Uranus and Neptune as Calibrators Glenn Orton Bryan Butler, Mark Hofstadter, Mark Gurwell, Leigh Fletcher

Outline Quick review of older models of Uranus and Neptune Implications of the Uranus T(p) developed from high-SNR Spitzer IRS data for predicting the submillimeter/millimeter spectrum. Models for the far-infrared/submillimeter spectrum of Uranus: implications for opacity sources Comparison of models of Mars (and other planets) with high-precision WMAP observations, implications for recalibration of Uranus and Neptune data (summary of the talk Bryan Butler would have given – he’s recovering from surgery) Comparisons between Uranus and Neptune, implications for Neptune’s spectrum: cross-comparison with ISO/LWS, SPIRE ratios and with PACS (Lellouch et al. 2010)

Previous models were based on: Voyager-2 radio occultations of Uranus and Neptune T(p)– at a single point H2, He mixing ratios and constraints on CH4. Neptune:

Uranus Earth-Based Observations & Voyager IRIS: Used by Griffith and Orton (1991) to define a standard spectrum. Uranus Uranus standard model, based on Voyager IRIS data (Pearl et al. 1990 Icarus 84, 12) Orton et al. (1987) Icarus 70, 1

Earth-Based Observations & Voyager IRIS: Used by Griffith and Orton (1991) to define a standard spectrum for Uranus. from Griffin and Orton (1993) Icarus 105, 537.

NEPTUNE Neptune: ISO/LWS and SWS spectrum calibrated using Uranus, based on the Griffith & Orton (1993) model spectrum. WAVELENGTH (µm) Burgdorf et al. (2003) NEPTUNE Orton et al. (1987) Icarus 70, 1 From Burgdorf et al. (2003) Icarus 164, 244.

Spectrum fit on basis of matching deeper T(p) to H2-H2 and H2-He collision-induced continuum. from Griffin and Orton (1993) Icarus 105, 537.

Spitzer IRS Observations of Uranus Taken 17 December 2007, ten days from equinox IRS composed of 4 different modules: Short High, Short Low, Long High, Long Low Spectral range of 5.2 - 37.2 microns (1920 cm-1 - 270 cm-1) High Res: /~600 from 10 - 36.5 m Low Res: /~90 from 5 - 21.5 m Uranus appears as a point source to Spitzer, so we are seeing a disk-averaged flux Four longitudes observed spaced evenly over ~17 hour equatorial period in all four Spitzer Space Telescope Infrared Spectrometer (IRS) modules

Zonally-averaged spectra: • Short-Low spectra and Long-Low zonal average spectra: consistent to well within ±3% variability, except for hydrocarbon emission. • We rely almost solely on these for accurate temperature information.

Reproducibility: • Spectral averages are good to within 3% for radiation emerging from the troposphere – verifies confidence level of the observations and shows that tropospheric temperatures are not variable on a hemispherical average. • But: hydrocarbon emission is up to ±10-12% variable

Problems with high-resolution modes - 1 The Q-branch of the ν5 band of C2H2 saturated SH Spectra Each spectral order had to be normalized to low-res spectrum ↑ ↑ Orders 18-20 were impossible to correct

Problems with high-resolution modes - 2 Many orders also had to be “pivoted” to match their low-resolution counterparts.

Problems with high-resolution modes – 3 Long-High (LH) orders were uncorrectable (1) no low-resolution spectrum to compare with (2) clear requirements for both scaling and pivoting We very reluctantly abandoned the LH data as trustworthy (they could have constrained He/H2 and para-/ortho-H2 ratios) LH data for Neptune (Cycle 1, Cycle 2) have the same problem.

Spectral regions controlled by H2 used to determine T(p): 1. collision-induced absorption (SL and LL modes only) 2. S(1) quadrupole (SH, but scaled to the SL continuum) H2 S(3) quadrupole ↓ CH4 C2H2 C2H6 H2 CIA H2 S(1) quadrupole H2 CIA

Wide vertical range of sensitivity: H2 S(1) quadrupole senses highest (some regions not covered well) H2 CIA continuum near quadrupole senses upper troposphere H2 CIA continuum near 1000 – 1100 cm-1 senses deepest

Because vertical coverage is not uniform, an “a priori” profile was adopted from the Voyager RSS occultation profile. H2 S(3) quadrupole ↓ CH4 C2H2 C2H6 ←RSS T(p) is too cold for 1-2 bar region ←a straight line smoothed through RSS stratospheric oscillations C2H2 Clearly the Voyager RSS occultation profile does not represent the disk-averaged temperatures all that well. H2 S(1) quadrupole RSS T(p) near tropopause does not match details of H2 CIA S(1) rotational line

Use a standard temperature retrieval approach to determine the tropospheric temperature structure. • Fit is extremely good . best fit RSS • Best-fit disk-averaged T(p) requires warmer troposphere than RSS results.

This “bottom up” approach provides a very good fit to both the quadrupole and CIA continuum. • the H2 S(1) quadrupole • the very broad H2 CIA S(1) rotational line

Initial Temperature Sounding Results: 7 mbar - 2 bars Variety of overlying lapse rates as upper boundary conditions Simultaneous fits to both continuum and H2 S(1) quadrupole Adiabat assumed for higher pressures Voyager ratios of H2, He, unsaturated CH4 H2 CIA fit strongly discriminates between various stratospheric lapse rate assumptions.

Alternative boundary condition: upper-level isotherm works equally well H2 CIA fit provided similar discrimination for these models

Best Temperature Profiles “warm” “cool”

Variations of thermospheric temperature base to match H2 S(2) line  Patterned after Herbert et al. Voyager UVS occultation temperatures  Variation required simultaneous adjustments of 10-4 – 10-5 temperatures to fit H2 S(1) and the H2 CIA.  As well as fitting C2H6 H2 S(2) quadrupole C2H6

Variations of thermospheric temperatures provide a match H2 S(3) line H2 S(3) quadrupole In reality the T(p) models were run with iterated variations to determine the best fits to H2 CIA continuum plus all three - S(1), S(2) and S(3) - quadrupole lines. H2 CIA “continuum”

Best-Fit Temperature Profiles However different these are in the stratosphere, they’re nearly identical in the troposphere.

Evaluation of the derived T(p)  Uncertainties in the absolute calibration ±3% from measurement ±5% is Spitzer’s quoted absolute calibration uncertainty Total estimated as ±6% This propagates into an absolute temperature uncertainty that is very small: ±0.2 K This uncertainty would propagate into a submm radiance uncertainty of ±0.4%. Errors are thus more likely to arise from other systematic uncertainties or assumptions.

Evaluation of the derived T(p) Retrieved profile is consistent with adiabat bottom of direct-sounding region: adiabatic lapse rate assumed below Note consistent lapse rates (not much sensitivity to composition) Test of different compositions: Solid 15.2% He Dashed 11.9% He Dotted 18.5% He (results from Conrath et al. 1987)

Uncertainties in the bulk composition are based on those of Voyager for mixing ratios of H2 , He ~ ±0.033 Biggest uncertainty is in far infrared: ±3% at 100 cm-1/100 μm, ±6% at 175 cm-1/ 57 μm Negligible influence at the longest wavelengths/ shortest frequencies. Para-/ortho-H2 uncertainty propagates an uncertainty in the spectral radiance that is very small. Mitigation of compositional uncertainty: successful SOFIA Basic Science proposal to constrain He/H2 ratio with multiple filters 19-37 μm in 2010 (to be followed by a spectroscopic study)

Nominal model does come close to - Voyager-2 IRIS observations of the disk of Uranus (1986, pole-on geometry) ____ - ISO SWS spectrum, calibrated relative to Mars (Burgdorf et al., unpublished) ☐☐☐☐ Solid 15.2% He Dashed 11.9% He Dotted 18.5% He (results from Conrath et al. 1987)

But we must be careful in making such comparisons But we must be careful in making such comparisons. There is a big change in the projected geometry over decades. In 1986, Voyager IRIS observed the whole disk of the planet pole-on; in 2010 we’re just past equinox. 1999 2004 Enhanced R, G, B colors from HST observations by Eric Karkoshka (1999); HST: enhanced R, G, B Keck AO: enhanced J, H, K (Karkoschka, press release) (de Pater and Hammel, press release)

VLT/VISIR 18-μm images of Uranus 1,2 Sept 2006 There is also genuine change in time whose influence we have to evaluate on the disk-averaged spectrum – the north polar region has cooled. VLT/VISIR 18-μm images of Uranus 1,2 Sept 2006 North pole should be warm, but it has cooled since 1986. How deep are these differences? - Assess using 19-21 μm imaging, sensing deeper Uranus obs. deconvolved IRIS-based simulation

Longer wavelengths need another opacity: revisiting a conclusion from an old calibration mini-workshop: ← a model with only H2 CIA absorption is too warm Griffin and Orton (1993) Gurwell and Hofstadter, SMA data

- - - Including NH3 gas does better, with 1 ppm VMR at depth (Moreno’s suggestion) But with our derived T(p), it’s too bright for the 5-10 cm-1 (150 – 300 GHz) data, particularly the recent 1.4-mm (7 cm-1) SMA measurement

H2S does better, with a deep mixing ratio of 30 ppm (VMR = 3 x 10-5) and full saturation equilibrium

New data obtained since that time: - CSO and JCMT (Sandel and Dowell) - SMA and VLA (Butler, Gurwell, Hofstadter) - improved precision and accuracy New calibration (Butler, who could not attend this workshop): A summary of his presentation follows:

Flux Density Scale in cm Based on Baars et al. 1976, which was fundamentally based on absolutely calibrated (horn, e.g.) observations of strong sources like Cygnus A and Cassiopeia (which are either known not to vary or vary in a predictable way, based on physical arguments), transferred to secondary sources like 3C 295, 3C 286, 3C 48, etc. Absolute scale at frequencies less than about 10 GHz accurate to a few %, 5-10% at 15 GHz, not claimed to be accurate above that. Problem with absolute scale sources is that they are very large. Secondary sources are smaller, but accuracy of secondary scale depends on how much the secondary sources vary with time, which is not as predictable especially as frequency goes up.

VLA Observations For ~30 years Perley (and for the past 15 year Butler) have been measuring a large set of secondary sources against each other with the VLA to see how they vary. Below 10 GHz: 3C 295 is constant to fraction of % 3C 286 is constant to less than 1% 3C 48 and others can vary significantly Above 10 GHz: Most 3C sources vary (understood physically because core starts to dominate emission, and the AGN at the core is expected to vary). NGC 7027 has predictable secular decrease (but is marginally too large). Other “standard” sources have similar problems. Planets could be used, because they are good blackbodies (modulo sounding depth issues), but models of planets are either too fundamentally uncertain or based on Baars et al. (or other) scales that are too uncertain.

Enter WMAP WMAP measured Jupiter against the CMB dipole accurately (Page et al. 2003), but Jupiter too large. WMAP finally published measurements of Mars, Uranus, and Neptune (Weiland et al. 2010 in press). With this, we can use WMAP observations to “calibrate” emission models for the planets. Weiland et al. do this for Mars, using the Wright model. We choose to use the Rudy model because it incorporates more physics (polar caps, for instance).

WMAP Mars and Rudy Model WMAP observations in 5 bands (~23, 33, 41, 61, & 93 GHz) plotted against Rudy model. Model fits well except first data point, when there was a dust storm. Adjustment of predicted flux density from Rudy model downward by 2% is required; constant vs. frequency (Weiland et al. found they needed a frequency dependent correction).

The Mars/Baars Scale Using Baars scale for frequencies < 10 GHz (for 3C 295 transferred to other sources) and Mars for frequencies > 10 GHz we can now calibrate our 30 years of observations at cm wavelengths absolutely, determining the secular variation of the secondary sources. We then use these secondary sources to calibrate other observations (e.g, Uranus or Neptune).

Mars as Primary? We now have an absolutely calibrated model for Mars, from ~1 GHz to at least 100 GHz (and indications are that it is good at higher frequencies). Why not just use it for all calibrations? Can be near the Sun, causing problems (notably for spacecraft visibility). Can be large, meaning beam corrections need to be accurate. Still doesn’t incorporate all of the physics (dust storms, for instance).

Enter Uranus & Neptune Uranus & Neptune are recognized “standard” calibrators in mm/submm; they are strong, relatively small, less solar confusion, no (or little at least) “weather”, etc. Problem is that previous “models” for emission were not physically based; notably they didn’t account for known secular changes in Uranus geometry.

Uranus Secular Changes 2 cm 6 cm 1981 1985 1989 1994 2002 Hofstadter & Butler 2003

Work on Uranus & Neptune Our group has been working on creating such physically based emission models for Uranus and Neptune from the infrared to centimeter wavelengths (and Jupiter, but it is less useful in this context). A key input to these models is characteristics of the deep (~bars and deeper) atmospheres of the planets. We can now absolutely calibrate long wavelength (λ > 1mm) Uranus and Neptune observations done with the VLA (and other radio telescopes) over the past 30 years at least, which probe to those depths, and hence constrain the models and how they vary with geometry (hence time).

H2S, same model (PH3 might also work but is a less likely candidate on the basis of chemical equilibrium) Sandel & Dowell data (2005)

Mark Hofstadter’s collection of UKIRT, CSO and VLA data with the WMAP correction, and fitting H2S and NH3 with the Spitzer-IRS T(p)

Mark Hofstadter’s collection of UKIRT, CSO and VLA data with the WMAP correction, and fitting H2S and NH3 with the Spitzer-IRS T(p)

Uranus spectral models suggested for Herschel calibration: Moreno’s model: based on colder RSS profile, with NH3 absorption below ~300 GHz “Orton’s model” should be testable with detection or non-detection of H2S absorption features, We already know that variability across the 1.4-mm SMA image is too big to be explainable by T alone. ~4 Kelvin maximum difference between the two models (maximum 5% difference in radiance prediction)

Moreno model (Voyager-2 RSS) Orton model (Spitzer IRS), continuum only shown maximum of 8% difference in radiance prediction near 167 cm-1 / 60 μm

Spitzer IRS Spectrum of Neptune CH4 C2H6 Note minimal region where H2 CIA spectrum is available in SL spectrum  CH3D Not much of spectrum is useful to constrain the global-mean T(p), but an attempt was made by student, Mike Line (2008). C2H2

Neptune: T(p) from Spitzer IRS data (Line et al. 2008 DPS) …forging ahead anyway… Perturb smoothed Voyager radio occ. temperature structure around the tropopause to match 510 – 640 cm-1. 2. Vary the overlying stratosphere to fit to the H2 S(1) quadrupole. 3. Vary the upper stratospheric temperature structure to fit the observed CH4 emission, accounting for the unknown stratospheric VMR of CH4

Akari data analysis (Fletcher et al. 2010) CH3D C2H6 Also limited to stratospheric emission features: tropospheric T(p) from Moses et al. (2005), of unknown provenance but looks like Burgdorf et al. (2003). C2H4

ISO Heritage: LWS and SWS studies of Neptune: application to FIRST Herschel (Burgdorf et al. 2003) LWS ground-based: IRTF (Orton et al. 1989, 1990) SWS calibration based on the Griffin & Orton “standard model” for Uranus

Herschel PACS: HD & T(p) ∆ λ = 0.02 μm H2O HD (6-9-12 x 10 -5) + best P(T) HD (910 -5) + Colder P(T) H2O (profile A) Observations ∆ λ = 0.12 μm Nominally smaller but consistent with ISO value (D/H = (6.5±2) x 10-5 , Feuchtgruber et al 1999 ) D/H = (4.5±1) x 10-5

T(p) models for Neptune Herschel PACS (Lellouch et al. 2010) Spitzer IRS (Line et al. 2008) ISO + ground-based (Burgdorf et al. 2003) Akari (Fletcher et al. 2010) Voyager RSS - - - (Lindal et al. 1990)

Comparison of PACS T(p) model spectrum with Spitzer IRS PACS T(p) does not fit the Spitzer H2 continuum, it’s too cold Even with an overall ΔT=0.8 K at all levels, the shape isn’t quite right Credible SL points 

   Herschel PACS ISO + ground-based Akari Voyager RSS N.B. Herschel PACS (green) spectrum is in error and must be re-done Herschel PACS ISO + ground-based Akari Voyager RSS   

Another approach: ratio the data back to the radiance of Uranus, which is fairly well established. SPIRE LWS Herschel PACS ISO + ground-based Akari Voyager RSS N.B. Herschel PACS (green) spectrum is in error and must be re-done

Conclusions Neptune Uranus Use ISO derived T(p) Use Spitzer-based T(p) Modify to fit PACS HD in stratosphere, Spitzer IRS Modify to fit UKIRT, JCMT data (other data) Compare with Planck data Constrain He/H2 from mid-infrared (SOFIA) Assess time variability Uranus Use Spitzer-based T(p) Modify with H2S or PH3 opacity Compare with Planck data Modify to fit any other data Constrain He/H2 from mid-infrared (SOFIA) Determine seasonal variability

Supplemental Information

Best-Fit Temperature Profile for Neptune

Best-Fit Temperature Profile for Neptune

Temperature sounding using Spitzer IRS data - 1 Start with smoothed version of the Voyager RSS profile (Lindal et al. 1990) Perturb the temperature structure around the tropopause to match 510 – 640 cm-1.

Neptune temperature sounding - 2 Vary the overlying temperature structure as shown to determine an optimal fit to the H2 S(1) quadrupole at 587 cm-1

Neptune temperature sounding - 3 Vary the upper stratospheric temperature structure to fit the observed CH4 emission, accounting for the unknown stratospheric VMR of CH4

Comparison with simulated image based on Voyager IRIS T(p,lat.) VLT/VISIR imaging: 1-2 September 2006 Uranus deconvolved Uranus simulation based on IRIS 2-3 September 2006 Uranus deconvolved Uranus simulation based on IRIS

2009 Thermal Images of Uranus Gemini-S/T-Recs 13 2009 Thermal Images of Uranus Gemini-S/T-Recs 13.2-μm images on Aug 29; Sep 18, 19, 20, 21, 22 VLT/VISIR 13.2 μm images on Aug 5, 6: FIRST IMAGES OF STRATOSPHERIC (acetylene) EMISSION

Spitzer IRS Disk-Averaged Spectrum of Uranus Short High Long High Long Low Short Low 2 Short Low Bonus Short Low 1

Plot vs wavenumber (cm-1) shows spectral features better. Plot of brightness temperature indicates approximate temperatures of depths sounded to determine T(p) ↑ The spectrum shortward of 7 μm is a mixture of thermal and reflected solar radiances so we’ll defer analysis of this portion of the spectrum

• Best-fit disk-averaged T(p) has warmer troposphere than RSS results. • Fit is extremely good

Good news / Bad news: The hydrocarbon variability is consistent with rotational changes, i.e. views of different longitudes with features of different brightnesses - implies a bright or dark feature(s) on one hemisphere No variability in emission from troposphere, so we’re only fitting one temperature profile, and variations are unlikely in the far-ir/mm region (so, as a Herschel calibrator, Uranus will not have rotational modulation) There is no variability in emission from the H2 S(1) quadrupole line, so we seem also to be fitting only one stratosphere, but it should be sensing the same altitudes as the hydrocarbons. Cycle-1 data: Similar behavior but ½ the amplitude.

Spitzer IRS Spectrum of Neptune

Cross-Comparison: Spitzer T(p) Spitzer troposphere is too warm to fit PACS HD, no matter how much we vary the HD/H2 ratio

New data obtained since that time: - CSO and JCMT (Sandel and Dowell) - SMA and VLA (Butler, Gurwell, Hofstadter) - improved precision and accuracy New calibration (Butler, who could not attend this workshop): Transfer of Wilkinson Microwave Anisotropy Probe (WMAP) calibration ±0.2% to several planets, including Mars Transfer to general observations through thermophysical models of Mars (Rudy, Lellouch) Uniform downward adjustment of both models by small percentages matches the CMB The adjustment using these models replaces the wavelength- dependent adjustment required by Weiland et al. (2010) who used a simpler model for Mars - Correction on Rudy is ~ -2%, Lellouch smaller (-1%).

Note: For years, this spectral region was unable to be fit by CIA models for H2, without resorting to bizarre models for bulk composition.

Breakthrough came in collaboration with Magnus Gustafsson, who generated a new ab initio model for H2 CIA and discovered a mistake in the original ab initio work by Borysow et al. (1986) revised H2 absorption model (Orton et al. 2007) H2 S(3) C2H2 C2H6 old H2 absorption model (Borysow et al. 1986)