Grassmann Variables and Berezin Integration

⚡ Key Takeaways

Grassmann variables are anticommuting generators ( $\theta_i\theta_j = -\theta_j\theta_i$ ) of an exterior (Grassmann) algebra, and Berezin integration is the unique linear functional on that algebra that behaves like both differentiation and integration. Key result: $\int d\theta\, \theta = 1$ , $\int d\theta\,1 = 0$ , and for a quadratic form $\int d^n\bar\theta\, d^n\theta\, e^{-\bar\theta^T A\theta} = \det A$ (note: determinant, not its inverse). Applies to: fermionic path integrals, supersymmetry, supermanifolds, random matrix theory, and the algebraic theory of determinants. Who/when: Hermann Grassmann (1844); formal integration by Felix Berezin (1966).

💡 The Core Insight

A Grassmann algebra is finite-dimensional as a vector space because the generators square to zero: any function of $n$ Grassmann variables is a polynomial whose highest term is $\theta_1\theta_2\cdots\theta_n$ . Consequently the only meaningful linear operation respecting translation invariance is to pick off the top coefficient — integration = extracting the coefficient of the top-form. The same operation differentiates. This collapse of analysis into algebra is what makes fermionic path integrals exact and combinatorial.

DefinitionGrassmann Algebra

Let $V$ be an $n$ -dimensional vector space over a field $K$ (usually $\mathbb{R}$ or $\mathbb{C}$ ) with basis $\{\theta_1,\dots,\theta_n\}$ . The Grassmann algebra (exterior algebra) $\Lambda(V)=\bigoplus_{k=0}^{n}\Lambda^k(V)$ is the associative unital $K$ -algebra generated by the $\theta_i$ subject to:

1

\quad \theta_i\theta_j + \theta_j\theta_i = 0 \quad\text{for all } i,j, \qquad \text{in particular } \theta_i^2 = 0.

Derivation hint: Impose the anticommutation relation as the defining quotient of the free tensor algebra $T(V)$ by the two-sided ideal generated by $\theta_i\otimes\theta_j+\theta_j\otimes\theta_i$ . A general element ("superfunction") has the finite expansion

2

\quad f(\theta) = \sum_{k=0}^{n}\ \sum_{i_1<\cdots<i_k} c_{i_1\cdots i_k}\,\theta_{i_1}\cdots\theta_{i_k}, \qquad c\in K,

Derivation hint: Reorder any monomial using [1]; repeated indices vanish by nilpotency, so distinct ordered index sets form a basis of $\dim\Lambda(V)=2^n$ . The algebra is $\mathbb{Z}_2$ -graded: $f = f_{\text{even}} + f_{\text{odd}}$ , where even (commuting) elements form the body+nilpotent center and odd elements anticommute among themselves.

DefinitionGrassmann Differentiation

The left derivative $\partial/\partial\theta_i$ is the odd linear operator defined on monomials by anticommuting $\theta_i$ to the far left and stripping it:

3

\quad \frac{\partial}{\partial\theta_i}\big(\theta_{j_1}\cdots\theta_{j_k}\big) = \sum_{a=1}^{k} (-1)^{a-1}\,\delta_{i j_a}\,\theta_{j_1}\cdots\widehat{\theta_{j_a}}\cdots\theta_{j_k},

Derivation hint: The sign $(-1)^{a-1}$ counts the transpositions needed to move $\theta_{j_a}$ past $\theta_{j_1},\dots,\theta_{j_{a-1}}$ ; the hat denotes omission. These satisfy the graded Leibniz rule and the nilpotency $\big(\partial/\partial\theta_i\big)^2=0$ and anticommutation $\{\partial_{\theta_i},\partial_{\theta_j}\}=0$ , plus $\{\partial_{\theta_i},\theta_j\}=\delta_{ij}$ .

DefinitionBerezin Integral

The Berezin integral $\int d\theta_i$ over a single Grassmann variable is the unique (up to normalization) linear functional satisfying translation invariance $\int d\theta\, f(\theta+\eta)=\int d\theta\, f(\theta)$ . It is fixed by:

4

\quad \int d\theta_i\, 1 = 0, \qquad \int d\theta_i\, \theta_i = 1.

Derivation hint: Translation invariance forces $\int d\theta\,1=0$ ; the normalization $\int d\theta\,\theta=1$ is convention. Hence Berezin integration coincides with left differentiation: $\int d\theta_i = \partial/\partial\theta_i$ acting from the left. For multiple variables one orders the measures and integrates from right to left:

5

\quad \int d^n\theta\ \theta_1\theta_2\cdots\theta_n = 1, \qquad d^n\theta := d\theta_n\,d\theta_{n-1}\cdots d\theta_1,

Derivation hint: Apply [4] iteratively, $\int d\theta_n\cdots\int d\theta_1$ , each acting on the leftmost remaining factor; the prescribed ordering gives $+1$ with no extra sign. Crucially, the measure carries odd parity: $d\theta_i$ anticommutes with all Grassmann generators and with other measures, $d\theta_i\,d\theta_j = -d\theta_j\,d\theta_i$ .

TheoremBerezin Gaussian Integral

Let $A$ be an $n\times n$ matrix and $\{\theta_i,\bar\theta_i\}$ be $2n$ independent Grassmann generators. Then

6

\quad \int \prod_{i=1}^n d\bar\theta_i\, d\theta_i\ \exp\!\Big(-\sum_{i,j}\bar\theta_i A_{ij}\theta_j\Big) = \det A.

Derivation hint: Expand the exponential — it terminates because each $\theta_i,\bar\theta_i$ appears at most once. Only the top term $\frac{1}{n!}(-\bar\theta A\theta)^n$ survives integration; collecting it gives $\sum_{\sigma}\mathrm{sgn}(\sigma)\prod_i A_{i\sigma(i)}=\det A$ . For a single set of $n$ Grassmann variables with antisymmetric $A=-A^T$ ,

7

\quad \int d^n\theta\ \exp\!\Big(\tfrac{1}{2}\theta^T A\,\theta\Big) = \mathrm{Pf}(A), \qquad \mathrm{Pf}(A)^2 = \det A,

Derivation hint: The top term is $\frac{1}{2^{n}n!}(\theta^T A\theta)^{n}$ ; matching against $\theta_1\cdots\theta_{2n}$ reproduces the Pfaffian's signed sum over perfect matchings.

ProofBerezin Gaussian = det A

Step 1. Write the bilinear exponent and expand. Because $\bar\theta_i^2=\theta_j^2=0$ , the Taylor series of the exponential truncates:

\exp(-\bar\theta A\theta)=\sum_{m=0}^{n}\frac{(-1)^m}{m!}\Big(\sum_{ij}\bar\theta_i A_{ij}\theta_j\Big)^m.

Step 2. The Berezin measure $\prod_i d\bar\theta_i d\theta_i$ returns zero on any term not containing every $\theta_i$ and every $\bar\theta_i$ exactly once (by [4]). Only the $m=n$ term qualifies.

Step 3. Expand $\big(\sum_{ij}\bar\theta_i A_{ij}\theta_j\big)^n$ . Surviving monomials use each $\bar\theta$ and each $\theta$ once, so the $\theta$ -indices and $\bar\theta$ -indices each range over a permutation. The combinatorial factor $n!$ cancels the $1/n!$ .

Step 4. Reordering the Grassmann factors into the canonical order $\bar\theta_1\theta_1\cdots\bar\theta_n\theta_n$ produces the signature $\mathrm{sgn}(\sigma)$ of the permutation relating $\theta$ -indices to $\bar\theta$ -indices. The $(-1)^n$ from the prefactor combines with the measure ordering to leave the standard determinant sign.

Step 5. Collecting, $\int = \sum_{\sigma\in S_n}\mathrm{sgn}(\sigma)\prod_i A_{i,\sigma(i)} = \det A.$ $\blacksquare$

∎

ProofChange of Variables / Jacobian Inversion

Let $\theta_i = \sum_j M_{ij}\theta'_j$ be an invertible linear change of Grassmann variables. The top-form transforms as $\theta_1\cdots\theta_n = (\det M)\,\theta'_1\cdots\theta'_n$ .

Step 1. Since integration extracts the coefficient of the top monomial, $\int d^n\theta\, f =$ (coeff of $\theta_1\cdots\theta_n$ ).

Step 2. For $\int d^n\theta\, f = \int d^n\theta'\, f$ to give consistent values, the measure must transform inversely to the coordinates:

8

\quad d^n\theta = (\det M)^{-1}\, d^n\theta'.

Derivation hint: Demand $\int d^n\theta\,(\theta_1\cdots\theta_n)=\int d^n\theta'\,(\theta'_1\cdots\theta'_n)=1$ ; substitute the top-form scaling by $\det M$ and require cancellation. This is the opposite of the ordinary Jacobian rule ( $d^nx=|\det M|\,d^nx'$ ), and is why the Berezin–Gaussian integral yields $\det A$ rather than $(\det A)^{-1}$ . $\blacksquare$

∎

FA

Felix Alexandrovich Berezin

1931–1980

A Soviet mathematical physicist at Moscow State University, Berezin systematized the calculus of anticommuting variables in his 1966 monograph The Method of Second Quantization. He introduced what is now called the Berezin integral, the supertrace, the superdeterminant (Berezinian), and laid foundations for the geometry of supermanifolds and geometric quantization. His insistence that integration over fermions be defined purely algebraically — abandoning any measure-theoretic picture — proved exactly right for quantum field theory. Berezin drowned in a kayaking accident on the Kolyma River in 1980; the integral, the Berezinian, and Berezin–Toeplitz quantization carry his name.

🧠 Intuition

Think of a Grassmann variable as a "coordinate that has only two states": absent or present, with no notion of magnitude (since $\theta^2=0$ kills any second power). A function of $\theta$ is therefore linear: $f(\theta)=a+b\theta$ . There is nothing to integrate in the Riemann sense — there is no interval, no area under a curve. The only translation-invariant linear readout of $f$ is its slope $b$ . So " $\int d\theta$ " is just "read off the coefficient of $\theta$ ," which is the same as differentiating. Integration and differentiation being the same operation is the signature weirdness of Grassmann calculus. The exclusion principle ( $\theta^2=0$ : you cannot put two fermions in the same state) is built into the algebra, not imposed afterward.

Figure

Boson vs. Fermion Gaussian integration: same quadratic form, inverse powers of the determinant

ExampleOne-variable integral with explicit coefficients

Given: $f(\theta)=3 + 5\theta$ , a function of a single Grassmann variable.

Step 1. By linearity, $\int d\theta\, f(\theta) = 3\int d\theta\,1 + 5\int d\theta\,\theta$ .

Step 2. Apply rules [4]: $\int d\theta\,1=0$ and $\int d\theta\,\theta=1$ .

Step 3. Therefore $\int d\theta\, f(\theta) = 3\cdot 0 + 5\cdot 1 = 5$ .

\boxed{\int d\theta\,(3+5\theta) = 5 \quad(\text{dimensionless})}

Note this equals $f'(\theta)$ — integration picked off the coefficient of $\theta$ , identical to differentiation.

Example

2\times 2

Berezin Gaussian = determinant

Given: Two Grassmann pairs $(\theta_1,\theta_2,\bar\theta_1,\bar\theta_2)$ and matrix

A=\begin{pmatrix} 2 & 1\\ 3 & 4\end{pmatrix}, \qquad S=\bar\theta_1(2\theta_1+\theta_2)+\bar\theta_2(3\theta_1+4\theta_2).

Step 1. Compute $\int d\bar\theta_1 d\theta_1\, d\bar\theta_2 d\theta_2\, e^{-S}$ . Expand $e^{-S}=1 - S + \tfrac12 S^2$ (higher terms vanish, only $4$ generators).

Step 2. Only the term with all four generators survives. In $\tfrac12 S^2$ the surviving cross-terms are $\bar\theta_1\theta_1\bar\theta_2\theta_2$ -type. Collecting signs gives $\det A$ .

Step 3. Directly: $\det A = (2)(4)-(1)(3) = 8 - 3 = 5$ .

\boxed{\int d\bar\theta_1 d\theta_1\, d\bar\theta_2 d\theta_2\, e^{-\bar\theta A\theta} = \det A = 5 \quad(\text{dimensionless})}

Contrast: the bosonic Gaussian with the symmetric part of $A$ would give $(2\pi)^{1}/\sqrt{\det A}=2\pi/\sqrt 5$ , the inverse power.

⚠️ Integration is NOT the inverse of differentiation here

Although $\int d\theta = \partial/\partial\theta$ as operators, integration does not "undo" differentiation in the analytic sense. There is no constant of integration and no antiderivative. Likewise, do not import the absolute-value Jacobian $|\det M|$ from real analysis: the Grassmann measure transforms by $(\det M)^{-1}$ with sign (equation [8]), because the measure is itself an odd object that transforms inversely to the coordinates. Forgetting this inversion is the single most common error in fermionic path-integral calculations.

CounterexampleThe "obvious" measure scaling is wrong

Naively one might guess that rescaling $\theta\to\lambda\theta$ scales the measure the same way, $d\theta\to\lambda\,d\theta$ , as in ordinary calculus. This is false. Demand consistency: $1=\int d\theta\,\theta = \int d(\lambda\theta)\,(\lambda\theta)$ . If $d(\lambda\theta)=\lambda^{-1}d\theta$ , then $\lambda^{-1}\cdot\lambda\int d\theta\,\theta = 1$ , consistent. The measure scales as $\lambda^{-1}$ , inversely to the variable. A student who writes $d(\lambda\theta)=\lambda\,d\theta$ will obtain $\lambda^2$ and a wrong answer — and in the multivariate case will get $\det M$ instead of $(\det M)^{-1}$ , flipping the determinant to the bosonic side.

connection

Grassmann calculus is the algebraic core of several neighboring structures. The exterior algebra it defines is exactly the home of Differential Forms, where $d\theta$ plays the role of $dx^i$ and the wedge product realizes [1]. The Berezin–Gaussian formula [6] is the fermionic analogue of the Gaussian Integral and underlies the Path Integral Formulation of fermionic quantum fields. The superdeterminant (Berezinian) generalizes Determinant to graded matrices, and the whole apparatus is the local model for Supermanifolds and the foundation of Supersymmetry. The Pfaffian formula [7] connects directly to the combinatorics of perfect matchings and to Random Matrix Theory.

🔬 Where this appears

⏳ History

- 1844 — Hermann Grassmann publishes Die lineale Ausdehnungslehre, introducing the exterior (anticommutative) algebra of "extensive quantities," largely ignored for decades.

- 1888 — Giuseppe Peano and later Élie Cartan revive and formalize the exterior algebra and differential forms, giving

\theta_i\wedge\theta_j=-\theta_j\wedge\theta_i

geometric meaning.

- 1925–1928 — The Pauli exclusion principle and Dirac/Jordan–Wigner anticommutation relations

\{a_i,a_j^\dagger\}=\delta_{ij}

make anticommuting objects central to physics.

- 1933 — Eddington and others use Pfaffians; the determinant–Pfaffian relation

\mathrm{Pf}(A)^2=\det A

becomes standard.

- 1966 — Felix Berezin's The Method of Second Quantization defines integration over anticommuting variables, equations [4]–[6], inventing the modern Berezin integral.

- 1971–1975 — Salam–Strathdee, Wess–Zumino, and Akulov–Volkov develop superspace; Grassmann coordinates become physical superspace dimensions.

- 1975–1977 — Faddeev, Popov, and others recast gauge-theory quantization with Grassmann ghost fields; Berezin integration becomes routine in QFT.

- 1979–1983 — Efetov's supersymmetric method applies Berezin integration to disordered electronic systems and random matrices.

- 1987 — DeWitt's Supermanifolds and Rogers' work give a rigorous geometric framework, embedding Berezin integration in the theory of

\mathbb{Z}_2

-graded manifolds.

open-problem

Infinite-dimensional rigor. For finitely many generators, Berezin integration is purely algebraic and unproblematic. In quantum field theory one needs infinitely many Grassmann variables, and the resulting fermionic functional integral $\int\mathcal{D}\bar\psi\,\mathcal{D}\psi$ has no underlying measure-theoretic foundation comparable to the (already delicate) bosonic case; it is defined by its formal expansion and the determinant identity, with regularization required to make $\det(\slashed D+m)$ finite. Curved Grassmann measures. On a general supermanifold, defining a global, invariant Berezin volume element requires the Berezinian and a choice of "integral forms"; the theory of integration on supermanifolds (Voronov, Witten's recent work on superstring measures and integral forms) involves genuine subtleties absent in ordinary geometry — e.g., the failure of naive integration-by-parts and the role of the "picture-changing" structure. The sign problem. When $\det A$ is complex or negative, Monte Carlo methods built on Berezin integration lose their probabilistic interpretation; whether a general efficient algorithm exists is a deep open question linked to computational complexity.

Grassmann Variables and Berezin Integration

Formal Statement

Why It Holds / Derivation

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also