betaFeedback

WorksPhysics and Astronomy

Two-Dimensional Euclidean Free Fermi Fields from Grassmannian Stochastic Analysis

Abstract

Motivated by Nelson's Euclidean Markov-field construction, we develop a counterpart for the massive two-dimensional free Dirac field. Starting from a Gaussian white noise with values in a Grassmann algebra on a non-commutative probability space, we construct the associated Brownian motion and a stationary Ornstein–Uhlenbeck field. The resulting Euclidean Fermi field is realized as a tempered distribution on spacetime test functions, with values in the underlying algebra. Its mixed two-point functions reproduce the Euclidean Dirac fundamental solution. Time-zero fields are obtained by conditional expectation onto the past filtration generated by the noise, while their covariance is the reflected Euclidean Dirac kernel. We show that conditional expectation is compatible with Wick ordering and maps Euclidean Wick polynomials to the corresponding time-zero Wick polynomials. The Euclidean time-translation semigroup is then constructed on time-zero fields and Wick sectors, with generators given by the one-particle Hamiltonian and by its second quantization.

Introduction

Constructive quantum field theory was originally shaped by the idea that Euclidean fields should not merely be described in terms of functional integrals, but as probabilistic objects. This has the merit of making it possible to establish criteria for a rigorous definition of Wick rotation, the key tool to reconstruct a corresponding Lorentzian theory. In the Bosonic case this viewpoint is embodied by the Osterwalder–Schrader axioms [13] and by Nelson’s construction of quantum fields from Markov fields [11], see also [16, 8]. In this last approach one starts from an Euclidean random field rather than from the canonical commutation relations. This allows to identify a distinguished time-zero algebra by means of conditional expectations recovering in addition the Hamiltonian dynamics from the Euclidean time-translation semigroup.

This programme has acquired renewed momentum in recent years thanks to the interaction between constructive quantum field theory and singular stochastic partial differential equations, which are revolving mainly around regularity structures and paracontrolled calculus [10, 9]. In turn, establishing these new frameworks has allowed to put on firm mathematical grounds the Parisi–Wu idea of stochastic quantization [14], which allows for the construction and analysis of Euclidean models beyond the perturbative level. Yet, most of the advances mentioned above concern Bosonic theories, while Fermionic models are less studied. The main structural difference lies in the anti-commuting nature of the underlying variables. This prevents a direct realization of Fermionic Euclidean fields as random variables on an ordinary commutative probability space, as already stressed in the analysis of Osterwalder and Schrader [12]. Consequently, the probabilistic side of the Nelson programme has no immediate Fermionic analogue.

A way around this obstruction has recently emerged in [1, 5]. Herein the idea consists of replacing the commutative probability space by a non-commutative counterpart. This is a unital C∗C^*-algebra endowed with a state, that is a positive and normalized functional thereon. Grassmann random variables are therefore regarded as algebra homomorphisms from an exterior algebra into this non-commutative probability space. In this setting one can define key probabilistic notions ranging from a Grassmannian white noise to conditional expectations. These allow to devise a natural stochastic language to encode anti-commuting variables, that is a framework in which Nelson’s construction can be revisited for Fermionic fields. On the one hand, it is worth mentioning that this non-commutative setting has already proved useful in the study of weakly coupled subcritical Euclidean Fermionic field theories [6]. On the other hand, it suggests that this might be the right framework to analyze interacting two-dimensional models, among which the Thirring model is a natural test case, see [3] for a recent analysis in the framework of singular stochastic partial differential equations.

The purpose of the present paper is to carry out part of this programme, focusing on the non-interacting, massive Euclidean Dirac field in two dimensions. More precisely, we construct a Euclidean free Fermi field starting from a Grassmannian white noise on a non-commutative probability space. This reconstruction may be viewed as a non-commutative fermionic analogue of Nelson’s construction from Markov fields [11]. Here, the Markov structure comes from an auxiliary stationary Grassmannian Ornstein–Uhlenbeck field φ\varphi and its driving white noise. An algebra-valued Euclidean Fermi field ψ\psi, whose mixed two-point functions reproduce the Euclidean Dirac fundamental solution, is derived from φ\varphi by applying a differential operator, local in time and space, to its second component. Hence, we do not require ψ\psi itself to be Markovian. Subsequently, we exploit the Markovianity of the auxiliary field and the associated conditional expectation to show that the resulting evolution coincides with the second quantization of the one-particle semigroup. A Markov property has been established also in [7] in their gauge-space construction, where the relevant algebra is generated by Euclidean fermionic gauge fields. Their proposal to introduce interactions through a normal central change of state encounters an obstruction. Our noise-based construction offers instead a different possible starting point for interacting models. We restrict the present analysis to the free theory so as to isolate the Markovian construction from renormalization issues. The main results can be summarized as follows:

  • We construct a Grassmannian white noise and the associated Grassmannian Brownian motion on a non-commutative probability space. The covariance is specified by a bounded operator whose associated two-point function is an antisymmetric bilinear form, as required by the fermionic nature of the variables. The resulting process satisfies the corresponding Gaussian Wick rule and has independent increments.

  • We define a stationary Grassmannian Ornstein–Uhlenbeck field and, from it, a two-dimensional Euclidean free Fermi field. Its only non-vanishing two-point functions are the mixed ones. These coincide with the integral kernel of the Euclidean Dirac propagator.

  • We construct zero-time Fermionic fields by conditional expectation onto the algebra generated by the past noise. Their covariance is the reflected Euclidean Dirac kernel and the conditional expectation is compatible with Wick ordering. In other words, Euclidean Wick polynomials are mapped to the corresponding Wick polynomials of the zero-time fields.

  • We show that Euclidean time translations induce a strongly continuous semigroup on the one-particle zero-time space. Its generator is the positive one-particle Hamiltonian Ax=(−∂x2+m2)1/2A_x = (-\partial_x^2 + m^2)^{1/2}. On Wick polynomials the induced generator is the corresponding second quantization.

The paper is organized as follows. In the first part of Section 1 we fix the analytic conventions for two-dimensional Euclidean Dirac fields and recall succinctly the Osterwalder–Schrader framework. In Section 2 we introduce the non-commutative probability space, Grassmann random variables, Gaussian covariances and conditional expectations. In Section 3 we discuss the construction of the Grassmannian white noise, the Ornstein–Uhlenbeck field, the Euclidean free Fermi field, the time-zero fields and the Euclidean time-evolution semigroup. The appendix contains the proof of the compatibility between conditional expectation and Wick ordering.

Analytic Structures

In this section we fix the notation and conventions used throughout. We take as background the two-dimensional Euclidean space (R2,δ)(\mathbb{R}^{2},\delta), with signature (+,+)(+,+) and Cartesian coordinates (t,x)(t,x). We shall refer to tt as Euclidean time, in contrast to its Lorentzian counterpart tMt^{M}, related to it by the Wick rotation

t=−itMt=-it^{M}

To simplify the notation we shall denote a generic point of the Euclidean space by x≡(t,x)x\equiv(t,x) and its Lorentzian counterpart by xM≡(tM,x)x^{M}\equiv(t^{M},x).

Function spaces. On R2\mathbb{R}^{2} we denote by S(R2)\mathcal{S}(\mathbb{R}^{2}) the space of rapidly decreasing smooth functions, endowed with its standard locally convex Fréchet topology. Its topological dual, the space of tempered distributions, is denoted by S′(R2)\mathcal{S}'(\mathbb{R}^{2}). Given τ∈R\tau\in\mathbb{R}, we introduce the subspaces

Sτ−(R2):={v∈S(R2)∣v(t,x)=0, for t>τ},Sτ+(R2):={v∈S(R2)∣v(t,x)=0, for t<τ}.(1)\mathcal{S}_{\tau}^{-}(\mathbb{R}^{2}):=\{v\in\mathcal{S}(\mathbb{R}^{2})\mid v(t,x)=0,\text{ for }t>\tau\},\qquad\mathcal{S}_{\tau}^{+}(\mathbb{R}^{2}):=\{v\in\mathcal{S}(\mathbb{R}^{2})\mid v(t,x)=0,\text{ for }t<\tau\}. \tag*{(1)}

Equivalently, the elements of Sτ+(R2)\mathcal{S}_{\tau}^{+}(\mathbb{R}^{2}) (resp. Sτ−(R2)\mathcal{S}_{\tau}^{-}(\mathbb{R}^{2})) are Schwartz functions supported in the closed half-space {t≥τ}\{t\geq\tau\} (resp. {t≤τ}\{t\leq\tau\}). Since the Fourier transform will play an important role, we fix the convention

F:S(R2)⟶S(R2),v(t,x)⟼v^(ω,k)=12π∫R2dtdx v(t,x)e−i(ωt+kx).\mathcal{F}:\mathcal{S}(\mathbb{R}^{2})\longrightarrow\mathcal{S}(\mathbb{R}^{2}),\qquad v(t,x)\longmapsto\widehat{v}(\omega,k)=\frac{1}{2\pi}\int_{\mathbb{R}^{2}}\mathrm{d}t\mathrm{d}x\,v(t,x)e^{-i(\omega t+kx)}.

Fourier transforms with respect to only one of the two variables are defined analogously.

Euclidean Dirac Operator – We consider distributional Euclidean spinors and cospinors, namely elements respectively of

S′(R2;C2),S′(R2;(C2)∗).\mathcal{S}'(\mathbb{R}^{2};\mathbb{C}^{2}),\qquad\mathcal{S}'(\mathbb{R}^{2};(\mathbb{C}^{2})^{*}).

Throughout the paper, using the standard Hermitian pairing, we shall freely identify C2\mathbb{C}^{2} and (C2)∗(\mathbb{C}^{2})^{*} whenever necessary and if no confusion can arise. The Euclidean Dirac operator and its formal adjoint are

DE=γEμ∂μ+m:S′(R2;C2)⟶S′(R2;C2),DE∗=−γEμ∂μ+m:S′(R2;C2)⟶S′(R2;C2).(2)D_{E}=\gamma_{E}^{\mu}\partial_{\mu}+m:\mathcal{S}'(\mathbb{R}^{2};\mathbb{C}^{2})\longrightarrow\mathcal{S}'(\mathbb{R}^{2};\mathbb{C}^{2}),\qquad D_{E}^{*}=-\gamma_{E}^{\mu}\partial_{\mu}+m:\mathcal{S}'(\mathbb{R}^{2};\mathbb{C}^{2})\longrightarrow\mathcal{S}'(\mathbb{R}^{2};\mathbb{C}^{2}). \tag*{(2)}

Here m≥0m\geq0 is a mass parameter, while γEμ∈M(2;C)\gamma_{E}^{\mu}\in\mathcal{M}(2;\mathbb{C}), μ=0,1\mu=0,1, are Euclidean gamma matrices. They are Hermitian and satisfy the Clifford relations

{γEμ,γEν}=2δμνI2,μ,ν=0,1.\{\gamma_{E}^{\mu},\gamma_{E}^{\nu}\}=2\delta^{\mu\nu}\mathbb{I}_{2},\qquad\mu,\nu=0,1.

A convenient choice is given by the Pauli matrices

γE0=σ1=(0110),γE1=σ2=(0−ii0).\gamma_{E}^{0}=\sigma^{1}=\begin{pmatrix}0&1\\1&0\end{pmatrix},\qquad\gamma_{E}^{1}=\sigma^{2}=\begin{pmatrix}0&-i\\i&0\end{pmatrix}.

For later use, we recall the identities

DEDE∗=(−Δ+m2)I2,DE∗DE=(−Δ+m2)I2.\begin{aligned} D_{E}D_{E}^{*}&=(-\Delta+m^{2})\mathbb{I}_{2},\\ D_{E}^{*}D_{E}&=(-\Delta+m^{2})\mathbb{I}_{2}. \end{aligned}

where Δ:=∂t2+∂x2\Delta:=\partial_{t}^{2}+\partial_{x}^{2} is the Euclidean Laplacian on R2\mathbb{R}^{2}, while I2\mathbb{I}_{2} denotes the 2×22\times2 identity matrix. For m>0m>0, we denote by G∈S′(R2)G\in\mathcal{S}'(\mathbb{R}^{2}) the translation-invariant fundamental solution of −Δ+m2-\Delta+m^{2} which decays at infinity. Its integral kernel reads

G(t,x)=12πK0(mt2+x2),G(t,x)=\frac{1}{2\pi}K_{0}\left(m\sqrt{t^{2}+x^{2}}\right),

where K0K_{0} is the modified Bessel function of second kind of order 0. Equivalently, in Fourier space, we have

G^(ω,k)=12π1ω2+k2+m2.\widehat{G}(\omega,k)=\frac{1}{2\pi}\frac{1}{\omega^{2}+k^{2}+m^{2}}.

Observe that, in the massless case m=0m=0, no fundamental solution vanishing at infinity exists in two dimensions and one must instead choose an infrared prescription. We avoid entering into the details leaving this choice understood. Consequently, the Euclidean Dirac fundamental solution is the matrix-valued tempered distribution

SE:=DE∗(GI2)∈S′(R2;End⁡(C2)).(3)S_{E}:=D_{E}^{*}(G\mathbb{I}_{2})\in\mathcal{S}'(\mathbb{R}^{2};\operatorname{End}(\mathbb{C}^{2})). \tag*{(3)}

whose expression in Fourier space is

S^E(ω,k)=12π−iγE0ω−iγE1k+mI2ω2+k2+m2.\widehat{S}_{E}(\omega,k)=\frac{1}{2\pi}\frac{-i\gamma_{E}^{0}\omega-i\gamma_{E}^{1}k+m\mathbb{I}_{2}}{\omega^{2}+k^{2}+m^{2}}.

Observe that this expression is valid also for m=0m=0 since, in this case, SES_{E} is locally integrable in a neighbourhood of (ω,k)=(0,0)(\omega,k)=(0,0).

Osterwalder–Schrader Distributions and Reconstruction

In this subsection we recall the Osterwalder–Schrader framework for a two-dimensional Bosonic scalar theory. We follow mainly [13] and [16]. We omit the proofs, in particular of the reconstruction theorem, since they can be found in the mentioned references. Actually our main goal is to fix the functional-analytic setting and to state the class of Euclidean correlation functions to which the theorem applies, as a preparation for the discussion of the Fermionic counterpart in § 3. We identify the large diagonal

Δn:={(x1,…,xn)∈R2n∣xi=xj for some i≠j}.\Delta_n := \{(x_1,\ldots,x_n) \in\mathbb{R}^{2n} \mid x_i = x_j \text{ for some } i \ne j\}.

At the same time we set

Sn:=S(R2n)and0Sn:={u∈Sn∣∂αu∣Δn=0, for every multiindex α},\mathcal{S}_n := \mathcal{S}(\mathbb{R}^{2n}) \quad\text{and} \quad{}^{0}\mathcal{S}_n := \{u \in\mathcal{S}_n \mid\partial^\alpha u|_{\Delta_n} = 0,\text{ for every multiindex } \alpha\},

where 0Sn{}^{0}\mathcal{S}_n is a closed subspace of Sn\mathcal{S}_n, see [13].

Remark 1 These spaces come accompanied with their topological duals Sn′≡S′(R2n)\mathcal{S}'_n \equiv\mathcal{S}'(\mathbb{R}^{2n}) and (0Sn)′({}^{0}\mathcal{S}_n)'. Although in [13] the latter is considered, in this work we shall assume from the outset that the Schwinger functions are represented by tempered distributions Sn∈Sn′S_n \in\mathcal{S}'_n satisfying the Osterwalder–Schrader axioms. If one starts instead from an element lying in (0Sn)′({}^{0}\mathcal{S}_n)', an extension to Sn′\mathcal{S}'_n exists by the Hahn–Banach theorem. Yet, such extension is nonunique and it might not preserve the structural properties encoded by the Osterwalder–Schrader axioms unless these are imposed. Nonetheless, the ambiguity is supported on the large diagonal Δn\Delta_n and, once a compatible extension has been fixed, this does not affect the reconstruction.

Two additional ingredients are needed:

  1. the Euclidean time reflection

ϑ:R2→R2,x=(t,x)↦ϑx=(−t,x),(4)\vartheta: \mathbb{R}^2 \to\mathbb{R}^2,\qquad x=(t,x) \mapsto\vartheta x=(-t,x), \tag*{(4)}

which extends to Sn\mathcal{S}_n as the map

References

  1. [1][ABDG22] S. Alberverio, L. Borasi, F. De Vecchi, and M. Gubinelli. Grassmannian stochastic analysis and the stochastic quantization of Euclidean fermions. Probability Theory and Related Fields, 183:909–995, 2022.arxiv.org/abs/2004.09637
  2. [2][AM82] H. Araki and T. Masuda. Positive cones and Lₚ-spaces for von Neumann algebras. Publ. Res. Inst. Math. Sci., 18(2):339–411, 1982.
  3. [3][BCDR24] A. Bonicelli, B. CosterI, C. Dappiaggi, and P. Rinaldi. A microlocal investigation of stochastic partial differential equations for spinors with an application to the Thirring model. Mathematical Physics, Analysis and Geometry, 27:16, 2024.arxiv.org/abs/2309.16376
  4. [4][BR97] O. Bratteli and D. W. Robinson. Operator Algebras and Quantum Statistical Mechanics 2: Equilibrium States, Models in Quantum Statistical Mechanics. Springer, 2nd edition, 1997.
  5. [5][DFGG25] F. De Vecchi, L. Fresta, M. Gordina, and M. Gubinelli. Non-commutative Lᵖ spaces and Grassmann stochastic analysis. Probability Theory and Related Fields, 192:949–1029, 2025.DOI
  6. [6][DFG25] F. De Vecchi, L. Fresta, and M. Gubinelli. A stochastic analysis of subcritical Euclidean fermionic field theories. The Annals of Probability, 53(3):906–966, 2025.arxiv.org/abs/2210.15047
  7. [7][FO74] J. Fröhlich and K. Osterwalder. Is there a Euclidean field theory for fermions? Helv. Phys. Acta, 47:781–805, 1974.
  8. [8][GJ87] J. Glimm and A. Jaffe. Quantum Physics: A Functional Integral Point of View. Springer, 2nd edition, 1987.
  9. [9][GIP15] M. Gubinelli, P. Imkeller, and N. Perkowski. Paracontrolled distributions and singular PDEs. Forum of Mathematics, Pi, 3:e6, 2015.arxiv.org/abs/1210.2684
  10. [10][Hai14] M. Hairer. A theory of regularity structures. Inventiones Mathematicae, 198(2):269–504, 2014.
  11. [11][Nel73] E. Nelson. Construction of quantum fields from Markoff fields. Journal of Functional Analysis, 12(1):97–112, 1973.DOI
  12. [12][OS73a] K. Osterwalder and R. Schrader. Euclidean Fermi fields and a Feynman–Kac formula for boson–fermion models. Helv. Phys. Acta, 46:277–302, 1973.DOI
  13. [13][OS73b] K. Osterwalder and R. Schrader. Axioms for Euclidean Green’s functions. Communications in Mathematical Physics, 31(2):83–112, 1973.DOI
  14. [14][PW81] G. Parisi and Y.-S. Wu. Perturbation theory without gauge fixing. Scientia Sinica, 24(4):483–496, 1981.DOI
  15. [15][Rud87] W. Rudin. Real and Complex Analysis. McGraw-Hill, 3rd edition, 1987.
  16. [16][Sim74] B. Simon. The P(Φ)₂ Euclidean (Quantum) Field Theory. Princeton University Press, 1974.
  17. [17][SW64] R. F. Streater and A. S. Wightman. PCT, Spin and Statistics, and All That. W. A. Benjamin, New York, 1964.

Paper details

vlnc:W_2b7fp63r
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents