- Email: [email protected]
Development of Si microcalorimeters for a neutrino mass experiment
JLti!li!l __
Nuclear Instruments and Methods in Physics Research A 370 (1996) 244-246
NUCLEAR INSTRUMENTS 8 METHODS IN PHYSICS RESEARCH
.Bfi j-=
&I
SectIon 4
ELSEVIEK
Development
of Si microcalorimeters experiment
for a neutrino mass
A. Alessandrello”. C. Brofferio”, D.Y. Camin”, C. Cattadori”, R. Cavallini”, 0. Cremonesi”, E. Fiorini”, A. Giuliani”, A. Maglioneb, B. Margesinb, A. Nucciottia,*, S. Parmeggiano”, M. Pavan”, M. Perego”, G. Pessina”, G. Pignatel’, E. Previtali”, M. Sistia,‘, L. Zanotti” “Diparfimenfo di Fiskade/[’Univrrsirir hlRST. I-38050 ‘Dipartirnento
di h~egneria
dei Materiali.
di Milano,
Pmv
Unirvrsitci
I-200133 Miho,
Italy
(TN ). Italy di Trerzto. I-.?80_70 Mesiano
(TN). Ital\
Abstract We are developing high resolution Si-implanted thermistors for a calorimetric neutrino mass experiment. The production process is being tuned to reach high performance and reproducibility. We discuss the properties of devices prepared with different process parameters and different geometries. We also present the results obtained using these thermistors for detecting low energy X-rays.
1. Introduction For a neutrino mass experiment, high energy resolution and high statistics are two major requirements. We believe we can pursue both by utilising Si-implanted thermistors. The NASA/Wisconsin collaboration has shown that energy resolutions below IOeV FWHM at 6 keV are achievable [2]. The demand for high statistics near the end-point of the beta spectrum can be satisfied to some extent by using low Q beta emitters like “‘Re. Moreover, reproducibility of the Si-implanted thermistors would allow to run a multi-element experiment and therefore to partially overcome the limitation given by intrinsic slowness of thermal detectors.
2. Experimental Eighteen wafers had been implanted to define the best process. Table I gives a summary of the processes. The warmer post-implant anneal was introduced to reduce the transient enhanced diffusion phenomena. All implants are about 0.5 pm deep. On every wafer fifteen different geometries are reproduced 320 times by photolithography. The devices have length-to-width ratios ranging from 3: 1 to 1: 10 and sizes ranging from 30 X 30 pm’ to 300 X *Corresponding author. Tel. f39 2 2392695, 70609512, e-mail [email protected].
fax
+39
2
’ Present address: Max-Planck-Institut fur Physik. D-80805, Mtinchen, Germany. 0168-9002/96/$15.00 0 1996 Elsevier Science B.V. All rights reserved SSDI 0168.9002(95)01099-8
300km’. For characterisation, dice with several devices cut from the wafers are glued with silver epoxy on the gold plated bottom of a DIL 40 pins ceramic package. The low temperature device characterisation is divided in two steps. Between 4 and 1.2 K the package is mounted in a pumped ‘He cryostat and a computer controlled AC resistance bridge measures the resistance of the devices while cooling or warming. From 1.2 to 0.05 K the package is mounted in a dilution refrigerator. A computer controlled system steps the temperature and for every step collects a complete Z-V curve for each device. In both the systems the temperature is given by calibrated germanium thermometers mounted on the heat-sinking of the package and measured with AC resistance bridges. Some of the devices were tested also as a single particle detectors. A dice with a single thermistor is cut from the wafer and hung by aluminium wedge-bonded wires in a light-tight copper holder. First test detectors were instrumented with a room temperature Si JFET preamplifier (3 nV/dHz at 1 kHz). Better results were obtained with a preamplifier having the front-end Si JFETs cooled at 80 K and at a distance of about 10 cm from the detector. Two JFETs are connected in parallel and placed inside a box suspended between the mixing chamber of the refrigerator and the detector. Thermal decouplings between the JFETs (at 80 K) and the box (at 4 K) and between the box and the mixing chamber were obtained using tensioned Kevlar fibres: the electrical connections between the JFETs and the detectors are realised with tensioned 30 Pm manganin wires. This configuration permitted a lower microphonic noise and a smaller signal integration due to parasitic
A. Alessandrello
et al. I Nucl. hstr.
and Meth. in Phys. Rex. A 370 (1996) 244-246
245
Table I Thermistor static properties summary Wafer
P dose
B
T anneal.
[lO’X/cm’]
camp.
[“Cl
I 2 3 4 5 6 I 8
3.25 3.30 3.35 3.40 3.45 3.50 3.50 3.50
108 I58 20%
920 920 920 920 920 920 920 920
9
3.50
25%
920
IO
3.50
30%
11
3.25
1000
I2
3.30
1000
920
I3
3.35
1000
14
3.40
1000
&is9 Wsql
T,I [RI
Y fixed
R(T) fit T range
0.13 x 10’ 0.21 x IO’ 0.28 X IO’ 0.45 x IO1 1.4x 10’
139.4 94.6
0.5 0.5
I.4 x 10’ 1.3 x 10’ 1.5 x IO’
[W/m’/K”]
14.3
0.5
45.3 6.1
0.5 0.5
IK
10.6
0.5
OZK
14.7
0.5
O.?K
16.6
0.5
0.2K
0.22 x IO1
102.5
05
lK
0.40 x 10’
59.2
0.5
IK
15
3.45
1000
0.77 x IO’
24.9
0.5
I K
16
3.50
10%
1000
1.6 X IO’
4.7
0.5
I7 I8
3.50 3.50
20% 30%
1000 1000
I.4 x 10’
8.6
0.5
O.ZK
capacitance. The JFETs are in source follower configuration with gain close to unity and the total drain current at the operating point is 120 p,A with a power dissipation of about 0.25 mW. The measured noise with shorted inputs is about 5 nV/dHz at I kHz.
3. Thermistor
static properties
The reproducibility of the devices within the same wafer depends on their size. The 300 X 300 pm” devices reproduce very precisely the same p(T) curve. Considering all geometries the experimental resistivity at a fixed temperature differs at most 40%. mainly because of geometrical uncertainties. In the following only the larger devices will be considered. The static analysis of the thermistors is different for the two temperature ranges where data were collected. Also the compensated and the uncompensated thermistors show completely different behaviour. In Fig. I the R(T) curves of some compensated samples are shown. In the high temperature range the R(T) data were simply fitted to the law R(T) = R, exp(T,,lT)Y [I]. The parameter y was both left free or fixed to 0.5 and 0.25. In this temperature range the compensated devices have an average y around 0.35. Between I and 0.15 K they have R(T) curves with y of 0.5. For a temperature lower than about 0.15 K the R-T curves flatten. The collected set of I-V curves allows a more detailed analysis below 1 K. Unfortunately not for all wafers complete data are available, only for four of them it was possible to fit the data to the so called “hot-electron” model [2]. Good fits to the model were obtained only taking the data between I and about 0.15 K. The parameters of the internal thermal
a
RI,
2.6 X 10’ K
2.3 x IO’
5.6 5.6
K
2.6 X 10’ 2.3 X IO’
5.2 5.3
conductance given by the fit to the “hot-electron” model are reported in Table I, g,, and rw are the parameters of the electron-phonon thermal conductance described by the law Pe = g,,(Tz - rl;). These data are very similar to those quoted in Ref. [2]. The flattening below 0.15 K is not explainable neither with the presence of a background power nor with other more sophisticated models (i.e. introducing a “field-effect” or other internal thermal conductances). The analysis of the data shows also that below 0.15 K the electrons of the thermistors seem to become more decoupled from the thermal bath than what is predicted by the parameters given by the “hot-electron” model fit. Studying the data available for different geometries and sizes we concluded that the thermistors behave as if there were resistive shunts along their edges. This effect could be due to an undercompensation of the samples at
0.6
1.0
1.4
1.8
2.2
Temperature-112
2.6
3.0
3.4
3.8
[K-‘/2]
Fig. 1. R(T) curves of 1 : 1 devices from waters 8, 9, 16 and 17.
VI. APPLICATIONS
246
A. Alessundrello
et al.
I Nucl. Instr.
and Meth.
the borders 121, but the uncompensated sample seems to contradict this hypothesis. Uncompensated sample R(T) curves can be fitted with y equal to 0.5 over the whole temperature range but with two different T,,, below I K the R(T) curves become less steep and T,, is about two orders of magnitude smaller. Also the internal thermal conductance is about two orders of magnitude smaller than the compensated sample. Geometrical considerations again suggest a problem of resistive shunts, but for these samples it is more dramatic than in the compensated ones and it could have a different origin.
4. Particle detection Several detectors were realised with 300 X 300 km’ square thermistors from wafer 6. 8, 9, 16. 17 and IS. The best thermal link was obtained with four pure aluminium bonding wires (2 mm X 027 km). The bottom of the Si dice was exposed to a weak “Fe source without collimator. Because of the direct absorption by the silicon dice the line width is expected to be limited by statistical fluctuations to about 200 eV FWHM. Therefore the analysis is restricted to the signal-to-noise (SIN) ratio. The best SIN ratio was reached with a thermistor of the wafer 17 (Fig. 2). Biased with 3 mV it worked at about 0.135 K with a resistance of 4 MR and a thermal coupling to the bath of about 7 X lo-” W/K. After optimal filtering the baseline width was about 130 eV FWHM. The two lines in the energy spectrum have pronounced high energy tails and
80
60 2 : s
40
in Phys. Rrs. A 370 (1996)
244-246
have a FWHM of 235 eV The pulse height for a 6 keV pulse is about 50 ~.LVat the preamplifier input. Three sources contribute to the observed noise: (1) the cold preamplifier noise, (2) the Johnson noise of the thermistor (5.3 nV/gHz at 0.13 K). and (3) some unavoidable microphonic noise at low frequency, probably due to the bonding wires holding the thermistor. On a frequency band of 4.6 kHz the total RMS noise at the preamplifier input is 0.8 PV. Two main reasons prevent from getting better results. (1) As a consequence of the problems described in the previous section, the temperature sensitivity of the thermistors drops to zero below 0.15 K. The drop is even more steep when the thermistors are biased, because of the anomalous decoupling of the electrons. At the usual working temperature the temperature sensitivity is reduced by a factor of about 3 with respect to the extrapolation of the “hot-electron” model fits. (2) Both from the pulse height and from the pulse decay time constant it appears that the heat capacity is about three times larger than expected. This fact could be due to an excess heat capacity in the aluminium bonding wires.
5. Conclusions To fully exploit the power of Si implanted thermistors and to be able to apply them to a neutrino mass experiment, better devices must be prepared. Thermistors without the problems we observed below 0.15 K would be capable of a FWHM resolution of about 20 eV The fabrication of a new batch of thermistors has been completed in July 1995. The new devices have been prepared to test the possibility of reducing the resistive shunts along the border, for example by making the implantation of B wider than the P one. Preliminary measurements on some of the new samples do not show any anomalous electron decoupling down to at least 80 mK. As soon as the thermistors will allow resolutions below 50 eV tests with absorbers (Sn, Re, .) will be carried out.
20
References 0.0
Energy Fig. 2. “Fe X-ray source thermistor from wafer.
10
6
2
energy
[keVl spectrum
collected
with
a
[I] B.I. Shlovskii and A.L. Efros, Electronic Properties of Doped Semiconductors (Springer, Berlin. 1984). [2] J. Zhang, Ph.D. thesis; _I.Zhang et al.. Phys. Rev. B4 48 (1993) 2312.
Nuclear Instruments and Methods in Physics Research A 370 (1996) 244-246
NUCLEAR INSTRUMENTS 8 METHODS IN PHYSICS RESEARCH
.Bfi j-=
&I
SectIon 4
ELSEVIEK
Development
of Si microcalorimeters experiment
for a neutrino mass
A. Alessandrello”. C. Brofferio”, D.Y. Camin”, C. Cattadori”, R. Cavallini”, 0. Cremonesi”, E. Fiorini”, A. Giuliani”, A. Maglioneb, B. Margesinb, A. Nucciottia,*, S. Parmeggiano”, M. Pavan”, M. Perego”, G. Pessina”, G. Pignatel’, E. Previtali”, M. Sistia,‘, L. Zanotti” “Diparfimenfo di Fiskade/[’Univrrsirir hlRST. I-38050 ‘Dipartirnento
di h~egneria
dei Materiali.
di Milano,
Pmv
Unirvrsitci
I-200133 Miho,
Italy
(TN ). Italy di Trerzto. I-.?80_70 Mesiano
(TN). Ital\
Abstract We are developing high resolution Si-implanted thermistors for a calorimetric neutrino mass experiment. The production process is being tuned to reach high performance and reproducibility. We discuss the properties of devices prepared with different process parameters and different geometries. We also present the results obtained using these thermistors for detecting low energy X-rays.
1. Introduction For a neutrino mass experiment, high energy resolution and high statistics are two major requirements. We believe we can pursue both by utilising Si-implanted thermistors. The NASA/Wisconsin collaboration has shown that energy resolutions below IOeV FWHM at 6 keV are achievable [2]. The demand for high statistics near the end-point of the beta spectrum can be satisfied to some extent by using low Q beta emitters like “‘Re. Moreover, reproducibility of the Si-implanted thermistors would allow to run a multi-element experiment and therefore to partially overcome the limitation given by intrinsic slowness of thermal detectors.
2. Experimental Eighteen wafers had been implanted to define the best process. Table I gives a summary of the processes. The warmer post-implant anneal was introduced to reduce the transient enhanced diffusion phenomena. All implants are about 0.5 pm deep. On every wafer fifteen different geometries are reproduced 320 times by photolithography. The devices have length-to-width ratios ranging from 3: 1 to 1: 10 and sizes ranging from 30 X 30 pm’ to 300 X *Corresponding author. Tel. f39 2 2392695, 70609512, e-mail [email protected].
fax
+39
2
’ Present address: Max-Planck-Institut fur Physik. D-80805, Mtinchen, Germany. 0168-9002/96/$15.00 0 1996 Elsevier Science B.V. All rights reserved SSDI 0168.9002(95)01099-8
300km’. For characterisation, dice with several devices cut from the wafers are glued with silver epoxy on the gold plated bottom of a DIL 40 pins ceramic package. The low temperature device characterisation is divided in two steps. Between 4 and 1.2 K the package is mounted in a pumped ‘He cryostat and a computer controlled AC resistance bridge measures the resistance of the devices while cooling or warming. From 1.2 to 0.05 K the package is mounted in a dilution refrigerator. A computer controlled system steps the temperature and for every step collects a complete Z-V curve for each device. In both the systems the temperature is given by calibrated germanium thermometers mounted on the heat-sinking of the package and measured with AC resistance bridges. Some of the devices were tested also as a single particle detectors. A dice with a single thermistor is cut from the wafer and hung by aluminium wedge-bonded wires in a light-tight copper holder. First test detectors were instrumented with a room temperature Si JFET preamplifier (3 nV/dHz at 1 kHz). Better results were obtained with a preamplifier having the front-end Si JFETs cooled at 80 K and at a distance of about 10 cm from the detector. Two JFETs are connected in parallel and placed inside a box suspended between the mixing chamber of the refrigerator and the detector. Thermal decouplings between the JFETs (at 80 K) and the box (at 4 K) and between the box and the mixing chamber were obtained using tensioned Kevlar fibres: the electrical connections between the JFETs and the detectors are realised with tensioned 30 Pm manganin wires. This configuration permitted a lower microphonic noise and a smaller signal integration due to parasitic
A. Alessandrello
et al. I Nucl. hstr.
and Meth. in Phys. Rex. A 370 (1996) 244-246
245
Table I Thermistor static properties summary Wafer
P dose
B
T anneal.
[lO’X/cm’]
camp.
[“Cl
I 2 3 4 5 6 I 8
3.25 3.30 3.35 3.40 3.45 3.50 3.50 3.50
108 I58 20%
920 920 920 920 920 920 920 920
9
3.50
25%
920
IO
3.50
30%
11
3.25
1000
I2
3.30
1000
920
I3
3.35
1000
14
3.40
1000
&is9 Wsql
T,I [RI
Y fixed
R(T) fit T range
0.13 x 10’ 0.21 x IO’ 0.28 X IO’ 0.45 x IO1 1.4x 10’
139.4 94.6
0.5 0.5
I.4 x 10’ 1.3 x 10’ 1.5 x IO’
[W/m’/K”]
14.3
0.5
45.3 6.1
0.5 0.5
IK
10.6
0.5
OZK
14.7
0.5
O.?K
16.6
0.5
0.2K
0.22 x IO1
102.5
05
lK
0.40 x 10’
59.2
0.5
IK
15
3.45
1000
0.77 x IO’
24.9
0.5
I K
16
3.50
10%
1000
1.6 X IO’
4.7
0.5
I7 I8
3.50 3.50
20% 30%
1000 1000
I.4 x 10’
8.6
0.5
O.ZK
capacitance. The JFETs are in source follower configuration with gain close to unity and the total drain current at the operating point is 120 p,A with a power dissipation of about 0.25 mW. The measured noise with shorted inputs is about 5 nV/dHz at I kHz.
3. Thermistor
static properties
The reproducibility of the devices within the same wafer depends on their size. The 300 X 300 pm” devices reproduce very precisely the same p(T) curve. Considering all geometries the experimental resistivity at a fixed temperature differs at most 40%. mainly because of geometrical uncertainties. In the following only the larger devices will be considered. The static analysis of the thermistors is different for the two temperature ranges where data were collected. Also the compensated and the uncompensated thermistors show completely different behaviour. In Fig. I the R(T) curves of some compensated samples are shown. In the high temperature range the R(T) data were simply fitted to the law R(T) = R, exp(T,,lT)Y [I]. The parameter y was both left free or fixed to 0.5 and 0.25. In this temperature range the compensated devices have an average y around 0.35. Between I and 0.15 K they have R(T) curves with y of 0.5. For a temperature lower than about 0.15 K the R-T curves flatten. The collected set of I-V curves allows a more detailed analysis below 1 K. Unfortunately not for all wafers complete data are available, only for four of them it was possible to fit the data to the so called “hot-electron” model [2]. Good fits to the model were obtained only taking the data between I and about 0.15 K. The parameters of the internal thermal
a
RI,
2.6 X 10’ K
2.3 x IO’
5.6 5.6
K
2.6 X 10’ 2.3 X IO’
5.2 5.3
conductance given by the fit to the “hot-electron” model are reported in Table I, g,, and rw are the parameters of the electron-phonon thermal conductance described by the law Pe = g,,(Tz - rl;). These data are very similar to those quoted in Ref. [2]. The flattening below 0.15 K is not explainable neither with the presence of a background power nor with other more sophisticated models (i.e. introducing a “field-effect” or other internal thermal conductances). The analysis of the data shows also that below 0.15 K the electrons of the thermistors seem to become more decoupled from the thermal bath than what is predicted by the parameters given by the “hot-electron” model fit. Studying the data available for different geometries and sizes we concluded that the thermistors behave as if there were resistive shunts along their edges. This effect could be due to an undercompensation of the samples at
0.6
1.0
1.4
1.8
2.2
Temperature-112
2.6
3.0
3.4
3.8
[K-‘/2]
Fig. 1. R(T) curves of 1 : 1 devices from waters 8, 9, 16 and 17.
VI. APPLICATIONS
246
A. Alessundrello
et al.
I Nucl. Instr.
and Meth.
the borders 121, but the uncompensated sample seems to contradict this hypothesis. Uncompensated sample R(T) curves can be fitted with y equal to 0.5 over the whole temperature range but with two different T,,, below I K the R(T) curves become less steep and T,, is about two orders of magnitude smaller. Also the internal thermal conductance is about two orders of magnitude smaller than the compensated sample. Geometrical considerations again suggest a problem of resistive shunts, but for these samples it is more dramatic than in the compensated ones and it could have a different origin.
4. Particle detection Several detectors were realised with 300 X 300 km’ square thermistors from wafer 6. 8, 9, 16. 17 and IS. The best thermal link was obtained with four pure aluminium bonding wires (2 mm X 027 km). The bottom of the Si dice was exposed to a weak “Fe source without collimator. Because of the direct absorption by the silicon dice the line width is expected to be limited by statistical fluctuations to about 200 eV FWHM. Therefore the analysis is restricted to the signal-to-noise (SIN) ratio. The best SIN ratio was reached with a thermistor of the wafer 17 (Fig. 2). Biased with 3 mV it worked at about 0.135 K with a resistance of 4 MR and a thermal coupling to the bath of about 7 X lo-” W/K. After optimal filtering the baseline width was about 130 eV FWHM. The two lines in the energy spectrum have pronounced high energy tails and
80
60 2 : s
40
in Phys. Rrs. A 370 (1996)
244-246
have a FWHM of 235 eV The pulse height for a 6 keV pulse is about 50 ~.LVat the preamplifier input. Three sources contribute to the observed noise: (1) the cold preamplifier noise, (2) the Johnson noise of the thermistor (5.3 nV/gHz at 0.13 K). and (3) some unavoidable microphonic noise at low frequency, probably due to the bonding wires holding the thermistor. On a frequency band of 4.6 kHz the total RMS noise at the preamplifier input is 0.8 PV. Two main reasons prevent from getting better results. (1) As a consequence of the problems described in the previous section, the temperature sensitivity of the thermistors drops to zero below 0.15 K. The drop is even more steep when the thermistors are biased, because of the anomalous decoupling of the electrons. At the usual working temperature the temperature sensitivity is reduced by a factor of about 3 with respect to the extrapolation of the “hot-electron” model fits. (2) Both from the pulse height and from the pulse decay time constant it appears that the heat capacity is about three times larger than expected. This fact could be due to an excess heat capacity in the aluminium bonding wires.
5. Conclusions To fully exploit the power of Si implanted thermistors and to be able to apply them to a neutrino mass experiment, better devices must be prepared. Thermistors without the problems we observed below 0.15 K would be capable of a FWHM resolution of about 20 eV The fabrication of a new batch of thermistors has been completed in July 1995. The new devices have been prepared to test the possibility of reducing the resistive shunts along the border, for example by making the implantation of B wider than the P one. Preliminary measurements on some of the new samples do not show any anomalous electron decoupling down to at least 80 mK. As soon as the thermistors will allow resolutions below 50 eV tests with absorbers (Sn, Re, .) will be carried out.
20
References 0.0
Energy Fig. 2. “Fe X-ray source thermistor from wafer.
10
6
2
energy
[keVl spectrum
collected
with
a
[I] B.I. Shlovskii and A.L. Efros, Electronic Properties of Doped Semiconductors (Springer, Berlin. 1984). [2] J. Zhang, Ph.D. thesis; _I.Zhang et al.. Phys. Rev. B4 48 (1993) 2312.