Transit Photometric Analysis

• Transit photometric analysis is the process of extracting exoplanetary and stellar parameters from light curves of transit events.
• It integrates noise mitigation techniques such as wavelet denoising and Gaussian process regression to enhance measurement precision.
• Bayesian inference and analytic modeling enable detailed characterization of orbital geometry and transit timing variations.

Transit photometric analysis is the quantitative extraction of exoplanetary and stellar parameters from time-series observations of periodic stellar flux decrements—transits—caused by an exoplanetary companion occulting a portion of its host star as seen by the observer. The transit method provides direct constraints on fundamental system properties—including planet-to-star radius ratios, orbital inclination and impact parameter, scaled semimajor axis, and limb-darkening—via light-curve modeling, noise diagnostics, detrending, and Bayesian inference. Recent advances exploit both high-cadence space-based photometry, such as from TESS, and rigorous noise modeling to push precision to sub-millimagnitude scales and robustly quantify transit-timing effects and system configurations (Saha et al., 2021).

1. Photometric Data Preparation and Baseline Correction

Transit photometric analysis begins with rigorous calibration and preprocessing of light curves. For space-based surveys like TESS, this typically involves selection of the Presearch Data Conditioning Simple Aperture Photometry (PDCSAP) flux and extraction of transit segments, padding each window to capture out-of-transit baselines (Saha et al., 2021). Baseline flux removal is performed via low-order polynomial fitting—commonly linear or quadratic in time—to the out-of-transit points,

$B(t) = a_0 + a_1 t + a_2 t^2$

yielding normalized fluxes

$$F_{\rm norm}(t_i) = \frac{F_{\rm raw}(t_i)}{B(t_i)}$$

so that the out-of-transit flux is unity.

Ground-based photometry requires additional steps: construction of an optimized artificial comparison star from the ensemble of reference field stars, differential aperture photometry, and correction for airmass and extinction trends. Simultaneous polynomial detrending is typically performed as part of the transit light-curve fitting, rather than as a standalone preprocessing step (Raetz et al., 2015).

2. Noise Characterization: Wavelet Denoising and Gaussian Processes

High-precision photometric time-series data are contaminated by multiple sources of noise, both uncorrelated (e.g., shot noise, detector readout) and temporally correlated ("red" noise; e.g., instrumental systematics, stellar activity, granulation). Two techniques are prevalent for robust noise mitigation:

Wavelet Denoising: A @@@@1@@@@ (DWT) is used to decompose the light curve, and high-frequency coefficients below a universal ("VisuShrink") threshold

$\lambda_j = \sigma_j \sqrt{2\ln N}$

(σ_j: scale-dependent noise estimate; N: number of data points) are set to zero. The signal is reconstructed with reduced high-frequency noise yet retains sharp transit features (Saha et al., 2021). This process improves point-to-point scatter by ~5–10% while preserving transit morphology.

Gaussian Process Regression: Time-correlated noise is modeled as a Gaussian process (GP) with kernel

$$k(t_i, t_j) = \alpha^2 \exp\left(-\frac{|t_i - t_j|}{\tau}\right) + \sigma_w^2 \delta_{ij}$$

where α is the characteristic amplitude, τ the correlation timescale, and $\sigma_w^2$ a white-noise term. This formalism captures astrophysical (e.g., quasi-periodic stellar variability), instrumental, and residual systematic errors (Saha et al., 2021, Chakrabarty et al., 2019). GP hyperparameters and astrophysical parameters are sampled jointly to account for model covariances and return realistic uncertainties.

3. Analytic and Numerical Transit Modeling

The standard forward modeling approach is the analytic Mandel & Agol (2002) model, which describes the fractional stellar flux decrement as a function of projected separation $z(t)$, planetary radius ratio $p=R_p/R_\star$, scaled semimajor axis $a/R_\star$, impact parameter $b$, and quadratic limb-darkening coefficients $(u_1, u_2)$: $$F_{\rm tr}(t) = 1 - \Delta F(z(t); p, b, R_\star/a, u_1, u_2)$$ with

$$I(\mu) = 1 - u_1(1-\mu) - u_2(1-\mu)^2, \qquad \mu = \cos\gamma$$

where $\gamma$ is the angle from the line of sight. The model implements piecewise analytic expressions depending on the transit geometry (full, partial, or out-of-transit).

Alternative implementations such as the Giménez model use Jacobi-polynomial expansions for limb-darkening integrals, optimized for speed and multi-band fitting in software like PyTransit (Parviainen, 2015).

Long-exposure binning (e.g., Kepler’s 29.4-min cadence) smears ingress and egress, inflating uncertainties—most acutely for ingress/egress duration (by up to ×34 for Earth-size planets)—and must be accounted for by explicit integration or supersampling (Price et al., 2014).

4. Bayesian Parameter Inference and Uncertainty Estimation

Parameter inference typically employs a Markov Chain Monte Carlo (MCMC) framework, in which the likelihood combines the transit model and the noise model (including GP or wavelet terms). Uniform or weakly informative priors are adopted for geometric and noise parameters, while Gaussian priors (mean and width from model grid predictions) are usually adopted for limb-darkening coefficients to regularize poorly constrained bands (Saha et al., 2021).

Sampling is carried out with Metropolis-Hastings or affine-invariant ensemble samplers. Convergence diagnostics include the Gelman–Rubin statistic ($R<1.05$) and visual inspection of the chains. Posterior percentiles yield the median and credible intervals for all parameters.

Parameter vector: $$\theta = \{T_0,\, p,\, a/R_\star,\, b,\, u_1,\, u_2,\, \alpha,\, \tau,\, \sigma_w\}$$ Posterior intervals are quoted as 16th, 50th, and 84th percentiles.

The combination of wavelet denoising and GP regression with simultaneous fitting of geometric and noise parameters yields a factor of 2–5 improvement in measurement precision compared to legacy ground-based analyses (Saha et al., 2021, Chakrabarty et al., 2019).

5. Scientific Output: Physical and Orbital Properties

Transit photometric analysis robustly determines the planet–star radius ratio, orbital geometry, and—given stellar parameters—planetary radii, equilibrium temperatures, and densities. For example, $R_p=1.293\pm0.060\,R_J$ for KELT-7 b, with relative uncertainty better than 5%, and orbital inclination resolved to within 0.01° (Saha et al., 2021).

Improved precision in $a/R_\star$ and $b$ translates into tighter constraints on planetary density and insolation. High-cadence, high-S/N light curves allow for detailed timing analysis, transit timing variation (TTV) diagnostics, and refinement of the transit ephemeris. The workflow enables a 30–50% reduction in photometric scatter and support for dynamical modeling and atmospheric inference.

6. Best Practices, Limitations, and Overview

Critical elements for robust transit photometric analysis include:

• Removal of long- and short-term instrumental/astrophysical trends prior to modeling.
• Sequential application of wavelet denoising and GP regression for uncorrelated and correlated noise, respectively.
• Careful handling of outliers, gaps, and normalization for each transit segment.
• Adoption of physically informed priors and simultaneous, joint sampling of all relevant parameters and noise hyperparameters.
• Sufficient baseline before/after transit to constrain systematics.

Space-based photometry—free from atmospheric noise—maximizes achievable precision, but instrumental systematics and intrinsic stellar variability remain important noise sources. Even with rigorous analysis, uncertainties can be dominated by systematics for partial transits or low-S/N regimes.

In summary, contemporary transit photometric analysis follows a sequential pipeline: (i) preprocessing and normalization, (ii) wavelet-based suppression of high-frequency noise, (iii) Gaussian-process regression for correlated residuals, and (iv) simultaneous analytic transit plus noise modeling in a Bayesian parameter estimation framework. This integrated approach achieves high-fidelity measurements of planetary and orbital parameters from transit time-series, enabling maximal scientific yield from missions like TESS (Saha et al., 2021).