TANSPEC: TIFR-ARIES Near Infrared Spectrometer

A calibration unit is used for the flat fielding and the wavelength calibration of the raw spectra taken through this spectrograph. As shown in Fig 2, it has an assembly of calibration lamps, integrating sphere, camera lenses, and a fold mirror (mounted on a linear translation stage that moves into the telescope beam and picks off the beam from the calibration lamps when required). The pick-off mirror and lamp control can be remotely operated from the instrument software. The calibration optics are designed in a way to match the telescope’s f/number and to evenly illuminate the TFP inside the TANSPEC with a 15 mm diameter flat field. This is the size required to cover the longest slits (60 arcsec). A 12.5 mm diameter exit port (field stop) of an integrating sphere serves as the uniform (Lambertian) flat field. This field is magnified to 15 mm diameter by a 246 mm focal length lens (Lens 1) placed 82 mm from the exit port. A 125 mm focal length CaF2 lens (Lens 2) re-images this field at 1:1 onto the f/9 TFP inside the cryostat. At the same time, Lens 2 images the 15 mm aperture stop of the Lens 1 onto the telescope secondary mirror to create an exit pupil identical to that of the telescope so that the calibration light sources illuminate the instrument identically to the telescope. The root mean square (RMS) spot diameter in the TFP from the calibration optics is about 0.5 mm. This is sufficient enough to satisfy the uniformity requirement of better than 1% across the slits, which is limited by the integrating sphere. The calibration unit has Argon and Neon Discharge lamps⁹⁹9https://www.newport.com/f/pencil-style-calibration-lamps and two Tungsten continuum (of different brightness) lamps which cover the entire wavelength range (0.55-2.5 $\mu$m) of the spectrograph.

The function of the fore-optics is to re-image the TFP onto the slits with the necessary magnification required for the spectrograph. At the same time an image of the entrance pupil needs to be formed where a cold stop can be located to baffle thermal flux and off-axis scattered light from the telescope. An optimized cold stop cannot be located in the spectrograph since the slit diffractively blurs the entrance pupil image. In the TANSPEC, this is achieved with an Offner relay (cf. Fig 3), which consists of three on-axis spherical mirrors. The primary and secondary of the offner relay consists of two concave and one convex mirrors, respectively. The two mirrors of the primary are a little different in focal length so that we can convert from f/9 beam to f/12 beam which is the desired F-number of the beam at the slit entrance to the spectrograph optics. This results in a slight reduction of the Offner optical performance that is not significant for the system performance. The Offner-relay is shown in Fig 3 with the folding flat mirrors removed. It re-images the f/9 $60^{\prime\prime}\times 60^{\prime\prime}$ TFP onto the slit plane at f/12. We painted the center of the cold secondary of Offner relay (onto which telescope pupil is imaged) black to work as a cold stop. Being all reflecting, the Offner relay is completely achromatic and has high and constant throughput across the desired wavelength range.

The IR slit viewer allows spectra of very red and embedded objects through NIR guiding. It is also a fairly capable imager. The slit viewer is a collimator-camera re-imaging system consisting of BaF2-LiF collimator and LiF-BaF2-ZnSe camera lenses that re-image the $60^{\prime\prime}\times 60^{\prime\prime}$ slit plane onto H1RG array with an image scale of $\sim 0.25^{\prime\prime}$ (cf. Fig 2). For Nyquist sampling of the field of view (FOV) reflected of the slit, only the central $480\times 416$ pixel² area (starting from x=416 and y=288 pixels) is used. The collimator doublet forms another image of the system entrance pupil where slit-viewing filters are placed (each 25.4 mm diameter). This second stop is not optimized to cold baffle since the cold stop in the fore-optics does this. However, it is the best location for the imaging filters (small and uniform pupil image). The slit viewer has a filter-wheel with 12 slots to accommodate filters of different wavelengths. This wheel have position accuracy of 50 $\mu$m. The details of the filters-wheel and filters used in the TANSPEC slit-viewer are given in Table 1 and Table 2, respectively. The $Y,J,H,K_{s}$ filters are in Mauna Kea Observatory (MKO) system [2] make by Asahi Spectra, Japan¹⁰¹⁰10https://www.asahi-spectra.com/. The $r^{\prime},i^{\prime},H_{2},$ and $Br\gamma$ filters are made by Andover Corporation, USA¹¹¹¹11https://www.andovercorp.com/, where the $r^{\prime},i^{\prime}$ filters are in the Sloan digital sky survey (SDSS) system [3]. The response curves for all the filters are provided in Fig 4.

The entire mechanical assembly of the TANSPEC is shown in Fig 8. The TANSPEC instrument has an envelope size of 1900 mm (width) x 915 mm (length) x 1148 mm (height) which is within the telescope instrument envelop. The weight of the TANSPEC and its mounting plate are 650 kg and 350 kg, respectively. Therefore, to balance this instrument and maintain the center-of-gravity of the 3.6-m DOT, an extra frame with counterweights (1000 kg) on it is added to the instrument. Electronics cabinet containing power-supplies,

Lakeshore, USA temperature controller¹¹1https://www.lakeshore.com/products/categories/overview/temperature-products/cryogenic-temperature-controllers/model-336-cryogenic-temperature-controller and Western Telematic Inc. (WTI), USA Ethernet power switches²²2https://www.wti.com/collections/power-transfer-switch are also mounted on this frame. The TANSPEC is attached to the telescope through a mounting plate via a rigid interface box. This box is mounted in between the top plate and the vacuum jacket. A calibration unit is placed inside this box. The box is 280 mm deep and the walls are of 25 mm thickness. The top plate of the vacuum jacket contains the liquid nitrogen fill port, entrance window, and the mount for closed-cycle cooler. The TANSPEC along with its frame mounted on the 3.6-m DOT is shown in Fig 1.

The vacuum jacket of the TANSPEC has a four-sided center section weighing about 205 kg and is bolted by two large O-ring sealed covers, each weighing about 38 kg. The four-sided center section is made from four plates of 6061-T6-Aluminium which are electron-beam welded together. The vacuum jacket has a external dimensions of 1070 mm (length) x 860 mm (height) x 630 mm (width). The vacuum jacket has no hard connections except at the rigid top plate (25 mm thick). Therefore, if we can keep the flexures below the yield strength of the Aluminium along with the flexure of shafts (connected to the warm motors mounted outside of the vacuum jacket) within the specifications, we can achieve a very stable configuration by using a thick vacuum jacket walls of 25.4 mm. Although, the large vacuum jacket covers are with ribs that taper up to 70 mm. All the internal surfaces are mirror finished to provide good radiation shielding and to improve the liquid Nitrogen (LN₂) hold time. Inside this there is an radiation shield which is closed-cycle cooler cooled shield that removes most of the thermal radiation from the vacuum jacket. The wall of the LN₂ cooled cold bench/box is 12.5 mm thick and hangs inside the vacuum jacket through fiber-glass V-trusses (3) which are connected to the top plate inside the radiation shield. The detectors are thermally clamped to the cold structure. The radiation shield is mounted around cold structure through four shear webs connected to the top of the vacuum jacket.

The TANSPEC has mechanisms for the slit wheel, guide channel filter wheel, grating wheel, spectrograph channel detector focus and calibration unit pick-off mirror. It uses external warm smart motors manufactured by Animatics Corporation, USA.³³3https://www.animatics.com/ mounted on one of the two covers of the vacuum jacket and are coupled to the wheels through ferrofluidic vacuum feed-throughs. The optics and detectors can be accessed by unbolting the opposite covers. Thees is no need to disassembled motors for majority of the troubleshooting inside the cryostat. The array controllers are mounted on the outer surface of the cryostat. The mechanisms have three different categories, i.e., continuously variable wheels, discrete position wheels with detents, and a linear flex stage. For the position sensing, we have used the non-contact Hall effect devices similar to the SpeX instrument [4]. All the wheels are equipped with a stationary Hall effect sensor. These sensors are used to encod wheels position by placing the magnets at their edges. Opposite polarities signifies the homing and other positions. The Hall effect sensors outputs are passed through the comparator circuits and function like limit switches. Stepper motors of 400 steps/revolution in the worm-gear mechanisms is used to drive all the wheels with detented positions and the final position is determined by counting the steps taken from the home position. The H2RG array focus stage consists of an Aluminium alloy flex stage with anti-backlash linear position mechanism which provides $\pm$2.5 mm of focus travel through a screw coupled with a stepper motor. The calibration lamp assembly has a plane mirror which can be moved by using an motor and linear stage arrangement.

The required cooling of this instrument is achieved through a hybrid system with closed-cycle cooling for the radiation shields and LN₂ cooling for the optical bench and detectors. In order to keep the instrument background below the HxRG detector dark current ($\sim$0.1 electron/sec), LN₂ cooling of the cold structure is done. The radiation shield which surrounds the cold structure is cooled by using a closed-cycle cooler. A cooling temperature of $\sim$150 K is sufficient for the radiation shield to reduces the radiation load on the cold structure which can be managed very easily by the LN₂ can. The surface areas of the cold structure, radiation shield and the vacuum jacket are 1.8 m², 2.6 m², and 4.3 m², respectively.

We have uses a CTI model 9600 high voltage compressor⁴⁴4https://www.idealvac.com/files/manuals/Brooks9600High_1.pdf manufactured by Brooks, USA⁵⁵5http://www.brooks.com along with two long (40 meters each) pressurized Helium lines to cool the highly polished 5 mm thick 6061-aluminum radiation shield to less than 150 K. The estimated heat load on the cold structure consists of 2.1 W heat from the radiation shield having temperature of 150 K, 1.9 W from the holes in it (entrance window etc), 1.7 W from support V-trusses, 1 W from the wiring, and 0.25 W from the motor shafts. The expected variation in the performance of the closed-cycle cooler does not effect the temperature of the cold structure as its heat load is negligible. The temperature variation due to the boil-off of the LN₂ ($\sim$0.5 K) is also small to affect the instrument performance (e.g., instrument background or focus).

The instrument can cool-down in $\sim$60 hours and the hold time for the 14.16 L capacity LN₂ can is $\sim$100 hours within the ambient temperature range of 283 to 301 K. Optical bench and guide detector are maintained at $76\pm 2$ K. Warm-up also takes 60 hours, but by bowing out the LN₂ by circulating dry air, the cryostat can be warmed up faster .

The array is continuously reset pixel by pixel when an exposure is not being taken. When an exposure is requested the present background reset is completed and then the requested exposure is done. This reduces the after image and avoids hard saturating array and also it sets up a consistent thermal cadence and reduces the thermal anomaly effect. The detector can be read out in various modes, i.e. Single Read, Double Correlated Sampling and the Sample-Up-The-Ramp (SUTR). The SUTR is the recommended mode of readout as there will be reduction in the readout noise due to the multiple sampling of the ramp values which reduces the readout noise to about 5 electrons. The SUTR readout technique is also useful when the pixels saturated or are hit by cosmic rays (see for details, [5]). The flux rate in each pixel is calculated by fitting a line to the counts along the time axis on the SUTR readout data cube. The H2RG array has sub-array capabilities in Y-axis, whereas, the H1RG has this capability in both the X and Y axes of the array. There is also in-built software to guide on slit image and send corrections to TCS Software to guide on-field objects. In H2RG array, four vertical-strips are read out simultaneously at a rate of $\sim$5 $\mu$s per pixel and in total, 5.263 seconds are required to complete one full frame read out. Therefore, the exposure time is adjusted automatically to the nearest multiple of 5.263 second (non-destructive readouts of the cumulative counts) and is written to the header of the image file. For H1RG array, it take 1.877 seconds to readout the frame and therefore, the actual exposure time is adjusted to the multiple of 1.877 second. In the top panel of Fig 12, we show the image and distribution of the counts for the first readout of the H2RG array. As the H2RG is readout via 4 channels, we can see four strips for that in the image. A broad intensity distribution with high values as expected in the bias level of NIR detectors is seen. In the lower panel of Fig 12, the same has been shown for the second readout by subtracting the first readout. The sampling clearly shows a Gaussian distribution with mean around zero and sigma of $\sim$25 Analog to Digital Unit (ADU) which is equivalent to the readout noise of the H2RG array in low-gain setting.

We have taken many frames of different exposures with the array exposed to the continuum lamp (for H2RG) and to the telescope enclosure with $Br\gamma$ filter (for H1RG). Then, for each pixel, we have plotted the counts and its variance (also know as photon transfer curve) and calculated the slop. The inverse of this slope will be the gain value (see for details, [5]). We have selected different regions of box size 5 $\times$ 5 pixels of the exposed area, totaling to $\sim$500 pixels, of both H1RG and H2RG arrays for the gain calculation. Fig 13 shows the histograms of the gains obtained for both H2RG and H1RG arrays. We have fitted a Gaussian function on these distributions and found the peak gain values as $4.5\pm 0.2$ $e^{-}$/ADU and $4.3\pm 0.2$ $e^{-}$/ADU for the H2RG and H1RG arrays, respectively. In high-gain setting of the H2RG array, we have found peak gain value of $\sim 1.12\pm 0.03$ $e^{-}$/ADU.

Dark current is estimated by putting cold block filter in the filter wheel (for H1RG) or the mirror in the slit wheel (for H2RG) and reading out the array. In the left panel of Fig 14, we show the variation of the counts in the dark frame as a function of time for the H2RG (for both high-gain and low-gain settings) and the H1RG arrays. A straight line fit to this distribution gave a dark current value of $\sim$0.016 ADU/sec and $\sim$0.013 ADU/sec for H2RG (low-gain) and H1RG arrays, respectively. This multiplied by gain-values gives us an estimate of the dark current as $\sim$0.07 e^-/sec and $\sim$0.06 e^-/sec. Similarly, in the high-gain setting of the H2RG array, the dark current is $\sim$0.07 e^-/sec. Few sets of dark frame readouts at the beginning and end of each night would be enough for its correction.

Since dark current is very small, for the readout noise calculation, we have used set of frames generated from the difference of consecutive frames of a dark exposure. We have calculated the standard deviation in the count values for each pixel in the above set of frames. This standard deviation is actually $\sqrt{2}$ times readout noise value of each pixel (see for details, [5]). The distribution of the readout noise for all pixels in the arrays are shown in the right panel of Fig 14. The readout noise for the H2RG (low-gain) and H1RG arrays is found be $5.3\pm 0.7$ and $3.5\pm 0.1$ ADUs, respectively. In the high-gain setting of the H2RG, the readout noise is estimated as $20.7\pm 3.5$ ADUs. Here, it is worthwhile to mention that the above readout noise values are for a single readout. The readout noise is a function of the number of readouts and will decrease with increase in the number of readouts, e.g., in case of H2RG, the readout noise will be 1.2 ADUs for 284 readouts (or 5.3 e^-, low gain setting). For more details please see [6].

Using the SUTR readout data of continuum lamps (for H2RG array) and Argon lamp with Br$\gamma$ filter (for H1RG), we have estimated the median saturation level of these arrays as $\sim$28000 ADU ($\sim$65000 ADU for high gain) and $\sim$32000 ADU, respectively, for low gain setting. In Fig 15, we show the variation of the median intensity value as a function of exposure time for these arrays for low gain setting. We can clearly see the saturation levels along with the non-linearity in the curves near these limits. We have estimated the median value of the upper limit for the linear regime of H2RG and H1RG arrays as 26000 ADU (57000 ADU for high gain) and 30000 ADU, respectively. As the median bias level for the H2RG and H1RG arrays are at $\sim$20000 ADU (20000 ADU for high gain) and $\sim$13000 ADU, the effective (useful) well depths for these arrays are found to be at the level of $\sim$6000 ADU ($\sim$37000 ADU for high gain) and $\sim$17000 ADU, respectively, in low gain setting. Here, it is worthwhile to mention that very bright spectral lines or stars can result in residual images that can contaminate the darks.

The pixels which deviate more than 8$\sigma$ from the median value in the image which is generated by dividing two flats (taken in low and high incident flux), are defined as bad pixels. The percentage of the bad pixels in the H1RG array is 0.03%, most of which are along the edges. For the H2RG array, which is used for spectroscopy, the different orders observed in continuum lamps were used to detect bad pixels which are negligible in numbers.

The pixels having counts more than 8$\sigma$ above the median value in a dark readout of 100 sec are defined as hot pixels. The fraction of hot pixels for the H1RG and H2RG arrays was found to be 0.05% and 0.03%, respectively. They are distributed randomly over the arrays. Similarly, cold pixels (pixels which have counts less than 8$\sigma$ below median value in a dark readout of 100 seconds) fraction was found to be negligible for both the arrays.

The main parameters of the H2RG and H1RG arrays used in the TANSPEC instrument are listed in Table 4.

The detailed procedure for observing through the TANSPEC is provided in the TANSPEC observation manual.⁹⁹9https://aries.res.in/sites/default/files/files/3.6-DOT/Tanspec-Observation-Manual.pdf Briefly, the observer selects the required slit and spectroscopy mode (XD mode or prism mode) through slit wheel and grating wheel, respectively. A desired filter can be selected through the filter wheel and the target source is then imaged in the slit viewer/guider. Afterwards, the target source is placed in the center of the slit by giving off-sets to the telescope. Guiding can also be done with the slit viewer through any filter (optical to NIR) on an object in its FOV. The magnitude limit for auto-guiding is roughly about J $\sim$10 mag in an exposure of 10 s. Telescope’s visible guider can also be used to guide stars. Once guiding is started, long integration of several hundred seconds can be done on the faint sources. In order to correct the spectra for the sky lines, telluric standards should be observed near to the target source at about similar airmass. Calibration frames such as arc-lamps, dark, flats are usually taken at standard star position by executing a calibration macro. All updates and information about the TANSPEC instrument are provided at ARIES 3.6-m DOT instrument page.¹⁰¹⁰10https://aries.res.in/facilities/astronomical-telescopes/360cm-telescope/Instruments The initial version of the data reduction pipeline, pyTANSPEC¹¹¹¹11https://github.com/astrosupriyo/pyTANSPEC, which supports the XD mode data reduction for slits of width 0.5 and 1.0 arcsec is also released.

By using the slit-viewer in the TANSPEC mounted on the 3.6-m DOT, we can image celestial objects on the H1RG array.

For the efficient throughput, the entire optics needed to work efficiently across the entire 0.55-2.5 $\mu$m range of the instrument. The theoretical estimates of the throughput of the TANSPEC instrument in different modes at different wavelengths are given in Table 5. These numbers are estimated for the best optics performance of the system including the telescope mirrors, instrument mirrors/lens/prism/grating/substrate and quantum efficiencies of the arrays at different wavelengths. However, these estimates can be treated as upper limits as they do not include the atmospheric transmission and the slit losses. As expected, there is a decrease in throughput at shorter wavelengths ($r^{\prime}$ and $i^{\prime}$ bands) due to the combination of decreasing quantum efficiency (QE) of the array (including its own single-layer anti-reflection coating) and decreasing efficiency of the Broadband Anti-Reflection (BBAR) coatings on the transmissive optics. The peak throughput of the XD mode is about $\sim$30%. On the other hand, the peak throughput of the prism mode ( $\sim$50%) is far more than that of the XD mode The imaging mode peak throughput is estimated as $\sim$60%.

The throughput of the entire system in the imaging mode was also measured from the total flux of standard stars in an area of 4 times of the FWHM of their profile. The imaging mode throughput percentage is $30\pm 5$%, $45\pm 5$%, $44\pm 5$% and $35\pm 5$% for the $r^{\prime}$, $J$, $H$ and $K_{s}$ bands, respectively. The lower $K_{s}$ band throughput is may be due to the higher background at these wavelengths, i.e. thermal background and atmospheric lines. These numbers also includes other parameters such as QE of the detector, instrument/telescope optics and the atmospheric turbulences. The throughput in spectroscopy mode is very difficult to measure as it involves placing the star at the exact center of the slit every time. However, the throughput in the prism and XD modes was measured on an A0V star through 1 arcsec wide slit. We used the $J$ band region of the spectrum to calculate the throughput. The theoretical $J$ band flux for the star was calculated using Pogson’s relation¹¹1M=2.512*log(F) + C where the zero point of the magnitude is taken from literature [10]. This value was converted to photon counts by dividing it with the energy corresponding to the $J$ band wavelength. The observed photon counts were converted to units of counts/m²/s by dividing it with the effective collecting area and the total integration time. The ratio of the observed values to the theoretical values are obtained as 21% and 37%, which are the throughput values of the spectrograph in the XD and prism modes, respectively. As expected, the measured throughput values are lower than the theoretical estimates due to the atmospheric transmission and the current reflectivity of the primary mirror of the 3.6-m DOT, i.e., 45-60% in optical bands. Slit losses can also contribute to the low throughput values and a wider slit, e.g., 4${{}^{\prime}}{{}^{\prime}}$, can help to minimize those losses.

During the commissioning nights of the TANSPEC instrument, we have observed various astrophysical sources in the spectroscopy as well as imaging modes. We have done observations of M53 globular cluster using the slit-viewer in the TANSPEC on the 3.6-m DOT. As the FOV of the slit-viewer is small ($\sim 60^{\prime\prime}\times 60^{\prime\prime}$), we have done observations of M53 in five pointings in $J$, $H$ and $K_{s}$-band filters with 25 Non Destructive Reads (NDRs). The readout time of each NDR is 1.877 secs. In Fig 23, we show the comparison of the mosaiced image of the slit-viewer with that of 2MASS survey. We can easily see a better resolved and deeper image of the TANSPEC. Fig 24 shows color-composite images created from the $J$, $H$ and $K_{s}$ images and $J$ band images of some of the sources observed with the TANSPEC. Fig 25 shows the optical-NIR spectra obtained from stars with different spectral types in both the XD and prism modes of the TANSPEC. The black curve is the atmospheric transmission at Mauna Kea.²²2http://twiki.cis.rit.edu/twiki/bin/view/Main/MaunaKeaTo100kmAtmosphericTransmissions The shaded regions are thus the regions where atmospheric features dominate the spectra and hence, we have not shown those wavelength ranges in our extracted spectra. We can easily identify various spectral features (marked as blue vertical lines) in both the spectra.

We thank the staff of the 3.6-m DOT and 1.3-m DFOT, Devasthal (ARIES), for their co-operation during the observations. It is a pleasure to thank the members of the IR astronomy group (Department of Astronomy and Astrophysics) at TIFR for their support during observations. We acknowledge the Spex instrument³³3http://irtfweb.ifa.hawaii.edu/~spex/overview/overview.html which we relied heavily on for the initial concept and the many practical lessons learned from that instrument. The Project In charge (PI) and Co-PI of TANSPEC project are thankful to the funding agencies (Department of Science and Technology (DST) and Department of Atomic Energy (DAE), Government of India, India) for approving the project and Directors of TIFR and ARIES for their support and encouragement. Thanks to the ARIES Governing council and Project Management Board (PMB) (3.6-m DOT) for reviewing time-to-time the progress of the project, and finally, the DOT and TIFR teams for their support during installation and commissioning of the TANSPEC on 3.6-m DOT. Finally, TIFR team acknowledges the support of the DAE, Government of India, under project Identification No. RTI 4002.