Gamma matrices & the Clifford algebra

The Dirac equation can be written in the Hamiltonian form $i\partial_t\psi=(\boldsymbol\alpha\cdot\mathbf p+\beta m)\psi$ , which puts spin on display but hides the symmetry relativity demands: time and space are torn apart, one derivative multiplied by $\beta$ , the others by $\boldsymbol\alpha$ . There is a better packaging. Multiply through by $\beta$ , use $\beta^2=I$ , and define four constant matrices $\gamma^0\equiv\beta$ , $\gamma^k\equiv\beta\alpha^k$ . The whole equation collapses into one manifestly covariant line, $(i\gamma^\mu\partial_\mu-m)\psi=0$ , with a single derivative $\partial_\mu$ contracted against a single object $\gamma^\mu$ . That repackaging is not cosmetic. The properties the $\alpha$ 's and $\beta$ were built to satisfy — squaring to the identity and mutually anticommuting — become, for the $\gamma$ 's, a single compact statement: the Clifford algebra $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}I$ . This one relation carries the entire content of the theory. Every trace identity a cross-section needs, the spin- $\tfrac12$ structure, the existence of antiparticles, the four-vector nature of the current — all of it follows from anticommutation and the metric, with no reference to any explicit array of numbers. The matrices are a means; the algebra is the physics. This article is the hinge of the relativistic-electron chapter. It takes the $\alpha,\beta$ matrices from aq-pp-dirac-equation, the metric from aq-pp-four-vector-notation, and the density–current pair from aq-pp-dirac-probability-current, and forges the covariant toolkit — $\gamma^\mu$ , the adjoint $\bar\psi$ , the Feynman slash, the Clifford relation — that every downstream QED calculation simply assumes.

Key takeaways

Defining $\gamma^0=\beta$ and $\gamma^k=\beta\alpha^k$ recasts the Dirac equation into the compact covariant form $(i\gamma^\mu\partial_\mu-m)\psi=0$ .
The four matrices obey one master relation, the Clifford algebra $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}I$ , which alone fixes the physics — no explicit representation required.
Hermiticity is mixed: $\gamma^0$ is Hermitian, the spatial $\gamma^k$ are anti-Hermitian, packaged as $\gamma^{\mu\dagger}=\gamma^0\gamma^\mu\gamma^0$ .
Despite the index $\mu$ , the $\gamma$ -matrices are not a four-vector — they are constant arrays, identical in every frame; the transformation lives in the spinor $\psi$ .
The adjoint spinor $\bar\psi\equiv\psi^\dagger\gamma^0$ turns the probability current into $j^\mu=\bar\psi\gamma^\mu\psi$ , a genuine conserved four-vector with $j^0=\psi^\dagger\psi\ge0$ .
Key idea: packaging the $\beta$ - and $\alpha$ -built matrices into $\gamma^\mu$ turns the Dirac equation into a single Lorentz-covariant line, and the entire content of the theory is encoded in one algebraic rule, $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}$ .

The core insight

Packaging the four $\beta$ - and $\alpha$ -built matrices into $\gamma^\mu$ turns the Dirac equation into a single Lorentz-covariant line, and the entire content of the theory is encoded in one algebraic rule: $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}$ .

Where this lives today

Every time a collider physicist computes a cross-section, the gamma matrices are doing the work — but invisibly, because the Clifford algebra lets them disappear before any number is evaluated. An amplitude like $e^+e^-\to\mu^+\mu^-$ is a string of gammas sandwiched between spinors; squaring it and summing over spins turns the whole thing into traces of gamma products, which the identities here — $\mathrm{tr}(\gamma^\mu\gamma^\nu)=4g^{\mu\nu}$ , the four-gamma trace, $\gamma^\mu\slashed a\gamma_\mu=-2\slashed a$ — reduce to dot products of momenta. No explicit $4\times4$ matrix is ever multiplied out, and this is exactly the machinery automated inside the symbolic tools physicists rely on: FeynCalc, FORM, and the trace routines in FeynArts/MadGraph implement these rules verbatim. Beyond QED, $\gamma^5$ and the projectors $P_{L,R}=\tfrac12(1\mp\gamma^5)$ are the algebraic seat of weak parity violation, where only left-handed fields couple to the $W$ ; the Majorana representation houses neutrino-mass models; and lattice QCD discretizes the same $\gamma^\mu$ to put quarks on a computer. A relation Dirac wrote down to fix a sign in a probability density is now the silent arithmetic behind the entire Standard Model.

The physical picture

The four gamma matrices are not four numbers or four vectors — they are four operators whose only essential property is how they multiply, captured by one anticommutation table: $\gamma^\mu\gamma^\nu+\gamma^\nu\gamma^\mu$ gives twice the metric entry $g^{\mu\nu}$ . On the diagonal this reproduces the Minkowski signature $(+,-,-,-)$ — $\gamma^0$ squares to $+I$ , each $\gamma^k$ to $-I$ — and off the diagonal the matrices anticommute to zero. The first figure is that table: the whole algebra on one grid, the metric read straight off the diagonal.

Figure

The Clifford anticommutator table {γ^μ,γ^ν} = 2g^{μν}I: the diagonal carries the Minkowski signature +2, −2, −2, −2 while every off-diagonal entry is 0, so distinct gammas anticommute and each squares to ±I

To turn the table into a working object one needs an explicit set of matrices, and the standard choice — the Dirac–Pauli representation — writes each $\gamma^\mu$ as $2\times2$ blocks: $\gamma^0$ block-diagonal, $\mathrm{diag}(I,-I)$ , and each $\gamma^k$ off-diagonal, built from the Pauli matrix $\sigma^k$ . The block layout makes the non-relativistic limit transparent, splitting the components into "large" and "small" pieces. The second figure shows that structure and its payoff — a positive probability density $j^0=\psi^\dagger\psi$ , curing the defect that had sunk the Klein–Gordon equation.

Figure

The Dirac–Pauli representation as 2×2 blocks: γ⁰ = diag(I, −I) is block-diagonal so (γ⁰)² = I, while γ^k = offset Pauli blocks gives (γ^k)² = −I; the adjoint ψ† γ⁰ then makes j⁰ = ψ†ψ ≥ 0

The core equations

The starting point is the Hamiltonian form of the Dirac equation, $i\partial_t\psi=(\boldsymbol\alpha\cdot\mathbf p+\beta m)\psi$ , whose four Hermitian matrices satisfy $\{\alpha^i,\alpha^j\}=2\delta^{ij}I$ , $\{\alpha^i,\beta\}=0$ , $(\alpha^i)^2=\beta^2=I$ . Multiplying by $\beta$ and defining the gammas below folds those relations into one covariant equation.

\gamma^0\equiv\beta,\qquad \gamma^k\equiv\beta\alpha^k,\qquad (i\gamma^\mu\partial_\mu-m)\psi=0 \qquad (1)

where: $\beta,\alpha^k$ — the Dirac Hamiltonian matrices; $\gamma^\mu$ — the four constant $4\times4$ gamma matrices, $\mu=0,1,2,3$ ; $\partial_\mu=(\partial_t,\nabla)$ — the four-gradient; $m$ — the fermion mass ( $\hbar=c=1$ ). Repeated indices summed. Reading: one derivative $\partial_\mu$ contracted with one object $\gamma^\mu$ — a single covariant line replaces the split time/space form. The relations of $\alpha$ and $\beta$ propagate into the gammas as one anticommutation rule — the Clifford algebra of Minkowski space. It is representation-independent, and everything downstream (traces, spin, the current) follows from it.

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

where: $g^{\mu\nu}=\mathrm{diag}(+1,-1,-1,-1)$ — the Minkowski metric; $I$ — the $4\times4$ identity. Reading: $\mu=\nu$ gives $(\gamma^0)^2=I$ , $(\gamma^k)^2=-I$ ; $\mu\ne\nu$ gives $\gamma^\mu\gamma^\nu=-\gamma^\nu\gamma^\mu$ . The metric on the right is the sole place spacetime geometry enters. Because $\beta$ and the $\alpha^k$ are Hermitian while the $\alpha^k$ anticommute with $\beta$ , the gammas inherit a mixed Hermiticity, neatly packaged using $\gamma^0$ itself as a "metric" on the spinor indices.

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

where: $\dagger$ — Hermitian conjugate. Reading: $\gamma^0$ is Hermitian (eigenvalues $\pm1$ ), each $\gamma^k$ anti-Hermitian (eigenvalues $\pm i$ ); the compact form $\gamma^{\mu\dagger}=\gamma^0\gamma^\mu\gamma^0$ is what makes the adjoint spinor and the current behave correctly under conjugation. For explicit calculation one picks a representation. The Dirac–Pauli choice writes each gamma as a $2\times2$ block matrix built from the identity and the Pauli matrices $\sigma^k$ ; it is the representation in which the non-relativistic limit is cleanest.

\gamma^0=\begin{pmatrix} I & 0 \\ 0 & -I\end{pmatrix},\qquad \gamma^k=\begin{pmatrix} 0 & \sigma^k \\ -\sigma^k & 0\end{pmatrix} \qquad (4)

where: $I$ — the $2\times2$ identity; $\sigma^k$ — the Pauli matrices, $\{\sigma^i,\sigma^j\}=2\delta^{ij}I$ . Reading: block-diagonal $\gamma^0$ gives $(\gamma^0)^2=I$ ; off-diagonal $\gamma^k$ with $(\sigma^k)^2=I$ gives $(\gamma^k)^2=-I$ , reproducing the full Clifford algebra. Any other valid set is related to this by a similarity transformation (Pauli's theorem). Two further constructs make QED compact: the Feynman slash $\slashed a\equiv\gamma^\mu a_\mu$ and the chirality matrix $\gamma^5$ . Splitting a slash product into symmetric and antisymmetric parts turns gamma strings into dot products plus a spin-tensor piece.

\slashed a\,\slashed b=a\cdot b\,I-i\sigma^{\mu\nu}a_\mu b_\nu,\qquad \sigma^{\mu\nu}\equiv\tfrac{i}{2}[\gamma^\mu,\gamma^\nu],\qquad \gamma^5\equiv i\gamma^0\gamma^1\gamma^2\gamma^3 \qquad (5)

where: $a\cdot b=a^\mu b_\mu$ — the four-vector dot product; $\sigma^{\mu\nu}$ — the antisymmetric spin tensor (six independent matrices); $\gamma^5$ — Hermitian, $(\gamma^5)^2=I$ , and $\{\gamma^5,\gamma^\mu\}=0$ . Reading: setting $b=a$ gives the crucial identity $\slashed a\,\slashed a=a^2 I$ ; $\gamma^5$ is the extra generator needed to build chirality projectors $P_{R,L}=\tfrac12(1\pm\gamma^5)$ . Finally, the gammas rebuild the probability current. Defining the adjoint spinor $\bar\psi\equiv\psi^\dagger\gamma^0$ converts the density–current pair into a single conserved four-vector.

\bar\psi\equiv\psi^\dagger\gamma^0,\qquad j^\mu=\bar\psi\gamma^\mu\psi,\qquad \partial_\mu j^\mu=0 \qquad (6)

where: $j^\mu=(\rho,\mathbf j)$ — the conserved probability four-current; $j^0=\bar\psi\gamma^0\psi=\psi^\dagger\psi\ge0$ — the positive-definite density. Reading: the adjoint $\bar\psi$ (not bare $\psi^\dagger$ ) makes $j^\mu$ a four-vector and $\bar\psi\psi$ a Lorentz scalar; $\partial_\mu j^\mu=0$ follows from the Dirac equation and its adjoint.

Derivation AFrom the alpha and beta matrices to the covariant Dirac equation

The goal is to show that the covariant form $(i\gamma^\mu\partial_\mu-m)\psi=0$ is not a new postulate but an exact rewriting of the Hamiltonian Dirac equation $i\partial_t\psi=(\boldsymbol\alpha\cdot\mathbf p+\beta m)\psi$ . We multiply by $\beta$ , use $\beta^2=I$ to symmetrize the time and space derivatives, and read off $\gamma^0=\beta$ , $\gamma^k=\beta\alpha^k$ that collapse it into a single index-contracted line.

Step 1. Write the Hamiltonian form with momentum realized as a derivative, $\mathbf p=-i\nabla$ : the equation reads $i\partial_t\psi=-i\alpha^k\partial_k\psi+\beta m\psi$ , where $\partial_k$ is the spatial gradient and the sum over $k=1,2,3$ is implied. Why: its square gives $E^2=\mathbf p^2+m^2$ , so the kinematics are already correct — we only want the Lorentz symmetry made manifest.

Step 2. Bring everything to one side and left-multiply the entire equation by $\beta$ : $i\beta\partial_t\psi+i\beta\alpha^k\partial_k\psi-\beta^2 m\psi=0$ . Since $\beta^2=I$ , the mass term becomes simply $-m\psi$ , and the time derivative now carries $\beta$ exactly as the space derivatives carry $\beta\alpha^k$ . Why: multiplying by $\beta$ puts the time and space derivatives on the same footing, each preceded by one matrix — the structural requirement for covariance.

Step 3. Name the matrices: define $\gamma^0\equiv\beta$ and $\gamma^k\equiv\beta\alpha^k$ . With $\partial_0\equiv\partial_t$ the equation is $i\gamma^0\partial_0\psi+i\gamma^k\partial_k\psi-m\psi=0$ , and combining the time and space terms into the four-index sum $\gamma^\mu\partial_\mu$ gives $(i\gamma^\mu\partial_\mu-m)\psi=0$ . Why: the definitions are forced — whatever multiplies each derivative is $\gamma^\mu$ — and the payoff is one covariant object replacing the split $\boldsymbol\alpha,\beta$ pair.

Step 4. Confirm the rewriting is exact and reversible: left-multiplying the covariant form by $\gamma^0=\beta$ and using $(\gamma^0)^2=I$ returns $i\partial_t\psi+i\gamma^0\gamma^k\partial_k\psi-\beta m\psi=0$ ; since $\gamma^0\gamma^k=\beta\beta\alpha^k=\alpha^k$ , this is exactly the Hamiltonian form. Why: the map running both ways certifies no physics was added or lost — the two forms are the same statement in different costumes.

Go deeper · Derivation B · 1 minThe Clifford algebra as the shadow of the anticommuting alphas and beta

Derivation BThe Clifford algebra as the shadow of the anticommuting alphas and beta

The goal is to derive the master relation $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}I$ from the elementary algebra of $\alpha$ and $\beta$ , checking all four index cases so the Minkowski signature $(+,-,-,-)$ emerges rather than being assumed. The only inputs are $\beta^2=(\alpha^k)^2=I$ , $\{\alpha^i,\alpha^j\}=2\delta^{ij}I$ , $\{\alpha^k,\beta\}=0$ , and the definitions $\gamma^0=\beta$ , $\gamma^k=\beta\alpha^k$ .

Step 1. The timelike diagonal: $(\gamma^0)^2=\beta^2=I$ , so $\{\gamma^0,\gamma^0\}=2I=2g^{00}I$ since $g^{00}=+1$ . Why: a timelike direction must square to $+I$ in the mostly-minus metric, and this fixes the sign of the whole signature.

Step 2. The spacelike diagonal: $(\gamma^k)^2=\beta\alpha^k\beta\alpha^k$ . Move the second $\beta$ leftward past $\alpha^k$ using $\alpha^k\beta=-\beta\alpha^k$ , giving $-\beta^2(\alpha^k)^2=-I$ . Hence $\{\gamma^k,\gamma^k\}=-2I=2g^{kk}I$ with $g^{kk}=-1$ . Why: the sign flip from anticommuting $\alpha^k$ past $\beta$ is what makes a spacelike gamma square to $-I$ — space enters through that minus sign.

Step 3. The mixed time–space case: $\{\gamma^0,\gamma^k\}=\beta(\beta\alpha^k)+(\beta\alpha^k)\beta=\beta^2\alpha^k+\beta(-\beta\alpha^k)=\alpha^k-\alpha^k=0=2g^{0k}I$ . Why: off-diagonal metric entries vanish, so time and space gammas anticommute — the spacetime axes are independent.

Step 4. The distinct-spatial case ( $i\ne j$ ): $\{\gamma^i,\gamma^j\}=\beta\alpha^i\beta\alpha^j+\beta\alpha^j\beta\alpha^i=-\alpha^i\alpha^j-\alpha^j\alpha^i=-\{\alpha^i,\alpha^j\}=-2\delta^{ij}I=0$ . Collecting all four cases into one formula gives $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}I$ . Why: this case rides on the alphas anticommuting among themselves; assembling all four shows the Clifford relation contains exactly the Minkowski metric — no more, no less.

Go deeper · Derivation C · 1 minThe conserved four-current from the Dirac equation and its adjoint

Derivation CThe conserved four-current from the Dirac equation and its adjoint

The goal is to prove that $j^\mu=\bar\psi\gamma^\mu\psi$ is conserved, $\partial_\mu j^\mu=0$ , and that its time component is the positive density $\psi^\dagger\psi$ — the property that rescued the single-particle interpretation from the negative-density failure of Klein–Gordon. The route is to conjugate the Dirac equation, convert $\psi^\dagger$ into the adjoint $\bar\psi$ using $\gamma^{\mu\dagger}=\gamma^0\gamma^\mu\gamma^0$ , and combine the two equations.

Step 1. Take the Dirac equation $i\gamma^\mu\partial_\mu\psi-m\psi=0$ and its Hermitian conjugate, $-i\partial_\mu\psi^\dagger\gamma^{\mu\dagger}-m\psi^\dagger=0$ . Right-multiply the conjugate by $\gamma^0$ and insert $\gamma^{\mu\dagger}\gamma^0=\gamma^0\gamma^\mu$ (which follows from $\gamma^{\mu\dagger}=\gamma^0\gamma^\mu\gamma^0$ and $(\gamma^0)^2=I$ ). Why: bare conjugation produces non-covariant $\gamma^{\mu\dagger}$ ; sandwiching with $\gamma^0$ restores a plain $\gamma^\mu$ and defines the adjoint $\bar\psi=\psi^\dagger\gamma^0$ .

Step 2. The result is the adjoint Dirac equation $i\partial_\mu\bar\psi\,\gamma^\mu+m\bar\psi=0$ , or equivalently $\partial_\mu\bar\psi\,\gamma^\mu=im\bar\psi$ . Alongside it the original equation gives $\gamma^\mu\partial_\mu\psi=-im\psi$ . Why: casting both equations in $\bar\psi,\psi$ form readies them to combine, with mass terms of opposite sign.

Step 3. Differentiate the current: $\partial_\mu j^\mu=(\partial_\mu\bar\psi)\gamma^\mu\psi+\bar\psi\gamma^\mu(\partial_\mu\psi)$ . Substitute the two boxed results: the first term is $(im\bar\psi)\psi$ and the second is $\bar\psi(-im\psi)$ , which cancel exactly, so $\partial_\mu j^\mu=0$ . Why: the mass cancels identically — conservation holds for any $m$ , guaranteed by the Dirac structure, not any fine-tuning.

Step 4. Read off the density: $j^0=\bar\psi\gamma^0\psi=\psi^\dagger\gamma^0\gamma^0\psi=\psi^\dagger\psi=|\psi_1|^2+|\psi_2|^2+|\psi_3|^2+|\psi_4|^2\ge0$ . Why: because $(\gamma^0)^2=I$ , the adjoint construction returns the non-negative spinor norm — the positive density Klein–Gordon could not provide, and the reason Dirac's equation was taken seriously.

Limiting cases

Case	Result	Meaning
Diagonal, timelike ( $\mu=\nu=0$ )	$(\gamma^0)^2=+I$	eigenvalues $\pm1$ ; Hermitian and unitary
Diagonal, spacelike ( $\mu=\nu=k$ )	$(\gamma^k)^2=-I$	eigenvalues $\pm i$ ; anti-Hermitian, still unitary
Distinct indices ( $\mu\ne\nu$ )	$\gamma^\mu\gamma^\nu=-\gamma^\nu\gamma^\mu$	pure anticommutation; underlies every gamma-string reduction
Full contraction	$\gamma^\mu\gamma_\mu=4I$	the sum $g_{\mu\nu}\gamma^\mu\gamma^\nu=g^\mu{}_\mu I$ in $d=4$
Sandwich contraction	$\gamma^\mu\gamma^\nu\gamma_\mu=-2\gamma^\nu$	the workhorse of one-loop and cross-section traces
Massless limit ( $m\to0$ )	$i\slashed\partial\psi=0$ , chirality conserved	$\gamma^5$ commutes with the dynamics; $\psi_L,\psi_R$ decouple
Non-relativistic limit	large/small components split	Dirac–Pauli blocks give the Pauli equation, $g=2$

Common trapThe gamma matrices form a four-vector

The tempting error is to read the raised index $\mu$ on $\gamma^\mu$ as meaning the four matrices transform as a contravariant four-vector — that a boost mixes $\gamma^0$ into $\gamma^1$ the way it mixes $t$ into $x$ . They do not. The gamma matrices are constant arrays of numbers, identical in every inertial frame; you write the same four $4\times4$ tables in the lab as on the moving train. What actually transforms is the spinor: under a Lorentz transformation $\Lambda$ , $\psi\to S(\Lambda)\psi$ with $S(\Lambda)=\exp(-\tfrac{i}{4}\omega_{\mu\nu}\sigma^{\mu\nu})$ , and the key identity is $S^{-1}(\Lambda)\gamma^\mu S(\Lambda)=\Lambda^\mu{}_\nu\gamma^\nu$ . The fixed matrix $\gamma^\mu$ , conjugated by the spinor rotation, reproduces the rotated matrix — which is exactly what lets $\bar\psi\gamma^\mu\psi$ come out a genuine four-vector even though $\gamma^\mu$ never moved. The index $\mu$ tells you how the surrounding spinors must transform; it does not make the matrix a vector.

Worked examples

1 · Verifying the Clifford algebra by explicit matrix multiplication (confidence builder)

Using the Dirac–Pauli representation, show directly that $(\gamma^1)^2=-I$ and that $\gamma^1$ anticommutes with $\gamma^2$ , then confirm $\mathrm{tr}(\gamma^1\gamma^1)=4g^{11}=-4$ .

Step 1. Write out $\gamma^1$ with $\sigma^1=\begin{pmatrix}0&1\\1&0\end{pmatrix}$ . In $4\times4$ form,

\gamma^1=\begin{pmatrix} 0 & \sigma^1 \\ -\sigma^1 & 0\end{pmatrix}=\begin{pmatrix} 0&0&0&1\\ 0&0&1&0\\ 0&-1&0&0\\ -1&0&0&0\end{pmatrix}.

The upper-right block is $+\sigma^1$ , the lower-left is $-\sigma^1$ ; all diagonal blocks vanish.

Step 2. Square it block-wise: $\left(\begin{smallmatrix}0&\sigma^1\\-\sigma^1&0\end{smallmatrix}\right)^2=\left(\begin{smallmatrix}-(\sigma^1)^2&0\\0&-(\sigma^1)^2\end{smallmatrix}\right)$ . Since $(\sigma^1)^2=I$ , this is $\begin{pmatrix}-I&0\\0&-I\end{pmatrix}=-I_{4}$ . So $(\gamma^1)^2=\mathbf{-I}$ , matching $g^{11}=-1$ .

Step 3. For the anticommutator, $\gamma^1\gamma^2$ has diagonal blocks $-\sigma^1\sigma^2$ and $\gamma^2\gamma^1$ has $-\sigma^2\sigma^1$ in the same slots. Adding, the diagonal blocks are $-\{\sigma^1,\sigma^2\}=0$ (Pauli matrices anticommute), so $\{\gamma^1,\gamma^2\}=\mathbf{0}$ .

Step 4. Trace: $\gamma^1\gamma^1=-I_4$ has trace $-4$ , matching $\mathrm{tr}(\gamma^1\gamma^1)=4g^{11}=-4$ . The explicit arrays reproduce the abstract Clifford relation exactly, as they must. $(\gamma^1)^2=-I$ , $\{\gamma^1,\gamma^2\}=0$ , $\mathrm{tr}(\gamma^1\gamma^1)=-4$ .

2 · A trace with real numbers: $\tfrac14\mathrm{tr}(\slashed p\,\slashed q)$ for two on-shell momenta

Let $p^\mu=(5,0,0,4)$ and $q^\mu=(5,0,0,-4)$ (natural units), both with mass $m=3$ since $m^2=E^2-|\mathbf p|^2=25-16=9$ . Evaluate $\tfrac14\mathrm{tr}(\slashed p\,\slashed q)$ and confirm it equals the invariant $p\cdot q$ .

Step 1. Use the fundamental trace identity $\mathrm{tr}(\gamma^\mu\gamma^\nu)=4g^{\mu\nu}$ , which follows from $\mathrm{tr}(\tfrac12\{\gamma^\mu,\gamma^\nu\})=\tfrac12\,\mathrm{tr}(2g^{\mu\nu}I)=4g^{\mu\nu}$ (the commutator part is traceless). Then $\mathrm{tr}(\slashed p\,\slashed q)=p_\mu q_\nu\,\mathrm{tr}(\gamma^\mu\gamma^\nu)=4\,p_\mu q_\nu g^{\mu\nu}=4\,p\cdot q$ . So $\tfrac14\mathrm{tr}(\slashed p\,\slashed q)=p\cdot q$ .

Step 2. Compute the invariant with the mostly-minus metric: $p\cdot q=p^0q^0-\mathbf p\cdot\mathbf q=(5)(5)-(4)(-4)=25+16=\mathbf{41}$ .

Step 3. Therefore $\tfrac14\mathrm{tr}(\slashed p\,\slashed q)=41$ and the full trace is $\mathrm{tr}(\slashed p\,\slashed q)=\mathbf{164}$ . The trace of an odd number of gammas vanishes ( $\mathrm{tr}(\slashed p\,\gamma^\mu)=0$ ); only even strings survive, which is why amplitude-squared calculations always pair the gammas up. $\tfrac14\mathrm{tr}(\slashed p\,\slashed q)=p\cdot q=41$ ; full trace $=164$ .

Physics of the portrait

The gamma matrices are inseparable from the man forced to invent them. Dirac was not seeking matrices; he wanted a first-order relativistic wave equation with a positive probability density — and the only way to linearize $\sqrt{\mathbf p^2+m^2}$ was to let the coefficients anticommute. Out fell four-component spinors, spin one-half, and, two years later, the positron.

Origins · 1 minPaul Dirac and how this was discovered

Paul Dirac

1902–1984 · relativistic quantum mechanics and the electron

Paul Adrien Maurice Dirac was an engineer turned mathematician turned physicist, and it shows in the gamma matrices — pure algebra deployed with ruthless economy. In 1928, dissatisfied with the negative probabilities of the Klein–Gordon equation, he sought an equation linear in the time derivative, hence (by relativity) linear in the space derivatives: $i\partial_t\psi=(\boldsymbol\alpha\cdot\mathbf p+\beta m)\psi$ . For its square to return $E^2=\mathbf p^2+m^2$ , the coefficients $\boldsymbol\alpha,\beta$ could not be numbers — they had to anticommute and square to the identity, forcing them to be at least $4\times4$ matrices. Dirac had stumbled onto a Clifford algebra without naming it. The resulting four-component wavefunction described spin one-half automatically with magnetic moment $g=2$ , and its negative-energy solutions, reinterpreted through "hole theory", led him to predict the positron in 1931 — found by Anderson the next year. The covariant $\gamma^\mu$ repackaging, $(i\gamma^\mu\partial_\mu-m)\psi=0$ , became the grammar of quantum field theory. Dirac shared the 1933 Nobel Prize with Schrödinger. Spare in speech and notation alike, he prized "mathematical beauty" as a guide to truth — and few equations have earned the phrase more than the one that carries his name.

History

1878

— Clifford defines the geometric algebras of anticommuting square roots of a quadratic form

1928

— Dirac publishes the relativistic electron equation with anticommuting α and β matrices

1931

— Dirac interprets the negative-energy solutions and predicts the positron

1936

— Pauli proves the fundamental theorem on the equivalence of gamma representations

1937

— Majorana constructs a real (imaginary-gamma) representation and real spinor fields

1949

— Feynman introduces the slash notation and gamma-trace technology of modern QED

Going further

Texts. Peskin & Schroeder, An Introduction to Quantum Field Theory, §3.2–§3.4 and Appendix trace identities · Thomson, Modern Particle Physics, §4.5–§4.5.1 · Griffiths, Introduction to Elementary Particles, 2nd ed., §7.2–§7.3 · Halzen & Martin, Quarks and Leptons, Ch. 5 and Appendix A · Schwartz, Quantum Field Theory and the Standard Model, Ch. 10–11. The key results to keep in mind: the definitions $\gamma^0=\beta$ , $\gamma^k=\beta\alpha^k$ reproduce $(i\gamma^\mu\partial_\mu-m)\psi=0$ ; all four cases of $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}I$ hold; $(\gamma^1)^2=-I$ and $\{\gamma^1,\gamma^2\}=0$ follow from explicit Dirac–Pauli multiplication; $\mathrm{tr}(\gamma^\mu\gamma^\nu)=4g^{\mu\nu}$ and in particular $\tfrac14\mathrm{tr}(\slashed p\,\slashed q)=p\cdot q=41$ for $p=(5,0,0,4)$ , $q=(5,0,0,-4)$ ; the contraction identities $\gamma^\mu\gamma^\nu\gamma_\mu=-2\gamma^\nu$ and $\gamma^\mu\gamma_\mu=4I$ ; and current conservation $\partial_\mu(\bar\psi\gamma^\mu\psi)=0$ with positive density $j^0=\psi^\dagger\psi\ge0$ .

Prerequisites. The Dirac Equation · Four-Vector Notation · The Dirac Probability Current. Leads to. The Covariant Dirac Equation · The Adjoint Spinor and Bilinear Covariants · Trace Techniques · Weyl Spinors and Chirality.

Ask the tutor: Ask me to run any gamma-matrix trace by hand — give me a string of slashes and momenta and I will reduce it to dot products step by step using only the Clifford algebra.

Why must the gamma matrices be at least 4×4?

Because the Clifford algebra $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}I$ cannot be satisfied by anything smaller. You need four mutually anticommuting matrices, three squaring to $-I$ and one to $+I$ . Numbers cannot anticommute; $2\times2$ matrices give only three anticommuting objects (the Pauli matrices), one short of what four spacetime dimensions demand. The smallest matrices supporting four such generators are $4\times4$ — the general result being that the irreducible Clifford representation in $d$ dimensions has dimension $2^{\lfloor d/2\rfloor}=2^2=4$ for $d=4$ . So the four-component spinor is not a modelling choice; it is forced by the algebra, and with it come spin one-half and antiparticles.

Are the gamma matrices unique?

No — only the algebra is unique, not the matrices. Any four $4\times4$ matrices satisfying $\{\gamma^\mu,\gamma^\nu\}=2g^{\mu\nu}I$ form an equally valid representation, and Pauli's fundamental theorem (1936) guarantees any two such sets are related by a similarity transformation $\gamma'^\mu=S\gamma^\mu S^{-1}$ , $S$ unique up to a constant. The common choices are Dirac–Pauli (best non-relativistically), Weyl/chiral ( $\gamma^5$ diagonal, best for high-energy fermions), and Majorana (imaginary gammas, real Dirac equation). Because physics is representation-independent, every observable comes out identical whichever you pick.

Why is γ⁰ Hermitian but the spatial γ^k anti-Hermitian?

It traces back to the reality requirements on the Dirac matrices. Both $\beta$ and the $\alpha^k$ must be Hermitian so that $H=\boldsymbol\alpha\cdot\mathbf p+\beta m$ is Hermitian and energies are real. Then $\gamma^0=\beta$ inherits Hermiticity directly, while $\gamma^k=\beta\alpha^k$ is a product of two Hermitian matrices that anticommute — hence anti-Hermitian: $(\beta\alpha^k)^\dagger=\alpha^k\beta=-\beta\alpha^k=-\gamma^k$ . The mixed behaviour is no blemish; it packages into the covariant statement $\gamma^{\mu\dagger}=\gamma^0\gamma^\mu\gamma^0$ , precisely what makes $\bar\psi=\psi^\dagger\gamma^0$ the right adjoint and $j^\mu=\bar\psi\gamma^\mu\psi$ a real, conserved four-vector.

What is the Feynman slash and why is it convenient?

The slash is shorthand: $\slashed a\equiv\gamma^\mu a_\mu=\gamma^0 a^0-\boldsymbol\gamma\cdot\mathbf a$ . Its power is the identity $\slashed a\,\slashed a=a^2 I$ , immediate from the Clifford algebra — the gamma structure squares into a plain Lorentz invariant. You meet the slash everywhere: the momentum-space Dirac equation is $(\slashed p-m)u(p)=0$ ; the fermion propagator is $(\slashed p+m)/(p^2-m^2)$ ; vertices are $\gamma^\mu$ contracted with photon polarizations. Because slash products reduce via $\slashed a\,\slashed b=a\cdot b-i\sigma^{\mu\nu}a_\mu b_\nu$ to dot products plus a spin tensor, whole amplitudes collapse to scalar functions of momenta. It is the notation that turns spinor calculations from matrix bookkeeping into algebra.

How does the Dirac current fix the Klein–Gordon negative-probability problem?

The Klein–Gordon equation, second order in time, yields a "density" $\rho\propto i(\phi^*\partial_t\phi-\phi\partial_t\phi^*)$ that depends on the time derivative and can be negative — fatal for a probability. Dirac's equation is first order in time, so its conserved density needs no time derivative: $j^0=\bar\psi\gamma^0\psi=\psi^\dagger\psi=\sum_i|\psi_i|^2\ge0$ . The gamma matrices make this work: the adjoint $\bar\psi=\psi^\dagger\gamma^0$ combines with $\gamma^0$ to give exactly $\psi^\dagger\psi$ , and the Clifford algebra guarantees $\partial_\mu j^\mu=0$ for any mass. Historically this positivity was the whole point — why Dirac's linearization was taken as a genuine single-particle wave equation, until the negative-energy sea forced its reinterpretation as a field theory.

Is γ⁵ a fifth gamma matrix, and what does it do?

Not a fifth spacetime direction — there are only four $\gamma^\mu$ . The object $\gamma^5\equiv i\gamma^0\gamma^1\gamma^2\gamma^3$ is a derived matrix, the product of all four, and the "5" is historical labelling. It is Hermitian, squares to the identity, and — crucially — anticommutes with every $\gamma^\mu$ : $\{\gamma^5,\gamma^\mu\}=0$ . Those properties let you build the chirality projectors $P_{R,L}=\tfrac12(1\pm\gamma^5)$ that split a Dirac spinor into right- and left-handed Weyl pieces. In the massless limit chirality is conserved, and $\gamma^5$ becomes the engine of weak-interaction parity violation, since the $W$ couples only to $P_L\psi$ . It also builds the axial current $\bar\psi\gamma^5\gamma^\mu\psi$ and the pseudoscalar $\bar\psi\gamma^5\psi$ , two of the sixteen bilinear covariants.

Gamma matrices & the Clifford algebra

Where this lives today

The physical picture

The core equations

Limiting cases

Worked examples

1 · Verifying the Clifford algebra by explicit matrix multiplication (confidence builder)

2 · A trace with real numbers: $\tfrac14\mathrm{tr}(\slashed p\,\slashed q)$ for two on-shell momenta

Physics of the portrait

Going further

Before this

Where this leads

Gamma matrices & the Clifford algebra

Where this lives today

The physical picture

The core equations

Limiting cases

Worked examples

1 · Verifying the Clifford algebra by explicit matrix multiplication (confidence builder)

2 · A trace with real numbers: 14tr(p̸ q̸)\tfrac14\mathrm{tr}(\slashed p\,\slashed q)41​tr(pq) for two on-shell momenta

Physics of the portrait

Going further

Before this

Where this leads

2 · A trace with real numbers: $\tfrac14\mathrm{tr}(\slashed p\,\slashed q)$ for two on-shell momenta