The low temperature performance of CsI(Na) crystals for WIMPs direct searches

Original Paper



   Previous studies showed that CsI(Na) crystals have significantly different waveforms between \(\alpha \) and \(\gamma \) scintillations.


   In this work, the light yield and pulse shape discrimination capability of CsI(Na) scintillators as a function of the temperature down to 80 K have been studied.


   As temperature drops, the fast component increases and the slow component decreases. By cooling the CsI(Na) crystals, the light yield of high-ionization events is enhanced significantly, while the light yield of background \(\gamma \) events is suppressed. At 110 K, CsI(Na) crystal achieves the optimal balance between low threshold and good background rejection performance.


   The different responses of CsI(Na) to \(\gamma \) and \(\alpha \) at different temperatures are explained with self-trapped and activator luminescence centers.


   PSD capability of CsI(Na) reaches the peak at 110 K, which can be the optimum operating temperature for future CsI(Na)-based dark matter detector.


CsI(Na) Crystal Dark matter Nuclear recoil Particle discrimination Self-trapped exciton 


It is generally accepted that dark matter makes up about 27% of the universe [1]. Weakly interacting massive particles (WIMPs) are considered as a kind of new particles, which are beyond the standard model (SM) and can be used to explain the dark matter. WIMPS, neutrons, and \(\alpha \) can scatter with the nuclei of target material and then produce nuclear recoil events. Electrons and \(\gamma \) can interact with the electrons of target material and generate electron recoil events. The nuclear recoil energy (1–100 keV) will be deposited in the target material. The energy threshold of detector material should be low enough to respond to such low energy.

Since such nuclear recoil event is extremely rare (estimated to be under 0.1 events/kg/day), it is vital to build shields with very low background materials surrounding the detector. However, there still exist background \(\gamma \) events from the shielding material or the detector material itself inevitably. Thus, a background rejection power better than 10e-5 of the detector will be critical to distinguish nuclear recoils from the background electron recoils. In this paper, we use \(\alpha \) instead of neutron to analyze the phenomena of nuclear recoils.

Cesium iodide is studied extensively and used to model crystal scintillator. Doped CsI such as CsI(Tl) and CsI(Na) exhibits high light yield (LY) and good pulse shape discrimination (PSD) capability. Previous studies showed that CsI(Na) exhibits very different responses to nuclear recoils and electron recoils [2]. There are two kind of light components. The fast component of CsI(Na) from self-trapped exciton (STE) luminescence can be significantly excited by particles with a high ionization density, like nuclear recoil events. While the slow component of CsI(Na) can be attributed to the activator luminescence (Na emission), which is mainly excited by particles with a low ionization density, like electron recoil events. At room temperature, the decay time of fast component of CsI(Na) is about 20 ns while the slow component is around 670 ns. This significant difference between nuclear and electron recoil pulse shapes makes CsI(Na) a possible candidate for dark matter searches.

Since nuclear recoils are dominated by STE emission, the detection threshold for nuclear recoil events depends on the light yield of fast light of CsI(Na). It is possible to enhance the STE emission by lowering the temperature of the crystal. The scintillation light yield of pure CsI crystal is under 5000 photons/MeV at room temperature and reaches 100,000 photons/MeV at 77 K [3, 4].

The aim of the present work is to study the influence of low temperature (80 K to room temperature) on the light yield and PSD capability of CsI(Na) and find the optimum operating temperature of CsI(Na) to detect WIMPs directly. Additionally, pure CsI crystal is tested as comparison.


In this study, the response of CsI(Na) and pure CsI crystals to \(\gamma \) and \(\alpha \) particles under room temperatures was tested. The emission was excited by \(\gamma \)-rays of 661.7 keV from a 0.5 micro-Curie \({}^{137}\hbox {Cs}\) source and \(\alpha \)-rays of 5244 keV from a 5 micro-Curie \({}^{239}\hbox {Pu}\) source. The samples were placed into a homemade cryostat that is cooled by liquid nitrogen.

Experimental setup

Our setup is illustrated in Fig. 1. A cubic-shaped crystal sample was placed inside a cold copper base of the homemade cryostat. The dimensions of both CsI(Na) and pure CsI crystals were 25 mm \(\times \) 25 mm \(\times \) 25 mm. The doping concentration of Na in CsI(Na) was about 0.02%. The left and right surfaces of crystal served as the light-emitting surfaces, and the other four surfaces were cooled by the cold base. The cold base was cooled by a copper rod constantly. The bottom end of the copper rod was soaked in liquid nitrogen in a dewar. A heater with proportion integral derivative (PID) power control was mounted in the joint part of the cold base and the copper rod. A Pt100 thermal sensor was attached to the bottom surface of the crystal cube to monitor the temperature and give feedback to the temperature control system. The temperature control system can reach a precision of ± 0.1 K. The crystal and the cold base were encapsulated in a 2-mm-thick fused quartz vessel, which had an evenly distributed 80% transmittance within a wide spectrum range. Since CsI(Na) crystals are deliquescent, the surfaces were polished to eliminate the deliquescent outermost layer by abrasive paper and then the crystal was put into oven to reduce water. Pure CsI just needs to be polished. The inside of fused quartz vessel was pumped to maintain the vacuum continually, which was vital to keeping low temperature as well as preventing further deliquescence of the crystal. During the experiment, the \(\alpha \)-source was put in the quartz vessel and the \(\gamma \)-source was put out of the quartz vessel. Two 2-inch PMTs (R8778, Hamamatsu) were placed outside the quartz vessel facing the two light-emitting sides of the crystal to collect the scintillation light. The R8778 PMTs feature a high quantum efficiency of more than 30% and an ultra-low background [5]. The quantum efficiency of R8778 PMT at 310 nm is close to its quantum efficiency at 410 nm.
Fig. 1

Experimental setup

Data acquisition

Figure 2 depicts the data acquisition system. The signals from the two PMTs were duplicated in a linear fan-in fan-out module (CAEN N625). One group of the two-channel signals from the PMTs were fed to the waveform acquisition device. The other group of the signals were fed into a discriminator (CAEN N840) module and converted to logical pulse signals. The threshold of the discriminator was set to be 0.5 single photon amplitude level. The two-channel logical pulses from the discriminators were fed into a gate generator and generate two-channel 1000 ns gate signals. The coincidence of the two gate signals triggered the acquisition of the two-channel signals from the PMTs by the waveform acquisition device. The full waveforms were transferred to a PC and saved on hard disk. The waveform acquisition device in this study was either four synchronized digitizers (CAEN V1729a) or an oscilloscope (Tektronix DPO3054C) depending on the acquisition window width. Each digitizer had a memory depth of 1.26 \(\upmu \)s with a sampling frequency of 2 GS/s. The four synchronized digitizers can acquire 5 \(\upmu \)s-long waveforms at 800 Hz. The oscilloscope can acquire 400 \(\upmu \)s-long waveforms with a sampling frequency of 2.5 GS/s at less than 1 Hz.
Fig. 2

Data acquisition system


The temperature of pure CsI and CsI(Na) samples was first brought down to 80 K and gradually rose to room temperature.

Waveform of CsI(Na)

Figure 3a, b shows the waveform of \(\alpha \) and \(\gamma \) for CsI(Na) at room temperature. The top of each picture is the original waveform, and the bottom one is the waveform that got by averaging 200 original waveforms. From these pictures, it is clear that there are both fast and slow components for CsI(Na). For \(\alpha \), fast component dominates the waveform, which is opposite to the \(\gamma \). Figure 4a is the fast and slow components decay time (use \(\alpha \) as an example) of CsI(Na) that varies with temperature. Since the LY is too weak at room temperature, it is hard to fit the waveform and we just show the data from 80 to 155 K. The time width of both fast component and slow component broadens as the temperature decreases. At 80 K, the decay time is 700 ns (fast) and 2750 ns (slow), respectively.

Waveform of pure CsI

Figure 3c shows the waveform of \(\alpha \) for pure CsI. We just observe a fast component in the waveform. From Fig. 4b , the decay time varies with temperature between 80 and 150 K. At 80 K, the decay time is 750 ns. The \(\alpha \)-scintillations and \(\gamma \)-scintillations from pure CsI crystal are very similar to the fast components of scintillations from CsI(Na) crystal.
Fig. 3

Waveforms of CsI(Na) and pure CsI(Na). a The waveforms of \(\alpha \)-scintillations from CsI(Na) at room temperatures. b The waveforms of \(\gamma \)-scintillations from CsI(Na) at room temperatures. c The waveforms of \(\alpha \)-scintillations from pure CsI at room temperatures

Fig. 4

The decay time varies with temperature. a The fast decay time and slow decay time of \(\alpha \) vary with temperature from 80 to 150 K for CsI(Na). b The fast decay time of \(\alpha \) varies with temperature from 80 to 150 K for pure CsI

As a comparison, waveform of CsI(Na) has both fast and slow components while that of pure CsI has only fast component, suggesting that the slow component can be attributed to the Na activator. As a dopant, Na activators have a larger average spacing than those in CsI(Na). And the fact that \(\gamma \) events yield significantly more slow component than alpha events is consistent with the discussion on the mechanism in our previous papers [2, 6].

Light yields

We can integrate the waveform over fast decay time to get LY of fast component and over slow decay time to get LY of both fast and slow components, and the result is shown in Fig. 5. As temperature falls, the light yield of both fast and slow components of \(\alpha \)-scintillations from CsI(Na) crystal increases while the \(\gamma \)-scintillation one decreases. From 298 to 80 K, the light yield of \(\alpha \)-scintillations was increased by 22 and 26 times for the fast component window and full waveform respectively, while the all light yield of \(\gamma \)-scintillation decreases and the fast component nearly the same. For pure CsI, the light yield of both \(\alpha \)-scintillations and \(\gamma \)-scintillations increases as temperature falls. At 80 K, the light yield of \(\alpha \)-scintillations and \(\gamma \)-scintillations was increased by a factor of 18 and 11, respectively, compared to that at 298 K.
Fig. 5

The fast and all light yield (LY) of \(\alpha \)-scintillation and \(\gamma \)-scintillation for CsI(Na) and pure CsI

Pulse shape discrimination

We use the ratio that the LY of fast component to the total light yield of CsI(Na) to characterize the pulse shape discrimination factor. The decay time of fast component varies with the temperature; hence, the integral window to get the light yield also changes at different temperature. For \(\alpha \)-scintillations, the fast component dominates the waveform, and the ratio approaches 1. For \(\gamma \)-scintillations, the ratio is far less than 1, because the slow component is stronger than the fast one. A suitable window width can contribute to get optimal ratio to distinguish \(\gamma \)/\(\alpha \). Figure 6 shows the most proper integral window time at different temperature. As temperature falls, the decay time of fast component increases. There is a rapid change between 120 and 180 K.
Fig. 6

Optimized fast component window time at different temperatures. The window time is the time of intersection of the normalized waveforms of \(\gamma \)-scintillation and \(\alpha \)-scintillation from CsI(Na)

Figure 7 shows the PSD scatter plot at 160 K with a fast component integral window of 1.26 \(\upmu \)s. The Y-axis is the ratio of the fast component to the total light yield. The X-axis is the number of photoelectrons. The \(\alpha \)-scintillation and the \(\gamma \)-scintillation were divided into two separate belt zones. The top belt is nuclear recoil events, and the bottom is electron recoil events. The electron recoil events bend up sharply below 10 p.e. and cross the nuclear recoil event zone. Moreover, there are several electron recoil events mixed in the nuclear recoil event zone, which means the nuclear and electron recoil events cannot be separated totally.
Fig. 7

PSD scatter plot of gamma and alpha events at 160 K. Red dots are gamma events, and blue dots are alpha events (color figure online)


According to Payne [7] and Gridin [8], after the creation of electron-hole pairs, there are many possible branches that the created electrons and holes can interact with each other and with the activators. The electrons and holes can form excitons. The holes can be self-trapped to recombine with an electron and STE. Activators can capture electrons and holes and jump to excited states.

In pure CsI, STE is the dominated luminescence center. If the distance between an electron and a hole is r, the electron will diffuse under the Coulomb field and recombine with the hole with the probability \(p=1-e^{(-R_{Ons}/r)}\) \((R_{Ons}=e^2/(4_0 k_B T)\), where \(R_{Ons}\) is the Onsager radius, e is the electron charge, \(\varepsilon \) is the static dielectric permeability, \(k_B\) is the Boltzmann constant, and T is the temperature). As the temperature falls, the Onsager radius \(R_{Ons}\) rises and electrons at a longer distance from the hole can diffuse to recombine with the holes. This means more electrons and hole can form excitons and enhances the STE emission and fast components as temperature drops.

In CsI(Na), both STE luminescence and activator (Na) luminescence can take place. In order to form the excited state of activator, the activator will have to capture an electron and a hole, or a hole and an electron sequentially, which limits the rate of production of excited states of activator. The slow component is generated by activator luminescence. Since the total number of electrons and holes is fixed, the STE luminescence and activator luminescence are two competing processes. The prevailing process depends on the concentration of activator (\(n_A\)), electrons (\(n_e\)), and holes (\(n_h\)). The electrons and holes are formed in pairs, so \(n_e=n_h\). For particles with high ionization density, \(n_h\gg n_A\), the electrons are most likely to be captured by self-trapped holes, so the STE luminescence dominates. That explains why the waveform of \(\alpha \)-scintillation is dominated by fast component and very similar to waveforms from CsI.

For particles with low ionization density like \(\gamma \) or electrons, \(n_h\ll n_A\), the electrons have a much higher chance to be captured by activators than recombining with self-trapped holes and form STE. So the slow component prevails in the waveform of \(\gamma \)-scintillation from CsI(Na).

The capture radius of activator changes little as temperature drops, because it is caused by the dipole polarization of the neutral activator center in the electron Coulomb field [7]. As the temperature falls, Onsager radius rises but the activator captures radius changes little. So the STE emission rises and the activator emission drops.

By cooling the CsI(Na) crystals, we can enhance the quenching factor for high ionization density events, while suppressing the quenching factor for \(\gamma \) background. The significant increase in light yield of \(\alpha \)-scintillations from CsI(Na) at low temperatures lowers the energy threshold to detect nuclear recoil events.


CsI(Na) crystals exhibit discriminability between \(\alpha \) and \(\gamma \) scintillation at temperatures as low as 80 K. As temperature goes down, the light yield of alpha scintillation increases but that of gamma scintillation decreases. The response of CsI(Na) and pure CsI to \(\gamma \) and \(\alpha \) at different temperatures can be explained with the competing processes of forming STE and activator luminescence centers. By cooling the CsI(Na) crystals, light yield of high ionization density or nuclear recoils events is enhanced, while light yield of the gamma background events is suppressed. PSD capability of CsI(Na) reaches the peak at 110 K, which can be the optimum operating temperature for future CsI(Na)-based dark matter detector.


  1. 1.
    E. Komatsu, K. Smith, J. Dunkley, C. Bennett, B. Gold, G. Hinshaw, N. Jarosik, D. Larson, M. Nolta, L. Page et al., Astrophys. J. Suppl. Ser. 192(2), 18 (2011)ADSCrossRefGoogle Scholar
  2. 2.
    S. Xi-Lei, L. Jun-Guang, H. Tao, Z. Li, C. Jun, W. Yi-Fang, Z. Liang, Y. Bo-Xiang, C. Xiao, F. Jian et al., Chin. Phys. C 35(12), 1130 (2011)ADSCrossRefGoogle Scholar
  3. 3.
    M. Moszyński, W. Czarnacki, W. Klamra, M. Szawlowski, P. Schotanus, M. Kapusta, Nucl. Instrum. Methods Phys. Res. Sect. A Accel. Spectrom. Detect. Assoc. Equip. 504(1), 307 (2003)ADSCrossRefGoogle Scholar
  4. 4.
    S.S. Gridin, A.N. Belsky, N.V. Shiran, A.V. Gektin, IEEE Trans. Nucl. Sci. 61(1), 246 (2014)ADSCrossRefGoogle Scholar
  5. 5.
    D. Akerib, X. Bai, E. Bernard, A. Bernstein, A. Bradley, D. Byram, S. Cahn, M. Carmona-Benitez, D. Carr, J. Chapman et al., Nucl. Instrum. Methods Phys. Res. Sect. A Accel. Spectrom. Detect. Assoc. Equip. 703, 1 (2013)ADSCrossRefGoogle Scholar
  6. 6.
    X. Sun, J. Lu, T. Hu, L. Zhou, J. Cao, Y. Wang, L. Zhan, B. Yu, X. Cai, J. Fang et al., Nucl. Instrum. Methods Phys. Res. Sect. A Accel. Spectrom. Detect. Assoc. Equip. 642(1), 52 (2011)ADSCrossRefGoogle Scholar
  7. 7.
    S.A. Payne, N.J. Cherepy, G. Hull, J.D. Valentine, W.W. Moses, W.S. Choong, IEEE Trans. Nucl. Sci. 56(4), 2506 (2009)ADSCrossRefGoogle Scholar
  8. 8.
    S. Gridin, A. Vasil’ev, A. Belsky, N. Shiran, A. Gektin, Phys. Status Solidi (B) 251(5), 942 (2014)ADSCrossRefGoogle Scholar

Copyright information

© Institute of High Energy Physics, Chinese Academy of Sciences; China Nuclear Electronics and Nuclear Detection Society and Springer Nature Singapore Pte Ltd. 2018

Authors and Affiliations

  1. 1.State Key Laboratory of Particle Detection and ElectronicsInstitute of High Energy Physics, CASBeijingChina

Personalised recommendations