1. Planning and reducing spectral line projects Anita Richards UK ALMA Regional Centre Jodrell Bank Centre for Astrophysics University of Manchester Many thanks to Rob Beswick and other ALMA/JBCA/e-MERLIN colleagues

2. Choose your line ● Science (up to you...) ● Receiver availability ● Atmospheric transmission ● ALMA Early Science bands

3. Protected radio astronomy bands

4. ● V depends on date – used for broad/extragalactic lines ● V LSR with respect to local 'stationary' point Reference frames

5. ● Older VLA, MERLIN data observed in V LSR ● Most modern arrays observe at fixed frequency ● Check very carefully what you need to specify ● If shift during observations > 1 channel, need to correct during data processing – allow enough edge chans

6. Velocity conventions ● Relationship between measured n or l of line and recessional velocity ● Relativistic: ● V/c = (n 0 2 – n 2 )/(n 0 2 + n 2 ) ● 'Radio' (V << c): ● V rad /c = (n 0 – n)/n 0 ● 'Optical' (V <<< c): ● V opt /c = (l 0 – l)/l 0 ≡ z CO n0 115 GHz ● Even at z~0.05, (V opt -V rad ) ≡ velocity shift > 16 GHz ● My first continuum paper :-(

7. Correction to V LSR Methanol spectrum Each point = 5 mins, one baseline (source is resolved)

8. Corrected to V LSR Corretected spectral features are much sharper as well as at correct velocity

9. Spatial resolution ● Everyone wants good resolution but.... ● Thermal emission at cm wavelengths ● >arcsec resolution (EVLA, GMRT...) ● Thermal absorption ● Must have suitable background source(s) ● Masers ● Can get (sub-)mas resolution – VLBI may resolve-out some emission within ~kpc ● Large area/surveys ● Small dishes/multi-beam/FPA (MeerKAT, WSRT etc.) ● Pointing strategy (Wedding cake, Nyquist, sparse...) – Optimise sensitivity, probably means >>arcsec resolution T b = S 10 29 l 2 /(2k B W B ) S mJy/beam area W B steradians

10. Spectral linewidths ● Individual spectral feature widths ● Thermal lines usually >1 km s -1 – Absorption or emission – Galactic SFR CO few km s -1 – Entire galaxies tens km s -1 ● Masers usually ≲ 1 km s -1 ● ~0.2 km s -1 in cool regions ● Extra-galactic OH can be ~10 km s -1 ● Total velocity spans ● Typically tens km s -1 for individual galactic objects ● Hundreds km s -1 for entire galaxies

11. Broad extragalactic lines OH megamaser >1000 km s - 1 total (Richards+,Yates+) Markarian 273 CO Downes & Solomon 1998

12. Spectral linewidths & sensitivity ● S ∝ 1/√(dn dt ) ● e.g. want 1 mJy/bm line wings ● Linewidth 0.1 MHz, span 3 MHz ● 4 MHz available – s rms4 0.05 mJy/bm in 1 hr – 128 channels gives 0.03 MHz/ch – s rms0.03 = 0.05 x √4/0.03 = 0.55 ● Want s rms0.03 0.2 mJy/bm ● If possible at least 3 channels across the faintest line ● Absorption? Enough continuum to subtract accurately – 128 channels gives 0.03 MHz/channel – s rms0.03 = 0.01 x √(4/0.03) = 0.55 mJy/bm ● Want s rms0.03 0.2 mJy/bm (i.e. 5s rms = 1 mJy/bm) – Time on target 1 x (0.55/0.2) 2 = 7.5 hr

13. Spectral calibration ● Bandpass calibration ● Must reach best dynamic range of target source ● e.g. line peak 100 mJy, s rms 0.2 mJy/bm per chan – Need signal-to-noise at least 100:0.2 = 500 on BP cal source ● Pointlike or good model; must reach required flux density on at least some baselines to all antennas – OQ208 known to be ~1 Jy at 1420 MHz ● Need s rms on OQ208 ~2 mJy/beam – Time on BP cal 1 x (0.55/2) 2 ~ 5 min – Prudent to repeat this several times during observation ● (sub-)mm observing: use BP cal at similar elevation to target ● Phase referencing ● Often use maximum available b/w for best sensitivity

14. Correcting wide-narrow offset ● Bandpass cal 3C84 ● 0.5 MHz bandwidth N – Config. used for target ● 14 MHz bandwidth W – Used for phase ref. ● Phase offsets seen W N ● Derive W solutions and apply to N N ● Derive residual solutions for N ● Apply just this offset to target ● See 2 nd part line tutorial

15. ALMA Correlator

16. Correlator limitations ● ALMA: 4 SB's of 4096 channels (FDM, dual polarization) ● SB widths 32.5 MHz to 2 GHz (including edges) ● WIDAR (EVLA, eMERLIN): ● Narrower SBs can have more channels ● Data rate may create further limitations

17. Full ALMA spectral example

18. e-MERLIN configurations ● Examples for full correlator ● OH megamaser lines and continuum

19. Summary of spectral data reduction ● Edit obvious bad data including RFI ● Apply instrumental corrections (e.g. WVR) ● Process continuum calibration sources as appropriate ● Bandpass and phase-referencing solutions ● Derive wide-narrow phase offset corrections if needed – Check all solutions as you go, image the phase-ref ● Apply solutions from calibration sources to target ● Subtract continuum if appropriate ● Correct target to constant velocity if needed ● Self-calibrate brightest spectral feature if possible ● Apply solutions to all channels ● Image the data cube

20. Editing using autocorrelations AutoCorrelations ● RFI spikes show up best in AutoCorrelations if present XC ● But dont' worry, faint stuff mostly vanishes in XC

21. Spot the RFI ● Some real features are hard to distinguish from RFI

22. 2-D editing ● XC ● XC show spike is weakest when SNR on real lines best ● RFI is time-variable, often polarized ● Terrestrial or peculiar, not sidereal velocity drifts XCXC ACAC Channe l Tim e

23. Calibrated data ● BP table (left) applied along with other calibration to data (right)

24. ● Zoom in on the channels with masers

25. ● Image the cube ● 5-chan average shown U Her OH masers observed with EVN

26. ● Errors in bandpass calibration source cause artefacts in the target spectrum ● Check before you write that Nature discovery paper! Bandpass artefacts Bandpass calibrator Target spectrum Oops !

27. Ringing (Gibbs phenomenon) ● Channels are made by Fourier transforming finite lags ● FT of top hat is sinc ● Worse if bright line near band edge ● Avoid this in planning obs.! ● Cure by Hanning smoothing ● CVel correction may suffice if shifts smear channels by the right amount

28. Strategies for continuum subtraction ● Apply calibration first but not CVel ● If continuum is bright enough to self-cal, select line-free channels and apply that to all target data ● Both line and continuum are bright: uv-plane subtraction ● CASA allows short solution intervals – no smearing ● Continuum and line must be in same spectral config. ● Line is very faint: could subtract in map plane ● Line emission is very extended and source is too faint to use short enough solution intervals in uv-plane ● Or if continuum was observed using broader channels – Interpolate line-free channels in dirty map, then clean – Or just subtract continuum map from each cube plane

29. 2 GHz 20 K Bandpass calibration: hard cases ● Removing instrumental wiggles from amp and phase ● Fring solves for delay (df/dn) and rate (df/dt) together ● Might have to smooth amplitudes to improve SNR Edit Time-dependent f/a calibration Wide/multi-chan/ smoothed cal Narrow-channel calibration ITERATIVE

30. Delay correction ● Delay errors are any non-astrophysical change of phase as a function of frequency ● Caused by instrumental timing errors and troposphere ● Usually smooth/linear – except near strong atmospheric lines ● 1/(frequency interval for a full wrap of phase) ● e.g. wrap in 1 GHz of b/w is 1 ns delay error Corrected Raw

31. Ultra-wide lines Morganti et al. 2003 ● Thousands km/s ● Few km/s resolution ● Very wide, shallow wings ● Need absolutely flat band ● Fitting polynomials is risky ● WSRT flattens bandpass by shifting central frequencies

32. Tutorials ● HI emission from a nearby galaxy ● ATCA low resolution, wide field ● Bandpass and time-varying calibration ● Continuum subtraction ● Total intensity cube and moments ● OH 1667 MHz maser in full polarization (if time) ● MERLIN high resolution, line only ● Bandpass and time-varying calibration – Correct for wide-narrow phase offset ● Self-calibrate polarizations separately ● Cubes in all Stokes parameters