Developments in superconducting tunnel junction detectors

Nuclear Instruments and Methods in Physics Research A314 (1992) 252-262 North41olland

NUCLEAR INSTRUMENTS &METHODS IN PHYSICS RESEARCH Section A

evelo en s in superconducting tunnel junction detectors . Kltrdktldo

."1duanred Material d Technology Research Laboratories, Nipport Steel Corporation, 1618 Ida, Nakahara-ku, Kawasaki-shi, KanagaN'a 211 . Japan

Superconducting tunnel junctions (STJs) have a large possibility of being utilized as radiation detectors of ultrahigh energy resolution . In fact, Sn/SnO. , /Sn junctions as X-ray detectors have already shown about three times higher energy resolution than Si detectors . Recently, in addition to Sri junctions, studies of new-type STJ detectors are in progress . Research into STJ detectors, including our recent studies on Nb/AI-A10, /Nb junctions with a single crystal Nb film and on series-connected STJ detectors, are briefly reviewed .

1. Introduction

The energy resolution of a radiation detector is essentially limited by the statistical fluctuation (AN) of the number (N) of electrons excited by a radiation of energy (E). The statistical limit of energy resolution AE is given as AE/E = AN/(N) a (E/E) t12 , where ( N ) is the mean value of N, and E ( = E ; ( N )) is the mean energy required to excite one electron in the

detector . Thus, a smaller E can result in a higher energy resolution . For the case of semiconductor detectors, it is known that phonons emitted from the electrons excited by radiation do not excite other electrons beyond the energy gap . For the case of superconductor detectors, 11(t)t) (to D : Debye frequency) is usually much larger than é°,'4 (= 23) so that phonons can break Cooper pairs and excite electrons (quasiparticles) . According to a numerical simulation of the cascade excitation process of quasiparticles and phonons in bulk Sn at 0 K, E is 1 .73 (= 1 meV) which is three orders smaller than the case of semiconductors . Therefore, ultrahigh

resolution detectors become feasible with superconducting tunnel junctions (STJs) if the excess quasiparticles efficiently pass through the tunnel barrier and contribute to a signal .

2.

eview of earlier studies In 1969, G .H . Wood and B .L . White reported that

a particles were detected with a superconducting tun-

nel junction (STJ) [1,2] . They used a crossed-film-type Sn/SnO . ,/Sn junction cooled to 1 .2 K . The do Josephson current was suppressed by applying a magnetic f : Id parallel to the junction . The STJ was obliquely 0168-9002/92/S05 .00

irradiated with 5 .1 MeV a particles, of which the energy loss in the junction was 140-500 keV . Regarding the mean energy E required to excite one quasiparticle in the detector, they estimated E <_ 8.2 meV by taking into account the quasiparticles which might have recombined before passing through the tunnel barrier .

The energy Ecff per quasiparticle which actually tunneled was estimated as E,, ff < 0.145 cV. They suggested that the STJs can be utilized as high energy resolution detectors because E for STJs was much smaller than that of semiconductor detectors . No clear peak corre-

sponding to the deposited energy could be observed in the pulse-height spectrum of the signals . They concluded that the actual limit of energy resolution was probably determined by the stochastic variations in

insulator thickness across the junction overlap area. The mechanism of signal appearance was considered as follows . A superconductor is brought into a nonequilibrium state immediately after the energy de-

position, but the superconductor recovers to an equilibrium state within a period of - 10 -" s . The temperature of the equilibrium state is higher than that before the incidence of radiation and it gradually decreases . Consequently the change of tunnel current following the temperature rise and fall causes the signal . After the study of Wood and White the next step in research on the STJ detectors was performed at Kyoto

University from 1977 to 1982 [3-101 . The Sn/SnO,,/Sn junctions were also adopted as detectors . First, a junction was perpendicularly irradiated with 5 .3 MeV a particles ; the energy loss in the junction was 100-140 keV . Typical pulse height spectra are shown in fig . 1 ; the peaks in the spectra are as broad as those obtained by Wood and White . From an investigation of the

temperature and the bias current dependence of the height of signals, it was suggested that the excess

1992 - Elsevier Science Publishers 13 .V . All rights reserved

M. Kurakado / Superconducting tunnel junction detectors 10 2

zz O

10'

W

10P

11

cn

E-

I

0 .5pA 1 .37K

1 .0 ~, A

1 .37K

e o

-

é

D 0 0

I .O ~LA_ 1 .88K

o

A small gap crergy Ep, of a detector is necessary to obtain a small E . A small E., however, does not always guarantee a small value of E . For example, an empirical relation between E and E. of semiconductor detectors is expressed as [I I ]

?.

-~

E = 2 .8 Et, + (0 .5-1 .0) eV .

O d

00

10-31

o é.o

20

.

.

--I

.

50

100

20

CHANNEL

50

100

20

50

NUMBER

Fig . 1 . Typical pulse-height spectra obtained with a Sn junction . Solid circles give the spectrum with a-particle irradiation and open circles are without irradiation [3] .

quasiparticles, i .e ., electron-like ones in a superconducting layer and also hole-like ones in the other layer, excited by a radiation are essential to the signal [3]. Secondly, by irradiating an STJ fabricated on a glass substrate of low heat conductivity with pulsed light of

300 ns duration time from a semiconductor laser, the pulse shape of the signals at about 1 .4 K was studied using a wide band amplifier of a low input impedance . The spectra A and B in fig . 2 were obtained in different pressures of heat-exchange He gas . When the thermal contact between the STJ and its environment was strong, in case A, the signal decayed with a short time constant ( - 180 ns) . On the other hand, when the

thermal contact was weak, in case B, the signal had two decay-time constants ( - 350 ns, - 3500 ns). The two short decay-time constants ( - 180 ns and - 350 ns) correspond to the energy relaxation from the electron system to the phonon system in the STJ, while the longer decay component (- 3500 ns) corresponds to the energy dissipation from the STJ to its environment [4] . This experiment made it more clear that the excess quasiparticles in a nonequilibrium state are essential to the signal appearance from the STJ detectors .

If one applies this equation to the cases of superconductors, E of the superconductors is 0 .5-1 .0 cV. According to the theories of noncqüii®briüiii supcr`vi®dü`tivity [12], E and the Fano factor F of superconducting

Sn were estimated in our previous works [6,8] . Quasiparticles excited by an incident radiation mostly have energies greater than several eV . The dominant interaction of the quasiparticles having cnergy greater than ßi'51 a (4cv n E( )' 1' - , of the order of 100 meV] is the Cooper pair breaking, i .e . the excita-

tion of other quasiparticles, where E F is the Fermi energy and ho), (of the order of 10 meV) is the Debyc energy . Quasiparticles with energy of 3 <_ E < N5 expend their energy mainly in phonon excitation, where 2A is the gap energy defined as the minimum energy necessary to excite two quasiparticles by destroying a Cooper pair . Therefore, if the phonons do not excite quasiparticles, as is the case with semiconductor detectors, E may be nearly equal to hw even when the gap energy E g is much smaller than 4ci . This probably explains the E g independent part in eq . (1) for the

semiconductors detectors . Since the gap energy of a superconductor is usually much smaller than ftcw ® , the phonons of energy f2 >_ 2A, which are emitted by higher energy quasiparticles, can produce other quasiparticles by breaking Cooper pairs (fig. 3). The resulting quasiparticles can also emit phonons . A Monte Carlo simulation of the cascade

excitation process between phonons and quasiparticles was performed to evaluate the ultimate limitation of energy resolution of the STJ detectors. The E value and the Fano factor F obtained for bulk Sn at 0 K were

E= -E/(N>=1 .68 10

=0 .969meV,

F--- ((N- 1(1\7> =0 .2,

s 6

o

®m o

®

o o

° "°

2 91

0 0

500

1000

1500

2000

2500

(nsec) Fig . 2 . Time spectra of laser-induced signals under different He gas pressures. (r,) - 0.1 Torr : (13) < 10 - ' Torr [4]. TIME

(2) (-1 )

where N is the number of quasiparticles excited by a radiation of energy E, and ( N > is the mean value of N . The result of the Monte Carlo simulation is shown in fig . 4 . These E and F mean that the limitation of energy resolution due to the statistical fluctuation of N is one or two orders smaller than those of semiconductor detectors, e .g ., FWHM = 2 .5 eV for 5 .9 kcV radia-

"

w 4

a ô

253

tion in the STJ [8] . The E value obtained by the computer simulation (nonequilibrium model) is completely different from that expected from an equilibrium model . If quasiparticles are excited by the temperature increase due to the 11 . SOURCES/I3ETECTORS

M. Kumkado / Sulvwonducting tunnel junction detectors

254

S2 [mev ]

E [meV[

rn -

COOPER-PAIR BREAKING (C. P. B .) C6(oWEr) I I` = 300 meV

E

PHONON EMISSION (P. E.)

0.5 0

a2(Q)F(Q)

G _

(C. P. B .)

0 .970 0 .965

110D = 17 .2 meV

-W D - A = 16 .6 meV

0 .975

m

Fig . 5 .

E

v

I

m

1~~a~ImoImnImW n

I

m

n

I

w

n

I

m

n

I

m

n

I

m

as a function of phonon injection energy 12 (21 :5 ,f1 _< ft ce n) [8] .

2A '=1 .15 meV 0 COOPER PAIRS QUASIPARTICLES

PHONONS

Fig . 3 . Main excitation processes of excess quasiparticles and phonons initiated by a radiation in Sn [6]. E is the energy of a quasiparticle and dl is the energy of a phonon . a 2 W) is an effective electron-phonon coupling function and F(D) is the density of state fer phonons . incidence of a radiation, the energy E of the radiation will be shared between the electron system and the phonon system according to their specific heats Ce and C p . Assuming that the energy of quasiparticles is 4, we obtain N=E[C e /(Cc +C P )]/3

80

=1 .68 (=0.969 meV )

6

70 60

< N) = 5932 .4

50 v

40

"

® 30

F

=0 .195

°ée4 *

20 0 5810

5900

6000

N

Fig. 4. A spectrum of N for 3000 radiations of energy 5 .75 eV ( = 10000J). The spectrum is obtained by a Monte Carlo method [8] .

and then E _ [ (Ce +C p )/Cc IA .

(5)

The specific heat Ce of a superconductor decreases exponentially ( a exp( -A/k B T)) with decreasing temperature because of the energy gap, while C p is proportional to T ; . Hence the temperature dependence of E is considerably large at low temperatures (T/Tc << 1) .

Furthermore, E becomes infinitely large at lower temperatures (T/Tc = 0) because Cc /Cp = 0 at such low temperatures . On the contrary, eq . (2) means that about 60% of the incident energy is assigned to the electron system immediately after the cascade excitation process of quasiparticles, in spite of its small Cc

value at 0 K. Fig . 5 shows the E value obtained numerically as a function of the injection energy of phonons . The phonons of energy D ? 24 can produce quasiparticles as efficiently as a higher energy radiation . From this

result, it was pointed out that the STJ detectors may be highly sensitive not only to ionizing events but also to nonionizing events [8,10] . The prior condition of the calculations was that the density of excess quasiparticles is too small to promote their recombination during the cascade process [6,8]. It was pointed out that the understanding of the recombination of the excess quasiparticles in the highly-local-

ized highly-dense excitation state at the initial energydeposition spot might be of importance in actual cases [10] . If the stochastic variation in the thickness of the tunnel barrier had been the main origin of the broad-

ness of the energy spectra obtained in the previous works [1,3], it might be extremely difficult to fabricate a high energy resolution detector with an STJ . On the basis of the studies mentioned above [3-8], an experi-

M. Kurakado / Superconducting tunnel junction detectors

40

a particles 200 X 300pm 2 X 3000 A 10000S (a)

30 20

diffusion of excess quasiparticles away from the junction are sufficiently depressed . Independently, Barone et al . demonstrated that an island type Nb/NbO A/Pb junction with narrow leads can achieve a higher energy resolution than a conventional crossed-film-type one [131.

with collimator dia . 304m

10 0

3. Review of recent studies

o

Pulser

0 16L0 () 1400

200X 300,gm 2 X3000 A

1200- 100s 1000800 600 400 0

20

40

60

25 5

60

100

120

140

CHANNEL NUMBER Fig. 6. Pulse-height spectra obtained with a Sn junction . (a) With a-particle irradiation at the central part of junction through a collimator ; (b) a pulse-height spectrum of signals induced by a pulser [9].

mental study to achieve a high energy resolution with STJs was carried out [9,101. To attain a higher energy resolution than those in the previous experiments, the following two subjects are considered to be of fundamental importance : (1) measurements at lower temperatures than before and (2) suppression of the degradation of energy resolution due to the diffusion of excess quasiparticles from the junction to its leads. STJs were cooled down to 0.32 K using a 3He cryostat, instead of a 4He cryostat, in order to decrease the density of thermally excited quasiparticles. Although the current through junctions hardly decreased with decreasing temperature below about 0.7 K probably because of leakage current, the signal-to-noise ratio was much improved. The peak in the spectrum obtained without a collimator was broad. When a hole of a collimator was set between the center and the edge of a junction, a sharp spectrum having a tail on the lower energy side was obtained. On the contrary, the sharp pulse-height spectrum of fig . 6a was obtained by irradiating a particles only to the central part of a junction, making use of a collimator. The width of the peak was almost determined only by the broadening due to noise (fig. 6b) . This study proved that the main origin of the broadness of the pulse-height spectra obtained before was not the stochastic variation in the thickness of the tunnel barrier . It also showed that the energy resolution of STJ detectors can be greatly improved if the leakage current of the junction and the

After the researches mentioned above [1-10,131, many sophisticated studies have been carried out [141. In 1986, using Sn/SnO .,/Sn junctions with small leakage current, Kraus et al. [151 of Technische Universit5t München (TUM) and Twerenbold [161 of former SIN (presently Paul Scherrer Institute : PSI) succeeded in detection of Mn Ka and Kß X-rays from "Fe. The junctions were operated at temperatures around 0.3 K. Kraus et al. obtained an energy resolution of 250 eV for 5.9 keV X-rays with a junction that has an area of 0.088 mm= and two electric leads with widths of 0.35 mm and 0.28 mm, respectively . The effective E value (E,,ff --- E/(Q/e)) was 20J, where Q is the actually collected signal charge and e is the charge of an electron [151. Later, the group of TUM achieved a higher energy resolution of 88 eV with an Sn junction of which the electric leads were made of Pb. The energy gap of Pb is larger than that of the junction material Sn. Thus, the difference of the energy gaps suppresses the diffusion of quasiparticles from the junction to the electric leads [171. Twerenbold detected the 5 .9 keV X-rays with an energy resolution of 90 eV and obtained E,ff - 4.2:1 2.4 meV. The junction area was 100 X 100 Rm2 and the width of the leads was 20 p,rn [161 . The PSI group achieved an energy resolution of 48 eV at 5.9 keV, which is about three times higher energy resolution than that of high energy resolution semiconductor detectors (= 150 eV). The high resolution was obtained with an Sn junction having 4 pLrn-wide electric leads [181. The thickness of the tunnel barrier of STJs is only a few nanometers, and therefore the resulting large electric capacitance (typically 10 WF/cm - ) makes it difficuït to utilize large-area Siis as high resolution detectors. Since the sensitive area and volume of the STJs that showed high resolution were small, i.e., of the order of 100 X 100 p,m 2 X 100 nm, their detection efficiencies were too low for many practical uses. Three kinds of STJ detectors were proposed up to now in order to overcome the drawback of the usual STJ detectors : (i) quasiparticle-trapping detectors [19,201; (ü.) acoustic detectors composed of a single-crystal radiation absorber and STJ phonon sensors [21,221 ; and (iii) series-connected STJ detectors [23,241 . 11. SOURCES/ DETECTORS

256

l1-1. Kurakado / Supe rconducting

Booth of the University of Oxford proposed a quasiparticle-trapping-detector [191 . Tile original motivation was to measure the energy spectrum of solar neutrinos via P-decay reaction oil 115 111, v, + "'In , "'Sri-' + e [25]. The detector consists of a large superconductor with an energy gap ,1, e .g ., single-crystal In, and a small STJ with a smaller energy gap A,, e.g ., AI STJ . The excess quasiparticles produced in the large superconductor diffuse into the STJ and will be trapped III' :'C IW 8tiàe the gap energy -A, is smaller than ,I, . This trapping results in highly efficient tunneling of the excess quasiparticles excited by a radiation in the large superconductor. The TUM group has shown that excess quasiparticles excited in a central part of a strip of a till film (3 s,, -- 0.58 meV) between two Al STJs (J A1 = O. 1S mcV) give rise to signals from the STJs . The two STJs are attached to the strip and are set (1 .94 mm apart from each other. The sample was cooled to 60 mK . The energy resolution obtained for 5 .9 keV X-rays was 6(l eV and the position resolution was determined to be better than 5 ~Lnl over a sensitive length of 450 l.Lm [24] . The E as a function of phonon injection energy f1 is almost constant, 13A, for D > 4A (fig . 5) [S]. Electrons in a superconductor are efficiently excited by nonthermal phonons of energy Q >_ 2J . It may, therefore, be possible to pleasure the energies of radiations by means of a radiation absorber and STJs, namely as a radiation-to-phonon converter and as nonlthermal-phonon sensors, respectively . A group at Stanford University have proposed, for detection of darkmatter, a new type of ml4ticlc detector based on ballistic phonon generation in a large single crystal of Si or other low acoustic loss insulating crystals . The distributions of the phonons at the surfacl.1, of the large single crystal can be measured with a large number of STJs prepared on the surfaces [21] . The TUM group succeeded in the detection of nonthcrmal phonons produced by a particles at the rear surface of a silicon substrate on which Sri or Al STJ detectors are prepared [22] . However, the E, tt for the cv particles was about two orders larger than that for X-rays that directly excite quasiparticles in the junction .

4. Nb / Al-Al laver [26]

twincl junction detectors

possible ; for example, the mean free path of excess quasiparticles in superconducting layers should be so long that most of them can pass through the tunneling barrier before they recombine into Cooper pairs. Good resolutions have been achieved with Sn/SnO . ,/Sn junctions [9,15-17] probably because of their small leakage current and the high quality of Sn layers . However, it is well known that the Sn/SnO .,/Sn junctions are easily destroyed by thermal cycles . Since Nb/Ai-AlO . ,./ Nb junctions arc more r.sistant to therSri mal cycles than junctions, extensive studies of Nb/Al-AIO.,/Nb junctions as an X-ray detector have recently been carried out [27-30]. Single crystal Nb films can be grown epitaxially on (1102) sapphire [311 . The mean free path of excess quasiparticles in single crystals is longer than in polycrystals . However, the leakage current of the STJs with a bottom superconductor of single crystal had been larger than those of usual STJs with a polycrystalline bottom layer [32,33]. A large leakage current is fatal to the STJ as a radiation detector . Single-crystal Nb films of 200-600 nm thickness were prepared on (1102) sapphire substrates by means of ultrahigh vacuum evaporation, the so called "molecular beam epitaxy (MBE)" technique. A source Nb was evaporated by electron-beam heating. The reflection high energy electron diffraction (RHEED) patterns of the Nb films were streaky, which suggests that the surfaces of the films were flat on the atomic scale. One of the RHEED patterns is shown in fig. 7. For cleaning of the surfaces, the Nb films were etched by 3 nm by means of reverse sputtering. Then, a 20 nm-thick polycrystalline Nb film and Al-Al0_,/Nb layers were added to the single crystal films at room temperature by means of sputtering deposition . Nb/Al-AIO,/Nb junctions with a single crystal Nb bottom layer were fabricated from the multilayers, by means of photolithography processes. The subgap current of the STJs at 4 .2 K was almost the same as or smaller than those of usual polycrystalline junctions.

junction using a single-crystal Nb

STJs nnay be used as radiation detectors of ultrahigh energy resolution if excess quasiparticles excited by a radiation efficiently contribute to the signal . High quality of superconducting layers is required to receive signal charges from STJ as effectively as

Fig . 7. Reflection high-energy electron diffraction pattern of a single-crystal Nb film on a (11()2) sapphire substrate [26] .

M. Kurukudo / Superconduc1ing tunnel junction detectors

25 7

A normal polycrystalline STJ a and the STJs with the single-crystal bottom layer, b-e were used for the detection of X-rays . All layers of the polycrystalline junction a were prep fired by sputtering at room tempcraturc . Properties and E,, f,. of the samples are listed in table 1 . The X-ray source is ""Fe which emits Mn K

X-rays (5 .9 keV : 88%, 6 .5 keV : 12%) . The operating temperature of STJs was about 0 .4 K. The quality of the bottom Nb layers of samples a-e was evaluated in terms of the residual resistance ratio: RR° R(300 K)/R(10 K)) . Higher RRR means longer mean free path of quasiparticles . The signals from an STJ were fed to a charge-sensitive preamplifier . The energy resolutions were not good ; i .e ., - 1 keV. Because of

the smallness of the present junctions, an appreciable amount of energy absorbed diffuses from the junction to the signal leads, of which width is 5 p,m for samples a-c, e and 20 gm for sample d . The amount of diffused energy depends on the incident position in the junction [9,34] . The major portion of the broadness of the peaks is probably caused by the effect of energy diffusion . A high signal-to-noise ratio is essential for a high

energy resolution detector . The signal charge from sample e was the largest, and the leakage current of the sample was the smallest among all samples a-e ; the dynamic resistance dV/d I was 2 .6 k S2 at the bias point where the signals were measured . The amplitude

of noise of sample e was measured . The broadness of the spectrum due to the noises was measured by supplying pulser signals to the test input of the preamplifier, and the full width at half maximum (FWHM) of the pulser signal peak was only 30 eV . This is the lowest noise level attained with Nb/Al-AIO,/Nb STJs to date, and it is about 3 times smaller than those of high resolution Si semiconductor detectors .

The signal charge Q is larger for higher RRR samples, and the dependence of Q on RRR was exTable i Properties and E,,  of each sample [26]

Bottom Nb poly single single single Area [~Lm 50' 80 2] 20 , 50 2 Width of It-ads

5 5.0 2000 8 550

Ve

.:

Fig . h . Pulse shape of the preamplifier output signals arising from the sample e 126] .

tremely large . At present, it is '.eot assured that the dependence can be attributed only to the mean free path of excess quasiparticles . For example, the recombination lifetime of the excess quasiparticles in a Nb film at a very low temperature might be dominated by the film quality. the grain boundaries [181 or other defects in the film might be trapping or recombination

centers of the quasiparticles, especially in the case of Nb of which coherence length is short . The value of E,,, ( = E/(Q/e)) of sample e was about 6 .6 meV = 4 .33, which is the smallest value so far obtained with Nb,/Al-Al®, /Nb STJs and is about 11500 that of semiconductor Si detectors . The capability of high rate detection is important in many practical uses of radiation detectors . Fig . 8 shows the shape of the preamplifier output signals induced by X-rays in the sample e . The rise time, that is the charge collection time, of the STJ was about 200 ns . This indicates the high capability of STJs in the detection of high rate radiations .

5 . Series-connected STJ detector

Sample

4Lm] RN IM :' d V/ d l [111 n RRR ` E,  [meV]

5 SIdiv,

5 0.13 600 13 100

5 0.11 500 21 28

20 0.044 460 32 21

single ~0 ,

5 0.91 2600 33 -6 .6

'' Normal state resistance: . t' Dynamic resistance at the bias point where X'-rays were measured . ` RRR = R(300 K)/R(10 K).

Since the sensitive area of STJs used in the prcvious works was very small . i .c ., of the order of or less than 100 x 100 Rm 2 , their detection efficiencies were too low for practical uses. For further development, it is clearly preferable to make STJs with a much larger sensitive region . However, the low resistivity and the S2 cm 2 and 10 large electric capacitance (typically 10 1eF/cm , ) make it very difficult to utilize large super-

conductor detectors . One of the possible methods to increase the sensitive area of superconductor detectors is to use an asscmbly of STJs connected in series on one chip.

11 . SOURCES//DETECTORS

258

M. Kurakado / Superconducting tunneljunction detectors

which is hereafter called a series-connected STJ detector (SCSD) [23]. 5.1. Electric resistance and effective capacitance C, if of. SCSD Sine the leakage current due to defects in STJ is practically unavoidable, the resistance of an STJ is usually very low . Electric charges induced by radiations are so quickly released through a low resistance that output signals are much depressed [27]. The resistance of the series-connected STJ detector is greatly increased by the connection of STJ elements in series . This enhancement of the resistivity greatly improves its property as a detector, as seen later. The capacitance of any detector is desired to be as small as possible. From a simple model, here we estimate the height of voltage signals from the series-connected STJ detector. It is assumed that each STJ element has the same capacitance C, the leak current is sufficiently small and the resistivity is nearly infinite. The bias voltage V is applied to the detector through a large resistor Rg. When one of the STJ elements in the detector is irradiated, the excess quasiparticles excited by the radiation penetrate the tunnel barrier through the tunnel effect and consequently produce the charge Q in the element . In practical uses, the capacitance parallel to the detector C should be taken into account in the estimation, as shown by fig . 9. The parallel capacitance usually comes from she capacitance of the field effect transistor (FET) at the input stage of the preamplifier and from the capacitance of signal cables. The same charge Q'(= VC) as stored in the parallel capacitance C is induced in each STJ element, as seen in fig . 9. Since the voltage at the detector is equal to the voltage at C', V, is given by two equations : ` = (QI + Q2 + . . . +QJIC - 1=Q'/C (6)

= Q'/ C,

where n is the number of STJ elements in the detector and Qk (k = 1, 2, - - - , n) is the charge directly produced at k th element. Eliminating Q', we obtain V = Q/(C + nC' ),

(8)

where Q=Q I +Q2 + --- +Q,, . Note that the total capacitance of the present series-connected STJ detector is given by (C/n + C) in the usual meaning when a charge Q is added to the system from the outside of the system. However, eq. (8) defines an effective capacitance C,ff of the detector, which can be used in discussions on the signal-to-noise ratio of radiation detectors : C,fr = C+nc' = SC/n + n C , where S is the total junction area and C is the specific capacitance of the junctions . The C~ff is minimized when n is equal to (SC,,/C')' )'12, and then v/,.. (11) Ceff = 2(SC(IC') / (12) Cet f a S' . On the other hand, the capacitance of an STJ with an n times larger area or an assembly of n STJ elements connected in parallel, Char , is given by Cp :,r = nC + C = SC + C. The capacitance C is ordinarily 10-200 pF depending on the preamplifier employed and C is typically 6 ~tF/cm -' for a Nb/Al-AIO.,/Nb junction . Thus, eq. (11) indicates that the effective capacitance of the series-connected STJ detector can be much smaller than that of one STJ detector with the same sensitive area. Furthermore, the eqs . (10) and (11) hold for to parallel connections of n series-connected STJs. This gives a wide choice of the size of a junction, S/rnn, in the series-connected STJ detectors [35]. 5.2. Detection of nonthermal phonons

Fig. 9. Equivalent circuit for the present series-connected STJ detector [231 .

Electrons in the superconductive state of STJ are efficiently excited by nonthermal phonons with energy .iZ > 2a to]- As shown in fig . 5, the number of gUasipai'ticles excited by a phonon is approximately proportional to the energy 0. This high sensitivity to nonthermal phonons can be used to measure full energies of high energy radiations such as a particles. The seriesconnected STJ detector has been used as a sensor to detect nonthcrmal phonons created by irradiation of its rear substrate with nuclear radiations . In addition to ionizing events, such phonons are also created through nonionizing events, such as elastic scattering processes, which cannot be sensed by ordinary radiation detectors. Note that the series-connected STJ detector is

At. Kurakadu / Supcrconductirtg tunrtcl jcutctiott detectors Multi-STJ

21ct 4%4mm'

4 min I

Jill

5.3-MeV c1 ray from 110% Fig. 11 . Experimental setup for detecting cif (% particles with ihr series-connected STJ detector [23] .

b)

Fig. 1(l. Schematic drawings of the series-connected STJ detector ; (a) a plane figure, (b) the enlargement of a part of the STJ assembly, (c) the cross-sectional view of the STJ assembly. Notations in the figures : (A) sapphire substrate (lox 1(1 mm 2 ); (B) assembly of STJ elements connected in series (4 x 4 mm -'); (C) electrode ; (D) aluminum oxide barrier; (E) top layer (Nb) . (F) contact leads (Nb); (G) insulator (SiOO); (N) bottom layer (Nb) . (1) electrical lead [23] .

sensitive to the nonionizing events as well as the ionizing events. In this case, the series-connected STJ detector is similar to the Si crystal acoustic detector (SiCAD) 121,221, which is also a phonon sensor composed of many STJs. 1r;hch STJ in SiCAD is operated indcpendcrtly, whip :-11 STJs in the present detector work as one phonon sensor through their connection in series . 5.3. Experiment 5.3.1. Erperhnental setup An array of micro STJ elements in the detector is schematically shown in fig . 10a ; the sensitive area covered with the STJ elements is 4 x 4 mm -' while the

area of one STJ clement is 20 x 20 or 10() x 1(H) wm 2 . All STJ elements in the detector are connected in series, as shown in fig . 10b . Each STJ clement consists of Nb/AI-AIO,/Nb layers, which arc shown in fig . lOc. The experimental setup to examine the performance of the series-connected STJ detector is shown in fig . 11 . The rear side of the detector is irradiated by 5.3 MeV - "'Po a particles which are collimated by a 1 mm diameter hole in a 1 mm thick copper plate. The whole energy of a particles is absorbed at the rear surface of the sapphire substrate and is immediately transferred to the energy of phonons. Signals from the detector cooled at 0.4 K in a cryostat are fed to a preamplifier mounted outside the cryostat. 5.3.2. Sample's prepared Three samples (A, B, and C) of the series-connected STJ detector were prepared . The bottom Nb layers of all samples were epitaxially formed on sapphire substrates as mentioned in section 2 [261. This may be preferable for sensing phonons coming from the substrate as well as for collection of excess quasiparticles excited by the phonons . Some properties of each sample are listed in table 2, i .e., the size of an STJ . the number of STJs n, the total area of STJs S. resistivity of an STJ in the normal state p, the dynamic resistance of sample R,, and the effective capacitance of the sample C,,,. The dynamic resistance R,, is Table 2 Properties of each sample [23] Sam- Size of pie an STJ [ p,m 2 ] A B C

t)  Number Total [il cm=] of STJs area of et junctions

20 x 20 8000 100 X 100 960 1(1(1 x 100 960

R,j C", [ktt] [~LFJ

[mm`] 3 .2 9.6 9 .6

- 5 x lt) -- " 100 5x1(1 -" 2 - -:; x M " 100

2 .0 () .'_4

0.24

Il . SOURC'ES DETECTORS

At Kurukcalo / Superconducting tunnel junction detectors

2611 Sample B Outputs from preamplifier 40

Tr = 0 .6111sec

20

40

00 1 =1944 sec Td ---I

r

0

1 0

I

1

I

1

I

2

1

3

I

1

20 30 40 ~i see Fig. 12 . A typical pulse shape of output signals from the charge-sensitive preamplifier (with sample i3), which are induced by 5 .3 MeV cv particles [23] . 10

defined by the differential dV/d 1 at V = 0, which was determined by the 1-V curve measured at 0.4 K. This resistance can be used as a parameter to indicate the leakage current of each sample . The effective capacitance Crtf was estimated from eq . (9) using C' = 250 pF and C = 24 pF for sample A and 600 pF for B and C. The rather large value for C' came from the preamplifier used in the work, which is designed for surfacebarrier semiconductor detectors with large input capacitance : the noise level is 2 keV for Si (E = 3 .6 eV and C = 0 pF) and the input capacitance is about 200 pF . 5.4. Results an(I cliscussions 5.,x.1 . Output signals Signals from the series-connected STJ detector (outputs from the preamplifier) were carefully obscived with oscilloscopes. The typical pulse shape of tho signal is given in fig. 12 . Some features of signals from each sample are summarized in table 3. Outputs from sample A were too small to observe directly with an oscilloscope . The pulse height of sample A was estimated with the voltage gain of the spectroscopy amplifier employed . The risetime and the decay time for samples B and C are determined from the pulse shapes obtained by a digital oscilloscope, as given in

fig. 12 . The risetime is rather fast (= 1 p,s) ; this indicatcs that nonthermal ballistic phonons excited by radiations are transmitted sufficiently fast in the single crystal of a sapphire substrate . The decay time for sample B is 19 p,s while that for sample C is 200 Rs ; the shorter decay time of sample B results from the larger leakage current, as indicated by the dynamic resistance in table 2. The E,rr (= Ee/Q) values estimated for all samples are also. listed in table 3. Note that those values are about three orders smaller than the E values of semiconductor detectors, e.g., 3.6 eV for Si detectors and 2.8 eV for Ge detectors; values for semiconductor detectors are almost equal to E values. The mean energy required to excite one electron in Nb, i.e ., the E value, is 2.5 meV assuming E = 1 .7-1 for Nb . As seen in table 3, the E/E~n for all samples are much smaller than unity . This indicates that most parts of energy (5 .3 MeV) are dispersed without contributing to signal charge Q; for example, only about 30% of the incident energy contributes to Q in the case of sample C. Because of the structure of the series-connected STJs, which is shown in fig. 10, the energy diffusion through leads may not be the dominant origin of the reduction of Erff values . The dispersion of absorbed energy is probably caused by relaxation of phonon energies, escape of nonthermal phonons outside STJ elements, and recombination of quasiparticles before tunneling a tunnel barrier. It should be, however, noted that the Ecff (5 .5 meV) obtained from sample C is nearly equal to the E,,f of 6.6 meV, which is obtained by the direct irradiation of a Nb/Al-AIO,/Nb STJ with X-rays (see table 1) . 5.4.2. Electronic noise The pulse-height spectra of 5 .3 MeV a particles obtained with the samples are shown in fig. 13 . The broadness due to electronic noises is indicated by the pulser peak in each spectrum, which was obtained by supplying pulser signals to the test input of the preamplifier. The widths of the pulser peak W, defined by the full width at the half maximum (FWHM), were 1140 keV, 200 keV and 34 keV for samples A, B and C, respectively . The width IV depends on the signal

Table Preamplifier outputs of each sample: [23]

Sample

13 C

Ileight [mV]

Rise timt'

-1 -II? 45

-0 .6 -0 .6

IRS]

' Intrinsic decay time of the preamplifier is 5511 p.s . _ 1 . 7.1 .

t, E

Decay time '' [ILS] 19 2(ll)

[PC] 1!i y() 1l)()

ECfl

[meV] 47 9.4 8.5

E IE'IC 1

1 1

;

1

5.3 26 29

1,

M. Korakaclo / Superconducting tonncl jttnrtiun dctectors X 10 3

5 .3-MeV a rays from z,0po

6

J

w Z Z

a

U w N Z

pulser -

4

0 4

sample B

0

100

200

300

400

PULSE HEIGHT Fig . 13 . Pulse-height spectra (if electric signals induced by 5 .3 MeV a particles 123) .

charge Q, the effective capacitance C0 , and the leakage current (or the resistivity of the detector R). The number of STJs in sample A, 80(10, is so large that C,,-, (2 .0 RF as seen in table 2) becomes one order larger than the capacitance Cp,, r (nC + C = 0.2 RF). In spite of the large C,  of this sample, the large resistance caused by the connection of STJs in series practically made it possible to detect a particles with the defective STJs (only about 10 11 for a 20 x 20 p,m -' STJ; 1 .5 mil if they are connected in parallel). The width G1;, of sample A is 1140 keV and that of sample B is 200 keV. The leakage current of sample B is about six times larger than that of sample A, as estimated with Ra and it in table 2 . In spite of the larger current and lower resistance of sample B, its width IV is decreased from 1140 keV to 200 keV. This improvement clearly results from the decrease of the effective capacitance C0ff , C,,,f for sample A is 2.0 R, F while that of sample B is 0.24 p.F . The total junction area S of sample B is 9.6 mm -, while that of sample A is 3.2 mm -. The main difference between samples B and C is the leakage current, as seen in table 2; the leakage current of sample C is decreased to one fiftieth that of sample B. This is the reason why the width R"n of sample C is improved from 200 to 34 keV and also why the decay time of output signals is prolonged from 19 to 200 p,s, as seen in table 3. 5 .-ß..i. Escape of pltonons and nontUttforrntty of'STJs The energy resolutions were remarkably worse than the broadness of spectra due to noises. A part of the nonthermal phonons created in the rear substrate escape from the detector region without contributing to the signal charge . The ratio of escaping

phonons to all phonons depends can the position on the rear substrate, where the phonons are produced by the irradiation of nuclear radiations. This indicates that the signal charge Q induced in the series-connected STJ detector used in the measurements is strongly dependent on the incident position of radiations at least when the sensor region is not sufficiently large compared with the sale of the collimator hole [231 . The leakage current per unit area of sample A is two hundred times larger and that of sample B is fifty times larger than that of sample C, as estimated from the data in table 2. This implies that there exists a large fluctuation in properties of STJ elements in such samples with large leakage currents . If the similarity of the d-V characteristics of each STJ element is not sufficient, the series connection does not bias some part of the STJ elements at the proper point for adequate: sensitivity . This effect clearly degrades the sensitivity of the detector . Phonons in a single crystal have a tendency to propagate in specific directions [21,22,361 . When junction properties in the detector are not uniform, the probability of absorbing phonons in STJ elements with adequate sensitivities depends on the incident position of radiations. This effect also results in the deterioration of the resolution of the detector . The energy resolution obtained by the best sample C is still not sufficient, at least it is suffering from the escape of phonons from the detector as discussed before . A method to attenuate the effect of phonon escape is to make samples with larger sensor regions . The effective capacitance C, 1f can be greatly decreased by employing a preamplifier system with lower input capacitance and samples precisely optimized for the number of STJ elements, which may make a great improvement in the signal-to-noise ratio of pulse height spectra. For example, with the values of C= 20 pF and C = 6 p,F/cm 2 . an effective capacitance C~ f, of

0.022 RF for S = 1 cm- and 0.22 RF for S = 100 cm-'

are possible . Series-connected STJ detectors with optimized geometry and low leakage current will have excellent properties . Quite recently, we have obtained an energy resolution of about 170 eV for 5.9 keV X-rays with an 178 x 178 p,m, Nb~'Al-A1(),./AI,iNb junction . The normal resistance R,,; was 0.14 St, and the dynamic resistance at the bias point was about 261 k 51 [371 .

Acknowledgements The author expresses his sincere gratefulness to r. Y. Isozumi and emeritus Prof. Sakae Shimizu of Kyoto University and Dr. Tohru Inoue of NSC . Our study of the STJ could not be possible without their warmhearted encouragements and stimulatine discussions . 11 . SOL°1ét'ES,'DETEL'I-ORS

262

Al. Kurakado / Superconducting trenne! jiutction detectors

He would like to thank Mr. A. Matsumura and Mr. T. Takahashi, who have shown their splendid abilities in the preparation of STJ detectors . The author also wishes to thank Dr. J. Ohono, Dr. T. Iuchi, Dr. K. Sawano and Dr. T. Manabe of NSC and Dr. S. Ito and Dr. R. Katano of Kyoto University for their encouragements and useful discussions.

References [1] G.H . Wood and B.L . White, Appl . Phys . Lett . 15 (1969)

237. [2] G.H . Wood and B.L . White, Can. J. Phys . 51 (1973) 2032 . [3] M. Kurakado and H. Mazaki, Phys. Rev. B22 (1980) 168. [4] M. Kurakado, S. Tachi, R. Katano and H. Mazaki, Bull . Inst . Chem . Res. Kyoto Univ . 59 (1981) 106. [51 M. Kurakado and H. Mazaki, Bull . Inst . Chem . Res. Kyot o Univ . 60 (1982) 243. [6] M. Kurakado and H. Mazaki, Nucl . Instr. and Meth . 185 (1981) 141. [7] M. Kurakado and H. Mazaki, Nucl . Instr. and Meth . 185 (1981) 149. [81 M. Kurakado, Nucl. Instr. and Meth . 196 (1982) 275. [9] M. Kurakado, J. Appl . Phys . 55 (1984) 3185 . [101 M. Kurakado, Oyo Buturi, 53 (1984) 532, in Japanese . [11] C.A. Klein, J. Appl . Phys. 39 (1968) 2029. [121 See, for example, J.J . Chang and D.J . Scalapino, Phys . Rev . Bl5 (1977) 2651, S.B . Kaplan, C.C . Chi, D.N. Langenberg, J.J . Chang and D.J . Scalapino, Phys . Rev. B14 (1976) 4554 . 1131 A. Barone, G. Darbo, S. De Stefano, G. Gallinaro, A. Seri, R. Vaglio and S. Vitale, Nucl . Instr. and Meth . A234 (1985) 61 . 1141 See, for example, Superconductive Particle Detectors, ed . A. Barone (World Scientific, Singapore, 1958), and references therein. [151 H. Kraus, Th . Peterreins, F. Pröbst, F. v. Feilitzsch, R.L . Missbauer, V. Zacek and E. Umlauf, Europhys. Lett . 1 (1986) 161 . [161 D. Twerenbold, Europhyti. Lett . 1 (1986) 209. [17] F. v. Feilitzsch, T. Hertrich, H. Kraus, Th . Peterreins, F.

Pröbst and W. Seidel, Nucl . Instr. and Meth . A271 (1988) 332. [181 W. Rothmund and A. Zehnder, ibid . ref. [14), p. 52 . [191 N.E . Booth, Appl . Phys. Lett . 50 (1987) 293 . [20] It . Kraus, F. v. Feilitzsch, J. Jochum, R.L . Miissbauer, Th . Peterreins and F. Pröbst, Phys. Lett . 13231 (1989) 195 . [211 B. Neuhauser, B. Cabrera, C.J . Matoff and B.A . Young, IEEE Trans. Magn . MAG-25 (1987) 469. [22] Th . Peterreins, F. Pröbst, F. v. Feilitzsch, R.L. M6ssbauer and H. Kraus, Phys . Lett . 13202 (1988) 161 . [23] M. Kurakado, A. Matsumura, T. Takahashi, S. Ito, R. Katano and Y. Isozumi, Rev. Sci. Inst . 62 (1991) 156, ibid. ref. [24], p. 76 . [24] D.J . Goldie, Proc. Workshop on Superconducting Tunnel Junctions for X-ray Detection, December 1990, ed . A. Barone (World Scientific, Singapore, 1991) p. 98 . [25] N.E . Booth, Sci. Prog. Oxford 71 (1987) 563. [26] M. Kurakado, T. Takahashi and A. Matsumura, Appl . Phys. Lett . 57 (1990) 1933 . [271 M. Kurakado and A. Matsumura, Jpn. J. Appl . Phys. 28 (1989) L459 . [281 P. Gare, R. Engelhardt, A. Peacock, D. Twerenbold, J. Lumley and R.E . Somekh, IEEE Trans. Magn . MAG-25 (1989) 1351 . [291 K. Ishibashi, K. Takeno, Y. Oae, T. Sakae, Y. Matsumoto, A. Katase, S. Takada, H. Akoh and N . Nakagawa, IEEE Trans. Magn . MAG-25 (1989) 1354. [301 M. Kurakado and A. Matsumura, Sensors and Actuators A21-23 (1990) 33 . [311 G. Oya, M. Koishi and Y. Sawada, J. Appl . Phys . 60 (1986) 1440 . [321 J.M . Lumley, R.E . Somekh, J .E . Evetts and J .H . James, IEEE Trans. Magn . MAG-21 (1955) 539. [33] R.B . Laibowitz and J.J . Cuonw, J. Appl . Phys . 41 (1970) 2748. [341 F. Pröbst, H. Kraus, Th . Peterreins and F. v. Feilitzsch, Nucl . Instr. and Meth . A280 (1959) 251 . [35] M. Kurakado, to be published. [361 See, for example, B. Taylor, H.J . Maris and C. Elbaum, Phys . Rev. Lett. 23 (1969) 416, G.A . Northrop and J .P . Wolfe, in : Nonequilibrium Phonon Dynamics, ed . W.E . Bron (Plenum, New York, 1985) p. 165. [371 A. Matsumura, T. Takahashi and M. Kurakado, Nucl . Instr. and Meth . A309 (1991) 350.