Notes from SNAP Collaboration Meeting Feb 10, 2002 Michael Richmond 9:00 SNAP orbit requirements keep dose rates < 1 rad/year want less than 0.1 hits/exposure keep total non-ionizing energy loss < 1 rad over mission average IR Earthglow < 1 W/m^2 must be << sunlight = 14 W/m^2 slew rate must be reasonable, approx 1 deg/sec avoid sun: look > 70 deg away avoid Earth/moon: look > 45 deg away choice of orbits aluminum shielding stops electrons, but not protons consider dose rate at different points in orbit can vary quite a bit if orbit eccentric, inclined South Atlantic Anomaly can raise rates by approx 10,000 other less extreme changes by factors approx 100 high orbits have less Earthglow but highly eccentric orbits come closer to Earth at perigee, so overall Earthglow increases but only by a little (factor 2 or so) eccentric orbits also increase temp fluctuations due to Earthglow variations only about 0.5 deg K so eccentric orbits are okay as semi-major axis increases, go farther from radiation belts so overall radiation dose decreases 10 Re vs. 5 Re drops rad dose by factor 1000! summary LEO is bad, bad, bad Molnyia (6 Re) bad: big rad dose extended Molnyia to 10 Re helps a lot higher orbits are better HEO are very eccentric 15-30 Re apogee prometheus orbits use lunar assists very eccentric orbit is inclined, too BUT takes extra 200 kg fuel long eclipses -- 5 hours requires more ground stations requires stronger transmission system requires gimballed antenna (or phased arrays?) so HEO-type elliptical orbits are best lock orbit to Earth's rotation dump data to Berkeley station once per day one ground station require little fuel to reach body-mounted fixed antenna BUT thermal stress during eclipse perigee ought to be above proton belt perigee > 12,000 km decreases rad dose factors > 10^5 relative to LEO perigee = 10,000 km is probably okay ground track can be made to pass over Berkeley for 5 hours orbit has apogee about 1/3 lunar orbital radius watch out for lunar perturbations to minimize, set orbit to lunar factors calculate perturbations over 3-year mission can make stable within acceptable limits longest shadows 2.5 hours, but most about 1 hour can keep perigee fixed over Berkeley during mission -- good need to adjust period slightly during mission Q: what if mission lasts more than 3 years? A: doesn't take much extra fuel Q: but thrusters must still be working A: of course. also, required to be able to control re-entry is US Govt requirement? can we get waiver? this orbit doesn't cross geosync volume (i.e. cross inclination = 0) launch delta4 is target launch vehicle can make it with amount of fuel we can fit on board rah! 9:25 Pankow: SNAP mechanics we have found a very good orbit "modified Chandra orbit" launches 26-deg incl, so Cape Kennedy OK in plane of ecliptic orbit (North/South fields okay) requires only launch at specific time of day launch vehicle comparison NASA standard is Delta II we can't fit into it SeaLaunch possibility Boeing didn't purchase so it's out Delta III is the current plan RSDO bus options look in catalog of existing vendors just about any vendor has appropriate product typical bus dry mass approx 500-600 kg IMDC risk assessment SNAP risk on spacecraft bus is low -- good baseline config 700 kg instr, 500 kg bus, 350 kg fuel (hydrazine) power 500 W bus config 3-axis stabile need reaction wheels to steer need star guiders fuel 180 kg to raise perigee 10 kg/yr station keeping 100 kg post-mission disposal light baffle is good solar sail forces momentum wheels to work const. must unload about once/day communications need 74 Gbits SSR need storage for spectr. data need Ka-band downlink not off-the-shelf now ACS driving requirements point 1 arc-sec attitude 0.02 ar-sec jitter 0.02 arc-sec over 200 sec exposure ACS issues bottom line: no new technology Q: what about gyros? A: is star-guider covered by shutter? if so, spacecraft could drift may require pick-off mirrors if star guider not on focal plane, can be problem star guider unsettled OTA primary, secondary, folding flat, tertiary GigaCam, spectrometer 140K camera dewar secondary structure composite materials can set CTE within pretty big range, as needed need lampshade and stovepipe to keep out stray light looks solvable -- good lose 0.8-meters diam in center of aperture baffling is important for SNAP we think we understand it secondary metering structure is tough studies show N-pod structure isn't very stiff N could be 3 to 8 using HST-like structure conflicts with baffles current plan: hybrid N-pod + HST structure active mechanisms focus knob to move secondary similar to flown designs: HST, etc. may need adjuster on tertiary, not sure shutter: boomerang style is attractive but must compensate for momentum may need flywheel thermal build, test, fly warm -- like HST use GaAs OSR striping of hot solar array panels gets rid of heat striking solar panels solar array will by approx 350 K large stray light baffle 180 K smaller internal light baffle 210 K use low-emissivity silvered mirrors primary mirror at 280 K requires only 25 W to maintain secondary mirror 280 K requires 10 W to maintain want to keep secondary structure warm not sure how much power it will take HST takes about 60-120 W camera at 140 K design leaves 8 W margin for maintaining must watch cabling carefully no new technology recent experience with HESSI helps contamination concerns moisture/frost on mirror is primary issue teflon blanketing is stock item use structured cool-down approach make focal plane last item to cool so water freezes elsewhere cyanate ester resins absorb approx 7 ppm water much less than typical epoxies, 70 ppm summary Delta 3/4 launch vehicle not ready today no new tech in SC Bus build, test, fly warm is appealing 9:55 end Q: how fixed is 280 K for temp of mirrors? A: higher temp requires more power Q: dumping momentum daily via gas jets? must worry about contaminating mirrors! A: we're not a UV mission, so not important propellent is mono-hydrazine jets are at base of telescope, pointing away from optics gas leaves with Mach 5 9:58 Nugent Supernova Simulations from theoretical side 1. core collapse working group 2. Ia working group want to keep systematics down to acceptable levels using information from Hoeflich document will be produced on WWW in next few weeks Core Collapse WG how to eliminate confusion with Ia? how to use SNAP core-collapse data set for cosmology? SN Ia WG what are observational requirements for SNAP to meet cosmo params? Confusion with core-collapse core-collapse rate approx 5x Ia rate goes up as SFR increases half are IIP, half are IIL note that core-collapse Ib/c occur at 0.5x Ia rate locally need multi-color photometry on rising side of light curve colors split types light curve shapes IIP very obvious, unlike Ia easy to distinguish very blue early fainter than Ia some IIL have rise times similar to Ia very blue in UV Ib/c are a major concern lost most of envelope similar rise times to Ia can be as bright as Ia spectra not unlike Ia occur in late-type galaxies not a lot of good light curves now comparison of spectra Type II have strong H Ia have strong abs Ib/c have fewer features but can resemble Ia at peak need obs at different times to distinguish Type IIP cosmology use Expanding Photosphere Method (EPM) need photometry to few percent in several filters need spectra at several times, to get expansion velocity R = 250 during plateau phase is needed spectral features narrower than Ia 75-100 Angstroms wide IIP are approx 2 mag fainter than Ia large dispersion in peak brightness, sigma > 1 mag can only study at z < 1 IR obs nice removes dust signature IR SED closer to blackbody than visible Type IIn have narrow lines can be very bright > 1 mag brighter than Ia idea: core-collapse which smashes into circumstellar envel almost perfect blackbodies very hot, 40,000 K could be used to measure dust evolution Chugai has method for finding distance but not many of them Q: can we be quantitative with II? A: tough, not many objects observed well need more data Cosmology with Type Ia what do we currently know? what must we understand to make SNAP success? Temporal evolution see different parts of object at different times can parametrize light curve shape based on "stretch" or other diff between peak/tail brightness not same in all Ia not many SNe with good data on this diff can be 30% indication of secondary parameters in explosion Differences between SNe Ia in light curves, in spectra understand theoretically pretty well Si II ratios show differences between objects correlates well with "stretch" factor don't know if provides extra information hotter ones take longer to rise need to measure redshift accurately velocities of expansion vary between Ia diffs up to 5,000 - 10,000 km/sec need to measure velocities accurately correlates with diffs in UV spectra if metallicity of progenitor changes, UV flux changes high metals push flux UV -> red have models, but not enough observations yet Parameters which control Ia SED total mass of Ni56 total mass of progenitor KE of explosion overall opacity all are function of 3D position and of time observables spectroscopy requirements spectral feature velocities: 150-250 km/s well, maybe 400 km/s okay typical velocity 10,000 km/s so this is 2-4% measurement spectral feature widths: 10 Angstroms spectral feature ratios: 5 percent Q: doesn't current state of art use only light curve shape and color to parametrize? A: yes, true but there are other observables which might work so we must be ready to use them, if they pan out Q: yes, but these speculative needs are driving mission! A: true enough photometry requirements "stretch": 1% light-curve rise time: 0.3 days peak-tail ratio: 0.05 mag Available simulation data Arnett's bolometric light curve code temporal evolution tools can use spectrum to judge age velocity at max light theoretical ranges 5000 - 25000 km/s temperature evolution metallicity evolution all of this code now availble on WWW What we need SN 2000cx lght curve is wierd doesn't fit any stretch relationship could be due to aspherical explosion has peculiar spectrum, too has strange color evolution -- much bluer need more simulations to see how wierd Ias can be used should discard them? or can use them? should we worry about them? 10:30 end Q: how many other wierd SNe? A: only one other like 2000cx comment: SN factory must collect big sample of these and see how they fit into Hubble diagram A: yes Q: how can we "sell" the diversity of SNe Ia? isn't this a problem for review? A: diversity helps us to understand systematics Q: but diversity today may not be same as diversity at high z A: true Q: is there an early observable which distinguishes these wierdos? so that you can look at them closely A: polarization might be one Q: the point is to measure everything you can possibly measure and then use big multi-factor fit to account for dispersion in absolute magnitude comment: skeptic might argue - SN community not at that stage yet so don't build SNAP yet A: with multicolor photometry at few percent, can diagnose the wierd SNe Ia so SNAP can do the job Q: cautionary tale: a few years ago, planetary formation theorists thought they understood things but then we discovered strange big planets close to stars so do you REALLY understand SNe? Q: how many SNe Ia with good data today? A: about 10 nearby ones, 0 distant ones Q: well, once we have 100 very well measured SNe, can be more confident that most or all behave properly Q: to get 100 very well measured SNe in next few years, need more big telescope time, because you need to move out to larger redshifts A: yes A: must be able to use observables, not theory, to reduce dispersion Q: one year ago, SNAP requirement on spectroscopy was R=2000 A: no, it never was [hmmm] comment: is it okay as long as one can break SNe into groups, based on observables, such that dispersion with each group is at 1 percent level? that is, show 1 percent empirical dispersion with no theoretical explanation A: is it good enough? not sure A: must be able to distinguish outliers using observables so can discard them from analysis comment: well, is SNAP a dedicated Ia-only mission, or is it a multi-function mission (including lensing)? is better to have multiple functions A: no, because can use Type II SNe as well as Ia comment: different populations of SNe are okay, BUT only if they don't evolve with time i.e. ratio of diff populations don't evolve with time comment: SNAP provides so much data, one can break it up into many diff groups and analyze separately A: yes, but concern is conspiracy of systematics comment: no one will believe cosmological parameters based on one method alone, no matter what that method is A: yes, but SNe may be one of the very few tools which does test dark energy in efficient way A: need to assemble a panel of skeptics which will give SNAP members a tough time comment: I've looked at 200 SNe, and only 1-2 percent of them are wierdos A: yes, we must emphasize that most SNe are "normal" 11:01 break starts 11:20 Bernstein: SNAP spectroscopy why are we doing it? SNe have some distribution of properties manifest observationally mainly as stretch factor plus smaller terms we measure broadband magnitudes distance, intrins. mag, extinction turns mags into colors need to know extinction law requirement 0 distinguish Ia vs. other ESSENTIAL requirement 1 use spectral information to constrain properties to determine intrinsic color, so we can constrain extinction so we can determine abs mag ESSENTIAL set requirement in extinction sigma < 0.02 implies learn metallicity of SN to 0.32 dex this is based on theory allowable random error in abs mag per SN is still unknown Q: can't you average SN in redshift bins? A: maybe requirement 2 use spectral lines to constrain params which might affect intrinsic properties at secondary levels ESSENTIAL if such params exist requirement 3 use spectral lines to check if intrinsic params vary with redshift at fixed stretch check on evolution DESIRABLE may use only on a subset so, what S/N ratios in spectra are required to do this? use theoretical spectra, Fisher matrix analysis results: for perfect, noiseless detectors as R increases 30 -> 100, increase information but beyond R = 100, no more signal gain don't know which piece of spectrum is most important could be important to decide spectrograph design 1-channel vs. 2-channel conclude: R > 100 doesn't help look at count rate on detector for SNe at diff redshift assume R = 70 assume read noise integrated over 7 pixels z=1.5 beats zodiacal light at lambda > 1.0 micron beats read noise at lambda > 1.3 micron beats dark current at lambda > 1.5 micron z=1.7 beats zodiacal light at lambda > 1.3 micron beats read noise at lambda > 1.5 micron beats dark current at lambda > 1.7 micron noise sources comparable to signal from distant SNe so, changing spatial pixel scale doesn't help conclude: major source of noise is the pixels use fewer pixels per spectral elem! what if detector has noise, is not perfect? now, signal goes DOWN as spectral R increases for metallicity, optimal point at R = 70-100 conclude: best design has R = 70 per pixel = 35 per resolution element? = or 140 per resolution element? where R defined as (delta lambda/lambda) spatial size of pixels doesn't matter too much at z=1.7, require exposure 15 hours to get S/N = 20 per pixel with R = 70 spectrograph rough guide: exposure time rises as (1 + z)^6 doesn't agree with Frogel's values but still big uncertainty in these estimates (some discussion of how much gain by making 2-channel spectrometer) (maybe gain by factor 2, not 10) further necessary steps analyze relation between photometric and spectroscopic info what is really needed from spectroscopy? need to find how intrinsic physical factors lead to photometric observables lead to spectroscopic observables want to diagnose degeneracies in photometric-only Q: must remember that theoretical spectra are not TRUTH these are just guides, are not perfect A: yes, so must be conservative in analysis and count on more uncertainty in real life Q: what about independent group to make spectra? A: nope, Hoeflich and Nugent are the only producers A: SNAP needs better models to be made to guide analysis need better templates Q: suppose SN factory produces lots of SNe and can measure correlations very well reviewer will say: wait for SN factory A: yes, this will come up Hoeflich believes dependencies are good to factor 10 can't be off by factor 100 is based on solid basic physics choice of reference rest-frame filter B is not optimal, due to influence of dust and strong effects of metallicity in blue moving the blue edge of filter is most important observationally, SNe more homogeneous in the red sub-luminous and extinguished SNe require longer exp times can one discard them without introducing bias? must couple time estmates with mission operations and cosmo-param-estmation code in order to calculate mission duration vs. telescope aperture and operational strategies 12:05 end Q: what if S/N in spectra drops from 20 to 15 or 10? A: estimate of distance modulus would have larger uncerainty grows linearly with S/N [I think] Q: systematics which are understood aren't a problem A: yes, but is important to verify that spectra at high-z are really similar to spectra at low-z Given current Bebek model for camera, and observing plan what is S/N for point source in 1 pass? for canonical Type Ia SN z = 0.5 S/N = 100 z = 0.9 S/N = 50 z = 1.5 S/N = 30 what if you survey 300 sq. deg in one year, full-time imaging S/N = 10 at AB mag = 28 has table of point source S/N under various operation scenarios is important for weak lensing science 300 sq. deg 1-year survey S/N = 10 at AB mag 27.5 point source 95 galaxes per sq. arcmin at ellipticity limit yields total 100 x 10^6 galaxies with ellipticities 10 sq. deg 2-year SN-only survey S/N = 10 at AB mag 29.3 point source 230 galaxies per sq. arcmin at ell. limit yields total 8 x 10^6 galaxies with ellipticities turns out SNAP isn't greatly better than LSST for weak lensing studies (using some metric) Q: why is SNAP not overwhelmingly deeper than HDF? A: SNAP gains in area, not greatly in depth Q: what about photo-z issues? [Ellis keeps asking about this] A: takes more analysis going into near-IR helps for red objects -- good Q: does this mean we should add more CCDs to focal plane with a single filter? A: no, for lensing, need photo-z, so need diff filters 12:19 Frogel: Spectroscopic exposure times: SNAP, NGST, VLT statistical formalities spectroscopic exposure times for each instrument statistical formalities basic S/N equation ignoring background due to host galaxy! when background is significant, noise contribution increases unless you use many pixels to find its level remember: ignoring host galaxy! issues: resolution of spectrograph signal-to-noise needed for science don't confuse 'per pixel' with 'per res elem' what is required accuracy of redshift for SN and host galaxy? can redshift of galaxy be determined with SNAP? what is max exposure time? need to split to discard cosmic rays exposure times with SNAP instrumental noise assume dark 0.016 e-/sec/pix assume read 4- per pix similar to NGST values boundary conditions background mainly readnoise, dark current, not zodiacal SN Ia at z=1.7 used in tables below key features: peak of spectrum, trough of Si II assume 2-meter telescope assume throughput = 0.35 neglect background from parent galaxy what are the exposure times? if use 4 pixels in aperture 2 in spatial, 2 in spectral each exp 1000 sec S/N = 20 requires 5 hours each exp 1000 sec S/N = 15 requires 3 hours noise dominated by instrumental effects if use 8 pixels in aperture 4 in spatial, 2 in spectral assumes that 4 spatial pixels gets all the signal each exp 1000 sec S/N = 20 requires 9 hours each exp 1000 sec S/N = 15 requires 5 hours if use 14 pixels in aperture (like Bernstein) then estimate each exp 1000 sec S/N = 20 requires approx 19 hours so Frogel and Bernstein estimates are reasonably close need to have people hash all this out later conclude: R = 150 and S/N = 20 would take too long if z=1.3 instead of z=1.7, exp times drop to 3 hours total what about NGST exposure times? is larger telescope z=1.7, S/N = 15, R = 100 per pix, need less than one hour conclude: exposure times less than one hour Frogel and NGST calculations agree pretty well would require between 800 and 4700 hours of NGST to get spectra of 2000 SNe found by SNAP (depending on S/N per pixel) doesn't include setup time doesn't include slew time 90% of the integration time due to SNe at z > 1 remember, these are optimistic estimates NGST is 10x more efficient than SNAP for most distant SNe what about ground-based observing? seeing atmospheric extinction OH emission need to go to near-IR for distant SNe predictions for z=1.7 SNe using VLT with ISAAC using R = 1600 using seeing = 1 arcsec using 4-hour exposure big atmospheric absorption at some important places does adaptive optics (AO) help? still depends on site seeing limited FOV example of AO Gemini has FWHM go 0.8 -> 0.09 arcsec in K' helps to use multiple guide stars which hasn't been done yet in practice helps to use OH suppression system or to use high resolution spectrograph but ground is always factor 10 worse than in space in near IR conclude: ground-based observing is tough AO may help, but only limited field don't know if photometry good to 1% OH suppresion doesn't solve Q: what if you only need spectra for features, not colors? is ground-based AO spectroscopy practical? A: would require 100% time on Keck killed by times when seeing is poor 12:55 end [Michael skips GO discussion at lunch]