arXiv subjects

Tiancheng Sun

Publications and source records attributed to Tiancheng Sun.

At least 19 recordsLinked to original sources

arXiv2026

Host--Non-host Differences in Stellar Chemistry, Activity, and Birth Radius: Hints of Distinct Formation Environments for Earth-like Planets and Sub-Neptunes

Host-star properties provide important clues to planet formation and evolution, yet it remains unclear whether the observed differences between planet-hosting stars and stars without detected planets reflect genuine signatures of planet formation or underlying differences in stellar populations. Using a homogeneous sample of 28,383 Kepler-LAMOST-Gaia dwarf and subgiant stars, including 629 host stars with 865 planets, we compare host stars with age-mass matched non-host stars after correcting for distribution-induced matching biases. Most host-non-host differences disappear when the full planet sample is considered. However, separating planets by radius reveals distinct trends. At the high-abundance end of the [O/Fe], [Mg/Fe], and [Si/Fe] distributions, Earth-like hosts tend to be more O-rich but more Mg-poor and Si-poor than their age-mass matched non-host stars, whereas sub-Neptune hosts tend to show the opposite behavior. Sub-Neptune hosts also tend to exhibit lower chromospheric activity and smaller birth radii than comparable non-host stars. These results suggest that host-non-host differences become apparent primarily after separating planets by size and that Earth-like planets and sub-Neptunes may be associated with distinct formation environments and evolutionary pathways. We also find that hot-Jupiter hosts are tentatively more metal-rich, chromospherically active, and formed at smaller Galactic birth radii than hosts of longer-period Jupiters.

astro-ph.SR↗

arXiv2026

Online Neural Space Time Memory for Dynamic Novel View Synthesis

Online novel view synthesis from multi-view streaming videos faces a fundamental trade-off: maintaining a persistent, long-horizon memory to reconstruct temporarily occluded regions while operating under strict real-time constraints. While Test-Time Training (TTT) offers a powerful memory mechanism, standard models mandate gradient-based memory updates at every frame to adapt to the changing motion in dynamic scenes. The computational cost of heavy memory updates precludes real-time application and can lead to instability over long contexts. Given that memory updates are more demanding than memory application and video content is largely redundant, we propose to decouple the frequencies of these two processes. Our approach performs periodic memory updates while applying the memory on a per-frame basis, using cross-view attention to manage deformations between the prior memory state and the current frame. To lock in the historical context, we introduce two critical mechanisms: an auxiliary Memory Loss that forces persistent internalization of the scene, and a Memory Caching strategy that regularizes active weights against catastrophic drift. Our method demonstrates real-time, state-of-the-art performance on scenes with dynamic human motion as well as minute-scale online memorization.

cs.CV↗

arXiv2026

Insights into the Exoplanet Radius Valley from Host-Star Ages, Activity, Chemistry, and Birth Radius

The radius valley, a bimodal feature in the size distribution of close-in small exoplanets, is widely interpreted as a signature of atmospheric loss and therefore provides a key constraint on the formation and atmospheric evolution of these planets. We investigate its dependence on host-star properties using 769 planets orbiting 558 stars, for which we derive stellar ages, chromospheric activity, and Galactic birth radius, together with elemental abundances. We find that the radius valley is not fully established at ages $\sim 3$ Gyr and evolves over gigayear timescales, with its prominence strongly affected by stellar population mixing. The dependence on magnetic activity is non-monotonic: a clear valley is present even among magnetically quiet stars, while highly active systems do not show a systematically stronger depletion. The valley morphology also varies with stellar composition: the valley is strongest in metal-poor stars, weakens near solar metallicity, and partially strengthens again at the highest metallicities. In addition, the valley shows sensitivity to refractory element ratios such as [Mg/Si], while correlations with [C/O] are weaker, indicating a dependence on planetary interior structure. Our results are more consistent with a dominant role for core-powered atmospheric mass loss than with purely irradiation-driven photoevaporation. Finally, the radius valley also depends on the Galactic birth environment, with systems near the estimated solar birth radius $\sim 4.5$ kpc showing a high fraction of Earth-like planets and a well-defined bimodal structure, suggesting that the Solar System formed in a region with a well-developed Earth-sized planet population.

astro-ph.SR↗

arXiv2025

Selfi: Self Improving Reconstruction Engine via 3D Geometric Feature Alignment

Novel View Synthesis (NVS) has traditionally relied on models with explicit 3D inductive biases combined with known camera parameters from Structure-from-Motion (SfM) beforehand. Recent vision foundation models like VGGT take an orthogonal approach -- 3D knowledge is gained implicitly through training data and loss objectives, enabling feed-forward prediction of both camera parameters and 3D representations directly from a set of uncalibrated images. While flexible, VGGT features lack explicit multi-view geometric consistency, and we find that improving such 3D feature consistency benefits both NVS and pose estimation tasks. We introduce Selfi, a self-improving 3D reconstruction pipeline via feature alignment, transforming a VGGT backbone into a high-fidelity 3D reconstruction engine by leveraging its own outputs as pseudo-ground-truth. Specifically, we train a lightweight feature adapter using a reprojection-based consistency loss, which distills VGGT outputs into a new geometrically-aligned feature space that captures spatial proximity in 3D. This enables state-of-the-art performance in both NVS and camera pose estimation, demonstrating that feature alignment is a highly beneficial step for downstream 3D reasoning.

cs.CV↗

arXiv2025

Learning Efficient Fuse-and-Refine for Feed-Forward 3D Gaussian Splatting

Recent advances in feed-forward 3D Gaussian Splatting have led to rapid improvements in efficient scene reconstruction from sparse views. However, most existing approaches construct Gaussian primitives directly aligned with the pixels in one or more of the input images. This leads to redundancies in the representation when input views overlap and constrains the position of the primitives to lie along the input rays without full flexibility in 3D space. Moreover, these pixel-aligned approaches do not naturally generalize to dynamic scenes, where effectively leveraging temporal information requires resolving both redundant and newly appearing content across frames. To address these limitations, we introduce a novel Fuse-and-Refine module that enhances existing feed-forward models by merging and refining the primitives in a canonical 3D space. At the core of our method is an efficient hybrid Splat-Voxel representation: from an initial set of pixel-aligned Gaussian primitives, we aggregate local features into a coarse-to-fine voxel hierarchy, and then use a sparse voxel transformer to process these voxel features and generate refined Gaussian primitives. By fusing and refining an arbitrary number of inputs into a consistent set of primitives, our representation effectively reduces redundancy and naturally adapts to temporal frames, enabling history-aware online reconstruction of dynamic scenes. Our approach achieves state-of-the-art performance in both static and streaming scene reconstructions while running at interactive rates (15 fps with 350ms delay) on a single H100 GPU.

cs.CV↗

arXiv2025

Insights into Planet Formation from the Ages, Masses, and Elemental Abundances of Host Stars

How planetary systems form and evolve is a key question in astronomy. Revealing how host star properties, such as elemental abundances, age, and mass, differ from those of non-host stars, and how they correlate with planetary characteristics such as radius, provides new insights into the formation and evolutionary pathways of planetary systems. We determine precise ages for 18890 dwarfs and subgiants from the LAMOST-Kepler-Gaia sample with a mean age uncertainty of about 15 percent (median about 10 percent). Within the framework of Galactic chemical evolution, we find that about 86 percent of planet-hosting stars younger than 8 Gyr occupy the upper branch ([Fe/H] > -0.2) of the characteristic V-shaped age-metallicity relation of the Galactic disk. Based on guiding radii (Rg), we further infer that about 19 percent of these young hosts likely originated in the inner disk and subsequently migrated to the solar neighborhood. Among stars older than 10 Gyr, host stars tend to be more metal-rich, with nearly 59 percent having [Fe/H] > -0.2. This suggests that both young and old planet-hosting stars preferentially form in relatively metal-rich environments. However, for host stars with [Fe/H] < -0.2, we find that their metallicities are on average lower by about 0.16 dex compared to non-host stars of similar age and mass, indicating that [Fe/H] is unlikely to be the dominant factor governing planet formation in metal-poor environments. We also identify a systematic depletion of volatile elements, especially carbon, in planet hosts. Moreover, host star [Fe/H] exhibits a weak correlation with planet radius, while [alpha/Fe] primarily support the formation of small planets.

astro-ph.SR↗

arXiv2025

Potential Impact of the Sagittarius Dwarf Galaxy on the Formation of Young O-rich Stars

The Milky Way underwent significant transformations in its early history, characterised by violent mergers and satellite galaxy accretion. However, recent observations reveal notable star formation events over the past 4 Gyr, likely triggered by perturbations from the Sagittarius dwarf galaxy. Here, we present chemical signatures of this accretion event, using the [Fe/H] (metallicity) and [O/Fe] (oxygen abundance) ratios of thin-disc stars. In the normalised age-metallicity plane, we identify a discontinuous V-shape structure at z$_{\rm max}$ (maximum vertical distance from the disc plane) $<$ 0.4 kpc in the local disc, interrupted by a star formation burst between 4 and 2 Gyr ago. This event is characterised by a significant increase in oxygen abundance, resulting in a distinct [O/Fe] gradient and the formation of young O-rich stars. These stars have larger birth radii, indicating formation in the outer disc followed by radial migration to the Solar neighbourhood. Simulations of late satellite infall suggest that the passage of the Sagittarius dwarf galaxy may have contributed to the observed increase in oxygen abundance in the local disc.

astro-ph.GA↗

arXiv2024

Quark: Real-time, High-resolution, and General Neural View Synthesis

We present a novel neural algorithm for performing high-quality, high-resolution, real-time novel view synthesis. From a sparse set of input RGB images or videos streams, our network both reconstructs the 3D scene and renders novel views at 1080p resolution at 30fps on an NVIDIA A100. Our feed-forward network generalizes across a wide variety of datasets and scenes and produces state-of-the-art quality for a real-time method. Our quality approaches, and in some cases surpasses, the quality of some of the top offline methods. In order to achieve these results we use a novel combination of several key concepts, and tie them together into a cohesive and effective algorithm. We build on previous works that represent the scene using semi-transparent layers and use an iterative learned render-and-refine approach to improve those layers. Instead of flat layers, our method reconstructs layered depth maps (LDMs) that efficiently represent scenes with complex depth and occlusions. The iterative update steps are embedded in a multi-scale, UNet-style architecture to perform as much compute as possible at reduced resolution. Within each update step, to better aggregate the information from multiple input views, we use a specialized Transformer-based network component. This allows the majority of the per-input image processing to be performed in the input image space, as opposed to layer space, further increasing efficiency. Finally, due to the real-time nature of our reconstruction and rendering, we dynamically create and discard the internal 3D geometry for each frame, generating the LDM for each view. Taken together, this produces a novel and effective algorithm for view synthesis. Through extensive evaluation, we demonstrate that we achieve state-of-the-art quality at real-time rates. Project page: https://quark-3d.github.io/

cs.CV↗

arXiv2024

The temporal and spatial variations of lithium abundance in the Galactic disc

This study investigates the temporal and spatial variations in lithium abundance within the Milky Way using a sample of 22,034 main-sequence turn-off (MSTO) stars and subgiants, characterised by precise stellar ages, 3D NLTE (non-local thermodynamic equilibrium) lithium abundances, and birth radii. Our results reveal a complex variation in lithium abundance with stellar age: a gradual increase from 14 Gyr to 6 Gyr, followed by a decline between 6 Gyr and 4.5 Gyr, and a rapid increase thereafter. We find that young Li-rich stars (ages $<$ 4 Gyr, A(Li) $>$ 2.7 dex) predominantly originate from the outer disc. By binning the sample according to guiding center radius and z$_{\rm max}$, we observe that these young Li-rich stars migrate radially to the local and inner discs. In addition, the stars originating from the inner disc experienced a rapid Li enrichment process between 8 Gyr and 6 Gyr. Our analysis suggests that the age range of Li-dip stars is 4-5 Gyr, encompassing evolution stages from MSTO stars to subgiants. The Galactic radial profile of A(Li) (with respect to birth radius), as a function of age, reveals three distinct periods: 14-6 Gyr ago, 6-4 Gyr ago, and 4-1 Gyr ago. Initially, the lithium abundance gradient is positive, indicating increasing Li abundance with birth radius. During the second period, it transitions to a negative and broken gradient, mainly affected by Li-dip stars. In the final period, the gradient reverts to a positive trend.

astro-ph.GA↗

arXiv2024

Relations of rotation and chromospheric activity to stellar age for FGK dwarfs from Kepler and LAMOST

The empirical relations between rotation period, chromospheric activity, and age can be used to estimate stellar age. To calibrate these relations, we present a catalog, including the masses and ages of 52,321 FGK dwarfs, 47,489 chromospheric activity index $logR^{+}_{HK}$, 6,077 rotation period $P_{rot}$ and variability amplitude $S_{ph}$, based on data from LAMOST DR7, Kepler and Gaia DR3. We find a pronounced correlation among $P_{rot}$, age, and [Fe/H] throughout the main-sequence phase for F dwarfs. However, the decrease of $logR^{+}_{HK}$ over time is not significant except for those with [Fe/H] $<$ $-$0.1. For G dwarfs, both $P_{rot}$ and $logR^{+}_{HK}$ are reliable age probes in the ranges $\sim$ 2-11 Gyr and $\sim$ 2-13 Gyr, respectively. K dwarfs exhibit a prominent decrease in $logR^{+}_{HK}$ within the age range of $\sim$ 3-13 Gyr when the relation of $P_{rot}-τ$ is invalid. These relations are very important for promptly estimating the age of a vast number of stars, thus serving as a powerful tool in advancing the fields of exoplanet properties, stellar evolution, and Galactic-archaeology.

astro-ph.SR↗

arXiv2024

Detection of Solar-like Oscillations in Sub-giant and Red Giant Stars Using 2-minute Cadence TESS Data

Based on all 2-minute cadence $TESS$ light curves from Sector 1 to 60, we provide a catalog of 8,651 solar-like oscillators, including frequency at maximum power ($ν_{\rm max}$, with its median precision, $σ$=5.39\%), large frequency separation ($Δν$, $σ$=6.22\%), seismically derived masses, radii, and surface gravity. In this sample, we have detected 2,173 new oscillators and added 4,373 new $Δν$ measurements. Our seismic parameters are consistent with those from $Kepler$, $K2$, and previous $TESS$ data. The median fractional residual in $ν_{\rm max}$ is $1.63\%$ with a scatter of $14.75\%$, and in $Δν$ it is $0.11\%$ with a scatter of $10.76\%$. We have detected 476 solar-like oscillators with $ν_{\rm max}$ exceeding the $Nyquist$ frequency of $Kepler$ long-cadence data during the evolutionary phases of sub-giant and the base of the red-giant branch, which provide a valuable resource for understanding angular momentum transport.

astro-ph.SR↗

arXiv2023

Text2Immersion: Generative Immersive Scene with 3D Gaussians

We introduce Text2Immersion, an elegant method for producing high-quality 3D immersive scenes from text prompts. Our proposed pipeline initiates by progressively generating a Gaussian cloud using pre-trained 2D diffusion and depth estimation models. This is followed by a refining stage on the Gaussian cloud, interpolating and refining it to enhance the details of the generated scene. Distinct from prevalent methods that focus on single object or indoor scenes, or employ zoom-out trajectories, our approach generates diverse scenes with various objects, even extending to the creation of imaginary scenes. Consequently, Text2Immersion can have wide-ranging implications for various applications such as virtual reality, game development, and automated content creation. Extensive evaluations demonstrate that our system surpasses other methods in rendering quality and diversity, further progressing towards text-driven 3D scene generation. We will make the source code publicly accessible at the project page.

cs.CV↗

arXiv2023

Neural Free-Viewpoint Relighting for Glossy Indirect Illumination

Precomputed Radiance Transfer (PRT) remains an attractive solution for real-time rendering of complex light transport effects such as glossy global illumination. After precomputation, we can relight the scene with new environment maps while changing viewpoint in real-time. However, practical PRT methods are usually limited to low-frequency spherical harmonic lighting. All-frequency techniques using wavelets are promising but have so far had little practical impact. The curse of dimensionality and much higher data requirements have typically limited them to relighting with fixed view or only direct lighting with triple product integrals. In this paper, we demonstrate a hybrid neural-wavelet PRT solution to high-frequency indirect illumination, including glossy reflection, for relighting with changing view. Specifically, we seek to represent the light transport function in the Haar wavelet basis. For global illumination, we learn the wavelet transport using a small multi-layer perceptron (MLP) applied to a feature field as a function of spatial location and wavelet index, with reflected direction and material parameters being other MLP inputs. We optimize/learn the feature field (compactly represented by a tensor decomposition) and MLP parameters from multiple images of the scene under different lighting and viewing conditions. We demonstrate real-time (512 x 512 at 24 FPS, 800 x 600 at 13 FPS) precomputed rendering of challenging scenes involving view-dependent reflections and even caustics.

cs.GR↗

arXiv2023

Age of FGK Dwarfs Observed with LAMOST and GALAH: Considering the Oxygen Enhancement

Varying oxygen abundance could impact the modeling-inferred ages. This work aims to estimate the ages of dwarfs considering observed oxygen abundance. To characterize 67,503 LAMOST and 4,006 GALAH FGK-type dwarf stars, we construct a grid of stellar models which take into account oxygen abundance as an independent model input. Compared with ages determined with commonly-used $α$-enhanced models, we find a difference of $\sim$9% on average when the observed oxygen abundance is considered. The age differences between the two types of models are correlated to [Fe/H] and [O/$α$], and they are relatively significant on stars with [Fe/H] $\lesssim$ -0.6 dex. Generally, varying 0.2 dex in [O/$α$] will alter the age estimates of metal-rich (-0.2 $<$ [Fe/H] $<$ 0.2) stars by $\sim$10%, and relatively metal-poor (-1 $<$ [Fe/H] $<$ -0.2) stars by $\sim$15%. Of the low-O stars with [Fe/H] $<$ 0.1 dex and [O/$α$] $\sim$ -0.2 dex, many have fractional age differences of $\geq$ 10%, and even reach up to 27%. The fractional age difference of high-O stars with [O/$α$] $\sim$ 0.4 dex reaches up to -33% to -42% at [Fe/H] $\lesssim$ -0.6 dex. We also analyze the chemical properties of these stars. We find a decreasing trend of [Fe/H] with age from 7.5-9 Gyr to 5-6.5 Gyr for the stars from the LAMOST and GALAH. The [O/Fe] of these stars increases with decreasing age from 7.5-9 Gyr to 3-4 Gyr, indicating that the younger population is more O-rich.

astro-ph.SR↗

arXiv2023

Characterising abundance-age relations of GALAH stars using oxygen-enhanced stellar models

Main Sequence Turn-off stars (MSTO) and subgiant stars are good tracers of galactic populations. We present a study of 41,034 MSTO and subgiant stars from the GALAH survey. Using a grid of stellar models that accounts for the variation of O abundances, we determine their ages with a median age uncertainty of $\sim$9.4 per cent. Our analysis reveals that the ages of high-O stars based on O-enhanced models (OEM models) are smaller than those determined with $α$-enhanced models, resulting in a mean fractional age difference of -5.3 per cent at [O/$α$] = 0.2 and -11.0 per cent at [O/$α$] = 0.4. This age difference significantly impacts the age distribution of thick disc and halo stars, leading to a steeper downward trend in the [Fe/H]-age plane from 8 Gyr to 14 Gyr, indicating a shorter formation time-scale and a faster chemical-enhanced history for these populations. We confirm the V-shape of the normalized age-metallicity distribution $p$($τ$$\mid$[Fe/H]) of thin disc stars, which is presumably a consequence of the second gas infall. Additionally, we find that the halo stars in our sample can be divided into two sequences, a metal-rich sequence (Splash stars) and a metal-poor sequence (accreted stars), with the Splash stars predominantly older than 9 Gyr and the accreted halo stars older than 10 Gyr. Finally, we observe two distinct sequences in the relations between various chemical abundances and age for disc stars, namely a young sequence with ages $<$ $\sim$8 Gyr and an old sequence with ages $>$ $\sim$8 Gyr.

astro-ph.GA↗

arXiv2021

NeLF: Neural Light-transport Field for Portrait View Synthesis and Relighting

Human portraits exhibit various appearances when observed from different views under different lighting conditions. We can easily imagine how the face will look like in another setup, but computer algorithms still fail on this problem given limited observations. To this end, we present a system for portrait view synthesis and relighting: given multiple portraits, we use a neural network to predict the light-transport field in 3D space, and from the predicted Neural Light-transport Field (NeLF) produce a portrait from a new camera view under a new environmental lighting. Our system is trained on a large number of synthetic models, and can generalize to different synthetic and real portraits under various lighting conditions. Our method achieves simultaneous view synthesis and relighting given multi-view portraits as the input, and achieves state-of-the-art results.

cs.CV↗

arXiv2021

Neural Light Transport for Relighting and View Synthesis

The light transport (LT) of a scene describes how it appears under different lighting and viewing directions, and complete knowledge of a scene's LT enables the synthesis of novel views under arbitrary lighting. In this paper, we focus on image-based LT acquisition, primarily for human bodies within a light stage setup. We propose a semi-parametric approach to learn a neural representation of LT that is embedded in the space of a texture atlas of known geometric properties, and model all non-diffuse and global LT as residuals added to a physically-accurate diffuse base rendering. In particular, we show how to fuse previously seen observations of illuminants and views to synthesize a new image of the same scene under a desired lighting condition from a chosen viewpoint. This strategy allows the network to learn complex material effects (such as subsurface scattering) and global illumination, while guaranteeing the physical correctness of the diffuse LT (such as hard shadows). With this learned LT, one can relight the scene photorealistically with a directional light or an HDRI map, synthesize novel views with view-dependent effects, or do both simultaneously, all in a unified framework using a set of sparse, previously seen observations. Qualitative and quantitative experiments demonstrate that our neural LT (NLT) outperforms state-of-the-art solutions for relighting and view synthesis, without separate treatment for both problems that prior work requires.

cs.CV↗

arXiv2020

Light Stage Super-Resolution: Continuous High-Frequency Relighting

The light stage has been widely used in computer graphics for the past two decades, primarily to enable the relighting of human faces. By capturing the appearance of the human subject under different light sources, one obtains the light transport matrix of that subject, which enables image-based relighting in novel environments. However, due to the finite number of lights in the stage, the light transport matrix only represents a sparse sampling on the entire sphere. As a consequence, relighting the subject with a point light or a directional source that does not coincide exactly with one of the lights in the stage requires interpolation and resampling the images corresponding to nearby lights, and this leads to ghosting shadows, aliased specularities, and other artifacts. To ameliorate these artifacts and produce better results under arbitrary high-frequency lighting, this paper proposes a learning-based solution for the "super-resolution" of scans of human faces taken from a light stage. Given an arbitrary "query" light direction, our method aggregates the captured images corresponding to neighboring lights in the stage, and uses a neural network to synthesize a rendering of the face that appears to be illuminated by a "virtual" light source at the query location. This neural network must circumvent the inherent aliasing and regularity of the light stage data that was used for training, which we accomplish through the use of regularized traditional interpolation methods within our network. Our learned model is able to produce renderings for arbitrary light directions that exhibit realistic shadows and specular highlights, and is able to generalize across a wide variety of subjects.

cs.GR↗