- Email: [email protected]
Annealing kinetics during rapid thermal processing of excimer laser-induced defects in virgin silicon
Applied Surface North-Holland
371
Science 46 (1990) 371-374
Annealing kinetics during rapid thermal laser-induced defects in virgin silicon B. Hartiti,
A. Slaoui,
J.C. Muller
Centre de Recherches Nucltkires Received
11 June 1990; accepted
(IN2P3).
and
Luhoratotre
for publication
processing
of excimer
P. Siffert PHASE
(UPR no 292 du CNRS),
B.P. 20, 67037 Strashourg
Cedex. France
5 July 1990
The annealing kinetics of two dominant defects (E(0.34 eV) and E(0.60 eV)) induced in virgin Fz n-type silicon by ArF excimer laser (193 nm, 0.75 J/cm*) irradiation have been studied within the temperature range 500-650 o C using rapid thermal processing (RTP). With deep-level transient spectroscopy measurements. we have observed that the reaction rates for both defects obey first-order kinetics. The E(0.34 eV) defect is shown to disappear with an activation energy of 0.7 * 0.1 eV and a frequency factor of 2 x 10’ s-‘. In this case. the annealing is enhanced by the ionization of this defect during RTA processing. The E(0.60 eV) trap is annealed at a faster rate exhibiting an activation energy of 1.15 + 0.1 eV and a frequency factor of lOI s- ‘.
1. Introduction In the past few years, extensive research has been carried out on the use of excimer laser radiation to process semiconductor materials [1,2]. The interest underlying this research is largely motivated by the realisation that laser processing may have significant advantages over more conventional processing steps for the fabrication of many semiconductor devices. Indeed, a pulsed excimer laser offers the possibility of confining high-temperature processing to the near-surface region of solids (< 0.3 pm) within well defined areas of good spatial resolution and during very short periods. However, one of the major obstacles to the development of pulsed-laser processing for practical applications is the formation of electrically active defects generated by laser treatment. Recently, we reported that high concentrations (10’“~1015 cm-‘) of electrically active traps were detected by deep-level transient spectroscopy (DLTS) in silicon samples irradiated with a pulsed ArF excimer laser [3,4]. We have shown in an earlier work [3] that these defects were similar to those reported for solid lasers (ruby and YAG), and that they were thought to be frozen-in during 0169.4332,‘90,‘$03.50
G 1990 - El sevier Science Publishers
the rapid quenching procedure. We have also shown in a second report [4] that a rapid thermal processing can be used to anneal such defects at around 650°C for a duration of the order of 1 min without the risk of introducing other defects into the processed wafers. Here we investigate in more detail the annealing kinetics of these excimer-laser-induced defects in silicon using rapid thermal processing and the DLTS technique. From these results, we discuss the mechanism of defect formation.
2. Experiments The samples were optically polished float-zone (Fz) (100) silicon (P-doped) 2 Q cm wafers. Irradiation was performed in air using a pulsed ArF excimer laser (X = 193 nm, 7 = 20 ns). The laser energy density was fixed at 0.75 J/cm’, which is well above the melting threshold of c-Si [51. After further rinsing in dilute hydrofluoric acid, some of the laser treated samples were annealed in a temperature/ time sequence ranging from 500 to 650°C in steps of 50” C, for durations of 10 to
B.V (North-Holland)
312
B. Hartitr et 01. / Annealrng krnetics durrng RTP
120 s. using a commercially available rapid thermal annealing system. Following irradiation, 1 mm gold dots were evaporated onto the n-type silicon surfaces to form Schottky diodes. The deep-level transient spectroscopy (DLTS) system used for measurements of the trap concentration consisted of a 1 MHz capacitance bridge and a double-phase lock-in detector method providing a square weighting function.
cm’ are detected at an emission a,=5X10P4 rate of 41 ss’ [3,4]. The defects E,-ES have previously been observed by many authors in solid-state laser annealing of virgin n-type silicon [6-lo]. The corresponding energies coincide within 0.04 eV with those found for a ruby laser (694 nm) and within 0.02 eV for YAG (1060 or 530 nm) lasers. However, in the case of excimer laser irradiation an additional trap E, located at about 0.60 eV from the conduction band is observed. We suggest that the difference in the re-growth velocity between excimer and solid state lasers [ll] is responsible for the appearance of this level. Curves 2 to 5 in fig. 1 show the evolution of the concentration of these traps during rapid thermal annealing (RTA) under different conditions of temperature and time. The increase of the annealing temperature results in appreciable reduction of all the defects.
3. Results and discussion A typical DLTS spectrum (curve 1) of the defect states observed in ArF excimer laser irradiated n-type silicon is shown in fig. 1. Six electron traps, E,(0.18 eV), E,(0.25 eV), E,(0.34 eV) u3 = 5 X 10-‘h cm2, E,(0.43 eV) a, = 1 x 10P” cm2, E,(0.53 eV) a, = 1 x lo-l5 cm2 and E,(0.60 eV)
1
1
1
OLTS Si N
I
1
of defrcrs ,n r’iqin .s~hcon
1
I
I
ANALYSIS
Fz
I
%
l-5Rcm
l- ArF Laser irradiation 0.75J1cm2 Rapid
thermal
annealing
after Laser Irradiation
2- 500°C , $Osec. 3- 600°C,
6Osec
4- 65O"C,
6Osec
s- 650°C.
120sec.
I
I 150
I 100
I 200
TEMPERATURE Fig.
1. DLTS
spectra
for ArF
excimw
treatments
laser-irradiated
(0.75
J/cm’.
193 nm)
I
I 300
I 250
(OKI
n-type silicon:
(I)
as-irradiated,
at (2) 500 o C. 60 s, (3) 600 o C. 60 s, (4) 650 o C. 60 s and (5) 650 o C.
and after
120s.
post-laser
K-PI
B. Hartrti et al. / Annealing
kinetics during RTP
ofdefectsin
virgin silicon
373
At 650 o C, the quenched defects are almost completely removed. In figs. 2 and 3, we report the variation of trap concentration as a function of time for E,(0.34 eV) and E,(0.60 eV) defects, respectively, annealed at 500, 550 and 600°C in the RTA furnace. We can propose no explanation at this time for the two-stage behaviour exhibited by the annealing rate for both traps. However, annealing with two stages has been also observed by Hirata et al. [12] and Kimerling et al. [13] during annealing of vacancy complexes. These results suggest that the origin of excimer-laser-related defects can be assigned to vacancy clusters or vacancy-impurity associations. In the second stage the annealing rate of the defects can be approximated by linear first-order kinetics and is given by
(1)
N(t) = No exp(-t/TO),
where N,) is the initial defect concentration and Q is the process time constant. When reported as a function of reciprocal temperature, l/r,, obeys an
I
I
20
I
/
1
60 Annealing
1
I
I
100 Time
(set
)
Fig. 2. Rapid thermal annealing rates of the E,(0.34 at different temperatures.
eV) defect
20
60 Anneal Ing Time
Fig. 3. Rapid
100 (set)
thermal annealing rates of the E,(0.60 at different temperatures.
eV) defect
Arrhenius law from which we deduced the activation energy and the frequency factor for the process. Fig. 4 shows these plots for RTA processing for the traps E,(0.34 eV) and Er(0.60 ev> giving E,=0.7+0.1 eV (~,=2X10’ s-‘) and E,= 1.15 + 0.1 eV (v,) = 1014 s-l), respectively. Unfortunately, to the best of our knowledge, only one work [14] gives the activation energy of a donor centre located at E, = 0.32 eV induced in low-resistivity n-type silicon after a YAG laser treatment. It is probable that it is the same level as our trap at E, = 0.34 eV. The 2 x 10’ pre-exponential factor deduced from the annealing rate of the E,(0.34 eV) trap strongly suggests long-range migration of a defect (- lo5 steps as an entity, occurring at a rate of - 10” ss’) and requires a capture of a free carrier for its annealing [15]. During the RTA processing, due to the higher concentration of photons whose energy hv B- Eg (where E,: = 1.1 eV, the forbidden gap in Si), a large number of electron-hole pairs are created, which affects the
374
B. Hartitr et al. / Annrahng
I
kinetrcP during RTP
4. Conclusion
n -type
SI
Rapld
Thermal
In summary, our results show that considerable ionization occurs when a rapid thermal furnace is used to anneal the excimer laser-induced defect located at 0.34 eV below the conduction band. This ionization induces a reduction in migration energy thus enhancing defect migration to sinks. On the other hand, the donor defect located at 0.6 eV presents a higher frequency factor value which allows its annealing in shorter times.
Annealing
t
100 c)
'0
eV
1 \’ I t
x c
r
-8
of defects m virgin .silron
r
EA=1;15i
O,le;
‘--
IO
References
+
E,
r
l
EC - 0.6 eV
PI 1
Chemical Processing with Lasers, Vol. 1 of Springer Series in Materials Science (Springer. Berlin. 1986). I.W. Boyd. Laser Processing of Thin Films and Microstructures. Vol. 3 of Springer Series in Materials Science (Springer. Berlin. 1987). B. Hartiti. A. Slaoui. J.C. Muller and P. Siffert. J. Appl. Phys. 66 (1989) 3934. B. Hartiti, A. Slaoui. J.C. Muller and P. Siffert. Mater. Sci. Eng. B 4 (1989) 257. F. Foulon. E. Fogarassy. A. Slaoui. C. Fuchs, S. Unamuno and P. Siffert, Appl. Phys. A 45 (1988) 361. L.C. Kimerling and J.L. Benton, in: Laser and Electron Beam Processing of Materials, Eds. C.W. White and P.S. Peercy (Academic Press. New York, 1980) p. 383. G.A. Kachurin, E.V. Nidaev and N.V. Danyushklna. Sov. Phys. Semicond. 14 (1980) 386. Z.K. Fan. V.Q. Ho and T. Sugano, Appl. Phys. Lett. 40 (1982) 418. A. Mesli. J.C. Muller and P. Siffert, in: Laser-Solid Interactions and Transient Thermal Processmg of Materials Proc.. Proc. 1st E-MRS Conf. (Les Editions de Physiques, Les Ulis. C.44. 1983) p. 281. K.L. Wang. Y.S. Lin, G.E. Possin, J. Karins and J. Corbett, J. Appl. Phys. 54 (1983) 3839. S. Unamuno and E. Fogarassy, Appl. Surf. Sci. 36 (1989) 1. M. Hirata and H. Saito, J. Appl. Phys. 37 (1966) 1867. L.C. Kimerling, H.M. DeAngelis and C.P. Carries. Phyy. Rev. B 3 (1971) 427. W.0. Adekoya. J.C. Muller and P. Siffert, Appl. Phys. Lett. 49 (1986) 1429. A. Chantre, in: Defects in Electronic Materials. Eds. M. Stavola and S.J. Pearton, Mater. Res. Sot. Conf. Proc.. Vol. 104 (MRS. Pittsburgh, PA. 1987). L.C. Kimerling, H.M. DeAngelis and J.W. Diehold, Solid State Commun. 16 (1975) 171.
[II D. Bluerle. - 0,34eV
I
1
I
11
12
I,3
I [31
i
1000’kT
)
Fig. 4. Annealing kinetics for E,(0.34 eV) and E,(0.60 eV) defects. The activation energy, E,, and frequency factor, v,,. are determined from the observed first-order kinetics according to the relation dN/dt = v,, exp( - E,/kT).
[41 [51 [61
occupation statistics of the 0.34 eV trap. In such a considerable ionization occurs. This situation, phenomenon is supported by an increase in the annealing rate about 15 times greater for samples RTA treated than in those annealed using a conventional furnace [4,14]. These results suggest that an important ionization occurs during RTA processing which reduces the activation energy required for the process and enhances defect migration to sinks [14,16]. As for the E,(0.60 eV) trap, the activation energy is high (1.15 eV) whereas the frequency factor (lOI s-’ ) is comparable to the lattice vibration frequency. It means that this defect requires shorter times to be annealed without supply of free carriers. We have no data concerning the charge state of these defects. Additional experiments, for example under reverse bias, are needed to yield information on the defect charge states.
[71 PI [91
[lOI PII [121 [I31 [I41 P51
[I61
371
Science 46 (1990) 371-374
Annealing kinetics during rapid thermal laser-induced defects in virgin silicon B. Hartiti,
A. Slaoui,
J.C. Muller
Centre de Recherches Nucltkires Received
11 June 1990; accepted
(IN2P3).
and
Luhoratotre
for publication
processing
of excimer
P. Siffert PHASE
(UPR no 292 du CNRS),
B.P. 20, 67037 Strashourg
Cedex. France
5 July 1990
The annealing kinetics of two dominant defects (E(0.34 eV) and E(0.60 eV)) induced in virgin Fz n-type silicon by ArF excimer laser (193 nm, 0.75 J/cm*) irradiation have been studied within the temperature range 500-650 o C using rapid thermal processing (RTP). With deep-level transient spectroscopy measurements. we have observed that the reaction rates for both defects obey first-order kinetics. The E(0.34 eV) defect is shown to disappear with an activation energy of 0.7 * 0.1 eV and a frequency factor of 2 x 10’ s-‘. In this case. the annealing is enhanced by the ionization of this defect during RTA processing. The E(0.60 eV) trap is annealed at a faster rate exhibiting an activation energy of 1.15 + 0.1 eV and a frequency factor of lOI s- ‘.
1. Introduction In the past few years, extensive research has been carried out on the use of excimer laser radiation to process semiconductor materials [1,2]. The interest underlying this research is largely motivated by the realisation that laser processing may have significant advantages over more conventional processing steps for the fabrication of many semiconductor devices. Indeed, a pulsed excimer laser offers the possibility of confining high-temperature processing to the near-surface region of solids (< 0.3 pm) within well defined areas of good spatial resolution and during very short periods. However, one of the major obstacles to the development of pulsed-laser processing for practical applications is the formation of electrically active defects generated by laser treatment. Recently, we reported that high concentrations (10’“~1015 cm-‘) of electrically active traps were detected by deep-level transient spectroscopy (DLTS) in silicon samples irradiated with a pulsed ArF excimer laser [3,4]. We have shown in an earlier work [3] that these defects were similar to those reported for solid lasers (ruby and YAG), and that they were thought to be frozen-in during 0169.4332,‘90,‘$03.50
G 1990 - El sevier Science Publishers
the rapid quenching procedure. We have also shown in a second report [4] that a rapid thermal processing can be used to anneal such defects at around 650°C for a duration of the order of 1 min without the risk of introducing other defects into the processed wafers. Here we investigate in more detail the annealing kinetics of these excimer-laser-induced defects in silicon using rapid thermal processing and the DLTS technique. From these results, we discuss the mechanism of defect formation.
2. Experiments The samples were optically polished float-zone (Fz) (100) silicon (P-doped) 2 Q cm wafers. Irradiation was performed in air using a pulsed ArF excimer laser (X = 193 nm, 7 = 20 ns). The laser energy density was fixed at 0.75 J/cm’, which is well above the melting threshold of c-Si [51. After further rinsing in dilute hydrofluoric acid, some of the laser treated samples were annealed in a temperature/ time sequence ranging from 500 to 650°C in steps of 50” C, for durations of 10 to
B.V (North-Holland)
312
B. Hartitr et 01. / Annealrng krnetics durrng RTP
120 s. using a commercially available rapid thermal annealing system. Following irradiation, 1 mm gold dots were evaporated onto the n-type silicon surfaces to form Schottky diodes. The deep-level transient spectroscopy (DLTS) system used for measurements of the trap concentration consisted of a 1 MHz capacitance bridge and a double-phase lock-in detector method providing a square weighting function.
cm’ are detected at an emission a,=5X10P4 rate of 41 ss’ [3,4]. The defects E,-ES have previously been observed by many authors in solid-state laser annealing of virgin n-type silicon [6-lo]. The corresponding energies coincide within 0.04 eV with those found for a ruby laser (694 nm) and within 0.02 eV for YAG (1060 or 530 nm) lasers. However, in the case of excimer laser irradiation an additional trap E, located at about 0.60 eV from the conduction band is observed. We suggest that the difference in the re-growth velocity between excimer and solid state lasers [ll] is responsible for the appearance of this level. Curves 2 to 5 in fig. 1 show the evolution of the concentration of these traps during rapid thermal annealing (RTA) under different conditions of temperature and time. The increase of the annealing temperature results in appreciable reduction of all the defects.
3. Results and discussion A typical DLTS spectrum (curve 1) of the defect states observed in ArF excimer laser irradiated n-type silicon is shown in fig. 1. Six electron traps, E,(0.18 eV), E,(0.25 eV), E,(0.34 eV) u3 = 5 X 10-‘h cm2, E,(0.43 eV) a, = 1 x 10P” cm2, E,(0.53 eV) a, = 1 x lo-l5 cm2 and E,(0.60 eV)
1
1
1
OLTS Si N
I
1
of defrcrs ,n r’iqin .s~hcon
1
I
I
ANALYSIS
Fz
I
%
l-5Rcm
l- ArF Laser irradiation 0.75J1cm2 Rapid
thermal
annealing
after Laser Irradiation
2- 500°C , $Osec. 3- 600°C,
6Osec
4- 65O"C,
6Osec
s- 650°C.
120sec.
I
I 150
I 100
I 200
TEMPERATURE Fig.
1. DLTS
spectra
for ArF
excimw
treatments
laser-irradiated
(0.75
J/cm’.
193 nm)
I
I 300
I 250
(OKI
n-type silicon:
(I)
as-irradiated,
at (2) 500 o C. 60 s, (3) 600 o C. 60 s, (4) 650 o C. 60 s and (5) 650 o C.
and after
120s.
post-laser
K-PI
B. Hartrti et al. / Annealing
kinetics during RTP
ofdefectsin
virgin silicon
373
At 650 o C, the quenched defects are almost completely removed. In figs. 2 and 3, we report the variation of trap concentration as a function of time for E,(0.34 eV) and E,(0.60 eV) defects, respectively, annealed at 500, 550 and 600°C in the RTA furnace. We can propose no explanation at this time for the two-stage behaviour exhibited by the annealing rate for both traps. However, annealing with two stages has been also observed by Hirata et al. [12] and Kimerling et al. [13] during annealing of vacancy complexes. These results suggest that the origin of excimer-laser-related defects can be assigned to vacancy clusters or vacancy-impurity associations. In the second stage the annealing rate of the defects can be approximated by linear first-order kinetics and is given by
(1)
N(t) = No exp(-t/TO),
where N,) is the initial defect concentration and Q is the process time constant. When reported as a function of reciprocal temperature, l/r,, obeys an
I
I
20
I
/
1
60 Annealing
1
I
I
100 Time
(set
)
Fig. 2. Rapid thermal annealing rates of the E,(0.34 at different temperatures.
eV) defect
20
60 Anneal Ing Time
Fig. 3. Rapid
100 (set)
thermal annealing rates of the E,(0.60 at different temperatures.
eV) defect
Arrhenius law from which we deduced the activation energy and the frequency factor for the process. Fig. 4 shows these plots for RTA processing for the traps E,(0.34 eV) and Er(0.60 ev> giving E,=0.7+0.1 eV (~,=2X10’ s-‘) and E,= 1.15 + 0.1 eV (v,) = 1014 s-l), respectively. Unfortunately, to the best of our knowledge, only one work [14] gives the activation energy of a donor centre located at E, = 0.32 eV induced in low-resistivity n-type silicon after a YAG laser treatment. It is probable that it is the same level as our trap at E, = 0.34 eV. The 2 x 10’ pre-exponential factor deduced from the annealing rate of the E,(0.34 eV) trap strongly suggests long-range migration of a defect (- lo5 steps as an entity, occurring at a rate of - 10” ss’) and requires a capture of a free carrier for its annealing [15]. During the RTA processing, due to the higher concentration of photons whose energy hv B- Eg (where E,: = 1.1 eV, the forbidden gap in Si), a large number of electron-hole pairs are created, which affects the
374
B. Hartitr et al. / Annrahng
I
kinetrcP during RTP
4. Conclusion
n -type
SI
Rapld
Thermal
In summary, our results show that considerable ionization occurs when a rapid thermal furnace is used to anneal the excimer laser-induced defect located at 0.34 eV below the conduction band. This ionization induces a reduction in migration energy thus enhancing defect migration to sinks. On the other hand, the donor defect located at 0.6 eV presents a higher frequency factor value which allows its annealing in shorter times.
Annealing
t
100 c)
'0
eV
1 \’ I t
x c
r
-8
of defects m virgin .silron
r
EA=1;15i
O,le;
‘--
IO
References
+
E,
r
l
EC - 0.6 eV
PI 1
Chemical Processing with Lasers, Vol. 1 of Springer Series in Materials Science (Springer. Berlin. 1986). I.W. Boyd. Laser Processing of Thin Films and Microstructures. Vol. 3 of Springer Series in Materials Science (Springer. Berlin. 1987). B. Hartiti. A. Slaoui. J.C. Muller and P. Siffert. J. Appl. Phys. 66 (1989) 3934. B. Hartiti, A. Slaoui. J.C. Muller and P. Siffert. Mater. Sci. Eng. B 4 (1989) 257. F. Foulon. E. Fogarassy. A. Slaoui. C. Fuchs, S. Unamuno and P. Siffert, Appl. Phys. A 45 (1988) 361. L.C. Kimerling and J.L. Benton, in: Laser and Electron Beam Processing of Materials, Eds. C.W. White and P.S. Peercy (Academic Press. New York, 1980) p. 383. G.A. Kachurin, E.V. Nidaev and N.V. Danyushklna. Sov. Phys. Semicond. 14 (1980) 386. Z.K. Fan. V.Q. Ho and T. Sugano, Appl. Phys. Lett. 40 (1982) 418. A. Mesli. J.C. Muller and P. Siffert, in: Laser-Solid Interactions and Transient Thermal Processmg of Materials Proc.. Proc. 1st E-MRS Conf. (Les Editions de Physiques, Les Ulis. C.44. 1983) p. 281. K.L. Wang. Y.S. Lin, G.E. Possin, J. Karins and J. Corbett, J. Appl. Phys. 54 (1983) 3839. S. Unamuno and E. Fogarassy, Appl. Surf. Sci. 36 (1989) 1. M. Hirata and H. Saito, J. Appl. Phys. 37 (1966) 1867. L.C. Kimerling, H.M. DeAngelis and C.P. Carries. Phyy. Rev. B 3 (1971) 427. W.0. Adekoya. J.C. Muller and P. Siffert, Appl. Phys. Lett. 49 (1986) 1429. A. Chantre, in: Defects in Electronic Materials. Eds. M. Stavola and S.J. Pearton, Mater. Res. Sot. Conf. Proc.. Vol. 104 (MRS. Pittsburgh, PA. 1987). L.C. Kimerling, H.M. DeAngelis and J.W. Diehold, Solid State Commun. 16 (1975) 171.
[II D. Bluerle. - 0,34eV
I
1
I
11
12
I,3
I [31
i
1000’kT
)
Fig. 4. Annealing kinetics for E,(0.34 eV) and E,(0.60 eV) defects. The activation energy, E,, and frequency factor, v,,. are determined from the observed first-order kinetics according to the relation dN/dt = v,, exp( - E,/kT).
[41 [51 [61
occupation statistics of the 0.34 eV trap. In such a considerable ionization occurs. This situation, phenomenon is supported by an increase in the annealing rate about 15 times greater for samples RTA treated than in those annealed using a conventional furnace [4,14]. These results suggest that an important ionization occurs during RTA processing which reduces the activation energy required for the process and enhances defect migration to sinks [14,16]. As for the E,(0.60 eV) trap, the activation energy is high (1.15 eV) whereas the frequency factor (lOI s-’ ) is comparable to the lattice vibration frequency. It means that this defect requires shorter times to be annealed without supply of free carriers. We have no data concerning the charge state of these defects. Additional experiments, for example under reverse bias, are needed to yield information on the defect charge states.
[71 PI [91
[lOI PII [121 [I31 [I41 P51
[I61