WorksPhysics and Astronomy
Can hint at a parity-odd sector of gravity?
Abstract
We study slow-roll inflation in the context of metric-affine gravity. We consider the most general projectively invariant metric-affine theory at most quadratic in the distortion and linear in curvature with no propagating torsion and non-metricity and couple it non-minimally to a scalar field. We focus on the class of -attractors, in which a single function fixes the whole scalar sector. Integrating out the non-dynamical distortion generates a correction to the kinetic term which is times a ratio of polynomials in . The new kinetic term can develop poles at finite field values, where the distortion field equations degenerate. We classify the poles and derive the conditions on the effective couplings for poles of order and , the maximal order allowed in this setting. We then proceed to study inflation. Second-order poles realize -attractors with fixed by the residue and are already available in the parity-even sector of the theory. Third-order poles instead require the parity-odd sector: whenever the parity-odd quadratic couplings vanish, the kinetic function splits into two independent terms with poles of order at most . Third-order pole-inflation, as we confirm numerically for , yields with a tunable amplitude and , a spectral index mildly larger than the -attractor prediction and well centered in the region preferred by the current combination of ACT, SPT and DESI data. Moreover, when inflation happens far away from the pole we also find a further universal attractor, an inverse-linear plateau, with and , testable with the forthcoming BICEP/Keck observations.
Introduction
Observations of the Cosmic Microwave Background (CMB) show that the universe is extremely homogeneous and flat [1]. In order to explain such fine-tuned properties the Standard Cosmological Model, a.k.a. the ΛCDM, is supplemented by a period of quasi-exponential accelerated expansion that drives the universe to the observed flatness and homogeneity [2, 3, 4]. This period goes under the name of inflation and has the appealing property of seeding the current observed large-scale structure of the universe. In order to sustain inflation a new particle, the inflaton, is usually introduced to drive the early accelerated expansion. In the most simple realization this particle is usually taken to be a scalar field, embedded in General Relativity (GR).
However, in this setting, minimal models fail to predict the CMB observables, in particular the tensor-to-scalar ratio, which is currently bounded at at 95% confidence level by the combined observations of Planck, BICEP/Keck, ACT and SPT [5, 6, 7, 8]. Several inflation models have been studied in different contexts trying to address such a suppressed tensor-to-scalar ratio [9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25]. In particular Starobinsky [4] and Higgs [26] inflation predict a value , well inside the current bounds and measurable within the next decade CMB experiments [27, 28]. On the other hand, recent measurements from DESI [29, 30], combined with the ACT and SPT dataset [7, 8, 31], have shifted the observed spectral index to at 68% confidence level, pushing the two most popular models over away from the central value1. New mechanisms for inflation then might be necessary to explain the updated observations.
One of the most recently studied frameworks in this context is Metric-Affine Gravity (MAG) [33, 34, 35]. In the metric-affine formalism the metric and the affine-connection are treated as a priori independent objects (unlike in GR where the connection is assumed to be the Levi-Civita one). MAG contains in general several more degrees of freedom [36], encoded in the two geometrical quantities that characterize a general connection: torsion and non-metricity . Moreover in MAG, the so-called hypermomentum tensor naturally arises as a consequence of taking the variation of the action with respect to the independent connection [37, 38], analogously to how one gets the energy-momentum tensor from the variation of the action with respect to the metric. The energy-momentum tensor and the hypermomentum are then to be regarded as the sources of MAG field equations [39, 40]. The enriched particle spectrum of MAG carries, in general, several ghost degrees of freedom [41, 42, 43, 44, 45, 46]. To avoid instabilities, one can then assume that the torsion and non-metricity are non-dynamical. In this way and only appear algebraically and can be eliminated from the field equations. When the inflaton field is non-minimally coupled to torsion and non-metricity, however, the hypermomentum is sourced directly by the scalar field gradient inducing a modified inflationary dynamics with respect to the standard formulation in General Relativity [47, 48], even though torsion and non-metricity are non-dynamical. Several cases have been studied in this context2 [60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74] and the case of inflation in general MAG has only been recently addressed in [75].
In this recent paper, single-field slow-roll inflation was studied in the general case of projectively invariant metric-affine theory containing all operators at most linear in the curvature and quadratic in the distortion, non-minimally coupled to a real scalar field . There, it has been shown that integrating out the non-dynamical distortion modifies the Jordan-frame scalar kinetic term by a factor . For -attractors, i.e. when all the non-minimal couplings are linear in a single positive monotonic function and the potential is , is times a ratio of polynomials in , and in the strong-coupling regime the model universally produces an -attractor [76, 77, 78, 79, 80, 81] with
independently of the choice of , where is the non-minimal coupling with the Ricci scalar and collects the whole effect of the metric-affine couplings. That analysis was restricted to the case of regular at every finite field value, while its pole structure was left unexplored. In this paper we fill that gap.
In order to do so, we adopt the same -attractor parametrization of the couplings as in [75], keeping generic. Our results can be summarized as follows: (i) the order of poles at finite field values is bounded by , a direct consequence of the fact that a scalar gradient excites exactly three independent distortion modes; (ii) second-order poles arise whenever the denominator of has a double zero, and are already available in the parity-even sector of the theory: they realize -attractors in which is a free parameter fixed by the residue at the pole; (iii) third-order poles require the denominator to be a perfect cube, and, our main result, they are impossible unless the theory contains parity-odd operators: if these are vanishing, the kinetic function splits into two additive pieces whose denominators are respectively quadratic and linear in , so that no pole can exceed second order; (iv) the pole-inflation predictions do not depend on the choice of , which only enters at subleading order. Finally, we scan the parameter space to visualize the predictions of second-order and third-order pole inflation in the plane. For the numerical analysis we use, as a test case, the minimal choice , i.e. a quartic Jordan-frame potential, while all the attractors we identify are derived for a generic .
The paper is organized as follows. In § 2 we briefly recall the model and the outcome of integrating out the distortion. In § 3 we present the Einstein-frame action, organize the pole structure of the modified kinetic term and show that pole inflation does not depend on the choice of . In § 4 we derive the conditions for second- and third-order poles and the corresponding inflationary predictions in the slow-roll approximation. In § 5 we show that the parity-even sector alone is not sufficient to generate third-order poles. We furthermore show that adding a scalar-Holst coupling restores this possibility. In § 7 we scan the parameter space to understand the general predictions of MAG in presence of a pole in the kinetic term. In § 8 we discuss the validity of the results and draw our conclusions. In appendix § 10 we define all the projectively invariant quantities that enter the action, while in appendix § 11 we collect all the required coefficient definitions which correspond to those found in [75].
The projective-invariant model
We work on a four-dimensional non-Riemannian manifold with signature and an independent affine connection, following the conventions of [75]. The affine connection is split as , with the Levi-Civita connection and the distortion tensor, equivalently encoded in the torsion and the non-metricity ,
and related to the distortion through
As in [75] we impose projective invariance to reduce the number of independent operators. The invariance under
with an arbitrary vector, selects eight out of eleven parity-even scalars , five out of six parity-odd scalars and three out of four linear terms in the distortion present in the general MAG action [75, 82, 83]. Thus, the most general projectively invariant action at most quadratic in the distortion and linear in the curvature, non-minimally coupled to a real scalar, reads
with manifestly projectively invariant quantities defined in the appendix § 10 ([75] for the details). The terms linear in the distortion arise from the Levi-Civita total derivatives after integration by parts, hence the notation [75]. This action contains eighteen free functions of (including the scalar potential ). The distortion appears only algebraically: its field equations are linear and sourced by alone. It can be proven that the general solution is given by [47]:
It was shown furthermore in [75] that projective-invariance reduces the independent modes to three: by defining , with and using (2.5),(7) we can rewrite the algebraic sector of the Lagrangian as:
with
where the six effective couplings are the combinations of listed in appendix § 11. Note for later use that the parity-odd couplings enter only through
Solving the linear system and substituting back yields:
In [75] the analysis was restricted to at every field value; here we are interested precisely in the case , where develops a pole. To render the analysis concrete we adopt the -attractor parametrization of [75], in which a single dimensionless function fixes all the couplings and the potential,
with a constant mass scale and a smooth, monotonically increasing and unbounded function, 3 (from now on we drop the explicit dependence of to simplify the notation). Every coupling is therefore linear in and the potential is quadratic in . The minimal choice , for which , , and with , corresponds to a truncation of the action up to operators of dimension , and will be used as a test case in the numerical analysis of § 7. Under (12) we have:
and , with and given in appendix § 11.
Einstein-frame kinetic function and pole-inflation
By performing a conformal transformation of the metric
we can rewrite the theory in the equivalent Einstein frame4:
with
By defining the canonical Einstein-frame scalar field as
we finally get the action in terms of the canonical scalar field:
We can now introduce the usual machinery needed to compute the slow-roll observables. We first define the slow-roll parameters:
where the denotes the derivative with respect to its argument. The slow-roll parameters can be directly related to the CMB observables, in particular the tensor-to-scalar ratio , the spectral index , the amplitude of the scalar perturbations and the number of e-folds . The latter does in general depend on the history of the evolution of the universe and hence, on the reheating temperature. In this paper we do not assume any specific reheating mechanism and consider the interval . The aforementioned CMB observables can be expressed in terms of the slow-roll parameters , as
where is the canonical field at horizon crossing, and is set by .5 In the -attractor class the potential is fixed by (12), , so that . In general a simple analytic expression of (3.7)-(23), in terms of cannot be found, unless (17) admits a closed form expression once integrated. However, this is not necessary: since is monotonic it can be used as field variable [75], with
and we can express (3.7)-(23) in terms of using the chain rule for derivatives:
where is the value of at horizon crossing and . For these reduce to the expressions in terms of for a quartic potential . Under the parametrization (12) the metric-affine modification to the kinetic term takes the form
with , and the dimensionless coefficients and collected in appendix B. For the minimal choice (29) reads
The denominator is a cubic polynomial in : this degree is not accidental but counts the three distortion modes reached by a scalar gradient, whose structure in (9), with entries linear in , cannot generate anything higher. A source exciting more modes, or couplings of higher degree in , could still give a correspondingly richer pole structure. Inserting (29) into (24),
as in [75]: the first two terms depend on alone, while the functional form of enters only through in the last term. For one has , constant.
Since is required for the conformal transformation (14), all finite-field poles of descend from real zeros of . The constant term never vanishes, so no pole can sit at : poles sit at , with a root of the cubic
Since , a root of multiplicity in at is a zero of multiplicity in at , so the pole order of at is
with the multiplicity of as a zero of the numerator . Because (32) is cubic,
and only if is a perfect cube and . Together with the second-order pole at studied in [75], which yields the MAG -attractor, this exhausts the singularity structure of the kinetic function.
Since the Einstein-frame potential is regular in , it is convenient to analyse the poles in the variable . Near a pole the kinetic function behaves as
while is regular at , with and . Expanding
and canonically normalizing, the standard pole-inflation results [80] apply: for the pole is at infinite distance in the canonical Einstein frame field, the potential develops a plateau, which at large -fold number predicts
The same expressions hold in any other field variable in which the potential is regular at the pole, e.g. : under one has and , so the combination , and hence , is invariant. In the variable , the pole term of (31) depends only on , i.e. on and on the couplings, while the last term is regular at . Pole-inflation predictions are therefore independent of the choice of .
The functional form of first appears at order in the expansion of around the pole, through , so that near the pole its effect is suppressed by .
In practice its leading effect comes from the last -folds, where the field is far from the pole and is no longer large, and amounts to an shift of the -fold number, i.e. , well within the uncertainty already associated with reheating6.
For the canonical field remains at finite distance, no plateau forms and at the kinetic term flips sign, implying instability; we do not discuss this case further. In the following we focus on the cases and . Notice that odd-order poles need one extra assumption: changes sign across , so the pole is a boundary of field space and inflation must take place on the side where .
Conditions for second- and third-order poles
Second-order poles
A second-order pole requires a double root of (32) at with . Writing the discriminant of the cubic, the condition on the couplings is
which is a single relation among the (codimension one in coupling space) and the inequality excludes the perfect-cube configuration of § 4.2. The double root sits at
which must be positive, as well as the residue , i.e. . Close to the pole we have that:
with , where the residue of a second-order pole is the same in any field variable, . The resulting model is an -attractor,
but of a different nature than the asymptotic one of [75]: there was fixed by the large-field ratio and the non-minimal coupling , whereas here is set by the residue of the pole at finite . In particular, a small (and hence small ) is achieved with a mild tuning of the residue rather than a large (as we also confirm numerically in § 7).
It is worth stressing that second-order poles do not require the parity-odd sector but do require the parity-even quadratic one. Setting the determinant factorizes, , and the parity-even factor
is quadratic in and supports a double zero under the simple conditions
and a non-vanishing numerator at . Notice that a positive requires . This cannot be achieved if the quadratic parity-even sector is turned off i.e. if , in which case . Hence second-order poles are not available in absence of a quadratic parity-even sector (in the distortion). Equation (43) is a one-parameter tuning of the parity-even quadratic couplings alone: -attractors are already available in parity-even projective-invariant metric-affine gravity.
Third-order poles
The maximal pole order requires the cubic (32) to be a perfect cube. Since the constant term is fixed at 3, this determines the whole polynomial in terms of alone:
The inequality places the pole at a physical field value; via appendix B, , so a negative net quadratic coupling of the distortion sector is what pushes the degenerate zero onto the positive real axis. The pole survives in provided the numerator does not vanish there,
otherwise the order degrades. Conditions (44) define a codimension-two surface in the space of effective couplings; we comment on the associated tuning in section 5.1.
Near the pole, (44) gives and hence
so that a healthy kinetic term approaching requires . The predictions follow from (37) with :
Two features should be noted. First, unlike the case the tensor-to-scalar ratio retains a dependence on the potential through : the amplitude of is a free parameter scaling as , intermediate between the of monomial models and the of -attractors. Second, the spectral index is shifted upward with respect to the -attractor value: at – one finds –, to be compared with – for . Given that current data combinations including DESI BAO pull the preferred towards larger values [8], third-order pole inflation lands naturally in the middle of the allowed region, making the two universality classes distinguishable with forthcoming CMB data.
Third-order poles and the parity-odd sector
We now come to the main structural result of this paper: the conditions (44) cannot be satisfied, within the parametrization (12) and for any choice of , with a surviving third-order pole unless the parity-odd quadratic sector is switched on. The precise statement is the following. If , i.e.
then the kinetic function has poles of order at most , for arbitrary values of all remaining couplings. In particular, setting is sufficient (although not necessary) to forbid the third-order pole.
When the matrix (9) becomes block diagonal and the solution (11) splits additively into
as it can be straightforwardly checked. Under the choice (12) the first denominator, , is quadratic in and can have at most a double zero, while the second, , is linear in and can only have a simple zero at most. Since a sum of rational functions cannot have a pole of higher order than its addends, we have .
Note that the conditions (44) can still be met with and they reduce to
i.e. the double zero of the parity-even factor is made to coincide with the zero of , and indeed develops a triple zero at . But when the two fractions in (49) are brought over the common denominator , the numerator inherits a simple zero at from whichever factor multiplies each term, and the pole order remains at most. Note also that the parity-odd linear coupling is of no help: it enters only the source , i.e. only the numerator of (11), and hence can never raise the order of the pole. We conclude that in order to have a third-order pole, a parity-odd term must appear in the quadratic sector.
From to
An exact third-order pole lives on a codimension two subspace of the parameter space. It might then seem an unlikely configuration realized only in presence of fine-tuning of the parameters. Indeed, one must always keep in mind that a generic one-parameter deformation of the couplings does not even encounter , since the double root already costs one condition, let alone . The conditions (38) and (44) however, are not disjoint: the perfect-cube locus lies on the boundary of the double-root locus, and one can pass continuously from one to the other. It is then possible, within the -attractor family (38), to have third-order points occur at isolated values of the remaining parameters, so a generic path along the surface can flow through a point.
Such a path is easy to construct explicitly. Keeping the double root fixed at and letting the third root vary,
This fixes as functions of , which in turn can be solved for three of the couplings. Without loss of generality, we choose , holding and the source fixed. The family is regular: vary smoothly with and reach finite critical values at , where (51) becomes the perfect cube (44).
Note that two poles are present for : a double pole at and a simple pole at . Their ordering is of fundamental importance, as we don’t want to spoil the kinetic term in the region where inflation happens, hence we need , i.e. . Writing
the kinetic function near takes the form
with given in (46), which interpolates between the two universality classes in the single variable :
with
The inflaton must approach from below, i.e. from the side on which the simple pole lies beyond : on the far side changes sign across and the region will inevitably carry a ghost. The third-order point is therefore reached continuously only from the -attractor side, which is what makes the flow well defined.
Figure 1 shows this flow for the family (51) for the minimal choice . The specific example in the plot is obtained by choosing (i.e. ), , for (lighter orange), (darker orange), (red), (brown) and setting , for simplicity. The observables were obtained by evaluating numerically the exact slow-roll equations with the rational of (30) for Jordan frame potential, without using any pole approximation. Panel (a) shows the trajectory in the plane as : it runs continuously from the second-order pole prediction to the third-order pole prediction, with and monotonically increasing. Panels (b) and (c) show that the entire flow is controlled by , with the two plateaus separated by a crossover. Panel (d) shows the value of fixed in terms of the other parameters through the measured amplitude of the scalar perturbation .

Figure 1. Continuous flow from a second- to a third-order pole along the family (51), for , obtained by choosing , , for (lighter orange), (darker orange), (red), (brown) and setting , for simplicity. The observables were obtained by evaluating numerically the exact slow-roll equations (20)–(23) with the rational of (16) for Jordan frame potential, without using any pole approximation. Panel (a) shows the trajectory in the plane as (with defined in (52)), with both and monotonically increasing. Panels (b) and (c) show that the entire flow is controlled by , with the two plateaus separated by a crossover. Panel (d) shows the value of fixed in terms of the other parameters through the measured amplitude of the scalar perturbation . Details in the text.
We see that, in general, -attractor models are favored by the SPT+ACT+Planck+BK data, hence the current CMB-only observations show a preference for either second-order poles or the second-order pole at studied in [75] (in the context of MAG inflation). These configurations can be already achieved by the model without including parity-odd terms in the Lagrangian. However, with the inclusion of DESI data, observations seem to prefer either second-order poles with a large 7 or general third-order poles. The latter can only take place on a codimension two subspace of the full parameters space, but can also be reached continuously within the -attractor family by tuning the parameters. In either case, under parametrization (12) and for any choice of , third-order poles are a prediction of projective-invariant MAG inflation only in presence of a parity-odd gravitational sector and the current CMB+DESI dataset might represent an indirect hint at the presence of such non-minimally coupled operators. We stress that the presence of such operators could only be confirmed by a direct measurement of parity-violation in the gravitational sector, a feature that cannot be confirmed by measurements of the spectral index alone.
A minimal example: the Holst coupling
The condition can be straightforwardly realized, for example, by a scalar-Holst non-minimal coupling , with the contraction of the Riemann tensor with the Levi-Civita tensor:
In the context of inflation, such a term has been extensively studied in several recent works [65, 68, 71, 84]. In our basis it corresponds to the choice , , [75], which gives
so a single Holst coupling switches on both parity-odd entries at once (for , ). Third-order poles are then available and inflation with the Holst invariant thus emerges as one of the minimal projective-invariant settings in which the universality class (47) can be realized (in presence of a non-vanishing parity-even sector).
Parameter space and slow-roll inflation
Parameter space and slow-roll inflation
The projectively invariant action (5) contains, under (12), 17 independent couplings, . Only 14 of them are reached by the scalar gradient sourcing the distortion, since , and drop out of and (see appendix § 11). For the numerical scan we fix , i.e. . We scan the parameter space by assuming and allow all the other couplings to be positive or negative. We sample the couplings in the range – (in absolute value) on a base-10 log scale, enforcing condition (38) for second-order and condition (44) for third-order poles, and compute for independent configurations by evaluating numerically (20)–(23). The parameter scan assumes for second-order poles. Moreover, roughly 60–80% of the sampled points (depending on the chosen subset) propagate a healthy kinetic term, while the remaining are ghosts8. In Figure 2 we only plot the well-behaved configurations. We show in Figure 2 the slow-roll predictions for the Jordan-frame potential under the parametrization (12) with , for second- and third-order pole inflation with different sectors of the action (5) switched on: a) (top-left) second-order pole, parity-even only (PE); b) (top-right) second-order pole, parity-even and linear distortion couplings (PE+DC); c) (center-left) second-order pole, parity-even and parity-odd (PE+PO); d) (center-right) second-order pole, all sectors (PE+PO+DC); e) (bottom-left) third-order pole, parity-even and parity-odd (PE+PO); f) (bottom-right) third-order pole, all sectors (PE+PO+DC). All observables are computed at (triangles) and at (bubbles); blue empty circles mark the inverse-linear plateau reached at large (§ 7.3, eq. (75) with ), blue empty squares quadratic inflation (larger symbols for ). The orange contours show the 95% and 68% confidence levels based on the latest combination of Planck, BICEP/Keck, ACT and SPT [8], while the cyan contours show the 95% and 68% confidence levels after including DESI. We stress that the points in Figure 2 do not assume the pole approximation.

Figure 2. Slow-roll predictions for the Jordan-frame potential , i.e. in the parametrization (12), for second- and third-order pole inflation with different sectors of the action (5) switched on: a) (top-left) second-order pole, parity-even only (PE); b) (top-right) second-order pole, parity-even and linear distortion couplings (PE+DC); c) (center-left) second-order pole, parity-even and parity-odd (PE+PO); d) (center-right) second-order pole, all sectors (PE+PO+DC); e) (bottom-left) third-order pole, parity-even and parity-odd (PE+PO); f) (bottom-right) third-order pole, all sectors (PE+PO+DC). All observables are computed at (triangles) and at (bubbles); blue empty circles mark the inverse-linear plateau reached at large (§ 7.3, eq. (75) with ), blue empty squares quadratic inflation (larger symbols for ). The orange contours show the 95% and 68% confidence levels based on the latest combination of Planck, BICEP/Keck, ACT and SPT [8], while the cyan contours show the 95% and 68% confidence levels after including DESI. Details in the text.
Besides the predictions (41) and (47), obtained when , the scan displays limits in which horizon exit occurs far away from the pole. They can all be understood in the same way, for a generic . For , the denominator of (29) becomes , so that , with a polynomial in . If a single power dominates then
with . In this limit we can integrate expressions (20)–(23) exactly (after neglecting the contribution coming from the end of inflation):
with
These limits do not depend on the choice of , which only affects their range of validity through the condition . For we have
which then implies
corresponding to quadratic inflation for and linear inflation for . For instead horizon exit takes place on the Jordan-frame plateau where
which yields
i.e. an inverse-power-law plateau approached at . Here the field rolls down from the plateau towards decreasing , as in -attractors. We remind that, since we performed a random sampling, these attractor behaviors are to be considered as leading order approximations in specific regimes, so the approximation is never exact for the plotted points. In the next section we look closely at these attractors and understand which sector is responsible for each regime.
Effect of parity-even and linear sectors
First of all, we remark, as already mentioned in § 4.1 that poles with can be reached only if the quadratic parity-even sector is on. From (43) and (44) we see that a positive can only be obtained if we have respectively and from the coefficient definitions of appendix § 11 we see that this cannot happen if (this is why we show no plot without the parity-even sector). More generally, for one finds , whose zeros at are simple: the parity-odd sector alone can at most generate the unhealthy poles. Adding the linear sector does not allow to reach a higher pole order either: as discussed in § 5, the parity-even sector alone cannot generate third-order poles, and the only enter the source , i.e. the numerator of .
The linear sector does, however, produce a new limit away from the pole: quadratic inflation. When the dominate, so does the lowest power of 9,
which must be positive, i.e. driven by or . This corresponds to the case , of (59) i.e.
for any , which predicts
in agreement with the results of [74, 75]. The quadratic limit is shown in Fig.2 with blue empty squares for (smaller square) and (larger square) and only appears when the linear terms are switched on, as expected.
Effect of the parity-odd sector
If we add the parity-odd sector to the parity-even one, both second- and third-order poles are available, and both reach the corresponding plateau predictions (41) and (47). Away from the pole a new limit appears: linear inflation. It corresponds to , in (59):
for any , which yields:
This limit cannot be reached without parity-odd operators (see Figure 2, top-right). For , eq. (49) writes schematically as
Far from the pole both denominators are constant, since , and grows at most linearly in . Therefore in the parity-even sector for , and the limit of linear inflation, for which we need , cannot be reached.
All sectors and the large- attractor
With all sectors on we still find the plateau predictions of second- and third-order poles, together with the quadratic and linear limits, for different configurations of the parameters. Moreover, in all cases, for large and (and in particular ) the predictions converge to a new attractor, represented by the blue empty circles, corresponding to the inverse-power-law plateau (64) with , i.e. an inverse-linear plateau .
To show how this attractor is reached consider the following. In this limit inflation takes place between the onset of the Jordan-frame plateau, , and the pole at . For large enough the window opens up at large , and the value of the field at horizon exit falls inside it. There we still have , while .
Let us first consider the parity-even sector alone with . Expanding at first order in one finds , independently of the pole condition (43), so that from (31), considering , we get
Notice that this requires in order to be well-defined everywhere. At large the first term dominates (for , , eq. (72) gives ). On the plateau, , one has
which gives
which is the inverse-linear plateau of (64), up to corrections of order . This predicts:
The couplings only enter through in . For one has and the attractor is completely universal and at leading order in we have:
which is testable and falsifiable within the next generation of CMB experiments [27, 28]. Note that this result is independent of the order of the pole and of the choice of , so both a second-order and third-order pole can reach this attractor. What is required is throughout the window , i.e. a suppression of the coefficients , which in our scan is realized by placing the pole at . The parity-odd sector adds to the expansion of a term , i.e. a contribution to growing linearly in on the plateau (as in for ). At horizon exit from our numerical scan we found it to be always subdominant with respect to , leaving essentially unchanged (as one can see from Figure 2 in the center-left and bottom-left panels), unless the parity-odd couplings exceed by an order of magnitude or more, in which case the predictions drift toward linear inflation. Conversely, if the pole sits close to the onset of the plateau, , the -folds are generated near the pole and one recovers the pole-inflation limit. This concludes the analysis of the parameter space.
Overall, we observe four distinct attractors: pole inflation at finite , linear inflation, quadratic inflation, and the inverse-linear plateau, all of which are independent of the choice of . Any other point in the plane is either obtained from the competition of these four attractors or from ad-hoc fine-tuned choice of the couplings.
Discussion and conclusions
We studied single-field slow-roll inflation in the context of metric-affine gravity. We considered the most general projectively invariant metric-affine gravitational sector up to linear order in curvature and quadratic order in the distortion, without propagating torsion and nonmetricity. We coupled it with a real scalar field, through seventeen independent functions , , , . After integrating out the distortion and performing a conformal transformation we are left with an equivalent Einstein frame metric theory of gravity plus a propagating non-canonical scalar field and a modified scalar potential, so that all the contribution from the metric-affine gravitational sector is now encoded in the modified scalar field dynamics. Its kinetic term is determined by ten independent functions of the real scalar field which are combinations of the original seventeen free functions in the action. We focused on the class of -attractors, in which every coupling is linear in a single monotonic function and the potential is . In this class the kinetic term is determined by and by seven effective parameters . In a recent work, the case for which has no poles was considered. There it was found that such a model always predicts an -attractor behavior at strong coupling, independently of [75]. Here we focused on the complementary case where admits poles at finite field values. We characterized these poles and found that, within our assumption the model can only support finite-field poles of order , since the denominator of is a cubic polynomial in . In particular we found that poles of order are available only if at least one of the parity even sector coefficients , while only becomes available if in addition to that the parity-odd couplings switch on , i.e. or . We also found that , (or more generally ) do not affect the scalar field dynamics. Moreover the linear distortion couplings have no effect on the pole order (but they still affect scalar field dynamics). We concluded that within the -attractor class, for any choice of , second-order poles are already available in the parity even sector and third-order poles are realized only if we also include the parity-odd one. Moreover, the pole-inflation predictions do not depend on : they are fixed by the residue of the pole and by the slope of the potential at the pole, in a combination which is invariant under field redefinitions, while only enters at subleading order through the canonical kinetic term of , effectively shifting the number of -folds by . For the minimal choice we found that inflation happens close to the pole in the limit , . In this limit the model predicts:
with and given by (41) and (46), and . Comparison with data shows that with second-order poles are preferred by the CMB-only data, while the CMB+BAO (i.e. including DESI data) prefers third-order poles. Since within the -attractor class (12) third-order poles are only available in presence of a parity-odd sector we suggest that, if a larger value is selected by future observations, it could be a hint for the presence of a parity-odd sector of gravity, though, we remark, not conclusive. We furthermore underline that a third-order pole requires exact tuning. As proven in section 4 a second-order pole costs one exact condition between the parameters (38) while a third-order one two (44). As shown in section 5.1 the second condition for the third-order pole can be relaxed and a significant shift towards third-order pole behavior can still be achieved within the -attractor family with tuning of the parameters, as explicitly shown in Figure 1 for a benchmark example. The amount of tuning necessary for realizing such a scenario has only been shown qualitatively in this work and we postpone a more quantitative analysis for future work, in case larger values of are selected by future observations.
In section 6 we extended our analysis beyond pole-inflation and showed the general predictions for inflation, in presence of a pole in the kinetic term . We did so by performing a random scan of the parameter space for . Beyond the pole-inflation attractor we shown, for a generic , that the model contains three more competing attractor behaviors, which are triggered by specific sectors of the gravitational action (see Figure 2). First of all, we found a general large attractor which appears whenever and corresponds to an inverse-linear plateau which predicts, at leading order in :
the attractor is available independently of the pole-order and hence already available in the parity-even sector alone. Such a high value of the tensor-to-scalar ratio is in the observation range of Simons Observatory [27] and BICEP/Keck [6] within the next few years and hence testable and falsifiable. The model contains moreover two more attractors: linear inflation, available only when the parity-odd sector is present and quadratic inflation, which is triggered by the presence of linear distortion couplings as already shown in [74, 75] in related contexts. All these attractors are independent of the choice of , which only affects their range of validity. Any other point in the plane generated by our random scan is either obtained as a result of the competition of these four attractors or as fine-tuned configurations of the parameter space. In general we can conclude the following, depending on future observations of the tensor-to-scalar ratio. If our model can be compatible with data only outside the pole-inflation approximation. Hence if a pole is present, horizon exit in this model must happen when the scalar field is already far from . In this regime, our model can predict specific testable values of the observables if and .
On the other hand if our model can be compatible with data only if horizon exit happens close to , which implies , . In this case the predictions of our model are specific and fall into the -attractor (41) behavior if second-order pole inflation is realized, while they predict (47) if third-order pole inflation is realized. In particular, in our model third-order poles are the only configurations landing in the region preferred by the combination of CMB and BAO data for : second-order poles give , while the quadratic and linear limits predict too large a tensor-to-scalar ratio. Third-order pole inflation thus can provide a mechanism to explain the observed shift of and, since in projectively invariant MAG such poles can only be obtained by switching on the parity-odd sector (within the -attractor class, for any ), recent observations might be hinting at the presence of such operators in the gravitational sector.
Acknowledgments
This work was supported by the Estonian Research Council via the grants, TARISTU24-TK10, TARISTU24-TK3, and the CoE program TK202 “Foundations of the Universe”.
Projective invariants
We first define the projective invariants rank-3 tensors:
from which we can construct two invariant parity-even vectors
together with a single parity-odd invariant vector:
with the Levi-Civita tensor and , , the trace vectors. From these we can construct by contraction all the possible non-zero manifestly projectively invariant scalars (at most quadratic in the distortion). The parity-even ones read
while the parity-odd terms are
We refer to [75] for the details.
Coefficients
For the reader’s convenience we collect from [75] the definitions used in the main text. The effective couplings entering (9) are
Under the -attractor parametrization (12) the effective couplings and the source take the form (13), , with the dimensionless coefficients
With
the numerator coefficients of (29) are
and the denominator coefficients are
so that .
References
- [1]Planck Collaboration, N. Aghanim et al., Planck 2018 results. VI. Cosmological parameters, Astron. Astrophys. 641 (2020) A6, [arXiv:1807.06209]. [Erratum: Astron.Astrophys. 652, C4 (2021)].
- [2]A. H. Guth, The Inflationary Universe: A Possible Solution to the Horizon and Flatness Problems, Phys. Rev. D 23 (1981) 347–356.
- [3]A. D. Linde, A New Inflationary Universe Scenario: A Possible Solution of the Horizon, Flatness, Homogeneity, Isotropy and Primordial Monopole Problems, Phys. Lett. B 108 (1982) 389–393.
- [4]A. A. Starobinsky, A New Type of Isotropic Cosmological Models Without Singularity, Phys. Lett. B 91 (1980) 99–102.
- [5]Planck Collaboration, Y. Akrami et al., Planck 2018 results. X. Constraints on inflation, Astron. Astrophys. 641 (2020) A10, [arXiv:1807.06211].
- [6]BICEP, Keck Collaboration, P. A. R. Ade et al., Improved Constraints on Primordial Gravitational Waves using Planck, WMAP, and BICEP/Keck Observations through the 2018 Observing Season, Phys. Rev. Lett. 127 (2021), no. 15 151301, [arXiv:2110.00483].
- [7]Atacama Cosmology Telescope Collaboration, E. Calabrese et al., The Atacama Cosmology Telescope: DR6 constraints on extended cosmological models, JCAP 11 (2025) 063, [arXiv:2503.14454].
- [8]L. Balkenhol et al., Inflation at the End of 2025: Constraints on r and nₛ Using the Latest CMB and BAO Data, arXiv:2512.10613.DOI
- [9]L. Boubekeur and D. H. Lyth, Hilltop inflation, JCAP 07 (2005) 010, [hep-ph/0502047].DOI
- [10]J. P. Conlon and F. Quevedo, Kähler moduli inflation, JHEP 01 (2006) 146, [hep-th/0509012].DOI
- [11]M. Cicoli, C. P. Burgess, and F. Quevedo, Fibre Inflation: Observable Gravity Waves from IIB String Compactifications, JCAP 03 (2009) 013, [arXiv:0808.0691].DOI
- [12]J. Ellis, D. V. Nanopoulos, and K. A. Olive, No-Scale Supergravity Realization of the Starobinsky Model of Inflation, Phys. Rev. Lett. 111 (2013) 111301, [arXiv:1305.1247]. [Erratum: Phys.Rev.Lett. 111, 129902 (2013)].DOI
- [13]G. F. Giudice and H. M. Lee, Starobinsky-like inflation from induced gravity, Phys. Lett. B 733 (2014) 58–62, [arXiv:1402.2129].DOI
- [14]F. Bauer and D. A. Demir, Inflation with Non-Minimal Coupling: Metric versus Palatini Formulations, Phys. Lett. B 665 (2008) 222–226, [arXiv:0803.2664].DOI
- [15]A. Racioppi, New universal attractor in nonminimally coupled gravity: Linear inflation, Phys. Rev. D 97 (2018), no. 12 123514, [arXiv:1801.08810].DOI
- [16]I. Antoniadis, A. Karam, A. Lykkas, and K. Tamvakis, Palatini inflation in models with an R² term, JCAP 11 (2018) 028, [arXiv:1810.10418].DOI
- [17]V.-M. Enckell, K. Enqvist, S. Rasanen, and L.-P. Wahlman, Inflation with R² term in the Palatini formalism, JCAP 02 (2019) 022, [arXiv:1810.05536].DOI
- [18]T. Tenkanen, Tracing the high energy theory of gravity: an introduction to Palatini inflation, Gen. Rel. Grav. 52 (2020), no. 4 33, [arXiv:2001.10135].DOI
- [19]C. Dioguardi, A. Racioppi, and E. Tomberg, Slow-roll inflation in Palatini F(R) gravity, JHEP 06 (2022) 106, [arXiv:2112.12149].DOI
- [20]C. Dioguardi, A. Racioppi, and E. Tomberg, Beyond (and back to) Palatini quadratic gravity and inflation, JCAP 03 (2024) 041, [arXiv:2212.11869].DOI
- [21]C. Dioguardi and A. Racioppi, Palatini F(R,X): A new framework for inflationary attractors, Phys. Dark Univ. 45 (2024) 101509, [arXiv:2307.02963].DOI
- [22]N. Bostan, R. H. Dejrah, C. Dioguardi, and A. Racioppi, Natural inflation in Palatini F(R), JCAP 03 (2026) 038, [arXiv:2511.20557].DOI
- [23]C. Dioguardi and A. Karam, Palatini linear attractors are back in action, Phys. Rev. D 111 (2025), no. 12 123521, [arXiv:2504.12937].arxiv.org/abs/2504.12937
- [24]C. Dioguardi, A. J. Iovino, and A. Racioppi, Fractional attractors in light of the latest ACT observations, Phys. Lett. B 868 (2025) 139664, [arXiv:2504.02809].DOI
- [25]C. Dioguardi, F. Gianesello, and A. Racioppi, Palatini inflationary (non-)attractors: Universal mapping of observables, arXiv:2609.20622.arxiv.org/abs/2609.20622
- [26]F. L. Bezrukov and M. Shaposhnikov, The Standard Model Higgs boson as the inflaton, Phys. Lett. B 659 (2008) 703–706, [arXiv:0710.3755].DOI
- [27]Simons Observatory Collaboration, P. Ade et al., The Simons Observatory: Science goals and forecasts, JCAP 02 (2019) 056, [arXiv:1808.07445].
- [28]LiteBIRD Collaboration, M. Hazumi et al., LiteBIRD: JAXA’s new strategic L-class mission for all-sky surveys of cosmic microwave background polarization, Proc. SPIE Int. Soc. Opt. Eng. 11443 (2020) 114432F, [arXiv:2101.12449].
- [29]DESI Collaboration, A. G. Adame et al., DESI 2024 VI: Cosmological Constraints from the Measurements of Baryon Acoustic Oscillations, arXiv:2404.03002.
- [30]DESI Collaboration, M. Abdul Karim et al., DESI DR2 results. II. Measurements of baryon acoustic oscillations and cosmological constraints, Phys. Rev. D 112 (2025), no. 8 083515, [arXiv:2503.14738].
- [31]SPT-3G Collaboration, E. Camphuis et al., SPT-3G D1: CMB temperature and polarization power spectra and cosmology from 2019 and 2020 observations of the SPT-3G main field, Phys. Rev. D 113 (2026), no. 8 083504, [arXiv:2506.20707].
- [32]E. G. M. Ferreira, E. McDonough, L. Balkenhol, R. Kallosh, L. Knox, and A. Linde, BAO-CMB tension and implications for inflation, Phys. Rev. D 113 (2026), no. 4 043524, [arXiv:2507.12459].arxiv.org/abs/2507.12459
- [33]F. W. Hehl, P. Von Der Heyde, G. D. Kerlick, and J. M. Nester, General Relativity with Spin and Torsion: Foundations and Prospects, Rev. Mod. Phys. 48 (1976) 393–416.DOI
- [34]F. W. Hehl, G. D. Kerlick, and P. Von Der Heyde, On a New Metric Affine Theory of Gravitation, Phys. Lett. B 63 (1976) 446–448.DOI
- [35]F. W. Hehl, J. D. McCrea, E. W. Mielke, and Y. Ne’eman, Metric affine gauge theory of gravity: Field equations, Noether identities, world spinors, and breaking of dilation invariance, Phys. Rept. 258 (1995) 1–171, [gr-qc/9402012].DOI
- [36]A. Baldazzi, O. Melichev, and R. Percacci, Metric-Affine Gravity as an effective field theory, Annals Phys. 438 (2022) 168757, [arXiv:2112.10193].DOI
- [37]F. W. Hehl, G. D. Kerlick, and P. Von Der Heyde, On Hypermomentum in General Relativity. 2. The Geometry of Space-Time, Z. Naturforsch. A 31 (1976) 524–527.DOI
- [38]F. W. Hehl, E. A. Lord, and Y. Ne’eman, Hypermomentum in Hadron Dynamics and in Gravitation, Phys. Rev. D 17 (1978) 428–433.DOI
- [39]J. Beltrán Jiménez, L. Heisenberg, and T. Koivisto, The coupling of matter and spacetime geometry, Class. Quant. Grav. 37 (2020), no. 19 195013, [arXiv:2004.04606].arxiv.org/abs/2004.04606
- [40]C. Rigouzzo and S. Zell, Coupling metric-affine gravity to the standard model and dark matter fermions, Phys. Rev. D 108 (2023), no. 12 124067, [arXiv:2306.13134].DOI
- [41]J. Beltrán Jiménez and A. Delhom, Ghosts in metric-affine higher order curvature gravity, Eur. Phys. J. C 79 (2019), no. 8 656, [arXiv:1901.08988].DOI
- [42]R. Percacci and E. Sezgin, New class of ghost- and tachyon-free metric affine gravities, Phys. Rev. D 101 (2020), no. 8 084040, [arXiv:1912.01023]. [Erratum: Phys.Rev.D 111, 109902 (2025)].DOI
- [43]Y. Mikura, V. Naso, and R. Percacci, Some simple theories of gravity with propagating torsion, Phys. Rev. D 109 (2024), no. 10 104071, [arXiv:2312.10249].DOI
- [44]Y. Mikura and R. Percacci, Some simple theories of gravity with propagating nonmetricity, Eur. Phys. J. C 85 (2025), no. 4 377, [arXiv:2401.10097].DOI
- [45]W. Barker, C. Marzo, and A. Santoni, Can metric-affine gravity be saved?, Phys. Rev. D 112 (2025), no. 4 044032, [arXiv:2505.23894].arxiv.org/abs/2505.23894
- [46]R. Percacci and E. Sezgin, Massive spin 3 and Metric-Affine Gravity, JHEP 01 (2026) 042, [arXiv:2508.14211].arxiv.org/abs/2508.14211
- [47]C. Rigouzzo and S. Zell, Coupling metric-affine gravity to a Higgs-like scalar field, Phys. Rev. D 106 (2022), no. 2 024015, [arXiv:2204.03003].DOI
- [48]I. Andrei, D. Iosifidis, L. Järv, and M. Saal, Scalar field with nonminimal couplings to metric-affine geometry, arXiv:2609.07856.arxiv.org/abs/2609.07856
- [49]A. Salvio, Inflating and reheating the Universe with an independent affine connection, Phys. Rev. D 106 (2022), no. 10 103510, [arXiv:2207.08830].DOI
- [50]I. D. Gialamas and K. Tamvakis, Inflation in metric-affine quadratic gravity, JCAP 03 (2023) 042, [arXiv:2212.09896].DOI
- [51]A. Di Marco, E. Orazi, and G. Pradisi, Einstein–Cartan pseudoscalaron inflation, Eur. Phys. J. C 84 (2024), no. 2 146, [arXiv:2309.11345].DOI
- [52]I. D. Gialamas and K. Tamvakis, Inflation in Weyl-invariant Einstein-Cartan gravity, Phys. Rev. D 111 (2025), no. 4 044007, [arXiv:2410.16364].DOI
- [53]M. He, M. Hong, and K. Mukaida, Starobinsky inflation and beyond in Einstein-Cartan gravity, JCAP 05 (2024) 107, [arXiv:2402.05358].DOI
- [54]M. He, M. Hong, and K. Mukaida, Increase of ns in regularized pole inflation & Einstein-Cartan gravity, JCAP 09 (2025) 080, [arXiv:2504.16069].DOI
- [55]G. K. Karananas, Geometrical origin of inflation in Weyl-invariant Einstein-Cartan gravity, Phys. Lett. B 862 (2025) 139343, [arXiv:2501.16416].arxiv.org/abs/2501.16416
- [56]A. Salvio, Independent connection in action during inflation, Phys. Rev. D 112 (2025), no. 6 L061301, [arXiv:2504.10488].DOI
- [57]D. Iosifidis and S. Karamitsos, Pole inflation from extended metric-affine gravity, arXiv:2510.20585.arxiv.org/abs/2510.20585
- [58]T. Katsoulas and K. Tamvakis, Inflationary assessment of F(R,R') Einstein-Cartan models, Phys. Rev. D 113 (2026), no. 12 124010, [arXiv:2512.02847].DOI
- [59]C. Dioguardi, The metric-affine R̂2 scalaraon: parameter-free inflation, arXiv:2610.23522.
- [60]K. Shimada, K. Aoki, and K.-i. Maeda, Metric-affine Gravity and Inflation, Phys. Rev. D 99 (2019), no. 10 104020, [arXiv:1812.03420].DOI
- [61]M. Shaposhnikov, A. Shkerin, I. Timiryasov, and S. Zell, Higgs inflation in Einstein-Cartan gravity, JCAP 02 (2021) 008, [arXiv:2007.14978]. [Erratum: JCAP 10, E01 (2021)].DOI
- [62]M. Shaposhnikov, A. Shkerin, I. Timiryasov, and S. Zell, Einstein-Cartan gravity, matter, and scale-invariant generalization , JHEP 10 (2020) 177, [arXiv:2007.16158].DOI
- [63]Y. Mikura, Y. Tada, and S. Yokoyama, Conformal inflation in the metric-affine geometry, EPL 132 (2020), no. 3 39001, [arXiv:2008.00628].DOI
- [64]Y. Mikura, Y. Tada, and S. Yokoyama, Minimal k-inflation in light of the conformal metric-affine geometry, Phys. Rev. D 103 (2021), no. 10 L101303, [arXiv:2103.13045].DOI
- [65]A. Racioppi and A. Salvio, Natural metric-affine inflation, JCAP 06 (2024) 033, [arXiv:2403.18004].DOI
- [66]I. D. Gialamas, T. Katsoulas, and K. Tamvakis, Inflation and reheating in quadratic metric-affine gravity with derivative couplings, JCAP 06 (2024) 005, [arXiv:2403.08530].DOI
- [67]S. Capozziello and M. Shokri, Comparing inflationary models in extended Metric-Affine theories of gravity, Phys. Dark Univ. 46 (2024) 101698, [arXiv:2408.17415].arxiv.org/abs/2408.17415
- [68]A. Racioppi, ξ-attractors in metric-affine gravity, JCAP 04 (2025) 084, [arXiv:2411.08031]. [Erratum: JCAP 08, E01 (2025)].DOI
- [69]I. D. Gialamas and A. Racioppi, Symmetry-breaking inflation in non-minimal metric-affine gravity, JCAP 05 (2025) 072, [arXiv:2412.17738].arxiv.org/abs/2412.17738
- [70]A. Racioppi, Quasi-pole inflation in metric-affine gravity, Gen. Rel. Grav. 58 (2026), no. 4 44, [arXiv:2512.16815].DOI
- [71]K. Dimopoulos, C. Dioguardi, I. D. Gialamas, and A. Racioppi, Quasi-pole quintessential inflation in metric-affine gravity, arXiv:2603.15776.DOI
- [72]D. Kraiko and A. Racioppi, Natural Metric-Affine Inflation: Reloaded, arXiv:2605.23827.DOI
- [73]A. Hassan, Inflation with nondynamic distortion to leading order in slow roll, arXiv:2607.03566.arxiv.org/abs/2607.03566
- [74]I. Andrei, C. Dioguardi, D. Iosifidis, L. Järv, A. Racioppi, and M. Saal, Inflation with Nieh-Yan-like terms in metric-affine gravity, arXiv:2608.07386.arxiv.org/abs/2608.07386
- [75]I. Andrei and C. Dioguardi, Inflationary attractors in metric-affine gravity, arXiv:2611.06750.arxiv.org/abs/2611.06750
- [76]R. Kallosh and A. Linde, Universality Class in Conformal Inflation, JCAP 07 (2013) 002, [arXiv:1306.5220].DOI
- [77]S. Ferrara, R. Kallosh, A. Linde, and M. Porrati, Minimal Supergravity Models of Inflation, Phys. Rev. D 88 (2013), no. 8 085038, [arXiv:1307.7696].DOI
- [78]R. Kallosh, A. Linde, and D. Roest, Superconformal Inflationary α-Attractors, JHEP 11 (2013) 198, [arXiv:1311.0472].DOI
- [79]R. Kallosh, A. Linde, and D. Roest, Universal Attractor for Inflation at Strong Coupling, Phys. Rev. Lett. 112 (2014), no. 1 011303, [arXiv:1310.3950].DOI
- [80]M. Galante, R. Kallosh, A. Linde, and D. Roest, Unity of Cosmological Inflation Attractors, Phys. Rev. Lett. 114 (2015), no. 14 141302, [arXiv:1412.3797].pubmed.ncbi.nlm.nih.gov/25910108
- [81]J. J. M. Carrasco, R. Kallosh, A. Linde, and D. Roest, Hyperbolic geometry of cosmological attractors, Phys. Rev. D 92 (2015), no. 4 041301, [arXiv:1504.05557].DOI
- [82]D. Iosifidis, The full quadratic metric-affine gravity (including parity odd terms): exact solutions for the affine-connection, Class. Quant. Grav. 39 (2022), no. 9 095002, [arXiv:2112.09154].DOI
- [83]D. Iosifidis, Quadratic metric-affine gravity: solving for the affine-connection, Eur. Phys. J. C 82 (2022), no. 7 577, [arXiv:2109.13293].DOI
- [84]M. Långvik, J.-M. Ojaperä, S. Raatikainen, and S. Rasanen, Higgs inflation with the Holst and the Nieh-Yan term, Phys. Rev. D 103 (2021), no. 8 083514, [arXiv:2007.12595].