How Little Red Dots Turn Blue: Evolution of Kelvin-Helmholtz Contracting Black Hole Envelopes and Observational Diagnostics

abstract · abstract

James Webb Space Telescope (JWST) observations have revealed compact, red-colored, broad-line emitters known as little red dots (LRDs), while a bluer counterpart population has been referred to as little blue dots (LBDs). Their physical connection remains unclear: they may represent different evolutionary stages of the same phenomenon, distinct populations, or different phases of the broader evolution of early galaxies and their black holes (BHs). We propose a model in which a rapidly growing BH is embedded in a geometrically thick gaseous envelope accompanied by concurrent nuclear star formation. The model reproduces the observed spectral energy distributions (SEDs) of LRDs and LBDs, and naturally connects the two phases through Kelvin-Helmholtz contraction of the envelope and exposure of the nuclear accretion disk, as the gas supply declines due to external stellar activity or feedback. We predict the corresponding evolutionary tracks in the UV-optical slope ($β_{\rm UV}$-$β_{\rm opt}$) plane which bridge the apparent gap between LRDs and LBDs. These evolutionary tracks start from the LRD regime and consistently exhibit a red excursion that extends toward $(β_{\rm UV},β_{\rm opt})\sim(2,-1)$ before turning toward the LBD regime. This feature is largely insensitive to variations in the model parameters that govern the envelope structure, providing a robust signature of the transitional population. The transition lasts $\sim5$-$50\,\mathrm{Myr}$, depending primarily on the BH feedback efficiency and the gas-retention efficiency of the envelope. The abundance ratio of transitional systems to LRDs therefore provides a constraint on BH feedback and gas retention during the rapid growth of early supermassive BHs.

S6: VI Conclusion / block 1 · html

We use a time-dependent model of an optically thick gaseous envelope surrounding a massive BH, in which radiative cooling, declining ISM inflow, BH accretion, and accretion-powered feedback are coupled, and translate the resulting evolution into mock SEDs including a coevolving nuclear stellar population. Our main conclusions are as follows.

S6: VI Conclusion / block 1 · html

Kelvin–Helmholtz contraction of a gas-fed BHE can naturally transform an LRD into an LBD through an observable transitional phase. Sustained inflow initially builds a cool, extended envelope whose photospheric temperature is regulated near the Hayashi track. As supernova feedback reduces the gas supply, the envelope can no longer be replenished and begins to contract. Its photosphere shrinks and heats until the accretion disk emerges, producing a continuous sequence from an LRD to a transitional source and finally to an LBD. The proposed AGN HR diagram provides a complementary view of this sequence, connecting a Hayashi-track-regulated BHE to an exposed accretion disk through envelope contraction.

S6: VI Conclusion / block 1 · html

The LRD–LBD transition follows a robust path in the ( β UV , β opt ) (\beta_{\rm UV},\beta_{\rm opt}) plane. As the blackbody peak moves through the UV, the tracks make a characteristic red excursion near ( β UV , β opt ) ∼ ( 2 , − 1 ) (\beta_{\rm UV},\beta_{\rm opt})\sim(2,-1) before turning blue and entering the LBD region. The same trajectory passes near BBQSORS, whose ∼ 10 4 ​ K \sim 10^{4}\,{\rm K} blackbody-like continuum is consistent with an envelope caught during contraction. Although the track shape varies little across the feedback and envelope-retention parameters explored, these parameters strongly regulate the evolutionary timing.

S6: VI Conclusion / block 1 · html

The abundance of transitional systems relative to LRDs can constrain BH physics. The LRD and transitional lifetimes span approximately ≲ 5 \lesssim 5 – 50 50 Myr across the model grid. In our fiducial case, the system remains an LRD for ∼ 15 \sim 15 Myr and then spends ∼ 20 \sim 20 Myr in the transitional phase, with both timescales increasing toward higher feedback and gas-retention efficiencies. Under a statistical steady-state assumption, the relative abundance of transitional systems and LRDs directly traces their relative lifetimes, providing an observational constraint on the feedback efficiency and gas retention that govern BHE evolution.

S6: VI Conclusion / block 1 · html

BH growth during the transition produces a systematic mass offset between LRDs and LBDs. Intermediate feedback efficiencies produce maximum LRD-phase BH masses of ∼ 10 6 ​ M ⊙ \sim 10^{6}{\,\rm M_{\odot}} , while the BH can reach ∼ 10 6.5 ​ M ⊙ \sim 10^{6.5}{\,\rm M_{\odot}} by the end of the transitional phase. This corresponds to a characteristic increase in BH mass by a factor of ∼ 3 \sim 3 – 5 5 between the two stages, implying that the relative abundance of LBDs should increase toward higher BH masses.

S6: VI Conclusion / block 1 · html

In conclusion, by combining a minimal set of physical ingredients, our model reproduces the observed diversity of broad-line emitters within a coherent evolutionary framework. This picture makes clear and testable observational predictions. Measurements of the number densities, continuum slopes, photospheric temperatures, and BH masses of transitional sources can directly test the proposed evolutionary sequence and constrain the efficiency of feedback during rapid BH growth.

S6: VI Conclusion / block 1 · html

Our model also provides a simple and readily implementable subgrid prescription for semi-analytic and cosmological simulations, as it requires only the ISM inflow history onto the nuclear region once the model parameters are specified. Coupling this prescription to a more realistic, time-dependent gas supply from galactic scales would allow the duration and possible recurrence of the subsequent LBD phase to be followed naturally, including the effects of long-term gas replenishment and repeated accretion episodes. Such an implementation would also follow the coevolution of the host galaxy, providing an additional channel for the disappearance of the LRD signature [ 107 , 9 , e.g.,] that is not captured in the present model.

S6: VI Conclusion / block 1 · html

A further important step will be to extend the current framework by incorporating a self-consistent hydrodynamic treatment of the BHE structure, or ultimately to follow the BHE evolution directly with dedicated radiation-hydrodynamic simulations. Such calculations would enable a more realistic treatment of line formation and departures from spherical symmetry.

A3: Appendix C Spectral dependence of the absorption and stellar treatment / block 1 · html

In the main text, we adopt a simple prescription in which the Balmer absorption is smoothly weakened as the system evolves. Here, we test the sensitivity of our results to this phenomenological treatment. We consider two main uncertainties: the smoothness of the transition, characterized by Δ sm \Delta_{\rm sm} , and the functional form of the transmission curve, characterized by the hydrogen column density N H N_{\rm H} . Throughout this comparison, we adopt the fiducial BH parameters, η fbk = 0.01 \eta_{\rm fbk}=0.01 and f env = 0.5 f_{\rm env}=0.5 .

S7.SS1: 7.1 Hydrogen broad-line emission / block 1 · html

Together, the large decrement of the broad lines and the small decrement of the narrow lines have been interpreted as evidence for localized dust near the BLR in the AGN ( Nikopoulos et al., 2025 , e.g.,) , or the result of collisional excitation as the primary emission mechanism of the broad lines ( Ji et al., 2026b ; Torralba et al., 2025 ; Torralba et al., 2026 , e.g.,) . Case-B recombination ratios combined with known dust-attenuation curves have been ruled out in one case ( D’Eugenio et al., 2025c ; Nikopoulos et al., 2025 ) , based on the line ratios of broad H α \alpha , H β \beta , and H γ \gamma . However, observations with the necessary SNR are very few, hence it is unclear how this result generalizes to the wider population. It is clear, nevertheless, that the Blue Rosetta and quasars in general have Balmer decrements much closer to the Case-B values see Übler et al., 2023 for GS-3073, while Dong et al., 2008 find a H α \alpha /H β \beta ratio of 3.06 and 10% scatter for SDSS Seyfert 1 and quasars . However, the apparent agreement between the observed H α \alpha /H β \beta ratios of blue AGN and the Case-B value may be deceptive, since these ratios may occur even in optically thick regimes ( Ruff et al., 2012 ) , a scenario supported by several lines of evidence ( Glikman et al., 2006 ; Ruff et al., 2012 , including the low Pa α \alpha /Pa β \beta flux ratio;) . A key difference is that the H α \alpha /Pa β \beta ratios in LRDs span a broad range of 3–12 ( de Graaff et al., 2025a ) , while the BLR models of Ruff et al. (2012) remain confined to H α \alpha /Pa β ≳ 10 \beta\gtrsim 10 .

S7.SS9: 7.9 Lack of variability / block 1 · html

Table 1 : Line widths of hydrogen, helium and [O iii ] lines in GS-3073, based on the purely Gaussian emission line model described in Sect. 6 , for both single and total components. For comparison, the results of exponential parametrization (Sect. 6.1 ) are also displayed.

S2.SS2: II.2 Analytical Calculation / block 1 · html

for f c = 0.8 f_{c}=0.8 , which is the empirical result found in [ 26 ] . The mass accretion rate onto the central VMS is estimated as ∼ 3 % \sim 3\% of the star formation rate in the stellar cluster. This result holds across a wide range of stellar densities and metallicities and is the outcome of mass segregation and binary-induced collisions in the evolution of a compact cluster.

S3.SS2: III.2 Analytical Calculation: Results / block 1 · html

To perform the calculation, we need to estimate the star formation rate and the metallicity of a typical LRD. Regarding the former, estimates for the LRDs vary widely. For example, [ 36 ] estimate values between ∼ 12 ​ M ⊙ ​ yr − 1 \sim 12\,\,{\rm M_{\odot}}\,\mathrm{yr}^{-1} and ∼ 49 ​ M ⊙ ​ yr − 1 \sim 49\,\,{\rm M_{\odot}}\,\mathrm{yr}^{-1} . [ 75 ] obtain values between ∼ 11 ​ M ⊙ ​ yr − 1 \sim 11\,\,{\rm M_{\odot}}\,\mathrm{yr}^{-1} and ∼ 30 ​ M ⊙ ​ yr − 1 \sim 30\,\,{\rm M_{\odot}}\,\mathrm{yr}^{-1} , although instantaneous values may reach ∼ 150 ​ M ⊙ ​ yr − 1 \sim 150\,\,{\rm M_{\odot}}\,\mathrm{yr}^{-1} . [ 43 ] infer values in an ultraluminous LRD at z = 4.47 z=4.47 that are significantly higher. Given such a range of estimates, we assume a conservative value of ∼ 30 ​ M ⊙ ​ yr − 1 \sim 30\,\,{\rm M_{\odot}}\,\mathrm{yr}^{-1} . Regarding the metallicity, estimates from [ 36 ] and [ 75 ] provide values between 6 % ​ Z ⊙ 6\%\,\rm Z_{\odot} and 32 % ​ Z ⊙ 32\%\,\rm Z_{\odot} , while [ 43 ] provides a value of 10 % ​ Z ⊙ 10\%\,\rm Z_{\odot} . Hence, we reasonably assume a reference value of Z = 10 % ​ Z ⊙ Z=10\%\,\rm Z_{\odot} . From the star formation rate, we calculate a constant accretion rate of M ˙ acc ∼ 1 ​ M ⊙ ​ yr − 1 \dot{M}_{\rm acc}\sim 1\,\,{\rm M_{\odot}}\,\mathrm{yr}^{-1} .

S4: 4 Summary / block 1 · html

We have conducted a systematic search for low-redshift analogues to the high-redshift LRD population within the DESI DR1 spectroscopic survey. Our search identified five objects at 0.2 < z < 0.4 0.2<z<0.4 that share the defining characteristics of the LRD. Our findings are summarized below.

S4: 4 Summary / block 1 · html

Unlike high-redshift LRDs, our low-redshift analogues are unambiguously classified as AGN by BPT/VO diagnostics. This suggests they reside in more chemically mature, relatively metal-rich host galaxies.

S4: 4 Observational Constraints on LRD / block 1 · html

Multi-wavelength observations resulting in non-detections have extended to an increasingly larger sample of LRD. Studying 434 LRD in the JWST COSMOS-Web survey, Akins et al. (2025b) find no detections, even when stacking, in X-ray, mid-IR, far-IR/submillimeter, and radio bands. Their non-detections are consistent with prior work of smaller samples that found LRD are not detected, or quite weakly emitting, in X-ray ( Ananna et al., 2024 ; Yue et al., 2024 ; Maiolino et al., 2025 ) and in radio ( Mazzolari et al., 2024 ; Perger et al., 2025 ; Gloudemans et al., 2025 ) .

S6: 6 Discussion and Conclusion / block 1 · html

Despite recent progress, many caveats surround the work performed to-date attempting to identify the nature of individual LRD. Many LRD are X-ray and radio quiet even in stacks, which could be consistent with heavy obscuration and low coronal efficiency (as expected from super-Eddington accretion models; Secunda et al. 2025 ) or purely stellar solutions; stellar templates fit a large fraction or LRD with mid-IR data included ( Hainline et al., 2025 ) . Radio and X-ray constraints specific to compact high redshift sources show that non detections do not rule out accretion for the luminosities expected from thick trapping dominated flows, which raises the need for sensitive follow up tuned to low efficiency spectra rather than relying on shallow stacks ( Latif et al., 2025 ) . Deeper observations in radio bands may uncover a tail of detections if even a minority of LRD host jets or compact cores, which provides a near-term observational test of the mixed population picture ( Latif and Whalen, 2025 ) .

S3.SS1: 3.1 A low covering solution / block 1 · html

Figure 1 compares the resulting forward-model predictions with the observed UV LF measurements. The parent BLAGN LF is described by a fixed double power law (blue solid curve), while the predicted LRD LF (green dashed curve) is obtained by applying orientation-dependent dust interception and the LRD color selection to this parent population. The model reproduces the observed normalization and overall shape of the LRD LF over − 21 ≲ M 1500 ≲ − 17 -21\lesssim M_{1500}\lesssim-17 . As shown in the lower panel, it also predicts a strongly luminosity-dependent apparent LRD fraction, Φ LRD / Φ BLAGN \Phi_{\rm LRD}/\Phi_{\rm BLAGN} , rising from below 1 % 1\% at M 1500 ∼ − 22 M_{1500}\sim-22 to a maximum of ∼ 20 % \sim 20\% near M 1500 ∼ − 19 M_{1500}\sim-19 , with the subsequent faint-end behavior depending on the extrapolated BLAGN LF and the magnitude range sampled in the Monte Carlo draw. This behavior reflects the combined effects of a dust covering factor that declines toward high luminosities and a color-selection efficiency that depends on luminosity, inclination, and attenuation. The same mock catalog predicts a substantially larger apparent LRD/BLAGN ratio when the population is selected in the rest-frame optical rather than in the UV; this additional wavelength dependence is discussed below.

S4: 4 Summary / block 1 · html

The model reproduces the observed LRD LF using a modest luminosity-dependent dust covering factor, whose average value is of order ⟨ C dust ⟩ ≃ 0.2 \langle C_{\rm dust}\rangle\simeq 0.2 , together with characteristic cloud extinctions of a few magnitudes. These parameters are not independently fixed: the fit shows an elongated degeneracy in which lower cloud extinctions can be compensated by larger dust covering factors, while larger extinctions require a smaller intercepted fraction. The predicted apparent LRD/BLAGN fraction is strongly luminosity dependent, rising from a few percent at the bright end to ∼ 20 % \sim 20\% near M 1500 ≃ − 19 M_{1500}\simeq-19 .

abstract · abstract

Recent James Webb Space Telescope (JWST) observations have revealed a population of sources with a compact morphology and a characteristic `v-shaped' continuum, namely blue at rest-frame $λ<4000$A and red at longer wavelengths. The nature of these sources, called `little red dots' (LRDs), is still debated, as it is unclear if they host active galactic nuclei (AGN) and their number seems to drastically drop at z<4. We take advantage of the $63 °^2$ covered by the quick Euclid Quick Data Release (Q1) to extend the search for LRDs to brighter magnitudes and to lower redshifts than what has been possible with JWST. The selection is performed by fitting the available photometric data (Euclid, the Spitzer Infrared Array Camera (IRAC), and ground-based $griz$ data) with two power laws, to retrieve both the rest-frame optical and UV slopes consistently over a large redshift range (i.e, z<7.6). We exclude extended objects and possible line emitters, and perform a careful visual inspection to remove any imaging artefacts. The final selection includes 3341 LRD candidates at z=0.33-3.6, with 29 detected also in IRAC. The resulting rest-frame UV luminosity function, in contrast with previous JWST studies, shows that the number density of LRD candidates increases from high-z down to z=1.5-2.5 and decreases at lower z. However, less evolution is apparent focusing on the subsample of more robust LRD candidates having IRAC detections, which however has low statistics and limited by the IRAC resolution. The comparison with previous quasar (QSO) UV luminosity functions shows that LRDs are not the dominant AGN population at z<4 and $M_{\rm UV}<-21$. Follow-up studies of these LRD candidates are pivotal to confirm their nature, probe their physical properties and check for their compatibility with JWST sources, given the different spatial resolution and wavelength coverage of Euclid and JWST.

S2.SS2: 2.2 Energy feedback / block 1 · html

In conclusion, when the outflow reaches observable regions, various tensions will likely arise, suggesting that outflows are a “no-go.” These feedback problems arise mainly because the BH mass is unusually large compared to the galaxy mass (see Figure 1 ). In other words, this is a common issue in models with overmassive BHs ( Inayoshi et al., 2022 ; Hu et al., 2022b ) . Note that simply reducing the BH mass while keeping the luminosity fixed does not solve the issue, as it only increases the super-Eddington ratio m ˙ \dot{m} (and in fact makes things even worse by increasing ξ \xi ).

S3: 3 AGN BH envelope / block 1 · html

A BH envelope could also form in direct-collapse models. As a result of the collapse of a supermassive star, a seed BH may form at the centre of a gaseous envelope. This scenario has been discussed in the context of quasi-stars ( Begelman et al., 2006 ; Begelman et al., 2008 ) .

S2.SS1: 2.1 Dark Star Candidates Found / block 1 · html

Table 1 : Summary of best fit parameters for the four SMDSs spectroscopic candidates found in Ilie et al. (2025) . ∗

S2.SS1: 2.1 Dark Star Candidates Found / block 1 · html

Refs. Zhang et al. (2024) ; Ilie et al. (2025) ; Mahmud et al. (2025) ; Siddiqa et al. (2025) offer further details on Dark Stars and the potential to detect them (for a brief summary see Appendix B ; for a review see Freese et al. (2016) ). As JWST continues to discover more high redshift objects, some of them lensed to high magnification, it will become possible for future spectroscopic studies with the low-resolution spectrograph on JWST’s Mid-Infrared Instrument (MIRI) and the Near Infrared Spectrograph (NIRSpec) to differentiate between SMDS and galaxy interpretations ( Ilie et al., 2023 ) . The unambiguous discovery of even a single dark star, based on smoking gun spectral signatures such as a Helium absorption features (e.g. 1640 Å, 3 3 3 We note here that accretion onto Direct Collapse Black Holes (DCBHs) is expected to produce strong nebular emission due to Helium-II at the same 1640 ​ Å 1640~\mathrm{\AA} wavelength ( Johnson et al., 2010 ) . or 2511 Å) would constitute (indirect) detection of Dark Matter. Additionally, it would usher a new era in astronomy, where Dark Matter parameters estimations are extracted from observable properties of Dark Stars.