Image persistence flagging for SPHEREx

In the quiet, frozen silence of deep space, telescopes act as the most sensitive eyes humanity has ever built, capturing faint whispers of light from the earliest galaxies. To see these distant signals, astronomers rely on specialized cameras equipped with sensors that can detect infrared light, a form of radiation invisible to the human eye but rich with cosmic history. However, these sensors are not perfect; they have a memory. Just as a bright light can leave a lingering afterimage on your retina after you look away, these sensitive detectors can hold onto a ghost of a bright source long after the source has moved out of view. This phenomenon, known as image persistence or a latent signal, means that a pixel that was just flooded with light from a bright star might continue to register a faint, decaying signal even when it is looking at empty, dark space. For a mission designed to map the entire sky, this lingering ghost can muddy the waters, contaminating new images with the echoes of old ones and making it difficult to distinguish the true background of the universe from the sensor's own stubborn memory.

The SPHEREx mission, a planned space telescope designed to survey the entire sky, faces this exact challenge. It will use six of these infrared detector arrays to take thousands of pictures as the spacecraft turns and moves from one patch of sky to another. The team behind the mission needed a way to know which pixels in their images were being corrupted by this ghostly signal so they could either correct for it or mark them as unreliable. To solve this, researchers Candice Fazara and her colleagues at the Rochester Institute of Technology, the California Institute of Technology, and the Jet Propulsion Laboratory developed a new method to predict and flag these persistent signals. They did not have data from the actual flight cameras yet, so they turned to a test detector, a spare unit built to the same specifications, to see how the sensors behaved under controlled conditions.

The researchers illuminated this test detector with a bright, concentrated beam of light, simulating the intense glare of a bright star, and then watched what happened after they turned the light off and reset the sensor. They observed that the sensor did not simply go dark; instead, it produced a decaying signal that slowly faded over time. By analyzing how this signal changed, they discovered that the amount of lingering charge depended heavily on how bright the original light was and how long the sensor was exposed to it. They found that for pixels that were not overwhelmed by light, the lingering signal grew in a predictable way based on the total amount of light received. However, for the most intensely lit pixels, the behavior became more complex, with some pixels holding onto a much stronger signal than others, likely due to tiny imperfections in the sensor material that trapped electrical charges.

To create a tool that could be used in real-time during the mission, the team built a mathematical model that could estimate how much of this ghost signal would remain in any given picture. They tested this model against their experimental data and found that while a simple formula could describe the general decay, it sometimes underestimated the signal in the first few minutes after a bright exposure. To be safe, they adjusted their model to slightly overestimate the lingering signal, ensuring that any pixel potentially contaminated would be caught. They then translated this into a practical system: a set of rules that the spacecraft's computer could follow. This system looks at the brightness of the current image, calculates how much "memory" the sensor should still have from previous views, and flags any pixel where the predicted ghost signal is too strong to ignore. They set two levels of caution: a lower limit to catch even faint traces of the signal, and a higher limit to catch only the most obvious contamination.

To see if this system would work in the real world, the team ran computer simulations using a virtual version of the SPHEREx mission. They fed the model a sequence of two days' worth of simulated sky observations, including views of bright stars and the faint glow of the Zodiacal light. The simulation showed that the model worked as intended, successfully predicting which pixels would be affected. The results indicated that after a bright exposure, a small but steady percentage of pixels—about two percent—would remain flagged as contaminated by the ghost signal for a significant amount of time. Even more notably, the simulation suggested that a much larger portion of the detector would always have a tiny, lingering signal above the faintest background noise, creating a kind of "floor" of persistence that never fully disappears.

This work provides the mission team with a crucial safety net. By using this model, the SPHEREx team can identify and mark the pixels that are likely to be unreliable, allowing scientists to either correct the data or exclude those specific spots from their final maps. While the model was built using a test detector and relies on simulations, it offers a robust way to handle the inevitable memory of the sensors. The researchers acknowledge that the actual behavior of the flight cameras might differ slightly, and they plan to refine their model once the mission launches and they can measure the real detectors in space. For now, they have established a clear path forward, ensuring that the ghost of a bright star will not obscure the faint, ancient light of the universe that SPHEREx is designed to find.