A Pre-Main Sequence Binary Viewed Through its Circumbinary Disk: A New Phase for KH 15D

The binary T Tauri system, KH 15D, is poised to provide unprecedented detail of gas and ices located between ∼3-5 AU of its circumbinary disk over the next few years. We analyze for the first time a complete set of ground-based time series photometry of the KH 15D system for which the 2022 October - 2025 April data have never been presented. By combining the latest I-band data with historical measurements, we show that the era of large amplitude photometric variability has ended. We define a revised photometric period of P = 48.365 ± 0.005 days, which accurately phases data over a time span of seventy-five years. We adopt magnitudes and colors for the stars that define the epochs when either star A (1995-2008) or star B (2010-2012) is the primary source of the out-of-eclipse brightness of the system. We determine for the first time the effective optical depth for the system between 2002 and 2025 assuming gray extinction. The current out-of-eclipse system color is consistent with a combination of the light from both stars, supporting the previously suggested model in which the trailing edge of the circumbinary disk is more transparent than the leading edge. We continue to see evidence for “clumps” and variable transparency along ingress and egress when the stars are probing the occulting material and suggest that observations made during these phases should permit studies of the physical and chemical nature of the planet-forming zone with the circumbinary disk.

KH 15D is a pre-main sequence binary embedded in a circumbinary (CB) disk that is truncated at ∼1 AU by the action of the binary and at ∼5 AU, possibly by a protoplanet. The CB disk is inclined to the orbital plane and warped such that it precesses as a unit on a time scale of ∼10 cubed years. The period of the binary is ∼48.6 days and its eccentric orbit is viewed nearly edge-on. Since ∼1960, due to precession, the CB disk has been oriented such that it covers various portions of the binary orbit as seen from Earth. In Epoch 1 (1960), the entire binary is visible, and the system is at its maximum brightness. In Epoch 2 (1995), the leading edge of the CB disk has advanced such that nearly all of the orbit of star B is obscured. In Epoch 3 (2010), precession has now carried the warp in the CB disk across the entire binary orbit. In Epoch 4 (2025), the warp has precessed such that the orbit of star B is now being revealed by the trailing edge. As is evident from the figure and the recent data discussed here, the era of large amplitude photometric variability caused by the opaqueness of the leading edge of the disk has ended and we have entered a new phase for the system, characterized by much lower amplitude photometric variability, as the binary is now seen through the apparently semi-transparent trailing edge of its disk.

If not occulted by its disk, KH 15D would be known as an ordinary Class 3, weak T Tauri (WTTS) binary member of NGC 2264, at a distance of 780 pc, with an inferred age of ∼2 Myr and no disk signature in the infrared or radio. Its total mass is 1.3 M⊙ and the mass ratio of the components is ∼0.86. An important feature of the projected CB disk is that its leading edge is nearly opaque and can be well-modeled as a “knife edge”, but its trailing edge seems not as sharply defined and more transparent. M. Poon et al. (2021) identify the sharp leading edge with the inner edge of the disk at ∼1 AU, and the fuzzier trailing edge with its outer edge at ∼5 AU. The apparent transition from a nearly opaque leading (inner) edge to a more transparent trailing (outer) edge means that we have an enhanced opportunity now and in the years to come to probe the physical and chemical state of the disk. Interestingly, the age of the KH 15D disk, ∼2 Myr, and location probed, ∼3-5 AU from the central binary, correspond to the time and site inferred for chondrule formation in the Solar System.

In this contribution, we present photometric observations made over seven seasons from 2018 October through 2025 April. Multiple observatories provided data, including Cerro Tololo Inter-American Observatory (CTIO), Lowell Observatory, Michael L. Britton Observatory, Liverpool Telescope, Telescope Altiplano de Granada (TAGRA), and data collected through the HOYS program. The I-band light curve of KH 15D from 1951-2025 is shown in Figure 3 and can be interpreted in light of our illustration of the system throughout the years shown in Figure 1. It is clear that the era of large amplitude photometric variability for KH 15D has ended, and will probably not begin again for hundreds of years.

To further refine the photometric period, we examined photometric data that was representative of each epoch of light curve variability. We phased the data from these epochs using periods ranging from 48.36 to 48.37 days and assessed the alignment of the “central reversal” in 2001/2002, 2009/2010, and 2024/2025 by eye. We find that a photometric period of P = 48.365±0.005 days (estimated error) centers the reversals seen at mid-eclipse (phase = 0.0 in Figure 5) and evenly splits the eclipses into two halves. The ephemeris is given by JD0 = 24560719.625 + 48.365E, where JD0 is the Julian Date associated with mid-eclipse, or phase 0, and E is the epoch (or cycle) number.

In Figure 6, darker symbols show eclipses of star A by the leading edge of the screen during the late 1990’s and early 2000’s. They are quite steep and well-modeled by a “knife-edge” cutting across the stellar surface. The more recent data, shown by lighter colored symbols, records eclipses of star B by the trailing edge of the screen. These eclipses are clearly different in shape, appearing more rounded and gradual in nature, which led A. Garcı́a Soto et al. (2020) to argue that the trailing edge is not so much of a knife edge as the leading edge, but “fuzzier”. Also note that out of eclipse, the system light is not as flat as it was in the 1990-2000’s and has generally continued to rise in brightness, although not monotonically. This suggests that the occulting screen may not be as optically thick as it was and that light from both stars is contributing to the system light.

We adopt a value of I = 14.5 mag ±0.05 for star A, I = 14.0 mag ±0.05 for star B, and I = 13.5 mag ±0.05 for the combined brightness of both stars. The V − I color measurements show that outside of eclipse (phases > 0.4) the color was quite stable, averaging V − I = 1.55 mag in the early 2000’s, when star A dominated the light at those phases, and V − I = 1.15 mag around 2010 when star B took over. The predicted value of the system color if neither star were occulted, assuming I = 14.5 mag for star A and I = 14.0 mag for star B, is V − I = 1.29 mag, which we round to 1.30 mag. In recent years the out-of-eclipse colors indicate that we are, in fact, seeing a mixture of light from both stars.

We show the optical depth through which each star would need to be viewed to account for the observed magnitude and color in each year. Note that this assumes a uniform optical depth across the star’s photosphere and is, therefore, labeled as an “effective” optical depth. Between 2002 and 2010 star A progressively “set” behind the sharp horizon of the leading edge. The system did not, however, disappear, as early models predicted because, at first, star B became progressively more visible and then star A reappeared. As can be seen in the bottom panel, we are now in a situation where the light from both stars must contribute to the system brightness near aphelion, and the effective optical depth for both has dropped to optically thin values. Figure 8 shows that optical depths of 0 to 1 could explain the magnitude and colors of the system that are observed.

During eclipses, near phase 0.0, the color behavior is different. In the early days (circa 2004), the system became bluer during eclipse, probably because of the growing importance of the light of star B to the system. By 2012, however, the system was becoming distinctly redder during eclipse, redder even than the inferred intrinsic value of star A. D. Windemuth & W. Herbst (2014) first noted and discussed this reddening and posited the existence of a third body whose presence could be detected by excess red light when the system was faint. Finally, we note that in the most recent data there is a distinct trend of reddening during eclipse that sometimes extends beyond even the inferred color of star A. The simplest interpretation may be that we are observing selective extinction from relatively small grains in this part of the disk, which, if true, supports the view that the trailing edge of the occulting screen is not fully opaque.

The most recent and detailed model of the KH 15D system is by M. Poon et al. (2021). Their best model fit to the photometric and radial velocity data results in a period of P = 48.377 days. We phased all of our data with a period of 48.377 days and found that the central reversal at mid-eclipse, the most notable feature that we have followed for decades, shifts in phase to the left of 0.0, in the opposite direction of where we see the model’s inflection point. The difference between the two periods is ∆ 0.012 days. Thus, every cycle that is computed by the model occurs later than what would be expected based on our current ephemeris. By 2020/2021, the shift in phase is approximately 0.055, which, using a period of P = 48.377 days would correspond to a shift of approximately 2.661 days. The model does not predict the depth of the minima very well in 2020-2023, nor does it account for the variable brightness of the system out-of-eclipse in some seasons. As argued in Section 3.3, the screen through which we view the binary now is not opaque, but characterized by an optical depth of ∼ 1. The scattering “halos” around each star that have been employed in the various models to explain the shape of the light curves once the star was fully behind the (assumed) opaque screen may, in fact, not be necessary. It now seems likely that we have always viewed the stars through the screen, albeit with a much higher attenuation in the earlier days (1995-2008) due to the higher opacity of the leading (inner) edge.

The KH 15D system currently presents the astronomical community with an opportunity to study the dynamics and chemistry of a solar-like binary that sits within a precessing circumbinary disk that is of planet-forming age. This circumbinary disk is modeled to extend from an inner edge at 1 AU to an outer edge at 5 AU. A near- and mid-IR spectroscopic study of the KH 15D system should enable the community to disentangle the contributions from dust, ices, and gas along the disk mid-plane. As the binary system illuminates different disk regions (disk atmosphere, mid-plane, and soot line) on timescales of days, it provides the opportunity to measure material properties across locations that are otherwise inaccessible in (near) edge-on systems. A single star can only illuminate a single line-of-sight in a disk. Alternatively, as the stars in the KH 15D system orbit their common center of mass, they illuminate different sightlines within the surrounding disk. This predictable motion provides a unique opportunity to probe multiple lines-of-sight in a single disk allowing an investigation of how properties might vary with radius and altitude, without the complication of source versus source variations. Multiple ice species could be identified simultaneously, and their relative abundances compared across different disk regions. We expect to sample the H2 O ice line between 3-5 AU.

We have shown that the photometrically derived period of P = 48.365 days accurately accounts for the observed brightness changes of the KH 15D system over seventy five years. By adopting I magnitudes that represent star A and star B (I = 14.5 mag and I = 14.0 mag, respectively), we show that the combined light of the system in I can accurately match the brightness of the system (I = 13.5 mag) when both stars are believed to be unocculted. We posit that some light from both stars has been shining through the trailing edge of the disk since star B began to emerge from behind the optically thick part of the leading edge. We show that by adopting colors for star A (V − I = 1.55 mag) and B (V − I = 1.15 mag), the current color of the system (V − I ∼1.3 mag) can be explained by a combination of starlight from each star, with occasional selective reddening. If this interpretation is correct, the scattering halos employed in the models may no longer be necessary or useful moving forward. It will be very important to continue to monitor the system photometrically to assess the brightness and colors of the stars/system to sort out some of the issues discussed here. It is possible that the stars may continue to be obscured by the circumbinary disk for decades to come, providing observers with an excellent opportunity to use the binary as a moving probe behind a possibly planet-forming portion of a circumbinary disk. This would permit studies of the physical and chemical nature of the disk if variable absorption features in the stellar spectra can be found and monitored. Space-based near-infrared spectral monitoring would be highly desirable (e.g., JWST/MIRI-MRS).

Improvements to AI Systems Based on This Paper:

1. Dynamic Occultation Modeling for Exoplanet/Protoplanetary Disk Studies

• Improvement: Train an AI to model time-varying extinction (optical depth) from binary stars behind a precessing, warped circumbinary disk, using the KH 15D data as a benchmark.

• Capability: The AI can predict light curves and color evolution for similar edge-on binaries with non-uniform, semi-transparent screens, enabling automated classification of disk geometry (knife-edge vs. fuzzy trailing edge) and extraction of physical parameters (e.g., disk inner/outer radii, warp precession rate).

2. Multi-Epoch Period Refinement via Phase-Coherent Ephemeris Fitting

• Improvement: Implement an AI that automatically searches for the optimal photometric period by minimizing phase drift of key features (e.g., mid-eclipse reversals) across decades of heterogeneous data.

• Capability: The AI can robustly refine orbital periods for any eclipsing binary with long-baseline photometry, even when data quality and cadence vary, and flag inconsistencies with model-predicted periods (as done for Poon et al. 2021).

3. Two-Component Stellar Flux Decomposition with Time-Varying Extinction

• Improvement: Develop a neural network that decomposes observed system brightness and color into contributions from two unresolved stars, each subject to independent, time-dependent optical depths.

• Capability: The AI can disentangle stellar fluxes and effective optical depths from multi-band photometry, even when both stars are partially occulted, and identify epochs where scattering halos are unnecessary (replacing complex radiative transfer models).

4. Automated Detection of Variable Transparency and Clumpiness in Occulting Material

• Improvement: Train a model on ingress/egress light curve shapes to detect non-monotonic brightness variations (clumps, variable transparency) and distinguish them from stellar variability.

• Capability: The AI can flag regions of a disk with inhomogeneous dust distribution, enabling targeted follow-up spectroscopy to probe chemical composition (e.g., water ice lines) at specific radii and altitudes.

5. Predictive Scheduling for Spectroscopic Observations of Moving Probes

• Improvement: Build an AI scheduler that uses the refined ephemeris and current optical depth trends to predict when each star will be maximally or minimally obscured, optimizing JWST/MIRI-MRS or ground-based observations.

• Capability: The AI can autonomously plan observations to capture absorption features from gas and ices along different sightlines as the binary moves, maximizing scientific yield for time-critical, space-based spectral monitoring.

6. Cross-Epoch Color–Magnitude Anomaly Detection

• Improvement: Implement an anomaly detection AI that identifies unexpected reddening during eclipses (as seen in KH 15D) and distinguishes between selective extinction from small grains and contributions from a third body.

• Capability: The AI can automatically flag unusual color behavior in any eclipsing system, prompting deeper investigation into grain size distributions or hidden companions, and refine models of disk opacity.

7. Simulation-to-Observation Transfer Learning for Precessing Disk Geometries

• Improvement: Use the KH 15D dataset to train a generative model that simulates light curves for arbitrary binary+disk configurations, then apply transfer learning to other systems with unknown parameters.

• Capability: The AI can rapidly infer disk warp angles, precession periods, and edge sharpness from sparse photometric data, accelerating the characterization of similar circumbinary disks in star-forming regions.

8. Real-Time Data Assimilation for Long-Term Monitoring

• Improvement: Create an AI pipeline that ingests new photometric observations (e.g., from HOYS or robotic telescopes) and updates the ephemeris, optical depth, and stellar flux estimates in real time.

• Capability: The AI can provide continuous, up-to-date predictions of system brightness and color, alerting astronomers to deviations from expected behavior (e.g., sudden transparency changes) and triggering immediate follow-up observations.

The binary T Tauri system, KH 15D, is poised to provide unprecedented detail of gas and ices located between about 3-5 AU of its circumbinary disk over the next few years. We analyze for the first time a complete set of ground-based time series photometry of the KH 15D system for which the 2022 October - 2025 April data have never been presented. By combining the latest I-band data with historical measurements, we show that the era of large amplitude photometric variability has ended. We define a revised photometric period of P = 48.365 plus or minus 0.005 days, which accurately phases data over a time span of seventy-five years. We adopt magnitudes and colors for the stars that define the epochs when either star A (1995-2008) or star B (2010-2012) is the primary source of the out-of-eclipse brightness of the system. We determine for the first time the effective optical depth for the system between 2002 and 2025 assuming gray extinction. The current out-of-eclipse system color is consistent with a combination of the light from both stars, supporting the previously suggested model in which the trailing edge of the circumbinary disk is more transparent than the leading edge. We continue to see evidence for ``clumps'' and variable transparency along ingress and egress when the stars are probing the occulting material and suggest that observations made during these phases should permit studies of the physical and chemical nature of the planet-forming zone with the circumbinary disk.