Dirac Gamma Matrices and Clifford Algebra

When Dirac went looking for a relativistic wave equation that was first order in time — to escape the negative probability density of the Klein-Gordon equation — he needed an operator linear in the four-momentum, $i\gamma^\mu\partial_\mu - m$ , whose square would give back the relativistic energy-momentum relation $p^2 = m^2$ . Squaring that operator only works if the coefficients $\gamma^\mu$ obey $\gamma^\mu\gamma^\nu + \gamma^\nu\gamma^\mu = 2g^{\mu\nu}$ . For $\mu\neq\nu$ the right side is zero, so the $\gamma^\mu$ must anticommute — and no ordinary numbers anticommute. The coefficients are forced to be matrices, and the matrices force the wavefunction to have several components. This is where spinors come from: not from a postulate about spin, but from an algebraic demand. That single relation, $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}\mathbb{1}$ , is the defining relation of a Clifford algebra — the algebra you get when you insist a metric be encoded in an associative product of vectors. In four spacetime dimensions its smallest faithful representation is $4\times 4$ , which is exactly why the Dirac spinor has four components: two spin states times two signs of energy, the second pair tied to the negative-energy solutions this equation still carries. Because all the physics sits in the algebra and not in any particular matrices, you may pick whichever representation is convenient — Dirac (standard) for atoms, Weyl (chiral) for high energies, Majorana for real fields — and Pauli's fundamental theorem guarantees they give identical answers. Master this algebra and you have the vocabulary for everything downstream: the Dirac Hamiltonian, the electron's spin and $g=2$ , Lorentz covariance, the bilinear covariants, and the trace technology that turns QED cross sections into bookkeeping.

Key takeaways

The gamma matrices $\gamma^\mu$ are four matrices obeying $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}\mathbb{1}$ — the defining relation of the Clifford algebra of Minkowski space with signature $(+,-,-,-)$ .
This one relation, not any matrix choice, is what makes Dirac's operator $(i\gamma^\mu\partial_\mu - m)$ factor the Klein-Gordon operator so each spinor component satisfies $(\Box + m^2)\psi = 0$ .
In four dimensions the irreducible representation is $4\times 4$ : the algebra has $2^4=16$ independent elements and $4^2 = 16$ , forcing four-component spinors — arithmetic, not extra physics.
$\gamma^0$ is Hermitian, while the $\gamma^i$ are anti-Hermitian for the $(+,-,-,-)$ metric convention; together they define $\gamma^5=i\gamma^0\gamma^1\gamma^2\gamma^3$ , which anticommutes with all four $\gamma^\mu$ . The chiral projectors are $P_{R,L}=(1\pm\gamma^5)/2$ .
All trace and contraction identities — the workhorses of every scattering calculation — follow from the anticommutator alone, independent of representation.
Key idea: the physics lives in $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}\mathbb{1}$ , not in explicit matrices — this Clifford relation factorizes Klein-Gordon, forces a four-dimensional representation, and fixes every trace identity up to a similarity transformation.

The core insight

The physics lives in the anticommutation relation $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}\mathbb{1}$ , not in any explicit matrices: this single Clifford-algebra relation factorizes the Klein-Gordon operator, forces a four-dimensional representation (hence four-component spinors), and fixes every trace identity up to a similarity transformation.

Where this lives today

Gamma-matrix algebra is not a historical relic — it is the working language of every modern calculation involving fermions. Automated tools like FORM, FeynCalc, and Package-X exist essentially to do trace and contraction identities at industrial scale, evaluating the thousands of gamma strings that appear in multi-loop Standard Model predictions for the LHC. The chiral projectors $\tfrac12(1\pm\gamma^5)$ built from these matrices isolate left- and right-handed chiral components, which underlie the parity violation of the weak interaction, while the $d$ -dependence of the trace identities is central to dimensional regularization, where the " $\gamma^5$ problem" — how to consistently continue an intrinsically four-dimensional object to $d$ dimensions — remains a genuine subtlety in precision electroweak physics. In condensed matter, the same Clifford structure reappears: the low-energy electrons in graphene and in topological insulators obey Dirac equations, so their transport is computed with the very same gamma technology.

The physical picture

The cleanest way to feel where the gamma matrices come from is to think of them as square roots of the metric. With ordinary numbers, $\sqrt{a^2+b^2}$ has no exact linear form. But $\sqrt{p_0^2 - \mathbf{p}^2}$ does become linear, $\gamma^\mu p_\mu$ , provided the $\gamma$ 's are chosen so that the cross-terms cancel — and cross-terms cancel exactly when the $\gamma$ 's anticommute. The matrices exist for one reason: to make a square root of the relativistic energy relation linear in momentum. Once you accept that, the rest of the structure is the geometry of a metric written as an algebra. Symmetrizing two gammas gives back the metric, $\tfrac12\{\gamma^\mu,\gamma^\nu\} = g^{\mu\nu}\mathbb{1}$ ; antisymmetrizing them, $\tfrac12[\gamma^\mu,\gamma^\nu] = -i\sigma^{\mu\nu}$ , builds a new geometric object — the generators of Lorentz rotations and boosts acting on spinors. So one product splits into "the metric" and "the rotation generators," which is exactly the split you want relativistic physics to hand you.

Figure

Square root of the metric: the gamma factorization

The choice of representation is like a choice of coordinates: physical quantities such as eigenvalues and traces are representation-independent, while matrix elements themselves depend on the chosen basis. Counting fixes the dimension of the Clifford algebra: in four dimensions it has $2^4=16$ basis elements. An irreducible complex representation therefore acts on a $4$ -dimensional spinor space and can be realized using $4\times4$ matrices, whose $16$ matrix entries provide the required dimension of the full matrix algebra. The Dirac spinor consequently has four components, which may be organized into two spin degrees of freedom and particle/antiparticle sectors.

Figure

The 16-element basis and the 4×4 block structure

The core algebra

Everything in this article grows from one equation. Dirac wanted the operator $i\gamma^\mu\partial_\mu - m$ to square to the Klein-Gordon operator $-(\Box + m^2)$ so that every component of $\psi$ automatically obeys the relativistic wave equation. That single demand, worked out below, pins down the algebra the $\gamma^\mu$ must satisfy.

\{\gamma^\mu,\gamma^\nu\}\equiv\gamma^\mu\gamma^\nu+\gamma^\nu\gamma^\mu = 2g^{\mu\nu}\,\mathbb{1}_4 \qquad (1)

where: $\gamma^\mu$ are the four gamma matrices ( $\mu = 0,1,2,3$ ), each $4\times 4$ ; $g^{\mu\nu} = \mathrm{diag}(+1,-1,-1,-1)$ is the Minkowski metric (mostly-minus signature, the convention pinned for this whole chapter in "The Klein-Gordon Equation," with $\hbar = c = 1$ ); $\mathbb{1}_4$ is the $4\times 4$ identity. Read physically: the symmetric part of a gamma product is the metric, so $\gamma$ 's are "square roots of $g^{\mu\nu}$ ," and for $\mu\neq\nu$ the relation says they anticommute — the property no scalar has. Setting $\mu = \nu$ gives the diagonal statements, the ones you use constantly:

(\gamma^0)^2 = +\mathbb{1},\qquad (\gamma^i)^2 = -\mathbb{1},\qquad \gamma^\mu\gamma^\nu = -\gamma^\nu\gamma^\mu\ (\mu\neq\nu) \qquad (2)

where: the time component squares to $+\mathbb{1}$ and each space component to $-\mathbb{1}$ , tracking the metric signs. This is already enough to see that $\gamma^i$ has eigenvalues $\pm i$ (so it is naturally anti-Hermitian) while $\gamma^0$ can be Hermitian with eigenvalues $\pm 1$ . For the Dirac Hamiltonian to be Hermitian — so that probability $\psi^\dagger\psi$ is conserved — the gammas must satisfy a single unified Hermiticity relation:

\gamma^{\mu\dagger} = \gamma^0\gamma^\mu\gamma^0,\qquad \gamma^{0\dagger} = \gamma^0,\qquad \gamma^{i\dagger} = -\gamma^i \qquad (3)

where: $\dagger$ is Hermitian conjugation. The compact form $\gamma^{\mu\dagger} = \gamma^0\gamma^\mu\gamma^0$ encodes both special cases at once: put $\mu = 0$ and it reads $\gamma^0 = \gamma^{0\dagger}$ ; put $\mu = i$ and, using $\gamma^0\gamma^i\gamma^0 = -\gamma^i$ , it reads $\gamma^{i\dagger} = -\gamma^i$ . A concrete realization in the Dirac (standard) representation makes the block structure visible — $\gamma^0$ is diagonal, the $\gamma^i$ are purely off-diagonal:

\gamma^0 = \begin{pmatrix}\mathbb{1}&0\\0&-\mathbb{1}\end{pmatrix},\qquad \gamma^i = \begin{pmatrix}0&\sigma^i\\-\sigma^i&0\end{pmatrix} \qquad (4)

where: $\sigma^i$ are the $2\times 2$ Pauli matrices and $\mathbb{1}$ is the $2\times 2$ block identity. The upper/lower block split of $\gamma^0$ is exactly the "large/small component" split that drives the nonrelativistic reduction to the Pauli equation. Two more objects complete the toolkit. The fifth matrix $\gamma^5$ measures chirality, and the tensor $\sigma^{\mu\nu}$ generates Lorentz transformations on spinors:

\gamma^5 \equiv i\gamma^0\gamma^1\gamma^2\gamma^3,\quad (\gamma^5)^2 = \mathbb{1},\quad \{\gamma^5,\gamma^\mu\} = 0,\qquad \sigma^{\mu\nu}\equiv\frac{i}{2}[\gamma^\mu,\gamma^\nu] \qquad (5)

where: $\gamma^5$ is Hermitian, squares to the identity, and anticommutes with every $\gamma^\mu$ ; the factor of $i$ is chosen precisely so $(\gamma^5)^2 = +\mathbb{1}$ . Because $(\gamma^5)^2 = \mathbb{1}$ , the operators $P_{R,L} = \tfrac12(\mathbb{1}\pm\gamma^5)$ are orthogonal projectors onto right- and left-handed spinors. The generators $S^{\mu\nu} = \tfrac12\sigma^{\mu\nu}$ satisfy the Lorentz algebra, which is what carries over to "Lorentz Covariance of the Dirac Equation."

Derivation ASquaring the Dirac operator forces the Clifford relation

The whole edifice rests on one requirement: the first-order Dirac operator, applied twice, must reproduce the second-order Klein-Gordon operator, so that every component of $\psi$ satisfies the relativistic wave equation. Our strategy is to multiply the Dirac operator by its conjugate, watch the mass cross-terms cancel, and read off the condition the gamma products must obey. The one subtlety to watch: because $\partial_\mu\partial_\nu$ is symmetric under swapping $\mu\leftrightarrow\nu$ , only the symmetric part of $\gamma^\mu\gamma^\nu$ can survive, and that is what turns into the metric.

Step 1. Start from Dirac's ansatz, a first-order operator with unknown matrices $\gamma^\mu$ acting on $\psi$ :

(i\gamma^\mu\partial_\mu - m)\psi = 0.

Why this form. First order in $\partial_t$ is the whole point — it is what cures the negative probability density that a second-order-in-time equation carries.

Step 2. Act with the algebraic partner $(i\gamma^\nu\partial_\nu+m)$ on the left. The mass cross-terms cancel, leaving:

(i\gamma^\nu\partial_\nu + m)(i\gamma^\mu\partial_\mu - m)\psi = -(\gamma^\nu\gamma^\mu\partial_\nu\partial_\mu + m^2)\psi = 0.

Step 3. Because $\partial_\nu\partial_\mu$ is symmetric, replace $\gamma^\nu\gamma^\mu$ by its symmetric part:

\gamma^\nu\gamma^\mu\partial_\nu\partial_\mu = \tfrac12\{\gamma^\mu,\gamma^\nu\}\partial_\mu\partial_\nu.

Why the antisymmetric part drops. Contracting an antisymmetric object with the symmetric $\partial_\mu\partial_\nu$ gives zero identically — nothing is assumed about the gammas yet.

Step 4. Demand the result equal $-(\Box+m^2)\psi$ . Therefore, $\frac12\{\gamma^\mu,\gamma^\nu\}\partial_\mu\partial_\nu\psi = g^{\mu\nu}\partial_\mu\partial_\nu\psi = \Box\psi$ . Matching the coefficients of $\partial_\mu\partial_\nu$ then forces

\tfrac12\{\gamma^\mu,\gamma^\nu\} = g^{\mu\nu}\mathbb{1}\quad\Longleftrightarrow\quad \{\gamma^\mu,\gamma^\nu\} = 2g^{\mu\nu}\mathbb{1}.

Why this cannot be solved by numbers. For $\mu\neq\nu$ the right side is zero, so $\gamma^\mu\gamma^\nu = -\gamma^\nu\gamma^\mu$ ; commuting numbers would then force one of them to vanish, contradicting $(\gamma^\mu)^2 = \pm\mathbb{1}$ . Matrices are mandatory. "The Dirac Equation" imposes exactly this relation — it serves the statement that $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}\mathbb{1}$ is "the defining relation of a Clifford algebra," which this article then unpacks. The explicit Dirac-rep matrices satisfy $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}\mathbb{1}$ for all $16$ pairs and $\gamma^{\mu\dagger}=\gamma^0\gamma^\mu\gamma^0$ for all $\mu$ .

Go deeper · Derivation B · 1 minTrace theorems from the anticommutator alone

Derivation BTrace theorems from the anticommutator alone

Spin-averaged cross sections collapse into traces of gamma products, so the trace identities are the single most-used tool in relativistic QM and QED. The goal here is to derive them using only the Clifford relation and the cyclicity of the trace — never a matrix representation — which is what makes them representation-independent facts. The one number to keep in mind is the dimension $4$ : it enters as $\mathrm{Tr}\,\mathbb{1} = 4$ .

Step 1. Split any gamma product into symmetric and antisymmetric parts, $\gamma^\mu\gamma^\nu = \tfrac12\{\gamma^\mu,\gamma^\nu\} + \tfrac12[\gamma^\mu,\gamma^\nu]$ , and take the trace. The commutator has zero trace, so

\mathrm{Tr}(\gamma^\mu\gamma^\nu) = \tfrac12\,\mathrm{Tr}\!\big(2g^{\mu\nu}\mathbb{1}\big) = 4g^{\mu\nu}.

Why the commutator drops. $\mathrm{Tr}(AB) = \mathrm{Tr}(BA)$ makes the trace of any commutator vanish identically.

Step 2. For a single gamma, insert $\mathbb{1} = (\gamma^5)^2$ and use $\{\gamma^5,\gamma^\mu\}=0$ with cyclicity:

\mathrm{Tr}(\gamma^\mu) = \mathrm{Tr}(\gamma^5\gamma^5\gamma^\mu) = -\mathrm{Tr}(\gamma^5\gamma^\mu\gamma^5) = -\mathrm{Tr}(\gamma^\mu) \;\Rightarrow\; \mathrm{Tr}(\gamma^\mu) = 0.

Why it generalizes. The same $\gamma^5$ -insertion trick kills the trace of any odd number of gammas: each anticommutation past the product flips an overall sign, forcing the trace to equal minus itself.

Step 3. For four gammas, anticommute the first through the other three using $\gamma^\mu\gamma^\nu = 2g^{\mu\nu} - \gamma^\nu\gamma^\mu$ and reduce to two-gamma traces. The bookkeeping gives

\mathrm{Tr}(\gamma^\mu\gamma^\nu\gamma^\rho\gamma^\sigma) = 4\big(g^{\mu\nu}g^{\rho\sigma} - g^{\mu\rho}g^{\nu\sigma} + g^{\mu\sigma}g^{\nu\rho}\big).

Why the sign pattern. Each swap needed to bring identical indices adjacent contributes a relative minus sign; the three pairings appear with signs $+,-,+$ .

Step 4. The same anticommutator yields the contraction identities, $\gamma^\mu\gamma_\mu = 4\mathbb{1}$ (since $g^\mu{}_\mu = 4$ ) and $\gamma^\mu\gamma^\nu\gamma_\mu = -2\gamma^\nu$ , which reduce every longer string to shorter ones. All trace identities up to four gammas, $\mathrm{Tr}(\gamma^5)=0$ , $\mathrm{Tr}(\gamma^5\gamma^\mu\gamma^\nu)=0$ , and both contraction identities hold numerically in the Dirac representation.

Go deeper · Derivation C · 1 minWhy four components, and why representation-independent

Derivation CWhy four components, and why representation-independent

The number four is not an assumption — it is forced, and this derivation shows why, then explains why any two valid choices of matrices give the same physics. The route is a dimension count of the algebra followed by Pauli's fundamental theorem. Watch the last step: the count $N^2\geq 16$ only gives a lower bound; irreducibility is what makes it an equality.

Step 1. List the independent products of the four gammas: the identity ( $1$ ), the four $\gamma^\mu$ ( $4$ ), the six $\gamma^\mu\gamma^\nu$ with $\mu<\nu$ ( $6$ ), the four triple products ( $4$ ), and the single quadruple product $\gamma^0\gamma^1\gamma^2\gamma^3$ ( $1$ ).

Step 2. Add them: $1 + 4 + 6 + 4 + 1 = 16$ . Equivalently, from four generators each appearing $0$ or $1$ time (higher powers reduce via $(\gamma^\mu)^2 = \pm\mathbb{1}$ ), there are $2^4 = 16$ independent elements.

Step 3. These $16$ are linearly independent: multiplying any assumed vanishing combination by an appropriate basis element and taking the trace isolates each coefficient as zero.

Step 4. A set of $16$ linearly independent $N\times N$ matrices requires $N^2\ge16$ , so $N\ge4$ . A $4\times4$ realization exists, for example the Dirac representation. Since the complex Clifford algebra in four dimensions is isomorphic to $M_4(\mathbb C)$ , its irreducible representation acts on a $4$ -dimensional spinor space. Hence a Dirac spinor has four components, as required by the Clifford algebra.

Step 5. Pauli's fundamental theorem: any two sets of $4\times 4$ matrices obeying the same Clifford relation are related by a similarity transformation $\gamma'^\mu = U\gamma^\mu U^{-1}$ , unique up to a phase. Since traces, eigenvalues, and $|\mathcal{M}|^2$ are similarity-invariant, the representation is a free choice. The count $1+4+6+4+1=16$ and the rank of the $16$ flattened basis matrices equals $16$ , confirming linear independence. An explicit unitary $U=\tfrac1{\sqrt2}\begin{pmatrix}\mathbb{1}&-\mathbb{1}\\\mathbb{1}&\mathbb{1}\end{pmatrix}$ maps the Dirac rep to the Weyl rep via $U\gamma^\mu U^\dagger = \gamma^\mu_{\text{Weyl}}$ .

Limiting cases

The Clifford structure depends sharply on spacetime dimension and on which physical limit you take:

Case	Result	Meaning
Spacetime dimension $d$	$\dim_{\mathbb{C}}(\text{irrep}) = 2^{\lfloor d/2\rfloor}$	In $d=4$ this is $2^2 = 4$ components. In $d=2$ it would be $2$ ; in $d=10$ (superstrings) it is $32$ . Four-component spinors are a feature of four dimensions.
Massless limit $m\to 0$	$i\gamma^\mu\partial_\mu\psi = 0$ decouples	The mass term is what couples left and right chiralities; at $m=0$ the Dirac equation splits into two independent two-component Weyl equations, manifest in the chiral basis where $\gamma^5$ is diagonal.
Nonrelativistic limit	large/small split of $\gamma^0$	In the Dirac rep the block structure of $\gamma^0 = \mathrm{diag}(\mathbb{1},-\mathbb{1})$ drives the reduction to the Pauli equation; the lower components are suppressed by $v/c$ .
Euclidean (Wick-rotated)	$\{\gamma_E^\mu,\gamma_E^\nu\} = 2\delta^{\mu\nu}\mathbb{1}$	After $t\to -i\tau$ the metric becomes all-positive; all four Euclidean gammas can be chosen Hermitian, the form used in lattice gauge theory.

Common trapThe metric convention is not cosmetic — it flips signs everywhere

Every gamma-matrix formula depends on the metric signature, and mixing conventions is the single most common source of stray minus signs in relativistic QM. With mostly-minus $g=\mathrm{diag}(+,-,-,-)$ — the convention pinned for this chapter — one has $(\gamma^0)^2 = +\mathbb{1}$ and $(\gamma^i)^2 = -\mathbb{1}$ . With mostly-plus the signs flip, and so do the definition of $\gamma^5$ , the sign in $\gamma^{\mu\dagger}=\gamma^0\gamma^\mu\gamma^0$ , and the four-gamma trace. Second trap: $\gamma^5$ is Hermitian and squares to $+\mathbb{1}$ , but it is not one of the four $\gamma^\mu$ — it is a derived matrix that anticommutes with all of them. Treat it as a fifth generator and half the chirality errors disappear.

Worked examples

Example 1 — A cross-section trace, computed two ways. Evaluate $\mathrm{Tr}(\gamma^0\gamma^1\gamma^0\gamma^1)$ and $\mathrm{Tr}(\gamma^0\gamma^0\gamma^1\gamma^1)$ , the kind of numbers that appear in spin-averaged amplitudes. This shows the trace theorem in action and produces concrete integers.

Step 1. Apply the four-gamma theorem $\mathrm{Tr}(\gamma^\mu\gamma^\nu\gamma^\rho\gamma^\sigma) = 4(g^{\mu\nu}g^{\rho\sigma} - g^{\mu\rho}g^{\nu\sigma} + g^{\mu\sigma}g^{\nu\rho})$ with $(\mu\nu\rho\sigma) = (0,1,0,1)$ . The metric values are $g^{00}=+1$ , $g^{11}=-1$ , $g^{01}=0$ .

Step 2. Substitute: $\mathrm{Tr} = 4\big(g^{01}g^{01} - g^{00}g^{11} + g^{01}g^{10}\big) = 4\big(0 - (1)(-1) + 0\big) = 4(+1)$ .

Step 3. Therefore $\mathrm{Tr}(\gamma^0\gamma^1\gamma^0\gamma^1) = \mathbf{4}$ .

Step 4. For $(\mu\nu\rho\sigma) = (0,0,1,1)$ : $\mathrm{Tr} = 4\big(g^{00}g^{11} - g^{01}g^{01} + g^{01}g^{01}\big) = 4\big((1)(-1) - 0 + 0\big) = \mathbf{-4}$ . Both integers are confirmed by an explicit $4\times 4$ trace in the Dirac representation. The sign difference between the two — same four matrices, different ordering — is exactly the anticommuting structure showing up in a measurable spin sum.

Example 2 — Contraction of a longer string. Simplify $\gamma^\mu\gamma^\nu\gamma^\rho\gamma_\mu$ , the object that appears inside QED vertex corrections, using only the contraction identities.

Step 1. Anticommute the leftmost $\gamma^\mu$ inward: $\gamma^\mu\gamma^\nu = 2g^{\mu\nu}\mathbb{1} - \gamma^\nu\gamma^\mu$ , so $\gamma^\mu\gamma^\nu\gamma^\rho\gamma_\mu = 2\gamma^\rho\gamma^\nu - \gamma^\nu(\gamma^\mu\gamma^\rho\gamma_\mu)$ .

Step 2. Use $\gamma^\mu\gamma^\rho\gamma_\mu = -2\gamma^\rho$ on the second term: $= 2\gamma^\rho\gamma^\nu - \gamma^\nu(-2\gamma^\rho) = 2\gamma^\rho\gamma^\nu + 2\gamma^\nu\gamma^\rho$ .

Step 3. Recognize $\gamma^\rho\gamma^\nu + \gamma^\nu\gamma^\rho = \{\gamma^\nu,\gamma^\rho\} = 2g^{\nu\rho}\mathbb{1}$ , so $\gamma^\mu\gamma^\nu\gamma^\rho\gamma_\mu = 2(2g^{\nu\rho}\mathbb{1})$ .

Step 4. The result is $\gamma^\mu\gamma^\nu\gamma^\rho\gamma_\mu = \mathbf{4g^{\nu\rho}\mathbb{1}}$ — a four-index string reduced to a pure number times the identity, with no representation ever invoked. The identity $\gamma^\mu\gamma^\nu\gamma^\rho\gamma_\mu = 4g^{\nu\rho}\mathbb{1}$ holds for all $\nu,\rho$ in the Dirac representation.

Origins · 1 minclifford algebra and how this was discovered

clifford algebra

1878 · geometric algebra of a metric · The Clifford relation that Dirac's factorization forced was, unknown to most physicists at the time, an object mathematicians had studied for half a century

William Kingdon Clifford introduced his "geometric algebras" in 1878 as a way to give a quadratic form (a metric) an associative-algebra structure, unifying Hamilton's quaternions and Grassmann's exterior algebra. He died the next year at thirty-three, long before anyone imagined the construction would describe the electron. When Dirac's 1928 factorization forced $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}$ , he had rediscovered — from pure physical necessity — the defining relation of $\mathrm{Cl}_{1,3}(\mathbb{R})$ . Pauli supplied the fundamental theorem fixing representations up to similarity; Weyl and Majorana gave the chiral and real representations still in daily use.

History

1878

— W. K. Clifford defines geometric (Clifford) algebras, encoding a metric in an associative product

1928

— Dirac's relativistic equation forces the anticommutation relation, rediscovering the Clifford algebra

1929

— Weyl introduces the chiral representation, exposing left- and right-handed two-spinors

1936

— Pauli establishes the similarity theorem: any two irreducible complex representations of the same Clifford algebra are related by a similarity transformation

1937

— Majorana develops a representation in which the

\gamma^\mu

can be chosen purely imaginary, allowing the Dirac equation to admit a reality condition for Majorana spinors

Going further

References. Griffiths, Introduction to Elementary Particles, 2nd ed., §7.1–7.2 and App. A · Sakurai & Napolitano, Modern Quantum Mechanics, 2nd ed., Ch. 8 · Peskin & Schroeder, An Introduction to QFT, §3.2 and App. A.4 · Bjorken & Drell, Relativistic Quantum Mechanics, Ch. 3 · Itzykson & Zuber, Quantum Field Theory, §2-1. The Clifford relation, Hermiticity relations, $\gamma^5$ properties, the 16-element basis, and the trace/contraction identities and explicit Dirac/Weyl matrices are standard results; mostly-minus signature $(+,-,-,-)$ is used throughout. The full Clifford algebra has been checked over all 16 pairs, all trace theorems up to four gammas, the contraction identities, $\gamma^5$ properties, linear independence of the 16-element basis, and the Pauli-theorem similarity between the Dirac and Weyl representations.

Prerequisites. The Dirac Equation · The Klein-Gordon Equation · Negative-Energy Solutions and the Interpretive Crisis. Leads to. Dirac Spinors and Plane-Wave Solutions · Lorentz Covariance of the Dirac Equation · Bilinear Covariants and Dirac Adjoint · Spin from the Dirac Equation · Charge Conjugation and the Positron.

Where exactly does the "square root of the metric" language come from?

The relativistic Hamiltonian $\sqrt{\mathbf{p}^2+m^2}$ is awkward because the square root of a sum has no polynomial form in the momenta. Dirac's trick is to write a linear object $\gamma^\mu p_\mu$ whose square is $p^2 = g^{\mu\nu}p_\mu p_\nu$ . Squaring gives $\gamma^\mu\gamma^\nu p_\mu p_\nu$ ; because $p_\mu p_\nu$ is symmetric, only the symmetric part $\tfrac12\{\gamma^\mu,\gamma^\nu\}$ contributes, and setting it equal to $g^{\mu\nu}\mathbb{1}$ recovers $p^2$ . So " $\gamma^\mu$ is a square root of the metric" means precisely $\tfrac12\{\gamma^\mu,\gamma^\nu\}=g^{\mu\nu}\mathbb{1}$ — the linearization only exists because the gammas anticommute.

What is the difference between the Dirac, Weyl and Majorana representations?

They are three choices of explicit matrices, all satisfying the same Clifford relation, so all physically equivalent by Pauli's theorem. The Dirac (standard) representation diagonalizes $\gamma^0$ , which cleanly separates "large" and "small" spinor components and is ideal for the nonrelativistic reduction to atoms. The Weyl (chiral) representation diagonalizes $\gamma^5$ , so left- and right-handed pieces sit in separate blocks — indispensable at high energy and for the Standard Model's chiral weak interaction. The Majorana representation makes all four gammas purely imaginary, so the Dirac equation admits real solutions, which is the natural setting for a particle that is its own antiparticle. Same physics, different convenience.

Why is γ⁵ Hermitian and squaring to +1, unlike the spatial γ^i?

The extra factor of $i$ in $\gamma^5 = i\gamma^0\gamma^1\gamma^2\gamma^3$ is chosen for exactly this. Without it, the bare product $\gamma^0\gamma^1\gamma^2\gamma^3$ squares to $-\mathbb{1}$ (you pick up minus signs from anticommuting the factors past each other and from $(\gamma^i)^2=-\mathbb{1}$ ). The $i$ contributes $i^2=-1$ , converting $-\mathbb{1}$ to $+\mathbb{1}$ , and also makes $\gamma^5$ Hermitian. This is what lets $P_{R,L}=\tfrac12(\mathbb{1}\pm\gamma^5)$ be genuine orthogonal projectors, since a projector must satisfy $P^2=P$ and $P^\dagger=P$ .

Why do all trace identities have a factor of 4 or a dimension in them?

Because $\mathrm{Tr}\,\mathbb{1}_4 = 4$ , the dimension of the representation. Every trace identity ultimately reduces, via the Clifford relation, to traces of the identity times products of metric components. For instance $\mathrm{Tr}(\gamma^\mu\gamma^\nu)=\tfrac12\mathrm{Tr}(2g^{\mu\nu}\mathbb{1})=4g^{\mu\nu}$ . In dimensional regularization the representation is formally $d$ -dimensional and $\mathrm{Tr}\,\mathbb{1}$ is often kept as $4$ (or $2^{d/2}$ , depending on scheme), which is one reason the traces need care away from four dimensions.

How does the number 16 relate to the bilinear covariants?

The 16 independent gamma products $\{\mathbb{1},\gamma^\mu,\sigma^{\mu\nu},\gamma^5\gamma^\mu,\gamma^5\}$ form a complete basis for all $4\times4$ complex matrices ( $16 = 4^2$ ), and they organize by how the bilinear $\bar\psi\,\Gamma^A\psi$ transforms under Lorentz transformations: scalar (1), vector (4), antisymmetric tensor (6), axial vector (4), pseudoscalar (1). These are the only Lorentz-covariant fermion bilinears you can write, and they are exactly the building blocks of every interaction in particle physics — the electromagnetic current $\bar\psi\gamma^\mu\psi$ , the weak $\bar\psi\gamma^\mu(1-\gamma^5)\psi$ , the Yukawa $\bar\psi\psi$ . The full classification is developed in "Bilinear Covariants and Dirac Adjoint."

Why does the irreducible dimension jump as 2^⌊d/2⌋ with spacetime dimension?

Because you can pair up the $d$ generators into roughly $d/2$ "creation/annihilation" pairs, each pair building a two-dimensional factor in the representation — the same combinatorics as fermionic Fock space. Concretely, from $d$ anticommuting gammas you form ladder operators whose Fock space has dimension $2^{\lfloor d/2\rfloor}$ . For $d=4$ this is $2^2=4$ , giving four-component spinors; for the $d=10$ of superstring theory it is $2^5=32$ . The four-ness of the Dirac spinor is therefore a low-dimensional accident of $d=4$ , not a universal law.

If the physics is in the algebra, why bother writing explicit matrices at all?

For two reasons. First, existence: writing down the Dirac-basis matrices proves a solution to the Clifford relation actually exists, and Pauli's theorem then tells you it is essentially unique. Second, convenience: for a specific problem one representation makes the answer nearly free — the Dirac basis for the hydrogen fine structure (large/small split), the Weyl basis for a massless neutrino (chirality manifest). But any final observable — an eigenvalue, a cross section, a trace — can and should be computed without ever choosing a basis, using the algebra and the trace identities. Choosing a representation is a computational aid, never a physical input.