- Email: [email protected]
Hydrogen cleaning of silicon wafers. Investigation of the wafer surface after plasma treatment
Thin Solid Films, 228 (1993) 23-26
23
Hydrogen cleaning of silicon wafers. Investigation of the wafer surface after plasma treatment J. R a m m , E. B e c k a n d A. Z u e g e r Balzers AG, FL-9496 Balzers (Liechtenstein)
A. Dommann Neu-Technikum, CH-9470 Buchs (Switzerland)
R. E. Pixley Physik-Institut der Universitat, CH-8001 Ziirich (Switzerland)
Abstract A single-step cleaning procedure with a newly developed ultrahigh vacuum (UHV) compatible plasma source is described. Utilizing this source, an argon-hydrogen d.c. discharge between the heated filament (cathode) and the grounded chamber walls (anode) of an UHV system is established. The discharge is characterized by high currents (up to 100 A) and low voltages (20 V-35 V). Without additional wet chemical cleaning steps, the silicon wafer as obtained from the manufacturer is cleaned by exposure to the plasma. The influence of the plasma treatment on the hydrogen content and the perpendicular strain profile of the wafer are investigated by elastic recoil detection, Rutherford baekscattering channelling and high resolution X-ray rocking curve diffraction measurements.
1. Introduction
2. Experiments
The development of dry wafer cleaning procedures has been greatly stimulated by the advanced silicon technology need for low temperature epitaxial growth. As a promising candidate for such a process, the hydrogen plasma has been investigated by several groups [ 1-4]. It has been shown that excited or ionized hydrogen is able to remove oxygen as well as carbon contaminants from the wafer surface and that low temperature epitaxial growth can be achieved using this cleaning procedure. Some o f these procedures, however, suffer from low etch rates and need to be scaled up for production. In previous papers [5-7] a low energy argon-hydrogen gas discharge has been described in which the etch rates for thermally grown silicon dioxide, diamond-like carbon, boron, and amorphous silicon were determined. Further, it was shown that a short single exposure to the plasma prepares the silicon wafer surface for homoepitaxial growth at substrate temperatures down to 300 °C on (100) and 400 °C on (111) silicon. In the present work we investigate how this cleaning procedure influences the surface properties of the wafer. The hydrogen content and the perpendicular strain profiles of (111) silicon wafers are studied for different annealing temperatures after plasma Cleaning.
2.1. Plasma cleaning process description
0040-6090/93/$6.00
Figure 1 shows the ultrahigh vacuum (UHV) plasma cleaning module. Before the silicon wafer is brought into the chamber, the plasma discharge is initiated in order to condition the chamber walls. It is essential for good etch reproducibility that the gas desorption from the walls be stable during the etch process as the gas composition of the plasma strongly influences the
8-
[H2
2 ._L
I
Ar, H2___L
"i
5
Fig. 1. Experimental configuration consisting of the Ioad-I~k (1) end the UHV plasma cleaning module (2) with the attached UHV plasma source (4).
© 1 9 9 3 - Elsevier Sequoia. All rights reserved
J. Ramm et aL / H cleaning of Si wafers
24
reactions on the wafer surface. Minor additions of oxygen, for example, can result in a reoxidation of the wafer surface and limit the cleaning result. Ultraclean gas processing is required for the work with this highly reactive plasma at low temperatures. A UHV version of the plasma source was one prerequisite for successful substrate cleaning. For processing, the silicon wafer is placed in the load-lock (1) from which it is transferred to the cleaning module (2). This module contains a grounded substrate holder (3) and the plasma source (4). The source consists of a heated filament (5) placed in a separate cavity (6). An orifice connects the cavity to the chamber. A working gas (usually argon) is fed into the cavity at 7. For cleaning, the reactive gas (hydrogen) is fed into the source at 7 or directly into the chamber from the wafer back side at 8. A gas discharge is established by applying a potential between filament and ground. For a voltage of typically 30 V, electron currents between I0 A and 100 A from the power supply (9) can easily be obtained. The cleaning module can be connected to a molecular beam epitaxy (MBE) growth chamber via transport channel (10). The details of such a low temperature MBE process are given elsewhere [5]. The whole cleaning procedure consists of the substrate exposure to the plasma for only a few minutes. For the present measurements, (111) n-type (phosphorus doped, 1-10 f~ cm, Czochralski) silicon wafers were used. An argon-hydrogen flow of 18 standard cm3min-l-20 standard cm3min -~ and a current of 30 A (at 30 V) was chosen for the discharge. The cleaning duration for the substrates was 20 min, about 5 times longer than that needed to remove the native oxide from the surface. The longer cleaning time was employed in order to obtain more drastic effects. Some of the wafers (see Table 1) were annealed for 15 min after plasma exposure at different temperatures using the heater of the MBE chamber. All measurements of sample properties were performed ex situ. 2.2. Analytical methods
The hydrogen content of the wafer was profiled by elastic recoil detection [8] using 2 MeV He + and a forward recoil angle of 30 °. These scattered helium TABLE 1. Results of Rutherford backscatteringmeasurements Sample
Symbol (Fig. 2)
Annealing temperature
Integrated H content
Zmi.
1 2 3 4 5
• (3 • []
No anneal 400 °C 500 °C 600 °C Untreated wafer
5027 3798 2540 1897 155
17.6(9)% 11.5(6)% 5.2(4)% 3.3(2)% 2.8(2)%
particles were stopped by a Mylar foil. The hydrogen build-up during He + bombardment was corrected by two measurements on the same beam spot. Channelling experiments were performed to determine the crystal quality of the first 500 nm of the wafer surface. The value of Xminwas obtained from the minimum yield of the channelled spectrum compared with the random spectrum. The random spectrum was measured by rotational averaging over different sample directions. To obtain complementary information about the crystallinity of the wafer surface and the perpendicular strain profile, high resolution X-ray diffraction analysis was performed. The measurements were carried out on a Philipps MPD 1880-HR diffractometer for the symmetric (333) reflections using a step-scan of 0.001 °. The four-crystal monochromator (GaAs crystal, Bartel type) was configured in the 220 setting, providing an intense parallel monochromatic beam of Cu K0q radiation. In order to study the diffuse scattering, high resolution multiple-crystal multiple-reflection diffractometer studies using a single silicon crystal as a onereflection analyser were performed [9]. A set of co-20-scans at different co offsets were measured in a narrow region around the reciprocal lattice point. The narrow scan range of 0.5 ° allows a linear co scan.
3. Results and discussion 3.1. Hydrogen content
In Fig. 2(a) the hydrogen depth profiles for samples 1 - 4 are compared. The unannealed sample 1 exhibits a drastic broadening of the surface peak and an increase in the hydrogen content to a depth of about 150 nm. For annealing temperatures from 400 °C to 600 °C, the integrated hydrogen content decreases by a factor of 2 (as can be seen from the values of the integrated hydrogen content in Table 1). The hydrogen content of the 600 °C sample is still almost an order of magnitude greater than that of the untreated sample (not shown in Fig. 2(a)). The results of the Rutherford backscattering channelling measurements are given in Table 1. ~rnin values for the plasma-cleaned samples range from 17.6% for the unannealed sample down to 3.3% for the 600 °C sample. This is to be compared with 2.8% for an untreated sample. Interstitial hydrogen is obviously causing a strong dechannelling. The annealing at 600 °C removes a large fraction of hydrogen. This results in channelling comparable with that for the untreated sample. One can conclude that no appreciable damage has been produced by the cleaning. This has been verified by low temperature homoepitaxial growth reported in refs. 5-7.
J. Rammet al. / H cleaning of Si wafers 0 1000 -11 I00 Itllllllllllllllllllllllllllll'llllllll 1000
2000
3000
03 -i.-, C +'3 O O
a
c"
(D
25
10r
800
O~9 100
600
;+.1
......................... i............
.
.
.
.
.
.
. . . . . . . . . . . . .
4
cO
400
0 c'<1+) CY~
O
200
"O >, c-
10"-0.1 °0.05 0 O
0.05 0.1
A0 / deg
-- m
v
Fig. 3. Rocking curves from the (333) diffraction plane of (111)-cut silicon wafers for samples 1-5 (read in the direction indicated by the
b
arrow).
0.20
E 0 k_ 030.10
0.00 - 1000
,
0
.
.
. 1000
. O
depth/A
.
. 2000
3000
Fig. 2. (a) Hydrogen depth profiles for samples i-4. The data were obtained for the same experimental set-up and a total charge of 100 laC. (b) Simulated perpendicularstrain profilesextracted from the rocking curves based on the dynamical theory.
3.2. X-ray rocking curve measurements and the simulation of the strain profile The results of the (333) reflection measurements are presented in Fig. 3. The curves belong to samples 1 - 5 in the sequence indicated by the arrow. Nearly no difference in the tail o f the rocking curves can be detected between the unannealed samples and the samples annealed at 400 °C. 500 °C annealing influences mainly the Bragg peak tail for negative (9. For 600 °C, however, a clear narrowing of the tail is found for both regions o f (9. It is known from theory that the strain causes a broadening o f the Bragg peak tail for negative values of (9. In the data shown, a broadening is also observed for positive values due to the diffused scattered background. To obtain the correct strain profile, this has to be taken into account for the strain fit procedure. Triple-axis measurements were made to evaluate the diffused scattered background. The perpendicular strain
profiles were obtained using an improved version of an algorithm given in the literature [10]. The results are given in comparison with the hydrogen depth profiles in Fig. 2(b) as a function o f the same depth scale. A reduction in the hydrogen content between the unannealed sample and the sample annealed at 400 °C is clearly observed. The strain profiles of these two sampies, however, are very close to each other. For higher annealing temperatures, the strain profiles are correlated with the loss of hydrogen in the samples.
4. Conclusions The argon=hydrogen cleaning procedure was studied for an arbitrary set of discharge parameters. The duration of the cleaning was enlarged to study the influences of the plasma on the wafer. F o r this strong treatment, hydrogen was observed to be introduced to the silicon to a depth of about 200 nm. Annealing reduces the hydrogen content as well as the perpendicular strain in the wafer. The promising results from recently obtained low temperature homoepitaxial growth at 300 °C for (100) and 400 °C for (111) silicon [5-7] indicate that the hydrogen may be built in at interstitial sites. With this cleaning procedure silicon wafers can be prepared for low temperature epitaxial growth in a single step.
Acknowledgment The authors would like to thank H. Senn for his valuable support during the cleaning experiments.
26
J. Ramm et aL / H cleaning of Si wafers
References 1 B. Anthony, L. Breaux, T. Hsu, S. Banerjee and A. Tasch, J. Vac. Sci. Technol. B, 7(1989) 621. 2 Y. Nara, Y. Sugita, N. Nakayama and T. Ito, J. Vac. Sci. Technol. B, 10(l) (1992) 274, 3 S. V. Hattangady, R. A. Rudder, M. J. Mantini, G. G. Fountain, J. B. Posthill and R. J. Markunas, J. AppL Phys., 68 (3) (1990) 1233. 4 T.-R. Yew and R. Reif, Mater. Res. Soc. Syrup. Proc., 202 (1991) 401.
5 J. Ramm, E. Beck and A. Zueger, Mater. Res. Soc. Symp. Proc., 220(1991) 15. 6 J. Ramm, E. Beck, F.-P. steiner, R. E. Pixley and I. Eisele, Mater. Res. Soc. Syrup. Proc., 259 (1992) 249. 7 J. Ramm, E. Beck, A. Zueger, A. Dommann and R. E. Pixley, Thin Solid Films, 222 (1992) 126. 8 A. Turos and O. Meyer, Nucl. 1nstruml Methods B, 4 (1984) 92. 9 P. F. Fewster, AppL Surf. Sci., 50(1991)9. 10 V. S. Speriosu and T. Vreeland, Jr., J. Appl. Phys., 59 (1984) 1591.
23
Hydrogen cleaning of silicon wafers. Investigation of the wafer surface after plasma treatment J. R a m m , E. B e c k a n d A. Z u e g e r Balzers AG, FL-9496 Balzers (Liechtenstein)
A. Dommann Neu-Technikum, CH-9470 Buchs (Switzerland)
R. E. Pixley Physik-Institut der Universitat, CH-8001 Ziirich (Switzerland)
Abstract A single-step cleaning procedure with a newly developed ultrahigh vacuum (UHV) compatible plasma source is described. Utilizing this source, an argon-hydrogen d.c. discharge between the heated filament (cathode) and the grounded chamber walls (anode) of an UHV system is established. The discharge is characterized by high currents (up to 100 A) and low voltages (20 V-35 V). Without additional wet chemical cleaning steps, the silicon wafer as obtained from the manufacturer is cleaned by exposure to the plasma. The influence of the plasma treatment on the hydrogen content and the perpendicular strain profile of the wafer are investigated by elastic recoil detection, Rutherford baekscattering channelling and high resolution X-ray rocking curve diffraction measurements.
1. Introduction
2. Experiments
The development of dry wafer cleaning procedures has been greatly stimulated by the advanced silicon technology need for low temperature epitaxial growth. As a promising candidate for such a process, the hydrogen plasma has been investigated by several groups [ 1-4]. It has been shown that excited or ionized hydrogen is able to remove oxygen as well as carbon contaminants from the wafer surface and that low temperature epitaxial growth can be achieved using this cleaning procedure. Some o f these procedures, however, suffer from low etch rates and need to be scaled up for production. In previous papers [5-7] a low energy argon-hydrogen gas discharge has been described in which the etch rates for thermally grown silicon dioxide, diamond-like carbon, boron, and amorphous silicon were determined. Further, it was shown that a short single exposure to the plasma prepares the silicon wafer surface for homoepitaxial growth at substrate temperatures down to 300 °C on (100) and 400 °C on (111) silicon. In the present work we investigate how this cleaning procedure influences the surface properties of the wafer. The hydrogen content and the perpendicular strain profiles of (111) silicon wafers are studied for different annealing temperatures after plasma Cleaning.
2.1. Plasma cleaning process description
0040-6090/93/$6.00
Figure 1 shows the ultrahigh vacuum (UHV) plasma cleaning module. Before the silicon wafer is brought into the chamber, the plasma discharge is initiated in order to condition the chamber walls. It is essential for good etch reproducibility that the gas desorption from the walls be stable during the etch process as the gas composition of the plasma strongly influences the
8-
[H2
2 ._L
I
Ar, H2___L
"i
5
Fig. 1. Experimental configuration consisting of the Ioad-I~k (1) end the UHV plasma cleaning module (2) with the attached UHV plasma source (4).
© 1 9 9 3 - Elsevier Sequoia. All rights reserved
J. Ramm et aL / H cleaning of Si wafers
24
reactions on the wafer surface. Minor additions of oxygen, for example, can result in a reoxidation of the wafer surface and limit the cleaning result. Ultraclean gas processing is required for the work with this highly reactive plasma at low temperatures. A UHV version of the plasma source was one prerequisite for successful substrate cleaning. For processing, the silicon wafer is placed in the load-lock (1) from which it is transferred to the cleaning module (2). This module contains a grounded substrate holder (3) and the plasma source (4). The source consists of a heated filament (5) placed in a separate cavity (6). An orifice connects the cavity to the chamber. A working gas (usually argon) is fed into the cavity at 7. For cleaning, the reactive gas (hydrogen) is fed into the source at 7 or directly into the chamber from the wafer back side at 8. A gas discharge is established by applying a potential between filament and ground. For a voltage of typically 30 V, electron currents between I0 A and 100 A from the power supply (9) can easily be obtained. The cleaning module can be connected to a molecular beam epitaxy (MBE) growth chamber via transport channel (10). The details of such a low temperature MBE process are given elsewhere [5]. The whole cleaning procedure consists of the substrate exposure to the plasma for only a few minutes. For the present measurements, (111) n-type (phosphorus doped, 1-10 f~ cm, Czochralski) silicon wafers were used. An argon-hydrogen flow of 18 standard cm3min-l-20 standard cm3min -~ and a current of 30 A (at 30 V) was chosen for the discharge. The cleaning duration for the substrates was 20 min, about 5 times longer than that needed to remove the native oxide from the surface. The longer cleaning time was employed in order to obtain more drastic effects. Some of the wafers (see Table 1) were annealed for 15 min after plasma exposure at different temperatures using the heater of the MBE chamber. All measurements of sample properties were performed ex situ. 2.2. Analytical methods
The hydrogen content of the wafer was profiled by elastic recoil detection [8] using 2 MeV He + and a forward recoil angle of 30 °. These scattered helium TABLE 1. Results of Rutherford backscatteringmeasurements Sample
Symbol (Fig. 2)
Annealing temperature
Integrated H content
Zmi.
1 2 3 4 5
• (3 • []
No anneal 400 °C 500 °C 600 °C Untreated wafer
5027 3798 2540 1897 155
17.6(9)% 11.5(6)% 5.2(4)% 3.3(2)% 2.8(2)%
particles were stopped by a Mylar foil. The hydrogen build-up during He + bombardment was corrected by two measurements on the same beam spot. Channelling experiments were performed to determine the crystal quality of the first 500 nm of the wafer surface. The value of Xminwas obtained from the minimum yield of the channelled spectrum compared with the random spectrum. The random spectrum was measured by rotational averaging over different sample directions. To obtain complementary information about the crystallinity of the wafer surface and the perpendicular strain profile, high resolution X-ray diffraction analysis was performed. The measurements were carried out on a Philipps MPD 1880-HR diffractometer for the symmetric (333) reflections using a step-scan of 0.001 °. The four-crystal monochromator (GaAs crystal, Bartel type) was configured in the 220 setting, providing an intense parallel monochromatic beam of Cu K0q radiation. In order to study the diffuse scattering, high resolution multiple-crystal multiple-reflection diffractometer studies using a single silicon crystal as a onereflection analyser were performed [9]. A set of co-20-scans at different co offsets were measured in a narrow region around the reciprocal lattice point. The narrow scan range of 0.5 ° allows a linear co scan.
3. Results and discussion 3.1. Hydrogen content
In Fig. 2(a) the hydrogen depth profiles for samples 1 - 4 are compared. The unannealed sample 1 exhibits a drastic broadening of the surface peak and an increase in the hydrogen content to a depth of about 150 nm. For annealing temperatures from 400 °C to 600 °C, the integrated hydrogen content decreases by a factor of 2 (as can be seen from the values of the integrated hydrogen content in Table 1). The hydrogen content of the 600 °C sample is still almost an order of magnitude greater than that of the untreated sample (not shown in Fig. 2(a)). The results of the Rutherford backscattering channelling measurements are given in Table 1. ~rnin values for the plasma-cleaned samples range from 17.6% for the unannealed sample down to 3.3% for the 600 °C sample. This is to be compared with 2.8% for an untreated sample. Interstitial hydrogen is obviously causing a strong dechannelling. The annealing at 600 °C removes a large fraction of hydrogen. This results in channelling comparable with that for the untreated sample. One can conclude that no appreciable damage has been produced by the cleaning. This has been verified by low temperature homoepitaxial growth reported in refs. 5-7.
J. Rammet al. / H cleaning of Si wafers 0 1000 -11 I00 Itllllllllllllllllllllllllllll'llllllll 1000
2000
3000
03 -i.-, C +'3 O O
a
c"
(D
25
10r
800
O~9 100
600
;+.1
......................... i............
.
.
.
.
.
.
. . . . . . . . . . . . .
4
cO
400
0 c'<1+) CY~
O
200
"O >, c-
10"-0.1 °0.05 0 O
0.05 0.1
A0 / deg
-- m
v
Fig. 3. Rocking curves from the (333) diffraction plane of (111)-cut silicon wafers for samples 1-5 (read in the direction indicated by the
b
arrow).
0.20
E 0 k_ 030.10
0.00 - 1000
,
0
.
.
. 1000
. O
depth/A
.
. 2000
3000
Fig. 2. (a) Hydrogen depth profiles for samples i-4. The data were obtained for the same experimental set-up and a total charge of 100 laC. (b) Simulated perpendicularstrain profilesextracted from the rocking curves based on the dynamical theory.
3.2. X-ray rocking curve measurements and the simulation of the strain profile The results of the (333) reflection measurements are presented in Fig. 3. The curves belong to samples 1 - 5 in the sequence indicated by the arrow. Nearly no difference in the tail o f the rocking curves can be detected between the unannealed samples and the samples annealed at 400 °C. 500 °C annealing influences mainly the Bragg peak tail for negative (9. For 600 °C, however, a clear narrowing of the tail is found for both regions o f (9. It is known from theory that the strain causes a broadening o f the Bragg peak tail for negative values of (9. In the data shown, a broadening is also observed for positive values due to the diffused scattered background. To obtain the correct strain profile, this has to be taken into account for the strain fit procedure. Triple-axis measurements were made to evaluate the diffused scattered background. The perpendicular strain
profiles were obtained using an improved version of an algorithm given in the literature [10]. The results are given in comparison with the hydrogen depth profiles in Fig. 2(b) as a function o f the same depth scale. A reduction in the hydrogen content between the unannealed sample and the sample annealed at 400 °C is clearly observed. The strain profiles of these two sampies, however, are very close to each other. For higher annealing temperatures, the strain profiles are correlated with the loss of hydrogen in the samples.
4. Conclusions The argon=hydrogen cleaning procedure was studied for an arbitrary set of discharge parameters. The duration of the cleaning was enlarged to study the influences of the plasma on the wafer. F o r this strong treatment, hydrogen was observed to be introduced to the silicon to a depth of about 200 nm. Annealing reduces the hydrogen content as well as the perpendicular strain in the wafer. The promising results from recently obtained low temperature homoepitaxial growth at 300 °C for (100) and 400 °C for (111) silicon [5-7] indicate that the hydrogen may be built in at interstitial sites. With this cleaning procedure silicon wafers can be prepared for low temperature epitaxial growth in a single step.
Acknowledgment The authors would like to thank H. Senn for his valuable support during the cleaning experiments.
26
J. Ramm et aL / H cleaning of Si wafers
References 1 B. Anthony, L. Breaux, T. Hsu, S. Banerjee and A. Tasch, J. Vac. Sci. Technol. B, 7(1989) 621. 2 Y. Nara, Y. Sugita, N. Nakayama and T. Ito, J. Vac. Sci. Technol. B, 10(l) (1992) 274, 3 S. V. Hattangady, R. A. Rudder, M. J. Mantini, G. G. Fountain, J. B. Posthill and R. J. Markunas, J. AppL Phys., 68 (3) (1990) 1233. 4 T.-R. Yew and R. Reif, Mater. Res. Soc. Syrup. Proc., 202 (1991) 401.
5 J. Ramm, E. Beck and A. Zueger, Mater. Res. Soc. Symp. Proc., 220(1991) 15. 6 J. Ramm, E. Beck, F.-P. steiner, R. E. Pixley and I. Eisele, Mater. Res. Soc. Syrup. Proc., 259 (1992) 249. 7 J. Ramm, E. Beck, A. Zueger, A. Dommann and R. E. Pixley, Thin Solid Films, 222 (1992) 126. 8 A. Turos and O. Meyer, Nucl. 1nstruml Methods B, 4 (1984) 92. 9 P. F. Fewster, AppL Surf. Sci., 50(1991)9. 10 V. S. Speriosu and T. Vreeland, Jr., J. Appl. Phys., 59 (1984) 1591.