Notes from SNAP Collaboration Meeting Feb 9, 2002 Michael Richmond (afternoon session) 1:34 PM Prieto: Spectrograph preliminary specs 0.35 - 1.7 microns note that NGST goes down only to 0.6 microns R < 150 efficiency CCDs lose response at nearIR takes into account detector plus optical elements note that mirror reflectivity decreases to blue in visible down to about 0.5 at blue edge includes prism about 0.8 at all wavelengths includes slicer about 0.9 at all wavelengths overall response SNAP doesn't have to do as well as NGST design skepticism that NGST spectrograph will meet specs LBL CCDs have peak farther the red than other CCDs [sorry, I couldn't follow this] performance vs. wavelength BK7 prism can reach R = 100 in nearIR (?) field of view 2.16 x 2.16 arcsec square 18 slices at 0.12 arcsec wide place 18 slices side-by-side on a 1024x1024 detector like echelle two channel or one channel for spectrograph? 2 channels better QE better spectral R detector noise better for CCD same complexity costs more need dichroics introduces break in spectrum 1 channel one detector can't perform as well across the entire range of wavelengths (some discussion of the effect of dichroic break (bad) vs. improved QE, R in blue and red (good) ) comment: noise in near-IR detectors is less well known so IR-side involves more risk development implement dithering optimize optical design decide on implementation R&D needed on slicer unit dithering mechanism Q: can one just move spacecraft? A: no, that's not good enough for spectral dithering Q: can one move charge electronically on CCD? A: that's not difficult it's the IR device which is tough Q: why do you need such a large field of view? A: using a slit would mean spacecraft would have to place SN on the slit precisely is very difficult for spacecraft so, take spectrum across a large area A: also, a slit leads to lost light make spectrophotometry very hard conceptual design involves 10 mirrors [I think, I count 8 in picture] uses prism with reflective coating at back prism makes device more compact device is size of a sheet of paper all mirrors are spherical could put a dichroic at back of prism, and then place one camera behind prism in a 2-camera design so 2-channel design not much bigger than 1-channel design where does spectrograph go in the spacecraft? is perpendicular to main focal plane think has a good place to put it development Work Breakdown Structure (WBS) has 8 main paths can make optical bench ourselves, or purchase it slicer must be made ourselves, is complex detectors mechanisms can be made ourselves have facilities to do integration and testing [good] philosophy make breadboard first then structural thermic model, optical engr model then proto-fly model run full tests then final modifications ... end up with flight hardware slicer progress validate 2001 manufacture validate slice stack make pupil mirror diffraction analysis some light from central slice falls into neighboring slices only 2-3 percent (?) effectively cross-talk must measure this effect in progress 2002 pupil mirror line assembly mechanical design scattering light analysis to be done 2002-2003 ESA proto tests space environment test performance tests results so far appears to yield efficiency > 93% at 0.5 um good news 2:14 end Q: your system stretches image 2/1 along axes A: idea is to put each element of sky onto two detector pixels otherwise, lose spectral resolution (?) Q: what is increase in cost if one places TWO spectrographs on SNAP? redundancy use one spectrograph for high-z SNe, other for low-z SNe and have two different designs optimized for each A: costs twice as much need two optical models there is plenty of room in the spacecraft for two boxes 2:18 Linder: Simulation/Theory Alex Kim, group leader, not here today big group 15 people teleconfs every 2 weeks meetings once per year active hypernews forum: read the minutes Objectives 1. short term address CD0 questions 2. long term create simulated pipeline of experiment to test lots of possible effects theoretical issues with dark energy Theory working group will deal with them how to determine cosmological parameters listen to talks on Monday using Fisher matrix and Monte Carlo optimal redshift range need z=1 to z=2 the high-z SN community agrees open question about how low one must go z=0.4? z=0.6? lots of ways to play with distribution with z Q: note that z=1.7 takes lots more time than z=1.5! A: yes, of course must make tradeoff connection with SNAP error budget need to include existing light curve simulators and fitters SNAP error budget listen to talks tomorrow (Sunday) Alternative missions how well can ground-based project do? how well can NGST do? Spectroscopy trigger when to switch to spectroscopy? see talk tomorrow survey strategy long/thin vs. square North, South field focal plane layout shutterless design is dead. [rah!] observing requirements see talks later during the meeting see written contributions Peter Hoeflich has good models of SNe can a 2-m telescope really meet the spectroscopic spec? depends on the spec undersampled photometry see PASP paper by Gary 2:27 Linder: Theory working group results [Michael misses some of this] [in fact, Michael tunes out of this one almost entirely. Sorry] anisotropy most models predict isotropy within visible universe other models do predict anisotropy but SNAP isn't the right tool don't need to go to high redshift use CMB or full sky survey sound speed doesn't affect SNAP observables other wierd theories can test some of them by observing a few SNe at z=2.5 summary: theorists have caught up with and passed observers need data! 2:44 Ellis: Weak Lensing CalTech, Cambridge, Edinburgh, Michigan weak lensing can reveal a lot about distribution of mass in universe current lensing experiments are primitive can only do a good job in clusters complements other methods in constraining Omega, Lambda, etc. test of models of structure formation with time observing strategies deep survey 20 sq. deg HDF depth > 100 gal per sq. arcmin wide survey 300 sq deg R = 27 60 gal per sq. arcmin super-wide survey 1000 sq. deg R = 27 60 gal per sq. arcmin need to know what is redshift distribution of background galaxies -- don't know now know ellipticities of galaxies at HDF depth have some methods for inverting the signal to get map 3-d maps in area where signal is strong concerns over using photo-z can break degeneracies in Omega-w plane SNAP does better than WHT [no surprise] Taylor: use strong signal to make 3-D map if clusters lie in front of each other, can reconstruct potential in 3-D better from space than ground lensing power spectrum can reveal details in non-linear regime with faint galaxies issues raised by review panel to what extent can SNAP wide-field lensing complement the SN data? have some answers what are advantages of SNAP over ground-based facilities? have spent lots of time recently thinking about this what are design tradeoffs for SNAP? good for lensing vs. good for SN how to make cost-benefit analysis? Ground-based efforts vs. space-based efforts are currently in agreement but errorbars are huge can't distinguish interesting models from each other just confirms what we already knew actually, a bit of discrepancy vs. cluster counts in some recent results what are limitations on the method from the ground? PSF anisotropies across frame can remove by using stars as true round objects is most important systematic error on the ground at very faint levels, very few stars to use reach limit of ability to remove anisotopic effects and always far above actual signal for areas > 100 sq. deg, ground based surveys limited by systematics these residual systematics are comparable to signals corresponding to 10% change in Omega on arcmin scales so it doesn't help to do larger surveys from ground space-based systems are better larger surface density of resolved galaxies reduced PSF smearing lower background more shape information examples of simulations of space-based images summary good progress in addressing "how complementary is lensing to SNe?" ditto for "what are limitations to ground-based telescopes?" will work on HDF-based simulations So, how best to sell lensing in context of SNAP? is it a way to break degeneracies in cosmological params? or is it a way to map dark matter? 3:08 end Q: why nervous about photo-z? A: [I couldn't figure out his answer] comment: tradeoff between area (for lensing) vs. repeated measurements (for SNe) could be strong comment: would like to have quantitative values for observation time, area, etc. needed for weak lensing science Q: rotation of field throughout orbit -- could it cause systematic errors in weak lensing? A: haven't addressed that yet Q: to obtain 1-2% on Omega-M, how much area do you need? A: need 300 sq. deg so CFHT can't do it, right? true -- not to that precision but other projects _will_ improve our knowledge Q: should we consider 40,000 sq. deg field == entire sky? like EUVE A: from lensing point of view 20 sq. deg. very deep is terrific 300 sq. deg. necessary for some purposes complete sky coverage not so important for lensing comment: can cover 2 sq. deg. per day with SNAP [oops, did I get that right?] comment: can improved ground-based telescopes do a better job than your current projections? A: we don't understand the source of the PSF anisotropies they aren't robust or reproducible comment: isn't it due to the atmosphere? [Ellis doesn't say] A: Subaru would be best telescope to investigate these effects Q: what is seeing in the current ground-based tests? A: about 0.6 arcsec at Keck 3:58 Destua: Calibration crucial point: ratio of fluxes across visible and IR is not (yet) available so can't (yet) calibrate SNAP photometry: broad passbands doesn't give information on SED within each broad band many different existing photometric systems spectrophotometry: narrow passbands or spectrograph can use it to produce synthetic phometry in any photometric system we want to know the ratio of the instrumental response between two wavelengths to < 1% example: Landolt UBV(RI)c system covers visible only internal precision to better than 0.01 mag can stellar atmospheric models to provide calibration fluxes? yes and no yes can use to interpolate between observed data in atmospheric absorption bands white dwarf models are consistent with data HST uses white dwarf models only in visible no complicated stellar atmospheres aren't fitted to within a few percent moral: interpolation is okay extrapolation is not steps to establish spectrophotometric calibration for SNAP set up bright standard stars with spectrophotometry transfer calibration to fainter stars requires 3-4 steps want wide range of stellar types use existing standards as sanity checks such a set of standards would be valuable for other ground-based and space-based programs need to have KNOWN instrumental response functions will have several instruments in the calibration process small telescopes for bright stars big telescopes for faint stars may use balloon, space in addition to ground observe primaries from balloon and/or ground observe secondaries on ground observe tertiaries and fainter on ground and SNAP ground vs. balloon ground: telescope can be larger can see fainter stars must interpolate across absorption can use existing facilities but cross-platform systematics cross-instrument systematics atmospheric effects balloon: telescope must be smaller above most of the H2O and aerosols removes most atmospheric effects but can only see bright stars V = 6 still strong OH background in near-IR reference source must go up in balloon R&D build/characterize telescope + spectrograph reference lamp balloon issues systematic errors during transfer process will start tests summer 2002 must identify vendors for instruments on-board calibration R&D focal plane configuration affects frequency of calibration fixed filters can increase calibration needs brightness limits how bright can one go with SNAP? how faint? shutter systematics diagnostic tests must monitor over time how does calibration program affect science cadence? may not be best for SN science or wide field science trade studies optical telescope design Mufson will discuss balloon, lamp also have ground-based lamp design telescope platforms spectrograph design use existing device? which possibilities for calibration will we use? understand the costs includes time in orbit status have developed some telescope concepts have test run at WIYN, KPNO in summer 2002 have started identification of sources of systematic error schedule hardware design and construction 1-2 years design observing program 1 year carry out observing program 3-4 years develop calibration diagnostics, and database continues during operations must check stability of instruments during mission note that immediate comparisons to SNe are mag = 25, takes a long long time to observe even with big telescopes personnel have 8 people part-time, need more 4:19 end Q: what about stellar models? A: current WD models are pretty good, though absolute calibration is still unsettled no plan for us to do stellar models ourselves comment: 2% errors in HST calibration means ... [didn't catch it] Q: will you bring SNAP cameras to ground-based telescopes? A: not sure yet they won't exist for a while, anyway Q: what happens if the calibration fails? A: we've looked the consequences - they are severe depends on exactly you want to know Q: using several filters can help to average out errors A: yes, it helps Q: what about flatfielding? A: can do it, but it steals time from other observations may conflict with other SNAP programs Q: can we use the time we're not observing to do some calibration? Q: has anyone placed an interference filter into radiation to see how it degrades? A: we aren't sure, should find out 4:32 Carithers: Computing and data handling Chris Day will talk, too concentated so far on a framework prototype evaluting STScI products so far, have concentrated on creating simulated images if we save every exposure we take, is 1 PetaByte so other science is not necessarily "free" plan: use scalable array of commodity PCs next steps need to involve larger collaboration set up communications with collaborating groups to match needs with resources let's discuss these issues Tuesday morning joint meeting of Sim and Comp groups 50A-5132 then Comp group meets in 50B-4205 must learn platforms, languages, resources needed 4:37 Day now speaks: SNAP software activities satellite -> data farm -> archive -> trigger on SNe -> scheduling SNAP will not discuss Archive now will be necessary to re-run data through flowchart multiple sets of data running through system data capacity (models below assume data flows from satellite continuously) (current idea is that data flows only a few hours per day) (computing team hasn't caught up to latest changes to mission) steady state model 20,000 CCD images per day calibration: 1 min co-add: 2 min search: 2.5 min use 200 processors gigabit Ethernet current PDSF-class CPUs this isn't an insane model, using today's technology what if power goes out for 10 minutes? network jams CPUs idle while waiting for data doesn't recover well bottom line is that one must budget on things going wrong data farm: OPUS 2 pipeline model wait for input data process modifies input depending on result, writes to new location other process grabs data multiple pipelines can exist at once multiple pipelines can run at once multiple nodes can run processes easy to scale this model OPUS used for several existing satellite systems at STScI long history supported free have some questions about scaling to larger data flow current max number nodes = 4 SNAP might require > 100 nodes Day shows OPUS running on his computer, right now data farm: OPUS 3 hitch: current programs use FTP is security sieve not allowed at LBNL STScI is waiting for next release of Java in meantime, consider Grid as alternative approach is popular with funding agencies right now Grid software: Globus goal is to cause groups which don't trust each other to work together have network ID certificate for users user with certificate is granted access to resources in particular, enhanced FTP software an use certificates for secure file transfer is a political issue between sites has set up a system on local machine incorporating into OPUS takes some changes not as easy as hoped? investigating a hack to use as proof of principle SNAP campaign sketch input from trigger into SNe Campaign Factory refers to library of SNe Curves and Spectra, archive of images produces request for spectrum request goes to Observation Planner Campaign Manager deals with photometric observation cadence SN Trigger manager see presentations by others tomorrow must be able to withdraw requests if learn not a SN Ia Observation planner takes in proposed observations applies constraints produces command stream to satellite have STScI/FUSE demo running here at LBL called "SPIKE" specialized versions for each satellite is well known is written in LISP (!) both long-term and short-term planning GO programs archive GO is a source of archive requests archive scheduler doesn't care satellite GO observation scheduler doesn't care archive scheduler doesn't care could impact on satellite operations GOs are their own campaign managers Q: SDSS underestimated manpower needs for computing is not glorious, doesn't get attention A: yes, happens in High-Energy Physics, too will SNAP own its computers? probably not could be a problem Q: do you have a plan for this effort? A: have a project plan, yes Q: campaign manager is mission-critical how to ensure that it works properly? will mockup of data help? will it help to have ground campaign data? A: possibly we thought of embedding signal into real images eventually, could make use of real data Q: computing could impact science plan of SNAP could put constraints on operations important to get pipeline working early Q: should try to factor in all known effects must not put spacecraft into safe mode! Q: LBL has experience building end-to-end simulators Q: NGST will spend $$ to create a test system that will run the actual ground command system software will build up at same time as hardware Q: HESSI did something similar (techical difficulties) 5:25 Pain: SNAP in France who are we? what are we doing for SNAP? what else could we do? SNAP in France funding Who are we? FRench Observing Group of Supernovae (FROGS) Physicists at CNRS astronomers from INSU scientists from CEA/Saclay long-time collaborators with UC Berkeley Babar, SCP What are we doing? 1997 in Paris high-z SN with SCP 1999: SN factory project at Berkeley very detailed spectroscopy of 300 low-z SNe UH 88-inch telescope + SNIFS instrument SNIFS: 2 channel IFU spectrometer R = 1500 blue arm: 2000x4000 Marconi CCD red arm: 2000x4000 LBL CCD imaging and guiding: 2 2000x4000 CCDs 2000: joined SNAP SNAP: electronics R&D immediate goal: test CDS1 prototype (DMILL) see talks by Barrelet and Genat later during meeting SNAP: building the IFU R&D slicer INSU + IN2P3 consortium optical design (INSU) readout electronics design and software (IN2P3) see talk by Prieto SNAP: science and software science: SN Ia for measuring dark energy deep galaxy survey software: simulation of rates, spectrometer performance, lightcurves, fits to cosmo params image processing algorithms What else could we do? optical system competent French Cies, Astrium/Matra, Sagem/Reosc electronics experience in building large CCD arrays EROS, MegaCam (1 sq. deg. = 40 2Kx4K CCDs) software: experience in image processing SN factory CFHT LS = 4 sq. deg. every 2nd night for 5 years expect approx 1000 SNe Ia at 0.3 < z < 0.9 science: weak, strong lensing SNAP in France currently approx 15 scientists = approx 5 FTE support from CNRS for science and electronics first presentation at CNES Nov 2000 seed support for science and IFU design second presentation Nov 2001 considered French space missions long-term given a set of questions tech, science, organizations, etc. review in Oct 2002 full support? 5:36 PM end Q: what is the single most important issue to get funding? A: we have a space agency, not an energy agency Q: anyone from Saclay involved? A: only one person so far Q: what happens after Oct 2002? A: could be given lots of $$ if ranked high on list depends on what happens in US funding Q: would it help to have letters from NASA? comment: no, you won't get them from NASA right now wait until Nov 2002, when Road Map is created NASA has plan for next decade, SNAP isn't in it Q: it might be dangerous to have French agency contact NASA about SNAP Q: I attended GLAST baseline review recently still no signed agreement between DOE and NASA to do GLAST (which is years beyond SNAP) it was clear that major issue was uncertainties in French funding do you know what those were? A: I think French support for GLAST is pretty well defined