Inverse Chiral Catalysis (ICC) in QCD

• Inverse chiral catalysis is the phenomenon where an external magnetic field lowers the chiral restoration temperature in QCD, contrary to standard magnetic catalysis.
• Various effective models, including linear sigma, NJL, and holographic approaches, reveal competing effects between valence enhancement and sea quark backreaction.
• The underlying mechanisms involve thermo-magnetic corrections, bosonic screening, and interaction weakening, which jointly shape the chiral phase transition.

Inverse chiral catalysis (ICC), often called inverse magnetic catalysis in the cited literature, denotes the phenomenon that in QCD-like theories subject to an external magnetic field $B$, the critical temperature for chiral symmetry restoration decreases as the field strength grows. It is therefore the opposite of ordinary magnetic catalysis, in which the field enhances chiral symmetry breaking and raises the transition scale. In the crossover region around $T_c$, the net effect of the magnetic background is to suppress the chiral condensate $\langle \bar\psi\psi\rangle$, whereas at low temperature—or, in several approaches, at asymptotically large $B$—the usual catalyzing behavior can re-emerge (Ayala et al., 2014, Bruckmann et al., 2013, Mueller et al., 2015).

1. Definition, observables, and diagnostic criteria

In field-theoretic terms, the condensate is an order parameter for spontaneous chiral symmetry breaking. In the Euclidean QCD formulation with a static homogeneous magnetic field, the partition function can be written as

$Z(B,T)=\int DA\,\det[\slashed D_A(B)+m]\,e^{-S_g[A]},$

and the chiral condensate is the mass derivative of the free-energy density,

$$\langle\bar\psi\psi\rangle(B,T)=\frac{1}{V}\,\frac{\partial \log Z(B,T)}{\partial m},$$

or equivalently the gauge-field average of $\mathrm{Tr}[(\slashed D_A(B)+m)^{-1}]$. In the chiral limit it is proportional to the spectral density at the origin via the Banks–Casher relation (Bruckmann et al., 2013).

The operational definition of the critical temperature depends on the framework. In the two-flavor linear sigma model with quarks, the chiral critical temperature $T_c(B)$ is located by the vanishing of the curvature of the effective potential at $v=\langle\sigma\rangle=0$,

$$\left.\frac{\partial^2V_{\rm eff}}{\partial v^2}\right|_{v=0}=0.$$

In Dyson–Schwinger and related continuum approaches, the pseudo-critical temperature is defined by the inflection point of $T_c$0 as a function of $T_c$1, or equivalently by the peak of the chiral susceptibility. In AMM-based NJL studies one also uses the inflection point of the constituent mass $T_c$2 (Ayala et al., 2014, Mueller et al., 2015, Wang et al., 2021).

A central phenomenological feature is the regime dependence. Continuum-extrapolated lattice QCD with physical quark masses shows that at $T_c$3 the condensate increases monotonically with $T_c$4, while around $T_c$5 it can turn over and decrease with further increasing $T_c$6 (Bruckmann et al., 2013). Full two-flavor QCD solved through coupled quark–gluon Dyson–Schwinger equations exhibits the same pattern: for small-to-moderate fields up to $T_c$7, $T_c$8 rises with $T_c$9 but $\langle \bar\psi\psi\rangle$0 decreases, whereas for $\langle \bar\psi\psi\rangle$1 both $\langle \bar\psi\psi\rangle$2 and $\langle \bar\psi\psi\rangle$3 eventually rise again (Mueller et al., 2015).

2. Effective-theory realizations

ICC has been realized in a range of low-energy and semi-microscopic models. What varies across these constructions is not the definition of the phenomenon, but the interaction channel through which the magnetic field weakens chiral order near the transition (Ayala et al., 2014, Pagura et al., 2016, Abreu et al., 2022, Wang et al., 2021, Mei et al., 2020, 2206.12054, Yu et al., 2014).

Framework	Magnetic ingredient emphasized	Reported ICC signature
Linear sigma model with quarks	thermo-magnetic corrections to $\langle \bar\psi\psi\rangle$4 and $\langle \bar\psi\psi\rangle$5, plus ring resummation	$\langle \bar\psi\psi\rangle$6 decreases monotonically with $\langle \bar\psi\psi\rangle$7
Nonlocal chiral quark model	$\langle \bar\psi\psi\rangle$8-dependent nonlocal form factor $\langle \bar\psi\psi\rangle$9	$B$0 decreases with $B$1 beyond $B$2
Finite-size two-flavor NJL	magnetized coupling $B$3 on a torus	magnetic field lowers $B$4, $B$5, and $B$6 near the transition in APBC
$B$7-flavor NJL with AMMs	$B$8	$B$9 and $Z(B,T)=\int DA\,\det[\slashed D_A(B)+m]\,e^{-S_g[A]},$0
PNJL with linear-in-$Z(B,T)=\int DA\,\det[\slashed D_A(B)+m]\,e^{-S_g[A]},$1 AMM	Pauli term in Landau-level dispersion	sufficiently large $Z(B,T)=\int DA\,\det[\slashed D_A(B)+m]\,e^{-S_g[A]},$2 yields $Z(B,T)=\int DA\,\det[\slashed D_A(B)+m]\,e^{-S_g[A]},$3
NJL with repulsive axial-vector channel	dynamical $Z(B,T)=\int DA\,\det[\slashed D_A(B)+m]\,e^{-S_g[A]},$4 from $Z(B,T)=\int DA\,\det[\slashed D_A(B)+m]\,e^{-S_g[A]},$5	$Z(B,T)=\int DA\,\det[\slashed D_A(B)+m]\,e^{-S_g[A]},$6 and then $Z(B,T)=\int DA\,\det[\slashed D_A(B)+m]\,e^{-S_g[A]},$7 decrease with $Z(B,T)=\int DA\,\det[\slashed D_A(B)+m]\,e^{-S_g[A]},$8

In the linear sigma model, the effective potential is assembled as

$Z(B,T)=\int DA\,\det[\slashed D_A(B)+m]\,e^{-S_g[A]},$9

with the ring term resumming the leading infrared bosonic self-energy insertions. Numerical evaluation for typical parameters such as $$\langle\bar\psi\psi\rangle(B,T)=\frac{1}{V}\,\frac{\partial \log Z(B,T)}{\partial m},$$0 and $$\langle\bar\psi\psi\rangle(B,T)=\frac{1}{V}\,\frac{\partial \log Z(B,T)}{\partial m},$$1 gives a monotonically decreasing $$\langle\bar\psi\psi\rangle(B,T)=\frac{1}{V}\,\frac{\partial \log Z(B,T)}{\partial m},$$2, often summarized by

$$\langle\bar\psi\psi\rangle(B,T)=\frac{1}{V}\,\frac{\partial \log Z(B,T)}{\partial m},$$3

The salient feature is that ICC appears only after going beyond mean field by including one-loop thermo-magnetic corrections to couplings and bosonic plasma screening through ring diagrams (Ayala et al., 2014).

Nonlocal chiral quark models provide a distinct route. The Euclidean action is built from nonlocal currents $$\langle\bar\psi\psi\rangle(B,T)=\frac{1}{V}\,\frac{\partial \log Z(B,T)}{\partial m},$$4, gauged with Wilson lines in the Landau gauge. At $$\langle\bar\psi\psi\rangle(B,T)=\frac{1}{V}\,\frac{\partial \log Z(B,T)}{\partial m},$$5 the models reproduce magnetic catalysis in good quantitative agreement with lattice QCD, but at finite $$\langle\bar\psi\psi\rangle(B,T)=\frac{1}{V}\,\frac{\partial \log Z(B,T)}{\partial m},$$6 they generate inverse magnetic catalysis because the effective nonlocal regulator becomes $$\langle\bar\psi\psi\rangle(B,T)=\frac{1}{V}\,\frac{\partial \log Z(B,T)}{\partial m},$$7-dependent and weakens the interaction in higher Landau levels (Pagura et al., 2016).

In finite-volume NJL implementations, ICC is not produced by the standard local model alone but by a magnetized coupling

$$\langle\bar\psi\psi\rangle(B,T)=\frac{1}{V}\,\frac{\partial \log Z(B,T)}{\partial m},$$8

whose decrease with $$\langle\bar\psi\psi\rangle(B,T)=\frac{1}{V}\,\frac{\partial \log Z(B,T)}{\partial m},$$9 suppresses the constituent mass $\mathrm{Tr}[(\slashed D_A(B)+m)^{-1}]$0. The same study shows that the qualitative outcome depends strongly on the spatial boundary conditions: APBC volumes act similarly to temperature and cooperate with ICC, whereas PBC retain an unsuppressed zero mode and can counteract it (Abreu et al., 2022).

AMM-driven variants replace the weakening of a collective coupling by a direct modification of single-particle spectra. In $\mathrm{Tr}[(\slashed D_A(B)+m)^{-1}]$1-flavor NJL, the AMM coupling is taken as $\mathrm{Tr}[(\slashed D_A(B)+m)^{-1}]$2, and the gap equations admit solutions such that $\mathrm{Tr}[(\slashed D_A(B)+m)^{-1}]$3 for moderately large $\mathrm{Tr}[(\slashed D_A(B)+m)^{-1}]$4 (Wang et al., 2021). In Pauli–Villars-regularized PNJL, a large enough constant AMM reverses the slope of the critical line so that inverse catalysis occurs throughout the $\mathrm{Tr}[(\slashed D_A(B)+m)^{-1}]$5 plane (Mei et al., 2020, 2206.12054). A separate NJL construction attributes ICC near $\mathrm{Tr}[(\slashed D_A(B)+m)^{-1}]$6 to a repulsive iso-scalar axial-vector interaction $\mathrm{Tr}[(\slashed D_A(B)+m)^{-1}]$7, induced by polarized instanton–anti-instanton molecules, which generates a dynamical chiral chemical potential $\mathrm{Tr}[(\slashed D_A(B)+m)^{-1}]$8 that competes with $\mathrm{Tr}[(\slashed D_A(B)+m)^{-1}]$9 (Yu et al., 2014).

3. Microscopic mechanisms

The best-known conceptual decomposition separates magnetic effects into a direct “valence” contribution and an indirect “sea” contribution. The valence effect is the fixed-background enhancement of low Dirac modes by $T_c(B)$0, which by itself increases $T_c(B)$1. The sea effect is carried by the quark determinant in the path integral: around $T_c(B)$2, the determinant orders the Polyakov loop and thereby suppresses low modes. Because the pure-gauge Polyakov-loop effective potential is especially flat near the crossover, this ordering effect becomes efficient precisely where ICC is seen on the lattice (Bruckmann et al., 2013).

This valence/sea distinction reappears in backreacted holographic QCD. In the Veneziano-limit model with

$T_c(B)$3

the magnetic field enters the flavor DBI sector through $T_c(B)$4, generating the factor

$T_c(B)$5

The $T_c(B)$6-dependence of the tachyon equation can then be split into explicit dependence through $T_c(B)$7, interpreted as a valence effect and always catalyzing, and implicit dependence through the backreacted geometry and dilaton, interpreted as a sea effect. In the large-$T_c(B)$8 limit, the explicit $T_c(B)$9 cancels from the tachyon equation, leaving the condensate governed purely by backreaction; this is the basis for the claim that flavor backreaction decatalyzes the condensate (Gürsoy et al., 2016).

A second widely used mechanism is interaction screening. In the linear sigma model, the ring contribution depends on the screened boson masses

$v=\langle\sigma\rangle=0$0

At the same time, one-loop thermo-magnetic corrections make $v=\langle\sigma\rangle=0$1, so that $v=\langle\sigma\rangle=0$2 decreases with $v=\langle\sigma\rangle=0$3, while $v=\langle\sigma\rangle=0$4 only increases mildly. ICC then emerges from competition between Landau-level enhancement and the weakening of the effective bosonic self-coupling together with stronger screening (Ayala et al., 2014).

Dyson–Schwinger and FRG-inspired analyses formulate the same competition in terms of the quark–gluon interaction. The quark-loop contribution to the gluon self-energy induces Debye-type screening, suppressing $v=\langle\sigma\rangle=0$5 for $v=\langle\sigma\rangle=0$6. At intermediate fields this drives the effective four-fermi coupling $v=\langle\sigma\rangle=0$7 to smaller values, which lowers $v=\langle\sigma\rangle=0$8 even though dimensional reduction tends to enhance pairing (Mueller et al., 2015). A recent functional-QCD calculation states the same mechanism in terms of a positive magnetic contribution to the gluon screening mass,

$v=\langle\sigma\rangle=0$9

which reduces the infrared quark–gluon interaction. In that framework the pseudo-critical temperature behaves as

$$\left.\frac{\partial^2V_{\rm eff}}{\partial v^2}\right|_{v=0}=0.$$0

so that the enhancement of the gluon screening mass becomes dominant near the chiral phase transition (Gao et al., 21 Jun 2026).

Other mechanisms are more model-specific but structurally similar. In nonlocal chiral quark models, the transverse part of the effective regulator is suppressed as $$\left.\frac{\partial^2V_{\rm eff}}{\partial v^2}\right|_{v=0}=0.$$1 grows, so the interaction strength decreases in higher Landau levels (Pagura et al., 2016). In AMM constructions, the Zeeman-like shift $$\left.\frac{\partial^2V_{\rm eff}}{\partial v^2}\right|_{v=0}=0.$$2 lowers the effective excitation threshold in each Landau level and can dominate over the usual magnetic enhancement (Mei et al., 2020, 2206.12054). In the axial-vector scenario, the generated $$\left.\frac{\partial^2V_{\rm eff}}{\partial v^2}\right|_{v=0}=0.$$3 acts like a chirality mismatch and destroys the condensate with pairing quarks between different chiralities (Yu et al., 2014).

4. Holographic formulations

Holography provides several nonperturbative realizations of ICC, and these are notable because they can incorporate full flavor backreaction rather than treating the magnetic field as a probe.

In improved holographic QCD in the Veneziano limit, the flavor action

$$\left.\frac{\partial^2V_{\rm eff}}{\partial v^2}\right|_{v=0}=0.$$4

contains a single free function controlling the $$\left.\frac{\partial^2V_{\rm eff}}{\partial v^2}\right|_{v=0}=0.$$5-coupling, $$\left.\frac{\partial^2V_{\rm eff}}{\partial v^2}\right|_{v=0}=0.$$6. The parameter $$\left.\frac{\partial^2V_{\rm eff}}{\partial v^2}\right|_{v=0}=0.$$7 tunes the strength of flavor backreaction: small $$\left.\frac{\partial^2V_{\rm eff}}{\partial v^2}\right|_{v=0}=0.$$8 implies large $$\left.\frac{\partial^2V_{\rm eff}}{\partial v^2}\right|_{v=0}=0.$$9, strong backreaction, and inverse catalysis; large $T_c$00 weakens backreaction and yields magnetic catalysis. Numerically, for $T_c$01, $T_c$02 gives decreasing $T_c$03, decreasing $T_c$04, and decreasing $T_c$05 near $T_c$06, whereas $T_c$07 yields the opposite trend (Gürsoy et al., 2016).

Bottom-up holographic QCD reaches ICC through a different construction. The scalar dual to $T_c$08 is coupled to the magnetic field by two lowest-order operators,

$T_c$09

At $T_c$10, the $T_c$11-term encourages the scalar profile to bend off zero and supports magnetic catalysis. Near the thermal transition, however, the factor $T_c$12 suppresses the $T_c$13-term near the horizon, so the $T_c$14-term dominates and can shift the effective mass upward, weakening the instability that drives chiral symmetry breaking. In the massless theory there is a region of parameter space, in particular $T_c$15, with magnetic catalysis at zero temperature but inverse magnetic catalysis at temperatures of order the thermal phase transition (Evans et al., 2016).

A lower-dimensional example occurs in the $T_c$16-dimensional soft-wall model. There the condensate $T_c$17 decreases with increasing $T_c$18 for $T_c$19, reaches a minimum at $T_c$20, and then grows linearly for $T_c$21. The pseudocritical field is defined by

$T_c$22

with numerical values $T_c$23 at $T_c$24, $T_c$25 at $T_c$26, and $T_c$27 at $T_c$28. The stated mechanism is competition between the confining dilaton profile and magnetic-field-induced warp-factor backreaction (Rodrigues et al., 2018).

Top-down holography also exhibits inverse catalysis in dense matter. In the Sakai–Sugimoto model at low temperature and finite chemical potential, adding a magnetic field decreases the critical chemical potential for chiral restoration, $T_c$29, over a substantial range before ordinary magnetic catalysis is recovered at larger $T_c$30. The paper estimates that this inverse effect persists up to $T_c$31 (Preis et al., 2010).

A related but distinct construction is inverse anisotropic catalysis in V–QCD. There, anisotropy acts destructively on the condensate near the transition temperature and lowers the chiral transition temperature. The authors suggest that the cause for inverse magnetic catalysis may be the anisotropy caused by the magnetic field rather than the charge dynamics created by it (Gursoy et al., 2018). This is not itself ICC, but it broadens the class of backreaction-driven mechanisms that produce the same phenomenology.

5. Regime structure: temperature, field strength, size, and density

ICC is not a statement about all temperatures or all magnetic-field strengths. Rather, the cited literature consistently presents it as a regime-specific outcome of competing effects.

At low temperature, many models recover ordinary magnetic catalysis. In the Dyson–Schwinger picture, the lowest Landau level effectively reduces the phase space from $T_c$32 to $T_c$33 dimensions, enhancing the quark self-energy and generating

$T_c$34

or more simply $T_c$35 when $T_c$36 runs only logarithmically. Thermal fluctuations then set $T_c$37, which restores magnetic catalysis in the asymptotic regime (Mueller et al., 2015). The same low-$T_c$38 increase of the condensate appears in nonlocal quark models, AMM-based NJL/PNJL models, and lattice-guided analyses (Pagura et al., 2016, Wang et al., 2021, Bruckmann et al., 2013).

Near the crossover, the sign can reverse. In the linear sigma model, $T_c$39 decreases monotonically with the dimensionless ratio $T_c$40 once thermo-magnetic coupling corrections and ring diagrams are included (Ayala et al., 2014). In nonlocal chiral quark models, all parametrizations studied yield $T_c$41 decreasing with $T_c$42 beyond $T_c$43 (Pagura et al., 2016). In improved holographic QCD with $T_c$44, the critical behavior is reported as

$T_c$45

with $T_c$46 negative for $T_c$47 (Gürsoy et al., 2016).

Finite size adds another competing scale. On a torus with APBC, shrinking the volume lowers the constituent mass and shifts both thermal and spatial susceptibilities toward restoration; in that case finite size acts similarly to temperature and cooperates with ICC. With PBC, however, the spatial zero mode enhances infrared correlations, so decreasing $T_c$48 can increase $T_c$49 and raise the pseudo-critical scale, thereby counteracting the inverse effect of the magnetic field (Abreu et al., 2022).

At finite density, inverse catalysis need not be parameterized by $T_c$50 alone. In the Sakai–Sugimoto model the key observable is $T_c$51, and at low temperature the magnetic field initially lowers the critical chemical potential for restoration (Preis et al., 2010). In PNJL with large AMM, both $T_c$52 at $T_c$53 and $T_c$54 at $T_c$55 decrease with increasing $T_c$56, producing inverse catalysis throughout the $T_c$57 plane; for smaller AMM the phase boundaries can cross, and the behavior depends on the region of the phase diagram (2206.12054).

6. Conceptual issues, misconceptions, and related extensions

A common misconception is to treat ICC as a contradiction of Landau-level enhancement. The cited works do not support that reading. Instead, they repeatedly describe ICC as the outcome of competition between a catalyzing tendency—dimensional reduction, lowest-Landau-level enhancement, or direct valence coupling—and a decatalyzing tendency—sea-quark backreaction, Polyakov-loop ordering, screening of bosonic or gluonic modes, weakening of effective couplings, or AMM-induced lowering of excitation energies (Bruckmann et al., 2013, Mueller et al., 2015, Ayala et al., 2014, Gürsoy et al., 2016).

A second misconception is that ICC should appear automatically in any low-energy chiral model. Several papers make the opposite point. If one uses a constant four-fermi coupling or a constant $T_c$58, only monotonic magnetic catalysis arises in the simplified NJL-type description of QCD (Mueller et al., 2015). Likewise, in local NJL one generally has to impose an ad hoc $T_c$59-dependent coupling to reproduce IMC, whereas in nonlocal models the behavior follows naturally from the field dependence of the regulator (Pagura et al., 2016).

The terminology itself is slightly variable. “Inverse chiral catalysis” emphasizes the decrease of the critical temperature for chiral restoration, while “inverse magnetic catalysis” emphasizes the suppression of $T_c$60 by an external magnetic field. In the cited papers these phrases refer to the same chiral phenomenon under $T_c$61, not to distinct effects (Ayala et al., 2014, Bruckmann et al., 2013).

Several extensions broaden the concept without altering its core definition. “Axial inverse magnetic catalysis” concerns the $T_c$62 sector and is found to correlate with chiral IMC in a $T_c$63-flavor NJL model with quark AMMs (Wang et al., 2021). “Inverse anisotropic catalysis” in holographic QCD lowers the chiral transition temperature through anisotropy alone and has been proposed as a broader sea-type mechanism related to ICC (Gursoy et al., 2018). Ongoing directions explicitly named in the cited literature include matching $T_c$64 to electromagnetic susceptibility data, extending holographic models to finite quark mass $T_c$65, baryon chemical potential $T_c$66, and $T_c$67 corrections, and adding quark wave-function renormalization, momentum-dependent vector interactions, Polyakov-loop dynamics, and beyond-mean-field corrections in nonlocal quark models (Gürsoy et al., 2016, Pagura et al., 2016).

Taken together, these studies present ICC not as a single mechanism but as a recurrent pattern in magnetized chiral matter: near the restoration region, the magnetic field can weaken the effective interaction responsible for $T_c$68, even though the same field enhances low-energy fermionic phase space. The specific dynamical realization varies across lattice-motivated effective theories, continuum functional methods, and holography, but the defining signature remains the same: a lowering of the chiral restoration scale with increasing magnetic field (Ayala et al., 2014, Mueller et al., 2015, Gürsoy et al., 2016, Gao et al., 21 Jun 2026).