EUROPEAN SOUTHERN OBSERVATORY Organisation Européenne pour des Recherches Astronomiques dans l'hemisphère Austral Europäische Organisation für astronomische Forschung in der südlichen Hemisphäre VLT PROGRAMME VERY LARGE TELESCOPE Test Report on HgCdTe Issue: Version 1.0 Date: 28 November 2002 Prepared: G. Finger... Name Date Signature Approved: A. Moorwood... Name Date Signature Released: G.Monnet... Name Date Signature
CHANGE RECORD Issue/Revision Date Part Affected Reason/Remarks Draft 5 November 2002 new document Revision 1 28 November 2002 final document
page 1 of 16 Table of Contents 1. Scope of the work 2 2. Introduction 2 3. Hawaii2 2Kx2K LPE arrays l c =2.5 mm 2 3.1. Measurement Setup 2 3.2. Quantum Efficiency 3 3.2.1. Engineering grade array: 4 3.2.2. Science grade array: 5 3.3. Dark current 5 3.4. Readout Noise and Glow 7 3.5. Problems with Hawaii2 7 4. Picnic 256x256 MBE array l c =1.7 mm 9 4.1. Calibration of transfer gain 10 4.2. Quantum Efficiency 10 4.3. Darkcurrent 11 4.4. Readout noise 11 5. Picnic 256x256 MBE array l c =2.5 mm 12 5.1. Calibration of transfer gain 12 5.2. Quantum Efficiency 13 5.3. Readout Noise 14 6. Continuous Flow Cryostat 14 6.1. Cryogenic Performance 15 6.2. Photon background 15 7. Conclusions 16
page 2 of 16 1. Scope of the work Test results obtained with HgCdTe arrays which have cut-off wavelengths = 2.6 µm are presented and the status of the detector development of short cut-off HgCdTe arrays is reviewed. 2. Introduction Four HgCdTe arrays have been tested for their potential use to equip the focal planes of the VLT instruments SPIFFI, NIRMOS and FINITO. Two arrays were grown by LPE (liquid phase epitaxy) on a sapphire substrate. They have a cut-off wavelength of 2.5 µm and are hybridized to the Hawaii2 multiplexer which has a format of 2Kx2K. Both an engineering grade and a science grade Hawaii2 array have been tested. These arrays were the result of a contract jointly funded by the University of Hawaii, Subaru, and ESO in 1997 to develop a large format 2048x2048 HgCdTe array. So far these arrays were the first and only 2Kx2K arrays available for evaluating their performance. Because the four MBE (molecular beam epitaxy) grown 2Kx2K arrays on order for NIRMOS, which have a cut-off wavelengths of λ c =1.9µm, have not yet been delivered, the evaluation of its continuous flow test cryostat was carried out using two 256x256 Picnic arrays grown by molecular beam epitaxy. The arrays are double layer planar heterostructures grown on a CdZnTe substrate. These arrays have been borrowed from Rockwell to ensure an early start of the continuous flow cryostat evaluation with MBE arrays. The first MBE Picnic array had a cut-off wavelength of 1.7 µm. This array showed excellent dark current performance but had excess noise. It was replaced by a MBE Picnic array having a cut-off wavelength of 2.5 µm. This array was tested in the cryostat of the VLTI fringe sensor FINITO. Test results of these arrays and some problems encountered with the test camera are summarized in this document. 3. Hawaii2 2Kx2K LPE arrays l c =2.5 m m 3.1. Measurement Setup A prototype continuous flow test cryostat was built to start the evaluation of Hawaii2 arrays and the cryogenic camera concept of NIRMOS. The NIRMOS cameras are based on a continuous flow cryostat. The minimum temperature of 95 K, which can be achieved with the continuous flow system under normal operation, is adequate for having a cutoff wavelength of λ c =1.9 µm but too high for the first two Hawaii2 arrays having a cut-off wavelength of λ c =2.6 µm. For this reason the test camera was upgraded with a pulse tube two stage closed cycle cooler. The moving displacer of a standard Gifford Mac Mahon closed cycle cooler is replaced by an oscillating pressure wave in a pulse tube. The advantage of the pulse tube is the fact
HgCdTe page 3 of 16 that it has no moving cryogenic parts and is completely free of vibrations. In our setup the minimum detector temperature achieved with the pulse tube is 41 K. Cryogenic clamp for reference Cooling braid Temperature sensor and heater Nitrogen transfer line Figure 1 Left: Schematics of unit cell and symmetrical operational amplifiers located on the fan-out board at cryogenic temperature. Right: Detector fan-out board with 36 cryogenic amplifiers, clamp circuit for reference output, filters and antistatic protection for bias and clock voltages. The arrays are read out by accessing the internal bus directly. All 32 analog outputs are used and fed into a symmetrical operational amplifier located near the focal plane on the fan-out board which is at cryogenic temperatures. The set-up is shown in Figure 1. The readout multiplexer has an on-chip output source follower for each of the 32 video channels to which the internal video signal bus is connected. The user has also direct access to the internal signal bus. The bus can be connected to +5 volt by an external 200KΩ load resistor. The internal bus is directly connected to the input of an external operational amplifier, located as close as possible to the detector signal pins for maximum immunity to noise pickup. A linear CMOS operational amplifier was selected which contributes ~ 3 electrons rms to the readout noise of a double correlated clamp. The design of differential data line drivers using these cryogenic CMOS amplifiers is shown in the schematics above and has been described in an SPIE paper. In this configuration the on-chip output source followers are not used. Hence, their glow is eliminated when multiple sampling is applied to reduce the readout noise. This is one of the biggest advantages of off-chip amplifiers. 3.2. Quantum Efficiency The quantum efficiency was determined by varying the temperature of a blackbody and observing the photon response of the detector for different levels of photon flux. The filter transmission curve used to derive the photon flux was measured at liquid
page 4 of 16 nitrogen temperature. The pixel transfer function needed to convert the detector signal to electrons was obtained by the standard shot noise method. 3.2.1.Engineering grade array: The cut-off wavelength of a first engineering grade array was measured to be λ c =2.58 µm. Since the dark current of the engineering grade array shown in Figure 2 was higher than expected and showed a temperature dependence which yields an effective temperature being one third of Teff = Eg/2Kb, the sensitivity of the array was investigated at wavelengths longer than λ c. A narrow band filter centered at λ=3.22 µm was mounted in the test camera and the quantum efficiency at λ=3.22 µm was determined to be 1.1 10-5. In order to exclude the existence of filter leaks and to confirm the spectral response at λ=3.22 µm, the transmission of the narrowband filter was measured with the engineering array using a scanning grating monochromator. The result is displayed in Figure 3. As the photon response was confirmed by the measurement, an extrinsic transition from impurity donors to the conduction band in the multiplexer, the buffer layer or the diode array must be assumed to generate the excess dark current. Figure 2 Dark current versus inverse temperature for engineering grade Hawaii2 LPE array. Teff = 897 K.
page 5 of 16 Figure 3 Transmission of narrow band filter measured with λ c =2.58 µm engineering grade array. 3.2.2.Science grade array: The quantum efficiency of the science grade array is above 0.68 over the whole spectral range of the detector and peaks in K-band at 0.84. The wavelength dependence of the quantum efficiency was measured with a grating monochromator. The result is represented in Figure 4. The efficiency of the monochromator was calibrated with a pyroelectric detector which is assumed to be spectrally flat. The accuracy of the measurement shown below is limited by the sensitivity of the pyroelectric detector. The cutoff wavelength is λ c =2.6 µm. The Hawaii2 science grade array has the highest QE of all LPE arrays ever evaluated at ESO. 3.3. Dark current The detector was blinded by a black aluminium cover. Since the detector is cooled by the central pins of a pin grid array (PGA) package, great care was taken to calibrate the detector temperature and take into account the temperature gradient between the cold finger and the detector. In a separate experiment a temperature sensor was mounted on an empty PGA carrier and cooled to measure this temperature gradient. For the science grade array the temperature dependence of the dark current yields the correct effective temperature Teff = 2360 K. The dark current is 0.004 e/s at 60K and 0.20 e/s at 79.1K as can be seen in Figure 6. The cosmetic quality degrades substantially by raising the temperature form 60K to 79 K as can be seen by the dark exposures shown in Figure 5.
page 6 of 16 Figure 4 Quantum efficiency versus wavelength for Hawaii2 LPE science grade array. The spectral bandpass of the broadband filters is indicated at the bottom with relative transmissions. Figure 5 Dark exposure of Hawaii2 LPE science grade array at T = 79 K ( left image) and T = 60 K ( right image ). Scale min = -0.025, max = 0.25 e/s/pixel. Detector integration time 4 hours. Mean dark current at T=79 K : 0.2 e/s/pixel. Mean dark current at T = 60 K : 0.004 e/s/pixel.
page 7 of 16 Figure 6 Dark current versus temperature for HgCdTe LPE and MBE arrays and different cut-toff wavelengths. Green Circles: Hawaii1 1Kx1K LPE, λ c =2.5µm, dark current 6.8310-4 e/s at T=43 K. Red Squares: Hawaii2 2Kx2K LPE, λ c =2.6µm, dark current 4.9310-3 e/s at T=59.7 K..V iolet Triangles: Picnic 256x256 MBE, λ c = 2.5µm, dark current 6.4510-3 e/s at T=80 K. V iolet Triangles: Picnic 256x256 MBE, λ c = 1.7µm, dark current 7.910-3 e/s at T=101 K. 3.4. Readout Noise and Glow The readout noise has been measured with the unbuffered output reading out all 32 channels. The time required to read out a full frame is 600 ms. The readout noise for double correlated sampling is 12 electrons rms. With multiple nondestructive sampling using 16 Fowler pairs the readout noise can be reduced to 5.2 electrons rms as shown by the noise histogram of Figure 7. In the best quadrant the readout noise is as low as 4.4 erms. The glow of the Hawaii2 multiplexer is substantially reduced in comparison to the Hawaii1 multiplexer. At the last row of each channel close to the edges of the array the glow is 1.5 electrons/frame and peaks in the corners to 3 electrons/frame. The glow is negligible at a distance of 30 pixels from the edges as can be seen in Figure 8. 3.5. Problems with Hawaii2 Although in most respects the Hawaii2 science grade array exhibits excellent performance, there are some problems associated with the Hawaii2 multiplexer. Interleaved clocking of the fast shift register is not possible. The 4 clocks of the fast shift register have to be pairwise complementary (CLK1 with CLK2,
page 8 of 16 CLKB1 with CLKB2) and skewed to less than 50 ns. If several arrays have to be read in parallel using one single clock driver, these clocks should be generated on the fan-out board using CMOS buffers ( HEF 4041) Use of the reference outputs requires four additional video channels or a cryogenic clamp circuit for the symmetrical cryo-opamps. The reference output is not available for the unbuffered output mode. 2Kx2K LPE, λ c =2.6µm, dark current 4.93 10-3 e/s at T=59.7 K. Violet Triangles: Picnic 256x256 MBE array, λ c =2.5µm, dark current 6.45 10-3 e/s at T=80 K. Violet Triangles: Picnic 256x256 MBE array, λ c =1.7µm, dark current 7.9 10-3 e/s at T=101 K Figure 7 Histogram of readout noise: Black: double correlated readout, noise 12 electrons rms, integration time 600 ms. Red: multiple sampling, readout noise 5.2 electrons rms with 16 Fowler pairs, integration time 11.54 sec.
page 9 of 16 Multiplexer glow Figure 8 Electroluminescence of Hawaii2 multiplexer with 32 channel unbuffered output mode. Intensity at the edges 1.5 electrons / full frame. 4. Picnic 256x256 MBE array l c =1.7 m m The arrays originally ordered to be installed in NIRMOS are double layer planar heterostructure MBE arrays grown on CdZnTe substrate. The cut-off wavelength of these arrays was set to λ c =1.9 µm. Since these arrays have not been delivered in time to test the cryogenic concept of NIRMOS, a smaller format engineering grade array grown by this technology was hybridized to a PICNIC 256x256 multiplexer and has been lent to ESO in order to validate continuous flow cryostats for infrared applications. The delivered engineering grade Picnic MBE array has 40 µm pixels and a cut-off wavelength of λ c =1.75 µm. The detector mount is shown in Figure 9. Figure 9 Detector mount for the 256x256 PINIC array. On the right image the cryogenic operational amplifiers for four channels can be seen.
page 10 of 16 4.1. Calibration of transfer gain The conversion factor or transfer gain, which gives the number of electrons corresponding to one ADU, was measured by the standard shot noise method. In Figure 10 the variance is plotted versus the detector signal. From the inverse slope of a straight line fitted to measured data points a transfer gain of 14.6 e/adu is derived. One Volt corresponds to 45540 ADU s at the video output of the detector. Hence, the transfer gain is 665 e/mv and corresponds to a capacity of the integrating node of 106 Ff, which is about 2.6 larger than the capacity of 41 ff obtained with the Hawaii 1Kx1K LPE array which has smaller 18 µm pixels. Figure 10 Calibration of transfer gain of Picnic λ c =1.7 µm MBE array. Tranfer gain is 14.61 e/adu corresponding to a capacity of the integrating node of 106 ff. 4.2. Quantum Efficiency Using this transfer gain the quantum efficiency was obtained as described above in section 3.2.2. The quantum efficiency of the engineering grade array peaks at 0.41 at a wavelength of λ=1.5 µm and drops to 0.21 at λ=1.2 µm. The cutoff wavelength is λ c =1.75 µm as can be seen in Figure 11.
page 11 of 16 Figure 11 Quantum efficiency of λ c =1.75µm Picnic 256x256 MBE array. 4.3. Dark Current The Picnic engineering grade array shows excellent dark current performance as shown by the violet triangles in figure 10. At a temperature of 100 K the dark current scaled to a pixel size of 18 µm is 0.008 e/s. The persistence is strongly reduced in comparison to LPE grown HgCdTe material demonstrating the potential of MBE on CdZnTe. Further tests with a science grade Picnic MBE array having a cut-off wavelength of λ c =1.9 µm will be carried out in FINITO, the fringe sensor of the VLT interferometer. The FINITO array will operate as part of an active control loop to stabilize the fringes of combined telescope beams. 4.4. Readout noise Applying the measured transfer gain determined in section 4.1 the readout noise derived for a simple double correlated clamp was 58 electrons rms. The Histogram of the readout noise for the λ c =1.7µm MBE Picnic array is represented by the blue curve in Figure 12. This value is much higher than expected. To test the noise contribution of our set-up, the reset switch of the array was permanently closed. In this configuration the gate of the unit cell source follower is permanently connected to the reset voltage bypassing the noise effects of the detector diode. The obtained noise histogram peaks at 20 electrons rms as shown by the red histogram in Figure 12 and indicates that the short cut-off HgCdTe diodes generate the excess noise. The existence of excess noise in λ c =1.7 µm MBE detector material was independently confirmed at the Space Telescope Science Institute and the University of Hawaii. With the Hawaii-1RG array having a pixel size of 18 µm the readout noise measured at STScI was 40 erms.
page 12 of 16 After several tests and discussions with the manufacturer the λ c =1.7 µm MBE Picnic array was replaced by a λ c =2.5 µm MBE array. 5. Picnic 256x256 MBE array l c =2.5 m m Because of the excess noise the λ c =1.7 µm PICNIC MBE engineering grade array was replaced by a MBE PICNIC array with a cut-off wavelength of λ c =2.5 µm. This array was also delivered on a loan basis. The array was tested in the FINITO cryostat which is a conventional bath cryostat. The detector temperature was 78.36 K. 5.1. Calibration of transfer gain As above, the transfer gain was obtained by the standard shot noise method as shown in Figure 13. The transfer gain obtained is 16.5 e/adu or 751 e/mv. This corresponds to a capacity of the integrating node of 120 ff, which is about three times larger than the capacity of 41 ff obtained with the Hawaii 1Kx1K LPE array. The transfer gain of this Picnic device was measured at Rockwell prior to shipment to ESO. The Rockwell value is 335 e/mv and corresponds to a capacity of 53 ff which is only slightly larger than the capacity of the Hawaii 1Kx1K array. The pixel size of the Hawaii 1Kx1K Figure 12 Comparison of Readout Noise of Hawaii2 2kx2K LPE array and Picnic 256x256 MBE array with single double correlated readout. Black: Hawaii2 2Kx2K LPE, λ c =2.6µm, noise = 11 erms. Red: Picnic 256x256 array with reset switch permanently closed, noise = 20 erms. Green: Picnic 256x256 MBE λ c =2.5µm, noise=35 erms. Blue: Picnic 256x256 MBE λ c =1.7µm, noise = 58 erms.
page 13 of 16 LPE array is 18.5 µm, the pixel size of the Picnic array is 40 µm. The area of the Picnic pixel is by a factor of A Picnic /A Hawaii1 = 4.7 larger than the area of the Hawaii2 pixel. The measurement of Rockwell implies, that the pixel capacity only slightly increases with increasing pixel area (C Picnic /C Hawaii1 = 1.29), whereas the ESO measurement indicates that the capacity scales with pixel area by a larger factor (C Picnic /C Hawaii1 = 2.93). The integrating node capacity is the sum of a fixed capacity, which is the gate capacity of the unit cell source follower and a variable capacity, which is the junction capacity of the depletion region of the infrared diode. The junction capacity depends on the doping level of the detector material and on the detector bias voltage which changes during the detector integration, while the diode capacity is discharged by absorbed photons. 5.2. Quantum Efficiency Using this transfer gain measured at ESO the quantum efficiency in the H-band was determined in the FINITO bath cryostat. The photon flux was varied by changing the blackbody temperature. In Figure 14 the detector signal is plotted as a function of integrated photons. The value derived for the quantum efficiency in H is 46 %. Figure 13 Transfer gain of Picnic λ c =2.5 µm MBE array 16.5 e/adu.
page 14 of 16 Figure 14 Quantum Efficiency of Picnic λ c =2.5 µm MBE array in H Band 46 %. 5.3. Readout Noise The readout noise for a simple double correlated readout is 35 electrons rms as represented by the green noise histogram in Figure 12. One Volt corresponds to 45540 ADU s at the video output of the detector. The measured readout noise of 35 erms corresponds to 48 µv rms at the video output, which is consistent with 50 µv rms measured at Rockwell Scientific. Applying the transfer gain obtained at Rockwell Scientific the noise voltage measured at ESO would correspond to a readout noise of 15.5 electrons rms, but to an extremely poor peak quantum efficiency of 20 %. Further tests will be carried out both at Rockwell Scientific and at ESO to clarify this discrepancy. 6. Continuous Flow Cryostat The detector test camera is based on a continuous flow cryostat. Cold nitrogen gas is flushed through transfer lines inside the cryostat; the gas cools the instrument with heat exchangers mounted at specific points in the cryostat. In Figure 15 the heat exchanger for the detector is shown. The detector is heat-sinked to the heat exchanger by flexible silver foils.
page 15 of 16 Figure 15. Heat exchanger of the continuous flow cryostat cooling the detector. 6.1. Cryogenic Performance The lowest detector temperature achieved with a continuous flow cryostat is 84 K. Operating at this temperature is not practical, since the valve at the end of the nitrogen transfer line is permanently open. Since the nitrogen gas at the outlet is still cold, the nitrogen consumption is extremely high. At temperatures above ~95 K the outlet valve which is controlled by a temperature sensor in the instrument is switching and regulates the instrument temperature. This reduces the nitrogen consumption to less than 30 litres per day. 6.2. Photon background The test camera has a warm entrance window, a cold filter wheel and detector. The filter wheel has 12 filter positions. The photon background was measured in J and H and with a dark filter position. All the dark current measurements of the have been made with the blinded by a special cover with the filter wheel put in the dark position. The photon background of the test camera was measured with the PICNIC MBE λ c =1.7 µm array. All fluxes have been scaled from the PICNIC pixel size of 40 µm to the Hawaii2 pixel size of 18 µm. In the H band the detector accepts room temperature photons from a solid angle corresponding to a focal ratio of f/0.56. The measured detector signal scaled to a pixel size of 18 µm is 272 e/s. This flux is consistent with the estimated thermal photon flux at a temperature of 18.5 C taking into account the measured filter transmission curve and the measured spectral quantum efficiency. The measured detector signal with the narrow band J filter was 0.4 e/s. The filter is centred at λ=1.195 µm and has a FWHM of λ=0.01 µm. At a wavelength of
page 16 of 16 λ=1.2 µm the photon flux should be negligible for room temperature radiation. With the J band filter the measured detector signal was 0.33 e/s. In an effort to suppress the photon background the filter baffling was improved by reducing the clear aperture of the narrow band J filter from 43.0 mm to 37.8 mm. This resulted in a decrease of the measured detector signal from 0.4 e/s to 0.093 e/s. 7. Conclusions In most respects the Hawaii2 λ c =2.5 µm LPE science grade arrays exhibit excellent performance. The quantum efficiency is 0.84 in K, the readout noise is 12 erms and the dark current is 6.8 10-4 e/s at T=43 K and 0.6 e/s at 80 K. The array can be used in a bath cryostat for imaging applications but should be cooled to 60 K for the use in a spectrograph. There are however also some problems associated with the Hawaii2 multiplexer such as rise times of < 50 ns for clocks of the fast shift register and the complexity of using the reference output. Consequently, Rockwell redesigned the Hawaii2 multiplexer resulting in the new Hawaii-2RG, which will be delivered with all MBE arrays presently on order by ESO. Tuning the bandgap of HgCdTe to shorter cut-off wavelengths is a viable method to increase the operating temperature of the detector for low background applications. For a cut-off wavelength of λ c =1.7 µm the dark-current measured with a MBE Picnic array at an operating temperature of 101K was 7.9 10-3 e/s when scaled to the pixel size of 18 µm. This is a convenient temperature for a continuous flow cryostat. A problem with short cut-off arrays is the noise performance. For a simple double correlated clamp a readout noise of 58 erms was achieved. The system noise of the set-up for this measurement was 20 erms. The excess noise is attributed to the short cut-off MBE detector material. There is some discrepancy with respect to the transfer gain measured by ESO and by the manufacturer, who obtains a lower transfer gain. However, if a lower transfer gain is adopted the quantum efficiency falls to very low values. ESO is working on the development of two new methods to determine the transfer gain. The first method uses X-rays and the second method relies on the comparison of a calibrated capacity with the unknown capacity of the integrating node. MBE having a cut-off wavelength of λ c =2.5 µm achieve dark currents of 6.4 10-3 e/s at a temperature of T=80 K and 3.35 10-2 e/s at T=95 K, which is a substantial gain in comparison to LPE material having the same cut-off wavelength. MBE arrays with λ c =2.5 µm can be used in bath cryostats even for spectroscopic applications. The baffling of a window-filter-detector configuration is demanding and requires oversized filters to suppress radiation leaks at the edges of the filters. The lowest detector temperature in a continuous flow cryostat under realistic operating conditions is = 95K.