- Email: [email protected]
Application of epitaxial CoSi2⧸Si⧸CoSi2 heterostructures to tunable Schottky-barrier detectors
Journal of Crystal Growth 127 (1993) 659—662 North-Holland
~
CRYSTAL GROWT H
Application of epitaxial CoSi 2/ Si/CoSi2 heterostructures to tunable Schottky-barrier detectors C. Schwarz, U. Schärer, P. Sutter, R. Stalder, N. Onda and H. von Känel Laboratorium für Festkörperphysik, ETH, Honggerberg, CH-8093 Zurich, Switzerland Epitaxial CoSi2/Si/CoSi2 heterostructures on Si(111) have been grown by MBE. STM measurements during the growth process showed suicide layers below 45 A to be coherent apart from dislocations associated with the wafer misorientation, and the 1000 A thick Si spacer to be pinhole-free. Diode structures were fabricated by standard photolithography and a combination of wet chemical and plasma etching. Both the upper and the buried suicide film could be contacted separately for electrical measurements. Vertical transport measurements showed blocking behaviour confirming the Si spacer layers to be continuous. The photoresponse of the diode structures was measured as a function of bias at 77 K. A maximum shift of the cutoff energy of 0.1 eV was obtained at an electrical field of 200 kV cm
1. Introduction Schottk~’-barrierinfrared (IR) detectors on Si are of current interest for application in JR imaging. IrSi/Si and PtSi/Si detector arrays with a response range between 1 and 10 i~mhave been integrated in cameras [1,2]. The principle of a Schottky-barrier IR detector is based on internal photoemission at the metal—semiconductor interface. Usually the cutoff wavelength of a Schottky barrier can be tuned over very limited range (—~50 meV [3]) by an applied electric field. This is due to the Schottky effect [3]. Using a metal— semiconductor—metal structure with a semiconductor thin enough to be fully depleted has the advantage of a broader tuning range of the cutoff wavelength, since the minimum electric field is zero at zero bias. In this work we discuss the first successful attempt to fabricate and measure such metal—semiconductor—metal sandwich structures from epitaxial CoSi2/Si/CoSi2 heterostructures on Si(100). Our results bear a close relationship to the ones obtained recently by Pahun et al. on Pt/Si/ErSi17 heterostructures [4]. 2. Growth and characterization The substrates used were Si wafers (n-doped, 1—2000 (2 cm), (111) oriented within better than 0022-0248/93/$06.00 © 1993
—
0.1°—0.3°. Substrate preparation and suicide growth have been described in detail elsewhere [5,61. Of particular relevance for this work are parallel monolayer steps on the substrate surface associated with the unintentional wafer misorientation. This step structure can be preserved in the silicide by using a template method in which a 10 A thin CoSi2 film is first deposited at room temperature (RT) [5—7].After a short anneal of this template to 350°C the final thickness was reached by MBE below 100°C followed by annealing to 550°C. The film thickness was kept below the critical thickness h~ 45 A [5] for biaxial strain relaxation. In situ STM as well as TEM images of these CoSi2 films confirmed that below h~they are coherent apart from partial dislocations associated with the wafer misorientation. They exhibit, thus, uniaxial strain relaxation. Overgrowth with Si on a totally strained silicide would give the best result due to perfect lattice matching. Because of the very large surface diffusion of Si on CoSi2, Si overgrowth was started with 5 A deposited at RT. Then Si was grown to the desired thickness in a cycle of deposition and annealing at gradually increasing substrate temperature with the last anneal just above the 7 x 7 —s 1 x 1 transition. Si films with thicknesses up to 1300 A were grown. STM studies showed that the surface step structure on these films is again
Elsevier Science Publishers B.V. All rights reserved
660
C. Schwarz et al.
/ Application of epitaxial
_____
~iJ~
CoSi
2 / Si / CoSi2 to tunable Schottky-barrier detectors
°~
~
incident RF power of 200 W [9]. Si etch rates varied between 25 and 30 A s~ depending on the quality of the Si. Under these etching conditions the buried suicide. CoSi2 Contacts is etched the rates CoSi2 in in the layer range were of 1. This largetoat difference etch rates 0.7 to 3 Ainsa clear etch stop, when reaching the resulted applied by lift-off of sputtered Cr/Au (50/2000 A). Electrical insulation to the Si substrate was provided by applying a 2 ~rm thick polymer coating before the metallization step (see insert fig.
2). 4. Electrical measurements I
1j.tm
Fig. 1. STM image of 14 A thick CoSi2 film overgrown with 1050 A Si. Final annealing temperature was 810°C. Parallel monolayer steps in the direction of the wafer misorientation (0.14°towards 2 pictures could [121])there are clearly be detected resolved. any In pinholes, none ofwhereas such 4 ~im some threading dislocations could always be seen.
Electrical resistivities of both CoSi 2 layers were determined by four-point measurements from RT down to 4 K. Although they both showed metallic behaviour the residual resistivities of the lower suicide, ured with e.g.the5.9upper ~i 12 CoSi cm for a 42 A film (meas2 removed) turned out to be higher than previously reported [10,11]. We associate this with the higher annealing temperatures used in the present case which may lead to
determined by the wafer misorientation. Moreover, the Si films appeared pinhole-free (fig. 1). In addition, however, they contain a 2) highprobably density of threading dislocations (> 108 cm due to the uniaxial strain relaxation of the underlying suicide. Overgrowth of Si with CoSi 2 was carried out in the same way as for the first layer.
3. Preparation of metal—semiconductor—metal
structures ments, Forsilicide test the structures electric and were photoelectric fabricatedseparately. such measurethat both layers could be contacted Circular diode structures with diameters varying between 100 and 1000 ~ were prepared by using standard photolithographical techniques. A mesa structure was etched with a HF—CrO3—H2O etchant [81.The CoSi2 on top was removed with a BHF etch. The intermediate Si was plasma-etched in a CF4/02 atmosphere (—‘ 80 mTorr with 96 mass% of CF4 and 4 mass% of 02) at RT with
6 106
CoSi~
3 10.6
~U
‘Cr/Au’
~i
__________________ _____________________
‘
I
insu ator
0
-3 106 -6 106
----77 — RI K 0 1 2 3 4 voltage [VI Fig. 2. Current—voltage curves of a CoSi2 /Si/CoSi2 heterostructure with a 1300 A thick Si spacer layer at RT and 77 K (diode diameter 100 jim) The blocking behaviour and the positive temperature coefficient cross of current shown. The insert shows a schematic sectiononset of aare processed heterostructure. The silicides were contacted with Cr/Au which is insulated by a polymer against the Si substrate. .‘~
~4
3
-2
-1
C. Schwarz et a!.
/ Application of epitaxial
CoSi
Si excess in the films. Current—voltage measurements across the CoSi2/Si/CoSi2 heterostructures were made at various temperatures between RT and 4 K. I—V curves (fig. 2) exhibited blocking behaviour with a nearly exponential increase of the current at about ±3 V for a 1300 A thick Si spacer layer at 77 K. Current densities in the region of the plateaus in fig. 2 were at least three orders of magnitude higher than the reverse current densities expected for a simple CoSi2/Si Schottky barrier. This can possibly be explained by leakage along threading dislocations. In fact the current is proportional to the device area, showing that edge effects can be neglected. Specific contact resistance zero voltage to increased 2 at RTat exponentially 106 (2 from 100 11 cm cm2 at 4 K indicating that the current is not flowing through metallic pinholes which would have the reverse effect. The voltage at which the current begins to increase exponentially has a negative temperature coefficient in contrast to the behaviour expected for avalanche breakdown. 5. Photoelectrical measurements
2 / Si / CoSi2 to tunable Schottky-barrier detectors
661
5
~
V
4
Si
C
~
3
~Si.
9 0
,~
~
i
2.OV 1.5 V p0.5 ~
a 0.5
0.6
0.7 0.8 energy [eV]
0.9
1
Fig. 3. Fowler plots obtained on a heterostructure consisting of 42—45 A thick suicide layers and a 1300 A thick Si spacer layer. voltages upcutoff to 2.5energy V wereofapplied. clear of theBias extrapolated at leastA0.1 eVdecrease can be seen. A schematic sketch of the band diagram of the heterostructure, with the bias condition used for the measurements, is shown in the insert. The applied bias leads to Schottky barrier lowering. Hot electrons and holes can both contribute to the photocurrent.
behaviour was found. For higher bias the current increases rapidly (fig. 2), which makes the mea-
The photoelectric response of the Schottkybarrier diodes was measured at RT and 77 K. The spectrum of a halogen lamp which passed a single prism monochromator was calibrated with a pyroelectric detector. The photocurrent was measured with lock-in technique. Schottky-barrier heights were obtained from Fowler-plots [12]. The Schottky effect predicts a decrease of the Schottky-barrier by ______
=
47rEd
(1)
with a voltage V applied across the Si layer of thickness d ( = dielectric permittivity) [4]. Since the latter must be assumed to be fully depleted, the constant electric field is given by V/d. In fig. 3 are shown typical Fowler plots obtained at 77 K on a heterostructure consisting of 42—45 A thick suicide layers separated by a 1300 A thick Si barrier. The bias conditions are indicated in the insert of fig. 3. For reversed voltages a similar
surement of the small photocurrent impossible. Because of the weak absorption of the thin silicide films, hot electrons and holes are generated in both of them [13]. The barrier for holes is lower (0.55 eV) than the one for electrons and independent of the temperature [14]. By contrast, the barrier for electrons was found to increase from 0.65 to 0.70 eV at 77 K [14]. For our symmetric heterostructures there is no way to distinguish unambiguously between hot electron and hot hole contributions to PB. (In principle an asymmetry due to the different structures of the two CoSi 2/Si interfaces [11] cannot be exeluded, although there does not exist any evidence for significantly different barrier heights at type A and type B interfaces). The measured decrease of the barrier height deduced from the Fowler plots of fig. 3 is actually larger than expected from eq. (1). One possible explanation could be fluctuations in the field strength across the heterostructures for which, however, we have no direct evidence.
662
C. Schwarz et at.
/ Application of epitaxial CoSi2 / Si / CoSi2
6. Conclusion We determined the cutoff energy for internal photoemission of CoSi2/Si/CoSi2 heterostructures as a function of bias voltage. The photothreshold could be tuned from 0.65 eV down to 0.55 eV by applying a maximum bias across the 1300 A thick Si spacer layer of 2.5 V. This is appreciably higher than the Schottky barrier lowering dud to the electric field in a conventional Schottky diode.
Acknowledgements We would like to express our sincere thanks to Professor Melchior and his group for their valuable technical assistance for device fabrication. Financial support by the FSRM is also gratefully acknowledged.
to tunable Schottky-barrier detectors
[2] J. Silverman, J.M. Mooney and F.D. Shepherd, Sci. Am. 3 (1992). [3] S.M. Sze, C.R. Crowell and D. Khang, J. AppI. Phys. 35 (1964) 2534. [4] L. Pahun, Y. Campidelli, F. Arnaud d’Avitaya and P.A. Badoz, Appl. Phys. Letters 60(1992)1166. [5] R. Stalder, H. Sirringhaus, N. Onda and H. von Känel, Surface Sci. 258 (1991) 153. [6] R. Stalder, N. Onda, H. Sirringhaus H. von Känel and C.W.T. Bulle-Lieuwma, J. Vacuum Sd. Technol. B 9 (199i~2307. [7] R.T. Tong and F. Schrey, AppI. Phys. Letters 54 (1989) 852. [8] E. Sirtl and A. Adler, Z. Metallk. 52 (1961) 529. [9] W. Hansch, Dissertation, Universität der Bundeswehr, München (1991). [10] H. von Kane! and G. Fishman, Phys. Rev. B 45 (1992) 3929. [11] H. von Känel, J. Henz, M. Ospelt, J. Hugi, F. Muller and N. Onda, Thin Solid Films 184 (1990) 295. [12] J.Y. RH. Duboz, Fowler, P.A. Phys.Badoz, Rev. 38J. (1931) [13] Henz 45. and H. von Känel, J. AppI. Phys 68 (1990) 2346. [14] J.Y. Duboz, P.A. Badoz, F. Arnaud d’Avitaya and E. Rosencher, Phys. Rev. B 40 (1989) 10607.
References [1] B.Y. Tsaur, M.J. McNutt, R.A. Bredthauer and R.B. Mattson, IEEE Electron Device Letters EDL-10 (1989) 362.
~
CRYSTAL GROWT H
Application of epitaxial CoSi 2/ Si/CoSi2 heterostructures to tunable Schottky-barrier detectors C. Schwarz, U. Schärer, P. Sutter, R. Stalder, N. Onda and H. von Känel Laboratorium für Festkörperphysik, ETH, Honggerberg, CH-8093 Zurich, Switzerland Epitaxial CoSi2/Si/CoSi2 heterostructures on Si(111) have been grown by MBE. STM measurements during the growth process showed suicide layers below 45 A to be coherent apart from dislocations associated with the wafer misorientation, and the 1000 A thick Si spacer to be pinhole-free. Diode structures were fabricated by standard photolithography and a combination of wet chemical and plasma etching. Both the upper and the buried suicide film could be contacted separately for electrical measurements. Vertical transport measurements showed blocking behaviour confirming the Si spacer layers to be continuous. The photoresponse of the diode structures was measured as a function of bias at 77 K. A maximum shift of the cutoff energy of 0.1 eV was obtained at an electrical field of 200 kV cm
1. Introduction Schottk~’-barrierinfrared (IR) detectors on Si are of current interest for application in JR imaging. IrSi/Si and PtSi/Si detector arrays with a response range between 1 and 10 i~mhave been integrated in cameras [1,2]. The principle of a Schottky-barrier IR detector is based on internal photoemission at the metal—semiconductor interface. Usually the cutoff wavelength of a Schottky barrier can be tuned over very limited range (—~50 meV [3]) by an applied electric field. This is due to the Schottky effect [3]. Using a metal— semiconductor—metal structure with a semiconductor thin enough to be fully depleted has the advantage of a broader tuning range of the cutoff wavelength, since the minimum electric field is zero at zero bias. In this work we discuss the first successful attempt to fabricate and measure such metal—semiconductor—metal sandwich structures from epitaxial CoSi2/Si/CoSi2 heterostructures on Si(100). Our results bear a close relationship to the ones obtained recently by Pahun et al. on Pt/Si/ErSi17 heterostructures [4]. 2. Growth and characterization The substrates used were Si wafers (n-doped, 1—2000 (2 cm), (111) oriented within better than 0022-0248/93/$06.00 © 1993
—
0.1°—0.3°. Substrate preparation and suicide growth have been described in detail elsewhere [5,61. Of particular relevance for this work are parallel monolayer steps on the substrate surface associated with the unintentional wafer misorientation. This step structure can be preserved in the silicide by using a template method in which a 10 A thin CoSi2 film is first deposited at room temperature (RT) [5—7].After a short anneal of this template to 350°C the final thickness was reached by MBE below 100°C followed by annealing to 550°C. The film thickness was kept below the critical thickness h~ 45 A [5] for biaxial strain relaxation. In situ STM as well as TEM images of these CoSi2 films confirmed that below h~they are coherent apart from partial dislocations associated with the wafer misorientation. They exhibit, thus, uniaxial strain relaxation. Overgrowth with Si on a totally strained silicide would give the best result due to perfect lattice matching. Because of the very large surface diffusion of Si on CoSi2, Si overgrowth was started with 5 A deposited at RT. Then Si was grown to the desired thickness in a cycle of deposition and annealing at gradually increasing substrate temperature with the last anneal just above the 7 x 7 —s 1 x 1 transition. Si films with thicknesses up to 1300 A were grown. STM studies showed that the surface step structure on these films is again
Elsevier Science Publishers B.V. All rights reserved
660
C. Schwarz et al.
/ Application of epitaxial
_____
~iJ~
CoSi
2 / Si / CoSi2 to tunable Schottky-barrier detectors
°~
~
incident RF power of 200 W [9]. Si etch rates varied between 25 and 30 A s~ depending on the quality of the Si. Under these etching conditions the buried suicide. CoSi2 Contacts is etched the rates CoSi2 in in the layer range were of 1. This largetoat difference etch rates 0.7 to 3 Ainsa clear etch stop, when reaching the resulted applied by lift-off of sputtered Cr/Au (50/2000 A). Electrical insulation to the Si substrate was provided by applying a 2 ~rm thick polymer coating before the metallization step (see insert fig.
2). 4. Electrical measurements I
1j.tm
Fig. 1. STM image of 14 A thick CoSi2 film overgrown with 1050 A Si. Final annealing temperature was 810°C. Parallel monolayer steps in the direction of the wafer misorientation (0.14°towards 2 pictures could [121])there are clearly be detected resolved. any In pinholes, none ofwhereas such 4 ~im some threading dislocations could always be seen.
Electrical resistivities of both CoSi 2 layers were determined by four-point measurements from RT down to 4 K. Although they both showed metallic behaviour the residual resistivities of the lower suicide, ured with e.g.the5.9upper ~i 12 CoSi cm for a 42 A film (meas2 removed) turned out to be higher than previously reported [10,11]. We associate this with the higher annealing temperatures used in the present case which may lead to
determined by the wafer misorientation. Moreover, the Si films appeared pinhole-free (fig. 1). In addition, however, they contain a 2) highprobably density of threading dislocations (> 108 cm due to the uniaxial strain relaxation of the underlying suicide. Overgrowth of Si with CoSi 2 was carried out in the same way as for the first layer.
3. Preparation of metal—semiconductor—metal
structures ments, Forsilicide test the structures electric and were photoelectric fabricatedseparately. such measurethat both layers could be contacted Circular diode structures with diameters varying between 100 and 1000 ~ were prepared by using standard photolithographical techniques. A mesa structure was etched with a HF—CrO3—H2O etchant [81.The CoSi2 on top was removed with a BHF etch. The intermediate Si was plasma-etched in a CF4/02 atmosphere (—‘ 80 mTorr with 96 mass% of CF4 and 4 mass% of 02) at RT with
6 106
CoSi~
3 10.6
~U
‘Cr/Au’
~i
__________________ _____________________
‘
I
insu ator
0
-3 106 -6 106
----77 — RI K 0 1 2 3 4 voltage [VI Fig. 2. Current—voltage curves of a CoSi2 /Si/CoSi2 heterostructure with a 1300 A thick Si spacer layer at RT and 77 K (diode diameter 100 jim) The blocking behaviour and the positive temperature coefficient cross of current shown. The insert shows a schematic sectiononset of aare processed heterostructure. The silicides were contacted with Cr/Au which is insulated by a polymer against the Si substrate. .‘~
~4
3
-2
-1
C. Schwarz et a!.
/ Application of epitaxial
CoSi
Si excess in the films. Current—voltage measurements across the CoSi2/Si/CoSi2 heterostructures were made at various temperatures between RT and 4 K. I—V curves (fig. 2) exhibited blocking behaviour with a nearly exponential increase of the current at about ±3 V for a 1300 A thick Si spacer layer at 77 K. Current densities in the region of the plateaus in fig. 2 were at least three orders of magnitude higher than the reverse current densities expected for a simple CoSi2/Si Schottky barrier. This can possibly be explained by leakage along threading dislocations. In fact the current is proportional to the device area, showing that edge effects can be neglected. Specific contact resistance zero voltage to increased 2 at RTat exponentially 106 (2 from 100 11 cm cm2 at 4 K indicating that the current is not flowing through metallic pinholes which would have the reverse effect. The voltage at which the current begins to increase exponentially has a negative temperature coefficient in contrast to the behaviour expected for avalanche breakdown. 5. Photoelectrical measurements
2 / Si / CoSi2 to tunable Schottky-barrier detectors
661
5
~
V
4
Si
C
~
3
~Si.
9 0
,~
~
i
2.OV 1.5 V p0.5 ~
a 0.5
0.6
0.7 0.8 energy [eV]
0.9
1
Fig. 3. Fowler plots obtained on a heterostructure consisting of 42—45 A thick suicide layers and a 1300 A thick Si spacer layer. voltages upcutoff to 2.5energy V wereofapplied. clear of theBias extrapolated at leastA0.1 eVdecrease can be seen. A schematic sketch of the band diagram of the heterostructure, with the bias condition used for the measurements, is shown in the insert. The applied bias leads to Schottky barrier lowering. Hot electrons and holes can both contribute to the photocurrent.
behaviour was found. For higher bias the current increases rapidly (fig. 2), which makes the mea-
The photoelectric response of the Schottkybarrier diodes was measured at RT and 77 K. The spectrum of a halogen lamp which passed a single prism monochromator was calibrated with a pyroelectric detector. The photocurrent was measured with lock-in technique. Schottky-barrier heights were obtained from Fowler-plots [12]. The Schottky effect predicts a decrease of the Schottky-barrier by ______
=
47rEd
(1)
with a voltage V applied across the Si layer of thickness d ( = dielectric permittivity) [4]. Since the latter must be assumed to be fully depleted, the constant electric field is given by V/d. In fig. 3 are shown typical Fowler plots obtained at 77 K on a heterostructure consisting of 42—45 A thick suicide layers separated by a 1300 A thick Si barrier. The bias conditions are indicated in the insert of fig. 3. For reversed voltages a similar
surement of the small photocurrent impossible. Because of the weak absorption of the thin silicide films, hot electrons and holes are generated in both of them [13]. The barrier for holes is lower (0.55 eV) than the one for electrons and independent of the temperature [14]. By contrast, the barrier for electrons was found to increase from 0.65 to 0.70 eV at 77 K [14]. For our symmetric heterostructures there is no way to distinguish unambiguously between hot electron and hot hole contributions to PB. (In principle an asymmetry due to the different structures of the two CoSi 2/Si interfaces [11] cannot be exeluded, although there does not exist any evidence for significantly different barrier heights at type A and type B interfaces). The measured decrease of the barrier height deduced from the Fowler plots of fig. 3 is actually larger than expected from eq. (1). One possible explanation could be fluctuations in the field strength across the heterostructures for which, however, we have no direct evidence.
662
C. Schwarz et at.
/ Application of epitaxial CoSi2 / Si / CoSi2
6. Conclusion We determined the cutoff energy for internal photoemission of CoSi2/Si/CoSi2 heterostructures as a function of bias voltage. The photothreshold could be tuned from 0.65 eV down to 0.55 eV by applying a maximum bias across the 1300 A thick Si spacer layer of 2.5 V. This is appreciably higher than the Schottky barrier lowering dud to the electric field in a conventional Schottky diode.
Acknowledgements We would like to express our sincere thanks to Professor Melchior and his group for their valuable technical assistance for device fabrication. Financial support by the FSRM is also gratefully acknowledged.
to tunable Schottky-barrier detectors
[2] J. Silverman, J.M. Mooney and F.D. Shepherd, Sci. Am. 3 (1992). [3] S.M. Sze, C.R. Crowell and D. Khang, J. AppI. Phys. 35 (1964) 2534. [4] L. Pahun, Y. Campidelli, F. Arnaud d’Avitaya and P.A. Badoz, Appl. Phys. Letters 60(1992)1166. [5] R. Stalder, H. Sirringhaus, N. Onda and H. von Känel, Surface Sci. 258 (1991) 153. [6] R. Stalder, N. Onda, H. Sirringhaus H. von Känel and C.W.T. Bulle-Lieuwma, J. Vacuum Sd. Technol. B 9 (199i~2307. [7] R.T. Tong and F. Schrey, AppI. Phys. Letters 54 (1989) 852. [8] E. Sirtl and A. Adler, Z. Metallk. 52 (1961) 529. [9] W. Hansch, Dissertation, Universität der Bundeswehr, München (1991). [10] H. von Kane! and G. Fishman, Phys. Rev. B 45 (1992) 3929. [11] H. von Känel, J. Henz, M. Ospelt, J. Hugi, F. Muller and N. Onda, Thin Solid Films 184 (1990) 295. [12] J.Y. RH. Duboz, Fowler, P.A. Phys.Badoz, Rev. 38J. (1931) [13] Henz 45. and H. von Känel, J. AppI. Phys 68 (1990) 2346. [14] J.Y. Duboz, P.A. Badoz, F. Arnaud d’Avitaya and E. Rosencher, Phys. Rev. B 40 (1989) 10607.
References [1] B.Y. Tsaur, M.J. McNutt, R.A. Bredthauer and R.B. Mattson, IEEE Electron Device Letters EDL-10 (1989) 362.