Home Who I Am What I Do Publications Projects Contact CV

Jing Lu (陆晶)

Hi, my name is Jing, pronounced exactly like "Jingle Bell" without the "-gle bell".

Jing Lu

Who I Am

DeerWhale

I am an astrophysicist studying the physics of supernovae using data-driven and machine learning approaches. By coincidence, my journey as a scientist has been closely tied to capital cities. I earned my bachelor's degree in Physics from Nanjing Normal University in China, located in Nanjing, the capital of Jiangsu Province and a historic capital of several dynasties. I then moved to the United States to pursue my PhD at Florida State University in Tallahassee, the capital of Florida. Afterward, I became a postdoctoral researcher at Michigan State University in Lansing, the capital of Michigan.

Next

What I Do

Over the past decade, I have worked on large-scale data analysis, generative modeling, deep learning, Monte Carlo methods, and Bayesian inference (primarily in scientific research on understanding the physical origin of supernovae). My work has led to 40+ publications and 1,000+ citations, and I have contributed to several open-source scientific software projects, including BYOST (developer), SNooPy (contributor), and TARDIS (maintainer).

Outside of my work, I enjoy a wide range of activities, such as board-related things (snowboarding, longboarding, paddleboarding and boardgames of course), sports evolving around hitting the ball with rackets (racquetball, tennis, badminton, table tennis), and activities that capture the beauty of nature and memory (photography, painting).

Next

Publications

A selection of the publication that I have led spanning peculiar Type Ia supernovae, near-infrared spectral templates, stripped-envelope supernova simulations, and helium detection in Type Ic supernovae. A full publication list can be found in my Google Scholar page.

ASASSN-15hy supernova image

ASASSN-15hy: An Underluminous, Red 03fg-like Type Ia Supernova

Lu et al. (2021) · ApJ 920, 107

ASASSN-15hy belongs to the peculiar "super-Chandrasekhar" 03fg-like subclass of Type Ia supernovae, yet is unusually underluminous and red compared to its groupmates. Optical and near-infrared observations reveal broad light curves, persistent carbon features, and a low silicon velocity. These properties are best explained by a "deflagration core-degenerate" scenario, the secular merger of a white dwarf with the degenerate core of an asymptotic giant branch star, requiring a low-metallicity progenitor.

Read paper →
Carnegie Supernova Project NIR spectra

Carnegie Supernova Project II: Near-infrared Spectral Diversity and Template of Type Ia Supernovae

Lu et al. (2023) · ApJ 948, 27

We present 339 near-infrared spectra of 98 Type Ia supernovae from the Carnegie Supernova Project II, the largest and most homogeneous NIR spectral dataset of its kind at the time. Using principal component analysis and Gaussian process regression, we construct a spectral template as a function of phase and light-curve shape. The new template reduces systematic K-correction uncertainties by ~90% over previous templates, with the potential of improving cosmological distance measurements that uses type Ia supernovae.

Read paper →
Stripped-envelope supernova simulations

Physics-driven Explosions of Stripped High-mass Stars: Synthetic Light Curves and Spectra with Broad Light Curves

Lu et al. (2025) · ApJ 979, 148

We compute synthetic light curves and spectra of the explosions of massive (45–120 M☉) Wolf–Rayet stars exploded with physics-driven neutrino-driven models. The resulting broad, faint light curves are marginally consistent with the ~15% of observed stripped-envelope supernovae with broad light curves. We find that even as little as 0.02 M☉ of helium produces persistent He I lines, and that standard ejecta mass estimators can overestimate the true value by up to a factor of 2.6.

Read paper →
Helium in Type Ic Supernova 2014L

Traces of Helium Detected in Type Ic Supernova 2014L

Lu et al. (2026) · ApJL 1002, L11

Type Ic supernovae are classified by the absence of helium in optical spectra, but whether helium is truly absent or merely spectroscopically hidden remains debated. We trained a deep learning model that emulates the radiative transfer code TARDIS, then apply Bayesian inference to infer the physical properties of SN Ic 2014L's ejecta providing both optical and NIR spectra. We constrain 0.018–0.020 M☉ of helium in its outer ejecta. A Bayesian statistical test confirms the spectra are inconsistent with a helium-free composition, providing direct observational evidence for hidden helium in a Type Ic supernova.

Read paper →
Next

Projects

A selection of interactive tools and visual explorations from projects that apply statistical analysis, predictive modeling, and emulation of scientific simulations.

Near-infrared Type Ia supernova spectral template explorer

Data-Driven Near-Infrared Spectral Templates

Generative Modeling · Principal Component Analysis · Gaussian Process Regression · Variational Autoencoder · Publication: Lu et al. 2023 →

I developed data-driven models to learn and reconstruct time-varying spectral patterns from observational datasets. I designed the full modeling workflow, including data preprocessing, dimensionality reduction with PCA, Gaussian Process regression (and a conditional variational autoencoder as alternative method) for generating continuous spectral templates. The main method is packaged into a open-source Python package BYOST. I evaluated the models against existing templates and quantified their impact on downstream analysis. The resulting framework reduced a key source of systematic error by approximately 90% relative to previous templates.

Check out the in-browser interactive explorer linked below.

Open interactive explorer →
Deep-learning emulator for radiative transfer simulation

Deep-Learning Emulator for Radiative Transfer Simulation

Deep-learning · DenseNet · Pytorch · Bayesian inference · Workflow · TARDIS · Publication: Lu et al. 2026 →

To infer physical properties from observations based on simulations, I developed a deep-learning emulator to replace computationally expensive radiative-transfer simulations. I built a configuration-driven workflow spanning training-data generation and processing, DenseNet-based emulator training, model validation, and integration into a broader inference pipeline. The resulting emulator accelerated spectral synthesis by roughly 100 million times while maintaining approximately 1% prediction error, enabling efficient exploration of previously impractical high-dimensional parameter spaces. I also leveraged HPC resources for generating large-scale simulation datasets and MPI to accelerate the inference process.

Check out the in-browser interactive explorer linked below.

Open interactive explorer →
Next

Contact

Reach out by email for any interest in my work. You can also request my services on Toptal, a freelance platform that connects client with screened independent talent.

Hire me on Toptal

Next

Jing Lu's CV

Last updated:

PDF preview is not available on mobile browsers.

Open CV (PDF)