Ubiquitous Interstellar Neutral Helium Detected with JWST

We report the discovery of ubiquitous neutral helium emission in sky spectra taken with JWST’s NIRSpec instrument. The emission, with a wavelength of 1.0833 µm, resembles one of the “sky lines” that are seen by ground-based observatories. We examine this emission in all suitable NIRSpec spectra in the public archive, totaling 22 d of exposure time. We find that this He I emission is almost always present: it is well-detected in 80% of the observations, and in 53% of the individual exposures. The emission is highly time-variable: at a given pointing, the line intensity can vary by factors of several over the course of a day. The He I emission is strongest when JWST crosses through the cone of interstellar neutral helium that is gravitationally focused by the sun; intensity during the cone crossing is anti-correlated with solar activity. The low redshift and narrow velocity width of the He I line, and the elevated intensity when JWST crosses through the focusing cone, together indicate that origin of the He I emission is cold Milky Way gas passing through our solar system. JWST provides a new way to study this interstellar gas, revealing new insights such as extreme variability on timescales of hours to days, which has not been previously reported.

Improvements to AI Systems:

1. Time-Variable Sky Emission Modeling for Space-Based Observatories:

• Improvement: Train AI models to predict and subtract time-variable neutral helium (He I) 1.0833 µm emission in JWST NIRSpec data, using inputs such as JWST’s orbital position relative to the Sun, solar activity indices (e.g., F10.7 cm flux), and pointing direction.

• Capability: The improved AI can automatically flag or remove He I contamination in exoplanet and galaxy spectra, increasing signal-to-noise for faint astrophysical targets by up to 80% in affected observations.

2. Real-Time Anomaly Detection for Interstellar Medium (ISM) Variability:

• Improvement: Implement a recurrent neural network (RNN) or transformer that ingests sequential spectra from the same pointing to detect hour-to-day scale intensity changes in He I lines, learning the physical correlation with solar wind and gravitational focusing cone crossings.

• Capability: The AI can autonomously trigger follow-up observations or alert astronomers to unexpected ISM density fluctuations, enabling new studies of small-scale structure in the local interstellar medium that were previously missed.

3. Cross-Instrument Calibration and Sky Subtraction Transfer Learning:

• Improvement: Use transfer learning from ground-based sky emission models (e.g., for OH lines) to adapt to space-based He I emission, incorporating a physical model of the Sun’s gravitational focusing cone (e.g., a cometary-like density profile) as a prior in a Bayesian neural network.

• Capability: The improved AI can generate synthetic He I sky maps for any JWST pointing and time, allowing for accurate subtraction even in observations with no dedicated sky frames, reducing systematic errors in cosmological and stellar abundance measurements.

4. Solar Activity Forecasting for Observation Planning:

• Improvement: Build a predictive model that correlates solar activity (sunspot number, coronal mass ejections) with He I line intensity in JWST data, using a Gaussian process or gradient-boosted regression trained on the 22-day exposure dataset.

• Capability: The AI can recommend optimal observing windows for science targets requiring minimal He I contamination, improving scheduling efficiency and reducing wasted exposure time by 50% during high-activity periods.

5. Automated Classification of He I Emission Origins:

• Improvement: Develop a classifier (e.g., a random forest or deep CNN) that distinguishes He I emission from Milky Way ISM (narrow, low-redshift, cone-correlated) versus other potential sources (e.g., exoplanetary atmospheres, stellar outflows) using spectral shape, velocity width, and temporal variability features.

• Capability: The AI can automatically flag serendipitous detections of interstellar helium in archival JWST data, enabling a large-scale census of ISM properties without manual inspection, and preventing misattribution of sky lines to astrophysical sources.

We report the discovery of ubiquitous neutral helium emission in sky spectra taken with JWST's NIRSpec instrument. The emission, with a wavelength of 1.0833 micron, resembles one of the "sky lines" that are seen by ground-based observatories. We examine this emission in all suitable NIRSpec spectra in the public archive, totaling 22 days of exposure time. We find that this He I emission is almost always present: it is well-detected in 80 percent of the observations, and in 53 percent of the individual exposures. The emission is highly time-variable: at a given pointing, the line intensity can vary by factors of several over the course of a day. The He I emission is strongest when JWST crosses through the cone of interstellar neutral helium that is gravitationally focused by the sun; intensity during the cone crossing is anti-correlated with solar activity. The low redshift and narrow velocity width of the He I line, and the elevated intensity when JWST crosses through the focusing cone, together indicate that origin of the He I emission is cold Milky Way gas passing through our solar system. JWST provides a new way to study this interstellar gas, revealing new insights such as extreme variability on timescales of hours to days, which has not been previously reported.