Last updated on 28-Aug-2001 2:11 PM

Michelle on UKIRT

The User Manual

 

A.Glasse

The Royal Observatory, Edinburgh

 

1   Introduction

2   Observing Strategies

3   Analytical Expressions

44   Calibration Concepts

 5   Instrument Configurations

6   Summary of Configurable Light Paths
  Model Observing Sequences
8   The Observing Tool
9   Model Observations
10   Sensitivity

Introduction

Michelle is an astronomical spectrometer and imager for use on both the UKIRT and Gemini telescopes at the thermal infrared wavelengths between 8 and 25mm. This document describes how Michelle operates on the UKIRT telescope. It includes a description of the instrument’s capabilities, and information on its spatial and spectral fields of view and sensitivity.

Observing Strategies

The traditional method for removing the large background flux that is seen when observing in the thermal infrared at a ground-based observatory, involves using the telescope secondary mirror to chop the detectors field of view from the target object to nearby background sky at a frequency of a few Hertz, and then taking the difference between the on-source and off-source measurements. Michelle follows this procedure for imaging and for spectroscopy at low spectral resolving powers.

At high spectral resolving powers, the effect of fluctuations in the sky’s thermal emission becomes less important compared to differences between the two optical paths used for chopping. It then becomes possible to make staring observations, where the telescope is only moved occasionally (every 20 seconds or so), to sample the background sky near the target object.

In order to understand the rationale behind the techniques of chopping and nodding that an observer will need to exploit when observing with Michelle, we must formulate expressions which describe these processes in terms of the sky, telescope and instrument background fluxes.

Analytical Expressions

The traditional method for detecting a faint astronomical target against the high thermal background flux of the sky and telescope involves chopping the telescope field of view from the target to nearby sky and taking the difference. At low spectral resolving powers and when imaging, Michelle will use this technique, with one or more exposures being taken in each chop position and the data for each chop position then coadded over several chop cycles (to form an ‘integration’) before being sent back to the host computer for processing and storage. Both on and off-source components of the integration will be stored.

The difference between on and off-target exposures contains spatial and spectral structure due to the combined effects of the atmosphere, the optics and the pixel to pixel variations in gain across the detector. To see this, we can write expressions for the detected photon signal 'f' in each pixel as a function of chop position 'C', in terms of the source functions of the sky and telescope, and assuming a model for the transport of radiation through the atmosphere. To begin with, we assume that the astronomical target is in chop position C1.

f(C1) = g(i,j) [Stel(C1) + Satm (1 - e) + Stgt e]

and, f(C2) = g(i,j) [Stel(C2) + Satm (1 - e)]

The meaning of the terms are described below,

g(i,j)

The product of the detector gain matrix with the instrument’s response function, or optical efficiency. It will vary both from pixel to pixel (row ‘i’ and column ‘j’) and more slowly with position across the entire array. It is assumed in the following that g(i,j) is not a strong function of time.

Stel(C) = etel Bl(l, 275K) + DStel(C)

The thermal emission from the telescope optics. In this model, it is assumed to take the value of an ambient temperature black body spectrum with an emissivity etel ~ 3% on Gemini’s boresight, with variations across the telescope focal plane included in g(i,j). Stel(C) is assumed to vary from this level with chop position by the telescope’s thermal offset, denoted DStel(C).

Stgt(a,d,l)

The signal from the target object which is assumed to be a function of RA, Declination and wavelength. It may be spatially extended and vary over a characterisitic scale (Da, Dd ~ l/D).

Satm(a,d,l) = (1 - etel ) Bl(l, Tsky)

The continuum, source function for the sky, which is approximated here by a black body at a temperature somewhat colder than ambient and with an emissivity given by the telescope throughput (~97% for a 3% emissive telescope).

τ(a,d,l) = τzen(l) Z(a,d)

The optical depth of the atmosphere, a strong and complicated function of wavelength, but with a spatial variation which can be approximated to first order by scaling the optical depth towards the zenith by the airmass Z=1/cos(z), where z is the target object’s zenith angle.

Using these approximations, we can rewrite the prescriptions for the on- and off-target signals for telescope beam (offset) position A, and the difference between them as,

f(A,C1) = g(i,j) [etel Bl(l, 275K) + DStel(C1) + Satm (1 - e) + Stgt e]

and, f(A,C2) = g(i,j) [etel Bl(l, 275K) + DStel(C2) + Satm (1 - e)]

ftgt(A) = f(A,C1) - f(A,C2) = g(i,j) [DStel(C1) - DStel(C2) + Stgt e]

The atmospheric term can be seen to cancel very well as long as the chop frequency is higher than the characteristic timescale over which the sky emission changes due to sky noise. The minimum chop frequency will be a few Hertz for imaging, but may drop to a fraction of a Hertz at high spectral resolving powers where the sky emission is restricted to narrow spectral regions. (If the atmosphere were static, at an airmass < 2, for chop throws < 30 arcseconds the airmass will change by < 0.05%.) It is intended to provide a means for measuring the noise power spectrum as part of the sequence of observations in order to estimate the optimum chop frequency and exposure times.

As long as the telescope thermal offsets are small and the gain matrix is reasonably well known, this raw chopped image can provide a good cosmetic measure of a target image or spectrum, as required by the quick look display. For spectroscopy, the target’s continuum spectrum will have dark atmospheric absorption features superimposed on it.

The telescope thermal offset terms can be removed by beamswitching, whereby the target is placed in chop position C2 and the blank sky in C1. The new beamswitch position is denoted by the letter 'B'.

ftgt(B) = f(B,C2) - f(B,C1) = g(i,j) [DStel (C2) - DStel (C1) + Stgt e]

Summing the two sets of observations gives,

ftgt = ftgt(A) + ftgt(B) = 2 g(i,j) Stgt e

The timescale for beamswitching can be longer than that for chopping because its effectiveness depends on the more slowly varying temperature variations across the telescope's optical surfaces.

The technique used for removal of the remaining gain matrix (flat fielding) and atmospheric optical depth terms (sky correction) differs between image and spectroscopy data; each will now be discussed.

Calibration Concepts

Flat Fielding and Flux Calibration of Image Data

When imaging, a flat field observation can be taken by staring at a region of blank sky near to the target (or extracting the sky positions from the chopped data set) to give an image,

fsky_flat = g(i,j) [etel Bl(l, 275K) + DStel (C2) + Satm (1 - e)]

Every pixel sees the same spectral band when imaging, and so the slow variation of Stel and Satm across the array make the gain matrix the only source of structure with high spatial frequency in the sky flat. Taking the ratio of the beamswitched image of the target object with the sky flat gives,

ftgt / fsky_flat = Stgt 2(1 - eatm(Z)) / [etel Bl(l, 275K) + DStel (C2) + eatm(Z) Satm]

where we have replaced the terms containing the optical depth by the effective emissivity of the sky (a function of airmass) averaged across the filter’s spectral band.

The resultant image will be a close approximation to the true target. It can be flux calibrated by observing a standard star at a similar airmass through the same filter, where the chief source of error will usually come from temporal variations in the sky emissivity between making the two observations. Aperture photometry of the standard star image and a knowledge of the filter profile will then allow the calculation of a scaling factor to convert from image data units (electrons per pixel per second) to calibrated flux units (Janskys per square arcsecond).

The accuracy of the flux calibration over an extended target when using a point-like standard star is dependent on the quality of the flat field.

Flat Fielding and Flux Calibration of Spectroscopic Data

Needless to say, the above techniques are inadequate for flat fielding long slit spectroscopy data, where the presence of numerous emission features make the sky a poor choice as a flat field object.

For flat fielding spectroscopic observations, Michelle will use a specially designed flat fielding plate which can be driven into the beam in front of the cryostat window. In terms of the formalism used above, an observation of the flat field plate can be written as,

fflat = g(i,j) [etel Bl(l, 275K) + Satm + DSflat]

Here, the flat plate is assumed to have a smooth spectrum which is approximately equal to the combined continuum spectra of the telescope and the sky source function, such that DSflat << Satm. When this spectral image is subtracted from a sky flat the result is,

fsky_flat - fflat = g(i,j) [DSflat + DStel (C2) + Satm e)]

Division of the beamswitched observation by this differenced image and some rearrangement then gives,

The strongly wavelength dependent optical depth now only appears in terms which are small. However, it may be necessary to remove the residual atmospheric emission structure by modelling and line fitting.

The flux calibration of spectroscopic data will, as for imaging, require the observation of a standard star and the subsequent calculation of a scaling factor to convert from data units (electrons per pixel per second) to spectroscopic intensity (Watts per square metre per micron per square arcsecond). The procedure is complicated for extended targets by the fact that the standard star spectrum will normally only cover a few rows of the array and so must be 'grown' or copied in the spatial direction along the slit before the division of target by standard can be carried out.

Wavelength Calibration of High Spectral Resolution Spectroscopic Data

The determination of an accurate wavelength scale is currently envisaged to be carried out in one of two ways, depending on the spectral resolving power of the observation. In each case, the grating must not be moved between carrying out the science observation and making the calibration observation. Evaluation of the calibration techniques over the next few months may result in alternative calibration procedures being adopted.

At high spectral resolving powers (when using the echelle), a set of cooled, solid state Fabry-Perot etalons, mounted in Michelle's filter wheels, will provide a pattern of emission features at known wavelengths (as determined by cross-calibration against sky emission features) and narrower (by a factor ~2) than the spectrometer’s spectral response function. The etalons are designed so that at least three features should be visible on any exposure, allowing the linearity of the wavelength scale to be measured. If this plan fails to work, the fall-back will be to use the atmospheric features on the target exposures.

The use of etalons should simplify the procedure, since the atmospheric emission spectrum will often be very complicated, making it difficult to unambiguously identify lines under conditions of differing atmospheric precipitable water vapour path length and airmass.

Wavelength Calibration of Low Spectral Resolution Spectroscopic Data

At low and moderate spectral resolving powers, it will not be possible to use atmosheric features (too spectrally broad) or solid state etalons (technically impractical). Instead, Michelle’s calibration unit is equipped with an integrating sphere illuminated by two arc lamps (probably argon and krypton), which will flood the instrument focal plane with a set of strong emission lines at wavelengths ranging up to 4mm. The low wavelength cutoff of the SiAs detector is around 2.5mm, and so these lines should be detectable using a K band (2.0 to 2.4mm) or L band (3.0 to 4.0mm) filter, with the diffraction gratings operating in diffraction orders of 4 or above.

Instrument Configurations

The Elements of a Michelle Data Set

The hierarchy and terminology used to describe the data sets produced by Michelle, whether chopping or staring are itemised and described below.

Exposure - A single on-chip exposure of the array. Where possible, the exposure time will be chosen so as to nearly fill the capacitative well associated with each pixel with charge, thereby maximising the ratio of signal (charge) to read noise. For low photon rates, where it takes a relatively long time to fill the wells, multiple non-destructive samples are made of the charge as the wells fill. A least squares fit of the plot of these sample values against time reduces the effective read noise for the exposure, in theory by a factor,

The anticipated range for single frame exposure times will be anything from a few milliseconds to tens of seconds.

Integration - The result of coadding one or more exposures in the array controller (EDICT). An integration results in the smallest data set that can be transmitted back to the host computer for quick look display, typically at a rate of a few integrations per second or less. When chopping, an integration will include the data from one or more chop cycles. It will then contain three separate images, the coadd of all on-source exposures the coadd of all off-source exposures, and their difference.

Observation - The smallest unit of data that can be requested using the Observing Tool, and the result of coadding one or more integrations. The result of an observation is a file that is written to disk, typically once every few seconds. An observation sequence should only be paused at the end of an observation.

Observation Sequence - The interleaved sequence of observations and telescope and instrument actions that are specified using the ORAC (or Gemini) Observing Tool. The result of an observation sequence will be a data set that constitutes a complete scientific programme, with all of the flat fielding and calibration images necessary for production of the final science product.

Summary of Configurable Light Paths

Figure 1 represents the configurations that can be selected by the operator from Michelle’s internal configurable systems, with the arrowed lines showing the paths along which light can be directed on its journey from the telescope to the detector array. It highlights some important features of the instrument, such as the fact that the imager path is completely independent of the spectrometer and its field rotator, and that the on-board calibration unit flat fielding targets cannot be viewed through the half-wave plates that are used for polarimetry.

Figure 1. Light paths through the Michelle spectrometer and imager that can be configured by the operator.

 

It omits those systems which are provided by the telescope, such as the target Acquisition and Guiding system, and the Peripheral Wavefront Sensors.

The following sections then deal with each of Michelle’s primary configurations in turn, outlining the actions that must be taken by the Telescope Control System during a typical observation and indicating the type of data that will be produced.

Model Observing Sequences

Imaging

The instantaneous field of view of the imager is approximately 48 x 64 arcseconds on the sky. The preferred observing scheme will be dictated by the angular size of the target astronomical object. For a compact target (i.e one for which the flux is negligible outside a region whose largest dimension is some 12 arcseconds), it will be possible to both chop and nod on the array, thereby maximising the time spent observing the science target.

The exposure time will be chosen so as to fill the pixel wells, typically a few 10's of milliseconds. If the telescope is chopped at 5Hz, then there will be 8 or 9 of these exposures in each stable chop position, and the coadd over two or three cycles could be used to form an integration which would be sent back to the host. An observation might request 150 of these integrations (taking some 30 seconds), after which period the telescope could be moved.

The following procedure assumes that the observation has been planned using the Gemini Observing Tool such that the positions of PWFS guide stars, and the direction and magnitude of the chop throw are all determined in advance.

  1. Select the desired band pass filter.
  2. Configure the image extraction and injection mirrors for imaging.
  3. Set the detector bias levels and clock waveforms for imaging. (Actions 1 to 3 should be performed as early as possible to allow the detector to stabilise.)
  4. Drive the telescope to the target region, using the pre-programmed guide stars as targets for the two PWFS’s.
  5. Set the Gemini instrument rotator to the preferred orientation (portrait or landscape) for Michelle’s rectangular science field.
  6. Start taking data with the telescope chopping and nodding, at the command of the Michelle ICS according to the sequence defined below.

Repeat until complete or interrupted by operator

For each Beamswitch position A, B, B, A

For each Chop position N, S

Coadd exposures E1, E2, E3,... En

 

For each A beam, coadd 'n' exposures in the N and S chop positions, then move on to the B beam. The smallest data set that would be transmitted back to the host from the EDICT array controller would then be a file containing two images, one for each chop position. There would typically be one such file per beam position.

Observing Sequence for Imaging Polarimetry

The procedure for polarimetric imaging closely follows that for imaging, except that the four components Q1, U1, Q2 and U2 of the polarised light must be measured at each beamswitch position. This is done by driving the birefringent half-wave plate that is mounted in the polarimetry slide in front of the cryostat window. The wave plate rotation angle 'q', measured relative to a fiducial position, must be changed in 22.5 degree (p/8 radians) increments, with the components assigned to each position as follows,

Polarised Component

q [radians]

Q1

0 + m p /2

U1

p/8 + m p /2

Q2

2 p /8 + m p /2

U2

3 p /8 + m p /2

Here, 'm' is an integer. Since the sequence of sampled components repeats for each 90 degree rotation of the wave-plate, the plate can be rotated continuously and the four repeated sequences that are obtained in each revolution will allow the correction of systematic variations in the plates optical properties.

The full observing sequence is then summarised in the following steps,

  1. Select the desired band pass filter in one filter wheel.
  2. Select a wire grid analyser in the other filter wheel.
  3. Configure the image extraction and injection mirrors for imaging.
  4. Set the detector bias levels and clock waveforms for imaging.
  5. Drive the telescope to the target region, using the pre-programmed guide stars as targets for the two PWFS’s.
  6. Set the Gemini instrument rotator to the preferred orientation (portrait or landscape) for Michelle’s rectangular science field.
  7. Start taking data with the telescope chopping and nodding, at the command of the Michelle ICS according to the sequence defined below.

Repeat until complete or interrupted by operator

For each Beamswitch position A, B, B, A

For each Wave-plate position Q1, U1, Q2, U2

For each Chop position N, S

Coadd exposures E1, E2, E3,... En

 

For each A beam, coadd 'n' exposures in the N and S chop positions, then rotate the wave-plate through 22.5 degrees to sample the next polarisation component. Once all four components have been sampled move on to the B beam and repeat the cycle.

There would now typically be one file for each position of the wave-plate, each containing a pair of coadded images, one per chop position.

Observing Sequence for Spectroscopy

One effect of the quasi-Littrow optical design that is used in Michelle’s spectrometer is that each pixel’s field of view on the sky is set by the angle of incidence of light on the diffraction grating as illustrated in Figure 2. As an example, with the entrance slit aligned with the Declination axis, and using the echelle near its blaze peak (q = 63 degrees), each square arcsecond on the sky will be mapped onto a lozenge shape extending 0.64 arcseconds in Dec., 1.56 arcseconds in RA and with an interior angle, y = 50.2 degrees.

The on-chip exposure time might be 20 seconds or so when using the echelle, with several hundred non-destructive samples being taken in order to reduce the effective read noise. At this low data rate there might then be only one exposure per integration, and only one integration per observation. It should be noted that the quick look display will only update at the end of the integration, once every 20 seconds or so in this case.

As for imaging, the following procedure assumes that the observation has been planned using the Gemini Observing Tool such that the positions of PWFS guide stars, and the direction and magnitude of the chop throw are all determined in advance.

The photon flux onto the detector will vary by factors ranging from 10’s to 1,000’s when switching beween imaging and spectroscopy configurations, and it is anticipated that the detector will take some time to recover and stabilise.

The usefulness of chopping at high spectral resolving powers where fluctuations in the sky emission become less important will need to be investigated during engineering commissioning,

The following scheme assumes that the observer wishes to set the spectrometer on a particular bright thermal infrared peak in a diffuse or complicated target field.

  1. Configure Michelle for imaging the target field; select a suitable band pass filter, configure the image extraction and injection mirrors for imaging and set the detector bias levels and clock waveforms. The quick look system should now be displaying current images from Michelle.
  2. Set the instrument rotator to the preferred orientation (portrait or landscape) for Michelle’s rectangular science field.
  3. Drive the telescope to the target region, using the pre-programmed guide stars as targets for the two PWFS’s, and achieve a stable chop and nod pattern, using the quick look display to position the target field on the array.
  4. Start taking image data with on-line data processing and data storage running. The quick look display is now used as a real-time monitor of data quality, and the latest processed image is displayed by the Observing Tool.
  5. Use the Observing Tool to overlay the planned position of the slit for the pre-stored spectroscopic observation onto the real image of the target region. Check by eye that it corresponds to the desired position, adjusting the telescope pointing or the planned slit angle as necessary.
  6. Re-configure the drive electronics and the optical train for spectroscopy, driving the image extraction and injection mirrors out of the optical path, and selecting the appropriate filter combination for spectroscopy. If necessary, wait for the array performance to stabilise.
  7.  

    Figure 2. Image shear in the Michelle spectrometer and its correction using a rotating entrance slit. ‘q’ and ‘g’ are the angles of incidence in and perpendicular to the plane of diffraction respectively;‘y’ is the slit rotation angle. See the text for further explanation.

     

  8. For low spectral resolving power spectroscopy, start taking data with the telescope chopping and nodding, at the command of the Michelle ICS according to the same sequence as defined for imagng. At high spectral resolving powers it may not be necessary to chop.

The Observing Tool: Behind the Buttons

The ORAC Observing Tool Component for Michelle (shown in Figure 3 and Figure 4) is used to configure the instrument for all subsequent observation iterators. The Camera and Polarimetry controls in Figure 3 select imaging or spectroscopic configurations, and their polarimetric counterparts.

Imaging

The front panel of the ORAC-OT for Michelle, with the user having selected the imager camera and not polarimetry, is shown in Figure 3.

In the section of the Editor labelled ‘Spatial and Spectral Filtering’, the three Filter Category radio buttons (labelled N Band, Q Band and Special), control which sub-set of Michelle’s filters are displayed in the text box to the right, and from which an individual filter can be selected for a subsequent imaging observation.

Figure 3. The component editor for imaging.

The user should not need to adjust any of the controls in the section of the OT frame labelled ‘Data Taking’. The first, Chop Frequency, can only take on a few discrete values (including 0 Hertz for no chopping at all). The magnitude and direction of the chop throw are both currently set explicitly at the telescope. The UKIRT chopping secondary can manage chop throws of < 20 arcseconds reliably at rates of up to 10 Hertz, with an absolute upper limit of 34 arcseconds possible. Telescope nodding is controlled using the ORAC-OT’s Offset Iterator.

The Exposure Time is primarily set by the time taken for the background flux from the warm sky and telescope (and warm polarimetry optics, if polarimetry is selected), to fill the pixel’s capacitor with charge. If the user does change this value, then they should be aware that the effects of saturation or insufficient charge in the well are not easy to detect when chopping. In addition, the array has been characterised to work best with the values of exposure time and chop frequency set here by the OT.

The final control, Observation Time, sets the time taken for a single observation, (after which a single frame will be saved to disk). Once again, the default value should be used, otherwise the value set should be consistent with the number of chop cycles and take account of the optimum timescales for nodding, selecting a new polarisation angle or translating the detector.

The display box labelled ‘Detector Duty Cycle’ at the bottom left of the Component Editor shows the efficiency of the selected instrument configuration in terms of the fraction of time spent integrating signal on the array compared to the elapsed time. This figure does not account for whether the astronomical target is in the beam or not.

Spectroscopy

The layout of the ORAC-OT component editor for spectroscopy with Michelle is shown in Figure 4. The reader will note that it is only the ‘Spatial and Spectral Filtering’ section of the Editor which changes between selection of the spectrometer and spectroscopy settings of the Camera control.

One of the five available diffraction gratings (LowN, LowQ, MedN1, MedN2 or Echelle) is selected by the user via the Grating control. The Wavelength box shows the wavelength which will be mapped onto the central column of the array, (to an accuracy of +- a few columns). For the LowN and LowQ gratings, the OT sets the Wavelength box itself to a default value when the grating is selected. This value is designed to provide full coverage of the N or Q atmospheric windows in a single exposure. For the other gratings, the currently selected wavelength is not changed. In all cases, once the grating has been selected, the user can set the wavelength to any value.

The Order control is set by the OT to a default value determined for the selected grating and wavelength using a look-up table. The user should only rarely want to change this box, specifically to select a non-default order for the echelle grating, since it will normally result in a low optical throughput. The adjacent Default button will reset the default order for the selected grating.

The Focal Plane Mask control sets the slit width to be used for forthcoming observations. The available slits (1pixel, 2pixel, 3pixel, 4pixel, 6pixel or 8pixel), are scaled in width from the 1pixel slit, which gives a 0.38 arcsecond wide field of view on the sky for the low dispersion gratings. The user can change the mask from the default value which is set initially by the OT, based on achieving the optimum sensitivity at the observing wavelength for a point source. For example, a narrower slit than the default will often (but not always) give a higher spectral resolving power than the default, but at the expense of sensitivity and the degree of spectral sampling for a single exposure.

 

Figure 4. The component editor for spectroscopy.

 

The Position Angle control can be used at this point to set the projected angle of the entrance slit on the sky. It can also be set later on in the observing sequence, for example using the Position Editor of the Observing Tool.

The Pixel Sampling control follows the familiar format of the equivalent control for CGS4. That is, the first number sets the number ‘N’ of movements that the detector will make in the spectral direction, and the second number sets the total number of pixels ‘M’ that the detector will move. So for example, a setting of 2x1 will result in 2 spectral images, separated by 0.5 pixels in the spectral direction.

The Spectral Resolving Power, Spectral Coverage and Spatial Field of View displays are set according to what has been selected in the above controls. The Selected Filter display is included to allow the user select the same filter for imaging as is being used for spectroscopy. This feature is intended to help the user to flux calibrate high dispersion (i.e R ³ 3000) spectra. (The low dispersion gratings use long pass blocking filters which are not really suitable for imaging).

Observing Tool Iterators: Flat, Arc, Bias and Observe

The iterators (the components which schedule data taking as opposed to instrument configuration changes), used for Michelle are the familiar Observe, Sky, Bias, Dark and the new iterator Michelle Calibration.

The Observe and Sky Iterators (shown in Figure 5) simply schedule the selected number of observes, each of which uses the exposure and chop parameters specified in the preceding Michelle Component, and flag them for reduction by the correct ORAC-DR recipe.

Figure 5. The Observe, Sky and Dark Iterators

 

The instrument dark current should only be a significant fraction of the background photon flux at high spectral resolutions. Although the Dark iterator, also shown in Figure 5, can be used to select exposure times and coadds (the number of exposures per observation), the user is strongly recommended to use the default values, since the longest time taken for a typical observation will be only 20 to 30 seconds.

The Michelle Calibration iterator shown in Figure 6 is used to configure components in Michelle’s calibration unit and calibration specific optical elements inside the instrument. It combines the functions of flat fielding and wavelength calibration.

 

Figure 6. The Calibration Iterator, set for flat fielding and wavelength calibration.

 

When imaging, Michelle is flat fielded by offsetting from the target object and then using the same simple ‘sky’ iterator as for TUFTI/IRCAM3.

Two flat fielding sources are available for spectroscopy. For low spectral resolutions, the window shutter can be used. By back-reflecting the cryostat interior, it provides a low flux, comparable to the sky and telescope background but without the sky emission lines.

For high spectral resolutions, a heated integrating sphere can be used to increase the flat fielding flux, and hence reduce the time required for flat fielding.

Wavelength calibration at high spectral resolutions can use the sky emission lines seen in the observed data prior to demodulation, in which case the Michelle Calibration iterator is not used. Alternatively, for echelle observations in the N Band, an internal solid-state Fabry-Perot etalon can be selected using the Line Source control with the iterator to provide a fixed pattern of fringes of known wavelength.

The Polystyrene selection of the Line Source control is used to wavelength calibrate the LowN grating with a cooled polystyrene film with features at known wavelength. The wavelength calibration of Q Band spectra is currently reliant on the sky emission line method.

 

The circumstances under which each iterator should be used and for what purpose are summarised in Table 1.

 

Calibration Activity

Dark

Flat Field

Arc

Imaging

Not Used

Sky

Not Used

LowN

Not Used

Mich. Cal. (Shutter)

Mich.Cal. (Polystyrene)

LowQ

Not Used

Mich. Cal. (Shutter)

Observe (Sky Lines)

MedN1

Dark

Mich. Cal. (Shutter or Hot Sphere)

Observe (Sky Lines)

MedN2

Dark

Observe (Sky Lines)

Echelle

Dark

Mich. Cal.

(Hot Sphere)

Mich. Cal. (Etalon) and Observe (Sky Lines)

Table 1. Michelle Iterators, what to use and when.

At the Telescope

The Instrument Status Display

The components which describe the state of the instrument during the execution of an observation are listed below;

ITEM EXAMPLE SETTING UNITS

Camera Spectrometer -

Filter E97B10 -

Grating LowN -

Wavelength 10.5 microns

Focal Plane Mask 2_pixel -

Position Angle 45 degrees

Pixel Sampling 2x1 -

Pixel Size 0.1 x 0.1 arcseconds

Coadds 10 -

Configuration Type Bias -

Science Field of View 30 x 20 arcseconds

Spectral Coverage 10.2 to 10.4 microns

Chop Frequency 5 Hertz

Exposure Time 0.05 seconds

Observation Time 30 seconds

Acquisition Mode ndchop -

Detector Duty Cycle 0.95 -

Model Observations

A Realistic Example - Gliese 229B

Gliese 229B is a brown dwarf at a distance of 5.7pc with a mass of some 40MJ. Spectroscopy in the thermal infrared promises (Noll et al., 1997, ApJ, 489,L87) to be a key probe of the atmospheric chemistry of this class of object, with implications for their origin as brown dwarf or giant exo-planet. A model observation with Michelle would then involve imaging the GJ229 system followed by broad band 10 and 20mm spectroscopy at low spectral resolving power and finally the study of selected spectral bands at moderate spectral resolving power.

 

Figure 7. The acquisition of GJ229 by the Michelle imager using the Gemini Observing Tool.

We start by assuming that the position of GJ229B (which is approximately 5.6 arcseconds south and 0.9 arcseconds east of its bright companion GJ229A), is not known with adequate certainty to position Michelle's 0.38 arcsecond wide entrance slit, and so it must first be found using the imager. Figure 7 then shows the appearance of the position editor component of the Gemini Observing Tool (a 'beta release' version, current on 20th February 1998), with the two arcs showing the patrollable area available to Gemini's two peripheral wavefront sensors. Guide stars at the north and west edges of the 8 arcminute square region have been selected and Michelle's imager field (the larger of the rectangles just to the south of the bright star), is shown to be set to map a region between 3 and 27 arcseconds to the south of the bright companion GJ 229A. Michelle is not currently equipped with a coronagraph and so this configuration is intended to avoid the scattered light from GJ 229A drowning out the image of GJ229B.

 

Figure 9. Michelle configured for imaging, with a 5Hz chop frequency and each 'observe' lasting 15 seconds.

The instrument’s configuration for this imaging observation is shown in Figure 9, where the array exposure parameters have been set for a chop frequency of 5Hz. The chop amplitude would be chosen as 12 arcseconds in the north-south direction (not shown) in order to keep the point-like target in the field of view of the array at all times.

The observing sequence is then set to start with a bias observation, that is, a measurement of the detector dark current. When imaging with Michelle, this could be done by taking an exposure of a blank plate in the cold filter wheel, but given that the dark current should be negligible it is likely to be an unnecessary operation. Flat fielding will be carried out using the sky images taken during normal observing.

The bulk of the observation is then made up of repetitions of sets of 'observe' components in the east (E) and west (W) beamswitch positions. In this example, the 'offset' component refers to a telescope movement of 16 arcseconds.

A single observation is set to last some 15 seconds, after which the telescope is offset by 16 arcseconds in RA to the next beamswitch position. The beamswitching sequence of observations at east, west, west, east positions is repeated 3 times to give a total integration time of 3 minutes. With a predicted 10mm flux of 4.2mJy (Matthews et al. 1997), and Michelle’s predicted 1 sigma 1 hour sensitivity for broad band 10mm imaging of a point source somewhere around 0.1mJy, this should give a 10 sigma detection, sufficient to position the 0.38 arcsecond wide spectrometer entrance slit on the peak of GJ229B’s thermal infrared emission.

The process by which the position of the entrance slit is refined after direct imaging of a target region is envisaged to involve pausing the observing sequence, displaying the reduced image in the Observing Tool (with calculated centroids of point source where appropriate), and then modifying the spectroscopy configuration to include the revised slit position, before re-starting the observing sequence.

 

Figure 10. The entrance slit positioned on the putative position of GJ229B for R=1,000 spectroscopy of the ammonia bands at 10.5mm. a CMa is selected as the standard star for flux calibration of the spectrum.

According to this scheme, the entrance slit would be overlayed on the Michelle broad band 10mm image, rather than the visible light catalogue image of the GJ229 system which is shown in Figure 10.

Some other points that should be noted regarding this observation include the chaining together of the imaging and spectroscopy observations which should help ensure that the slit's field of view falls on the intended part of the image. The spectroscopic observation is also followed directly by a slew to a standard star for flux calibration of the spectrum, which would in turn be followed by wavelength calibration using the on-board calibration unit (not shown in Figure 10). This sequence is required in order to avoid moving the grating between taking and calibrating a spectrum.

 

Performance

Spatial Coverage

Table 2 summarises the key dimensions associated with Michelle’s focal plane. The telescope is assumed to have a diameter of 7.899m of which the central 16% is obscured. The point source dimensions all scale linearly with wavelength ‘l’.

 

Ideal Diffraction Limited Point Source Dimensions for l = 10mm

Diameter of First Airy Minimum (=2.44l/D)

0.64 arcseconds

Full Width at Half Maximum Intensity

0.26 arcseconds

50% Encircled Energy Diameter

0.29 arcseconds

80% Encircled Energy Diameter

0.74 arcseconds

Detector Array, SBRC Si:As high flux IBC hybrid

Pixel pitch

50 microns (mm)

Format

320x240 pixels

Imager, detector focal plane

Imager Focal Plane Scale (chosen to match 2 pixels to the FWHM intensity of a point source at 8mm).

2.0 arcseconds/mm

0.10 arcseconds/pixel

Imager field of view

32x24 arcseconds

Spectrometer, detector focal plane

Spectrometer focal plane scale

3.6 arcseconds/mm

0.18 arcseconds/pixel

Spectrometer entrance slit length

43.2 arcseconds

Nominal spectrometer entrance slit width

0.36 arcseconds

Table 2 Key dimensions in the detector focal plane.

Spectral Coverage

When configured as an imager, the spectral pass band will be determined by which of the 40 or so available filters are selected, and will generally range from 1% of the observing wavelength up to the full N or Q band.

For the spectrometer, the combination of four conventional diffraction gratings or an echelle with seven entrance slits leads to a wide variety of spectral configurations. The baseline spectral parameters listed in Table 3 apply to the spectrometer when using a 2 pixel wide entrance slit, doubling the slit width will generally halve the spectral resolving power but leave the instantaneous spectral coverage unchanged. The slit wheel will be equipped with seven masks, currently these include entrance slits that have a field of view on the sky equal to multiples of 1,2,3,4,6,8 or 16 times the basic field of view of a single pixel (0.18 arcseconds).

It should be noted that when using the echelle, the number of pixels that will be matched to these slits will be increased by a factor normally lying between 1.2 and 1.5, whose value depends on the angle of incidence of light on the echelle and hence on the wavelength. For example, when using the nominal 2 pixel (0.36 arcsecond) slit with the echelle, the flux from a spectrally unresolved line will be shared between anything from 2.4 to 3 detector columns. This reduction in the pixel’s field of view in the spectral direction is mirrored by a corresponding increase in their field of view in the spatial direction, and a wavelength dependent variation in the instantaneous spectral coverage from the figure quoted in Table 3.

Grating Name

Low-N

Low-Q

Med-N1

Med-N2

Echelle

Diffraction Order

1

1

1

1

n=3 to 10

Blaze Wavelength [mm]

9.6

20.5

9.7

9.7

73.7/n

Instantaneous Spectral Coverage

7.7mm

10.0mm

1.5mm

0.5mm

1500km/sec at blaze wavelength

Wavelength range for good efficiency [mm]

8 – 13

17 – 25

8 - 13

8 – 13

8 – 25

Spectral Resolving Power (l/Dl) for a 2 pixel wide entrance slit.

200

160

1000

3100

10,000 to 30,000

Table 3 Baseline spectral coverages for the four 1st order gratings

Sensitivity

The ideal background limited fluctuation on the charge detected by a single pixel during a staring observation made using a BIB detector with negligible dark current or read noise (BIB detectors do not exhibit the generation-recombination noise seen in conventional photoconductors) is,

where,

G is the photoconductive gain (taken as =1)

b is the gain dispersion (b = <G2>/<G>2, also taken as =1)

ibg is the photocurrent generated by background radiation (electrons/second)

t is the integration time (seconds).

If the background photocurrent is assumed to be dominated by a single blackbody spectrum external to the instrument (the telescope and other warm optics), then its magnitude can be calculated from the formula,

where,

e is the combined emissivity of the telescope and other warm optics (3%),

Aw2 is the etendue of a single pixel, (A= 49m2 for Gemini, and for the Michelle spectrometer w=0.18 arcsec, giving Aw2 = 1.59 m2arcsec2),

t(l) is the optical efficiency (throughput) of the instrument fro the cryostat window to the detector (~0.6 for the imager, ~0.3 for the spectrometer),

h(l) is the quantum efficiency of the detector (< 0.5 photoelectrons/photon),

B(l,Twarm) is the Planck function for a black body at a temperature Twarm (=273K),

lband is the band of wavelengths illuminating the pixel, defined by the grating or filter.

 

Insertion of the above numeric values into the equation for ibg and substitution of l/R, for the integral over the pass band, where ‘R’ is the spectral resolving power gives, for l = 10mm,

The value of ibg scales with l2B(l,T), and approximately doubles in going from 10mm to 20mm. Its lowest value (~10,000 el/sec) occurs at l=8mm for the echelle operating at R~30,000, and this figure sets targets for Michelle’s rejection of scattered light and the maximum allowable temperature of the optics.

We can now write expressions for the 1 sigma 1 second sensitivity (i.e the flux required to achieve a signal to noise ratio of 1 in a 1 second integration time) in more useful units, by equating DQbg to the signal detected from a source at the top of the atmosphere. For example, for a spectrally unresolved line source where the background flux increases with spectral bandwidth but the signal stays constant,

Substitution of the parameters listed above gives,

The factor is introduced here to relate these sensitivity figures to real astronomical targets, for which a fraction f of the total available signal is spread over P pixels. For example, reference to Table 2 shows that when imaging a point source at 10mm, and numerically integrating over an aperture matched to the 80% encircled energy diameter, we should use the values f = 0.8 and P = 43. Table 4 lists values of f and P for a selection of slit widths and coaddition of spectra across several rows.

Fraction of light lost from a 10mm point source at the entrance slit ‘f’.

Slit Width

0.18"

0.36"

0.54"

Distance along

slit to coadd

0.18"

0.300

0.698

0.793

0.36"

0.460

0.750

0.840

0.54"

0.494

0.776

0.844

Table 4 The vignetted fraction 'f ' for some spectroscopic configurations

For a continuum source the background flux and signal both increase with the bandwidth and the formula for calculating the 1 sigma 1 second sensitivity becomes,

and the resultant value of F1s1s is then,

Table 5 illustrates the estimation of sensitivities for Michelle using a few more examples. It should be noted that the resulting figures apply to observations where the source image is always somewhere on the array; the effects of chopping off-chip, and overheads due to chopping and nodding are not included.

Observation

Sensitivity Limit

(S/N = 1 in 1 hour)

R=10 imaging of a point source at 10mm, with w=0.10" pixels coadded out to an 80% EED diameter of 0.74".

0.14 mJy

Broad band spectroscopy of an extended continuum source with a 2 pixel (f=1, P=2) wide slit, (w=0.18") and the LOW-N grating (R=190).

1.3 mJy / arcsecond2

As above (R=190) but for a point source, (f=0.70 and P=2 if the target is on the centre of a pixel).

0.09 mJy

N Band spectroscopy at R=3000 of an extended target, with the signal coadded from a 5 arcsecond length of the slit. (f=1, P=2x28=56)

3.0x10-18 Watt m-2 arcsecond-2

Echelle spectroscopy of a spatially and spectrally unresolved line source at 10mm, at R=20,000 with a 0.36" wide slit, f=0.7 and P = 2x1.3 = 2.6

2.2x10-20 Watt m-2

Table 5 Goals for Michelle’s sensitivity limits in key observing configurations.

 

A.Glasse