Photospheric Radius Expansion in Neutron Stars

• PRE is the temporary inflation of a neutron star’s radiative surface during Type I X-ray bursts when radiation pressure overcomes gravity at the Eddington limit.
• Observations reveal a canonical sequence—expansion, contraction, touchdown, and cooling—that is used to diagnose burst atmospheres and infer physical properties.
• PRE bursts serve as diagnostic tools for neutron-star mass–radius and equation-of-state studies despite challenges like non-static photospheres and composition-dependent effects.

Photospheric radius expansion (PRE) is the temporary inflation of a neutron star’s radiating surface during a thermonuclear Type I X-ray burst when the burst luminosity reaches, or slightly exceeds, the local Eddington limit. In accreting neutron-star low-mass X-ray binaries, unstable burning of accumulated H/He on the stellar surface can drive the radiative flux high enough that radiation pressure lifts the outer atmospheric layers; the effective photosphere then moves to larger radius and lower effective temperature while the observed spectrum remains approximately Planckian with evolving apparent radius and temperature (Zhang et al., 2012). In the current literature, PRE is both a phenomenological classification of luminous bursts and a diagnostic framework for burst atmospheres, radiation-driven winds, ignition geometry, and neutron-star mass–radius inference (Sala et al., 2012, Guichandut et al., 2021).

1. Physical basis and Eddington-limited expansion

The basic condition for PRE is the Eddington balance between outward radiative force and inward gravity. For electron-scattering opacity $\kappa$, the Eddington luminosity is

$L_{\rm Edd} = \frac{4\pi G M c}{\kappa},$

with $M$ the neutron-star mass and $c$ the speed of light (Zhang et al., 2012). In composition-dependent form used in PRE mass–radius analyses, the opacity is commonly written as $\kappa_e = 0.2(1+X)\ {\rm cm^2\,g^{-1}}$, where $X$ is the hydrogen mass fraction, so helium-rich atmospheres have a higher Eddington luminosity than hydrogen-rich atmospheres (Sala et al., 2012). For a distant observer, the flux is further modified by distance and gravitational redshift; one form used for the Eddington flux at infinity is

$$F_{\infty} = \frac{GMc}{0.2(1+X) D^2}\left(1-\frac{2GM}{Rc^2}\right)^{1/2},$$

where $D$ is the distance and $R$ the stellar radius (Sala et al., 2012).

During PRE, the apparent emitting radius inferred from blackbody fits is not the true stellar radius. A standard relation is

$R_{\rm bb} = \frac{R}{f_c^2}(1+z)^{-1},$

where $L_{\rm Edd} = \frac{4\pi G M c}{\kappa},$0 is the color-correction factor and $L_{\rm Edd} = \frac{4\pi G M c}{\kappa},$1 is the gravitational-redshift factor (Zhang et al., 2012). This dependence is central: changes in $L_{\rm Edd} = \frac{4\pi G M c}{\kappa},$2 can reflect physical expansion and contraction of the photosphere, changes in $L_{\rm Edd} = \frac{4\pi G M c}{\kappa},$3, or both.

Recent general-relativistic atmosphere calculations place PRE into two regimes. Static expanded envelopes give $L_{\rm Edd} = \frac{4\pi G M c}{\kappa},$4–$L_{\rm Edd} = \frac{4\pi G M c}{\kappa},$5 km, whereas radiation-driven winds give $L_{\rm Edd} = \frac{4\pi G M c}{\kappa},$6–$L_{\rm Edd} = \frac{4\pi G M c}{\kappa},$7 km (Guichandut et al., 2021). In these calculations, photospheric radii of less than $L_{\rm Edd} = \frac{4\pi G M c}{\kappa},$8 km can be explained by static envelopes, but only in a narrow range of luminosity, while higher luminosities lead naturally to a wind with photospheric radius $L_{\rm Edd} = \frac{4\pi G M c}{\kappa},$9 km (Guichandut et al., 2021). This distinction matters because it directly affects how touchdown fluxes and apparent radii are interpreted.

2. Observational sequence and defining signatures

Time-resolved spectroscopy identifies PRE through a characteristic evolution of bolometric flux, color temperature, and apparent blackbody radius. A canonical sequence is: rise to Eddington, expansion, contraction, touchdown, and cooling tail (Zhang et al., 2012). In 4U 1636–53, for a representative PRE burst, the rise over the first $M$0–$M$1 s brings the color temperature to $M$2–$M$3 keV while $M$4 is $M$5–$M$6 km; the expansion phase around the peak shows $M$7–$M$8 keV and $M$9–$c$0 km at flux saturation; the contraction then returns $c$1 to $c$2–$c$3 km while $c$4 rises again to $c$5–$c$6 keV (Zhang et al., 2012). The point where $c$7 reaches its minimum after contraction is the touchdown point.

Operationally, touchdown is often taken to mark the return of the photosphere to the neutron-star surface, but this assumption is not exact. Static-envelope calculations show that in the contraction phase the photosphere can still be $c$8 km above the surface when the effective temperature is only $c$9 away from its maximum value (Guichandut et al., 2021). A plausible implication is that touchdown fluxes carry a systematic uncertainty when they are used as exact measures of the Eddington flux at the stellar surface.

The 2018 NICER burst from 4U 1820–30 demonstrated how soft X-ray coverage changes PRE phenomenology. NICER measured a maximum apparent radius of $\kappa_e = 0.2(1+X)\ {\rm cm^2\,g^{-1}}$0 km and a minimum blackbody temperature of $\kappa_e = 0.2(1+X)\ {\rm cm^2\,g^{-1}}$1 keV during a strong expansion phase lasting $\kappa_e = 0.2(1+X)\ {\rm cm^2\,g^{-1}}$2 s, followed by a moderate-expansion plateau at $\kappa_e = 0.2(1+X)\ {\rm cm^2\,g^{-1}}$3 km and $\kappa_e = 0.2(1+X)\ {\rm cm^2\,g^{-1}}$4 keV, with touchdown at $\kappa_e = 0.2(1+X)\ {\rm cm^2\,g^{-1}}$5 s after burst onset (Keek et al., 2018). Because NICER covers 0.2–12 keV, the event appeared as a soft spike; instruments sensitive only above 3 keV would have registered a dip at the same stage (Keek et al., 2018).

A source-specific contrast comes from GRS 1741.9–2853. One NuSTAR burst showed clear PRE, with $\kappa_e = 0.2(1+X)\ {\rm cm^2\,g^{-1}}$6, a temperature decrease from $\kappa_e = 0.2(1+X)\ {\rm cm^2\,g^{-1}}$7 keV to $\kappa_e = 0.2(1+X)\ {\rm cm^2\,g^{-1}}$8 keV, and an apparent radius increase to $\kappa_e = 0.2(1+X)\ {\rm cm^2\,g^{-1}}$9 km for an assumed 7 kpc distance, followed by contraction to $X$0 km at touchdown (Pike et al., 2021). Another burst in the same source showed a double-peaked bolometric profile but no PRE signature, illustrating that multiple-peaked burst morphology is not equivalent to radius expansion (Pike et al., 2021).

3. Post-touchdown evolution, cooling, and burst oscillations

The post-touchdown phase is not uniform across PRE bursts. In 4U 1636–53, analysis of 1490 RXTE observations found 336 Type I bursts, including 69 PRE bursts, and showed that the post-touchdown evolution of $X$1 falls into two distinct classes (Zhang et al., 2012). The first class, seen in 52 of the 69 PRE bursts, has $X$2 decrease rapidly to touchdown, remain approximately constant for $X$3–$X$4 s, and then increase slowly. The second class, seen in the remaining 17 bursts, has $X$5 decrease to touchdown and then increase again quickly within $X$6 s, after which it either decreases slightly or remains nearly constant at the larger value (Zhang et al., 2012).

Zhang et al. defined the post-touchdown phase as the contiguous time interval after the burst peak during which $X$7 km (Zhang et al., 2012). The distributions of its duration differ sharply between bursts with and without tail oscillations: bursts with tail oscillations have post-touchdown durations longer by a factor $X$8, while bursts without tail oscillations cluster around $X$9–$$F_{\infty} = \frac{GMc}{0.2(1+X) D^2}\left(1-\frac{2GM}{Rc^2}\right)^{1/2},$$0 s, with a Kolmogorov–Smirnov probability of only $$F_{\infty} = \frac{GMc}{0.2(1+X) D^2}\left(1-\frac{2GM}{Rc^2}\right)^{1/2},$$1 that the two groups come from the same parent population (Zhang et al., 2012). In this sample, every PRE burst with tail oscillations had a prolonged interval of nearly constant $$F_{\infty} = \frac{GMc}{0.2(1+X) D^2}\left(1-\frac{2GM}{Rc^2}\right)^{1/2},$$2, whereas none of the short post-touchdown bursts showed tail oscillations (Zhang et al., 2012).

This phenomenology was interpreted in terms of a cooling-wake model in which the speed and width of the cooling front depend on latitude: near the equator the front moves faster and is wider, while toward the poles it is slower and narrower, with an equator-to-pole speed difference of a factor $$F_{\infty} = \frac{GMc}{0.2(1+X) D^2}\left(1-\frac{2GM}{Rc^2}\right)^{1/2},$$3 (Zhang et al., 2012). In that picture, long post-touchdown phases with tail oscillations correspond to slower, high-latitude cooling wakes, whereas short post-touchdown phases without oscillations correspond to fast equatorial wakes (Zhang et al., 2012). The same work connected this to accretion state through the source position $$F_{\infty} = \frac{GMc}{0.2(1+X) D^2}\left(1-\frac{2GM}{Rc^2}\right)^{1/2},$$4 in the color–color diagram, with PRE bursts showing tail oscillations clustering at higher $$F_{\infty} = \frac{GMc}{0.2(1+X) D^2}\left(1-\frac{2GM}{Rc^2}\right)^{1/2},$$5 (Zhang et al., 2012).

The cooling phase also complicates the standard use of PRE bursts as fixed-area radiators. In 4U 1636–53, the average flux–temperature relation during the cooling phase differs significantly from the canonical $$F_{\infty} = \frac{GMc}{0.2(1+X) D^2}\left(1-\frac{2GM}{Rc^2}\right)^{1/2},$$6 relation for PRE, hard non-PRE, and soft non-PRE bursts, and a single power law cannot fit the average relation for any of the three types (Zhang et al., 2010). The same study concluded that hard non-PRE bursts ignite in a hydrogen-rich atmosphere, whereas soft non-PRE and PRE bursts are helium-rich, and that the metal abundance in the atmosphere decreases as the bursts decay, probably because heavy elements sink faster than H and He (Zhang et al., 2010). This directly undercuts the assumption that a single, constant $$F_{\infty} = \frac{GMc}{0.2(1+X) D^2}\left(1-\frac{2GM}{Rc^2}\right)^{1/2},$$7 and constant emitting area apply throughout the cooling tail.

4. PRE as a mass–radius and equation-of-state probe

PRE bursts provide two observables that can be combined to constrain neutron-star structure: the touchdown flux, used as a proxy for the Eddington flux, and the apparent radius in the cooling tail. In one commonly used formulation,

$$F_{\infty} = \frac{GMc}{0.2(1+X) D^2}\left(1-\frac{2GM}{Rc^2}\right)^{1/2},$$8

so the apparent angular area depends on the true radius, distance, color correction, and compactness (Sala et al., 2012). The Rapid Burster provides a clear example: a Swift/XRT PRE burst from MXB 1730–335 was analyzed with a Bayesian method and, after marginalization over a likely distance of 5.8–10 kpc, yielded $$F_{\infty} = \frac{GMc}{0.2(1+X) D^2}\left(1-\frac{2GM}{Rc^2}\right)^{1/2},$$9 and $D$0 km at $D$1, with the distance identified as the dominant systematic (Sala et al., 2012).

The same logic has been extended to joint analyses with independent constraints. In Aql X-1, simultaneous use of PRE bursts and quiescent spectra implied a distance range of 4.0–5.75 kpc from the overlap of the two confidence regions, and the resulting mass–radius constraints were reported to be compatible with strange-star equations of state and conventional neutron-star models (Li et al., 2017). Conversely, source-dependent systematics can drive extreme inferences. In 4U 1746–37, three PRE bursts with low touchdown fluxes, combined with geometric corrections for a high inclination angle, produced low-mass solutions such as $D$2 and $D$3 km or $D$4 and $D$5 km, depending on the geometric scenario (Li et al., 2014). In XTE J1810–189, application of the direct cooling-tail method to a PRE burst gave $D$6 and $D$7 km in high-metallicity atmospheres, but $D$8–$D$9 and $R$0–13 km in low-metallicity, hydrogen-rich models (Ban et al., 13 Mar 2026). These examples show that PRE inference is composition-sensitive by construction.

Distance, composition, and $R$1 are not the only difficulties. In 4U 1636–53, only PRE bursts with long post-touchdown phases have genuinely stable radii and color factors during the early tail; bursts with rapid post-touchdown evolution likely violate the simplifying assumptions usually made in mass–radius inference (Zhang et al., 2012). Static-envelope calculations likewise indicate that the photosphere may still be modestly expanded at touchdown (Guichandut et al., 2021). A neutral reading of the PRE literature is therefore that PRE bursts remain powerful equation-of-state probes, but only a subset of events is likely to satisfy the assumptions required for precision inference.

5. Winds, superexpansion, and spectral diagnostics

A major development has been the shift from treating PRE solely as continuum evolution to treating it as wind physics with spectral diagnostics. In the 1999 superburst from 4U 1820–30, the precursor showed PRE and had a total fluence of $R$2, $R$3–$R$4 more energetic than ordinary short bursts from the same source, indicating that helium burning alone was insufficient and that shock heating likely contributed (Keek, 2012). The same analysis identified a later superexpansion phase with $R$5 km and expansion velocity $R$6, interpreted as a transient radiation-driven outflow or shell ejection (Keek, 2012).

NICER has since broadened the observed PRE regime. In 4U 1820–30, 15 bursts observed between 2017 and 2023 all showed PRE, including one superexpansion burst with $R$7 km and blackbody temperature of $R$8 keV, thirteen strong PRE bursts with $R$9 km, and one moderate PRE burst with $R_{\rm bb} = \frac{R}{f_c^2}(1+z)^{-1},$0 km (Yu et al., 2023). That work also found that the first $R_{\rm bb} = \frac{R}{f_c^2}(1+z)^{-1},$1 s of the bursts depart strongly from a single blackbody and can be described either by enhanced persistent emission attributed to Poynting–Robertson drag, an additional blackbody, or a reflection model, with the reflection model presented as the self-consistent explanation (Yu et al., 2023).

Line spectroscopy during PRE has opened a further dimension. NICER observations of 4U 1820–30 first showed narrow emission and absorption lines during PRE bursts, with co-added spectra displaying features near 1.0 keV in emission and 1.7 and 3.0 keV in absorption; the stronger PRE burst pair showed line centroids systematically blue-shifted by a factor $R_{\rm bb} = \frac{R}{f_c^2}(1+z)^{-1},$2 relative to the weaker pair (Strohmayer et al., 2019). A larger 2025 analysis then reported a 1.034 keV emission line at $R_{\rm bb} = \frac{R}{f_c^2}(1+z)^{-1},$3 significance and absorption lines at 1.64 and 3 keV at $R_{\rm bb} = \frac{R}{f_c^2}(1+z)^{-1},$4 and $R_{\rm bb} = \frac{R}{f_c^2}(1+z)^{-1},$5, respectively, in co-added maximum-radius spectra, and concluded that the observed energy shifts are consistent with the burst-driven wind model and that the 1 keV feature is likely a superposition of several narrower Fe L-shell lines (Yu et al., 11 Mar 2025). A parallel reanalysis of 12 NICER bursts from 4U 1820–303 found several significant absorption lines and confirmed the previously reported 2.97 keV feature, but did not confirm a consistent correlation between line energies and blackbody radii; instead, bursts with larger radii showed more lines and higher line strength (Barra et al., 2 Jan 2025). The combined implication is that PRE winds are spectroscopically accessible, but the relation between line centroids and continuum-inferred radius is not yet settled.

6. Methodological scope, extensions, and open questions

PRE studies are instrument-limited in highly source-dependent ways. RXTE/PCA established the classic phenomenology through dense burst samples and time-resolved spectroscopy in the 3–20 keV or 2–60 keV range (Zhang et al., 2012, Zhang et al., 2010). Swift/XRT provided soft-band PRE analyses such as the Rapid Burster case (Sala et al., 2012). NuSTAR observed a PRE burst in GRS 1741.9–2853 and used the peak flux to infer $R_{\rm bb} = \frac{R}{f_c^2}(1+z)^{-1},$6 kpc under the assumption of a pure-helium Eddington limit (Pike et al., 2021). NICER, with its soft response and lack of pile-up issues at burst peak, made possible complete tracking of strong and superexpansion phases as well as line detections (Keek et al., 2018, Yu et al., 2023, Yu et al., 11 Mar 2025). XMM-Newton and NuSTAR spectroscopy of 4U 1702–429 further showed that one NICER burst displays clear PRE with a maximum photospheric radius of $R_{\rm bb} = \frac{R}{f_c^2}(1+z)^{-1},$7 km and minimum temperature of 1.4 keV, while three XMM-Newton bursts from the same source show no PRE signatures (Mandal et al., 18 Aug 2025).

The nuclear physics of burst ignition and burning also feeds back into PRE phenomenology. A recent $R_{\rm bb} = \frac{R}{f_c^2}(1+z)^{-1},$8Mg$R_{\rm bb} = \frac{R}{f_c^2}(1+z)^{-1},$9Al measurement permitted a state-of-the-art model to reproduce light curves of the GS 1826–24 clocked burster with mean deviation $L_{\rm Edd} = \frac{4\pi G M c}{\kappa},$00 and revealed a strong correlation between the He abundance in the accreting envelope of a PRE burster and the dominance of the $L_{\rm Edd} = \frac{4\pi G M c}{\kappa},$01Mg$L_{\rm Edd} = \frac{4\pi G M c}{\kappa},$02 branch (Hu et al., 2021). This does not redefine PRE, but it does place the PRE threshold inside a broader thermonuclear network problem involving composition, $L_{\rm Edd} = \frac{4\pi G M c}{\kappa},$03-flow, and recurrence behavior.

A final extension concerns magnetars. A 2010 study argued that a PRE-like mechanism could plausibly operate in magnetar bursts, despite the different emission process, and that identifying the magnetic Eddington limit could constrain magnetic field strength and distance and, in principle, enable a measurement of gravitational redshift (Watts et al., 2010). The August 24, 2008 burst from SGR 0501+4516 was presented as consistent with this possibility, but the same work emphasized that conclusive confirmation would require more detailed radiative models (Watts et al., 2010). This remains the prudent position more generally: PRE is a mature observational classification, but the mapping from continuum fits and line features to true atmospheric radius, outflow structure, and neutron-star compactness remains model-dependent.

Taken together, the PRE literature now supports a layered view. PRE bursts are Eddington-limited thermonuclear events in which radiation pressure lifts the photosphere; they can proceed through static-envelope and wind regimes; their post-touchdown behavior encodes surface-cooling geometry; and, in favorable cases, they yield constraints on mass, radius, distance, and composition. At the same time, touchdown does not always mark a settled atmosphere, the cooling tail does not always obey $L_{\rm Edd} = \frac{4\pi G M c}{\kappa},$04, and line-rich PRE winds introduce additional structure beyond a single blackbody. These are not peripheral complications but defining features of the modern PRE problem (Zhang et al., 2012, Guichandut et al., 2021, Barra et al., 2 Jan 2025).