Fermionic, Gauge, and Lattice Reflection Positivity

Reflection positivity survives for fermions and lattice gauge fields only after the reflected algebra is changed to match grading, link orientation, and gauge invariance. The scalar formula cannot be copied unchanged: the reflection is antilinear and order reversing, physical positivity is tested on the appropriate even or gauge-invariant positive-time subalgebra, and the lattice action must admit a reflection-positive decomposition.

Required background. Osterwalder–Schrader axioms and reflection positivity supplies the scalar condition. Reflection positivity and Hilbert-space reconstruction supplies the quotient and transfer semigroup. Grassmann functional integrals for free fermions supplies Berezin integration and fermionic signs.

Helpful background. Reflection positivity and transfer-matrix criteria gives the regulated construction. The free Maxwell field and gauge redundancy explains why gauge-dependent elementary fields need not act on the positive physical Hilbert space.

Graded reflection

Let $\mathcal A_+$ be the Grassmann algebra generated by fermion fields supported at positive Euclidean time. A fermionic reflection $\Theta$ is antilinear and reverses order:

$$\Theta(FG)=\Theta(G)\Theta(F), \qquad \Theta(cF)=\overline c\,\Theta(F).$$

On Dirac generators it also exchanges $\psi$ and $\bar\psi$ and inserts the matrix implementing reflection in the Euclidean time direction. The exact matrix and phase depend on gamma-matrix conventions; they must be chosen so that $\Theta^2$ has the required action on the even physical algebra. The invariant content is the adjoint-like, order-reversing map, not a convention-specific component formula.

For even $F\in\mathcal A_+$, graded reflection positivity takes the form

$$\langle\Theta(F)F\rangle_E\geq0.$$

Odd sectors may require a twist or a paired formulation; one must state which algebra carries the positive form. A Grassmann functional is not a positive probability measure, so positivity is a property of this reflected pairing, not of a pointwise Berezin weight. The systematic Euclidean fermion construction and its relation to relativistic fields are developed in Fröhlich and Osterwalder 1974, especially §§2–4, pp. 781–805.

Link reflection and gauge-invariant observables

On a hypercubic lattice, choose a reflection plane either through a time slice or midway between adjacent slices. A link $U_{x,\mu}\in G$ is oriented from $x$ to $x+\hat\mu$. Reflection maps it to the oppositely oriented reflected link, hence to the group inverse or adjoint in a unitary representation. Links crossing the plane require their own assignment; treating them as scalar site variables loses orientation information.

Let $\mathcal A_+$ contain functions of links and matter fields on the positive side. For gauge theory, the physical form is tested on gauge-invariant $F$:

$$\int d\mu(U,\bar\psi,\psi)\,\Theta(F)F\geq0.$$

The restriction is substantive. Gauge-variant states include redundant directions, while gauge fixing and Faddeev–Popov ghosts commonly introduce an indefinite auxiliary state space. Failure of positivity for a gauge-fixed elementary propagator therefore does not by itself disprove positivity of gauge-invariant observables; conversely, positivity of gauge-invariant observables does not make every gauge-fixed correlator positive.

A standard sufficient proof splits the action into positive, negative, and crossing pieces. Schematically, the Boltzmann factor is expanded as

$$e^{-S}=e^{-S_+-\Theta S_+} \sum_\alpha c_\alpha\,\Theta(B_\alpha)B_\alpha, \qquad c_\alpha\geq0,$$

with $B_\alpha\in\mathcal A_+$. Haar integration and the nonnegative coefficients then make the reflected expectation a sum of squares. For compact gauge groups, character expansions provide this structure for the Wilson plaquette action. Adding extended loops or improvement terms changes the crossing interaction; reflection positivity must be rechecked coefficient by coefficient and is not guaranteed by gauge invariance alone.

First QFT application: Wilson lattice gauge theory

For the Wilson action, plaquettes wholly on either side pair under reflection, while a plaquette crossing the reflection plane factorizes into a positive-side matrix element and its reflected conjugate. Character orthogonality turns the gauge integral into a nonnegative sum. With Wilson fermions, graded order reversal and the hopping-term decomposition supply the corresponding fermionic factors.

Osterwalder and Seiler proved physical positivity for lattice Yang–Mills theories with fermions and used it to construct the physical Hilbert-space framework Osterwalder and Seiler 1978, pp. 440–471. Menotti and Pelissetto extended the Wilson-action proof to reflections through planes containing sites; in their four-dimensional Wilson-fermion conventions the gauge-invariant inequality is stated for hopping parameter $K<1/6$, and the combined site- and link-reflection argument yields a positive transfer matrix Menotti and Pelissetto 1987, §§1–2, pp. 369–373.

The reconstructed lattice inner product is formed from gauge-invariant positive-time functionals modulo null vectors. One-step time translation descends to the transfer operator $T$; reflection positivity gives $T\geq0$ in the stated construction, and gauge averaging projects to the Gauss-law state space. The detailed regulated criterion and its limitations are given in reflection positivity and transfer-matrix criteria.

This is a fixed-lattice theorem. A positive transfer matrix at nonzero lattice spacing neither proves that a continuum limit exists nor that the limiting theory is interacting. Those require uniform estimates and convergence of gauge-invariant correlations.

Adversarial fermion-sign test

Take two odd generators $\psi_1,\psi_2\in\mathcal A_+$ and $F=\psi_1\psi_2$. Correct reflection gives

$$\Theta(F)=\Theta(\psi_2)\Theta(\psi_1) =-\Theta(\psi_1)\Theta(\psi_2).$$

If one applies the bosonic scalar rule without reversing order, the reflected monomial differs by a minus sign. The proposed “norm” $\langle\Theta(F)F\rangle_E$ is therefore assigned the opposite sign in this two-fermion sector. No later quotient can repair a form whose sign was defined incorrectly; the missing hypothesis is the graded anti-automorphism.

The same diagnostic catches a second error. If a temporal link is reflected without reversing its orientation, $U$ appears where $U^\dagger$ is required. The Haar-integral factor is no longer a matrix element times its conjugate, so the sum-of-squares proof fails.

Independent checks

• Involution: verify antilinearity, order reversal, and $\Theta^2$ on the declared physical algebra.
• Support: specify site- or link-centered reflection and assign every crossing link once.
• Gauge sector: state whether positivity holds for all functions, only gauge-invariant functions, or an averaged subalgebra.
• Coefficients: inspect the character or hopping expansion; every crossing coefficient used as a square weight must be nonnegative.
• Transfer step: distinguish positivity of $T$, positivity of $T^2$, and mere self-adjointness.
• Limit claim: keep fixed-cutoff reflection positivity separate from continuum existence and universality.

Exercise

Why does an orientation-reversing reflection send a unitary link variable to the adjoint link?

Solution

A link is a parallel transporter from its initial to its final site. Reversing the geometric orientation exchanges those endpoints. Parallel transport along the reversed path is the inverse group element $U^{-1}$, which equals $U^\dagger$ in a unitary representation. This adjoint is also what turns the reflected crossing factor into the complex conjugate needed for a nonnegative square.

References

• Fröhlich, Jürg, and Konrad Osterwalder. “Is There a Euclidean Field Theory for Fermions?” Helvetica Physica Acta 47 (1974): 781–805. Open scan.
• Menotti, Pietro, and Andrea Pelissetto. “General Proof of Osterwalder–Schrader Positivity for the Wilson Action.” Communications in Mathematical Physics 113 (1987): 369–373. doi:10.1007/BF01221251. Open PDF.
• Osterwalder, Konrad, and Erhard Seiler. “Gauge Field Theories on a Lattice.” Annals of Physics 110 (1978): 440–471. doi:10.1016/0003-4916(78)90039-8.