Khee-Gan Lee

Scientific Interests

Lyman-alpha Forest Tomography

I have spent the past decade developing the technique of Lyman-alpha forest tomography, which uses distant galaxies and quasars (z=2-3, or 10-11 billion light years away) to illuminate the foreground hydrogen gas tracing the underlying cosmic web. With the largest telescopes in the world, we can observe dense grids of these background sources closely-separated with each other which allows us to 'tomographically' reconstruct the full 3D gas distribution. This provides a unique probe of 3D large-scale structure on small scales at this distant epoch. I lead the COSMOS Lyman-Alpha Mapping And Tomography Observations (CLAMATO) Survey, which deploys this technique on the 10.3 diameter Keck-I Telescope in Hawai'i to map a region of the sky called the COSMOS Field. The following are the science highlights from CLAMATO:

• First systematic use of galaxies (instead of QSOs) as Ly-alpha forest probes (Lee+2014)

• First detection of galaxy protoclusters in Ly-a forest absorption (Lee+2016)

• Highest-redshift detection of cosmic voids (Krolewski, Lee+2017)

• Possible detection of AGN feedback in proto-ICM of galaxy protoclusters at z~2.3 (Dong, Lee+2023, 2024)

• First 3D study of galaxy properties as a function of environment (Momose, Lee+2024)

• First estimation of LBG linear bias as a function of galaxy stellar mass (Zhang, Lee+2025)

Currently, the Subaru Prime Focus Spectrograph (PFS) is conducting the 360-night Subaru Strategic Program (SSP) survey. As part of the Galaxy Evolution component, it will carry out IGM tomography over a much wider area of ~12.5 sq deg with ~15,000 background LBGs and QSOs, which will allow the identification of the cosmic web over cosmological volumes along with multi-wavelength data on foreground galaxies. 

I am one of the Co-Chairs of the Galaxy Evolution working group, and the Survey Coordinator of the PFS SSP.

My strongest research interest is in using Subaru PFS to measure the intelligence of the avian species Gallus gallus domesticus, in the context of possible correlations with their delectability. I would be very eager to hire undergrad researchers to study this topic.

Fast Radio Bursts and the Cosmic Baryon Census

Another recent interest of mine are FRBs as a probe of the cosmic baryon distribution. The frequency dispersion of the FRB signal encodes the integrated line-of-sight electron density, which probes the IGM and CGM since most cosmic baryons in these regions are ionized. With samples of localized FRBs (i.e. host galaxy redshifts) as well as spectroscopy of foreground galaxies, me and my team have devised techniques to combine all the information to disentangle the relative distribution of the diffuse cosmic baryons in the Cosmos. See this paper for more details. 

I led the FRB Line-of-Sight Ionisation Measurement From Lightcone AAOmega Mapping (FLIMFLAM) Survey, which used the 4m Anglo-Australian Telescope in Australia to map out the cosmic web in front of ~20 localized FRBs (Huang+2025). Together with coordinated observations on Keck, Gemini and SOAR, our DR1 analysis yielded the first-ever constraints on the partition of IGM and CGM baryons (Khrykin+2024). FLIMFLAM is in collaboration with the Commensal Real-Time ASKAP Fast Transient (CRAFT) and the Fast and Fortunate FRB Follow-up (F^4) teams. 

With Subaru PFS, we will map the foregrounds of higher-redshift FRBs (beyond z>0.5) and probe the redshift evolution of the cosmic baryon distribution. 

Field-Level Inference

What is field-level inference? For most cosmological or astrophysical analyses of galaxies and large-scale structure, we compute summary statistics such as galaxy luminosity/mass distributions or N-point correlation functions. While these have served well for decades, we are pushing into the precision era with cosmology (<1% precision) as well as aiming for a comprehensive understanding of galaxy evolution and other astrophysical processes. Field-level inference seeks to move beyond these summary statistics by modeling and analyzing the full underlying fields—such as the 3D density and velocity fields— from which the observed galaxies arise. This retains substantially more cosmological and astrophysical information, allow for principled uncertainty quantification, and enable joint inference of both initial conditions and relevant physical parameters. This paradigm is particularly powerful in the era of large surveys, where the richness of the data demands methods that can capture non-Gaussian structure, exploit spatial correlations at all scales, and fully leverage the predictive power of modern simulations to bridge the gap between cosmology and astrophysics.

Me and my collaborators have pioneered the application of field-level inference to spectroscopic surveys at high redshifts (z>2), such as when we reconstructed the full cosmic evolution of z~2.3 galaxy protoclusters in the COSMOS field (Ata, Lee+2022), based on pre-existing data. 

The TARDIS density reconstructions are also a form of field-level inference which we used to infer the cosmic web at z~2.3 (Horowitz, Lee+2022). I believe that this technique is the future of cosmology and astrophysics, and I am pushing from the observational side to fuse observations and simulations in unprecedented ways.