Abstract
OSIRIS is a near-infrared integral field spectrograph operating behind the adaptive optics system at W. M. Keck Observatory. While OSIRIS has been a scientifically productive instrument to date, its sensitivity has been limited by a grating efficiency that is less than half of what was expected. The spatially averaged efficiency of the old grating, weighted by error, is measured to be 39.5% ± 0.8% at λ = 1.310 μm, with a large field-dependent variation of 11.7% due to efficiency variation across the grating surface. Working with a new vendor, we developed a more efficient and uniform grating with a weighted average efficiency at λ = 1.310 μm of 78.0% ± 1.6%, with field variation of only 2.2%. This is close to double the average efficiency and 5 times less variation across the field. The new grating was installed in 2012 December, and on-sky OSIRIS throughput shows an average factor of 1.83 improvement in sensitivity between 1 and 2.4 μm. We present the development history, testing, and implementation of this new near-infrared grating for OSIRIS and report on the comparison with the predecessors. The higher sensitivities are already having a large impact on scientific studies with OSIRIS.
Export citation and abstract BibTeX RIS
1. Introduction
In the last decade, the combination of a near-infrared integral field spectrograph (IFS) and adaptive optics (AO) has proven to be crucial in a range of astronomical studies from our solar system to galaxies in the early universe. Some example observations include the sulphur dioxide distribution on one of the Galilean moons, Io (Laver & de Pater 2009); morphology of novae ejecta (Lyke & Campbell 2009); the atmosphere of an extrasolar gas giant planet (e.g., Barman et al. 2011; Konopacky et al. 2013); the crowded stellar fields of the Galactic center (e.g., Trippe et al. 2008; Do et al. 2009, 2013); an active galactic nucleus (AGN) (e.g., Davies et al. 2007; McConnell et al. 2011; Contini et al. 2012); and high redshift galaxies (e.g,. Förster Schreiber et al. 2006; Law et al. 2009; Wright et al. 2009; Wisnioski et al. 2011). IFSs are also aimed to be the first light instruments for the next generation of extremely large telescopes, such as the InfraRed Imaging Spectrograph (IRIS) on the Thirty Meter Telescope (TMT) (Larkin et al. 2010), HARMONI on the European Extremely Large Telescope (E-ELT) (Thatte 2010), and the Giant Magellan Telescope Integral Field Spectrograph (GMTIFS) on the Giant Magellan Telescope (GMT) (McGregor et al. 2012).
OSIRIS (OH-Suppressing Infrared Imaging Spectrograph; Larkin et al. 2003, 2006), a moderate spectral resolution (R ∼ 3800) diffraction-limited IFS for the AO system at W. M. Keck Observatory, is one of a handful of IFS instruments in use with AO systems worldwide today. It was the first diffraction-limited IFS instrument to use a lenslet array as the sampling element on the sky and has plate scales ranging from 0.02'' to 0.1'' per spaxel.6 OSIRIS' optics and lenslet array produce low noncommon path error (< 30 nm rms), a factor of approximately 3 times less than any other IFS, preserving the diffraction-limited point-spread function of the Keck AO system (Wizinowich et al. 2006).
OSIRIS was designed with a single fixed diffraction grating to ensure spectral stability and make data reduction possible with very dense spectral packing on the detector (only 2 pixel spacing between spectra). The grating is used in multiple orders (m) to cover traditional near-infrared wavebands: K (λcen = 2.2 μm) is sampled in m = -3, H (λcen = 1.6 μm) in m = -4, J (λcen = 1.3 μm) in m = -5, and Z (λcen = 1.1 μm) in m = -6.
While OSIRIS has been a productive instrument to date, its performance has been limited by sensitivity, which is approximately 50% lower than its design prediction, particularly at shorter wavelengths (Z and J bands). Through our team's investigation, this performance limitation has been determined to be due to the quality of the spectrograph's diffraction grating. Since 2009, our team has actively pursued acquiring a new grating for the OSIRIS spectrograph. In 2011, we began to work with the Bach Research Corporation, Boulder, CO, to fabricate a new, more efficient grating for OSIRIS. Our goal was to improve the grating performance sufficiently to double the signal-to-noise ratio for detector limited observations.
In this paper, we describe our acquisition and testing of a new grating. In § 2, we summarize the history of the OSIRIS spectrograph grating. In § 3, we describe the laboratory setup used to measure the grating efficiency and the results of those measurements. In § 4, we report on the 2012 December installation of the new grating. In § 5, we discuss the on-sky performance of OSIRIS with the new grating. For those interested in equipment characterization, Appendices
2. History Of Osiris Grating
The OSIRIS spectrograph grating is a unique and unusual single fixed diffraction grating that has a coarse ruling of 27.93 grooves mm-1 at a shallow blaze angle of 5
76. The specifications of the grating are listed in Table 1. The grating design was done by Richardson Gratings, Rochester, NY, in collaboration with SSG Precision Optronics, Inc., Wilmington, MA, the designers and fabricators of the OSIRIS collimator and camera three-mirror anastigmats. Over the time OSIRIS has been in service at Keck Observatory, we have installed three different gratings. They are summarized in Table 2.
 |
 |
Originally, SSG manufactured two large aluminum grating blanks and provided these to Richardson Gratings for ruling. However, this first option for ruling was abandoned by them due to the large amount of tool pressure that would be required. A new vendor, Diffraction Products, Inc., Woodstock, IL, then agreed to take on the challenge of ruling this very coarse grating directly into a pure gold coating placed on the SSG aluminum blank. The resulting grating is identified as G1 in Table 2.
During laboratory testing of OSIRIS in 2004 October, it was determined that G1 had a slightly varying, incorrect (62 instead of 576) blaze angle. At high order, this puts the majority of the light into the wrong order and off the field of the detector. Efficiencies in the Z and J band were below 20% and even in the K band were below 30%. Due to time constraints, OSIRIS was shipped to the telescope with this imperfect grating while a replacement was ordered. Diffraction Products significantly improved their process, and a replacement grating with the correct blaze angle (called G2) was installed in OSIRIS in 2005 June.
Figure 1 is a photograph of diffraction spots on a wall by G1 (left image) and G2 (right image) at the Keck Observatory in 2005 June. The left image shows scattered light between the different orders due to incomplete ruling. This shows the improvement of the grating quality visually. The throughput measurement at the time of the servicing mission showed a gain of a factor of 3 to 4 in the J band with a smaller gain at longer wavelengths.
Fig. 1. Photograph of diffraction spots at the Keck Observatory in 2005 June. It shows a HeNe laser at 632.8 nm being diffracted by G1 (left image) and G2 (right image). The locations of diffraction spots were marked with black electrical tape above the spots. Note that G1 produces dramatic light loss between the orders compared to G2.
Unfortunately, even with G2, the throughput was still ∼50% of what was expected. This was later confirmed by our team during a 2009 October servicing mission. G2 was removed from OSIRIS, and its efficiency was measured at Keck Observatory using a 1.310 μm laser (close to fifth-order expected blaze wavelength) and an infrared camera. The resulting absolute efficiency measurements are shown in Figure 2. This was also verified with atomic force microscope (AFM) scans of G2. An AFM scan of one of the grating facets is shown in Figure 3. The grating facet shows a flat spot at the edge of the ruling, and the profile on the primary facet has at least two distinct angles. The effect of the curved profile on the steep side of the profile is to distribute some of the energy from the expected order into adjacent orders. The expected effect of the flat spot in the facets is that it causes some of the light to be scattered across all of the orders. We do not know if the same facet profile occurs throughout the grating, but both of these effects are clearly visible in our efficiency plot in Figure 2. Most likely, G2 has generally poor quality groove shapes, like Figure 3. If all of the energy between orders -5 and -6 in Figure 2 were concentrated in the expected order (m = -5), the efficiency at 1.310 μm would be >60%, as expected from the specifications.
Fig. 2. Measurements of absolute efficiency by order for G2 at the Keck summit in 2011. The blaze wavelength is 6.5 μm, so the fifth order should have the maximum power at ∼60% to 70%.
Fig. 3. AFM scan of one of the facets of G2 made by Diffraction Products, illustrating the curved profile on the facet, which decreases the overall sensitivity in each order.
At this point, the OSIRIS team began a search for a new vendor to manufacture a better quality grating. Bach Research Corporation, formed by the founding members of the Hyperfine company, began making custom astronomical gratings, and we selected them in 2011 to begin the process of ruling a new grating on the original SSG blank.
The first and second OSIRIS gratings were directly ruled into a gold coating applied to a machined one piece aluminum grating substrate and grating mount. Rather than directly ruling into the grating substrate, Bach Research suggested that we replicate the grating onto the aluminum substrate and then coat the replica with gold. The one-piece machined aluminum grating mount and substrate is an expensive component to machine, so we made use of the spare grating mount used during the first attempt of the grating by Diffraction Products. To produce the new grating, Bach Research removed the coating from the G1 mount and repolished it. This provided a new surface to apply a new ruling using a replication process. The fabrication was performed in two steps by Bach Research Corporation:
- 1.A new master grating was ruled onto a Zerodur substrate (a glass-ceramic composite material produced by Schott AG).
- 2.The new master was used for replication of the grating on a new coating on the spare substrate (called G3).
As part of the contract discussions for the new grating, Bach Research made a demonstration test ruling on a small 5 mm × 100 mm long substrate. A comparison of the diffraction of a HeNe laser at 632.8 nm with both this substrate and G1, taken at Bach Research, is shown in Figure 4.
Fig. 4. Diffraction of HeNe laser at 632.8 nm using G1 (left image) and the test ruling made by Bach Research (right image). The images were produced by Back Research.
The left image is the diffraction spots produced by G1, and the right image is by the test ruling. The light diffracted by the test ruling is well concentrated in spots, while G1 smears the light in the direction of dispersion.
One of the important challenges encountered in the manufacture of the previous gratings was that the grating maker could only evaluate the grating performance using a simple setup involving a HeNe laser with visual evaluation of the resulting dispersion and relative intensities in each order. The method used did not predict the grating's eventual performance at infrared wavelengths. Before installation of G3, we acquired an infrared laser source (1.310 μm) and infrared camera and created a setup that allowed measurement of the grating efficiency in a reliable fashion. This allowed us to evaluate the test ruling as well as the final grating before it was installed in OSIRIS.
3. Grating Efficiency Measurement
To investigate the grating performance in a more robust manner, we measure the direct efficiency of the grating at λ = 1.310 μm, which corresponds to a wavelength in the J band. In this section, we describe the measurement equipment, measurement setup, procedure, and discuss the measurement results.
3.1. Measurement Equipment and Stability
For the grating efficiency measurements in the infrared, we used an InGaAs camera (Raptor Photonics OWL SW 1.7 CL-320) and a 1.310 μm laser diode coupled to a SMF-28 fiber (a single mode fiber with a core diameter of 8.2 μm operating at 1.310 μm to 1.625 μm). The linearity of the camera and the stability of the laser are discussed in Appendices
3.2. Measurement Setup and Procedure
To measure the efficiency of the grating at the same configuration as in OSIRIS, we set up an optical path on an optical bench, where the angle of incidence (α) is α = -302, and the angle of diffraction (β) for m = -5 at 1.310 μm is β = 18°.4. We define the sign conventions and orientation for the OSIRIS grating used throughout our measurements in Figure 5. In our setup, the incident angle is accurate within 1°, which is theoretically a < 1% change in m = -5 efficiency at 1.310 μm (Fig. 6) using our rigorous coupled-wave analysis (RCWA; Moharam & Gaylord 1981) code.
Fig. 5. Orientation of OSIRIS grating blaze direction (black arrow), incident angle α (negative; blue line), and outgoing angle β (positive; red, dot-dashed line). Negative orders are defined to be in the direction towards the grating normal (GN) from m = 0 (pure reflection).
Fig. 6. RCWA prediction of the theoretical m = -5 efficiency change at λ = 1.310 μm due to the incident angle change. With a 1° change in the incident angle, the change in efficiency is less than 1%.
RCWA is a semianalytic computational method used to solve Maxwell's equations. Our code uses this method to calculate the portion of light being diffracted into different orders by a diffraction grating for the given grating specification, Table 1, and the incident angle.
A 1.310 μm laser diode that is coupled to a SMF-28 fiber is connected to an attenuator and a collimator. The collimated laser beam goes through two neutral density (ND) filters and hits the grating surface, where it is diffracted into its constituent orders. An achromatic lens pair focuses the beam on the InGaAs camera, which sits on a dovetail optical rail system. The beam's full width at half-maximum at m = -5 is about 1.7 mm on the camera. The schematic of the configuration is shown in Figure 7, and a photo of the setup is in Figure 8. An aluminum baffle box resides over the entire experiment to eliminate scattered background light. All the components, except the grating, were kept fixed onto the optical bench until all efficiency measurements were completed. Every time we left the lab, the grating was carefully packed and put away in a secured location.
Fig. 7. Schematic of the efficiency measurement configuration in the lab (not to scale). The entire setting is covered up by an aluminum baffle box to control background and scattered light.
Fig. 8. Photo of the efficiency measurement configuration in the lab.
The collimator focal length and the optical path length were determined by considering the divergence angle of the laser with the goal of keeping the final spot size well inside the detector field of view (FOV). The attenuator and combination of two ND filters are employed to ensure that the final spot on the detector is not saturated at a reasonable exposure time for the brightest order with good signal-to-noise ratio on the faintest orders. The achromatic camera lens pair is chosen so that only one spot falls on the camera's detector at a time.
The efficiency is defined as the flux of monochromatic light diffracted into the order being measured relative to the total flux. We measure the reflection of the same light source from an unruled area on the grating/test ruling (called pure reflection) and use this as the total flux. Since the unruled part of the grating/test ruling is outside of the clear aperture, and the quality of its surface is not guaranteed, we also use the total sum of all orders (called order sum) as a measure of the total flux as well. Efficiency measurements using both values for total flux are presented in this paper.
A typical efficiency measurement procedure is as follows: (1) set up the grating for the pure reflection; (2) close the bench baffle; (3) find the best exposure time and measure the pure reflection with the laser on; (4) measure the pure reflection with the laser off; (5) open the baffle; (6) set up the grating to place the first order to be measured in the camera's FOV; (7) close the baffle; (8) find the best exposure time and measure the flux with laser on and off; (9) move the IR camera to the next order; and (10) repeat steps 8 and 9 until all orders are measured. In these tests, we measured m = -13 to m = 8. It takes about 1 hr to complete this procedure. Since we know that the polarization state of the fiber does not change in 1 hr (Appendix
The ruled area of the OSIRIS grating is 205 mm × 230 mm. To assess the spatial dependence of the efficiency, measurements are made at nine locations (3 × 3 configuration) across the grating surface, as illustrated in Figure 9.
Fig. 9. Diagram showing the location of the grating efficiency measurements. The light yellow area is the region of the ruled area, 205 mm × 230 mm. The bright yellow area is the overall area of the substrate. The arrow indicates the direction of blaze angle.
After the efficiency at one location is measured, we move the grating sideways or change the height of the stage, where the grating sits, to move to the next location. The grating surface is kept parallel to the dovetail optical rail, and the optical path length is kept the same for all measurement. This allows us to keep the setup fixed as much as possible.
For each grating order, we acquire 100 frames with an additional 100 background frames. The 100 background frames are median combined to make a master background and subtracted from each science frame. Then 100 background-subtracted frames are median combined and divided by the exposure time to make the final reduced image. To conserve the optical alignment, we do not take flat frames. A histogram of a normalized flat field taken during testing shows a normal distribution with a standard deviation of 0.024. Instead of applying flat fielding to the final reduced image, we include the flat-field fluctuation of 2.4% as a part of measurement uncertainties.
A two-dimensional Gaussian function is fitted to the final reduced image to find the spot center, and the total flux in the spot is obtained by summing up the counts in the biggest circle that can be fit in the final reduced image centered at the spot center. A more detailed discussion of this method is found in Appendix
There were two efficiency requirements defined by Keck Observatory and our team that the new grating (G3) had to meet in order to be eligible for installation in OSIRIS:
- 1.Global efficiency requirement.—On average, the new grating has to be at least 50% more efficient than G2, which means >45% in the J band (1.310 μm).
- 2.Field dependent efficiency requirement.—The efficiency of the new grating has to be better than G2 efficiency (< 30%) at all locations across the grating.
In the next section, we report the results of the new grating efficiency measurements.
3.3. New (G3) and Old (G2) OSIRIS Grating Efficiencies
Before G3 was shipped to the Dunlap Institute for Astronomy and Astrophysics (Dunlap) in 2012 July, Bach Research assessed the quality of the wavefront of G3 surface with a 4 inch aperture Zygo interferometer. Bach Research took wavefront measurements along the center of the grating and moved the aperture from the start to the end of the ruling. They performed these measurements across several optical orders to yield an indication of wavefront error over the entire surface. The wavefront quality across the surface is roughly ± 0.5 wave at λ = 632.8 nm.
The peak efficiency (m = -5) of G3 with respect to the pure reflection at λ = 1.310 μm at nine spatial locations across the grating surface are summarized in Table 3, and Figure 10 shows detailed efficiency for -13 ≤ m ≤ 8.
Fig. 10. Efficiency of G3 by Bach Research with respect to the pure reflection (black lines) and with respect to the order sum (cyan lines), measured at -13 ≤ m ≤ 8.
 |
At all nine locations, the peak efficiencies are more than 75%, and the average efficiency, weighted by error, is 78.0%± 1.6%7 with respect to the pure reflection, and the nonweighted average is 77.8% with respect to the order sum (see footnotes in Table 3). This grating meets both global and field-dependent efficiency requirements stated in § 3.2, which led our team to install G3 in OSIRIS in 2012 December.
After G3 was installed and its on-sky performance was confirmed through engineering observations, G2 was shipped to Dunlap, and its efficiency was measured using the same setup used for G3.
Table 4 summarizes the peak efficiency at nine locations, and Figure 11 is the result of full efficiency measurements of G2. On average, G2 has a weighted efficiency of 39.5% ± 0.8% with respect to the pure reflection and 35.8% ± 0.7% with respect to the order sum.
Fig. 11. Efficiency of G2 by Diffraction Products with respect to the pure reflection (black lines) and with respect to the order sum (cyan lines), measured at -13 ≤ m ≤ 8.
 |
On average, G3 has a factor of about 2 greater efficiency at 1310 μm. We also find that G3 has a close-to-uniform efficiency across the surface compared to G2. The field-dependent standard deviation of G3 peak efficiencies is 2.23%, whereas the field-dependent standard deviation of G2 is 11.68%.
3.4. Measurement Uncertainties
To estimate the uncertainties in the efficiency measurement, we incorporate the configuration uncertainty of 1% and flat-fielding uncertainty of 2.4% (§ 3.2), random camera noise of ∼10 DN (Appendix
To confirm the repeatability of our measurements and the estimate of the error, we measured G2 efficiency at two locations, top-left and top-center (see Fig. 9 for the locations), twice, the second measurement about 2 weeks after the first measurement. The peak efficiency ratios of the first time to the second time are 0.982 (pure reflection) and 0.995 (order sum) for the top-left location, and 1.005 (pure reflection) and 1.011 (order sum) for the top-center location. They are both within the measurement uncertainties.
3.5. Polarization Effect on Grating
The polarization state of the incident light can affect the efficiency of a diffraction grating in many cases. To understand the polarization dependence on the OSIRIS grating efficiency, we modelled the transverse-electric (TE) (polarized parallel to the groove) and Transverse-magnetic (TM) (polarized perpendicular to the groove) efficiencies of the OSIRIS grating using RCWA. The RCWA model predicts that the peak TE to TM efficiency ratio for a 1.310 μm monochromatic light source at m = -5 is 1.041.
We conducted the polarized efficiency measurement experiments on the test ruling 3 times using: (1) the IR laser diode (no polarizer); (2) the IR laser diode in TE mode defined by the polarizer; and (3) the IR laser diode in TM mode defined by the polarizer. During the measurement, all the components, especially the fiber, were fixed to keep the same polarization state throughout. Figure 12 shows these results where the total efficiency is defined either by the pure reflection (left chart) or the order sum (right chart).
Fig. 12. Efficiency of the test ruling made by Bach Research with respect to the pure reflection (left chart) and with respect to the order sum (right chart), measured (-13 ≤ m ≤ 2) using TE/TM modes (magenta/cyan lines) and no polarizer (black lines).
The peak TE to TM efficiency ratio at m = -5 is 1.056 ± 0.092 and 1.029 ± 0.074 for the pure reflection and the order sum, respectively. These measurements are in agreement with the theoretical predictions.
We confirmed that the polarization state of the laser does not change during the full efficiency measurement at one spatial location (§ 3.2 and Appendix
4. New Grating Installation And Commissioning
OSIRIS was slowly warmed up to ambient temperature over a 1 week period, and G3 installation was performed in 2012 December by our team. G2 was removed from OSIRIS, and its alignment was measured and marked with respect to the mounting plate in the lab at the Keck Observatory summit facility. G2 was detached from the mounting plate, and G3 was aligned to the marks and installed on the mounting plate.
Figure 13 shows a HeNe laser at 632.8 nm being diffracted by G2 (top image) and G3 (bottom image) in the Keck summit lab prior to the installation of G3 in OSIRIS. The light is visibly more concentrated to 1 order for G3 while a higher fraction of light is diffracted to multiple orders by G2.
Fig. 13. Diffraction of a HeNe laser at 632.8 nm using G2 (top image) and G3 (bottom image). The test was performed in 2012 December. The light is more concentrated in 1 order for G3, whereas a high fraction of light is diffracted in multiple orders for G2.
On 2013 January 20 and 27, OSIRIS with G3 was commissioned on-sky using the Keck I AO system. Both nights started with a clear night, but unfortunately, within 1 or 2 hr, the weather conditions changed to thin/high cirrus with high wind. Some standard stars and blank sky were observed with the deformable mirror off. The resulting measurements are certainly affected by the varying weather conditions. The measurements of OSIRIS sensitivity with G3 in each spectrograph filter from these nights, as well as prior measurements with G2, are reported in § 5.
5. On-sky Performance And Throughput
In this section, we compare on-sky performance between G2 and G3. Our comparison is complicated by the fact that in early 2012 OSIRIS was moved from Keck II to Keck I to be the first dedicated science instrument for a new laser guide star (LGS) AO capability on Keck I that was installed in 2010. The LGS system on Keck I uses a significantly improved laser system compared to the existing Keck II laser system (Chin et al. 2010).
The final zero-point magnitudes at all broadband filters for OSIRIS are calculated using standard star observations. The data used to calculate G2 zero points were all taken on Keck II, and G3 data were all taken on Keck I. Although there are differences in throughput between the two telescopes and AO systems, as well as differing weather condition between our limited observations, there is a general zero-point improvement for G3 in comparison to G2.
We use as many standard star observations as we had access to, but the number of observations is fairly small in each particular band. For OSIRIS with G2 on Keck II, HD 105601, HD 106965, HD 201941, and HD 18881, taken between 2007 April and 2012 January, are used; and for OSIRIS with G3 on Keck I, HD 44612 and HD 18881, observed on 2013 January 20 and 27, are used. The zero points are calculated by applying a large rectangular aperture on a raw (nonreduced) image, over the entire spectrum. In all cases, approximately equal rectangles are used for the wavelength ranges. The resulting zero-point magnitude and the factors of improvement are shown in Table 5.
 |
6. Conclusion
OSIRIS at W. M. Keck Observatory is a particularly unique IFS instrument among other IFSs with AO capability today due to its use of a single, fixed, exceptionally coarse ruling (27.93 grooves mm-1) diffraction grating, which uses m = -3, -4, -5, and -6 to cover K, H, J, and Z bands. While OSIRIS has delivered a number of important scientific results, its sensitivity was limited by the performance of its spectrograph grating. Our team has worked with a new grating vendor, Bach Research Corporation, to produce a better quality grating for OSIRIS.
Bach Research manufactured a test ruling and the new grating (G3) in 2012, and we have carefully measured the direct efficiencies of both at 1.310 μm in the lab. The weighted field-averaged peak efficiencies of G3 are 78.0% ± 1.6% (pure reflection) and 77.8% (order sum) (see footnotes in Table 3) with field standard deviations of 2.23% (pure reflection) and 1.32% (order sum). After the G3 efficiency was confirmed to be high and close to uniform over the surface, G3 was installed in OSIRIS in 2012 December. G2 was shipped to Dunlap after G3 performance was tested and confirmed on-sky in 2013 January. G2 efficiency was measured as well using the same lab setup used for the G3 measurement. For G2, the weighted field-averaged peak efficiencies are 39.5% ± 0.8% (pure reflection) and 35.8% ± 0.7% (order sum) with field standard deviations of 11.68% (pure reflection) and 9.82% (order sum).
The new OSIRIS grating gives a factor of about 2 times increase in average efficiency at 1.310 μm with less field-dependent efficiency change across the surface. The final sensitivity improvement was difficult to assess because OSIRIS was moved from Keck II to Keck I in early 2012; however, we were able to determine the zero-point magnitudes and factors of improvement for each broadband filter. On average, on-sky throughput is 1.83 times better than when it was at Keck II with G2. This enables us to observe fainter objects and to use observing time more efficiently.
A single fixed diffraction grating with a coarse ruling can reach high efficiency and perform well on OSIRIS, but it is very difficult to fabricate such a grating today. For the next generations of IFS instrumentation, the more usual approach of using a finer ruling grating with m = 1 order would have less risk and would be cheaper. For example, IRIS on TMT, an IFS with some characteristics and design elements similar to OSIRIS, will instead have several gratings with finer groove densities of ∼150 to 900 grooves mm-1 (Moore et al. 2010; Larkin et al. 2010).
We particularly want to thank the Bach Research Corporation for all of their efforts and support during the selection and fabrication of the OSIRIS grating. We enjoyed working with the Bach Research Corporation team during this entire process. We also thank the Keck Observatory staff who helped with the planning and installation work for the new grating in OSIRIS. We would also like to thank the referee for their thorough read and valuable suggestions. The new OSIRIS grating was graciously funded by the Dunlap Institute for Astronomy and Astrophysics at University of Toronto. Funding for this project was also provided by the NSERC Discovery grant (RGPIN 419376) and the Canada Foundation for Innovation grant (31773). Our on-sky data presented herein were obtained at the W. M. Keck Observatory, which is operated as a scientific partnership among the California Institute of Technology, the University of California, and the National Aeronautics and Space Administration. The observatory was made possible by the generous financial support of the W. M. Keck Foundation. The authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of Mauna Kea has always had within the indigenous Hawaiian community. We are most fortunate to have the opportunity to conduct observations from this mountain.
Appendix A:: Camera Linearity
We evaluated the linearity of the InGaAs camera, Raptor Photonics OWL SW 1.7 CL-320, by measuring the average dark counts on the detector as a function of exposure time. The 100 dark frames per grating order taken at the time of efficiency measurements were median combined to make the master dark, and the average of the master dark was plotted as a function of exposure time on Figure 14. All camera settings were kept fixed for all times except for the exposure time.
Fig. 14. Average dark counts vs. exposure time at the time of efficiency measurements. The data were fit with a straight line, shown overplotted in magenta.
Figure 15 shows the absolute difference between data points and the straight line fit, with the Poisson noise overplotted in cyan. The result shows that regardless of the exposure time, there are about 10 DN fluctuations until the Poisson noise takes over at around texp = 10 ms. Hence, we include 10 DN in the noise calculation.
Fig. 15. Absolute difference between the average dark count and the linear fit plotted as a function of exposure time, with the Poisson noise overplotted in cyan.
We also looked at the effect of the camera temperature over 1 hr, which is the amount of time it takes for an efficiency measurement of a single spatial position on the grating surface over a large range of orders. In this test, we took images of the laser spot at texp = 0.2 ms every 60 s for 1 hr and repeated the process twice. During the test, the camera and the laser were kept on all the time. The laser was imaged to include the effect of the laser heating up the image sensor, in case that happened.
Figure 16 shows the changes of two different camera temperatures over a 1 hr period. Our input temperature (thermoelectric cooling (TEC) temperature) was always 15° C, and the sensor temperature was almost always constant around 15° C, but the PCB temperature increased about 10° C during the first test and about 6° C during the second test over a 1 hr period.
Fig. 16. Sensor and PCB temperature changes over 1 hr. With the TEC temperature set to 15° C, the sensor temperature stayed almost constant while the PCB temperature increased 5° C to 10° C. The magenta points are the measurements from the first test, and the black points are the measurements from the second test.
During G2 and G3 combined efficiency measurements, most samples were taken at texp = 30 ms. This is because G3 and G2 efficiency measurements were all taken with a frame rate of 25 Hz, and taking into account the trigger delay and data transfer, we set the maximum exposure time to be tmax = 30 ms. The exposure times for individual measurements were chosen to maximize signal level while maintaining the exposure below the saturation, but since we set a maximum exposure time threshold, many fainter spot images were taken with the maximum exposure time.
Figure 17 shows normalized average dark counts as a function of the PCB temperature change. The flux increases about 0.6% as the PCB temperature increases about 10° C. We do not have information on how long the camera was turned on during the experiments, but we know that to measure the full efficiency at one location, it takes about 1 hr, and in 1 hr the PCB temperature changes about 10° C (Fig. 16). Combining this information and Figure 17, we take 0.3% (a half of the full increase in flux) as noise due to camera PCB temperature changes during the efficiency measurements.
Fig. 17. Relative flux change of t = 30 ms (black crosses) and t = 0.2 ms (magenta circles) average dark counts due to the PCB temperature change.
Appendix B:: Infrared Laser Diode Stability Test
We used a 1.310 μm laser diode coupled to a SMF-28 fiber as a light source for infrared measurements. To ensure consistency in our measurements, we measured the stability of the laser by monitoring its intensity using the infrared camera tested in Appendix A. First, we fixed the exposure time to texp = 0.2 ms and took a series of images separated by three time intervals: 5, 20, and 60 s (Fig. 18). The camera and the laser were kept on until all measurements in a particular sequence were completed. These measurements allowed us to search for any time-dependent instabilities in the combined system of laser and camera. We found fluctuation in the system seems independent of the time interval of data taken but dependent on the duration of time the laser is on.
Fig. 18. Relative flux of the 1.310 μm laser diode vs. time for three time intervals: 5 (black circles), 20 (cyan crosses), and 60 (magenta stars) s.
The polarization of stimulated emission is parallel to the diode junction plane, and thus laser diodes are usually linearly polarized. When a laser is used in a polarization-dependent setup, intensity fluctuations can occur due to changing polarization states. Our laser diode is coupled to a SMF-28-J9 step-index fiber with a numerical aperture (NA) of 0.14 and an 8.2 μm diameter core. For this fiber, the dimensionless normalized frequency or normalized thickness of the guide (V):
2.75 at 1310 μm. k is the wavenumber, and d is the fiber core radius (e.g., Iizuka 2002). The first critical frequency (cutoff V) for single mode operation is 2.405, and therefore our fiber supports four polarization modes. To understand the polarization characteristics of our laser, we tested the laser stability with and without a calcite polarizer in the optical path. We took a series of images of the laser beam for 1 hr at 60 s intervals. We performed these measurements twice: once with a calcite polarizer in front of the laser and once without the polarizer. The polarizer was oriented so that the output beam had the maximum intensity (direction of the polarization of the laser is parallel to the direction of polarizer). After 1 hr of measurement, we verified that the peak intensity from the polarizer was still at the same angle. This means, over 1 hr, the polarization state of the laser did not change, and therefore we assume the laser polarization state does not change during the efficiency measurements.
Figure 19 shows that the laser flux fluctuates 3.5% (one-half of [highest—lowest] flux) over 1 hr. The initial discrepancy of up to 1000 s between the measurements with and without the polarizer is probably due to room/detector temperature differences because the two measurements were taken on different days. Since the polarization state of the laser did not change over 1 hr, this variation is probably from the laser itself, as shown in Figure 18. We take into account this 3.5% laser flux variation in the noise calculation.
Fig. 19. Relative flux of the 1.310 μm laser diode vs. time for a 60 s interval with a polarizer (black crosses) and without a polarizer (magenta stars).
Appendix C:: Total Flux In A Spot
The grating efficiency is measured by summing up the flux in an individual spot at a particular order and dividing by the total flux. To confirm that we collect all flux in an individual spot, we tested two methods. One is Gaussian fitting. A two-dimensional (2D) Gaussian function is fitted to the final reduced image, and aperture photometry is applied to both the reduced image and the 2D Gaussian function. Figure 20 shows the growth curves of the reduced image (black lines) and the Gaussian fit (magenta lines) for the test ruling. They are both normalized to the reduced image total flux. The solid line is the growth curve of the brightest spot, and the dashed line is of a dimmer spot. The optical path of the dimmer spot is the longest, and thus the spot size is biggest on the detector due to the divergence angle of the laser beam. The maximum radius on the plot is the radius of the biggest circle that can be fitted in the frame centered at the Gaussian fit center. For both bright and dim spots, the two lines are quite similar. Figure 21 shows the efficiency of the test ruling with respect to the pure reflection, using the aperture photometry of the reduced image (black line) and the integrated sum of the Gaussian fit to infinity (magenta line). The two plots are almost exactly the same.
Fig. 20. Growth curves of the reduced image (black lines) and 2D Gaussian fit (magenta lines) for the test ruling. The solid lines are for the brightest spot, and the dashed lines are for a dim spot whose optical path is the longest and whose size is therefore the biggest on the detector.
Fig. 21. Efficiency of the test ruling with respect to the pure reflection. For the two cases, a circular aperture applied to the reduced image (black line) and a Gaussian fit integral (magenta line), the efficiencies are almost on top of each other.
The growth curve and efficiency comparisons illustrate that our experiment and optical setup are optimized with respect to the laser spot size at the detector and the detector plate scale. Since the two cases give similar answers, we deploy the simple aperture photometry method to the reduced image to calculate the total counts in a spot.
Appendix D:: Pipeline Modification
One unique aspect of OSIRIS is that over 3000 spectra are all partially overlapped on the detector at staggered wavelength, and hence special reduction and calibration steps are required. The OSIRIS data reduction pipeline (DRP) reduces the science data to the level where a user can begin their custom scientific analysis. After OSIRIS was moved to Keck I and G3 was installed, the DRP had to be modified to account for the new AO system and the new grating.
On Keck I, the AO system optical path to OSIRIS has one less mirror than the Keck II system. This produces a flipping of the image in the y-direction. This axis flip is fixed in the DPR "Assemble Data Cubes" module. The Keck I AO system uses a different IR/visible splitting dichroic from that used on Keck II, and white light measurements using the AO fiber calibration source showed that the Keck I AO system dichroic produces essentially no instrumental dispersion. The "Correct Dispersion" module in the DRP that corrects for atmospheric dispersion and instrumental dispersion was appropriately modified for the Keck I AO path.
We determined a new wavelength solution using an arc lamp and OH lines in each broadband filter, and we found about four spectral channels of shift from the previous (G2) version. The field-dependent wavelength solution per spaxel was calculated using the cross-correlation of OH lines using the Kn3 filter with the 35 mas spaxel scale. Figure 22 shows relative/absolute pixel and wavelength offsets in angstroms between 2006 with G2 (top graphs) and 2013 with G3 (bottom graphs). It is a residual offset after using the "Assemble Data Cubes" and a new wavelength solution. The comparison of two confirms that the quality of the ruling on the new grating is more uniform across the surface. The field-dependent and global wavelength solution is now implemented in the new pipeline.
Fig. 22. Uncorrected relative pixel shift (left images), uncorrected absolute wavelength offsets in angstroms (middle images), and corrected absolute wavelength shift in angstroms (right images) for Kn3 35 mas in 2006 with G2 (top images) and in 2013 with G3 (bottom images).
A problem had been observed prior to commissioning of G3 that affected the lower right quadrant of the reduced data cube with varying shifts in measured intensity and shifts in the detector channel offsets. During the modification of the DRP, we found that there appeared to be a bad column on the detector, and this was biasing the wavelength solution. This problem started on 2011 September 17 when one of the spectrograph Hawaii-II detector Leach-ARC detector readout video boards was swapped. In the end, it was just a one pixel shift that skewed the timing of the entire channel. This channel offset is now fixed in the new pipeline within the "Subtract Frame" module.
All these modifications are implemented in the new version of the DRP, and the DRP full package is now available to download at the OSIRIS instrument webpage.8
Footnotes
- 6
Spectrum of each spatial element.
- 7
this is a weighted average of 8 values because of the reason in footnote in Table 3.
- 8