the science and technology of microcalorimeter arrays
TRANSCRIPT
ARTICLE IN PRESS
Nuclear Instruments and Methods in Physics Research A 520 (2004) 402–406
*Corresp
Tel.: +1-30
E-mail a
(C.A. Kilbo
0168-9002/$
doi:10.1016
The science and technology of microcalorimeter arrays
Caroline A. Kilbourne*
NASA Goddard Space Flight Center, Mail code 662, Greenbelt, MD 20771, USA
Abstract
I will review the present status and future prospects for arrays of microcalorimeters used for high-resolution photon
spectroscopy. The various applications impose different constraints on the array designs, but there are many common
physics and engineering issues as well. I will present both the unique and the shared aspects of their design, fabrication,
and operation. As we are well into the age of small microcalorimeter arrays, I will present the performance of a variety
of functioning arrays. Arrays with more than 1000 pixels are under development; I will report on their progress and on
the technical issues associated with arrays of this scale and larger.
r 2003 Elsevier B.V. All rights reserved.
PACS: 07.20.Fw
Keywords: Microcalorimeter arrays; Thermal detectors
1. Introduction
The microcalorimeter is proving to be a versatileand flexible device, able to take a myriad of formsacross a wide range of applications. Even limitingthe discussion to applications requiring precisemeasurement of the energy of individual photons(and thus excluding bolometers and particlecalorimeters), we are left with a diverse field.When we speak of arrays of microcalorimeters,however, there are certain physical and technicalissues common to all implementations.
A microcalorimeter is primarily a thermometerthat is partially decoupled from a heat sink.Energy deposited in the thermometer is sensed as
onding author. Formerly known as C.K. Stahle.
1-286-2469; fax: +1-301-286-1684.
ddress: [email protected]
urne).
- see front matter r 2003 Elsevier B.V. All rights reserve
/j.nima.2003.11.345
a temperature change against a background ofthermal fluctuations. Most microcalorimeters alsoneed a radiation absorber that is well coupled tothe thermometer, either because the thermometeritself does not have sufficient absorption efficiencyor because it does not reproducibly thermalize theenergy of the incident photons. An array ofmicrocalorimeters is not simply a collection ofsuch devices, but is a thermal and electrical systemthat must be designed as a unit for the bestperformance of each pixel and of all the pixels inaggregate.
I will limit my discussion to arrays of indepen-dent pixels that do not provide position informa-tion on scales finer than the pixel pitch, sinceimaging microcalorimeters are discussed elsewherein these proceedings [1]. The motivation for sucharrays of discrete microcalorimeters is to obtainthe best possible energy resolution within theconstraints of the application.
d.
ARTICLE IN PRESS
C.A. Kilbourne / Nuclear Instruments and Methods in Physics Research A 520 (2004) 402–406 403
The following issues need to be addressed ingoing from an isolated microcalorimeter to anarray of any scale:
* Fabrication and assembly (scale of integration),* uniformity,* thermal and electrical contacts,* thermal and electrical crosstalk.
I will return to these issues repeatedly as I discussboth existing arrays and those in development.
2. Design and performance of functioning arrays
2.1. Goddard XRS arrays
X-ray astronomy motivated the first microca-lorimeter arrays, those developed by the Goddard/Wisconsin collaboration for the soft diffuse X-raybackground sounding rocket experiment X-rayQuantum Calorimeter (XQC) [2] and the X-RaySpectrometer (XRS) [3,4] instruments of theAstro-E and Astro-E2 Japan/US X-ray observa-tories. These 36-pixel arrays were fully integratedin microfabrication except for absorber attach-ment. They comprise ion-implanted Si thermistors,Si thermal links, and HgTe absorbers. The state ofthe art is the 6� 6 array that will be a part ofAstro-E2 when it is launched in 2005. Each pixelcontains a (0.624�mm)2� 8 mm HgTe absorberattached to a suspended thermistor formed in a1.5-mm device layer. Each thermistor is suspendedover a separate well etched through the substrateby deep reactive ion etching (DRIE). Four sucharrays have been completed and tested, resulting ina resolution at 6 keV of 5.3–6.5 eV, with one to twooutliers per array.
The XRS arrays are the most thoroughlystudied of any calorimeter array to date. Theyshow no on-chip electrical crosstalk (though thereis capacitive crosstalk at the preamplifier feed-through) and no triggerable thermal crosstalk.Thermal crosstalk of 0.06% in adjacent pixels and0.015% between distant pixels has been measuredby averaging. Thermal crosstalk and X-rayshitting the silicon frame contribute to a count-rate-dependent noise. Any energy deposited in theframe (e.g., from a photon or a minimum ionizing
particle) results in pulses on multiple pixels of scaleB3000 less than if the energy had been absorbeddirectly in a pixel [5]. This scaling derives from theratio of the conductance of the pixel thermal linkto the conductance of the connection between thesilicon frame and the heat sink.
The XRS array is mounted to an alumina boardusing epoxy, providing a 90 nW/K link to the60 mK heat sink. For the highest resolution at thehighest count rates, better thermal anchoring ofthe die frame will be necessary. The alumina boardis connected to a gold-plated copper detectorbox via Au wire bonds providing a 10 mW/Kconnection.
2.2. Stanford optical arrays
The Stanford microcalorimeter group has devel-oped superconducting transition-edge sensor(TES) arrays for optical spectro-photometry ofcompact celestial objects. A 6� 6 optical arrayconsists of (20 mm)2 tungsten TES pixels with Alleads. The photons are absorbed directly into theW metal without an intermediate absorber. Theelectron–phonon coupling in the W serves as thethermal link to the heat sink, thus the metal filmslie on a solid silicon substrate. The loss of hotphonons to the substrate during thermalizationresults in a loss of 58% of the energy of eachphoton. So far only 4 pixels are read simulta-neously. The resolution is 0.15 at 1 eV [6].
Parasitic signals occur when photons areabsorbed directly in the substrate or the deviceleads. A reflective mask has been designed to blockthe leads and exposed substrate to scatter thephotons onto a TES. With this scheme, there is noneed to press for a high filling factor of the TESsensors themselves.
2.3. TES arrays for X-ray spectroscopy
The X-ray calorimeter groups at the NationalInstitute of Standards and Technology (NIST), theSpace Research Organization of the Netherlands(SRON), and the NASA Goddard Space FlightCenter (GSFC) are developing large TES arrays,and each has operated small prototype arrays.Much of this development is motivated by the
ARTICLE IN PRESS
C.A. Kilbourne / Nuclear Instruments and Methods in Physics Research A 520 (2004) 402–406404
future X-ray spectroscopy satellites Constellation-X (NASA) and XEUS (ESA). The NIST work isalso driven by applications in materials analysis.
NIST has tested 8� 8 arrays of (0.4 mm)2 Mo/Cu TES pixels on low-stress, off-stoichiometrysilicon–nitride (Si–N) membrane thermal linksdefined by front side patterning and DRIE wellsetched behind each TES. Electrical contacts run to20 of the pixels. They have measured 6.4 eVresolution on heat pulses in a test array withoutabsorbers [7].
SRON has tested 5� 5 arrays of (0.25mm)2 Ti/Au TES pixels with small Cu absorbers onpatterned Si–N membranes. They use [1 1 0] Siwafers so that a KOH etch can be used to place anentire row in a common substrate trench withvertical side walls. Only three pixels are wired forindependent bias and read-out, but the remainingchannels are connected in parallel so that biaspower can be applied to the whole chip. The bestresolution obtained in such an array is 6 eV at6 keV; the three pixels were very similar [8].
SRON has also made test structures to measurethe thermal properties of the 40 mm wide 0.4 mmdeep Si bars between the KOH trenches in thisarray geometry, since these are the only conduitsfor conducting heat from the pixels to their 20 mKheat sink. Their data indicate a phonon mean freepath of 0.15 mm, suggesting a crossover fromdiffuse to specular transport [8].
GSFC has tested 5� 5 arrays of (0.25 mm)2 Mo/Au TES pixels with Bi/Cu absorbers on Si–Nmembrane thermal links defined by DRIE wellsbehind each TES. A 3� 3 subset is wired for read-out and four pixels have been operated simulta-neously. The absorbers are mushroom-shaped,contacting the TES bi-layer in the middle butextending cantilevered over the membrane andwiring between pixels. There is a 6-mm gapbetween absorbers of adjacent pixels. The absor-bers consist of 4 layers each of Bi and Cu; the totalCu thickness is 0.8 mm and Bi thickness is 8.5 mm.The best resolution obtained thus far in such anarray is 5.5 eV at 1.5 keV and 7.0 eV at 6 keV. Thedifferences between pixels need to be understood.
GSFC has also looked for thermal crosstalk intheir TES arrays, distinguishing thermal fromelectrical crosstalk by reversing the polarity of
the bias on one channel. Nearest-neighbor thermalcrosstalk was about 0.1%; between diagonalneighbors it was about 10 times lower. Apreliminary determination of the conductancefrom the array frame to the heat sink wasB60 nW/K, of the same magnitude as the XRSarray heat sinking. Au heat sinking pads andwirebonds were included but were not sufficient.More measurements are needed.
3. Development of arrays for the future
The designs of both XEUS and Constellation-Xrequire 32� 32 arrays of close-packed (0.25 mm)2
microcalorimeter pixels providing 2–4 eV resolu-tion over the astronomical X-ray band. Oneconsideration is whether to micro-fabricate fullyintegrated arrays or to build arrays from modules.The modular approach allows pre-screening ofsegments for potentially higher yield, and thenatural assembly layers provide ample room forelectrical contacts. Potentially, the array canremain modular after integration, though, aslearned for the 4� 32 SHARC II bolometer arrays[9], this is difficult in practice, especially with tighttolerances on alignment. Heat sinking the multiplecomponents and making electrical contacts at theend of the array may be more complex. TheSmithsonian Astrophysical Observatory (SAO) isdeveloping neutron transmutation doped (NTD)germanium microcalorimeter arrays for Constella-tion-X based on assembling rows of pixels onvertical substrates [10]. The NTD technology lendsitself towards such a modular approach since anintegrated process for microfabrication of NTDarrays has not yet been developed. For the TESarray development, however, the development isbiased towards realizing fully monolithic arrayswith high-density contacts.
The small NIST, SRON, and GSFC arraysdescribed in the previous section can be scaled to32� 32 arrays. The essential features are mem-brane thermal links defined by etching through thesubstrate and mushroom-shaped absorbers. Withfine-line lithography, it should be possible to bringout all the leads on the substrate surface availablebetween pixels in a 32� 32 array. GSFC is also
ARTICLE IN PRESS
C.A. Kilbourne / Nuclear Instruments and Methods in Physics Research A 520 (2004) 402–406 405
developing ultra-low-resistance through-wafer mi-cro-vias [11] for bringing contacts to the backsideof a wafer, where they can be bump-bonded to afan-out board that connects to the first stage of theread-out.
NIST and SRON are developing differentmicromachining processes for bringing contactsout under the pixels in a TES array. The NISTapproach raises the pixel membranes on tablestructures above the wafer surface. The SRONapproach uses vias to a buried wiring layer, leavingthe pixel membranes in the plane of the wafersurface. In addition to providing space for wiring,the solid substrate will be stronger and morereadily thermally anchored than an etched sub-strate. Fig. 1 illustrates the bulk and surfacetechniques.
Electrical crosstalk is a concern for high-densityleads. SRON modeling had led to the conclusionthat magnetic screening is crucial for dense surfacewiring, either through use of a separate micro-stripline per channel or a common ground planebeneath the lines [8].
Fig. 1. Top: Bulk micromachining concept common to many
arrays, showing GSFC micro-via scheme. Middle: SRON
surface micromachining. Bottom: NIST surface micromachin-
ing.
4. Conclusion
We are well into the era of small (30-pixel)microcalorimeter arrays, and techniques andprocesses needed for 1000-pixel arrays are beingdeveloped and proven in smaller arrays. Electricaland thermal models of the entire array system(frame, mounting substrate, electrical contacts) arerequired for optimal performance of the array as awhole. Measurements and simulations of electricalcrosstalk and heat sinking being done now arecritical to the design of future arrays.
For TES arrays, minimizing thermal crosstalk isaided by extreme electrothermal feedback, whichoffsets most of the input energy of a photon byreducing the Joule power into the TES. Thethermal design of a 1000-pixel TES array will bedriven instead by the integrated bias power. Attypically 10 pW/pixel, array heat sinks will needmuch higher thermal conductance than presentlyobtained to keep the whole die from heating and tokeep each pixel connected to the same referencetemperature.
There is no fixed scale cut-off for the technol-ogies discussed, but for the present, the design of32� 32 arrays is sufficiently challenging.
Acknowledgements
The author is grateful for information providedby Henk Hoevers, Kent Irwin, and JenniferBurney, and for discussions with colleagues atGSFC.
References
[1] E. Figueroa-Feliciano, Position sensitive cryogenic detec-
tors, Nucl. Instr. and Meth. A (2004), these Proceedings.
[2] D. McCammon, et al., Astrophys. J. 576 (2002) 188.
[3] C.K. Stahle, et al., Proc. SPIE 3765 (1999) 128.
[4] C.K. Stahle, et al., The next-generation microcalorimeter
array of XRS on Astro-E2, Nucl. Instr. and Meth. A
(2004), these Proceedings.
[5] C.K. Stahle, et al., Cosmic ray effects in microcalorimeter
arrays, Nucl. Instr. and Meth. A (2004), these Proceedings.
[6] B. Cabrera, et al., AIP Conf. Proc. 605 (2002) 565.
ARTICLE IN PRESS
C.A. Kilbourne / Nuclear Instruments and Methods in Physics Research A 520 (2004) 402–406406
[7] K.D. Irwin, presented at Constellation-X Facility Science
Team meeting, Columbia University, New York, USA,
2003.
[8] M.B. Bruijn, presented at Fourth Round Table on
Micro/nano Technologies for Space, Noordvijk, The
Netherlands, 2003.
[9] S.H. Moseley, private communication
[10] J. Beeman, et al., AIP Conf. Proc. 605 (2002) 211.
[11] F.M. Finkbeiner, et al., Development of ultra-low
impedance through-wafer micro-vias, Nucl. Instr. and
Meth. A (2004), these Proceedings.