- Email: [email protected]
Electrical contacts to silicon
831
NOTES
character of those patterns is similar to that seen on etch pits which were formed on single-crystal CdS.(s) No “stepped” patterns were evident on the films tested in this Laboratory, although at least 50 samples have been examined with etch times varying from 10 set to 5 min. The vapours were in an equilibrium concentration HCl solution at room temperature. Unfortunately, no structures like those used by REYNOLDS(~) could be obtained for direct test. The conclusion to be reached is that the observation of vapour-phase HCl etch structures alone is not a reliable means for estimating minimum crystallite sizes in deposited semiconductor films. Acknowledgements-The authors wish to thank R. ZULEEC of the Hughes Microcircuit Division for comments, and for providing CdS film samples.
M. L. CONRAGAN R. S. MULLER Electrical University
Engineering
Department
of California,
Berkeley
EXPERIMENTAL
References 1. R. ZULEEC and R. S. MULLER, Solid-State Electron. 7, 57.5 (1964). 2. D. C. REYNOLDS, The Art and Science of Growing Cqjstals, p. 73 Wiley, N.Y. (1963).
Solid-State Electronics Pergamon Press 1965. Vol. 8, Printed in Great Britain pp. 831-833. Electrical
contacts
(Received 19 March 1965
determined for several evaporated metals and silicon of varying impurity concentration. The variables investigated were (a) pre-evaporation silicon cleaning, (b) metal evaporation conditions, and (c) post evaporation contact sintering conditions. SULLIVAN and EIGLER@) made measurements of the contact resistance between silicon and electroless nickel, but their measurement technique was not capable of accurate values below 10-s Q-cm2. The measurement method employed for this work is similar to that developed by MENGALI and SEILER(~) who studied metal contacts on thermoelectric materials and obtained values of specific contact resistance in the range of 10~Q-cma. The metals selected for this study include alumimolybdenum, nickel, cobalt, num, vanadium, palladium, and silver and they were chosen on the basis that they represent a group of potentially useful metals for use in semiconductor device manufacture.
to silicon
; in revised form 3 May 1965)
INTRODUCTION
DUE TO the increased usage and the more stringent specifications placed on silicon devices, it has become increasingly important to develop methods of constructing ohmic, very low-resistance contacts to silicon. Although information can be found in the literature concerning the nature of rectifying contacts of metal-semiconductor interfaces,(r) very little can be found concerning the very important, practical problem of obtaining ohmic lowresistance contacts. This work attempts to determine the effect of some important process variables on the contact resistance at the metal-silicon interface. The effect of these process variables was
The semiconductor material used in this investigation was typically Czochralski pulled, single crystal silicon. The samples were sawn into blocks 5 x 5 x 2.5 mm such that the 5 x 5 mm faces are the (111) plane. After lapping the 5 mm faces, the samples were chemically polished in a solution of nitric, hydrofluoric and acetic acid before metallic contacts are applied to the opposite (111) faces. Contact application was performed principally by vacuum deposition. Plated and alloyed metals were studied, but the results are not reported here. After the contacts were deposited, the sample was placed in a measurement fixture where a clean indium preform is mechanically pressed onto the contact surfaces deforming the preform to make intimate contact with the evaporated metal contact surface. After the 2.5 mm surface of the sample is polished flush with the mounting fixture, a sinusoidal current from an oscillator is amplified, measured and passed through the sample. A tungsten probe, etched to a very sharp point and mounted rigidly above the assembly, is lowered onto the sample to sense the voltage developed at any particular point along the sample. A micrometer table capable of 0.01 mm resolution is moved laterally in 0.10 mm increments, and the
832
NOTES
voltage is recorded at each step. This probe excursion is made normal to the plane of the contact interface. By using a low-noise, battery-powered, low-level preamplifier and carefully shielding all wiring, voltages as low as 2 x 10-7 V can be routinely detected. The indium-metal contact voltage drop is at least one order of magnitude lower than this and does not contribute to the measurement appreciably. By graphically plotting the measured voltage vs. the distance along the sample and extrapolating the line to the sample edge, the voltage across the interfacial barrier is determined. From this voltage the contact resistance in Q-cm2 is easily calculated using Ohm’s law. All measured values of contact resistance noted below are the average of at least four measurements. Generally, the values noted were reproducible to + 20 per cent.
resistance. On the other hand, a dilute hydrofluoric acid rinse which reduced the surface oxide thickness to a minimum resulted in a much lower value of contact resistance. For example, using aluminum contacts, the value of Rc was a factor of 50 lower after a hydrofluoric acid treatment than when nitric acid was employed. A similar effect was noted for the cases of nickel and molybdenum. One note with regard to this particular area of investigation was the necessity to shield the samples from the evaporating charge until vaporization of the metal charge was initiated. Unless this precaution was taken, the measured values of R, were scattered (usually increased R, resulted), indicating that impurities on the charge (possibly adsorbed organic material) were being driven onto the sample. Evaporation temperature The effect of substrate temperature during metal deposition was studied briefly using aluminum as the contacting metal. Substrate temperatures were varied from 25 to 600°C with only small variations in the values of R, resulting, but it is interesting to note that for 0.002 Q-cm p-type silicon the lowest value of R, (4.1 x 10-S Q-cm”) was found for the case in which the substrate temperature was in excess of the silicon-aluminum eutectic temperature (577°C).
RESULTS Pre-contact cleanup of the silicon surface It was found that the preparation of the silicon surface prior to metal deposition had a marked effect on the resulting value of contact resistance. A variety of pre-evaporation acid treatments and rinses were studied, and it was clear from the results that the effect of a relatively thick surface oxide, as would be present after a hot nitric or sulfuric acid treatment, is to increase the contact Table 1. Contact resistance in n-cm2
Si resistivity in R-cm and cond. type 0.002 p 0.005 p O,OlOp 0.050 p 0.10 p 0.25 p 0.50 p
3.7 x 10-5 5.2 x 10-s
0.005 0010 0.05 0.10 0.50
4.1 x 10-3 8.3 x 10-s 6.4 x10-s
n 11 ?I n n
V
Al
6.0 5.4 1.3 1.1
x lo-* x 10-h x 10-3 x lo--3
2.9 x 10-r
5.2 1.3 7.4 6.4 1.8
x 10-s x10-s x 10-s x 10-s x 10-r
of evaporated metal contacts to silicon.*
MO
Ni
Chrome1
Co
4.4 x10-s 1.4 x 10-j
7.8 x 10 -6 2.2 x 10-j 2.1 x10-s
1.3 x 10-a 1.1 x 10-r
2.3x10-j 4-2 x lo--’
2.2 x 10-2 1.4 7.3 2.6 2.1 1.1 2.9
9.4 x10-s x 10-5 x 10-s x 10-r x 10-r
7.8 x 10-s 6.1 x 10-l 2.0
1.8~10 7.0 x 10-d 5.6 x 1O-3 7.0 x10-s
a
1.0 x 10-4 1.4 x 10-a 1.6 x 10-r 1.0
2.2 x 10-s
26
* Samples were chemically polished and soaked in HF. substrate temperatures were 250-300°C.
The evaporations
were performed
at i 10-s Torr,
and
833
NOTES
Post evaporation heat treatment The effect of sintering on R, was studied for several metal-silicon couples. For the metals aluminum, silver, chromium, and molybdenum, sintering at various times and temperatures in nitrogen produced only small changes (generally less than 50 per cent) in the measured values of the contact resistance. On the other hand, sintering produced considerable lowering of R, for the metals palladium and nickel. For example, in the case of nickel contacts the value of R, decreased from 7.0 x 10-4 to 3.8 x 10-4 to 1.8 x 10-5 Q-cm2 for the conditions of no sintering, 450°C for 6 min, 550°C for 6 min respectively. Effects of silicon resistivity and the particular metals used as contacts Several different metals were evaporated onto both n- and p-type silicon. The silicon substrates included a wide range of resistivities, and Rc varied almost seven orders of magnitude between 4.4 x 10-s R-cm2 and 26.0 LLcm2. The results of the measurements are given in Table 1, where it can be seen that R, is highly dependent on the resistivity of the silicon regardless of the particular metal contact. It is interesting to note most of the metals studied yielded low resistance contacts if silicon was highly doped. The values of Rc obtained also indicate the difficulty encountered when attempting to construct low-resistance, ohmic contacts to n-type silicon of 0.01 Q-cm and above. During a measurement to n-type silicon of 0.01 Q-cm and above, the voltage across the contact interface was observed to be rectified, the degree of rectification increasing with resistivity. Since R, then would depend on the direction of current flow, the values of R, in Table 1 of the higher n-type resistivities are not accurate. R. C. HOOPER Semiconductor Research and J. A. CUNNINGHAM J. G. HARPER Development Laboratory Texas Instruments Incorporated Dallas, Texas References 1. H. K. HENISCH, Sylvan. Tech. IX, 73 (1956). 2. M. V. SULLIVAN and J. H. EIGLER, J. Eleclrochem. Sot. 104, 226 (1957): 3. 0. J. MENCALI and M. R. SEILER, Adwznc. Energy Conv. 2, 59 (1962).
Solid-State Electronics pp. 833-835.
Pergamon Press 1965. Vol. 8, Printed in Great Britain
Electra-optic observation effects in gallium
of space arsenide
charge
(Received 3 May 1965) A VISUAL examination of non-uniform electric fields in high resistivity GaAs has been made using the linear electro-optic effect. The nonuniform fields arise from space charge effects in the compensated semiconductor. Above a critical voltage, the space charge distribution becomes unstable and a high field region moves across the sample from cathode to anode, accompanied by oscillations in the current.(l32) The motion of the high field region is strikingly demonstrated by the electro-optic effect. This effect is observed by placing the sample between crossed polarizers and illuminating it with light of wavelength longer than the absorption edge (Fig. 1). The transmitted infrared light is viewed with an image converter. In zero field the sample appears uniformly dark. When a field is applied, the sample becomes birefringentts) and light is transmitted. The field distribution is mapped by the intensity variations over the sample. Any strain present in the sample will cause light to be transmitted in zero fields due to the photoelastic effect.(a) It should be noted that the method described here differs from a previous electro-optic method in which the Franz-Keldysh effect was used to study space charge effects in CdS.c4) No effect could be seen in GaAs without the polarizers. The electro-optic effect for crystals of class 33rn such as GaAs has been discussed in detail by SAMBA. In general the crystal becomes biaxial on application of the field and the phase difference I’ between the extraordinary and ordinary rays depends on both the direction of the field and the direction of viewing. However, if the field is applied along a (111) direction, the crystal becomes uniaxial with axis parallel to the field, and for any direction of viewing perpendicular to the field E the phase difference is I?=
n2/3 -lnir41E x
where no is the index
of refraction,
~41 is the
NOTES
character of those patterns is similar to that seen on etch pits which were formed on single-crystal CdS.(s) No “stepped” patterns were evident on the films tested in this Laboratory, although at least 50 samples have been examined with etch times varying from 10 set to 5 min. The vapours were in an equilibrium concentration HCl solution at room temperature. Unfortunately, no structures like those used by REYNOLDS(~) could be obtained for direct test. The conclusion to be reached is that the observation of vapour-phase HCl etch structures alone is not a reliable means for estimating minimum crystallite sizes in deposited semiconductor films. Acknowledgements-The authors wish to thank R. ZULEEC of the Hughes Microcircuit Division for comments, and for providing CdS film samples.
M. L. CONRAGAN R. S. MULLER Electrical University
Engineering
Department
of California,
Berkeley
EXPERIMENTAL
References 1. R. ZULEEC and R. S. MULLER, Solid-State Electron. 7, 57.5 (1964). 2. D. C. REYNOLDS, The Art and Science of Growing Cqjstals, p. 73 Wiley, N.Y. (1963).
Solid-State Electronics Pergamon Press 1965. Vol. 8, Printed in Great Britain pp. 831-833. Electrical
contacts
(Received 19 March 1965
determined for several evaporated metals and silicon of varying impurity concentration. The variables investigated were (a) pre-evaporation silicon cleaning, (b) metal evaporation conditions, and (c) post evaporation contact sintering conditions. SULLIVAN and EIGLER@) made measurements of the contact resistance between silicon and electroless nickel, but their measurement technique was not capable of accurate values below 10-s Q-cm2. The measurement method employed for this work is similar to that developed by MENGALI and SEILER(~) who studied metal contacts on thermoelectric materials and obtained values of specific contact resistance in the range of 10~Q-cma. The metals selected for this study include alumimolybdenum, nickel, cobalt, num, vanadium, palladium, and silver and they were chosen on the basis that they represent a group of potentially useful metals for use in semiconductor device manufacture.
to silicon
; in revised form 3 May 1965)
INTRODUCTION
DUE TO the increased usage and the more stringent specifications placed on silicon devices, it has become increasingly important to develop methods of constructing ohmic, very low-resistance contacts to silicon. Although information can be found in the literature concerning the nature of rectifying contacts of metal-semiconductor interfaces,(r) very little can be found concerning the very important, practical problem of obtaining ohmic lowresistance contacts. This work attempts to determine the effect of some important process variables on the contact resistance at the metal-silicon interface. The effect of these process variables was
The semiconductor material used in this investigation was typically Czochralski pulled, single crystal silicon. The samples were sawn into blocks 5 x 5 x 2.5 mm such that the 5 x 5 mm faces are the (111) plane. After lapping the 5 mm faces, the samples were chemically polished in a solution of nitric, hydrofluoric and acetic acid before metallic contacts are applied to the opposite (111) faces. Contact application was performed principally by vacuum deposition. Plated and alloyed metals were studied, but the results are not reported here. After the contacts were deposited, the sample was placed in a measurement fixture where a clean indium preform is mechanically pressed onto the contact surfaces deforming the preform to make intimate contact with the evaporated metal contact surface. After the 2.5 mm surface of the sample is polished flush with the mounting fixture, a sinusoidal current from an oscillator is amplified, measured and passed through the sample. A tungsten probe, etched to a very sharp point and mounted rigidly above the assembly, is lowered onto the sample to sense the voltage developed at any particular point along the sample. A micrometer table capable of 0.01 mm resolution is moved laterally in 0.10 mm increments, and the
832
NOTES
voltage is recorded at each step. This probe excursion is made normal to the plane of the contact interface. By using a low-noise, battery-powered, low-level preamplifier and carefully shielding all wiring, voltages as low as 2 x 10-7 V can be routinely detected. The indium-metal contact voltage drop is at least one order of magnitude lower than this and does not contribute to the measurement appreciably. By graphically plotting the measured voltage vs. the distance along the sample and extrapolating the line to the sample edge, the voltage across the interfacial barrier is determined. From this voltage the contact resistance in Q-cm2 is easily calculated using Ohm’s law. All measured values of contact resistance noted below are the average of at least four measurements. Generally, the values noted were reproducible to + 20 per cent.
resistance. On the other hand, a dilute hydrofluoric acid rinse which reduced the surface oxide thickness to a minimum resulted in a much lower value of contact resistance. For example, using aluminum contacts, the value of Rc was a factor of 50 lower after a hydrofluoric acid treatment than when nitric acid was employed. A similar effect was noted for the cases of nickel and molybdenum. One note with regard to this particular area of investigation was the necessity to shield the samples from the evaporating charge until vaporization of the metal charge was initiated. Unless this precaution was taken, the measured values of R, were scattered (usually increased R, resulted), indicating that impurities on the charge (possibly adsorbed organic material) were being driven onto the sample. Evaporation temperature The effect of substrate temperature during metal deposition was studied briefly using aluminum as the contacting metal. Substrate temperatures were varied from 25 to 600°C with only small variations in the values of R, resulting, but it is interesting to note that for 0.002 Q-cm p-type silicon the lowest value of R, (4.1 x 10-S Q-cm”) was found for the case in which the substrate temperature was in excess of the silicon-aluminum eutectic temperature (577°C).
RESULTS Pre-contact cleanup of the silicon surface It was found that the preparation of the silicon surface prior to metal deposition had a marked effect on the resulting value of contact resistance. A variety of pre-evaporation acid treatments and rinses were studied, and it was clear from the results that the effect of a relatively thick surface oxide, as would be present after a hot nitric or sulfuric acid treatment, is to increase the contact Table 1. Contact resistance in n-cm2
Si resistivity in R-cm and cond. type 0.002 p 0.005 p O,OlOp 0.050 p 0.10 p 0.25 p 0.50 p
3.7 x 10-5 5.2 x 10-s
0.005 0010 0.05 0.10 0.50
4.1 x 10-3 8.3 x 10-s 6.4 x10-s
n 11 ?I n n
V
Al
6.0 5.4 1.3 1.1
x lo-* x 10-h x 10-3 x lo--3
2.9 x 10-r
5.2 1.3 7.4 6.4 1.8
x 10-s x10-s x 10-s x 10-s x 10-r
of evaporated metal contacts to silicon.*
MO
Ni
Chrome1
Co
4.4 x10-s 1.4 x 10-j
7.8 x 10 -6 2.2 x 10-j 2.1 x10-s
1.3 x 10-a 1.1 x 10-r
2.3x10-j 4-2 x lo--’
2.2 x 10-2 1.4 7.3 2.6 2.1 1.1 2.9
9.4 x10-s x 10-5 x 10-s x 10-r x 10-r
7.8 x 10-s 6.1 x 10-l 2.0
1.8~10 7.0 x 10-d 5.6 x 1O-3 7.0 x10-s
a
1.0 x 10-4 1.4 x 10-a 1.6 x 10-r 1.0
2.2 x 10-s
26
* Samples were chemically polished and soaked in HF. substrate temperatures were 250-300°C.
The evaporations
were performed
at i 10-s Torr,
and
833
NOTES
Post evaporation heat treatment The effect of sintering on R, was studied for several metal-silicon couples. For the metals aluminum, silver, chromium, and molybdenum, sintering at various times and temperatures in nitrogen produced only small changes (generally less than 50 per cent) in the measured values of the contact resistance. On the other hand, sintering produced considerable lowering of R, for the metals palladium and nickel. For example, in the case of nickel contacts the value of R, decreased from 7.0 x 10-4 to 3.8 x 10-4 to 1.8 x 10-5 Q-cm2 for the conditions of no sintering, 450°C for 6 min, 550°C for 6 min respectively. Effects of silicon resistivity and the particular metals used as contacts Several different metals were evaporated onto both n- and p-type silicon. The silicon substrates included a wide range of resistivities, and Rc varied almost seven orders of magnitude between 4.4 x 10-s R-cm2 and 26.0 LLcm2. The results of the measurements are given in Table 1, where it can be seen that R, is highly dependent on the resistivity of the silicon regardless of the particular metal contact. It is interesting to note most of the metals studied yielded low resistance contacts if silicon was highly doped. The values of Rc obtained also indicate the difficulty encountered when attempting to construct low-resistance, ohmic contacts to n-type silicon of 0.01 Q-cm and above. During a measurement to n-type silicon of 0.01 Q-cm and above, the voltage across the contact interface was observed to be rectified, the degree of rectification increasing with resistivity. Since R, then would depend on the direction of current flow, the values of R, in Table 1 of the higher n-type resistivities are not accurate. R. C. HOOPER Semiconductor Research and J. A. CUNNINGHAM J. G. HARPER Development Laboratory Texas Instruments Incorporated Dallas, Texas References 1. H. K. HENISCH, Sylvan. Tech. IX, 73 (1956). 2. M. V. SULLIVAN and J. H. EIGLER, J. Eleclrochem. Sot. 104, 226 (1957): 3. 0. J. MENCALI and M. R. SEILER, Adwznc. Energy Conv. 2, 59 (1962).
Solid-State Electronics pp. 833-835.
Pergamon Press 1965. Vol. 8, Printed in Great Britain
Electra-optic observation effects in gallium
of space arsenide
charge
(Received 3 May 1965) A VISUAL examination of non-uniform electric fields in high resistivity GaAs has been made using the linear electro-optic effect. The nonuniform fields arise from space charge effects in the compensated semiconductor. Above a critical voltage, the space charge distribution becomes unstable and a high field region moves across the sample from cathode to anode, accompanied by oscillations in the current.(l32) The motion of the high field region is strikingly demonstrated by the electro-optic effect. This effect is observed by placing the sample between crossed polarizers and illuminating it with light of wavelength longer than the absorption edge (Fig. 1). The transmitted infrared light is viewed with an image converter. In zero field the sample appears uniformly dark. When a field is applied, the sample becomes birefringentts) and light is transmitted. The field distribution is mapped by the intensity variations over the sample. Any strain present in the sample will cause light to be transmitted in zero fields due to the photoelastic effect.(a) It should be noted that the method described here differs from a previous electro-optic method in which the Franz-Keldysh effect was used to study space charge effects in CdS.c4) No effect could be seen in GaAs without the polarizers. The electro-optic effect for crystals of class 33rn such as GaAs has been discussed in detail by SAMBA. In general the crystal becomes biaxial on application of the field and the phase difference I’ between the extraordinary and ordinary rays depends on both the direction of the field and the direction of viewing. However, if the field is applied along a (111) direction, the crystal becomes uniaxial with axis parallel to the field, and for any direction of viewing perpendicular to the field E the phase difference is I?=
n2/3 -lnir41E x
where no is the index
of refraction,
~41 is the