Eur. Phys. J. D (2025) 79 :89
https://doi.org/10.1140/epjd/s10053-025-01034-6
THE EUROPEAN
PHYSICAL JOURNAL D
Regular Article – Cold Matter and Quantum Gases
Performance of silicon photomultipliers from room
temperature to below 200 mK
Otto Hanski1,a , Tom Kiilerich1,b , Sampsa Ahopelto1,c, Aleksei Semakin1,d , Janne Ahokas1,e ,
Viacheslav Dvornichenko1,f, and Sergey Vasiliev1,g
1 Department of Physics and Astronomy, University of Turku, Vesilinnantie 5, 20014 Turku, Finland
Received 9 January 2025 / Accepted 26 June 2025 / Published online 25 July 2025
© The Author(s) 2025
Abstract.
Abstract We present experimental results of characterization of Silicon photomultipliers (SiPM) in a
temperature range from 90 mK to 40 K and compare them to room-temperature results. Two SiPMs, one
from ONSEMI and one from Hamamatsu Photonics, were tested. Operating voltage ranges, dark count
rates, afterpulsing effects and photon detection efficiencies (PDE) were determined with illumination by
275- and 470-nm light fed into the cryostat via an optical fiber. A cryogenic shutter provided a true dark
condition, where thermal radiation from room temperature is shielded and the thermal excitations in the
chips are frozen. A second tunneling breakdown was observed at this condition, which substantially limits
the operating voltage range for the temperatures 20–30 K. Below ∼5 K, both SiPMs recover to an operating
overvoltage range of 3–5 V. We found the chips function through the entire tested temperature range and
are capable of withstanding thermal cycling with no major performance degradation.
1 Introduction
A multitude of scientific experiments [1–4] require cryo-
genic single-photon detectors. The traditional choice for
a single-photon detector is a photomultiplier tube, but
in cryogenic conditions PMTs can be difficult to work
with due to their typically large form factors and high
operating voltages.
In the interest of finding a more suitable detec-
tor, we have characterized the operational parameters
and photon detection efficiencies of two commercially
available silicon photomultipliers from room tempera-
ture to the sub-Kelvin regime. These SiPMs are single-
photon avalanche diode (SPAD)-based multipixel pho-
ton counter (MPPC) devices.
Our motivation for studying SiPMs at low temper-
atures is to measure their characteristics as detectors
for our project in two-photon spectroscopy of the 1 S–
2 S transition in ultracold, magnetically trapped atomic
hydrogen [1]. The experiment requires detection of
Otto Hanski and Tom Kiilerich have contributed equally to
this work.
a e-mail: otolha@utu.fi (corresponding author)
b e-mail: tckiil@utu.fi
c e-mail: srahop@utu.fi
d e-mail: assema@utu.fi
e e-mail: jmiaho@utu.fi
f e-mail: viacheslav.dvornichenko@utu.fi
g e-mail: servas@utu.fi
extremely low fluxes of fluorescent Lyman-alpha radi-
ation. SiPMs are an attractive alternative to cryogenic
PMTs in this experiment due to their relatively low
operating voltages in the range of tens of volts, along
with high PDEs and small form factors.
2 Measurement setup
We tested and compared two commercially available
SiPM models, Hamamatsu S13370-6050CN [5] and
Onsemi MicroFJ-30035-TSV [6]. They were tested at
two wavelengths, visible blue at 470 nm and UV at
275 nm. The Hamamatsu chip was chosen because it
is specifically designed for UV detection, with a room-
temperature photon detection efficiency (PDE) of 17%
at 275 nm and 12% at 121 nm. This chip has also previ-
ously been tested at cryogenic and sub-Kelvin temper-
atures [7,8]. The Onsemi chip was selected due to also
having been tested at cryogenic temperatures [3] and its
significantly lower operational voltage compared to the
Hamamatsu chip. Additionally it is manufactured at a
significantly lower price point, making it an attractive
alternative especially for larger detector arrays.
A notable difference in the Hamamatsu chips com-
pared to Onsemi is the use of metallic thin film quench
resistors for the SPAD microcells. This helps reduce
the effect of temperature on pulse shape, as the metal
resistor resistance does not significantly change on cool-
123
89 Page 2 of 9 Eur. Phys. J. D (2025) 79 :89
Fig. 1 Overall measurement scheme, including a sketch of
the measurement design a and a picture of the SiPM instal-
lation within the dilution fridge b
ing into cryogenic region. The Onsemi MicroFJ chip
uses semiconductor quench resistors. Additionally, the
Onsemi chip has a built-in fast output mode, which uses
a high-pass filter to filter out the typical slow decaying
tail of an avalanche voltage pulse. For the Hamamatsu
chip, there is no built-in fast signal option, so we instead
used a self-made high-pass RC filter with a cutoff at
31.8 MHz (R = 50 Ω, C = 100 pF).
In our measurement setup (Fig. 1), each SiPM was
mounted to a biasing and measurement board (Fig. 2)
and placed on the mixing chamber of a homemade wet
dilution refrigerator system. The biasing board design
is based on the one used by Wiesinger [3]. In addition
to the biasing circuit itself, the bias board includes two
thermometers, a platinum chip and a RuO2 chip, in
order to cover the entire temperature range from room
temperature to below 100 mK. There is also a sec-
ond RuO2 chip on the board, used as a heater for fine
tuning of the SiPM temperature. The SiPM was illu-
minated with a LED from room temperature, coupled
into the cryostat via UV-compatible optical fiber (Thor-
labs UM22-400). The LED light was directly coupled
into the open fiber through an optical window separat-
ing the cryostat environment from room temperature.
The LED and the optical window to the cryostat were
encased in a lighttight box to shield the SiPM from any
stray light from the environment. Additionally, several
baffles were positioned along the neck of the cryostat
to shield the cryostat from room-temperature radiation
and to further shield the SiPM from stray photons.
Additionally, a rotatable shutter thermalized at the
dilution fridge mixing chamber was placed between the
fiber and the SiPM. The main purpose of the shutter
was to prevent room-temperature infrared photon leak-
age via the fiber during dark count measurements. The
fiber has also been thermalized at several temperatures
(4 K, 1 K, mixing chamber) along the way, to minimize
heat leak from the fiber connection at room tempera-
ture. The LED output was purposely left unfocused to
minimize total photon flux through the fiber. The out-
put from the fiber was also left uncollimated, resulting
in a beam diameter of roughly 4 to 5 mm at the SiPM.
Our ability to effectively stabilize the system tem-
perature during the cooldown and warm-up processes
is limited, which restricted our ability to perform mea-
surements between liquid helium (LHe, 4.2 K) and
liquid nitrogen (LN2, 77 K) temperatures, as well as
between liquid nitrogen and room temperatures. Due
to low thermal coupling of the SiPM to the dilution
fridge mixing chamber, we were able to overheat the
SiPMs and stabilize the temperature for measurements
up to ∼ 40 K.
Pulseform data acquisition was done via a Keysight
DSOX1102G oscilloscope, and the illuminating LED
was controlled via the same oscilloscope’s signal gener-
ator. The SiPM bias voltages are controlled, and pho-
tocurrents were measured via a Keithley 6487 current
source. In our amplifier chain, we used a ZFL-1000NL
and a ZFL-1000NL+ at room temperature. Code and
data are available on GitHub [9].
For breakdown voltage (Vbd) measurements, we used
several techniques. At room temperature, we measured
the breakdown voltage directly from the dark current
as the crossing point of zero current with a linear fit
on a
√
I/V plot [10]. At cryogenic temperatures, the
dark count rate is too low to generate a measurable
dark current, necessitating a different approach. We
used two methods to extract the breakdown voltage in
this regime; firstly, we measured the breakdown voltage
from the pulse height of our SiPM versus voltage [3]. Vbd
was subsequently determined as the crossover point of
123
Eur. Phys. J. D (2025) 79 :89 Page 3 of 9 89
Fig. 2 Picture (left) and circuit diagram (right) of the SiPM biasing and readout board
pulse height with zero. The second method is to use low
LED power, 10 µA current for the blue LED and 50 µA
for the UV LED, to illuminate the SiPM at a low signal
level in lieu of dark current. Vbd can then be measured
with the same methodology as in the room-temperature
measurements. The results from these different meth-
ods are in agreement with each other within the margin
of error.
A more detailed description of the measurement
setup and data analysis methodology is given in the
reference [11].
3 Results
3.1 Pulse shape
Similar to previous experiments [3], we saw pulse shape
changes due to quench resistor temperature dependence
when cooling down to cryogenic temperatures (Fig. 3a)
with our Onsemi chip. At room temperature, the pulse
shape is characterized by a fast voltage increase from
the avalanche, followed by an exponential decay back to
zero voltage with a characteristic time of around 100 ns.
At lower temperatures, the pulse shape is divided into
two components: a fast and a slow component. The fast
component decays with characteristic time below 10 ns,
whereas the slow component decays with characteristic
times above 100 ns. Additionally, the peak pulse volt-
age decreased by roughly 30% when cooled down from
room temperature to liquid nitrogen temperature. No
further degradation of maximum pulse voltage was seen
on cooling below this point, and below 10 K, the pulse
shape did not change at all.
For the Hamamatsu chip, we saw minor pulse shape
changes from room temperature to liquid nitrogen tem-
perature (see Fig. 3b). When further cooling the dilu-
tion fridge from liquid nitrogen down to 150 mK (equiv-
alent to 186 mK at the SiPM), no further changes were
detected. This is due to a metallic quench resistor hav-
ing orders of magnitude smaller temperature depen-
dence in the low-temperature regime than that of semi-
conductor resistors. For Hamamatsu, we additionally
utilized a high-pass RC filter to simulate a fast out-
Fig. 3 SiPM pulse shapes, measured as a 30 pulse average.
Filter refers to the high-pass RC filter described in Sect. 2.
For Hamamatsu, the pulse shape remains the same between
81 K, 5.7 K and 186 mK
put mode and increase our pulse counting resolution at
higher photon fluxes.
At room temperature, the Onsemi fast output chan-
nel could be used to generate extremely sharp rise time
pulse signals for pulse counting purposes without the
long-tail characteristic of a recharging SPAD. However,
the on-chip filters on the Onsemi chip did not function
properly when cooled down, and below 110 K, the fast
output channel created a lot of excess noise. This ren-
123
89 Page 4 of 9 Eur. Phys. J. D (2025) 79 :89
dered the channel inoperable at low temperatures. One
should instead use a room-temperature high-pass filter
on the normal output channel, as was done with the
Hamamatsu chip.
The relevance of the pulse shape changes depends
on the application the sensor is used for. When used
as a single-photon counter for low photon fluxes, the
shape changes do not affect the sensor functionality in a
major way, since in this type of measurement one should
count the pulses by rising edge. In these measurements,
a high-pass filter should regardless be implemented to
cut the slow signal tail for a better SNR. When used
for measuring large photon fluxes where the signal will
be calculated as an integral of the signal, one must take
into account the signal shape, as it strongly affects the
integral of the pulse shape and the detected photocur-
rent.
3.2 Dark count rate
An essential feature of the measurements at low tem-
peratures is that a true dark condition can be real-
ized when excitation via light or thermal effects are
fully frozen out. Starting from room temperature, we
observed a fast drop of the dark count rate (DCR),
roughly by the factor of 2 for each 10 K, consistent with
the manufacturers’ specifications. Already at ∼ 100 K,
we noticed that the dark count rate strongly depends
on the position of the shutter above SiPM. Closing the
shutter decreased the DCR by more than two orders
of magnitude, reaching dark count rates well below 1
Hz. With the closed shutter, we have a true dark con-
dition; the upper end of the optical fiber is located in a
room-temperature lighttight box, but there can remain
a tiny flux of photons leaking into the low-temperature
part of the apparatus due to imperfections in the shield-
ing at room temperature. These photons have sufficient
energy to create avalanches and generate nonzero pho-
tocurrent. This leakage can be blocked with a cold shut-
ter after the fiber.
Previous experiments have reported similar dark
count rates below 1 Hz [3], and it is not entirely clear
why they did not have similar issues with photon leak-
age, despite their lack of a cryogenic shield. Our main
assumption is that the room-temperature shielding of
the fiber in our setup is not as good as used in previous
experiments, which might allow more radiation to enter
the fiber, and the level of photon leakage with optimal
room-temperature shielding is sufficient to reduce the
effects of the leakage to negligible levels.
Once the chips were cooled below 100 K and the shut-
ter was closed, we could only see occasional pulses at
count rates below 1 Hz. This is expected, since at tem-
peratures below 200 K, the dark count rate is domi-
nated by band-to-band tunneling [12], which has a sig-
nificantly lower excitation rate compared to thermally
excited dark counts at higher temperatures. No further
reduction in the DCR was observed on cooling into
sub-Kelvin range. We suggest that the residual DCR
is caused by scattered thermal radiation photons which
leak inside the vacuum can of our refrigerator via imper-
fect radiation shields. A tiny fraction of them can reach
the chip through the gap between the shutter plate
and the chip surface. Another possibility could be, e.g.,
cosmic rays or high-energy elementary particles, which
penetrate through the metal enclosures of our cryostat.
We did not perform a detailed quantitative study of
DCRs below 1 Hz.
Closing the shutter to achieve a true dark condi-
tion was a key factor for observing the second break-
down and correct determination of the operating volt-
age range, as described in the next section.
3.3 Breakdown voltages and tunneling effect
The breakdown effect of a SiPM is seen as a rapid
increase in the photosensitivity of the detector above a
certain bias voltage, the breakdown voltage Vbd. At this
point, the electric field is high enough that a charge car-
rier generated by an incident photon will gain sufficient
kinetic energy over the mean free range in the bulk to
generate another charge carrier pair upon impact, lead-
ing to an avalanche. At room temperature, the charge
carrier may be created by thermal excitations, without
illumination by light. This leads to the fairly high dark
count rate, increasing proportionally to the overvoltage
above Vbd. The breakdown voltage is typically temper-
ature dependent, due to increased carrier mobility at
low temperatures [13].
At sufficiently low temperatures (below ∼ 100 K),
the thermal excitations are frozen out, and if no pho-
tons are coming to the SiPM, thermally excited break-
downs do not spontaneously occur even at fairly large
overvoltages. Therefore, under true dark conditions we
need to send a small flux of photons, which allows mea-
surements of the Vbd, as explained in Sect. 2.
When increasing the overvoltage further in dark con-
ditions at temperatures below 40 K, we found that
the photocurrent exhibited a sudden and steep growth
indicating the appearance of a second breakdown at
a voltage Vt, which depends strongly on temperature.
For the Onsemi SiPMs, we found that the difference
between these two breakdown voltages exceeds 12 V
at and above 40 K, which would indicate a cryogenic
operational voltage range substantially larger than the
5 V overvoltage recommended by the manufacturer
for room-temperature applications. We demonstrate
the typical behavior of both breakdown phenomena in
Fig. 4.
We suggest that the second breakdown could be
caused by a tunneling phenomenon, when the bias volt-
age is high enough for charge carriers to directly tun-
nel from the valence band (band-to-band tunneling),
or intermediate impurity-induced energy states to the
conduction band (trap-assisted tunneling).
This effect has not been previously reported, and the
physical basis for the second breakdown has not previ-
ously been explained. In their characterization exper-
iments, Zhang et al [7] reported a rapid rise in pho-
tocurrent at higher overvoltages, which they suggested
123
Eur. Phys. J. D (2025) 79 :89 Page 5 of 9 89
Fig. 4 Different types of breakdown behavior on the
Hamamatsu chip, measured at 1 K. Normal breakdown
occurs at Vbd ≈ 41.5 V, tunneling breakdown is seen at
Vt ≈ 44.2 V
to be due to increased afterpulsing probability causing
self-sustaining afterpulse trains. We were able to repro-
duce this behavior (orange line in Fig. 4) with an open
shutter. We suggest that by blocking photon leakage
from room temperature with our cryogenic shutter, we
create a true dark condition, necessary for observation
of the proposed tunneling breakdown effect.
3.4 Operational voltage range
After understanding the breakdown phenomena at
cryogenic and true dark conditions, we may now define
an operating range for the SiPMs as the voltage range
between Vbd and Vt. We characterize the operating
range at temperatures below 40 K, where the second
breakdown can be clearly identified. The temperature
range between 77 K and 40 K is problematic for reli-
able stabilization of the SiPM temperature inside the
vacuum can of the dilution refrigerator. As shown in
Fig. 5, Vt showed a pronounced minimum for both
chips at around 23 K. The first breakdown voltage
Vbd monotonously decreased during cooldown from its
room-temperature value, by approximately 5 V for the
Onsemi chip, and approximately 10 V for the Hama-
matsu chip, reaching its minimum around LN2 tem-
perature. Vbd remained nearly constant throughout our
primary measurement range below 40 K.
Operating the Hamamatsu SiPM near 25 K becomes
problematic, since overvoltages above 0.3 V cannot be
used. At the lowest temperatures below 5 K, the oper-
ating range increases to ∼ 3 V, which is comparable
with the values recommended by the manufacturer at
room temperature. The Onsemi device operating range
increases to > 5 V below 5 K and does not change
at lower temperatures. A remarkably large operating
range of more than 12 V is observed for the Onsemi
chip at ∼ 40 K. The overall behavior of both chips is
similar to reports on earlier experiments on the Hama-
matsu chip by Zhang et al [7].
3.5 Afterpulsing
Afterpulsing behavior is typically seen as an appear-
ance of secondary pulses followed the main pulse with
a delay comparable with the pulse relaxation time. An
afterpulse is a secondary avalanche of a SPAD typically
caused by charge carriers from the initial avalanche get-
ting trapped in impurities of the silicon [14] and caus-
ing a second avalanche when released. A single primary
event may have several afterpulses, either due to mul-
tiple trapped charge carriers from the primary pulse
released at different intervals or by having an afterpulse
trigger a second afterpulse.
Afterpulses are easiest to detect by observing dark
counts of the device, since due to the low dark count
rate of around 1 Hz at cryogenic conditions, the like-
lihood of detecting multiple dark counts in a short
time window is significantly lower than the afterpuls-
ing probability. See Fig. 6 for reference pulse shapes
seen in cryogenic measurements.
The afterpulsing effects essentially function as an
additional noise source for both integrating measure-
ments for high photon fluxes and pulse counting mea-
surements for low photon fluxes. Both the long and
short delay afterpulses distort the integral of a single-
detection-event signal and will therefore distort any
averaging measurements. The long delay pulses addi-
tionally create additional pulse counts, indistinguish-
able from primary detection events, creating a system-
atic error in a pulse count measurement. The short
delay pulses do not cause this issue, as they can be
easily distinguished by a significantly lower pulse volt-
age.
Afterpulsing creates a systematic error, which always
overestimates the real photon flow. For high-photon-
flux measurements where we are detecting a photocur-
rent, afterpulsing effectively functions as an extra gain
factor for the signal. For experiments where measuring
the absolute photon flux is important, the afterpuls-
ing effects need to be characterized and accounted for.
However, if one is only interested in the relative photon
flux determination, the afterpulses should have negligi-
ble effect on measurement error.
At temperatures below 40 K, we detected a signifi-
cant increase in the probability of afterpulsing and an
increase in the average time delay between the pri-
mary pulse and afterpulse(s). Together, these effects
occasionally combine to create long-lasting afterpulse
trains, where successive afterpulses cause further after-
pulses. This cascade of afterpulses can last for a long
time at cold temperatures, up to roughly 1 ms. As the
afterpulsing probability also increases with overvoltage,
these afterpulse trains can limit the functionality of the
SiPMs at low temperatures. This is due to increased
noise characteristics and heating effects from afterpulse-
train-induced photocurrent, as previously reported [7].
We did not do a quantitative analysis of the afterpuls-
ing effects in the scope of this study, as the effect has
low relevance to our intended application of the SiPM
chip as a fluorescence detector.
123
89 Page 6 of 9 Eur. Phys. J. D (2025) 79 :89
Fig. 5 Operational voltage ranges at temperatures T ≤ 40 K
Fig. 6 Afterpulsing pulse shapes, measured from the
Hamamatsu chip. a Afterpulse with a long delay and cross
talk on the secondary pulse. b Afterpulse with a short delay.
c Afterpulse with no cross talk
3.6 Photon detection efficiency
For the measurement of photon detection efficiency
(PDE) at cryogenic temperatures, the main difficulty is
having a calibrated photon flux to calculate the detec-
tion efficiency. As we do not have access to a calibrated
cryogenic photon source, we instead opted for a relative
measurement of PDE.
The room-temperature specifications of the sensors
are listed in Table 1. Using these data as a calibra-
tion point for our light source, we are able to calcu-
late the relative PDE of the sensor at lower tempera-
tures without knowing the absolute value of the incident
photon flux. We assume the transmission of our fiber
does not change significantly with temperature, and the
fiber alignment with respect to the SiPM also remains
constant during the cooldown. The former assumption
is supported by initial test measurements without a
fiber coupling yielding similar, albeit less accurate, PDE
numbers when compared with measurements using a
fiber coupling. The latter assumption is made on the
basis of the mechanically rigid mounting of both the
SiPM and the fiber output at the mixing chamber, min-
imizing movement of both components during and after
cooldown.
To measure the relative PDE, we pulse the LED
repeatedly with constant pulse parameters, and we
can determine from the statistics the mean number
of detected events via a Poissonian fit. From this, we
subtract the dark count rate to determine the mean
number of photons per pulse [3,11]. We then use the
manufacturer specifications for the room-temperature
PDE values to calibrate our absolute PDE value rela-
tive to the photon flux from our LED, and use this as
the comparison point for our low-temperature measure-
ments. In order to avoid afterpulsing effects, we chose
a low overvoltage of 2.5 V for the PDE measurements.
In the case of the Hamamatsu chip, the manufacturer
provided PDE values were measured at an overvoltage
of 4 V, which was taken into account by first using the
manufacturer value to calibrate our room-temperature
2.5 V overvoltage PDE, which was subsequently used
as the reference point for our low-temperature measure-
ments.
When cooled down from room temperature to tem-
peratures below 1 K, we see a large drop in PDE
(Fig. 7). For both tested sensors, the drop in PDE is
significantly larger for the UV than for the visible blue
light. Our measurement data are in good agreement
with previous measurements for both devices [3,7,8].
Both SiPMs are significantly more sensitive to visible
blue light than to UV light at cryogenic temperatures,
with a difference of 1 to 2 orders of magnitude. There-
fore, for detection of scattered or fluorescent UV light,
123
Eur. Phys. J. D (2025) 79 :89 Page 7 of 9 89
Fig. 7 PDE as a function of temperature for blue (left) and UV (right) light, with zoom window on the low-temperature
regime. Each point measured at a constant 2.5 V overvoltage
Table 1 Room-temperature specifications for our SiPMs [5,6]
Hamamatsu Onsemi
Vbd [V] 51.3 24.5
Overvoltage [V] 4.0 2.5 2.5
PDE(470 nm), [%] 29 26 31
PDE(275 nm), [%] 17 16 8
Breakdown voltages Vbd measured by us, due to high error margins in manufacturer specs
a measurement scheme with frequency converters based
on TPB [15] would remain significantly more efficient
even after conversion losses.
3.7 Thermal cycling tolerance
We monitored the thermal cycling tolerance of the chips
in order to get an idea of their useful lifetime in cryo-
genic experiments. Through several thermal cycles, we
saw no notable drop in PDE or other performance char-
acteristics, albeit with the Onsemi chip, which has a
protective glass layer covering the SiPM pixels, and we
saw damage in the form of cracks on the glass layer
(see Fig. 8). From our PDE data, we saw no perfor-
mance impact from these cracks when comparing room-
temperature PDE before and after thermal cycling.
The cracks in the glass cover are typically formed
during the first cooldown, and there is no notable pro-
gression in the damage over subsequent cooldowns. Due
to this and the aforementioned good retention of actual
performance characteristics, the Onsemi SiPMs appear
to be well tolerant to thermal cycling and can be oper-
ated over several thermal cycles without replacement.
Fig. 8 Thermal cycling surface damage. a Hamamatsu
without glass cover b Onsemi with glass cover & cracks
The Hamamatsu chips we tested have no protective
glass cover, and therefore, no cracking or other damage
related to thermal cycling was observed.
123
89 Page 8 of 9 Eur. Phys. J. D (2025) 79 :89
4 Conclusions
We have characterized the functionality of two com-
mercial SiPM models, Hamamatsu S13370-6050CN and
Onsemi MicroFJ-30035-TSV, as low-flux photon detec-
tors in a wide temperature range from room tempera-
ture down to 90 mK. We found that while their per-
formance drops with temperature, they are functional
with a PDE above 10% for the Hamamatsu chip and
above 5% for the Onsemi chip throughout the entire
temperature range in the visible blue light range. For
the Hamamatsu chip, we also measured the PDE to
remain above 10% down to roughly 100 K at UV wave-
lengths and confirmed the ability of both chips to detect
UV at temperatures at around 1 K, albeit at a signifi-
cantly lower PDE of less than 1%. Hamamatsu showed
better overall PDE performance than Onsemi, although
with a smaller operating voltage range. The operating
voltage for Hamamatsu detectors is also nearly a factor
of 2 larger than for Onsemi detectors, which results in
a larger Joule heating for the same photocurrent. This
can become an issue at low temperatures.
The relaxation after an avalanche became very slow
for Onsemi detectors at low temperature due to semi-
conductor quench resistors. However, well-resolved and
sharp signal peaks are observed in the beginning of
an avalanche pulse, which can easily be detected and
counted. Metal-based quench resistors utilized in the
Hamamatsu detectors therefore do not provide a signif-
icant benefit for pulse counting experiments at low tem-
peratures. However, for higher-photon-flux measure-
ments the increased area under each pulse (released
charge carriers per pulse) improves the SNR of any
integrating or photocurrent measurements. Addition-
ally, measurements with a large range of temperatures
will benefit from the constant pulse shape of the Hama-
matsu chip with regard to temperature due to its metal
quench resistors. Both chips also exhibited excellent
dark count characteristics at low temperatures.
For the purpose of low-temperature fluorescence
detection in the UV range, we would suggest using a
wavelength shifting material for conversion of incident
UV photons to the blue light for more efficient detec-
tion.
We have also detected a possibly novel behavior in
the form of a second breakdown voltage, detectable
as a sudden exponential increase in the photocurrent
at a high overvoltage, with notably different behavior
from the typical afterpulsing effects. We posit a possible
band-to-band tunneling effect as a possible explanation
for this behavior, albeit further study would be neces-
sary to ascertain this theory.
For follow-up studies, we would recommend a more
thorough analysis of the afterpulsing behavior to con-
firm the assumption of signal linearity and to more
accurately characterize the exact afterpulse probabil-
ity effect. Building a better theoretical understanding
of the proposed tunneling breakdown effect might pro-
vide information on how to optimize SiPM design for
maximum gain at cryogenic temperatures. Extensive
testing with large sample sizes of detectors from dif-
ferent production batches and over a larger number of
cycles would also provide worthwhile information on
the reproducibility of cryogenic SiPM characteristics.
Acknowledgements This project was supported by the
Jenny and Antti Wihuri foundation. Additional thanks to
the anonymous referees for their excellent feedback and sug-
gestions, which notably improved the quality of this publi-
cation.
Author contributions
All authors contributed to the concept and the design
of the experimental setup. Setting up and running
the experiment were performed by Otto Hanski, Tom
Kiilerich and Slava Dvornichenko. Data collection was
done by Tom Kiilerich, and data processing and anal-
ysis were done by Tom Kiilerich and Otto Hanski. The
first draft of the manuscript was written by Otto Hanski
and Tom Kiilerich. All authors commented on previous
versions of the manuscript and approved its final ver-
sion.
Funding Open Access funding provided by University of
Turku (including Turku University Central Hospital). Dilu-
tion fridge operation costs were covered by the Wihuri Phys-
ical Laboratory, with funding received from the Jenny and
Antti Wihuri foundation.
Data Availability Statement This manuscript has asso-
ciated data in a data repository. The datasets generated
and/or analyzed during the current study are available from
the corresponding author on request. Unsorted data are also
available in the related data collection uploaded in GitHub
[9].
Declarations
Conflict of interest All authors certify that they have no
affiliations with or involvement in any organization or entity
with any financial interest or non-financial interest in the
subject matter or materials discussed in this manuscript.
Open Access This article is licensed under a Creative Com-
mons Attribution 4.0 International License, which permits
use, sharing, adaptation, distribution and reproduction in
any medium or format, as long as you give appropriate credit
to the original author(s) and the source, provide a link to
the Creative Commons licence, and indicate if changes were
made. The images or other third party material in this arti-
cle are included in the article’s Creative Commons licence,
unless indicated otherwise in a credit line to the material. If
material is not included in the article’s Creative Commons
licence and your intended use is not permitted by statu-
tory regulation or exceeds the permitted use, you will need
to obtain permission directly from the copyright holder.
123
Eur. Phys. J. D (2025) 79 :89 Page 9 of 9 89
To view a copy of this licence, visit http://creativecomm
ons.org/licenses/by/4.0/.
References
1. J. Ahokas, A. Semakin, J. Ja¨rvinen, O. Hanski, A. Lap-
tiyenko, V. Dvornichenko, K. Salonen, Z. Burkley, P.
Crivelli, A. Golovizin et al., A large octupole mag-
netic trap for research with atomic hydrogen. Rev.
Sci. Instrum. 93(2), 023201 (2022). https://doi.org/10.
1063/5.0070037
2. A. So´te´r, K. Todoroki, T. Kobayashi, D. Barna,
D. Horva´th, M. Hori, Segmented scintillation detec-
tors with silicon photomultiplier readout for measur-
ing antiproton annihilations. Rev. Sci. Instrum. 85(2),
023302 (2014). https://doi.org/10.1063/1.4863648
3. M. Wiesinger, F. Stuhlmann, M. Bohman, P. Micke, C.
Will, H. Yildiz, F. Abbass, B.P. Arndt, J.A. Devlin,
S. Erlewein, M. Fleck, J.I. Ja¨ger, B.M. Latacz, D.
Schweitzer, G. Umbrazunas, E. Wursten, K. Blaum,
Y. Matsuda, A. Mooser, W. Quint, A. Soter, J.
Walz, C. Smorra, S. Ulmer, Trap-integrated fluores-
cence detection with silicon photomultipliers for sym-
pathetic laser cooling in a cryogenic penning trap. Rev.
Sci. Instrum. 94(12), 123202 (2023). https://doi.org/10.
1063/5.0170629
4. T. Heindl, T. Dandl, M. Hofmann, R. Kru¨cken, L. Ober-
auer, W. Potzel, J. Wieser, A. Ulrich, The scintillation
of liquid argon. Europhys. Lett. 91(6), 62002 (2010)
5. Hamamatsu: VUV-MPPC Datasheet. (2017). Hama-
matsu. https://hamamatsu.su/files/uploads/pdf/3_
mppc/s13370_vuv4-mppc_b_(1).pdf
6. Onsemi: Silicon Photomultipliers Datasheet. Onsemi.
Rev. 7. (2021). https://www.onsemi.com/download/
data-sheet/pdf/microj-series-d.pdf
7. J. Zhang, D. Goeldi, R. Iwai, M. Sakurai, A. Soter, Scin-
tillation detectors with silicon photomultiplier readout
in a dilution refrigerator at temperatures down to 0.2
k. J. Instrum. 17(06), 06024 (2022). https://doi.org/10.
1088/1748-0221/17/06/P06024
8. R. A´lvarez-Garrote, E. Calvo, A. Canto, J.I. Crespo-
Anado´n, C. Cuesta, A. de la Torre Rojo, I. Gil-Botella,
S. Manthey Corchado, I. Mart´ın, C. Palomares, L.
Pe´rez-Molina, A. Verdugo de Osa, Measurement of the
photon detection efficiency of hamamatsu vuv4 sipms at
cryogenic temperature. Nucl. Instrum. Methods Phys.
Res., Sect. A 1064, 169347 (2024). https://doi.org/10.
1016/j.nima.2024.169347
9. T. Kiilerich, SiPM Measurement Scripts & Data Collec-
tion. https://github.com/WaffleCoder4Life/SiPMs
10. Onsemi: Introduction to the silicon photomultiplier
(sipm). OnSemiconductors Application Notes (2021)
11. T. Kiilerich, Characteristics of silicon photomultiplier
at cryogenic temperatures. Master’s thesis, University
of Turku (2024)
12. G. Collazuol, M.G. Bisogni, S. Marcatili, C. Piemonte,
A. Del Guerra, Studies of silicon photomultipliers at
cryogenic temperatures. Nucl. Instrum. Methods Phys.
Res., Sect. A 628(1), 389–392 (2011). https://doi.org/
10.1016/j.nima.2010.07.008
13. E.A. Gutierrez-D, M.J. Deen, C.L. Claeys, Low Temper-
ature Electronics: Physics, Devices, Circuits, and Appli-
cations. Academic Press, ebook (2001). https://doi.org/
10.1063/1.1485590
14. B.K. Ridley, Lucky-drift mechanism for impact ion-
isation in semiconductors. J. Phys. C: Solid State
Phys. 16(17), 3373 (1983). https://doi.org/10.1088/
0022-3719/16/17/020
15. L. Baudis, G. Benato, R. Dressler, F. Piastra, I. Usolt-
sev, M. Walter, Enhancement of light yield and sta-
bility of radio-pure tetraphenyl-butadiene based coat-
ings for VUV light detection in cryogenic environments.
J. Instrum. 10(09), 09009 (2015). https://doi.org/10.
1088/1748-0221/10/9/P09009
123