Spectroscopy with LIFU
WEAVE's LIFU fibres feed a dual-arm (blue + red) spectrograph housed on one
Nasmyth platform of the WHT. A 5900-A
dichroic splits the light between the blue and red arms.
Dispersion is effected by inserting
one of three VPH gratings in the blue arm, and one of two in the red arm,
i.e. the two arms offer a total of five spectroscopic modes (blue low-res,
red low-res, blue high-res 1, blue high-res 2, red high-res).
The table below lists the
wavelength range, spectroscopic resolution and scale (Å/pixel)
expected for each. The spectroscopic resolution delivered by LIFU is
half that
delivered by the WEAVE MOS and mIFU modes, because the LIFU fibres have twice
the diameter of the MOS/mIFU fibres.
| Low-resolution |
High-resolution |
| Blue arm |
Red arm |
Blue arm |
Red arm |
| VPH grating |
'B LR' | 'R LR' | 'B HR1' | 'B HR2' | 'R HR' |
| Wavelength range (Å) |
3660-6060 | 5790-9590 | 4040-4650 | 4730-5450 | 5950-6850 |
| Inter-CCD gap (Å) |
5491-5539 | 7590-7669 | 4525-4536 | 5302-5315 | 6412-6431 |
| Spec. resolution R for LIFU |
2500 | 2500 | 10000 | 10000 | 10000 |
| R in km/s |
120 | 120 | 30 | 30 | 30 |
| Scale (Å pixel-1) |
0.30 | 0.48 | 0.076 | 0.090 | 0.11 |
| WHT/WEAVE throughput (expected) |
~ 0.25 |
~ 0.15 to 0.20 |
The spectroscopic resolutions currently being achieved routinely by LIFU
(at the
middle of each wavelength range) are R ~ 3000 in low-resolution mode
and R ~ 11000
in high-resolution mode.
In high-resolution mode on the blue arm,
observers have a choice of two
possible VPH gratings, for wavelength range 4040 - 4650 Å (HR1 = 'blue') or 4730 - 5450 Å (HR2 = 'green'); they cannot be deployed
simultaneously.
For any given observation, low- and high-resolution
modes can't be mixed, e.g. it's not possible to observe
at low resolution in the blue
arm and at high resolution in the red arm.
Also, a blue-arm VPH cannot be used on the red arm, or vice versa.
The three allowed combinations of
blue (B) + red (R) configurations for observing are thus:
'B LR' + 'R LR';
'B HR1' + 'R HR' and
'B HR2' + 'R HR'.
The pixel scales observed during commissioning are close to the values
given above.
The figure below (click to enlarge) shows the
the wavelength range (and inter-CCD gap) of each
of the 5 available spectroscopic configurations (in blue and red),
relative to the wavelength of the dichroic cut (5900 Å, green dotted line)
and
a typical
La Palma night-sky spectrum (black).
The vertical black dotted line indicates the wavelength of the most prominent
atmospheric absorption (A band, ~ 7594 Å), which falls (by design) into
the inter-CCD gap when observing at low-res on the red arm.
On each arm, the spectra fall on a mosaic of two 6144 x 6160 pixel
deep-depletion EEV CCDs,
i.e. a total of 12288 x 6160 15-micron pixels (total area ~ 90 mm x 180
mm).
The x direction
(physically, parallel to the surface of the optical bench) is
spectroscopic, the y direction is 'spatial', i.e. from fibre trace to fibre
trace (physically, vertical). In each arm, the
inter-CCD gap is about 2.5 mm (equivalent to 170 pixels in x) (TBC).
In each arm ~ 8200 x 6140 pixels are used for science.
On the blue arm, the spectra cover roughly 0 < x < 8k;
and on the red arm roughly 4k < x < 10k.
The images
below show LIFU low-resolution ThAr arc-lamp exposures taken with the blue arm
(top, wavelength increases to left)
and with the red arm (bottom image, wavelength increases to right).
The exposure times were 25 and 9 sec respectively.
These images have been cropped to show most of the useful wavelength
range in the x direction (7k, 8k pixels),
and approximately the central third of the slit (2k pixels) in the y direction.
Each white dot corresponds to an arc line (vertical bright line) detected in
the spectrum from one fibre (horizontal).
In low-resolution mode (on both arms) the arc lines are straight, and nearly parallel to the y axis.
A magnified view of the centre of an image shows how
the spectra are grouped in blocks of 21 in the
y direction, reflecting the grouping of fibres into
individual slitlets (29 of them) along the LIFU slit-head in the spectrograph.
About 2/3 of the way up this image,
we see the prominent dark band of the
overscans at the junction between the
upper halves (4 quadrants) and the lower halves (4 quadrants) of the two
6k x 6k CCDs. Below the overscan we see the 21 fibres in each of slitlets 13 and 14
and (just below the overscan region), the first fibre of slitlet 15.
The other 20 fibres of slitlet 15 are visible above the overscan.
The middle fibre of the 21 in slitlet 15 is #305, the fibre which is physically
at the centre of the LIFU science array.
The fibres within a slitlet are separated in the y direction by ~ 9.1 pixels.
The separation between the last fibre in one slitlet to the first in the next
(peak to peak) is about 19 pixels. The fibres thus span ~ (29 * 20 * 9.1)
+ (28 * 19) + 32 ~ 5840 pixels in the y direction.
Information about the mapping of fibre number to y pixel value,
and the location in y of the central fibre of the array (#305),
can be found on
a separate page.
In high-res mode, the format is similar, except that the
arc lines are curved (i.e. the wavelength
at a given x depends on fibre number). This is a high-res ThAr arc
exposure from the red arm, i.e. with VPH 'R HR' (click on image to enlarge):
On the blue arm, the hi-res spectra (i.e. with VPH 'B HR1'
or 'B HR2') are similarly curved, but in the opposite direction.
The figures above indicate schematically for the blue arm (left)
and the red-arm (right) the pixels
used for science (yellow), masked off (grey) and corresponding to
bias (8 vertical strips in green) and overscan (8 horizontal and 8 vertical
strips in red). The widths of the bias and overscan strips are ~ 40
and 15 pixels respectively. The overscan strips come in pairs so e.g. the
thickness in y of the central overscan strip, prominent on the detector array
(as noted above) is ~ 30 pixels.
The numbering of the quadrants 'TSn' reflects that used in the FITS headers
e.g. BIASSECn for n = 1 to 4.
In slow readout mode (used for most science exposures), the readout noise is typically
~ 2.8 electrons per pixel rms on the blue arm, and 2.4 rms on the red.
For this readout mode the gains are approximately 1.0 on both arms.
In slow mode, with no binning, the readout time is 150 sec.
The detectors can also be read out binned (x1, x2, x4 in each direction)
and/or with different readout speeds, but not all of these options are available to
observers, and most are suitable only for test/commissioning purposes.
See
here for a table of readout times.
Three binning options are available to observers: x1, x2 or x4 in the spectral direction (x).
Binning in the 'spatial' direction (y) is not offered because it would result in
blending of the fibre traces.
The detectors cannot be windowed on readout.
For reasons of observing efficiency and for optimal performance from the
data-reduction pipeline, there are some constraints on the allowed
lengths of observing blocks (OBs, typically 1 hour for survey observations),
on the lengths of individual exposures within each OB, and on the linkage betwee
different OBs.
For open-time observers, the rules are set out on the
WEAVE open-time pages,
and in particular on the
PROGTEMP-builder page, which tells open-time observers how to
specify the observations required.
Cross-talk between fibres
As noted above, the separation of the fibre traces on the
CCD is ~ 9.1 pixels (in the y direction). The delivered FWHM in that direction
is ~ 5 pixels, small enough that the spectra can easily be
traced and extracted by the CASU pipeline.
The cross-talk (at extraction) between adjacent fibres has not been
characterised accurately but, with an FWHM of ~ 5 pixels,
is likely be ~ few %.
Ghosting
In all 5 spectroscopic modes, each spectrum
on the detector is accompanied by a faint ghost image a few pixels
across, thought to be a recombination ghost.
In x, the ghost image is usually located near the mid-point of the
illuminated area.
In y, the locations of the spectrum and ghost are roughly symmetrical about
the centre (y = yc) of the illuminated area.
I.e. if the spectrum is at y = yc + dy,
the ghost will probably be within ~ 160 pixels of y = yc - dy.
The peak counts in the ghost are typically <~ 1% the peak
in the spectrum. The exact details may change as a result of the ongoing
realignment of the spectrograph.
More generally, the stray-light contribution from any one object to the
(non-adjacent)
spectrum of any other object is expected to be < 0.5%.
Chris Benn, Lilian Domínguez, Cecilia Fariña
Back to LIFU home page
