🔭 astrophysics

Updates to WFC3/UVIS Encircled Energy Values in Select Filters

This paper presents updated encircled energy (EE) curves for select HST/WFC3 UVIS filters, derived from improved PSF analysis and deep wing observations, which result in minor adjustments to zeropoints and inverse sensitivity tables to be released later in 2026.

Original authors: Anne O'Connor, Jennifer Mack, Varun Bajaj

Published 2026-08-13

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Anne O'Connor, Jennifer Mack, Varun Bajaj

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

=== DRAFT ===
Imagine you are trying to take a picture of a single, tiny firefly in a dark field using a giant, high-powered camera. You want to know exactly how much light that firefly is giving off. But here's the catch: the camera lens isn't perfect. Instead of all the light landing in one sharp, tiny dot, some of it spills out, creating a fuzzy glow around the center. This fuzzy glow is called a "Point Spread Function" (PSF). To measure the firefly's true brightness, you have to draw a circle around it and count the light inside. But if your circle is too small, you miss the spilling light; if it's too big, you might accidentally count the light from a neighbor.

Astronomers face this exact problem when looking at stars with the Hubble Space Telescope. They use a special tool called a "filter" to let in only specific colors of light. To get accurate measurements, they need a map called an "Encircled Energy" (EE) curve. Think of this curve as a recipe that tells them exactly what percentage of a star's total light is captured inside a circle of a specific size. If this recipe is slightly off, the calculated brightness of the star will be wrong, which throws off all the science built on those numbers. This report is about fixing that recipe for a specific set of Hubble filters, ensuring that when we count the light, we aren't missing any crumbs or counting the wrong stars.

The Great Light Hunt: Fixing Hubble's Star-Measuring Recipe

The team behind this report, Anne, Jennifer, and Varun, decided to take a fresh look at how the Hubble Space Telescope's UVIS camera measures starlight. They weren't just tweaking a knob; they were re-measuring the very shape of the starlight itself. Their goal was to update the "Encircled Energy" (EE) curves for a select group of filters. These curves are the mathematical bridges that help astronomers convert the raw pixels on a camera into the true brightness of a star.

The Problem: A Blurry Center and a Fuzzy Edge
Imagine trying to draw a perfect circle around a glowing balloon. If you don't know exactly where the center of the balloon is, your circle might be slightly off. In the past, the measurements of where the "center" of a star's light was located had a tiny error—sometimes off by as much as one whole pixel (a tiny dot on the camera sensor). For small circles (less than 10 pixels wide), this tiny mistake meant they were missing a significant chunk of the light.

Furthermore, measuring the light far away from the center (the "wings" of the star's glow) is like trying to hear a whisper in a noisy room. The old data was good near the center but got too "noisy" to trust when looking at the faint outer edges of the star's glow.

The New Detective Work
To solve this, the team acted like digital detectives. They used two different types of "snapshots" of the stars:

The "Short-Stack" Snapshots: These are made by stacking hundreds of short exposures. They are great for seeing the bright, sharp center of the star clearly.
The "Deep" Snapshots: These are made from very long exposures. They are so sensitive that the bright center gets "blown out" (saturated), but they reveal the faint, distant wings of the star's glow with incredible clarity.

By combining these two, the team could see the whole picture: the sharp center from the short stacks and the faint edges from the deep stacks.

The Big Fix: Sharper Centers and Better Edges
The team applied a new, super-precise method to find the exact center of each star. They fitted a smooth mathematical curve (a 2-D Gaussian) to the star's light to pinpoint its location with sub-pixel accuracy. This was a game-changer. For some filters, this new centering shifted the "center" by up to one pixel.

Why does this matter? Because for small circles (like the 10-pixel radius used to measure star brightness), being off by one pixel changes the amount of light you count. The new, sharper centers meant the team could capture more of the star's light accurately.

They also checked the outer edges of the star's glow. They compared their new measurements against a computer model of how the telescope's optics should work. They found that for ultraviolet (UV) filters, the old computer model was over-estimating how much light was in the outer wings. The real data showed slightly less light out there than the model predicted. For visible light filters, the model was pretty close, but the team still preferred to use their new, real-world measurements to be safe.

The Results: A New Recipe for Brightness
The team updated the EE curves for 14 filters, including popular ones like F275W (ultraviolet) and F814W (infrared). Here is what they found:

For Specific UV and Visible Filters: The new measurements show that slightly more light is captured in a small circle than previously thought for several specific filters (F218W, F225W, F275W, F775W, F814W, and F845M). The amount of light at a 0.4 arcsecond radius (10 pixels) is larger by about 0.5% to 2% for these filters, especially for the UVIS2 side of the detector. This might sound small, but in astronomy, it changes the brightness calculation by about 0.005 to 0.02 magnitudes.
For Long-Pass Filters: In contrast, for filters that let in a wide range of colors (like F200LP, F350LP, and F850LP), the new measurements show slightly less light in the small circle than before. The F850LP filter, for example, differs by about 2% (0.02 mag).
The "Amp" Difference: They noticed that the camera has two sides (UVIS1 and UVIS2), and they behave slightly differently. The UVIS2 side generally captures more light in small circles than UVIS1 for the filters that increased, while the trend varies for others. The team suggests using the UVIS2 numbers for most cases, unless you are specifically looking at the corner of the UVIS1 detector.

Why This Matters
These updated curves will change the "zeropoints" (the baseline numbers astronomers use to say how bright a star is) for these filters. The impact is small but significant: it brings the measurements from Hubble into much closer agreement with other independent studies of starlight.

The team is careful to note that while they have improved the measurements, there are still tiny differences between their new tables and measurements made from other types of star observations. They suspect this is because of how they measure the background "noise" (the sky) around the star. But overall, by sharpening the center and checking the edges with deep, high-quality data, they have created a more reliable map for measuring the universe's light.

In short, they didn't just tweak the numbers; they re-measured the shape of starlight itself, ensuring that when we look at the cosmos, our ruler is as straight and accurate as possible. These new tables will be released later in 2026, ready to help the next generation of astronomers count the stars with even greater precision.

Technical Summary: Instrument Science Report WFC3 2026-02

Problem Statement
This report addresses discrepancies identified between the current WFC3/UVIS photometric calibration and recent archival Point Spread Function (PSF) studies. Specifically, a 2025 study by Huynh et al. found large discrepancies (0.02–0.03 mag) in aperture corrections for UVIS2 in filters F275W and F814W when compared to the 2020 Encircled Energy (EE) tables. These discrepancies exceeded the expected variation across focus levels, motivating a re-examination of the empirical PSFs used to derive the 2020 EE solutions. Additionally, the 2020 EE curves for certain filters were derived using assumptions and scalings from other filters rather than direct, high-fidelity measurements for all cases.

Methodology
The authors re-computed EE curves for a subset of UVIS filters by combining two distinct types of drizzled PSF data with an updated analysis pipeline:

Data Sources:
- Short-Stack PSFs: Derived from stacking CALSPEC standard star observations (2009–2019) to achieve high signal-to-noise (S/N) in the PSF core (radii $r \le 2''$ ). These include both 2016 and 2020 datasets.
- Deep PSFs: Utilizing deep calibration observations (Programs 11438 and 17271) in six filters (F225W, F275W, F336W, F555W, F625W, F814W). These provide high S/N in the PSF wings ( $2'' < r \le 6''$ ) where short-stack data becomes noisy, though they are saturated in the core.
Analysis Improvements:
- Refined Centroiding: A 2-D Gaussian profile was fitted to a small cutout around the initial centroid for each star in the short-stack images to achieve sub-pixel accuracy. This corrected systematic centroid offsets of up to 1 pixel found in previous analyses.
- Curve of Growth: Using the Photutils package, count rates were measured in circular apertures from 1 to 150 pixels. Sky background was estimated using a 3 $\sigma$ -clipped mean in an annulus of 156–165 pixels to maintain consistency with existing zeropoint derivations.
- Hybrid Normalization:
  - For the six filters with Deep PSFs, the short-stack EE curves were normalized to match the empirical Deep PSF EE value at $2''$ .
  - For filters without Deep PSFs, the short-stack EE curves were normalized to the optical model (Hartig 2009) EE value at $2''$ .
  - The optical model was validated against Deep PSF data; it was found to agree with empirical data for visible (VIS) filters but over-estimate EE by $\sim0.5\%$ at $2''$ for UV filters (F225W, F275W, F336W). Consequently, UV filters were normalized to the empirical Deep PSF values rather than the model.

Key Contributions and Results

Improved Small-Radius Accuracy: The refined centroiding significantly improved EE measurements at small radii ( $r \le 10$ pixels). At $r=5$ pixels, updated EE values differ from 2020 values by up to 0.034 mag (wide filters) and 0.044 mag (long-pass filters). At $r=10$ pixels (the standard zeropoint aperture), differences reach up to 0.012 mag (wide) and 0.019 mag (long-pass).
Resolution of Discrepancies: The updated EE curves for F275W, F336W, and F814W show substantially better agreement with the focus-diverse PSF measurements of Huynh et al. (2025). The previous discrepancies were largely attributed to centroiding errors in the 2020 analysis.
UVIS1 vs. UVIS2 Differences: The new analysis confirms that UVIS2 (Amp C) consistently exhibits higher EE at small radii (5 and 10 pixels) compared to UVIS1 (Amp C corner), which is sensitive to astigmatism and focus variations.
Filter-Specific Updates:
- UV Filters (F218W, F225W, F275W, F336W): EE at 0.4'' ($10$ pixels) is generally larger by $\gtrsim 0.5\%$ ( $\sim0.005$ mag), particularly for UVIS2.
- Long-Pass Filters (F200LP, F350LP, F850LP): EE at 0.4'' is generally smaller. F850LP shows a significant difference of $\sim2\%$ ($0.02$ mag).
- VIS Filters: F475W, F555W, F606W, and F625W were included for validation but are not updated in this delivery, as their 2016/2020 solutions were consistent with archival PSFs.
Optical Model Validation: The study confirms that the Hartig (2009) optical model agrees well with empirical data for VIS filters but over-predicts EE for UV filters at large radii ( $2''$ ).

Significance and Impact
The revised EE solutions directly impact the WFC3/UVIS photometric zeropoints, which are derived from aperture photometry in a 0.4'' radius and corrected to a 6'' aperture using EE tables. While the overall impact is described as small, the updates correct systematic errors in the calibration:

Several filters (F218W, F225W, F275W, F775W, F814W, F845M) show EE increases of $\gtrsim 0.5\%$ ($0.005$ mag) at 0.4''.
F850LP shows a decrease of $\sim 2\%$ ($0.02$ mag).

The report concludes that the updated methodology, particularly the improved centroiding and the hybrid normalization strategy using Deep PSFs and the optical model, produces EE curves that are more accurate and consistent with independent archival PSF studies. Updated EE tables, along with revised inverse sensitivity tables and zeropoints, are scheduled for delivery later in 2026. In the interim, updated tables for commonly used filters are provided in the report's appendix. The authors caution that for small apertures ( $r < 7$ pixels), where EE is highly dependent on focus and orbital breathing, users should derive aperture corrections from isolated stars in individual science frames rather than relying solely on tabulated values.

More like this