the generators by

[θ_i, x_j] = δ_ij x_j, [θ_i, ∂_j] = -δ_ij∂_j,

(1.7)

so that $θ = ∑ _i θ_i$ satisfies $[θ, x^α ∂^β] = (|α| - |β|), x^α∂^β$ and grades $A_n$ by the integer $|α| - |β|$ .

Variable Definitions

$K$: the ground field, of characteristic zero
$A_n$: the $n$ -th Weyl algebra over $K$
$x_i$: the operator of multiplication by the variable $x_i$ on $K[X]$
$∂_i$: the operator of partial differentiation with respect to $x_i$
$[P,Q]$: the commutator $PQ - QP$
$ad P$: the inner derivation $Q ↦ [P,Q]$ of $A_n$
$δ_ij$: the Kronecker delta, $1$ if $i = j$ and $0$ otherwise
$e_i$: the $i$ -th standard basis vector of $N^n$ , so $β - e_i$ lowers the $i$ -th exponent by one
$σ(D)$: the commutative polynomial in $x_1, … ,x_n,ξ_1, … ,ξ_n$ obtained from the canonical form of $D$ by replacing $∂_i$ with $ξ_i$
$θ_i$ , $θ$: the Euler operators $x_i∂_i$ and their sum $∑ _i x_i∂_i$

Properties and Behaviour

The relations are complete

The relations (1.1) suffice: they generate every relation holding in $A_n$ . Precisely, the quotient of the free associative algebra on $2n$ generators by the two-sided ideal generated by the expressions in (1.1) is isomorphic to $A_n$ . This is the content of the presentation theorem, and it is what makes (1.1) a complete description rather than merely a true list.

No finite-dimensional representations

$A_n$ has no non-zero finite-dimensional modules. If $V$ were one, then $∂$ and $x$ would act by matrices whose commutator is the identity, and taking traces gives $0 = tr(∂ x - x∂) = tr(1) = _K V$ , so $V = 0$ . The argument needs characteristic zero — or at least that $_K V$ is not divisible by the characteristic. See the dedicated page.

Commutators are cheaper than products

Commutators drop filtration degree. With the Bernstein filtration $[B_r, B_s] ⊂ eq B_r+s-2$ , and with the order filtration $[F_r, F_s] ⊂ eq F_r+s-1$ . In both cases the associated graded algebra is therefore commutative, and the leading term of a commutator is computed by a Poisson bracket of symbols rather than by a product.

Relation to the Heisenberg Lie algebra

The $2n$ generators together with $1$ span a Lie algebra under the commutator: the $(2n+1)$ -dimensional Heisenberg Lie algebra $h_n$ , with $1$ spanning its centre. The Weyl algebra is the quotient of the universal enveloping algebra $U(h_n)$ by the two-sided ideal generated by $z - 1$ , where $z$ is the central generator. Every statement in this collection about $A_n$ can be translated into a statement about $U(h_n)$ -modules on which the centre acts by the scalar $1$ .

Symmetries of the relations

Relations (1.1) are invariant under more symmetry than is first apparent. The map $x_i ↦ ∂_i$ , $∂_i ↦ -x_i$ preserves them and extends to an automorphism of $A_n$ — the algebraic Fourier transform. So does the map induced by any element of $SL(2,K)$ acting on the pair $(x,∂)$ when $n = 1$ . These are Coutinho's exercises 4.7 and 4.8, and they are the reason $x$ and $∂$ can often be interchanged in proofs.

Examples and Special Cases

Moving a single generator through a power

Repeated use of $∂ x = x∂ + 1$ gives $∂^b x = x∂^b + b,∂^b-1$ and $x^a ∂ = ∂ x^a - a x^a-1$ . Both are the $n = 1$ case of (1.5). For instance $∂^3 x = x∂^3 + 3∂^2$ , which can be checked on $x^k$ : the left-hand side gives $(k+1)k(k-1)x^k-2$ and the right-hand side gives $k(k-1)(k-2)x^k-2 + 3k(k-1)x^k-2 = k(k-1)(k+1)x^k-2$ .

The Leibniz expansion in a small case

Take $f = x^3$ and $β = 2$ in (1.6). The expansion gives $∂^2 x^3 = x^3∂^2 + 2(3x^2)∂ + 6x = x^3∂^2 + 6x^2∂ + 6x$ . Test on $x^k$ : the left-hand side is $(k+3)(k+2)x^k+1$ , and the right-hand side is $k(k-1)x^k+1 + 6kx^k+1 + 6x^k+1 = (k^2 + 5k + 6)x^k+1$ . The two agree.

The grading by the Euler operator

$[θ, x^a∂^b] = (a - b),x^a∂^b$ for $θ = x∂$ , so the eigenspaces of $ad θ$ decompose $A_1$ into $ℤ$ -graded pieces. The degree-zero piece is the commutative polynomial algebra $K[θ]$ , which is a maximal commutative subalgebra of $A_1$ and the natural home for hypergeometric-style computations.

Local nilpotence is special to the generators

The relations do not make $A_n$ a Lie-nilpotent or Lie-solvable algebra in any useful sense. Although $[∂, x] = 1$ is central, iterated commutators of general elements do not terminate. Since $[x^2∂, f] = x^2 f'$ for $f ∈ K[x]$ , applying $ad(x^2∂)$ repeatedly to $x$ gives $x^2$ , then $2x^3$ , then $6x^4$ , and so on without ever reaching zero. Only the derivations $adx_i$ and $ad∂_i$ are locally nilpotent.

Worked Example

Computing $[x^2∂,; x∂^2]$ in $A_1$ three ways

Step 1 - expand both products in canonical form
Write $P = x^2∂$ and $Q = x∂^2$ . For $P Q$ , the only reordering needed is $∂ x = x∂ + 1$ :

$PQ = x^2∂ ⋅ x∂^2 = x^2(∂ x)∂^2 = x^2(x∂ + 1)∂^2 = x^3∂^3 + x^2∂^2.$

For $Q P$ the inner block is $∂^2 x^2$ , which the Leibniz expansion (1.6) evaluates as $x^2∂^2 + 4x∂ + 2$ :

$QP = x∂^2 ⋅ x^2 ∂ = x(∂^2 x^2)∂ = x(x^2∂^2 + 4x∂ + 2)∂ = x^3∂^3 + 4x^2∂^2 + 2x∂.$
Step 2 - subtract
The cubic terms cancel, which they must, because commutators lose degree:

$[P,Q] = PQ - QP = -3x^2∂^2 - 2x∂.$
(1.8)

In Bernstein degree, $P$ and $Q$ both lie in $B_3$ , their product lies in $B_6$ , and the commutator lies in $B_4 = B_3+3-2$ , exactly as $[B_r,B_s] ⊂ eq B_r+s-2$ predicts.
Step 3 - check by acting on a general monomial
Since elements of $A_1$ are operators, (1.8) can be verified outright. On $x^k$ one has $P(x^k) = kx^k+1$ and $Q(x^k) = k(k-1)x^k-1$ , so

$PQ(x^k) = k(k-1) ⋅ (k-1),x^k = k(k-1)^2 x^k, QP(x^k) = k ⋅ (k+1)k,x^k = k^2(k+1)x^k.$

Their difference is $k [(k-1)^2 - k(k+1) ]x^k = k(-3k+1)x^k = (-3k^2 + k)x^k$ . Applying the right-hand side of (1.8) to $x^k$ gives $-3k(k-1)x^k - 2kx^k = (-3k^2 + 3k - 2k)x^k = (-3k^2+k)x^k$ . The two agree for every $k ≥ 0$ , so (1.8) is an identity in $A_1$ .
Step 4 - predict the leading term with a Poisson bracket
The symbols are $σ(P) = x^2ξ$ and $σ(Q) = xξ^2$ . Because the commutator drops one order, its symbol is the Poisson bracket

$σ(P), σ(Q) = (∂ σ(P))/(∂ ξ)(∂ σ(Q))/(∂ x) - (∂ σ(P))/(∂ x)(∂ σ(Q))/(∂ ξ) = x^2 ⋅ ξ^2 - 2xξ ⋅ 2xξ = -3x^2ξ^2.$

That matches the leading term $-3x^2∂^2$ of (1.8), and it was obtained without a single reordering. The Poisson bracket predicts the top of a commutator; it says nothing about the term $-2x∂$ below it.
Step 5 - sanity-check with the grading
Under the Euler grading of (1.7), $P = x^2∂$ has weight $2 - 1 = 1$ and $Q = x∂^2$ has weight $1 - 2 = -1$ . Weights add under multiplication, so $[P,Q]$ must be homogeneous of weight $0$ . Both terms of (1.8), $x^2∂^2$ and $x∂$ , have weight $0$ . Had the answer contained a term of any other weight, it would have been wrong.

Result

$[x^2∂,; x∂^2] = -3x^2∂^2 - 2x∂$ , confirmed by direct expansion, by acting on the general monomial $x^k$ (both sides multiply it by $-3k^2 + k$ ), by the Poisson bracket for the leading term, and by weight considerations for the shape of the answer.

Applications and Industry Use

In a mathematics topic, this section covers downstream use inside mathematics, computing and engineering rather than a manufactured product.

Quantum mechanics. Relation (1.1) is the canonical commutation relation, and the structural facts derived here — trivial centre, no finite-dimensional representations — are the algebraic form of the uncertainty principle and of the Stone-von Neumann uniqueness theorem. See the quantum origins page.
Symbolic computation. Every implementation of the Weyl algebra reduces products to canonical form using (1.6); the relations are literally the rewriting rules of the data structure. The same applies to the Ore algebras used for difference and $q$ -difference equations, where (1.1) is replaced by a twisted analogue.
Constructing automorphisms and modules. Because (1.1) is a complete set of relations, a map out of $A_n$ can be defined by naming the images of the $2n$ generators and checking (1.1) for the images. This is how automorphisms are built and how module actions are specified.
Microlocal analysis and symplectic geometry. The drop in degree under commutators, and the resulting Poisson bracket on symbols, is the algebraic origin of the symplectic structure on the cotangent bundle and of Gabber's involutivity theorem.
Representation theory. Realising $A_n$ as $U(h_n)/(z-1)$ places it inside Lie theory, where the same relations govern oscillator representations, Fock spaces and the metaplectic representation.

Standards and Codes

For mathematics, the relevant standards are notation, numeric and markup standards together with reference implementations.

ISO 80000-2 fixes $∂$ for partial differentiation, $δ_ij$ for the Kronecker delta, and upright roman type for named operators such as $dim$ and $tr$ , all used above.
ISO/IEC 40314 (MathML 3.0) is the encoding used for the expressions on this page, which keeps commutator brackets machine-readable rather than rendering them as images.
There is no standard sign convention. Coutinho and most of the D-module literature write $[∂_i, x_j] = δ_ij$ ; the physics literature writes $[ q, p] = iℏ$ , which is the opposite ordering with an extra scalar. Both are consistent internally, and neither is more correct — but they must not be mixed within a computation.
Software conventions differ on the side of the canonical form. Macaulay2 and SageMath normalise with the multiplication operators on the left, matching this page; some interfaces to Singular use the opposite order for internal representation.
Result citation follows Coutinho's numbering, so the relations are Ch. 1 §1 and the exponentiated automorphism is (3.2) in that text.

Computational Notes

Read this as the manufacturing section of the template: how the object is actually built by machine, at what cost, and where the computation stops being decidable.

The relations are executable. A product of two canonical forms is computed by applying (1.6) to each pair of basis monomials and merging the results.

For a single pair $x^α∂^β ⋅ x^γ∂^δ$ the expansion produces $∏ _i ( (β_i,γ_i) + 1)$ terms, each with an integer coefficient $β_iκ_i γ_iκ_iκ_i!$ in each variable. Two practical consequences follow. First, the term count is worst when the two operands have comparable derivative and multiplication degrees, so it pays to reduce operands before multiplying. Second, the coefficients grow factorially, so over $ℚ$ the bottleneck is usually integer size rather than the number of terms.

Commutators specifically are cheaper than the difference of two products suggests, because the top-degree terms cancel. A good implementation computes $[P,Q]$ by expanding $ad P$ over the generators using (1.3) rather than by forming $P Q$ and $Q P$ separately, which avoids constructing and then cancelling the largest terms.

Systems implementing these rules directly include the WeylAlgebra construction in Macaulay2, nctools.lib and dmod.lib in Singular, DifferentialWeylAlgebra and ore_algebra in SageMath, and HolonomicFunctions.m in Mathematica. All fix a side for the canonical form; check which before comparing coefficients with a printed formula.

Limits of Validity

Relations (1.1) are true over any commutative ground ring; what fails outside characteristic zero is not the relations but the conclusions drawn from them.

Centre. The proof that the centre is $K$ needs a polynomial with vanishing partial derivatives to be constant, which fails in characteristic $p$ . There $x_i^p$ and $∂_i^p$ are central and $A_n$ is a finitely generated module of rank $p^2n$ over its centre.
No finite-dimensional modules. The trace argument gives $_K V ⋅ 1_K = 0$ , which forces $V = 0$ only when the characteristic does not divide $_K V$ . In characteristic $p$ there are modules of dimension $p$ , and indeed of dimension $p^n$ .
Completeness of the relations. That (1.1) generates every relation is a theorem about the characteristic-zero algebra. Over a field of characteristic $p$ the algebra of operators on $K[X]$ satisfies the extra relation $∂_i^p = 0$ , which is not a consequence of (1.1).
Scalars. The relations are stated over a field. Over a commutative ring $R$ containing $ℚ$ they still hold, but the centre is then $R$ rather than a field, and ideals of $R$ generate proper two-sided ideals, so simplicity is lost.

Deformations are different algebras

The relations also have $q$ -deformations, in which $∂ x - q,x∂ = 1$ or $yx = qxy$ . These generate genuinely different algebras with different ideal structure, and no theorem on this page should be assumed to survive the deformation. The behaviour usually depends on whether $q$ is a root of unity.

Failure Modes and Common Mistakes

Reading $[∂,x] = 1$ as $∂ x = 1$

$[∂, x] = 1$ says that the operator $∂ x - x∂$ equals the identity operator. It does not say $∂ x = 1$ , and it does not say the two operators are equal up to a scalar factor. Applying $∂ x$ to a polynomial $f$ gives $xf' + f$ , not $f$ .

Getting the sign of the relation backwards

Both $[∂, x] = 1$ and $[x, ∂] = -1$ appear in the literature, and physics texts often write $[ q, p] = iℏ$ , which is the second ordering. A sign error here propagates into every subsequent commutator and typically produces a Poisson bracket of the wrong sign. Fix a convention at the start; this collection uses $[∂_i, x_j] = δ_ij$ , following Coutinho.

Treating a commutator as multiplicative

$ad P$ satisfies the Leibniz rule $[P,QR] = [P,Q]R + Q[P,R]$ , not the multiplicative rule $[P,QR] = [P,Q][P,R]$ . The mistake is easy to make when the commutators happen to be scalars, as in $[∂, x^2] = 2x$ : the correct route is $[∂,x]x + x[∂,x] = x + x = 2x$ , not any product of $[∂,x]$ with itself.

Mistaking the Poisson bracket for the commutator

The identity $σ([P,Q]) = σ(P),σ(Q)$ holds only for the top graded piece with respect to the filtration in use. In the worked example the Poisson bracket returned $-3x^2ξ^2$ and missed the term $-2x∂$ entirely. Using the bracket to compute an exact commutator, rather than its principal symbol, is a common and silent error.

Losing a term in the Leibniz expansion

Reordering identities such as (1.6) contain binomial coefficients and factorials, and a term with $κ = β$ contributing $∂^|β|f/∂ x^β$ is easy to drop. The safest check is the one used throughout this page: apply the claimed identity to the general monomial $x^k$ and compare polynomials in $k$ . If the two sides do not agree identically in $k$ , the reordering is wrong.

Historical Notes

The relation was written down in physics before it was studied in algebra. In 1925 Born and Jordan, developing Heisenberg's matrix mechanics, obtained the commutation rule between position and momentum matrices; Dirac reached the same rule independently and identified it as the quantum counterpart of the classical Poisson bracket, which is the correspondence that the Core Concepts section states algebraically.

The immediate difficulty was that no finite matrices satisfy the rule. Wintner and Wielandt later gave clean proofs that the relation is unsatisfiable by bounded operators on a Banach space, a stronger statement than the trace argument, and von Neumann and Stone established that the exponentiated relations have, up to unitary equivalence, only one irreducible representation. Weyl's 1928 book worked systematically with those exponentiated relations, which is why his name attached to the algebra.

The purely algebraic study — simplicity, the trivial centre, the automorphism group — belongs to the ring-theoretic tradition, with Littlewood in 1933 and then Dixmier's 1968 paper, which fixed both the name and the questions. Dixmier's conjecture that every endomorphism of $A_n$ is an automorphism, which is a statement about maps preserving (1.1), remains open and is now known to be equivalent to the Jacobian conjecture.

Comparison

Four algebras generated by two elements, distinguished only by the relation between them.
	$K[x,ξ]$	$A_1$	$U(h_1)$	$q$ -plane, $q$ not a root of unity
Defining relation	$ξ x = xξ$	$∂ x = x∂ + 1$	$∂ x = x∂ + z$ , $z$ central	$yx = q,xy$
Centre	the whole ring	$K$	$K[z]$	$K$
Simple?	no	yes	no	no; $(x)$ is a two-sided ideal
Finite-dimensional modules	many	none except zero	only those with $z$ acting as $0$	only those with $x$ or $y$ acting as $0$
Domain?	yes	yes	yes	yes
Relation to $A_1$	associated graded ring of $A_1$	—	$A_1 = U(h_1)/(z-1)$	a deformation, not a degeneration

Key Takeaways

Key points

The generators of $A_n$ satisfy $[∂_i,x_j] = δ_ij$ and $[x_i,x_j] = [∂_i,∂_j] = 0$ ; only a derivative and its own variable fail to commute, and they fail by the identity.
All three relations are consequences of the product rule and the commutativity of $K[X]$ , so they are computed rather than postulated.
$adP = [P,-]$ is a derivation, so commutators obey the Leibniz rule and the Jacobi identity; $ad∂_i$ and $adx_i$ act on symbols as $∂/∂ x_i$ and $-∂/∂ξ_i$ .
Consequently the centre of $A_n$ is exactly $K$ in characteristic zero, and $A_n$ has no non-zero finite-dimensional modules.
Commutators lose filtration degree — $[B_r,B_s] ⊂ eq B_r+s-2$ — which makes the associated graded algebra commutative and turns the top term of a commutator into a Poisson bracket.
The relations are complete: $A_n$ is the free algebra on $2n$ generators modulo the ideal they generate, which is what licenses defining maps out of $A_n$ by checking (1.1) alone.

FAQs

Why do

∂_i

and

x_j

commute when

i ≠ j

Because differentiating with respect to $x_i$ treats $x_j$ as a constant. Concretely $∂_i(x_j f) = x_j,∂ f/∂ x_i$ with no extra term, since $∂ x_j/∂ x_i = 0$ . Only the diagonal case produces the identity operator.

Is the constant in

[∂, x] = 1

a choice?

Its value is, its non-vanishing is not. Rescaling $∂$ by $λ ≠ 0$ turns the relation into $[λ∂, x] = λ$ , so any non-zero scalar can be arranged, and physics arranges $i ℏ$ . What no rescaling can do is make the commutator zero, and every theorem about $A_n$ depends only on that.

Do the relations determine the algebra completely?

Yes in characteristic zero: $A_n$ is the free associative algebra on $x_1, … ,x_n,∂_1, … ,∂_n$ modulo the two-sided ideal generated by (1.1). That is the presentation theorem. It is what allows an algebra map out of $A_n$ to be defined by choosing $2n$ images and verifying (1.1).

How do I compute a commutator quickly?

Use $ad P$ as a derivation. Expand $[P, QR] = [P,Q]R + Q[P,R]$ until only generators remain, then apply (1.1). Equivalently, if only the leading term is wanted, take the Poisson bracket of the symbols — but remember that this discards everything below the top degree.

Why is the centre only the scalars?

Because commuting with $∂_i$ forces the symbol of an element to have zero derivative in $x_i$ , commuting with $x_i$ forces zero derivative in $ξ_i$ , and in characteristic zero a polynomial in $2n$ variables with all partial derivatives zero is a constant. The characteristic-zero hypothesis is essential; in characteristic $p$ the elements $x_i^p$ and $∂_i^p$ are central.

Does

[P,Q] = 0

mean

P

and

Q

are polynomials in a common element?

Not in general, though it is close to true in $A_1$ . The centraliser of a non-scalar element of $A_1$ is a commutative subalgebra and is known to be a free module of finite rank over $K[P]$ for suitable $P$ , but the statement that any two commuting elements are polynomials in a common one is a delicate result and not something to assume in higher $n$ .

Why does the trace argument rule out finite-dimensional representations?

The trace of any commutator of finite matrices is zero, while the trace of the identity on a $d$ -dimensional space is $d$ . Relation (1.1) would give $0 = d$ in $K$ , impossible in characteristic zero unless $d = 0$ . In characteristic $p$ the argument only fails to exclude $d$ divisible by $p$ , and indeed such representations exist.

What is the relationship to the Poisson bracket?

Taking the commutator lowers filtration degree by one (order filtration) or two (Bernstein filtration), so the operation induced on the associated graded ring is not multiplication but a bracket, and it is exactly the Poisson bracket of the symbols. This is the algebraic form of Dirac's quantisation correspondence; see the symbol algebra page.

References

S. C. Coutinho, A Primer of Algebraic D-modules, London Mathematical Society Student Texts 33, Cambridge University Press, 1995 - Ch. 1 §1 for the relations, Ex. 4.1, 4.3, 4.5 for the commutator identities, and (3.2) for the exponentiated automorphism.
J. Dixmier, Sur les algèbres de Weyl, Bulletin de la Société Mathématique de France 96 (1968), 209-242.
M. Born and P. Jordan, Zur Quantenmechanik, Zeitschrift für Physik 34 (1925), 858-888 - where the commutation rule first appears.
P. A. M. Dirac, The fundamental equations of quantum mechanics, Proceedings of the Royal Society A 109 (1925), 642-653 - the commutator and Poisson bracket correspondence.
H. Weyl, Gruppentheorie und Quantenmechanik, Hirzel, Leipzig, 1928.
J. C. McConnell and J. C. Robson, Noncommutative Noetherian Rings, Graduate Studies in Mathematics 30, revised edition, American Mathematical Society, 2001 - Ch. 1, for the presentation and the centre.
J.-E. Björk, Rings of Differential Operators, North-Holland Mathematical Library 21, North-Holland, 1979 - Ch. 1, for filtrations and the symbol calculus.
ISO 80000-2:2019, Quantities and units - Part 2: Mathematics, International Organization for Standardization.
ISO/IEC 40314:2016, Information technology - Mathematical Markup Language (MathML) Version 3.0, International Organization for Standardization.

AI Suggested Questions

Prove $[x_i, ∂^β] = -β_i∂^β - e_i$ by induction, using only the Leibniz rule for commutators.
Verify the Leibniz expansion (1.6) for $β = 3$ and $f = x^2$ by acting on the general monomial $x^k$ .
Compute $[x∂^2,, x^3∂]$ in $A_1$ and check the leading term with a Poisson bracket.
Show that $ad(x^2∂)$ is not locally nilpotent, and identify which elements it does annihilate.
Give the details of the trace argument and explain exactly which step fails in characteristic $p$ .
Show that $x ↦ ∂$ , $∂ ↦ -x$ preserves the relations and hence defines an automorphism of $A_1$ .
Determine the centraliser of $x∂$ in $A_1$ using the Euler grading.

Commutation Relations in the Weyl Algebra

Overview

Definition

Core Concepts

The commutator as a derivative

What the derivation differentiates

How far the non-commutativity reaches

Where the physics enters

Construction and Proof

Key Equations

Variable Definitions

Properties and Behaviour

Examples and Special Cases

Worked Example

Applications and Industry Use

Standards and Codes

Computational Notes

Limits of Validity

Failure Modes and Common Mistakes

Historical Notes

Comparison

Key Takeaways

Key points

FAQs

Related Engineering Topics

References

AI Suggested Questions

Continue learning