Quantum-Corrected Higgs Inflation

• Quantum-corrected Higgs inflation is a framework where the Standard Model Higgs, non-minimally coupled to gravity, drives slow-roll inflation via an RG-improved effective action with an exponentially flat plateau.
• Quantum corrections regulate key parameters like λ and ξ and induce curvature-squared operators, which modify the inflationary potential and can generate an extra scalaron degree of freedom.
• Different formulations—including metric, Palatini, and Einstein–Cartan—demonstrate how quantum effects adjust EFT cutoffs and impact observable predictions such as the spectral index and tensor-to-scalar ratio.

Quantum-corrected Higgs inflation is the class of inflationary scenarios in which the Higgs sector, usually identified with the Standard Model Higgs at large field values, is coupled non-minimally to curvature and analyzed with an RG-improved effective action rather than a purely tree-level potential. In its standard metric form, the Jordan-frame scalar-gravity sector is

$$S=\int d^4x\,\sqrt{-g}\left[\left(\frac{M_P^2}{2}+\xi H^\dagger H\right)R-|D_\mu H|^2-\lambda\left(H^\dagger H-\frac{v^2}{2}\right)^2\right],$$

and after a Weyl transformation the Einstein-frame potential approaches a plateau of Starobinsky type. Quantum corrections then renormalize $\lambda$ and $\xi$, generate curvature-squared operators, modify the effective cutoff and unitarity structure, and in generic large-$\xi$ settings can induce an additional scalaron degree of freedom so that Higgs inflation interpolates toward $R^2$ inflation (Salvio et al., 2015, Ghilencea, 2018).

1. Classical non-minimally coupled Higgs sector

In unitary gauge, $H\to (0,h)/\sqrt{2}$, so that $H^\dagger H=h^2/2$ and $|D_\mu H|^2\to (\partial h)^2/2$. The non-minimal Higgs-curvature coupling is removed by the conformal rescaling

$$g_{\mu\nu}\to g_{\mu\nu}/\Omega^2,\qquad \Omega^2(h)=1+\frac{\xi h^2}{M_P^2},$$

after which the Einstein-frame scalar is canonically normalized through

$$\frac{d\chi}{dh}\simeq \sqrt{\frac{1+\xi(1+6\xi)h^2/M_P^2}{\left(1+\xi h^2/M_P^2\right)^2}}.$$

In the large-field, large-$\lambda$0 regime, $\lambda$1, the potential becomes

$\lambda$2

which is exponentially flat for $\lambda$3 and gives slow-roll inflation (Salvio et al., 2015).

The classical slow-roll predictions on this plateau are

$\lambda$4

with $\lambda$5–$\lambda$6. Matching the scalar amplitude gives, for example,

$\lambda$7

with nearby values $\lambda$8 for $\lambda$9 and $\xi$0 for $\xi$1. For $\xi$2 at inflationary scales, this places $\xi$3 in the $\xi$4 range (Salvio et al., 2015). A closely related EFT analysis gives the amplitude relation $\xi$5, emphasizing the same large-$\xi$6 scaling (Bezrukov et al., 2010).

2. One-loop renormalization and RG improvement

The quantum-corrected description is built from the one-loop effective action including all operators up to dimension 4,

$\xi$7

with

$\xi$8

and

$\xi$9

In dimensional regularization and $\xi$0, the reduced Planck mass does not run at one loop in the regime where quantum gravity effects are neglected, while the dominant running of the Higgs potential is governed by the Standard Model couplings, especially $\xi$1, $\xi$2, $\xi$3, and $\xi$4 (Salvio et al., 2015).

The non-minimal coupling satisfies

$\xi$5

and the curvature-squared couplings satisfy

$\xi$6

with

$\xi$7

The $\xi$8 enhancement is the central large-$\xi$9 quantum effect: it makes curvature-squared operators unavoidable unless their boundary values are tuned (Salvio et al., 2015).

A complementary RG-improved formulation parameterizes the inflationary logarithmic correction by the anomalous scaling

$R^2$0

and writes the large-field effective action as

$R^2$1

In the inflationary regime, Higgs-graviton kinetic mixing is suppressed by

$R^2$2

so that the large-field Einstein-frame potential takes the RG-improved form

$R^2$3

This representation makes explicit that quantum effects mainly alter the plateau slope through logarithms, while the plateau height remains controlled by $R^2$4 (Barvinsky et al., 2022).

3. Radiatively induced scalaron and the Starobinsky-Higgs picture

The most consequential quantum effect at large $R^2$5 is the radiative generation of an $R^2$6 term. In the perturbative analysis of large-$R^2$7 Higgs inflation, naturalness implies

$R^2$8

up to $R^2$9 logarithmic factors. The corresponding effective action,

$H\to (0,h)/\sqrt{2}$0

contains an extra propagating scalar degree of freedom, the scalaron. After introducing an auxiliary field and transforming to the Einstein frame, the system is equivalent to a two-field model with Higgs and scalaron variables, and in the large-$H\to (0,h)/\sqrt{2}$1 limit inflation proceeds along the scalaron direction, reproducing Starobinsky inflation with

$H\to (0,h)/\sqrt{2}$2

In this regime the attractor predictions return to $H\to (0,h)/\sqrt{2}$3 and $H\to (0,h)/\sqrt{2}$4, but the inflaton is effectively no longer the Higgs alone (Salvio et al., 2015).

This mechanism can be expressed directly as an RG effect in the coefficient $H\to (0,h)/\sqrt{2}$5 of $H\to (0,h)/\sqrt{2}$6. For $H\to (0,h)/\sqrt{2}$7 real scalars with non-minimal coupling $H\to (0,h)/\sqrt{2}$8,

$H\to (0,h)/\sqrt{2}$9

Taking $H^\dagger H=h^2/2$0 for the Higgs doublet and evolving to $H^\dagger H=h^2/2$1, a large coupling $H^\dagger H=h^2/2$2 yields

$H^\dagger H=h^2/2$3

the value required by CMB normalization in Starobinsky inflation. In this “Higgs Starobinsky” mechanism, inflation is driven by the scalaron rather than large Higgs excursions, so Standard Model metastability does not directly obstruct inflationary dynamics (Calmet et al., 2016).

At two loops, the same unification appears in a more systematic form. Even if $H^\dagger H=h^2/2$4 at tree level, matter loops generate $H^\dagger H=h^2/2$5-dependent curvature-squared couplings,

$H^\dagger H=h^2/2$6

so the Einstein-frame potential

$H^\dagger H=h^2/2$7

depends nontrivially on both Higgs and scalaron. In this formulation, multi-field inflation is a quantum consequence rather than a model-building input. The one- and two-loop corrections shift $H^\dagger H=h^2/2$8 by about $H^\dagger H=h^2/2$9–$|D_\mu H|^2\to (\partial h)^2/2$0 and increase $|D_\mu H|^2\to (\partial h)^2/2$1 by a few to ten percent for representative $|D_\mu H|^2\to (\partial h)^2/2$2, while keeping the system close to the Starobinsky attractor (Ghilencea, 2018).

4. Frames, cutoff scales, and covariant renormalization

Quantum-corrected Higgs inflation is unusually sensitive to field redefinitions because the distinction between Jordan and Einstein frames is tied to a nontrivial field-space metric. A covariant treatment based on the Vilkovisky–DeWitt effective action shows that the one-loop divergences differ substantially from non-covariant background-field calculations, and yields frame-independent consistency bounds on the EFT cutoff: $|D_\mu H|^2\to (\partial h)^2/2$3 in the small-field regime and $|D_\mu H|^2\to (\partial h)^2/2$4 in the large-field regime (Moss, 2014). In parallel, a direct computation of the one-loop effective potential for the real scalar case finds no disagreement between Jordan and Einstein frames once all dimensionful scales are transformed consistently; constant cutoffs in either frame give the same effective potential provided masses and renormalization points are mapped by the Weyl factor (George et al., 2013).

The background dependence of the cutoff is central. In the usual metric formulation, the vacuum cutoff is

$|D_\mu H|^2\to (\partial h)^2/2$5

while on the inflationary background it rises to

$|D_\mu H|^2\to (\partial h)^2/2$6

The Hubble scale during inflation,

$|D_\mu H|^2\to (\partial h)^2/2$7

is therefore well below $|D_\mu H|^2\to (\partial h)^2/2$8 for perturbative $|D_\mu H|^2\to (\partial h)^2/2$9, so the inflationary background itself is under EFT control (Salvio et al., 2015). A related Jordan-frame analysis writes the background-dependent cutoff as

$$g_{\mu\nu}\to g_{\mu\nu}/\Omega^2,\qquad \Omega^2(h)=1+\frac{\xi h^2}{M_P^2},$$0

which interpolates between $$g_{\mu\nu}\to g_{\mu\nu}/\Omega^2,\qquad \Omega^2(h)=1+\frac{\xi h^2}{M_P^2},$$1, $$g_{\mu\nu}\to g_{\mu\nu}/\Omega^2,\qquad \Omega^2(h)=1+\frac{\xi h^2}{M_P^2},$$2, and $$g_{\mu\nu}\to g_{\mu\nu}/\Omega^2,\qquad \Omega^2(h)=1+\frac{\xi h^2}{M_P^2},$$3 in the small-, intermediate-, and large-field regimes, respectively (Bezrukov et al., 2010).

Functional RG calculations reinforce the same picture. When the coarse-graining scale is chosen consistently with inflation, $$g_{\mu\nu}\to g_{\mu\nu}/\Omega^2,\qquad \Omega^2(h)=1+\frac{\xi h^2}{M_P^2},$$4, the heat-kernel expansion requires

$$g_{\mu\nu}\to g_{\mu\nu}/\Omega^2,\qquad \Omega^2(h)=1+\frac{\xi h^2}{M_P^2},$$5

and the large-field threshold factor

$$g_{\mu\nu}\to g_{\mu\nu}/\Omega^2,\qquad \Omega^2(h)=1+\frac{\xi h^2}{M_P^2},$$6

suppresses graviton and Higgs contributions to the beta functions. In that regime $$g_{\mu\nu}\to g_{\mu\nu}/\Omega^2,\qquad \Omega^2(h)=1+\frac{\xi h^2}{M_P^2},$$7, the running of $$g_{\mu\nu}\to g_{\mu\nu}/\Omega^2,\qquad \Omega^2(h)=1+\frac{\xi h^2}{M_P^2},$$8 is weak, and the classical plateau predictions are stable against leading gravitational corrections (Saltas, 2015). A separate Standard Model computation in the Einstein frame reaches a compatible EFT conclusion: the theory is renormalizable in the effective-field-theory sense in the small-, mid-, and large-field regimes, although the large-field beta functions depend subtly on the treatment of Goldstone bosons (George et al., 2015).

5. Alternative formulations and UV extensions

Quantum-corrected Higgs inflation is not a single model but a family of formulations that differ in how loop effects, extra operators, and UV completion are handled. The most widely discussed cases are summarized below.

Formulation	Characteristic quantum structure	Representative consequence
Metric FRG	$$g_{\mu\nu}\to g_{\mu\nu}/\Omega^2,\qquad \Omega^2(h)=1+\frac{\xi h^2}{M_P^2},$$9, $$\frac{d\chi}{dh}\simeq \sqrt{\frac{1+\xi(1+6\xi)h^2/M_P^2}{\left(1+\xi h^2/M_P^2\right)^2}}.$$0 suppresses graviton and Higgs loops	$$\frac{d\chi}{dh}\simeq \sqrt{\frac{1+\xi(1+6\xi)h^2/M_P^2}{\left(1+\xi h^2/M_P^2\right)^2}}.$$1; plateau predictions remain close to classical values (Saltas, 2015)
Palatini	$$\frac{d\chi}{dh}\simeq \sqrt{\frac{1+\xi(1+6\xi)h^2/M_P^2}{\left(1+\xi h^2/M_P^2\right)^2}}.$$2, with RG scale below cutoff	$$\frac{d\chi}{dh}\simeq \sqrt{\frac{1+\xi(1+6\xi)h^2/M_P^2}{\left(1+\xi h^2/M_P^2\right)^2}}.$$3, $$\frac{d\chi}{dh}\simeq \sqrt{\frac{1+\xi(1+6\xi)h^2/M_P^2}{\left(1+\xi h^2/M_P^2\right)^2}}.$$4, and $$\frac{d\chi}{dh}\simeq \sqrt{\frac{1+\xi(1+6\xi)h^2/M_P^2}{\left(1+\xi h^2/M_P^2\right)^2}}.$$5 (Shaposhnikov et al., 2020)
Einstein–Cartan	Generalized $$\frac{d\chi}{dh}\simeq \sqrt{\frac{1+\xi(1+6\xi)h^2/M_P^2}{\left(1+\xi h^2/M_P^2\right)^2}}.$$6 counterterm induces scalaron with $$\frac{d\chi}{dh}\simeq \sqrt{\frac{1+\xi(1+6\xi)h^2/M_P^2}{\left(1+\xi h^2/M_P^2\right)^2}}.$$7	UV extension works except close to the Palatini limit (He et al., 2023)
Non-local self-healing	Higgs kinetic operator $$\frac{d\chi}{dh}\simeq \sqrt{\frac{1+\xi(1+6\xi)h^2/M_P^2}{\left(1+\xi h^2/M_P^2\right)^2}}.$$8	Loops become finite and hard amplitudes are exponentially damped without new massive degrees of freedom (Koshelev et al., 2020)

In the Palatini formulation, where the connection is independent of the metric, the Einstein-frame kinetic factor is $$\frac{d\chi}{dh}\simeq \sqrt{\frac{1+\xi(1+6\xi)h^2/M_P^2}{\left(1+\xi h^2/M_P^2\right)^2}}.$$9 rather than $\lambda$00. The higher cutoff allows a more direct link between collider-scale and inflationary parameters. The resulting tensor amplitude is exceedingly small,

$\lambda$01

and the paper’s numerical window $\lambda$02 implies $\lambda$03 between $\lambda$04 and $\lambda$05 (Shaposhnikov et al., 2020).

In Einstein–Cartan gravity, the additional coupling to the Nieh–Yan density introduces an interpolation parameter

$\lambda$06

with $\lambda$07 corresponding to the Palatini limit and $\lambda$08 to the metric limit. The generalized counterterm

$\lambda$09

then induces a scalaron for any $\lambda$10, raises the cutoff to

$\lambda$11

and gradually decouples as the Palatini regime is approached (He et al., 2023).

A different strategy is non-local self-healing. There the Higgs radial mode is given an analytic infinite-derivative kinetic term,

$\lambda$12

so that the Euclidean propagator becomes

$\lambda$13

This exponentially suppresses loop integrals and softens hard scattering, while preserving the large-field Higgs-inflation plateau provided $\lambda$14 and $\lambda$15 (Koshelev et al., 2020).

6. Initial conditions and quantum cosmology

At the purely classical level, Higgs inflation is an attractor system. In the Einstein frame, numerical evolution in a flat FRW background shows that even starting with kinetic energy density as large as $\lambda$16, the system relaxes to slow roll on the plateau provided the initial Higgs field is sufficiently large, with $\lambda$17 for $\lambda$18. Under the assumption of initial homogeneity, no worrisome classical fine-tuning is required (Salvio et al., 2015).

The quantum problem is different. Once the RG flow of $\lambda$19, $\lambda$20, and $\lambda$21 is included, maintaining pure Higgs inflation requires suppressing the scalaron sector despite the large $\lambda$22 contribution. In that sense the main tuning is not classical initial data but the initial values of the running couplings, particularly $\lambda$23, at the scale where RG evolution begins (Salvio et al., 2015). This suggests a sharp distinction between classical attractor robustness and quantum selection of the effective theory.

Several papers recast the initial-condition problem in quantum-cosmological terms. In the tunneling-state approach, the initial distribution for the inflaton field is

$\lambda$24

and the RG-improved Higgs potential develops a sharp probability peak at

$\lambda$25

with relative width $\lambda$26. In this picture, the inflationary background is probabilistically selected rather than imposed by hand (Barvinsky, 2010). A different quantum-cosmological construction replaces the wavefunction by a microcanonical density matrix,

$\lambda$27

whose periodic Euclidean path integral is dominated by garland-type instantons. In that formulation the upper bound on the effective cosmological constant is

$\lambda$28

and the saddle-point geometry selects hill-top initial data for inflation (Barvinsky et al., 2022).

7. Observational predictions and current status

The baseline quantum-corrected metric scenario remains close to the Starobinsky/Higgs attractor. For $\lambda$29–$\lambda$30, both classical Higgs inflation and scalaron-dominated limits give

$\lambda$31

comfortably within the Planck 2015 bound $\lambda$32 quoted in the literature (Salvio et al., 2015). In RG-improved analyses based on the anomalous scaling parameter $\lambda$33, one writes

$\lambda$34

and obtains

$\lambda$35

For $\lambda$36, one analysis finds $\lambda$37, $\lambda$38 during inflation, and a Higgs-mass window $\lambda$39 consistent with the CMB constraints used there (Barvinsky et al., 2022).

A recent one-loop reanalysis aimed at the ACT+DESI-preferred higher scalar tilt uses the same $\lambda$40-parametrization and, for $\lambda$41 and $\lambda$42, finds

$\lambda$43

In that treatment the P-ACT-LB interval translates to

$\lambda$44

and amplitude normalization gives

$\lambda$45

This is still a low-$\lambda$46 attractor regime, but with an upward quantum shift in $\lambda$47 relative to the tree-level plateau (Yuennan et al., 7 Oct 2025).

The principal unresolved issues are not the existence of slow roll but the UV and matching problem. If $\lambda$48 crosses zero, pure Higgs inflation fails unless threshold corrections or new physics stabilize the potential (Salvio et al., 2015). If $\lambda$49 is not tuned small, scalaron dynamics generically dominates (Salvio et al., 2015). Metric formulations remain more sensitive to unknown UV physics than Palatini ones (Shaposhnikov et al., 2020). Taken together, these results imply that quantum-corrected Higgs inflation is best viewed not as a unique model, but as a structured family of effective theories whose common classical plateau is modified in sharply different ways by RG flow, curvature-squared operators, and UV completion.