1. First On-orbit Calibration of WFC3-IR Count Rate-Dependent Non-Linearity Adam Riess WFC3 ISR 2010-07 Count-rate non-linearity (a.k.a. the Bohlin Effect, reciprocity failure Is not the same effect as count non-linearity! cr(x,y)  f(x,y) , cr is observed count rate f is the flux on pixel, source+sky  ≥ 1 (=1 for no effect)

2. Rate Dependent Non-Linearity Problem: Zeropoints calibrated with standard stars at m=12. sky-dominated sources * (e.g., UDF) are ~5 dex (10 5 ) fainter NICMOS had a rate-dependent non-linearity of 0.063 mag/dex (F110W) and 0.029 mag/dex (F160W), so photometry errors of ~0.3 mag result if uncorrected! In 2006 NICMOS used “count-rate boosting” (i.e., lamps) to fully calibrate the effect. WFC3 cannot observe with lamps so will use the bright Earth limb (GO 11933 July 23 rd ) to do same. However, we can compare the WFC3-IR linearity now to NIC2/ACS using photometry of identical sources and comparable filters. *sky limits minimum count rate to m~23 for WFC3-IR, m~24 for NICMOS

3. NIC2/ACS to WFC3 Cross-calibration All data processed using best current reference files (end 2009) as well as contemporaneous P330 calibration star (all cameras, all filters). Matched source list produced NIC2 photometry already calibrated for rate-dep non-lin, ACS corrected for CTE 47 Tuc (2 positions) NIC2(F110W/F160W) WFC3(F110W/F160W) NGC 3603 WFC3 (F098M) ACS(F850lp) ~220 stars in common ~1400 stars in common

4. Stellar Photometry All photometry from small, 0.2” apertures, measured relative to P330E Color terms to transform, e.g.: NIC(F110W)-WFC3(F110W) NIC(F110W)-NIC(F160W) between P330E and cluster stars calibrated using appropriate Castelli and Kurucz (2004) models (for 47 tuc Z=-0.2, gravity=g50) and calspec of P330E with synphot Solve for residual color dependence

5. Cross-calibration Model NIC or ACS (mag)=a 0 *WFC3 (mag) +Δa 1 [NIC or ACS Color (mag)] a 0 measures WFC3-IR non-linearity relative to NIC/ACS (from P330E to sources) cr=flux ,  =2-a 0. Δa 1 =residual color correction after synphot transform Multi-linear regression, 2 free parameters. Clipped points     (~ 2%) Non-linearity detected >4 sigma. Factor of ~4x smaller than NICMOS. Result: Photometry of sky-dominated sources (m>23) expected to be measured too faint by 0.04 +/ 0.01 mag % / dex 1.0 +/-0.2 1.1 +/-0.2 1.1 +/-0.3

6. 47 Tuc NIC2 vs WFC3-IR (example, F110W), ~220 stars Direct Comparison: NIC2 to WFC3, No color term Model Residuals Residuals vs. color F110W=23 (Vega), sky limit

7. ACS F850lp vs. WFC3-IR F098M NGC 3603, 1000 stars Straight difference, no color term Model Residuals Color residuals

8. What Next? Count-Rate Boosting Bright Earth skimming observations with WFC3 (like NICMOS lamp on/off), cleanest approach (model independent), should get strong result, scheduling July 23rd lamp off lamp on - lamp off NICMOS NIC2 F110W NICMOS ISR 2006-001

9. 0.042 mag/dex (dashed) NICMOS Count-rate Non-linearity model-independent calibration (F110W), 3 dex, to be 0.063 mag/dex (see ISR 2006-001, all cycles’s data now combined to extend 1 dex) Calibration extended by WFC3 (via ACS CCD) from 17 to 22 mag F110W (ISR 2010-007)  non-linearity correction remains consistent to σ=0.03 mag at faint end (10 to 22 rd mag) NICMOS Count-Rate Dependent Non-linearity Revisited Observed Count Rate (ADU/sec), NIC2 F110W Original range Sky Vega Mag P330E lamp on vs off Stellar phot. WFC3 vs NIC2 fractional residual Residual from 0.063 mag/dex calibration shown

10. Summary We have measured the count-rate dependent non-linearity of WFC3-IR to be 1% +/- 0.2% per dex, insensitive to wavelength, by comparing stellar photometry over 3 dex down to the sky limit We will re-measure this by boosting the count rate of a stellar field with bright Earth limb light later this month Using WFC3-IR cross-calibrated via ACS, we affirm the calibration of the NIC2 count-rate dependent non-linearity down to near the sky (0.1 ADU/sec/pixel or F110W=22 mag) to 0.03 mag, good agreement with Bob Hill’s lab measurements of 3 different HgCdTe devices Non-linearity splinter session at 3pm Friday, N310

11. Power Law Non-linearity in HgCdTe Observed Count Rate (ADU/sec), NIC2 F110W Original range Sky Vega Mag P330E lamp on vs off Stellar phot. WFC3 vs NIC2 fractional residual

12. WFC3 vs NICMOS Persistence NICMOS WFC3 Our persistence (blue curve) is down by a factor of few from NIC, just like our non-linearity, which makes sense Persistence Models (Smith et al 09) indicate persistence and cr non-linearity both result from trapping, degree should scale with trap density

13. Persistence and Non-Linearity: Catch and Release Count-rate Non-linearity and persistence appear to arise from the same phenomenon of charge trapping (Smith et al. 2007), former from capture, latter from the release. WFC3 persistence measured from stars in well dithered 47 Tuc IlluminationSubsequent Image minus registered, prior image

14. Persistence and Non-Linearity Persistence=P(t illum,f illum ) Total prior illumination Persistence Time Since we see persistence in WFC3, we must have some degree of non-linearity Time decay function (double exp or power law)Illumination Dependence (Fermi-Dirac) (I have a tool to remove persistence using these functions)

15. Next epoch

16. 21.0 (Vega) =2 ADU in F160W =0.029 mag/dex 24-21=1.2 dex =0.035 mag extrapolation 25 (sky limited) =0.046 mag extrapolation 14 NICMOS Count-rate Non-linearity calibrated from 14 to 21 st mag (F160W)

17. Synthetic Color Terms: NIC2 to WFC3-IR I’d prefer more stars with colors 0.0-0.5 (rel P330E), but I propagate generous error of 0.05 Good agreement with use of 47 tuc colors to measure, but less precision P330E color range of 47tuc vd8