Notes from SNAP Collaboration Meeting Feb 10, 2002 Michael Richmond [GO discussion at launch is devoted to "how can we get NASA to pay $$ for SNAP, esp. launch?"] 1:51 Mufson Strawman Ballon Experiment for SNAP Primary Calibrators goal is measurements with systematic errors < 1% is it possible? HST can do 2% in optical theoretical atmospheres bad in near-IR what is effect of errors on cosmo params? Monte Carlo calculations 3 or 4 cosmo parameters if obs good to 1% each, cosmo params still have considerable uncertainties see talks tomorrow morning why a balloon? avoid atmosphere nightly variations in extinction are 0.02-0.03 mag in extinction coeffs on ground must beat down errors over many nights even worse in IR, time variable but you also do it from the ground, of course use relative measurements, vs. a calibrated lamp how to build an instrument for balloon payload? compact 15-inch RC, maybe even 12-inch dominated by background, so bigger scope doesn't help fiber-fed spectrometer is okay to have lamp source off-axis because optical design has good PSF off-axis must compensate for motions of many arcsec in flight use star tracker + 2 gyros use rapidly moving mirror to keep light on fiber requires no new tech development what exposure time do you need? dominant background is OH emission in 2 hours, can get good S/N for star V = 8 light source is the hard part must have artificial star with known spectrum must make a lamp look like a star can be done use integrating sphere use off-axis paraboloid to make lamp more like star with a starlike source size look at lamp and star simultaneously switch fibers onto diff places on detector to get rid of variations across detector to calibrate light source to 1%, go to NIST must bring entire instrument to NIST, not just lamp NIST can do 0.5% from 0.4 - 1.7 microns CAN NOT DO BETTER THAN THIS may need to bring an environment with instrument to NIST to match environment at balloon altitude NIST needs 10^-5 Watts/cm^2 in 5-nm bandpass to give you 0.5% accuracy that means we need to provide a lamp < 100 W Q: how correlated with wavelength are uncertainties in NIST calibration? A: we don't know exactly The big problem NIST uses 10^-5 W/cm^2 but mag 5 star star we look at is 2.4 x 10^-17 W/cm^2 so we must attenuate lamp by factor 10^12! don't know yet exactly how to do this geometry subaperture illum shutter 1/r^2 examples of MSFC HERO balloon gondola system have discussed it with MSFC they will let us borrow their designs which stars to use as primary standards? use K giants bright in near IR common stable R&D issues spectrometer design how to make light source look like star? fiber studies active guiding 2:25 end Q: could you do the calibration on ground from South Pole? A: no, still water vapor there 2:28 Fruchter: Ideas for calibrating SNAP starting on the ground Earth's atmosphere has absorption bands in IR long-term ground-based optical photometric studies can provide internal precision 0.005 mag spectrophotometric approach no color terms run every clear night with robotic telescope build 2 at different sites can't use SNAP on 5-th mag star so must transfer calibration to faint stars use spectrophotometry from ground and with SNAP can use balloon-borne measurement to check model atmospheres it "fills in the gaps" due to atmosphere Q: could one use SOFIA? A: their instruments don't match our wavelengths would they let us put our instrument on their plane? Must self-calibrate the detectors requires tight correlation of pixels across detectors must dither in many directions long/thin SNAP field is bad big IR pixels can lead to big errors in photometry used NICMOS pictures of stars observed repeatedly looked at diff in measured mag vs. center position full effect +/- 0.10 mag can alleviate via PSF-fitting? maybe, if one understands the effect also leads to error in measured position comment: collection of charge, and diffusion, both are factors comment: future IR detectors should behave better Richmond: Calibration issues comment: filters from WF/PCI were NOT very different from original transmission after years in space so some filters may not change much comment: calibration brainstorm later today 3:12 McKee: Supernova Trigger SNAP has two modes survey (photometry) takes little time measurement (spectroscopy) takes lots of time need trigger based on photometry alone properties of SNe of different types Type Ia get bluer with time Type Ib/c get redder with time Type II generally dimmer and faster Type IIn can be very bright, like a blackbody can use template light curve for Ia on rise if new object looks similar, trigger BUT don't get stretch factor until after peak so can't use light curve shape as trigger instead, use color difference inferred magnitude based on photo-z of host galaxy shape of rising light curve -- weak rate of change of color differences current trigger software IDL repository generate artificial SNe PAW and KUMAC files simulate observations examples of simulated light curves filter selection chose 9 filters used center wavelength goes 1.15^N, N = 1 to 9 relative abundances of diff SN types if SNe Ia = 1 at z=0.4, then at z=0.4 Ib = 0.4 Ic = 0.4 II(bright) = 4 II(bright) = 4 assume need to suppress Type II by factor 100 shape of light curve weak photometric magnitude is useful but not enough by itself also need color information trigger details mag trigger need first point > 4 sigma over background 4 measures rising monotomic color trigger (B-V) < 0.25 how did it work? discards 99% of Type II -- good Q: how many diff input spectra do you have? A: only 1 input Type II spectrum so far in LC generator trying to get more spectra for input Future directions Type II seem easy to suppress dim, fast rise, color Q: with only 1 input spectrum, how can you know? Type Ib/Ic more difficult need more input data! distribution of colors, rise times, abs mags vs. Type need to ensure efficiency of Type Ia stays high with cuts host galaxy information may prove useful 3:30 end comment: not many type Ia with multicolor measurements before peak using abs mag as cut is like Malmquist bias can correct for it time evolution of color seems like very good discriminant is independent of extinction comment: remember, we don't know redshift of SN well before spectra so hard to get colors right for color evolution comment: earliest measurements of Ia show small change in color only 0.05 mag change in (B-V) A: use (B-R) or (B-I) instead 3:35 Cummins: Dust (this is his first Powerpoint presentation ever!) dust in our galaxy know much dust in host galaxies know little intergalactic dust know nothing Dust in MW confined to disk plane (0.1 kpc scale height) total mass approx 1% of interstellar gas sub-micron particles graphite, silicates, PAHs, ices created by red giants, HB stars absorbs UV, visible, re-radiates in IR History of one grain grain-grain collisions: shatter, or merge ion-grain collisions: sputtering gas-grain collisions: accelerate grain photo-ionization from UV Lorentz force due to mag field etc. complex history! how to describe dust? absorption coeff as func of wavelength optical depth over some line of sight extinction: basically optical depth * factor(lambda) reddening dust absorbs much more at blue than red Rv tends to be large in dense clouds tends to be small in low-density clouds so large grains in dense clouds UV extinction varies greatly in diff lines of sight strong feature at 217 nm weaker features elsewhere in spectrum evidence for silicates, graphite, etc. standard dust model reproduces observations reasonably dust grain sizes range 5nm to 2500 nm number of grains proportional to amount of H in line of sight BUT some real grains are elongated, not spherical as shown by polarization of starlight lab formation of grains from gas yields needles distribution of dust in the MW lots in center, less far out in radius is very patchy can one make model of dusty galaxy? must be crude try bulge + disk following Hatano, Branch, Deaton (1998) assume axisymmetric where do SNe occur? put some in bulge: 1 out of 7 SNe in disk assume follows distr of stars in disk 5 kpc scale length where is dust assume gaussian with scale height 0.1 pc in z radial scale length assumes big possible range of overall density over factor 10^4 Monte Carlo results if galaxy face-on, overall density like MW then big peak near extinction = 0 but long tail going to high extinction because not many long path lengths through disk when galaxy is face on is common result for any inclination of galaxy so, calc cumulative probability of exinction <= some value calc avg extinction as function of projected radius in kpc edge-on galaxies have much higher avg extinction Observational selection effects assume we want to observe SNe to z0=1.7 is a corresponding magnitude limit m0 then nearby SN would be brighter than limit m0 but could have extinction which could knock it out of sample as we look closer to limiting redshift z0, can accept less and less extinction limiting extinction changes 6 mags at z=0.2 0 mags at z=1.7 so can calculate probability of observing a SN (due to dust extinction knocking out of sample) as a function of its redshift can calculate avg. extinction vs. redshift there appears to be a "saturation" in average extinction as one moves to larger redshift conclusions so far observational selection limits avg extinction at high z avg extinction and sigma(extinction) saturate observations of SNe suggest typical host galaxy is slightly less dusty than MW Inter-galactic grey dust no direct evidence for it have tried to explain trends in SN mags with z as due to grey dust see Aguirre papers needle-like grains driven out of hosts by winds in early star formation is speculative radiation transport calculation focus on co-moving photon density, N dust is a sink on N -- aborbs radiation assume LTE between dust and radiation constraints on this idea observed UV-vis background flux measurements observed far-IR background (reradiation by dust) color excesses of Type Ia SNe as function of z SN 1997ff z=1.7 was inconsistent with grey dust but was not "gold-plated" event Cummins has re-done Aguirre's calculations used range of opacities, incl some grey ones calc flux vs. wavelength from background must know CMB accurately to subtract it used range of dust density values in UV, visible, flux SAME regardless of dust density but very different in far IR results dust temp ranges 14-6 K Ab at z=0.5 ranges 0.13 - 0.41 E(B-V) at z=0.5 ranges 0.04 to 0.00 only one model isn't almost ruled out by observations grey down to approx middle of visible conclusions reduction of uncertainties in deep IR background measurements would help a lot no easy way refute Aguirre's hypothesis yet, BUT SNAP high-z SN measurements can test theory severely Q: could you exploit correlation between mean extinction and sigma(extinction) in galaxies? A: yes, probably Q: you are assuming you know the albedo of the dust does not include scattering (as opposed to absorb/emit) if you look at scattering instead, cannot rule out the dust A: yes, I didn't include scattering Q: does it help to have distance-independent dust indicator, and then discard SNe with evidence for lots of dust? there can be systematic errors A: haven't thought about that 4:14 Deustua: Extinction of SNe Ia observed magnitude of SN depends on extinction in MW, host, intergalactic how much does host galaxy extinction affect high-z SNe? if underestimate host extinction, observed mags too dim affects Lambda results reddening makes decline rate slower affects stretch corrections star formation increases at high z so dust content may go up and dust composition may change, too do diff SN types have diff environments, hence diff dust effects? see Rowan-Robinson, astro-ph/0201034 claim is that extinction is underestimated excluded all SNe without pre-max data applied peculiar velocity corrections to low-z SNe considers various samples applies fixed extinction correction for all high-z SNe in SCP sample keeps Phillips corrections for high-z SNe (comes under criticism here) bottom line: Lambda results less than 5 sigma one can use colors of SNe to estimate extinction, and correct one can use projected distance from center of host shows E(B-V) for observed SCP SNe vs. distance from host center see no significant trend look at SCP Hubble diagram based on host galaxy no significant differences bottom line: host galaxy extinction appears not to be big deal but still worth running simulation to estimate expected effects and compare against actual measurements 4:25 end (break) [Michael doesn't pay attention to anything else today]