Comments on supersymmetric renormalization in two-dimensional curved spacetime

In this technical note we introduce a manifestly gauge-invariant and supersymmetric procedure to regularize and renormalize one-loop divergences of chiral multiplets in two-dimensional N\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \mathcal{N} $$\end{document} = (2, 2) theories in curved spacetime. We apply the method, a version of Pauli-Villars regularization, to known examples of supersymmetric localization and demonstrate that the partition functions are renormalized.

1 Introduction: supersymmetric regularization and countererms for a chiral multiplet In this note we explain how a version of Pauli-Villars (PV) regularization can be used for the one-loop determinant of an N = (2, 2) chiral multiplet coupled to background gauge and gravity multiplets. The relevant facts about N = (2, 2) rigid supersymmetry and supergravity are summarized in appendices A and B, respectively.
Though the gauge and the gravity multiplets are treated as non-dynamical here, one may promote them to dynamical fields. An application is discussed in [1].

Pauli-Villars regularization
A simple choice of supersymmetric regularization is a version of Pauli-Villars regularization. 1 Let us consider a chiral multiplet with charge +1 for the flavor symmetry group U(1) F and the complex central charge z. Often in a supersymmetric background, the complex scalar σ in the non-dynamical U(1) F vector multiplet is constant; in that case σ is called the twisted mass. In order to regularize the one-loop amplitudes, we add to the action supersymmetric kinetic terms for extra 2N PV −1 chiral multiplets such that N PV −1 of them have the usual statistics, and the remaining N PV have the opposite statistics. We let the index J run from 0 to 2N PV − 1, and j from 1 to 2N PV − 1. We use J = ±1 to denote the statistics, +1 and −1 for the usual and the opposite, respectively. We set 0 = +1. Let the j-th multiplet have an integer charge b j under U(1) F . In order to make the regulator fields very heavy, we couple them to another vector multiplet for an abelian symmetry U(1) PV , under which the j-th PV multiplet has charge a j . We take this symmetry group to be non-compact, U(1) PV = R rather than U(1), so that a j do not have to be an integer. 2 For convenience we take a j to be positive: a j > 0. For the vector R-symmetry, we assign R-charge c j q to the j-th PV multiplet. We set a 0 = 0, b 0 = 1, and q 0 = q. We denote the complex scalar in the U(1) PV vector multiplet by Λ. As we will see, σ and Λ can be JHEP12(2017)081 constants or non-trivial functions in spacetime, depending on the supergravity background.
For later convenience we define the following combinations of parameters. (1.6)

A counterterm action for the chiral multiplet
At the classical level, a field theory is specified by the action functional. To define a theory at the quantum level, one must specify the counterterms in addition to a regularization method. We now propose a counterterm action S ct which, combined with the Pauli-Villars regularization above, defines unambiguously the theory of a chiral multiplet in the background of gauge and gravity multiplets. Let us denote the renormalized action for the background fields by S ren and the regularized one-loop determinant by Z PV 1-loop . The partition function, which is a functional of the gravity multiplet and the U(1) F gauge multiplet, is given as The limit Λ → +∞ is taken pointwise when Λ is not constant in spacetime.

JHEP12(2017)081
We will give the counterterm action S ct in terms of a twisted superpotential W ct . As is well known a twisted chiral multiplet can be formed from a gauge multiplet. The scalars σ and Λ are the bottom components of the U(1) F and U(1) PV gauge multiplets. As reviewed in appendix B.3 the gravity multiplet also gives rise to a twisted chiral multiplet 4 whose scalar component is denoted as H.
Let µ > 0 be an arbitrary (renormalization) mass scale. Our counterterm action S ct is constructed from the twisted superpotential (1.8) by applying the rule (A.15). We choose the branch cut of the logarithm to be along the negative real axis. In later sections we demonstrate that the counterterm action given by (1.8) successfully regularizes the one-loop determinants obtained by supersymmetric localization. This is the main result of this note. The second term in (1.8) is essentially minus the familiar effective twisted superpotential from integrating out massive chiral multiplets [3]; here they are coupled to the gauge multiplets not only for the usual U(1) F symmetry, but also for the U(1) PV and R-symmetries. The requirement of supersymmetry and gauge symmetry does not fix the counterterm completely; in including ia j in the first term, we made a particular choice for the finite counterterm.
To exhibit more features of (1.8), we expand it in 1/Λ, assuming that Λ > 0. (1.9) The first term of (1.9) gives, via the relation (A.15), the supersymmetrization of the wellknown countertermĉ in the action, whereĉ UV = 1 − q is the UV central charge and R is the scalar curvature. 5 The second term describes the renormalization of the FI parameter and the theta angle for 4 The gravity multiplet contains a gauge field A R µ for the vector R-symmetry, and one can construct the corresponding gauge multiplet from the fields in the gravity multiplet. The R-symmetry gauge multiplet then gives rise to a twisted chiral multiplet whose scalar component is H (up to a multiplicative numerical factor). H differs from the graviphoton field strength H by terms involving gravitini. 5 In our convention R = +2/ 2 for the round sphere of radius .
In appendix C we demonstrate by a flat space computation that the counterterm indeed cancels the divergence in the effective FI-parameter. We also show that the relation (1.11) gives rise to the standard σ-dependent effective FI-parameter [3]. Using r(µ) and θ we define the complexified FI parameter (1.12) The part of the renormalized action S ren that survives in localization calculations below is the FI and the theta terms, both of which can be represented by the twisted superpotential −tΣ/4π.

SUSY sphere partition function
In this section we apply our renormalization procedure to the two-dimensional sphere partition function [2,4]. Let us write The supergravity background is given in our convention as and is otherwise arbitrary. For the PV gauge multiplet we have, in analogy with (2.4), The SUSY parameters in the convention explained in the appendix are (2.7) 6 The renormalized FI parameter r(µ) is the coefficient of D/2π in Lren, while the bare parameter r0(Λ) is the corresponding coefficient in Lren + Lct, where the action S and the Lagrangian L are related as

JHEP12(2017)081
The computations in [2,4,5] for a single chiral multiplet give a formal, i.e., nonconvergent, expression for the one-loop determinant as an infinite product: (2.8) With the Pauli-Villars ghost contributions, the infinite product (2.9) is absolutely convergent if the constraints (1.1) are satisfied. We can express (2.9) in terms of gamma functions using the infinite product formula (2.11) We remove the absolute value signs, |B| → B, by applying the equality Γ( where the second line is valid for Λ −1 . The final partition function for a chiral multiplet is . (2.13) A convenient choice that gives a simple formula is to take µ = 1/ . Then . (2.14) This is the formula that is often quoted in the literature. For large mass Re σ we can further integrate out the physical chiral multiplet and shift the theta angle. The sign of the shift is correlated with the sign of the mass (as in the discussion of topological insulators).

JHEP12(2017)081 3 A-twisted theory with omega deformation
The next example is a chiral multiplet in the omega-deformed A-twisted theory.
Let us consider the two-sphere with metric where z = re iϕ . Let us define the vector field Let us use the hat to indicate frame indices. Then Following [6] we consider the SUSY parameters where the omega-deformation parameter a is an arbitrary constant. The gravitino variations vanish when 8 In order to preserve SUSY, we require that δλ ± = 0, δλ ± = 0. These conditions can be rewritten, with rotational symmetry (W -invariance) assumed, as They are solved by with σ and σ being arbitrary functions of r. References [7,8] found the formal expression for the one-loop determinant Here the constant σ is defined by the relations σ(north pole(r = 0)) = σ + B 2 a, σ(south pole) = σ − B 2 a. We assume that the gauge flux for U(1) PV vanishes, so that 8 In our convention ωz = iωzẑẑ, ωz = −iωzẑẑ, where ω a b = dx µ ωµ a b is the usual spin connection determined by de a + ω a b ∧ e b = 0.

JHEP12(2017)081
the associated quantities Λ and Λ, corresponding to σ and σ respectively, coincide. The PV-regularized version of (3.8) is then As in the previous section we rewrite this using gamma functions: Let us assume here that Λ > 0. We find (3.12) The counterterm action can be evaluated in the background configuration by reducing it to the integral of a total derivative expression. The resulting on-shell value of the master counterterm is such that . (3.14)

Discussion
In this paper we applied the Pauli-Villars regularization to particular one-loop divergences arising from chiral multiplets in two dimensions. The Pauli-Villars method has been used in other supersymmetric contexts; see for example [9,10] and the references therein. It may be possible and useful to use the central charges, i.e., the charges for the graviphotons, as the large masses for the Pauli-Villars regulator fields. Such a scheme would be more conventional in the sense that the counterterms will depend on the physical vector and gravity multiplets, but not on the Pauli-Villars vector multiplets we introduced in section 1.1. A drawback is that we cannot completely recycle the one-loop computations that already exist in the literature because we need to know the dependence on the central charge z. The drawback may not be significant because the dependence on the central

JHEP12(2017)081
charge is expected to be essentially the same as the dependence on the twisted mass. In particular in the index theory approach, the one-loop determinant is computed from the square of the localizing supercharge, in which the central charge enters in the same way as the twisted mass [6]. Another complication is that the graviphotons and the central charge are not completely independent because they must obey flux quantization conditions [6].
In what sense are our results supersymmetric? In the given rigid limit of supergravity, the preserved supersymmetries are restricted and finite dimensional. Our regularization method and the counterterms are invariant under such rigid supersymmetry transformations. Our regularization scheme, however, has manifest invariance in somewhat more general circumstances as we now explain.
In principle, our regularizations and counterterms are applicable in general, not necessarily supersymmetric, backgrounds. Without regularization and countertrerms, integrating out matter supermultiplets produces, in an intermediate step, a cut-off dependent expression 9 that is manifestly invariant under general supersymmetry transformations. In general the UV cut-off Λ(x) itself transforms and is position dependent. In the limit Λ(x) → +∞, we expect to obtain a finite functional (1.7) for the background gravity and gauge supermultiplets. The point is that this functional is invariant under the general supergravity transformations of the multiplets. The result of a localization computation as described in the previous paragraph is a specialization of the functional to the specific supersymmetric background.

A Rigid N = (2, 2) SUSY in curved spacetime
In [6], quantum field theories coupled to a version of two-dimensional N = (2, 2) supergravity were studied. The version of supergravity coincides with the dimensional reduction of four-dimensional N = 1 new minimal supergravity, and as such it requires the presence of a U(1) R-symmetry. We adapt the convention such that it is the vector R-symmetry. The formulas in this section are for rigid SUSY: they are obtained from those of appendix B by setting gravitini, their variations, eẑ z , and eẑ z to zero.

A.1 SUSY transformations
The SUSY transformations are related to the supercharges as

JHEP12(2017)081
From [6] and in the convention of [11], we have the following supersymmetry transformations for a vector multiplet in the Wess-Zumino gauge: For chiral and anti-chiral multiplets the transformations are Similarly, fields in twisted chiral and anti-chiral multiplets transform as

A.2 Relations between multiplets
An abelian gauge multiplet gives rise to twisted chiral and twisted anti-chiral multiplets, indicated by superscript "ga": Holomorphic and anti-holomorphic functions W (v (j) ) and W (v (j) ) give rise to twisted chiral and twisted anti-chiral multiplets

A.3.1 Chiral multiplet
For the chiral multiplet, the supersymmetric UV action is Here The superpotential coupling for several chiral multiplets (φ i , ψ i ∓ , F i ) is

A.3.2 Vector multiplet
For the vector multiplet, the action is (A.12)

B.1 Supergravity transformations
The symbol D µ denotes the minimal derivative that is covariant with respect to general coordinate, local Lorentz, R-symmetry, and central charge transformations. As such it is constructed from the partial derivative and the appropriate gauge fields only.
In describing supergravity, we often use complex coordinates (x µ ) = (z, z) and the corresponding frame one-forms (vielbein) (e a ) = (eẑ, eẑ). Thus the Greek indices µ, ν, . . . are for coordinates, and the Latin indices a, b, . . . are for the frame (or flat, or tangent space) indices. The latter take a valueẑ orẑ. The metric takes the form

JHEP12(2017)081
Using the gamma matrices and the conventions we can also write For the gravitini and the R-symmetry gauge field, we have We introduced the following notations.
The equations (B.11) and (B.15) define a covariant derivative D µ , which acts on a general spinor with R-charge +1 and appears in other formulas such as (B.10). Similarly D µ defined by (B.12) and (B.16) acts on a spinor with R-charge −1. When acting on ψ σ and ψ σ , the covariant derivatives D ρ and D ρ are defined so that they do not contain Christoffel symbols. 10 10 The same is true for the four-dimensional new minimal supergravity [17]. If we included the torsionless Christoffel symbol constructed from the metric (or the vielbein) in Dρ and Dρ they would cancel out in ρσ Dρψσ. Thus the formula (B.10) is generally covariant. If we included the Christoffel symbol corresponding to ωµ + Kµ which has torsion, it would contribute, but this would be the wrong definition.

JHEP12(2017)081
B.3 Relations between multiplets B.3.1 Twisted chiral and twisted anti-chiral multiplets from the gravity multiplet The following fields constructed from the gravity multiplet forms a twisted chiral multiplet: form a twisted anti-chiral multiplet.

B.3.2 Gauge multiplet from the gravity multiplet
We can also construct a gauge multiplet (R-symmetry gauge multiplet) from the gravity multiplet: (B.29)

B.3.3 Twisted chiral multiplet from the abelian gauge multiplet
An abelian gauge multiplet gives rise to twisted chiral and twisted anti-chiral multiplets, indicated by superscript "ga":

B.3.4 Twisted chiral multiplet for a holomorphic function
Given a holomorphic function W (v (j) ) of scalars v (j) in several twisted chiral multiplets, we get a new twisted chiral multiplet

C Renormalizing the FI parameter in flat space
Since renormalization is a UV phenomenon, one should be able to understand it in flat space. In this appendix we study the one-loop renormalization of the FI parameter in flat space.
We first note that the auxiliary field D couples to the chiral multiplet only through the −φDφ term in (A.9). For simplicity, let us turn off the central as well as the R-charges, and assume that σ and Λ are constant. For small D and in flat space we have Here φφ PV is the two-point function at the coincident points, regularized by the Pauli-Villars method: (C. 2) The integral converges absolutely because J J b J = 0. We then find Thus the effective FI coupling (the coefficient of D/2π in − log Z) is The dependence of r 0 on Λ must be such that this combination is finte as Λ → ∞. We fix the finite part by demanding that out effective FI coupling r eff coincides with that found in [3]: r eff = r(µ) + log(|σ|/µ). This gives the relation (1.11) between the renormalized and the bare FI parameters. Though our regularization method differs from that of [3], the difference is accounted for by the counterterms, yielding the equivalent renormalized theory.
Open Access. This article is distributed under the terms of the Creative Commons Attribution License (CC-BY 4.0), which permits any use, distribution and reproduction in any medium, provided the original author(s) and source are credited.