A Textured Silicon Calorimetric Light Detector
ABSTRACT We apply the standard photovoltaic technique of texturing to reduce the reflectivity of silicon cryogenic calorimetric light detectors. In the case of photons with random incidence angles, absorption is compatible with the increase in surface area. For the geometrically thin detectors studied, energy resolution from athermal phonons, dominated by position dependence, is proportional to the surface-to-volume ratio. With the CaWO4 scintillating crystal used as light source, the time constants of the calorimeter should be adapted to the relatively slow light-emission times.
arXiv:physics/0307042v1 [physics.ins-det] 7 Jul 2003
A Textured Silicon Calorimetric Light Detector
P. C. F. Di Stefano∗, T. Frank, G. Angloher, M. Bruckmayer, C. Cozzini,
D. Hauff, F. Pr¨ obst, S. Rutzinger, W. Seidel, and L. Stodolsky
Max-Planck-Institut f¨ ur Physik, F¨ ohringer Ring 6, D-80805 Munich, Germany
Institut f¨ ur Solarenergieforschung Hameln/Emmerthal,
Am Ohrberg 1, D-31860 Emmerthal, Germany
(Dated: February 2, 2008)
We apply the standard photovoltaic technique of texturing to reduce the reflectivity of silicon
cryogenic calorimetric light detectors. In the case of photons with random incidence angles, absorp-
tion is compatible with the increase in surface area. For the geometrically thin detectors studied,
energy resolution from athermal phonons, dominated by position dependence, is proportional to
the surface-to-volume ratio. With the CaWO4scintillating crystal used as light source, the time
constants of the calorimeter should be adapted to the relatively slow light-emission times.
PACS numbers: 29.40, 84.60.Jt, 85.25.Oj, 95.55.Vj, 95.35.+d
∗E-mail: email@example.com, Permanent address : Institut de Physique Nucl´ eaire de Lyon, 4 rue Enrico
Fermi, F-69622 Villeurbanne Cedex, France
Cryogenic calorimeters, in which the phonons created by incoming particles are read out,
now rival longer-established techniques of particle detection such as ionization in semicon-
ductors and scintillation. They boast excellent thresholds and resolutions which can be
enhanced by measuring athermal phonons in addition to the thermal ones. Another of their
advantages, exploited by rare-event searches for which radioactive background is an issue, is
the ability to distinguish between particles interacting with electrons in matter (e.g. photons
and electrons) and those interacting with nuclei (e.g. neutrons and putative dark matter
particles). Until now this has been achieved mainly through a simultaneous measurement of
charge in semiconducting calorimeters [1, 2, 3]. Another technique is a simultaneous mea-
surement of scintillation, with a principal calorimeter made out of a scintillating material
which emits photons read in a secondary calorimeter [4, 5] (or some other light sensitive
device ). For instance, the next phase of the CRESST (Cryogenic Rare Event Search
with Superconducting Thermometers) dark-matter search will deploy up to 33 such mod-
ules with CaWO4as the main calorimeter . The challenge is that the emitted light is but
a small fraction of the deposited energy, and not all of it necessarily reaches the secondary
calorimeter. We report on optimization of these light detectors.
In the case of main calorimeters like CaWO4emitting light in the visible spectrum, and
for optical sources in general, silicon would appear well suited as an absorber for the light-
detection calorimeter, because of its band gap around 1 µm (1.17 eV at mK temperatures).
Moreover, Si has already been successfully used as an absorber in cryogenic calorimeters
(e.g. Ref. ). Its advantages include a high speed of sound (≈ 5760 m/s) which gives good
phonon properties, and a high melting point (≈ 1690 K) which facilitates deposition of thin
films made from materials with high melting temperatures, such as tungsten, when they are
chosen as thermometers. However, polished silicon has a high visible reflectivity. A similar
problem has been encountered in the field of photovoltaics, and solved by a combination
of texturing the surface of the silicon and coating it with anti-reflective layers . The
texturing squares the reflectivity for normal incident photons by providing them with two
chances to be absorbed. We first describe preparation of our textured light detectors before
discussing experimental results obtained.
PREPARATION OF THE LIGHT DETECTORS
Two 4 inch diameter, 525±35 µm thick, float-zone p-type silicon wafers with a resistivity
of between 10200 and 71030 Ωm were used. Orientation of both wafers was (100) for the
purpose of texturing. One side of each wafer was polished, the other lapped and etched.
A natural silicon oxide layer is assumed to have been present on all Si surfaces. A 150 nm
thick SiNxlayer was deposited by plasma-enhanced chemical vapor deposition through an
Al mask into 5 mm diameter disks on the polished surface of one of the wafers. This wafer
was then etched in a KOH-isopropanol mix at 75◦C in order to texture the exposed Si
into a random pyramid structure . Typical height of the pyramids is 2–5 µm, while
the pyramid angle of 70.5◦given by the crystalline structure of Si means that the textured
surface area is about 1.74 times greater than the original, planar, surface area. The SiNx
remained unaffected by the texturing.
Samples of size 20 × 20 mm2and 30 × 30 mm2were cut from both wafers. Tungsten
transition-edge sensors of the type depicted in Figure 1 were next deposited onto the samples
using a standard procedure developped by the CRESST collaboration  : tungsten films
about 300 nm thick were evaporated at 550◦C under 10−10mbar onto the samples; the W
was structured by photolithography and a KH2PO4- KOH - K3Fe(CN)6- H20 solution to
sizes of 2 × 2 mm2or 2 × 3 mm2. Electrical contact pads made of 200 nm thick aluminum
were then sputtered onto the tungsten for the readout, as was a 200 nm thick gold thermal
contact. Similar aluminum pads were deposited as contacts for a film heater used to stabilize
the operating temperature of the thermometer and to send periodic heat pulses to monitor
the stability of the detector response. Gold or aluminum wires of 25 µm diameter were
ultrasonically bonded to the pads to provide the thermal or electrical links. The W films
were placed near the center of the Si absorbers, and on the SiNxin the case of the textured
samples. On silicon, the tungsten reliably gave superconducting transitions near 20 mK,
once it was realized this transition appears to depend on the natural oxide on which the W
is deposited : when the W was evaporated onto samples which had been etched in HF just
before mounting in the deposition chamber, the transition temperature was of the order of
1 K; when the time lapse between HF bath and W deposition was of the order of a week
the transition was at about 60 mK. This is presumably linked to some chemical interaction
between Si and W which is inhibited by the presence of natural oxide. Such interactions
also appear to have been blocked by the SiNxlayer in the case of the textured absorber, as
a transition temperature near 20 mK was obtained.
Three detectors were selected for testing : 20 × 20 mm2and 30 × 30 mm2planar Si
absorbers, and a 20 × 20 mm2textured Si one. Their characteristics are summarized in
EXPERIMENTAL SETUP AND RESULTS
Preliminary tests and setup
All three light detectors were first cooled in a standard copper holder inside a dilution
fridge and exposed to a collimated55Fe source (5.9 keV photons) to estimate their intrinsic
energy resolution. Resolution, estimated as the full width at half the maximum (FWHM)
of the 5.9 keV line, was 350 eV for the textured detector and 180 eV for the smooth ones.
However, detector responses in terms of pulse height varied with the position of the colli-
mated spot. It is quite likely that these resolutions, especially that of the textured detector,
contain a contribution from the finite size of the collimated hole.
Next, the three detectors were each placed in a setup inside the fridge to measure their
light absorption. The setup (Figure 2), described in detail elsewhere [12, 13], consisted of
a CaWO4scintillating crystal of cylindrical shape (35 mm high with a 40 mm diameter)
placed in a concentric 50 mm diameter light collector lined with a polymer reflective foil .
The CaWO4crystal had a non-functioning 5 × 6 mm2W film on it. Both ends of the light
collector were lined with the same foil; however one of the ends had four Teflon pegs to hold
the light detector. In this manner, both sides of the light detector should have been exposed
to any available scintillating light. Care was taken to minimize thermal leaks between
the calorimeters and their environment while avoiding spurious light traps in the setup.
To provide an absolute energy reference, a55Fe source illuminated the light detector from
outside the light collector foil, through a hole in the light collector’s Cu structure behind
the light detector (for mechanical reasons, the whole large light detector was exposed to the
source, whereas the small detectors where illuminated through a 14 mm diameter hole). An
external60Co source (main photon lines at 1.17 MeV and 1.33 MeV) was used to stimulate
scintillating light from the CaWO4(peak of emission ≈ 440 nm, FWHM ≈ 100 nm ).
The detectors were operated in their superconducting transition by stabilizing their base-
line temperature through the film heaters. This proved a challenge at ground level because
the high rate of cosmic-ray-induced background interacting in the 266 g scintillator led to
pile-up in the light detector, especially in the large one. Pile-up, a nuisance in itself, can
also degrade the temperature stabilization of transition-edge sensors. The large device was
therefore operated with active thermal feedback  to shorten pulse times in some runs.
Stability of the small detectors was monitored with a pulser.
Pulse shapes in light detector
Two classes of light detector events were recognizable from their time constants (Fig. 3).
Fast pulses were caused by direct hits in the light detector (mainly due to the55Fe source).
Slower pulses were events of scintillating light (due to interaction in the CaWO4 of the
cosmic background and60Co when present). That the slow pulses originate in the scintillator
was previously verified by the coincidences between the light detector and an instrumented
The direct events in the detectors have been fitted with a model assuming an exponen-
tial rise (collection and thermalization of athermal phonons in the thermometer), a fast
exponential decay (relaxation of thermal phonons in the thermometer through its heat sink,
referred to as the athermal signal), and a slower exponential decay (relaxation of the entire
detector, referred to as the thermal signal) :
R(t) = H(t)
where H(t < 0) = 0 and H(t ≥ 0) = 1 and, to simplify, the pulses are assumed to start at
t = 0. Direct 5.9 keV hits gave typical rise times of τrise≈ 60 µs with a FWHM of 15 µs for
the distribution of these events. Both the athermal (τath≈ 0.7 ms) and thermal (τth≈ 6 ms)
components were clearly present, with the athermal component making up between 40 and
90 % of the total pulse amplitude. All these parameters depended on the detector and to a
certain extent the operating temperature.
Scintillation-induced hits with similar energy deposits in the light detector had typical
rise times of 250 µs with a FWHM of 50 µs for the distribution.This slow rise time
is interpreted as indicative of a relatively slow scintillation component in CaWO4. The