- Sky and Telescope emissivity
- Atmospheric Water Vapor
- Observing Strategy
- Image Quality
Sky and Telescope Emissivity
In this section, some general features of the mid-IR observing technique are reviewed. In most cases, practical examples and numbers used in CanariCam will be utilised to complement the explanation of the general concepts. This information is aimed particularly at those observers who are new to infrared astronomy.
The main difference between ground-based optical and infrared astronomy observations is that, at mid-IR wavelengths, the background is huge compared with the emission of any astronomical source. Besides, this background emission varies rather fast in time and in space. Nextv figure shows the Earth's atmospheric emission and the telescope emission (at 10 and 25 Celsius) in the CanariCam's wavelength operating range, assuming a telescope emissivity of 5%. Note that telescope emissivity can dominate over sky emission in the 10-micron wavelength range.
(*) Note that the ORM's mid-IR atmospheric emission and transmission curves are not available. Hence, the sky emission and transmission curves shown here were downloaded from the Gemini Observatory mid-IR resources webpage and simulate the conditions at the Cerro Pachón Observatory. The curves were calculated using ATRAN models (Lord, 1992).
Some of the factors that contribute to the telescope emissivity are the spiders of the secondary mirror (M2) holding structure, the interspace between segments of the primary mirror (M1), the telescope tube metallic structures that are around M1, and the dust accumulated on the mirror surfaces.
To avoid or even eliminate some of these components:
- The GTC secondary mirror (telescope pupil) is undersized, i.e. a small fraction of the primary mirror is vignetted so that no thermal emission from structures outside the primary mirror enter CanariCam.
- M2 also has retractable baffles that cover sky holes (e.g. the Cassegrain hole) when we observe with optical instruments, but leave them open to the sky when we observe in the mid-IR.
- CanariCam has Lyot masks that follow the shape of the GTC pupil. These prevent the thermal emission from the secondary mirror support spiders from reaching the detector.
- M1 is designed so that the physical separation between segments is kept to a minimum of 3 mm.
- GTC mirrors are cleaned regularly with CO2 snow to avoid accumulation of dust, which may be particularly important during the Summer, when the frequency of Saharian dust storms increases.
- GTC mirrors are re-aluminised when the aluminium coating starts to degrade, which ensures maximum reflectivity (minimum emissivity).
Thanks to the pupil imaging engineering mode of CanariCam, we can glimpse at the GTC pupil. The next figure shows an image of the pupil taken on January 2013, where the lighter orange (yellow) represents the warm emission, while the darker orange represents the cooler emission. Also note that it is possible to see the emission from the space between the M1 segments.
The telescope emissivity was measured on April 4th, 2013 using the same method as described in the Keck Observatory web site. In brief, we took a 10-micron spectum of the back of the M3 tower, which represents a 100% emissive reference black body at the ambient temperature. Immediately afterwards we took a spectrum of the background sky, which includes the emission not only from the sky but also from the metallic parts of the GTC that are visible to CanariCam (e.g. the segment gaps), the dirt in the mirrors and the entrance window. The ratio between the background spectrum and black-body spectrum is a measure of the emissivity from all background radiation sources. This total emissivity is represented by the black dots in the next figure. The figure also shows with solid lines the emissivity of a theoretical sky with a similar PWV to the actual observations plus a theoretical black body (Planck function) with no emissivity (i.e. only sky - red line), 10% emissivity (green line) and 20% emissivity (blue line), respectively.
By comparyng the observed emissivity with the theoretical emissivity curves, we can estimate the emissivity of the telescope. The previous figure shows that the emissivity between 8 and 9 micron is well below 10%. However, note that the sky models used in the figure do not represent correctly the Ozone feature at 9.7 micron.
Therefore, we decided to use a new set of sky models, downloaded from the ESO SKYCALC model calculator for Paranal page. These models represent the Ozone emission with far more accuracy than the ATRAN models for Cerro Pachon, as shown in the following 2 figures.
In the previous figure, the Ozone emission feature from the Paranal sky emission models match quite well the observed Ozone emission at the Roque de los Muchachos observatory. The measured emissivity data cross several of the lines at different PWV values. The 8 to 9 micron region seems to be better represented by the curve corresponding to 2.5mm of PWV, while the region above 11 micron is better represented by the curve corresponding to 7.5mm. The closest PWV value to the actual PWV value (6.6mm as measured by the IAC-run PWV monitor) during the observations is 7.5mm.
The next figure represents the same measurements as the two previous figure, but the theoretical models correspond to the sky at Paranal with black body emissivities od 0%, 10% and 20%, respectively.
The observed emissivity increases as the wavelength increases, being approximately 10%-18% between 10 and 12.5 micron. The reason why this increase emissivity at longer wavelengths occurs it is not clear yet and it is under investigation.
Atmospheric Water Vapor
The Earth's atmosphere not only emits strongly in the mid-IR, but also absorbs selectively the radiation from the celestial bodies, being almost transparent in only a few mid-IR observing windows. CanariCam is designed to work in two of such windows, the so called 10-micron (N) and 20-micron (Q) windows, respectively. Still, in these windows there are several telluric absorption lines that degrade the sensitivity of the observations. Some of the main contributors to the Earth's atmospheric absorptions are:
- Ozone, with an important absorption feature at 9.6 µm.
- CO2, whose strong absorption at ~ 15 µm defines the separation between the 10 and 20 µm observing windows.
- Water vapour, which shows several bands in the 10 µm window and strongly dominates the 20 µm window.
Next figures show the CanariCam filters (shadowed in green) and the sky transmission in the mid-IR(*) (in red). In both plots, all the absorption features mentioned above can be clearly distinguished.
The presence of water vapor in the atmosphere has an important impact on the sky transmission and emissivity in the mid-IR, particularly in the 20 micron window. The larger the amount of water vapor, the lower the sky transmission and the higher the emissivity. The figure below shows the model sky transmission at an airmass of 1.5 for three different values of the PWV (2.3, 4.3 and 10 mm). These ATRAN models were obtained from the Gemini Observatory web site and correspond to the conditions of Cerro Pachon. They are very useful to understand how the mid-IR sky background changes with PWV.
The figure shows that the sky transmission in the 10 micron window decreases by ~5% between 2.3 and 10mm of PWV. However, the sky transmission in the 20 micron band decreases between ~30% in the bluest part and ~80% in the reddest part, when the PWV increases from 2.3 mm to 10 mm.
The effect of PWV on the sky emissivity is the opposite to its effect on the transmission, and it is shown in the figure below.
The sky emissivity in the 10 micron window increases by ~5% when the PWV increases from 2.3 to 10 mm. The emissivity in the 20 micron window increases ~30% in the bluest part and ~80% in the reddest part, when the PWV increases from 2.3 mm to 10 mm. In the 20 micron window the width of the saturated water bands increases as the atmospheric water vapor content increases.
Hence, for practical purposes, the signal-to-noise ratio of a measurement is affected by high PWV through the combined effect of reduced tranparency of the atmosphere and its increased emissivity.
PWV Statistics at ORM
The PWV is highly variable and is one of the main parameters that define the infrared quality of a night. Based on its value, we can make the following rough classification:
- PWV <= 5 mm -- Good conditions to observe both at 10 and 20 µm.
- 5 mm < PWV <= 10 mm -- Good conditions to observe at 10 µm and poor conditions at 20 µm.
- PWV > 10 mm -- Possible to observe at 10 µm but very bad at 20 µm.
Since 2011 the IAC SKY team provides continuous real-time PWV measurements at the observatory, based on GPS measurements. These data can be found here. The following graph shows these measurement for nearly two years where the seasonal variation can clearly be seen.
Based on these data, the seasonal statistical behavior can be seen in the following figure. The difference of these recent measurements with the earlier study is not understood, but the existing on-line PWV measurements are used as the basis for the nightly assessment of the mid-IR quality of the sky, and appear to agree with the flux measurements made with CanariCam.
Data Quality and PWV
Even though the recent PWV statistics at the ORM shows realively high values, the impact that high water vapor conditions has on observations in the 10-micron window is not critical. The following figure shows the sensitivity in the Si5-11.6 filter as a function of PWV based on observations of standard stars taken from June 2011 to September 2013. The size of the symbols in the plot is proportional to the FWHM of the PSF in each image. The plot shows an important scatter in the sensitivity with a shallow positive slope, which can be fit with a straight line. Even though the plot shows a trend to worse sensitivity towards higher PWV values, there are other factors that are also contributing to the degradation in sensitivity. For instance, the larger symbols, i.e. data with worse image quality, tend to be located in the upper part of the plot, i.e. where the sensitiviy is worse. The straight line fit shows that when the PWV changes from 5mm to 15mm, the sensitivity is degraded by a factor of ~2. However, the data also show that a night with relatively low PWV and bad image quality can yield the same sensitivity as a night with high PWV and good image quality.
The situation in the case of observations in the 20-micron atmospheric window in high PWV condition is more critical. The following figure shows the sensitivity in the Q1-17.65 filter as a function of PWV based on observations of the same standard stars as in the previous figure. The plot clearly indicates a worsening of the sensitiviy as the PWV increases. In fact, in this case the sensitivity is degraded exponentially with the PWV, as shown in the exponential fit to the data. Roughtly, when the PWV varies from 5mm to 15mm the sensitivity is degradaded by a factor of approximately 10. Since the impact of the seeing at 20 micron is less than at 10 micron, most of the data points in the plot have the same diameter. Still, one can clearly see that when the seeing is so bad that can degrade the image quality even at 20 micron, the sensitivity is also clearly degraded (e.g. data point at approximately (11,290)).
The previous plots are consistent with the expected behavior of the atmospheric transmission and emission as a function of water vapor content, which was shown at the beginning of this section based on ATRAN models. Furthermore, the behaviour seen in the Si5-11.5 filter can be extended to most the 10-micorn band filters that are relatively clean of strong water vapor bands, as it is shown in the sensitivity comparison tables shown in the section Imaging Sensitivity.
The following plot shows the PWV as a function of the ratio between the flux of mid-IR standards in the Si5-11.6 filter and in the Q1-17.65 filter. Each data point corresponds to a different standar star. The standard represented in this plot are the same as in the previous two plots. The fluxes in each filter were calculated using aperture photometry (QPHOT task in IRAF) in the optimal aperture, i.e. the one that maximizes the SNR. The fluxes in ADU were converted into electrons using the nominal gain of the detector.
The Si5-to-Q1-ratio plot indicates that the PWV increases logarithmically as the Si5/Q1 increases, which is one would expect when looking at the bahaviour of the sensitivity in both filters as a function of PWV. This plot can be used during the observations to help astronomers to assess the quality of the night in the unlikely event the real time PWV monitor is off line. Note that to evaluate the quality of the night by measuring the Si5/Q1 ratio in a standard star has the disadvantage that one needs to spend time observing the standard star with CanariCam before actually knowing the quality of the night. However, the real-time PWV monitor based on the GPS technique operated by the IAC provides this information in advance to deciding using CanariCam and this is why the monitor is the main decision-making tool regarding the observing queue that we use at the telescope.
The ubiquitous presence of mid-IR background forces instrument builders to locate all the opto-mechanical components and the detector under extreme cryogenic conditions. In the case of CanariCam, the motors that drive the opto-mechanical components work at a temperature of ~70K, while the nominal working temperature of the detector is about 9.5K.
Fast detector electronics is required to evacuate pixels before they become saturated by the massive arrival of background photons. Contrary to most optical detectors, mid-IR detectors are shutterless. The detector is continuously exposing and every few tens of milliseconds there is a readout and a detector frame is completed. The electronics takes care of co-adding detector frames into a buffer, until all co-added frames are saved into a disk (save-set). Typical frame exposure times for CanariCam are ~25 milliseconds, and co-added frames are saved typically every 10 seconds.
Still, in each single frame, we would normally see only background emission and detector noise, with no trace of the astronomical source at all. To be able to see the source of interest, a technique called "chopping and nodding" is used. Chopping consists of swinging the telescope secondary mirror at a typical frequency of a few Hz between the position of the source (on-source) and the nearby sky (off-source). Since the background seen by the instrument is different when the secondary mirror is oriented on-source and off-source, the telescope axes are actuated (nodded), typically every 30 seconds, to swap on- and off-source positions. The light of the astronomical source can be unveiled by a given combination of summations and subtractions of the chop-nod frames (see more details in the CanariCam User Manual and in the section Observing Strategy.
To illustrate the chopping technique and the importance of the background in the mid-IR, next figure shows (a) one on-source save-set, (b) one off-source save-set and (c) the subtraction of both. These are actual CanariCam data of a mid-IR standard star taken with the Si2-8.7 filter during the commissioning run held on September, 2010. In this case, the observation was done in chop mode (without nodding) with a chop frequency of 3 Hz. The detector frame time was 25 milliseconds, and 44 frames were coadded into each save-set. Therefore, each save-set has an exposure time of 1.1 seconds. In the on- and off-source save-sets there is an average of ~420000 ADU/pix. No trace of the star is seen at all in the on-source save-set. CanariCam has 16-bit analog-digital converters, which means that individual frames saturate at 65000 ADU. Hence, pixel wells were filled in each 25 ms frame up to ~9500 ADU, i.e. 14% of their capacity. The vertical stripes in frames (a) and (b) are the traces of the 16 detector readout channels. Once the off-source frame is subtracted from the on-source frame (c), we see the light from the star, which has a peak of 8000 ADU. Therefore the peak star light amounts only 2% of the light arriving from the background, and this is a rather bright star (of several Jy). Normally, science targets of interest are tens to hundreds of times fainter.
Fortunately, when observing with CanariCam the user only has to worry about the total on-source exposure time. CanariCam software will take care of optimising the frame times and save times and of rejecting the appropriate number of frames while the secondary mirror is in motion between chop beams as well as while the telescope is in a nod transition.
Another important factor to bear in mind when preparing an observation with CanariCam@GTC is that at mid-IR wavelengths, the spatial resolution of the delivered images is no longer seeing-limited, but diffraction-limited. In next figure we see that the seeing decreases as ~ λ-0.2 while the diffraction limit of a telescopes grows as ~ λ/D.
The superior spatial resolution delivered by mid-IR instruments built for large aperture ground-based telescopes makes them very competitive when compared with space-based observatories. In space-based observatories, the sensitivity is higher due to the lack of background emission, but large apertures are technically very challenging. For instance, the typical FWHM of the MIPS@Spitzer PSF at 20 micron is ~5.5 arcsec, while CanariCam@GTC delivers near diffraction-limited images at 20 micron with a FWHM of 0.5 arcsec, i.e. a factor of 10 better than Spitzer.
We have used actual CanariCam data from the first years of scientific operation to determine the statistics of the image quality in several key filters in the 10 and 20 micron observing window. Statistical values of the FWHM of the PSF (in arcseconds) are presented in the following table corresponding to the filters, Si2-8.7, Si5-11.6 and Q1-17.65.
The Si5-11.6 and Q1-17.65 data correspond to the period between between June, 2011 and May, 2014. These data were extracted from the observations taken regularly in the Si5-11.6 and Q1-17.65 filter to determine the quality of the night in terms of water vapor presence in the atmosphere. The following two figures show the statistical distribution as well as the accumulated distribution of the FWHM in both filters.
The Si2-8.7 data correspond to the period between between December, 2012 and March, 2013. These data were extracted from the standard star observations taken as part of the CanariCam scientific programs.
Last modified: 14 August 2014