Thermal Instability in Astrophysics

• Thermal instability (TI) is the process by which small thermal perturbations amplify due to imbalanced heating and cooling, leading to multiphase structures.
• TI links local thermodynamics to condensation, cloud formation, turbulent dynamics, and episodic accretion by differentiating isobaric, isochoric, and isentropic behavior.
• The phenomenon is modulated by thermal conduction, magnetic fields, geometry, and nonequilibrium conditions, affecting environments from galaxy halos to protoplanetary discs.

Searching arXiv for recent and foundational papers on thermal instability across astrophysical contexts. Thermal instability (TI) is the amplification of thermal perturbations in a medium whose net heating and cooling make small departures from thermal balance grow rather than decay. In astrophysical plasmas and discs, TI links local thermodynamics to condensation, cloud formation, multiphase structure, turbulent morphology, and episodic accretion. The subject began with the linear theory associated with Parker and Field, but current work treats TI in nonequilibrium states, in the presence of thermal conduction, magnetic fields, cosmic rays, spherical geometry, and strongly time-dependent flows, so that “thermal instability” now denotes a broad class of related mechanisms rather than only the homogeneous, static limit (Falle et al., 2020). Modern analyses therefore emphasize not only the local sign of a stability criterion, but also whether gas can enter a TI zone, remain there long enough for perturbations to grow, and saturate by steady condensation, rebound, pulsation, or suppression (Stricklan et al., 19 May 2025).

1. Linear stability framework

The classical formulation of TI is expressed in terms of the net cooling function $\mathcal{L}(\rho,T)$, or, in cooling-function units, $L_{\rm net} = \Lambda(T) - \Gamma(n)$. Balbus’ nonequilibrium generalization of the Field criteria uses the thermodynamic derivatives

$$\left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_p < 0, \qquad \left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_\rho < 0,$$

for isobaric and isochoric instability, respectively (Stricklan et al., 19 May 2025). In the re-analysis of Field’s theory by Falle and collaborators, the physically relevant parameter in the absence of conduction and magnetic field is

$\alpha = \frac{a_e^2}{a_f^2},$

with $\alpha < 0$ corresponding to isobaric instability, $0 \leq \alpha \leq 1$ to stability, and $\alpha > 1$ to isentropic instability (Falle et al., 2020).

This linear framework already distinguishes several regimes that remain central in later work. The isobaric mode is the classical condensation mode, while isochoric and isentropic behavior become important when pressure equilibrium is not maintained. The same analyses also show that magnetic fields can be incorporated without changing the basic stability criteria, even though magnetic fields later prove decisive for morphology, anisotropic transport, and nonlinear evolution (Falle et al., 2020).

A recurring consequence is that TI is most naturally formulated as a competition between local thermal response and dynamical communication. In short-wavelength, pressure-balanced limits, isobaric growth dominates; when sound crossing is slow, nonisobaric behavior becomes important. This division underlies later distinctions between gentle condensation, cloud pulsation, and complete suppression of clump formation.

2. Nonequilibrium states, TI zones, and saturation

A major extension of the classical theory is the systematic treatment of nonequilibrium states. The 2025 stability analysis of cooling functions introduces a critical cooling rate $\Lambda_c$ that maps TI zones even away from thermal equilibrium, with instability written as

$\Lambda(T) > \Lambda_c.$

In this formalism, the isochoric threshold can be written

$$\Lambda_c = \Gamma(n) + T\frac{d\Lambda(T)}{dT} + E\left( \frac{\lambda_F}{\lambda} \right)^2,$$

and analogous expressions hold for the isobaric mode (Stricklan et al., 19 May 2025). This reformulation makes it possible to follow parcels through phase space and determine whether they cross into or out of TI zones without requiring them to remain on an equilibrium curve.

The nonlinear saturation of TI has also been recast in analytic form. Starting from the entropy equation

$L_{\rm net} = \Lambda(T) - \Gamma(n)$0

recent work derives general identities showing that saturation is triggered when a fluid element crosses the boundary of the instability region, the Balbus contour (Waters et al., 2023). The key event is a “pressure reversal”: the comoving time derivative of the pressure changes sign, the gas pressure force reverses direction, and continued growth of the condensation is slowed.

This saturation mechanism differs sharply between isobaric and nonisobaric evolution. For isobaric evolution, pressure reversal occurs nearly simultaneously for every fluid element in the condensation and a steady state is quickly reached. For nonisobaric evolution, the condensation is no longer in mechanical equilibrium; the contracting gas rebounds with greater force during the expansion phase, and the cloud then pulsates because the return to mechanical equilibrium becomes wave-mediated (Waters et al., 2023). The same analysis also identifies an isochoric TI zone and clarifies that, unless this zone intersects the equilibrium curve, isochoric modes can only become unstable if the plasma is in a state of thermal nonequilibrium.

3. Transport processes, magnetic fields, and geometry

Thermal conduction is the classical small-scale stabilizer of TI. In several of the cited studies, the relevant stabilizing scale is the Field length,

$L_{\rm net} = \Lambda(T) - \Gamma(n)$1

or equivalent forms differing only in notation (Stricklan et al., 19 May 2025). Perturbations with $L_{\rm net} = \Lambda(T) - \Gamma(n)$2 are suppressed, and increasing conduction shrinks TI zones. This effect is not confined to linear theory: in simulations with isotropic conduction and viscosity, cloud structure and dynamics become dominated by evaporation and condensation flows, while the Darrieus–Landau instability disrupts cloud surfaces (Jennings et al., 2020).

Magnetic fields reorganize rather than simply eliminate TI. In the magnetized interstellar medium, anisotropic conduction does not significantly alter the overall density and temperature statistics in the saturated state, but it strongly affects the shapes and sizes of the cold clouds formed by TI. Uniform initial fields produce long filaments of cold gas, whereas initially tangled fields do not (Choi et al., 2011). In related 2D MHD simulations, high-$L_{\rm net} = \Lambda(T) - \Gamma(n)$3 runs show dense filaments aligned with local magnetic fields, while strong magnetic fields suppress cross-field contraction and allow cold filaments to form along or perpendicular to the initial fields (Jennings et al., 2020).

The same sensitivity appears when pressure becomes anisotropic. In simulations with a polybaric pressure model, the strength of the background magnetic field only marginally affects the overall evolution, but the ratio of the parallel and perpendicular pressures can considerably alter the mass and volume fractions of the warm, cold, and unstable neutral components (Sarkar et al., 2024). The resulting turbulence is described as largely isobaric and incompressible, even though it is driven purely by time-dependent heating and cooling.

Geometry also modifies the instability boundary. In spherical interstellar clouds, using spherical perturbations on a quasi-static spherical cloud changes the instability criterion relative to flat geometry, so that sphericalness can increase the occurrence of TI. In that setting, perturbations with shorter wavelengths have more chance to grow via TI, with greater growth rates (Nejad-Asghar, 2024). TI is therefore not purely a local thermodynamic property; it is mediated by conduction, anisotropy, magnetic topology, and the geometry of the background state.

4. Multiphase halos, molecular clouds, and condensation

A prominent application of TI is the formation of multiphase structure in hot halos, the circumgalactic medium, and molecular gas. In clusters, groups, and galaxies, local TI can produce a multiphase medium with $L_{\rm net} = \Lambda(T) - \Gamma(n)$4 K cold filaments condensing out of the hot medium only when $L_{\rm net} = \Lambda(T) - \Gamma(n)$5. In the same simulations, the interplay among heating, cooling, and TI reduces the net cooling rate and the mass accretion rate at small radii by factors of $L_{\rm net} = \Lambda(T) - \Gamma(n)$6 relative to cooling-flow models, and the required feedback efficiency is $L_{\rm net} = \Lambda(T) - \Gamma(n)$7 for clusters (Sharma et al., 2011).

Environment	Control condition	Reported outcome
Clusters, groups, galaxies	$L_{\rm net} = \Lambda(T) - \Gamma(n)$8	$L_{\rm net} = \Lambda(T) - \Gamma(n)$9 K cold filaments condense
CGM with cosmic rays	$$\left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_p < 0, \qquad \left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_\rho < 0,$$0	trapped CRs inhibit collapse; fast escape reduces support
Molecular cloud cores	wavelengths greater than few astronomical units	perturbations can grow to form LMCs

In the circumgalactic medium, cosmic rays modify but do not uniformly suppress TI. Under purely advective CR transport, CR pressure significantly suppresses the collapse of thermally unstable regions. In more realistic transport models, CRs escape collapsing regions via streaming and diffusion along magnetic fields, diminishing their influence on the thermal and dynamic structure of the cold CGM. The key control parameter is the ratio of the CR escape time to the cloud collapse time; slow effective CR diffusion maximizes pressure support, whereas fast effective diffusion facilitates rapid CR escape (Weber et al., 30 Jan 2025).

In molecular cloud cores, the situation is different again. Including heating due to ambipolar diffusion can lead to the occurrence of isobaric TI on a timescale smaller than the dynamical timescale, and linear analysis places the unstable region in the outer part of the core. Perturbations with wavelengths greater than few astronomical units are protected from the destabilization property of thermal conduction, so they can grow to form low-mass condensations (Nejad-Asghar, 2011). In the outer half of quasi-static spherical molecular clumps and cores, the local thermal balance yields $$\left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_p < 0, \qquad \left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_\rho < 0,$$1 with $$\left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_p < 0, \qquad \left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_\rho < 0,$$2; increasing $$\left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_p < 0, \qquad \left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_\rho < 0,$$3 leads to flatter density profiles and to more thermally unstable Inside-Rush-Perturbation and Outside-Rush-Perturbation modes with smaller growth timescales (Nejad-Asghar, 2019).

5. Active galactic nuclei and radiatively driven outflows

Thermal driving on parsec scales in active galactic nuclei provides one of the clearest demonstrations that TI depends on background dynamics as much as on local cooling physics. In simulations of AGN outflows irradiated by an AGN, the first clumpy wind solutions in 1-D and 2-D were obtained only after identifying three obstacles to clump formation: insufficient radiative heating or other physical processes that prevent the outflowing gas from entering the TI zone, the stabilizing effect of stretching due to rapid radial acceleration, and a flow-speed effect in which gas passes through the TI zone too quickly for perturbations to grow. A necessary condition to trigger TI in an outflow is that the pressure ionization parameter decreases along a streamline once gas enters a TI zone (Dannen et al., 2020).

The same conclusion is sharpened in the analysis of highly supersonic outflows. During the launching phase, acceleration stretches fluid elements and suppresses cloud formation, making TI-driven clump formation unlikely in the acceleration zone. The most favorable conditions are found at radii beyond the acceleration zone, where the growth rate of entropy modes is set by the linear theory rate for a static plasma. Time variability in the radiation field is identified as a robust means of placing gas in a TI zone, while adiabatic cooling and heat advection tend to remove gas from it (Waters et al., 2021).

TI also constrains the warm, optically thick X-ray corona proposed to explain the soft X-ray excess in AGN. In a self-consistent vertical structure model with MRI-driven magnetic heating and radiative cooling by free-free processes and Compton scattering, magnetic pressure does not remove TI caused by radiative processes operating in X-ray emitting plasma. TI disappears only in the case of accretion rates higher than $$\left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_p < 0, \qquad \left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_\rho < 0,$$4 of Eddington and high magnetic field parameter $$\left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_p < 0, \qquad \left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_\rho < 0,$$5. In the permitted regime, the warm Compton-cooled corona has typical temperature in the range of $$\left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_p < 0, \qquad \left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_\rho < 0,$$6 keV and optical depth even up to $$\left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_p < 0, \qquad \left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_\rho < 0,$$7 (Gronkiewicz et al., 2023).

6. Discs, episodic accretion, and planetary atmospheres

In protoplanetary discs, TI is often tied to hydrogen ionisation and to the S-curve structure of local thermal equilibria. In 2D axisymmetric radiation hydrodynamic models of the inner dead-zone edge around a Class II T Tauri star, TI develops in discs with moderate accretion rates and results from the activation of the magnetorotational instability in the dead zone. The instability creates an extensive MRI active region around the midplane, disrupts the stable pebble- and migration trap at the inner edge of the dead zone, and consistently produces TI-reflares that generate pressure maxima within $$\left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_p < 0, \qquad \left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_\rho < 0,$$8 AU. On a timescale of a few thousand years, TI regularly disrupts the disc’s radial and vertical structure within $$\left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_p < 0, \qquad \left( \frac{\partial \mathcal{L}/T}{\partial T}\right)_\rho < 0,$$9 AU, destroying and later reinstating stable migration traps (Cecil et al., 2024).

The same hydrogen-ionisation TI is used to interpret episodic eruptions of young accreting stars. In one recent study, the TI scenario with realistic viscosity values explains intermediate type bursters and predicts a dearth, or “desert,” of bursts with peak accretion rates between $\alpha = \frac{a_e^2}{a_f^2},$0 and $\alpha = \frac{a_e^2}{a_f^2},$1. Most classic EXORs appear to be on the cold branch of the S-curve during peak light, whereas most classic FUORs appear to be on the hot branch. At the same time, TI is unable to explain how classic FUORs can last for up to centuries, and it over-predicts the occurrence rate of short FUORs by at least an order of magnitude; the conclusion is that TI is a required ingredient of episodic accretion operating at $\alpha = \frac{a_e^2}{a_f^2},$2 au, but additional physics must act at larger scales (Nayakshin et al., 2024). In high-mass young stellar objects, 1D simulations reproduce the durations and peak accretion rates of long outbursts, but they struggle with short-duration bursts and fail to reproduce multiple outbursts seen in some systems, which motivates alternatives such as gravitational instabilities and disc fragmentation (Elbakyan et al., 2024).

A distinct form of TI appears in hot Jupiter atmospheres as thermo-resistive instability. In radiatively dominated atmospheric regions, the instability occurs when the ohmic dissipation rate increases with temperature faster than the radiative cooling rate. The characteristic domain is typically $\alpha = \frac{a_e^2}{a_f^2},$3–$\alpha = \frac{a_e^2}{a_f^2},$4 mbar and $\alpha = \frac{a_e^2}{a_f^2},$5–$\alpha = \frac{a_e^2}{a_f^2},$6 K, making the mechanism a candidate explanation for dayside thermal inversions. The instability is suppressed by high levels of non-thermal photoionization, and its nonlinear outcome requires global non-linear atmospheric models with adequate MHD prescriptions (Menou, 2012).

7. Current extensions, reinterpretations, and open controversies

One of the most active current debates concerns whether observed condensations should be attributed to classical TI or to a related but distinct instability. In coronal-loop simulations, a $\alpha = \frac{a_e^2}{a_f^2},$7-based stability analysis shows that loops undergoing periodic episodes of coronal rain formation are linearly unstable to catastrophic cooling instability, while TI is stabilized by thermal conduction. In this interpretation, traditional analyses based only on the Field criterion fail to account for nonequilibrium evolution and conduction, and catastrophic cooling, rather than classical TI, is the dominant trigger (Stricklan et al., 19 May 2025).

Another extension proposes a new regime, “latent thermal instability,” for the intracluster medium outside cluster cores. In weakly collisional or collisionless plasmas, heat-flux-driven plasma instabilities can anomalously suppress thermal conduction, opening a parameter regime in which condensates form under conditions that would be stable in the classical conductive limit. In the reported simulations and analysis, this extends the regime of instability-driven fluctuations to over $\alpha = \frac{a_e^2}{a_f^2},$8 of the cluster, and condensates can reach a steady state as in the hydrodynamic limit (Choudhury et al., 24 Jun 2026).

These developments suggest a more restrictive and more precise view of TI. Thermal instability is not synonymous with every condensation event, every multiphase medium, or every accretion outburst. In some systems, the decisive issue is whether conduction or CR escape stabilizes the gas; in others, whether the gas can enter a TI zone at all; in still others, whether the relevant linear mode is TI or catastrophic cooling. This suggests that TI is best understood as a conditional mechanism embedded in a larger thermal-dynamical problem, with its observable outcome determined by cooling functions, transport coefficients, geometry, magnetic topology, gravity, and background flow history.