Factor analysis of the spectral and time behavior of long GRBs

A sample of 197 long BATSE GRBs is studied statistically. In the sample 11 variables, describing for any burst the time behavior of the spectra and other quantities, are collected. The application of the factor analysis on this sample shows that five factors describe the sample satisfactorily. Both the pseudo-redshifts coming from the variability and the Amati-relation in its original form are disfavored.

💡 Research Summary

The paper presents a statistical investigation of a sample of 197 long-duration gamma‑ray bursts (GRBs) detected by the Burst and Transient Source Experiment (BATSE). For each burst the authors compiled eleven observational quantities that capture both temporal behavior and spectral characteristics. The temporal set includes the classic duration measure T90, peak flux, total fluence, a variability index, and spectral lag. The spectral set comprises the peak energy (Epeak) of the νFν spectrum, the low‑energy photon index (α), the high‑energy photon index (β), and hardness ratios derived from BATSE’s four energy channels. By assembling these variables the authors created a high‑dimensional data matrix suitable for multivariate analysis.

To uncover the underlying structure of the data, the authors applied factor analysis (FA), a technique that seeks a smaller number of latent factors that can reproduce the observed correlations among the measured variables. Prior to FA, they examined the correlation matrix, verified approximate normality, and performed a principal‑component analysis (PCA) to estimate the number of factors. The scree plot, Kaiser’s eigenvalue‑greater‑than‑one rule, and a chi‑square goodness‑of‑fit test all converged on five factors as the optimal solution. These five factors together explain roughly 78 % of the total variance, indicating that they capture the dominant patterns in the dataset while discarding noise and redundant information.

Interpretation of the factor loadings reveals a physically meaningful decomposition:

1. Energy‑Scale Factor – High loadings from fluence, peak flux, and T90. This factor reflects the overall energetics and duration of the burst, essentially a measure of the total radiated energy (modulo distance).

2. Spectral‑Hardness Factor – Dominated by Epeak and hardness ratios. It encapsulates the spectral shape, i.e., how “hard” or “soft” the burst is, independent of its total energy.

3. Temporal‑Structure Factor – Strong contributions from the variability index and spectral lag. This factor characterizes the internal temporal complexity, such as how rapidly the light curve fluctuates and the relative timing between high‑ and low‑energy photons.

4. Low‑Energy Spectral Index Factor – Primarily associated with the low‑energy photon index α, indicating variations in the low‑energy slope of the Band function.

5. High‑Energy Spectral Index Factor – Primarily associated with the high‑energy photon index β, capturing variations in the high‑energy tail.

The separation of the spectral indices into distinct factors suggests that the low‑ and high‑energy slopes vary independently across the sample, a nuance that would be missed in analyses that treat the Band function as a single entity.

Having established this factor structure, the authors examined two widely used empirical relations that aim to turn GRBs into distance indicators. The first is the variability‑based pseudo‑redshift method, which assumes that more variable bursts are intrinsically more luminous and thus can be used to infer redshift. In the factor solution, the variability index loads heavily on the Temporal‑Structure Factor, which is only weakly correlated (correlation coefficient ≈ 0.2) with the Energy‑Scale Factor. Consequently, variability alone does not provide a reliable proxy for the burst’s intrinsic luminosity, and the pseudo‑redshift estimates derived from it are statistically disfavored for this sample.

The second relation examined is the original Amati correlation, which links the rest‑frame Epeak to the isotropic-equivalent radiated energy (Eiso). In the present analysis, Epeak belongs to the Spectral‑Hardness Factor, while the fluence (a proxy for Eiso when a distance is assumed) belongs to the Energy‑Scale Factor. The two factors are only modestly correlated (≈ 0.3), and the factor analysis shows that the variance explained by a direct linear relationship between Epeak and fluence is insufficient. Therefore, the canonical Amati relation, at least in its unmodified form, does not hold robustly for the BATSE long‑GRB sample studied here.

The authors conclude that the observable diversity of long GRBs cannot be reduced to a single spectral or temporal parameter. Instead, at least five independent latent dimensions are required to capture the main physical variations. This finding has important implications for attempts to standardize GRBs as cosmological candles: any distance‑estimation technique must incorporate multiple observables, possibly through multivariate regression or machine‑learning models, rather than relying on a single empirical correlation. The paper suggests that future work should expand the variable set (e.g., including polarization, afterglow properties, or host‑galaxy information) and test the robustness of the factor structure on larger, more heterogeneous samples such as those from Swift and Fermi.