Cross-spectral Analysis of the Type-C Quasi-periodic Oscillation Shoulder Component in GX 339-4

We revisit Rossi X-ray Timing Explorer (RXTE) observations of GX 339−4 during the rising phase of its 2006/2007 outburst and apply a joint power-density-spectrum (PDS)–cross-spectrum (CS) decomposition to the type-C quasi-periodic oscillation (QPO) region. Within this framework, the QPO region is described by a narrow QPO fundamental and a neighboring high-frequency shoulder, whose amplitudes and phase lags can be measured separately. The shoulder is first detected at MJD 54142.04, mainly through the imaginary part of the CS and a narrow local structure in the phase-lag spectrum, before becoming a resolved high-frequency shoulder in the PDS. It follows the QPO frequency evolution on the high-frequency side, with Rν = νsh /νQPO ≃ 1.04–1.18. The QPO lag remains small, typically below ∼ 0.17 rad, whereas the shoulder carries a larger hard lag of ∼ 0.5–0.8 rad. Energy-resolved fits show the same separation: the QPO lag is close to zero or only weakly positive across most of the energy band, while the shoulder lag is systematically larger and generally increases with photon energy. The two components have broadly similar rms–energy shapes, although their relative strengths evolve during the observed sequence. Although the shoulder remains broad, with Q ∼ 2–4, its lag and rms–energy behavior resemble those of the type-B QPO detected shortly after our observations. This similarity raises the interesting possibility that the shoulder is related to an earlier, broader stage of the variability later seen as the type-B QPO.

Improvements to AI systems:

1. Enhanced time-series decomposition for non-stationary signals

• The AI can now jointly model power-density spectra (PDS) and cross-spectra (CS) to separate overlapping variability components (e.g., a narrow QPO fundamental from a broad high-frequency shoulder) in noisy astrophysical data. This improves detection of weak, transient features that are invisible in PDS alone but appear in the imaginary part of CS or phase-lag spectra.

2. Automatic identification of precursor signatures

• The AI can learn to flag early-stage variability patterns (like the shoulder preceding a type-B QPO) by tracking frequency ratios (Rν = 1.04–1.18) and lag evolution over time. This enables predictive classification of state transitions in accreting systems, potentially forecasting outburst evolution.

3. Energy-resolved multi-band correlation analysis

• The AI can simultaneously fit rms–energy and lag–energy spectra for each decomposed component, distinguishing components with near-zero lags from those with increasing hard lags. This improves physical interpretation of emission regions (e.g., corona vs. jet) and can be generalized to other multi-wavelength time-domain datasets.

4. Robust phase-lag estimation with local structure detection

• The AI can detect narrow local structures in phase-lag spectra (e.g., a shoulder appearing as a bump) even when the amplitude in PDS is unresolved. This improves sensitivity to subtle frequency-dependent timing signatures in low-signal-to-noise observations.

5. Adaptive component tracking across frequency evolution

• The AI can track a component’s centroid frequency as it drifts (e.g., νQPO rising from 0.1 to 1 Hz) while maintaining a separate shoulder with a variable frequency ratio. This enables real-time monitoring of dynamical systems with multiple coupled oscillators.

What the improved AI system can do:

• Automatically decompose and characterize quasi-periodic oscillations and their sidebands in X-ray binaries, with separate measurements of amplitude, quality factor, and phase lag per component.

• Predict the emergence of type-B QPOs from earlier shoulder-like variability, aiding in early-warning systems for state transitions in black hole binaries.

• Provide a unified framework for analyzing time-lag and rms spectra across energy bands, applicable to other compact objects (e.g., AGN, pulsars) and terrestrial signal processing where multi-component oscillatory signals overlap.

• Operate on sparse or irregularly sampled time series, using cross-spectral information to recover features missed by standard periodogram methods.

We revisit Rossi X-ray Timing Explorer (RXTE) observations of GX 339 - 4 during the rising phase of its 2006/2007 outburst and apply a joint power-density-spectrum (PDS)--cross-spectrum (CS) decomposition to the type-C quasi-periodic oscillation (QPO) region. Within this framework, the QPO region is described by a narrow QPO fundamental and a neighboring high-frequency shoulder, whose amplitudes and phase lags can be measured separately. The shoulder is first detected at MJD 54142.04, mainly through the imaginary part of the CS and a narrow local structure in the phase-lag spectrum, before becoming a resolved high-frequency shoulder in the PDS. It follows the QPO frequency evolution on the high-frequency side, with R nu= nu sh/nu QPO 1.04 -- 1.18. The QPO lag remains small, typically below about0.17 rad, whereas the shoulder carries a larger hard lag of about0.5 -- 0.8 rad. Energy-resolved fits show the same separation: the QPO lag is close to zero or only weakly positive across most of the energy band, while the shoulder lag is systematically larger and generally increases with photon energy. The two components have broadly similar rms--energy shapes, although their relative strengths evolve during the observed sequence. Although the shoulder remains broad, with Q about2 -- 4, its lag and rms--energy behavior resemble those of the type-B QPO detected shortly after our observations. This similarity raises the interesting possibility that the shoulder is related to an earlier, broader stage of the variability later seen as the type-B QPO.