- Email: [email protected]
Gallium arsenide charged particle detectors; trapping effects
Nuclear Instruments and Methods in Physics Research A 342 (1994) 83-89 North-Holland
NUCLEAR INSTRUMENTS & METHODS IN PHYSICS RESEARCH Section A
Gallium arsenide charged particle detectors; trapping effects RD8 C01 aboration S.P . Beaumont d, R. Bertin 1), C.N. Booth ', C. Buttar `, L. Carraresi c . F . Cindolo a, M. Coloccf c, F.H. Combley S. D'Aufia a, C. del Papa a, M. Dogru ', A Edwards ", F. Foster , A. Francescato , R. Gray , G. Hill , Y. Hou , P. Houston ', G. Hughes f, B.K. Jones f, *, J.G. Lynch e, B. Lisowski a, J. Matheson e, F. Nava 9, M. Nuti c, V. O'Shea P.G . Pelfer e, C. Raine e, J. Santana f, P.H. Seller h, K. Shankar c, P.H. Sharp h, 1.0. Skillicorn e, T. Sloan f, K.M. Smith e, N. Tartoni ", 1 . ten Have e, R.M. Turnbull e, U. Vanni c, A. Vinattieri c, K. Zdansky k, A. Zichichi b
a
9
° Dipartimento di Fisica dell'Universita and INFN, Bologna, Italy b CERN, Geneva, Switzerland Dipartimento di Fisica dell'Universita and INFN, Florence, Italy d Department of Electrical and Electronic Engineering, University of Glasgow, UK 'Department of Physics and Astronomy, University of Glasgow, UK f Department of Physics, University of Lancaster, UK 9 Dipartimento di Fisica dell'Universita and INFN, Modena, Italy h Rutherford Appleton Labitwatory, Chilton, Didcot, Oxon, UK `Department of Physics, University of Sheffield, UK I Department of Electrical Engineering, University of Sheffield, UK k Institute of Radio-technology and Electronics, Prague, Czech Republic
The progress on the development of gallium arsenide particle detectors is reviewed . The limitation to the performance is the presence of traps . Studies of the trap properties using a particle DLTS measurements and C-V measurements are described .
1 . Introduction In previous publications [1-4] we have demonstrated that it is possible to make charged particle detectors from gallium arsenide . Such detectors have been routinely made. as microstrip devices as would be required for a large area detector at the proposed LHC accelerator . Such a large area detector is currently being designed for use by the ATLAS collaboration [5] . The advantages of gallium arsenide over silicon as a medium for charged particle detection are its -,need due to the higher carrier mobility and its radia-
tion hardness . l he latter is a particular advantage in the hostile, high radiation environment at the LHC. In our previous publications [2-4] the speed and the radiation hardness of the devices were demonstrated. However, they were found to be intrinsically more complicated than similar devices made in silicon with
Corresponding author.
somewhat variable charge collection efficiencies . The loss of charge collection efficiency has been attributed to the presence of traps in the material . In this paper we describe work which is currently in progress to investigate the traps in the gallium arsenide material and their effect on the design and properties of detectors .
2 . The detectors The requirements of a detector include fast speed, high conversion efficiency and good radiation damage resistance . The fast speed can be obtained if the length of the depleted area is short and the carrier mobilities, or their saturation velocities if a high field can be developed, are high . For a high conversion efficiency for minimum ionising particles the length of the depleted area should be long and all the free carriers created should leave this area rapidly . The damage
0168-9002/94/S07 .00- 1994 - Elsevier Science B.V . All rights reserved 5SDI 0168-9002(y3)E I 1 18-H
l . DESIGN . FABRICATION
S.P. Beaumont et aL /NucL Instr. and Meth . in Phys. Res. A 342 (1994) 83-89
84
done by radiation should not be such that the free carriers are prevented from operating efficiently. Of the many practical considerations perhaps the most important are that the leakage current must be small so that power dissipation is low and that any substrate, or non-active thickness, is thin to reduce multiple scattering and conversion of neutral particles. Ideally the material would be very pure, with few free carriers and impurity levels, donors or acceptors, with a Schottky contact and an ohmic contact. The purity levels of GaAs are not high. In practice the available high resistivity material is semi-insulating and has defects and impurities to form energy levels near mid-gap which trap most free carriers . Commercial material is variable ii. its properties since it is made for device substrate material and only simple inert characteristics are required for that use. The system is thus complex and one must be careful about the assumptions made in even a device with a Schottky and an ohmic contact. The presence of deep levels, donor or acceptor energy levels near mid-gap, can have several consequences. If full they can be ionised by the incoming radiation to release free carriers to create a signal. If empty they provide traps to capture the free carriers of a signal to reduce the efficiency and if the free carriers are then released slowly they can lengthen the signal pulse . Their charge state, or occupation, also alters the depletion region width and hence the active volume and the details of the charge capture. Although it has bLL . ; LL1.iunmfated that ~hesc de-
0.1
vices have acceptable properties, detailed studies are necessary in order to understand the materials and the device operation so that the properties and performance can be optimised. For this purpose two basic techniques have been used, capacitance-voltage (C-V ) measurements and ionisation deep level transient spectroscopy (DLTS). The experiments are performed over a range of temperatures between 77 K and above room temperature. At the low temperature limit the traps are inactive so that this is a base level. The experimental results are supplemented by modelling and current-voltage (I-V) measurements as well as radiation damage experiments. In the description here we will concentrate on two specific devices with different properties; these are denoted by K10 and N10. Both are made from LEC material with Schottky contact and an ohmic contact. Specimen K10 has a thickness of 440 p.m and is 2.37 x 3.375 mm in area with pressure contacts . It has a high collection efficiency. Specimen N10 has a 2 mm diameter Schottky contact and a thickness of 200 p.m . The ohmic contact is NiGeAu and the Schottky contact is TiAu . It has a lower collection efficiency . 3. Capacitance-voltage measurements The capacitance-voltage-frequency data were taken using an HP4275A LCR bridge over the ranges 0-10,` V (reverse bias on the Schottky diode) and 77-350 K. Typical results are shown in Fig . 1 . These show a x10 -10
a
0.55
b
0.5
0.09 0.08
0.45
0.06
0.35
0.05
0.3
0 .07
0.4
0.04
0.25
0.03
0.2
0.0-
0.15
0,01
0.1
RDUSE MWv)
US_ MAS (v)
Fig . 1 . The zapacitance-~oltagc: rneasurccmer ,i s at 29() K for (a) diode KW %hen the fregtwncses of measur . rnent are at 1(N) . 2( . and -t(g) 11z corresponding to dccreasing %alu,- . ()I .,jpacitan~t: and (h) di , Fdc N10 -4.bth the frcqucnm c5 loo. 2't() . 4M . ' .W and " l00(94E H/
S.P. Beaumont et al. /Nucl. Instr. and Meth. ist Pktis. Res . A 342 (1994) 83-89
decrease as the voltage increases, corresponding to an increase in the depletion width. There is a large frequency dependence; at this temperature with a d(-crease in the capacitance (increase in depletion width) as the frequency increases . X, the temperature increases the capacitance increases as the depletion width decreases . Although the system is complex and corrections are needed, a simple and c--)nsistent approximate analysis of the main features can be made by ass , urit:i that the material is a simple n-type semiconductor. This model can then by extended to include several trapping levels . With this assumption the capacitance per unit area for a Schottky diode made of a semicoriducto ; with uniform carrier concentration n C
A
Er E en -(
~ 1
/
0
C
2.25 2.225 2.2 2.175 2.15 2.125
2
2 a )
2 E r c,en
2.3 2.275
where d is the depletion width, E,E,, the dielectric constant and Vt i the bu+'t-iii voltage . For a semicondu-ton with a significant number of deep impurity levels the quantity measured is not equal to the carrier density, n, or the ionised donor density but Borne effective pace charge Neff . At small voltages this is of less importance We also assume to no current flow. Rearranging this gives A'
85
V 4- Vh .
se that the slope of t`tc 1 /C -'-L' curve gives ~ o_ zarrier concentration . A typ,; al curve of 1 /C -2 -~ at low voltage is shown in Fig . 2. It can he ,cen that these coo d he strai,ht line sections with decreasing slope as the depletion is increased . We will return to this later. If we use Eq. (1) we can obtain a plot of the depletion width against voltage and this is shown in Fig . 3 over a larger range of bars voltage for the tw ; ,, specimens . It can be seen that the slopes and the absolute vaittos vary between spe-tmens depeu]ing on the material and .h, trap details. The significant feature of trips is that they fill at a rate, usua'.!y fast, depende=nt on the lice car . ' ._ , dct,sity and the trap ere ; -.-s-sect iop, o,, and they empty by titer %t al i1ct1Vitt1Ü11 to cite lu',diJl:ïil~li l or valence) band rate gr,en by the Boltzmann facit~r c-,p( F,/kT ). where EA is the trap energy depth . 1 ae.h t .ap there!ore has a characteristic response time, T. If the mea!urcmeni is performed at a slower rate the trap responds, if at a faster rate then it is ina . - 'tvc . As the ten-i- ;crature is changed different traps wilt I ecome active .t the expcrimcr . :al Ircqu-mcy or, if the experiment is kept at one temperature, the traps become: active as the c-prim-ntal frequency is changed . Fig . 4 ill
%how ,, a tvpirRl ;)I(,t ¬
REVERSE BIAS (V) Fig . 2. The plot of 1 against voltage V for specimen K10 at 290 K and frequencies 100, 200, 400 Hz corresponding to decreasing values of capacitance . The reciprocal slope measures the carrier density.
/C
applied voltage, against the ineasutement frequency and a fit to the Debye curve which is expected . The frequency of the transition gives the characteristic trap time and the amplitude between the low and high frequency limits gives a measure of the number of active: traps via Eq . (1). The. fit to theory is good and the apparent number of traps is constant with temperatures, which gives confidence in the analysis. These data will be prc`ented biter and compared with value=s ohtained bv DL7 S. This, frequency analysis means that
-3 E â â 3 z
0 a 0 0 0
~t
t .tkcn at a km
LOG REVERSE BLAS(V) soltage at 29(I K lof diodes KI() (®) and N i( ) (tii ). I DFSKYN . FABKICATR ;.N
86
S.P. Beaumont et al. /Nucl. Instr. and Meth . in Phys. Res. A 342 (1994) 83-89
frequency and temperature are related variables and considerable time can be saved by taking data at only one frequency . The basic features of the results in Fig . 1 can now be understood . It should be noted that because of the lai ge effect of traps the speed at which all measurements, of all quantities, are made needs to be stated as well as the temperature so that it can be known which traps are active. We will now return to the I1C`-V curve of Fig . 2 and discuss the changes of slope as the voltage increases . This corresponds to apparent increases in the carrier density . For this material it is expected that there will be several deep and shallow levels and this is verified by the measurements presented here and later. This results in a complicated structure for the bands bent in the semiconductor depletion region . A schematic diagram is shown in Fig . 5 for the case of one electron trap and one shallow donor. It can be seen that the bands are bent so that the different traps intersect the Fermi energy at different depths into the senniconductor. The trap is assumed to be donor-like and it is full and uncharged if it is below the Fermi energy but is charged if above. Each trap has its own characteristic time depending on its energy depth and the temperature. The basic structure is one of layers parallel to the metal electrode with the layers containing more trap levels ionised as they are closer to the metal . We thus see the basic features of our results . As the frequency or temperature is changed the traps which contribute may change from those at the edge of one of these layers to include those at
065 0.6
4 4
0.55 0.5 0.45
Electron energy
Metal
Distance --
Fig. 5. A schematic diagram of the bands bent at the Schottky diode boundary. It is drawn for a shallow donor and one deep donor trap. The shaded areas of the charge density represent the charge taking part in the AC capacitance measurement if the trap can respond.
another edge, with a different density . Similarly at one frequency and temperature one trap will be active at low voltages . As the voltage is increased the bands bend and successive layers form with more traps ionised . This alters the internal field distribution and the capacitance and depletion width voltage cun~es [6,71. This C-V measurernent also has the potential for determining the number and properties of the traps from the breaks and slopes. However the analysis is not straightforward and the experiment is complicated by the presence of three characteristic frequencies: the AC modulation of the capacitance measurement, the ramp rate of the C-V data and the time the specimen has had to reach equilibrium in its trap occupation after any change in the conditions. 4. DLTS measurements
0.4 0.35 0.3 10 2
10 3
10 4
10 5
LOG FREQUENCY (Hz)
Fig
4 . The variation of d1'/d( 1 / (`) with frcquenc) lor specimen K1t) at temperatures of Zcïi) . 31)) and 144) K anO I V Tht cur4-e% are fits to the Dehye function . I! I 4tivrc aj is the measurement of fr cglscnc,~ and - i% the tsap `harat teri ,Jic wric con ,~tani
Deep level transient spectroscopy (DLTS) is a well cstabiNhed technique: to study the properties of traps. There are many possible variants depending on the sample being studied . In the basic technique the traps are filled by moving the Ferrai lead or by some other excitation and then sr,rne property of the %ample, such as the conductance or depletion capacitance . is monivired
a5
the traps empty by thermal excitation . Tie
amplitude c:f tLe transient gives a measure of the number of trap,, while the temperature dependence of th:-- ?late don4tant glares the actiiat'® , :, cnc.'rgy E,-, and the trap cri~%s-sect :on . if .
S.P. Beaumont et al. /NucL Instr. and Meth . in Phys . Res. A 342 (1994) 83-89 ideal behaviour trapless diode
0(t)
2.25 v W
traps
diode with traps
2
....... ... .. ....... , . ,... .
1.75
a
amplifier decay time
-
87
1.5
W N
a
1 .25
1
1 0.75
time
0.5
Bias supply
0.25
particles
Fig. 6. The circuit diagram of the a-DLTS measuring circuit and a schematic diagram of the transient seen in the charge amplifier. The depletion capacitance transient cart be studied but normally for devices made from semi-insulating material the transient in the d,-vice conductance is studied after an optical pulse to excite the carriers . We have developed a new technique which has the advantage of being very versatile and also reflecting the particle detection capabilities to which the devices being studied will be used .
c J
14
a
1
0
0.1 0.2 0 .3 0.4 0.5 0.6 0.7 0.8 0.9 1®-5 x
TIME (s) Fig. 7. A typical charge transient pulse from the circuit of Fig. 6. Specimen K10 at 290K using a-particle excitation into the Schottky contact end of the diode. The device is biased as a particle detector under current bias and the signal is passed into an integrating amplifier to produce a charge pulse. The circuit is shown in the inset to Fig. 6 which also shows the signal obtained . For an ideal device and amplifier there is a rapid rise after the incoming radiation pulse and the charge level remains constant . With real amplifiers the output droops over long times. The rise time is determined by the transit tame of the ionised carriers and the amplifier response time . In our detectors a component of the rise is exponential and determined by some
c
12
12
10
10
6 4 2 0 -2
2
4
6
6
10
12
I aoo/T(K)
Fig. K. The Arrhencus plot for the traps seen in the dicxie%. The "olid line~ are from the DLTS experiments, the broken unes are from the ('- V data and the data points are from the rec<
S.P. Beaumont et al. /Nucl. Instr. and Meth. in Phys. Res. A 342 (1994) 83-89
88
É v Z O z W U Z O U
1000/T(10
I OOO/T (K)
Fig. 9. The trap concentrations found using the C-V data for the traps shown in Fig. 8. (a) diode K10, (b) N10. The concentration of the trap providing the recovery time constant (shaded region) was deduced from the DLTS data .
trap time constant. This trapping slows the response, and if slow compared with the amplifier droop, reduces the pulse height and hence the detector efficiency . A typical experimental curve is shown in Fig . 7. We fit the response to a function of the form A= {A + J: A1[1-exp(-A 2 t ) ] ) exp(- A301 (3) t i where A is the primary ionisation step, A 1 is the amplitude propot tional to the number of traps, (A, )- t is the trap time constant and (A3) - I is the amplifier droop time constant which does not vary during the experiment. A number, i, of traps can be fitted . At low temperatures no traps are active but as the temperature is raised the traps speed up to become faster than the amplifier droop rate and hence become measurable. At higher temperatures they become faster than the amplifier rise time and then add to the fast primary response A ® so that the charge collection efficiency increases . From the temperature dependence of the trap transients we can obtain the trap activation energy EA and cross section, a. The Arrhenius plot for the C- V and DLTS data is shown in Fig . 8a for specimen K10 and in Fig. 8b for specimen N10 . It can be seen that there is excellent agreement between the two methods for EA (the slope of the curve) and a (the intercept). These measurements thus have the capability of completely characterising the traps within the devices . Modelling of the band bending is then possible using these data in Poisson's equation .
The trap densities for the same specimens measured using the C-V method are shown in Fig . 9. These DLTS results were obtained by using at-particles injected into the Schottky gate side. The small depth of penetration results in the ionisation occurring in the depletion region. Data have been obtained with the a's injected through the ohmic contact . Measurements are in progress using minimum ionising signals from a ß source . Short light pulses with photon energy above and below the band gap energy, so that there is short and deep penetration, have also been used to demonstrate the versatility of the technique and to verify the ionising radiation results . There is complete consistency. The results of all these measurements are summarised in Table 1 which shows the trap parameters and possible trap identification [8,9]. It appears that Table 1 Measured properties of the traps found from the C-V data Trap energy (eV)
Capture cross section (cm - ~)
Trap density (cm - `)
K10, CCE for alphas 80c7c at room temperature 0.70 1.6E - 14 1.65E16 0.61 1.9E-10 2.01E15 0.24 7.12E- 16 7.2E14 N10 : CCE for alphas 30%. at room temperature 0.70 1.6E- 14 1.9E16 0.62 2.07E - 10 1 .91 E 15 0.32 4.2E- 15 L2E15 0.119 1 .32E - 17 7E! 4
Label
EL2
EL2 ET2
S.P. Beaumont et al. /Nucl. Instr. and Meth . in Phys. Res. A 342 (1994) 83-89
the low efficiency and other undesirable properties shown by N10 is due ro the lower-energy traps, since these are the observed difference between it and the good diode K10. We would like to report an encouraging result which suggests that samples with poor efficiency, and perhaps those that have had their performance degraded by radiation damage, may have their efficiency increased . At low temperatures a device showing an efficiency of about 70% can be treated to give a nearly ideal characteristic with approaching 100% efficiency by applying a short forward bias pulse to fill all the traps. The inefficiency recovers w;th an exponential time constant corresponding to a trap. This time constant is plotted on the Arrhenius plot of Fig. 8b. This is thus evidence that trap removal, compensation or tilling can be used to improve the device performance . The accuracy of the recovery data using this present method is not yet high enough to identify the trap involved . 5. Discussion The results presented show that GaAs detectors have an acceptable performance. However there are still aspects of devices made from a maturing technology which require investigation to aid understanding so that consistent high quality and reliable devices may be fabricated . One aspect is the need to understand and characterise the material which is needed for the fabrication and can be expected to be available. The results presented here for the analysis and measurement of the properties of the traps in the devices show that much is known about the traps in the low temperature region . However there are no obvious explanations available for some of the effects seen in
89
the specimens at room temperature and above, where the devices will be expected to work. The C-V results at room temperature for some specimens, with poor efficiency, show a capacitance maximum at very low voltages. This is shown in Fig . 1 b. The trap densities measured from the a-DLTS transients also show unexpected variations near room temperature in the same samples. The behaviour of these samples needs to be understood so that all devices can be made to the standard of the good devices . Acknowledgements We wish to thank SERC (UK) and INFN (Italy) for financial support . References [1] R. Bertin et al., Nucl. Instr. and Meth. A 294 (1990) 211. [2] S.P. Beaumont et al., Nucl. Instr. and Meth. A 321 (1992) 172. [3] S.P. Beaumont et al ., Nucl. Instr. and Meth. A 322 (1992) 472 . [4] S.P. Beaumont et al., Nucl. Instr . and Meth. A 326 (1993) 313. [5] ATLAS Letter of Intent CERN LHCC 92/4 LHCC/12 (1992) . [6] P. Blood and J.W. Orton, Electrical Characterisation of Semiconductors : Majority Carriers and Electron States (Academic Press, 1992). [7] Z. Li and H.W. Kraner, IEEE Trans. Nucl. Sci . NS-38 (1991) 244 . [8] G.M. Martin, A. Mitonneau and A. Mircea, Electron. Lett. 13 (1977) 191 . [9] A. Mitonneau, G.M. Martin and A. Mircea, Electron. Lett. 13 (1977) 666.
I. DESIGN, FABRICATION
NUCLEAR INSTRUMENTS & METHODS IN PHYSICS RESEARCH Section A
Gallium arsenide charged particle detectors; trapping effects RD8 C01 aboration S.P . Beaumont d, R. Bertin 1), C.N. Booth ', C. Buttar `, L. Carraresi c . F . Cindolo a, M. Coloccf c, F.H. Combley S. D'Aufia a, C. del Papa a, M. Dogru ', A Edwards ", F. Foster , A. Francescato , R. Gray , G. Hill , Y. Hou , P. Houston ', G. Hughes f, B.K. Jones f, *, J.G. Lynch e, B. Lisowski a, J. Matheson e, F. Nava 9, M. Nuti c, V. O'Shea P.G . Pelfer e, C. Raine e, J. Santana f, P.H. Seller h, K. Shankar c, P.H. Sharp h, 1.0. Skillicorn e, T. Sloan f, K.M. Smith e, N. Tartoni ", 1 . ten Have e, R.M. Turnbull e, U. Vanni c, A. Vinattieri c, K. Zdansky k, A. Zichichi b
a
9
° Dipartimento di Fisica dell'Universita and INFN, Bologna, Italy b CERN, Geneva, Switzerland Dipartimento di Fisica dell'Universita and INFN, Florence, Italy d Department of Electrical and Electronic Engineering, University of Glasgow, UK 'Department of Physics and Astronomy, University of Glasgow, UK f Department of Physics, University of Lancaster, UK 9 Dipartimento di Fisica dell'Universita and INFN, Modena, Italy h Rutherford Appleton Labitwatory, Chilton, Didcot, Oxon, UK `Department of Physics, University of Sheffield, UK I Department of Electrical Engineering, University of Sheffield, UK k Institute of Radio-technology and Electronics, Prague, Czech Republic
The progress on the development of gallium arsenide particle detectors is reviewed . The limitation to the performance is the presence of traps . Studies of the trap properties using a particle DLTS measurements and C-V measurements are described .
1 . Introduction In previous publications [1-4] we have demonstrated that it is possible to make charged particle detectors from gallium arsenide . Such detectors have been routinely made. as microstrip devices as would be required for a large area detector at the proposed LHC accelerator . Such a large area detector is currently being designed for use by the ATLAS collaboration [5] . The advantages of gallium arsenide over silicon as a medium for charged particle detection are its -,need due to the higher carrier mobility and its radia-
tion hardness . l he latter is a particular advantage in the hostile, high radiation environment at the LHC. In our previous publications [2-4] the speed and the radiation hardness of the devices were demonstrated. However, they were found to be intrinsically more complicated than similar devices made in silicon with
Corresponding author.
somewhat variable charge collection efficiencies . The loss of charge collection efficiency has been attributed to the presence of traps in the material . In this paper we describe work which is currently in progress to investigate the traps in the gallium arsenide material and their effect on the design and properties of detectors .
2 . The detectors The requirements of a detector include fast speed, high conversion efficiency and good radiation damage resistance . The fast speed can be obtained if the length of the depleted area is short and the carrier mobilities, or their saturation velocities if a high field can be developed, are high . For a high conversion efficiency for minimum ionising particles the length of the depleted area should be long and all the free carriers created should leave this area rapidly . The damage
0168-9002/94/S07 .00- 1994 - Elsevier Science B.V . All rights reserved 5SDI 0168-9002(y3)E I 1 18-H
l . DESIGN . FABRICATION
S.P. Beaumont et aL /NucL Instr. and Meth . in Phys. Res. A 342 (1994) 83-89
84
done by radiation should not be such that the free carriers are prevented from operating efficiently. Of the many practical considerations perhaps the most important are that the leakage current must be small so that power dissipation is low and that any substrate, or non-active thickness, is thin to reduce multiple scattering and conversion of neutral particles. Ideally the material would be very pure, with few free carriers and impurity levels, donors or acceptors, with a Schottky contact and an ohmic contact. The purity levels of GaAs are not high. In practice the available high resistivity material is semi-insulating and has defects and impurities to form energy levels near mid-gap which trap most free carriers . Commercial material is variable ii. its properties since it is made for device substrate material and only simple inert characteristics are required for that use. The system is thus complex and one must be careful about the assumptions made in even a device with a Schottky and an ohmic contact. The presence of deep levels, donor or acceptor energy levels near mid-gap, can have several consequences. If full they can be ionised by the incoming radiation to release free carriers to create a signal. If empty they provide traps to capture the free carriers of a signal to reduce the efficiency and if the free carriers are then released slowly they can lengthen the signal pulse . Their charge state, or occupation, also alters the depletion region width and hence the active volume and the details of the charge capture. Although it has bLL . ; LL1.iunmfated that ~hesc de-
0.1
vices have acceptable properties, detailed studies are necessary in order to understand the materials and the device operation so that the properties and performance can be optimised. For this purpose two basic techniques have been used, capacitance-voltage (C-V ) measurements and ionisation deep level transient spectroscopy (DLTS). The experiments are performed over a range of temperatures between 77 K and above room temperature. At the low temperature limit the traps are inactive so that this is a base level. The experimental results are supplemented by modelling and current-voltage (I-V) measurements as well as radiation damage experiments. In the description here we will concentrate on two specific devices with different properties; these are denoted by K10 and N10. Both are made from LEC material with Schottky contact and an ohmic contact. Specimen K10 has a thickness of 440 p.m and is 2.37 x 3.375 mm in area with pressure contacts . It has a high collection efficiency. Specimen N10 has a 2 mm diameter Schottky contact and a thickness of 200 p.m . The ohmic contact is NiGeAu and the Schottky contact is TiAu . It has a lower collection efficiency . 3. Capacitance-voltage measurements The capacitance-voltage-frequency data were taken using an HP4275A LCR bridge over the ranges 0-10,` V (reverse bias on the Schottky diode) and 77-350 K. Typical results are shown in Fig . 1 . These show a x10 -10
a
0.55
b
0.5
0.09 0.08
0.45
0.06
0.35
0.05
0.3
0 .07
0.4
0.04
0.25
0.03
0.2
0.0-
0.15
0,01
0.1
RDUSE MWv)
US_ MAS (v)
Fig . 1 . The zapacitance-~oltagc: rneasurccmer ,i s at 29() K for (a) diode KW %hen the fregtwncses of measur . rnent are at 1(N) . 2( . and -t(g) 11z corresponding to dccreasing %alu,- . ()I .,jpacitan~t: and (h) di , Fdc N10 -4.bth the frcqucnm c5 loo. 2't() . 4M . ' .W and " l00(94E H/
S.P. Beaumont et al. /Nucl. Instr. and Meth. ist Pktis. Res . A 342 (1994) 83-89
decrease as the voltage increases, corresponding to an increase in the depletion width. There is a large frequency dependence; at this temperature with a d(-crease in the capacitance (increase in depletion width) as the frequency increases . X, the temperature increases the capacitance increases as the depletion width decreases . Although the system is complex and corrections are needed, a simple and c--)nsistent approximate analysis of the main features can be made by ass , urit:i that the material is a simple n-type semiconductor. This model can then by extended to include several trapping levels . With this assumption the capacitance per unit area for a Schottky diode made of a semicoriducto ; with uniform carrier concentration n C
A
Er E en -(
~ 1
/
0
C
2.25 2.225 2.2 2.175 2.15 2.125
2
2 a )
2 E r c,en
2.3 2.275
where d is the depletion width, E,E,, the dielectric constant and Vt i the bu+'t-iii voltage . For a semicondu-ton with a significant number of deep impurity levels the quantity measured is not equal to the carrier density, n, or the ionised donor density but Borne effective pace charge Neff . At small voltages this is of less importance We also assume to no current flow. Rearranging this gives A'
85
V 4- Vh .
se that the slope of t`tc 1 /C -'-L' curve gives ~ o_ zarrier concentration . A typ,; al curve of 1 /C -2 -~ at low voltage is shown in Fig . 2. It can he ,cen that these coo d he strai,ht line sections with decreasing slope as the depletion is increased . We will return to this later. If we use Eq. (1) we can obtain a plot of the depletion width against voltage and this is shown in Fig . 3 over a larger range of bars voltage for the tw ; ,, specimens . It can be seen that the slopes and the absolute vaittos vary between spe-tmens depeu]ing on the material and .h, trap details. The significant feature of trips is that they fill at a rate, usua'.!y fast, depende=nt on the lice car . ' ._ , dct,sity and the trap ere ; -.-s-sect iop, o,, and they empty by titer %t al i1ct1Vitt1Ü11 to cite lu',diJl:ïil~li l or valence) band rate gr,en by the Boltzmann facit~r c-,p( F,/kT ). where EA is the trap energy depth . 1 ae.h t .ap there!ore has a characteristic response time, T. If the mea!urcmeni is performed at a slower rate the trap responds, if at a faster rate then it is ina . - 'tvc . As the ten-i- ;crature is changed different traps wilt I ecome active .t the expcrimcr . :al Ircqu-mcy or, if the experiment is kept at one temperature, the traps become: active as the c-prim-ntal frequency is changed . Fig . 4 ill
%how ,, a tvpirRl ;)I(,t ¬
REVERSE BIAS (V) Fig . 2. The plot of 1 against voltage V for specimen K10 at 290 K and frequencies 100, 200, 400 Hz corresponding to decreasing values of capacitance . The reciprocal slope measures the carrier density.
/C
applied voltage, against the ineasutement frequency and a fit to the Debye curve which is expected . The frequency of the transition gives the characteristic trap time and the amplitude between the low and high frequency limits gives a measure of the number of active: traps via Eq . (1). The. fit to theory is good and the apparent number of traps is constant with temperatures, which gives confidence in the analysis. These data will be prc`ented biter and compared with value=s ohtained bv DL7 S. This, frequency analysis means that
-3 E â â 3 z
0 a 0 0 0
~t
t .tkcn at a km
LOG REVERSE BLAS(V) soltage at 29(I K lof diodes KI() (®) and N i( ) (tii ). I DFSKYN . FABKICATR ;.N
86
S.P. Beaumont et al. /Nucl. Instr. and Meth . in Phys. Res. A 342 (1994) 83-89
frequency and temperature are related variables and considerable time can be saved by taking data at only one frequency . The basic features of the results in Fig . 1 can now be understood . It should be noted that because of the lai ge effect of traps the speed at which all measurements, of all quantities, are made needs to be stated as well as the temperature so that it can be known which traps are active. We will now return to the I1C`-V curve of Fig . 2 and discuss the changes of slope as the voltage increases . This corresponds to apparent increases in the carrier density . For this material it is expected that there will be several deep and shallow levels and this is verified by the measurements presented here and later. This results in a complicated structure for the bands bent in the semiconductor depletion region . A schematic diagram is shown in Fig . 5 for the case of one electron trap and one shallow donor. It can be seen that the bands are bent so that the different traps intersect the Fermi energy at different depths into the senniconductor. The trap is assumed to be donor-like and it is full and uncharged if it is below the Fermi energy but is charged if above. Each trap has its own characteristic time depending on its energy depth and the temperature. The basic structure is one of layers parallel to the metal electrode with the layers containing more trap levels ionised as they are closer to the metal . We thus see the basic features of our results . As the frequency or temperature is changed the traps which contribute may change from those at the edge of one of these layers to include those at
065 0.6
4 4
0.55 0.5 0.45
Electron energy
Metal
Distance --
Fig. 5. A schematic diagram of the bands bent at the Schottky diode boundary. It is drawn for a shallow donor and one deep donor trap. The shaded areas of the charge density represent the charge taking part in the AC capacitance measurement if the trap can respond.
another edge, with a different density . Similarly at one frequency and temperature one trap will be active at low voltages . As the voltage is increased the bands bend and successive layers form with more traps ionised . This alters the internal field distribution and the capacitance and depletion width voltage cun~es [6,71. This C-V measurernent also has the potential for determining the number and properties of the traps from the breaks and slopes. However the analysis is not straightforward and the experiment is complicated by the presence of three characteristic frequencies: the AC modulation of the capacitance measurement, the ramp rate of the C-V data and the time the specimen has had to reach equilibrium in its trap occupation after any change in the conditions. 4. DLTS measurements
0.4 0.35 0.3 10 2
10 3
10 4
10 5
LOG FREQUENCY (Hz)
Fig
4 . The variation of d1'/d( 1 / (`) with frcquenc) lor specimen K1t) at temperatures of Zcïi) . 31)) and 144) K anO I V Tht cur4-e% are fits to the Dehye function . I! I 4tivrc aj is the measurement of fr cglscnc,~ and - i% the tsap `harat teri ,Jic wric con ,~tani
Deep level transient spectroscopy (DLTS) is a well cstabiNhed technique: to study the properties of traps. There are many possible variants depending on the sample being studied . In the basic technique the traps are filled by moving the Ferrai lead or by some other excitation and then sr,rne property of the %ample, such as the conductance or depletion capacitance . is monivired
a5
the traps empty by thermal excitation . Tie
amplitude c:f tLe transient gives a measure of the number of trap,, while the temperature dependence of th:-- ?late don4tant glares the actiiat'® , :, cnc.'rgy E,-, and the trap cri~%s-sect :on . if .
S.P. Beaumont et al. /NucL Instr. and Meth . in Phys . Res. A 342 (1994) 83-89 ideal behaviour trapless diode
0(t)
2.25 v W
traps
diode with traps
2
....... ... .. ....... , . ,... .
1.75
a
amplifier decay time
-
87
1.5
W N
a
1 .25
1
1 0.75
time
0.5
Bias supply
0.25
particles
Fig. 6. The circuit diagram of the a-DLTS measuring circuit and a schematic diagram of the transient seen in the charge amplifier. The depletion capacitance transient cart be studied but normally for devices made from semi-insulating material the transient in the d,-vice conductance is studied after an optical pulse to excite the carriers . We have developed a new technique which has the advantage of being very versatile and also reflecting the particle detection capabilities to which the devices being studied will be used .
c J
14
a
1
0
0.1 0.2 0 .3 0.4 0.5 0.6 0.7 0.8 0.9 1®-5 x
TIME (s) Fig. 7. A typical charge transient pulse from the circuit of Fig. 6. Specimen K10 at 290K using a-particle excitation into the Schottky contact end of the diode. The device is biased as a particle detector under current bias and the signal is passed into an integrating amplifier to produce a charge pulse. The circuit is shown in the inset to Fig. 6 which also shows the signal obtained . For an ideal device and amplifier there is a rapid rise after the incoming radiation pulse and the charge level remains constant . With real amplifiers the output droops over long times. The rise time is determined by the transit tame of the ionised carriers and the amplifier response time . In our detectors a component of the rise is exponential and determined by some
c
12
12
10
10
6 4 2 0 -2
2
4
6
6
10
12
I aoo/T(K)
Fig. K. The Arrhencus plot for the traps seen in the dicxie%. The "olid line~ are from the DLTS experiments, the broken unes are from the ('- V data and the data points are from the rec<
S.P. Beaumont et al. /Nucl. Instr. and Meth. in Phys. Res. A 342 (1994) 83-89
88
É v Z O z W U Z O U
1000/T(10
I OOO/T (K)
Fig. 9. The trap concentrations found using the C-V data for the traps shown in Fig. 8. (a) diode K10, (b) N10. The concentration of the trap providing the recovery time constant (shaded region) was deduced from the DLTS data .
trap time constant. This trapping slows the response, and if slow compared with the amplifier droop, reduces the pulse height and hence the detector efficiency . A typical experimental curve is shown in Fig . 7. We fit the response to a function of the form A= {A + J: A1[1-exp(-A 2 t ) ] ) exp(- A301 (3) t i where A is the primary ionisation step, A 1 is the amplitude propot tional to the number of traps, (A, )- t is the trap time constant and (A3) - I is the amplifier droop time constant which does not vary during the experiment. A number, i, of traps can be fitted . At low temperatures no traps are active but as the temperature is raised the traps speed up to become faster than the amplifier droop rate and hence become measurable. At higher temperatures they become faster than the amplifier rise time and then add to the fast primary response A ® so that the charge collection efficiency increases . From the temperature dependence of the trap transients we can obtain the trap activation energy EA and cross section, a. The Arrhenius plot for the C- V and DLTS data is shown in Fig . 8a for specimen K10 and in Fig. 8b for specimen N10 . It can be seen that there is excellent agreement between the two methods for EA (the slope of the curve) and a (the intercept). These measurements thus have the capability of completely characterising the traps within the devices . Modelling of the band bending is then possible using these data in Poisson's equation .
The trap densities for the same specimens measured using the C-V method are shown in Fig . 9. These DLTS results were obtained by using at-particles injected into the Schottky gate side. The small depth of penetration results in the ionisation occurring in the depletion region. Data have been obtained with the a's injected through the ohmic contact . Measurements are in progress using minimum ionising signals from a ß source . Short light pulses with photon energy above and below the band gap energy, so that there is short and deep penetration, have also been used to demonstrate the versatility of the technique and to verify the ionising radiation results . There is complete consistency. The results of all these measurements are summarised in Table 1 which shows the trap parameters and possible trap identification [8,9]. It appears that Table 1 Measured properties of the traps found from the C-V data Trap energy (eV)
Capture cross section (cm - ~)
Trap density (cm - `)
K10, CCE for alphas 80c7c at room temperature 0.70 1.6E - 14 1.65E16 0.61 1.9E-10 2.01E15 0.24 7.12E- 16 7.2E14 N10 : CCE for alphas 30%. at room temperature 0.70 1.6E- 14 1.9E16 0.62 2.07E - 10 1 .91 E 15 0.32 4.2E- 15 L2E15 0.119 1 .32E - 17 7E! 4
Label
EL2
EL2 ET2
S.P. Beaumont et al. /Nucl. Instr. and Meth . in Phys. Res. A 342 (1994) 83-89
the low efficiency and other undesirable properties shown by N10 is due ro the lower-energy traps, since these are the observed difference between it and the good diode K10. We would like to report an encouraging result which suggests that samples with poor efficiency, and perhaps those that have had their performance degraded by radiation damage, may have their efficiency increased . At low temperatures a device showing an efficiency of about 70% can be treated to give a nearly ideal characteristic with approaching 100% efficiency by applying a short forward bias pulse to fill all the traps. The inefficiency recovers w;th an exponential time constant corresponding to a trap. This time constant is plotted on the Arrhenius plot of Fig. 8b. This is thus evidence that trap removal, compensation or tilling can be used to improve the device performance . The accuracy of the recovery data using this present method is not yet high enough to identify the trap involved . 5. Discussion The results presented show that GaAs detectors have an acceptable performance. However there are still aspects of devices made from a maturing technology which require investigation to aid understanding so that consistent high quality and reliable devices may be fabricated . One aspect is the need to understand and characterise the material which is needed for the fabrication and can be expected to be available. The results presented here for the analysis and measurement of the properties of the traps in the devices show that much is known about the traps in the low temperature region . However there are no obvious explanations available for some of the effects seen in
89
the specimens at room temperature and above, where the devices will be expected to work. The C-V results at room temperature for some specimens, with poor efficiency, show a capacitance maximum at very low voltages. This is shown in Fig . 1 b. The trap densities measured from the a-DLTS transients also show unexpected variations near room temperature in the same samples. The behaviour of these samples needs to be understood so that all devices can be made to the standard of the good devices . Acknowledgements We wish to thank SERC (UK) and INFN (Italy) for financial support . References [1] R. Bertin et al., Nucl. Instr. and Meth. A 294 (1990) 211. [2] S.P. Beaumont et al., Nucl. Instr. and Meth. A 321 (1992) 172. [3] S.P. Beaumont et al ., Nucl. Instr. and Meth. A 322 (1992) 472 . [4] S.P. Beaumont et al., Nucl. Instr . and Meth. A 326 (1993) 313. [5] ATLAS Letter of Intent CERN LHCC 92/4 LHCC/12 (1992) . [6] P. Blood and J.W. Orton, Electrical Characterisation of Semiconductors : Majority Carriers and Electron States (Academic Press, 1992). [7] Z. Li and H.W. Kraner, IEEE Trans. Nucl. Sci . NS-38 (1991) 244 . [8] G.M. Martin, A. Mitonneau and A. Mircea, Electron. Lett. 13 (1977) 191 . [9] A. Mitonneau, G.M. Martin and A. Mircea, Electron. Lett. 13 (1977) 666.
I. DESIGN, FABRICATION