A metallicity sweet spot for disc fragmentation and planet formation

Abstract: Fragmentation of gravitationally unstable discs offers an alternate formation mechanism for gas giant planets and brown dwarfs on wide orbits. Metallicity plays a key role in disc evolution from the onset of gravitational instability to the formation of planets. We aim to determine the effect of metallicity on disc fragmentation and on the properties of disc-instability planets. We model gravitationally unstable discs with varying metallicity ($0.01-10 \,\rm Z_{\odot}$) using the Smoothed Particle Hydrodynamics code PHANTOM. Our simulations reveal a "sweet spot" for fragmentation at $0.3 \,\rm Z_{\odot}$, where cooling is most efficient, with fragmentation also happening less vigorously at higher and lower metallicities. However, further away from the sweet spot, fragmentation becomes more difficult and is eventually suppressed at extreme low and high metallicities ($0.01 \,\rm Z_{\odot}$ and $10 \,\rm Z_{\odot}$, respectively), where the disc cools inefficiently. Discs with metallicities close to the sweet spot form more planets per disc, faster, and with lower initial masses than fragmenting discs with higher or lower metallicities. Our results may explain the slight overabundance of wide-orbit giant planets observed around metal-poor stars; these planets may have formed via disc fragmentation.

• The paper finds that gravitationally unstable discs fragment most efficiently at 0.3 Z☉, producing 14 ± 2 protoplanets per disc and fragmenting within 1–2 kyr, while fragmentation is absent at 0.01 and 10 Z☉.
• The simulations show that cooling is maximized when the outer-disc optical depth approaches unity, with low-metallicity discs limited by weak emissivity and high-metallicity discs by photon diffusion, rather than by changes in spiral structure.
• The results suggest that disc instability may explain the prevalence of wide-orbit super-Jupiters around metal-poor stars, but dust evolution, additional molecular cooling, sink accretion, and varied disc properties must be tested before generalizing the prediction.

Motivation and context

The formation of giant planets on wide orbits remains contested between the core accretion and gravitational instability (GI) frameworks. Core accretion struggles to build cores beyond ~20 AU within disc lifetimes, and predicts a positive planet–metallicity correlation, whereas observations show that for planets with masses $\geq 4\,M_{\rm J}$ this correlation inverts: super-Jupiters on wide orbits are preferentially found around metal-poor stars (2608.18830). Carter et al. address whether metallicity regulates disc fragmentation directly, using three-dimensional SPH simulations of gravitationally unstable discs spanning $Z = 0.01$ to $10\,Z_\odot$.

Numerical method

The authors use the SPH code {\sc phantom} with the radiative cooling approximation of Lombardi et al. (2015), which improves upon the Stamatellos et al. (2007) flux-limited diffusion scheme by providing more efficient cooling. Opacities follow the Bell & Lin (1994) parameterisation, with dust and molecular opacities scaled by a factor $z$ proportional to metallicity; H$^-$, bound-free, free-free, and electron-scattering opacities are left unscaled. The fiducial system is a $0.25\,M_\odot$ disc around a $0.7\,M_\odot$ star, with disc-to-star mass ratio 0.36 chosen so that a solar-metallicity disc sits just at the fragmentation threshold. Discs extend from 1–120 AU with $5\times10^5$ SPH particles; protoplanets form as sink particles when central density reaches $10^{-3}\,\rm g\,cm^{-3}$. Multiple realisations per metallicity (10 runs each for $Z = 0.1, 0.3, 1.0$, and 20 for $Z = 0.01$0) allow statistical comparison via KS tests.

Two modelling choices deserve emphasis: dust dynamics (settling, grain growth, radial drift) are neglected, and additional low-metallicity coolants such as H$Z = 0.01$1/HD and fine-structure line emission from O I and C II are not included. Both assumptions bear directly on the low-metallicity results discussed below.

The metallicity sweet spot

Fragmentation occurs in all realisations at $Z = 0.01$2, $Z = 0.01$3, and $Z = 0.01$4, in only 75% of runs at $Z = 0.01$5, and in none at $Z = 0.01$6 or $Z = 0.01$7. The central result is non-monotonic: discs at $Z = 0.01$8 cool most efficiently, fragment earliest (within 1–2 kyr versus >1.5 kyr elsewhere), and produce the most protoplanets per disc ($Z = 0.01$9 versus $10\,Z_\odot$0, $10\,Z_\odot$1, and $10\,Z_\odot$2 at $10\,Z_\odot$3, 1.0, and 3.0 respectively). Formation-time distributions at $10\,Z_\odot$4 differ from all others with $10\,Z_\odot$5.

$10\,Z_\odot$6	Runs	Protoplanets/disc	Merged (%)	Surviving (%)	$10\,Z_\odot$7 (%)
0.1	10	$10\,Z_\odot$8	$10\,Z_\odot$9	$z$0	$z$1
0.3	10	$z$2	$z$3	$z$4	$z$5
1.0	10	$z$6	$z$7	$z$8	$z$9
3.0	20	$^-$0	$^-$1	$^-$2	$^-$3

The physical explanation is an optical-depth argument: cooling is maximised when $^-$4. At the typical fragmentation radius (~80 AU), where $^-$5, $^-$6 K, and $^-$7, this condition yields $^-$8. Higher-metallicity discs are optically thick there and cool by photon diffusion; lower-metallicity discs are optically thin and limited by local emissivity. In both regimes cooling cannot overcome stellar irradiation and PdV heating in spiral shocks, suppressing fragmentation. This result contradicts Boss (2002), who reported metallicity-independent fragmentation across $^-$9–$0.25\,M_\odot$0, but agrees with Cai et al. (2006) and Lee et al. (2025). It also disagrees with Matsukoba et al. (2022, 2023), who find fragmentation down to zero metallicity — but their models include molecular and fine-structure line cooling absent here, which may be decisive at low $0.25\,M_\odot$1.

Fourier analysis shows the spiral structure retains a dominant $0.25\,M_\odot$2 mode largely independent of metallicity, so the sweet spot reflects thermodynamics rather than differing spiral morphology.

Properties of disc-instability protoplanets

Protoplanets initially form with masses of a few $0.25\,M_\odot$3, consistent with the opacity-limited fragmentation mass $0.25\,M_\odot$4; lower-metallicity discs yield systematically lighter initial fragments ($0.25\,M_\odot$5). By contrast, final masses show no strong metallicity dependence overall, indicating that post-formation accretion, mergers, and scattering dominate the end-state mass function. An exception is $0.25\,M_\odot$6, where ~50% of surviving protoplanets have final masses $0.25\,M_\odot$7 ($0.25\,M_\odot$8) — plausibly because vigorous fragmentation distributes disc gas over more objects.

Formation radius depends on where $0.25\,M_\odot$9 lies in the disc: low-metallicity discs fragment closer in ($0.7\,M_\odot$0 AU), high-metallicity discs farther out (>120 AU). Final separations peak at 100–150 AU for objects $0.7\,M_\odot$1, while more massive objects concentrate within 100 AU after accreting in the inner disc. Migration statistics (40% inward, 60% outward) show no metallicity dependence, implying migration is driven by stochastic scattering rather than disc torques. Clump collapse timescales are a few hundred years — shorter than one orbital period even at wide separations — limiting tidal disruption, and increase with metallicity because optically thick clumps cool more slowly.

Accretion rates span $0.7\,M_\odot$2 to $0.7\,M_\odot$3, up to two orders of magnitude above estimates for PDS 70 b,c and far above WISPIT 2b. The authors note that SPH sink accretion rates are likely overestimated, so final protoplanet masses should be treated as upper limits.

Observational implications

The results offer a natural explanation for the observed excess of wide-orbit giants around sub-solar metallicity hosts: GI operates most efficiently near $0.7\,M_\odot$4, so disc-instability planet formation should peak in metal-poor environments, including the outer Galaxy. This complements the core-accretion picture, which dominates at solar and super-solar metallicities for Jupiter-mass planets. The claim that close-in and wide-orbit giant planets form through different channels is consistent with the metallicity distributions of host stars split at 50 AU separation.

Limitations and open questions

Several caveats qualify these conclusions. First, the neglect of dust evolution is significant: dust concentrates in spiral arms in self-gravitating discs, locally raising opacity and potentially shifting the sweet spot; the authors explicitly flag coupled dust-gas thermodynamic treatment as required future work. Second, missing low-temperature coolants (H$0.7\,M_\odot$5, HD, O I, C II lines) could permit fragmentation below $0.7\,M_\odot$6, reconciling these results with Matsukoba et al.; resolving this discrepancy requires simulations including those processes. Third, the single disc mass and stellar mass explored mean the sweet-spot location, derived from $0.7\,M_\odot$7 at ~80 AU, will shift with surface-density profile and irradiation temperature — whether it persists across the disc population is untested. Fourth, the interpretation regarding metal-poor environments assumes disc initial conditions do not themselves vary systematically with ambient metallicity, an assumption the authors acknowledge as uncertain. Finally, sink-particle accretion physics limits the accuracy of the reported growth rates and final masses.

Conclusion

This work establishes that disc fragmentation depends non-monotonically on metallicity, peaking at $0.7\,M_\odot$8 where the outer disc optical depth approaches unity, and being fully suppressed at $0.7\,M_\odot$9 and $5\times10^5$0. Within $5\times10^5$1–$5\times10^5$2, GI remains a viable channel for gas giants and brown dwarfs on wide orbits, with maximum efficiency at sub-solar metallicity — a prediction that aligns with the demographics of directly imaged super-Jupiters and motivates targeted observational tests in metal-poor stellar populations.