- Email: [email protected]
SIMS study of Cu trapping and migration in low-k dielectric films
Applied Surface Science 231–232 (2004) 791–795
SIMS study of Cu trapping and migration in low-k dielectric films Yupu Lia,*, Jerry Huntera, Tom J. Tateb a
Materials Analytical Services, Sunnyvale Laboratory, 285 North Wolfe Road, Sunnyvale, CA 94085, USA b Department of Electronic and Electrical Engineering, Imperial College of Science Technology and Medicine, London SW7 2BT, UK Available online 21 April 2004
Abstract A 545 nm thick low-k dielectric film was implanted at room temperature with 50 keV 63 Cuþ to a dose of 1:0 1014 atoms/ cm2. The film is a SiOx-based material and doped with about 8 at.% of flourine. Analyses by secondary-ion mass spectrometry show that Cu is fast diffuser in the low-k film, and after the RTA anneals Cu has redistributed within the film and some Cu has migrated to the interface between the low-k film and Si substrate. At 800 8C RTA, the apparent ‘‘diffusion’’ coefficient in the implanted film, DA, is estimated as 1:5 109 cm2/s. The 1100 8C RTA sample was re-analysed after stripping the low-k film and the result showed that 4:0 1012 atoms/cm2 Cu had moved into the silicon substrate to a depth of about 170 nm. # 2004 Elsevier B.V. All rights reserved. Keywords: Low-k dielectric film; Copper trapping and migration; SIMS
1. Introduction New 90 nm manufacturing process technology may combine the use of strained Si, copper interconnects and oxide-based low-k dielectric materials [1]. Low-k (dielectric constant 2–3.5) films are oxide (dielectric constant 4) based materials, doped with high levels of F or C, or which use H, C, O to replace the Si atoms in the oxide. Cu damascene interconnects with low-k dielectrics have become mainstream process due to the reduction in the interconnect delay for Cu and lowk films versus Al and SiO2. For this reason, it is an interesting issue to study Cu trapping and diffusion in low-k films during the thermal annealing process. In this paper, as first step, we report range data for 50 keV Cu implantation into a low-k film (i.e. a SiOxbased film doped with 8 at.% F) at room temperature * Corresponding author. E-mail address: [email protected] (Y. Li).
and the behaviour of trapping and migration of Cu in the low-k film during rapid thermal annealing (RTA) in a temperature range from 800 to 1100 8C.
2. Experimental A 545 nm thick SiOx-based low-k film was used for Cu implantation. After implantation with 50 keV 63 Cuþ ions to a dose of 1:0 1014 atoms/cm2 (78 tilt, at room temperature) at the Imperial College, UK, the as-implanted sample was divided into several pieces and annealed in flowing nitrogen ambient using a RTA oven, at various temperatures between 800 and 1100 8C for 30 s. Copper profiles for all samples were obtained on a Cameca 4f SIMS tool at Materials Analytical Services, using a 3.5 keV O2þ profiling energy at an incidence angle of 48.88. To minimize the sample charging the normal incidence e-gun (with a filament current about 1.2 A) was used. 63 Cuþ
0169-4332/$ – see front matter # 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.apsusc.2004.03.072
792
Y. Li et al. / Applied Surface Science 231–232 (2004) 791–795
secondary ions were monitored using high mass resolution (M/DM > 4000) to avoid the mass interferences with 28 Si þ F þ 16 O, etc. Depth calibration was achieved by a profilometer.
3. Results and discussion Fig. 1 shows the F depth profile in the starting material obtained using the Cameca 4f SIMS with a Csþ primary ion beam. The SIMS results show that the SiOx-based material contains 8 at.% F and the thickness is 545 nm. The films contain <0.1 at.% of other impurities (H, C, N and Cl) as shown in Fig. 1 (i.e. using atomic density as 6:7 1022 atoms/cm3). Cu trapping during implantation: Fig. 2 shows the as-implanted (50 keV 1:0 1014 atoms/cm2) 63 Cu concentration profile. Concentration calibration was achieved by generation of a relative sensitivity factor assuming that the implanted dose given above was accurate, this assumption was verified by quantification based on our Cu in an oxide standard. The distribution is Gaussian-like, and the depth of the Cu peak is 44 nm. The maximum Cu concentration Cmax is found to be 2:2 1019 atoms/cm3.
Fig. 2 also shows the simulated Cu concentration and vacancy concentration profiles for the Cu implantation into SiO2 film calculated using SRIM [2]. It should be noted that the initial film is an amorphous material and the vacancy profile only refers the displaced damage to target atoms. Fig. 2 shows the measured Cu peak is in very good agreement with the simulated Cu peak (Cmax ¼ 2:7 1019 atoms/cm3 at a depth of 42 nm), although the measured Cu profile is about 21% (i.e. DRp the straggle of the profile) wider than the simulated profile, defined as the half-width of the depth profile at Cu concentration C ¼ Cmax /e. Here DRp (SIMS) is 25.4 nm versus 21 nm for DRp (SRIM). Cu migration during annealing: Fig. 3 shows a change in the Cu distribution after RTA at 800, 1000 and 1100 8C for 30 s. It has been found that the annealing at 800 8C resulted in a noticeable change of the as-implanted Cu depth profile. Clearly, there is significant diffusion of Cu at temperatures as low as 800 8C. Fig. 3 shows that the implanted Cu starts to migrate within the film at temperatures below 800 8C. As a guideline, it is often convenient to roughly relate the anneal temperature in Kelvin to activation energy for migration by Um 25kTanneal [3], where k is the
1E+23
CONCENTRATION (atoms/cm3)
1E+22
1E+21
C
1E+20
H N 1E+19
Cl F
1E+18
1E+17
1E+16 0
1000
2000
3000
4000
5000
6000
7000
8000
DEPTH (Angstroms)
Fig. 1. SIMS profiles of H, C, N, Cl, and F in the starting material.
9000
10000
Y. Li et al. / Applied Surface Science 231–232 (2004) 791–795
793
1E+20
Cu-SIMS: 10.0e13 atoms/cm2 1E+19
Cu-SRIM Simulation
CONCENTRATION (atoms/cm3)
Vacancies x0.001: SRIM simulation
1E+18
1E+17
1E+16
1E+15
1E+14 0
1000
2000
3000
4000
5000
6000
DEPTH (Angstroms)
Fig. 2. SIMS 63 Cu profile after 50 keV 63 Cu implantation to a dose of 1:0 1014 atoms/cm3, and simulated Cu concentration profile and vacancy profile (multiplied by a factor of 0.001) using SRIM.
Boltzmann constant and Tanneal is the anneal temperature. Thus, the apparent activation energy for activating Cu in the film is estimated to be below 2.33 eV. It is possible to roughly estimate the apparent ‘‘diffusion’’ coefficient at given temperature through the SIMS profiles before and after the annealing. At 800 8C the implanted Cu is clearly mobile (see Fig. 3). The profile becomes broader and some of the implanted Cu migrated into the deeper layers of the film and the interface where the irradiation damage is less. Using DX 2 ¼ 2DA t (i.e. via Brownian movement, see Ref. [4]), the apparent diffusion coefficient in the implanted film, DA 1:5 109 cm2/s. Here we have defined the mean square displacement of the migrating Cu, DX2, equal to (300 nm)2 as we roughly ˚ from curves estimated the Cu has moved about 3000 A 1 and 2 in Fig. 3 (i.e. 63 Cu trace after 800 8C RTA and 63 Cu trace after implantation). As compared with Al and Cu in Si, from the literature [5], at 800 8C, the diffusivity of Al in Si is lower than 1014 cm2/s and diffusivity of Cu in Si is higher than 107 cm2/s. These results indicate that Cu is also a fast diffuser in the
low-k film, although we have not found the ‘‘diffusivity’’ of Cu in standard SiO2 for comparison. The general behaviour of Cu migration in the irradiated low-k film can be summarized as follows: 1. The surface region and the interface between the film and silicon substrate are sinks for migrated Cu. The surface region contains more irradiation damage (see curve 3 in Fig. 3) and so even after RTA, the surface region has more trapping centers for Cu. The interfaces in SiO2/Si systems always are trapping centers for impurities. 2. Fig. 3 shows the interface region has trapped a large percentage of remaining Cu after RTA. It can be seen from Fig. 3 that some Cu has also moved beyond the interface. In order to see this behaviour more clearly, the low-k film was stripped using HF acid and the Cu was profiled in the Si alone. As shown in Fig. 4, for the 1100 8C (30 s) RTA piece, after stripping the film, the silicon substrate contained about 4:0 1012 atoms/cm2 63 Cu to a ˚ . It should be noted that the depth of about 1700 A
794
Y. Li et al. / Applied Surface Science 231–232 (2004) 791–795 1E+21
1: as-impl. :10e13 atoms/cm2
CONCENTRATION (atoms/cm3)
1E+20
2: 800C RTA Cu=2.0e13 atoms/cm2 3: 1000C RTA Cu=1.2e13 atoms/cm2 4: 1100C RTA Cu=8.6e12 atoms/cm2
1E+19
1E+18
1E+17
1E+16
1E+15
1E+14 0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
DEPTH (Angstroms) 63
Fig. 3. SIMS Cu profiles: curve 1: after implantation; curve 2: plus 800 8C RTA for 30 s; curve 3: plus 1000 8C RTA for 30 s; curve 4: plus 1100 8C RTA for 30 s.
RTA time is so short, however, Cu diffusivity in silicon is so high. Thus, it is not surprising to see some of Cu in the Si substrate since the very high diffusivity of Cu in Si at 1100 8C (i.e. 4 106 cm2/s at 1100 8C, the diffusion length (DAt)0.5 for 30 s could be as large as 1:09 102 cm). 3. The retained 63 Cu dose after RTA are as follows: 2:0 1013 atoms/cm 2 after 800 8C RTA, 1:2 1013 atoms/cm2 after 1000 8C RTA and 8:6 1012 atoms/cm2 after 1100 8C RTA. It should be noted the above calculation is based on the assumption that all measured Cu is in the low-k film since the quantified concentration profiles shown in Fig. 3 are based on the sensitivity factor obtained from the low-k matrix. For curves 2–4 shown in Fig. 3, since there is a small percentage of trapped Cu atoms at the interface and in the silicon substrate, the measurement conditions used underestimate the concentration of Cu in the silicon substrate since we did
not use a Si RSF in the region beyond the interface. However the error is small since the majority of the Cu resides in the low-k film or at the interface. For the 1100 8C annealed sample, the error is larger since the silicon trapped near half of the remained Cu. For this sample, as shown in Fig. 4 after stripped the low-k with HF acid, Cu analysis on the silicon substrate only is a more accurate method for calculation of the Cu in Si. It is clear from the retained dose calculations that most of the implanted Cu has ‘‘evaporated’’ due to the RTA process. At this point we do not have an explanation for the phenomenon. It is interest to note that another possibility: because Cu is such a fast diffuser in Si it may be that the Cu did not ‘‘evaporate’’ but moved all the way to the back of the wafer. In summary, Cu migrates at a temperature below 800 8C in the F-doped low-k film and the surface region and the interface between the film and silicon
Y. Li et al. / Applied Surface Science 231–232 (2004) 791–795
795
1E+20
CONCENTRATION (atoms/cm3)
1E+19
Cu=4.0e12 atoms/cm2 in Silicon 1E+18
1E+17
1E+16
1E+15
1E+14 0
1000
2000
3000
4000
5000
6000
DEPTH (Angstroms)
Fig. 4. SIMS
63
Cu profile in silicon substrate for the 1100 8C RTA sample after removal of the low-k film.
substrate are sinks for migrating Cu during the annealing. Cu is a fast diffuser in the F-doped low-k film.
[2]
[3]
References
[4]
[1]
[5]
Does Advanced Technology Really Matter, Intel 2002 Annual Report. http://www.intel.com.
J.F. Ziegler, J.P. Biersack, U. Littmark, The Stopping and Ranges of Ions in Solids, Pergamon Press, New York, 1985. T.D. Townsend, J.C. Kelly, N.E.W. Hartley, Ion Implantation and their Applications, Academic Press, New York, 1976. W. Jost, Diffusion in Solids, Liquid, Gases, Academic Press, New York, 1960. H.F. Wolf (Ed.), Silicon Semiconductor Data, Pergamon Press, New York, 1969.
SIMS study of Cu trapping and migration in low-k dielectric films Yupu Lia,*, Jerry Huntera, Tom J. Tateb a
Materials Analytical Services, Sunnyvale Laboratory, 285 North Wolfe Road, Sunnyvale, CA 94085, USA b Department of Electronic and Electrical Engineering, Imperial College of Science Technology and Medicine, London SW7 2BT, UK Available online 21 April 2004
Abstract A 545 nm thick low-k dielectric film was implanted at room temperature with 50 keV 63 Cuþ to a dose of 1:0 1014 atoms/ cm2. The film is a SiOx-based material and doped with about 8 at.% of flourine. Analyses by secondary-ion mass spectrometry show that Cu is fast diffuser in the low-k film, and after the RTA anneals Cu has redistributed within the film and some Cu has migrated to the interface between the low-k film and Si substrate. At 800 8C RTA, the apparent ‘‘diffusion’’ coefficient in the implanted film, DA, is estimated as 1:5 109 cm2/s. The 1100 8C RTA sample was re-analysed after stripping the low-k film and the result showed that 4:0 1012 atoms/cm2 Cu had moved into the silicon substrate to a depth of about 170 nm. # 2004 Elsevier B.V. All rights reserved. Keywords: Low-k dielectric film; Copper trapping and migration; SIMS
1. Introduction New 90 nm manufacturing process technology may combine the use of strained Si, copper interconnects and oxide-based low-k dielectric materials [1]. Low-k (dielectric constant 2–3.5) films are oxide (dielectric constant 4) based materials, doped with high levels of F or C, or which use H, C, O to replace the Si atoms in the oxide. Cu damascene interconnects with low-k dielectrics have become mainstream process due to the reduction in the interconnect delay for Cu and lowk films versus Al and SiO2. For this reason, it is an interesting issue to study Cu trapping and diffusion in low-k films during the thermal annealing process. In this paper, as first step, we report range data for 50 keV Cu implantation into a low-k film (i.e. a SiOxbased film doped with 8 at.% F) at room temperature * Corresponding author. E-mail address: [email protected] (Y. Li).
and the behaviour of trapping and migration of Cu in the low-k film during rapid thermal annealing (RTA) in a temperature range from 800 to 1100 8C.
2. Experimental A 545 nm thick SiOx-based low-k film was used for Cu implantation. After implantation with 50 keV 63 Cuþ ions to a dose of 1:0 1014 atoms/cm2 (78 tilt, at room temperature) at the Imperial College, UK, the as-implanted sample was divided into several pieces and annealed in flowing nitrogen ambient using a RTA oven, at various temperatures between 800 and 1100 8C for 30 s. Copper profiles for all samples were obtained on a Cameca 4f SIMS tool at Materials Analytical Services, using a 3.5 keV O2þ profiling energy at an incidence angle of 48.88. To minimize the sample charging the normal incidence e-gun (with a filament current about 1.2 A) was used. 63 Cuþ
0169-4332/$ – see front matter # 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.apsusc.2004.03.072
792
Y. Li et al. / Applied Surface Science 231–232 (2004) 791–795
secondary ions were monitored using high mass resolution (M/DM > 4000) to avoid the mass interferences with 28 Si þ F þ 16 O, etc. Depth calibration was achieved by a profilometer.
3. Results and discussion Fig. 1 shows the F depth profile in the starting material obtained using the Cameca 4f SIMS with a Csþ primary ion beam. The SIMS results show that the SiOx-based material contains 8 at.% F and the thickness is 545 nm. The films contain <0.1 at.% of other impurities (H, C, N and Cl) as shown in Fig. 1 (i.e. using atomic density as 6:7 1022 atoms/cm3). Cu trapping during implantation: Fig. 2 shows the as-implanted (50 keV 1:0 1014 atoms/cm2) 63 Cu concentration profile. Concentration calibration was achieved by generation of a relative sensitivity factor assuming that the implanted dose given above was accurate, this assumption was verified by quantification based on our Cu in an oxide standard. The distribution is Gaussian-like, and the depth of the Cu peak is 44 nm. The maximum Cu concentration Cmax is found to be 2:2 1019 atoms/cm3.
Fig. 2 also shows the simulated Cu concentration and vacancy concentration profiles for the Cu implantation into SiO2 film calculated using SRIM [2]. It should be noted that the initial film is an amorphous material and the vacancy profile only refers the displaced damage to target atoms. Fig. 2 shows the measured Cu peak is in very good agreement with the simulated Cu peak (Cmax ¼ 2:7 1019 atoms/cm3 at a depth of 42 nm), although the measured Cu profile is about 21% (i.e. DRp the straggle of the profile) wider than the simulated profile, defined as the half-width of the depth profile at Cu concentration C ¼ Cmax /e. Here DRp (SIMS) is 25.4 nm versus 21 nm for DRp (SRIM). Cu migration during annealing: Fig. 3 shows a change in the Cu distribution after RTA at 800, 1000 and 1100 8C for 30 s. It has been found that the annealing at 800 8C resulted in a noticeable change of the as-implanted Cu depth profile. Clearly, there is significant diffusion of Cu at temperatures as low as 800 8C. Fig. 3 shows that the implanted Cu starts to migrate within the film at temperatures below 800 8C. As a guideline, it is often convenient to roughly relate the anneal temperature in Kelvin to activation energy for migration by Um 25kTanneal [3], where k is the
1E+23
CONCENTRATION (atoms/cm3)
1E+22
1E+21
C
1E+20
H N 1E+19
Cl F
1E+18
1E+17
1E+16 0
1000
2000
3000
4000
5000
6000
7000
8000
DEPTH (Angstroms)
Fig. 1. SIMS profiles of H, C, N, Cl, and F in the starting material.
9000
10000
Y. Li et al. / Applied Surface Science 231–232 (2004) 791–795
793
1E+20
Cu-SIMS: 10.0e13 atoms/cm2 1E+19
Cu-SRIM Simulation
CONCENTRATION (atoms/cm3)
Vacancies x0.001: SRIM simulation
1E+18
1E+17
1E+16
1E+15
1E+14 0
1000
2000
3000
4000
5000
6000
DEPTH (Angstroms)
Fig. 2. SIMS 63 Cu profile after 50 keV 63 Cu implantation to a dose of 1:0 1014 atoms/cm3, and simulated Cu concentration profile and vacancy profile (multiplied by a factor of 0.001) using SRIM.
Boltzmann constant and Tanneal is the anneal temperature. Thus, the apparent activation energy for activating Cu in the film is estimated to be below 2.33 eV. It is possible to roughly estimate the apparent ‘‘diffusion’’ coefficient at given temperature through the SIMS profiles before and after the annealing. At 800 8C the implanted Cu is clearly mobile (see Fig. 3). The profile becomes broader and some of the implanted Cu migrated into the deeper layers of the film and the interface where the irradiation damage is less. Using DX 2 ¼ 2DA t (i.e. via Brownian movement, see Ref. [4]), the apparent diffusion coefficient in the implanted film, DA 1:5 109 cm2/s. Here we have defined the mean square displacement of the migrating Cu, DX2, equal to (300 nm)2 as we roughly ˚ from curves estimated the Cu has moved about 3000 A 1 and 2 in Fig. 3 (i.e. 63 Cu trace after 800 8C RTA and 63 Cu trace after implantation). As compared with Al and Cu in Si, from the literature [5], at 800 8C, the diffusivity of Al in Si is lower than 1014 cm2/s and diffusivity of Cu in Si is higher than 107 cm2/s. These results indicate that Cu is also a fast diffuser in the
low-k film, although we have not found the ‘‘diffusivity’’ of Cu in standard SiO2 for comparison. The general behaviour of Cu migration in the irradiated low-k film can be summarized as follows: 1. The surface region and the interface between the film and silicon substrate are sinks for migrated Cu. The surface region contains more irradiation damage (see curve 3 in Fig. 3) and so even after RTA, the surface region has more trapping centers for Cu. The interfaces in SiO2/Si systems always are trapping centers for impurities. 2. Fig. 3 shows the interface region has trapped a large percentage of remaining Cu after RTA. It can be seen from Fig. 3 that some Cu has also moved beyond the interface. In order to see this behaviour more clearly, the low-k film was stripped using HF acid and the Cu was profiled in the Si alone. As shown in Fig. 4, for the 1100 8C (30 s) RTA piece, after stripping the film, the silicon substrate contained about 4:0 1012 atoms/cm2 63 Cu to a ˚ . It should be noted that the depth of about 1700 A
794
Y. Li et al. / Applied Surface Science 231–232 (2004) 791–795 1E+21
1: as-impl. :10e13 atoms/cm2
CONCENTRATION (atoms/cm3)
1E+20
2: 800C RTA Cu=2.0e13 atoms/cm2 3: 1000C RTA Cu=1.2e13 atoms/cm2 4: 1100C RTA Cu=8.6e12 atoms/cm2
1E+19
1E+18
1E+17
1E+16
1E+15
1E+14 0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
DEPTH (Angstroms) 63
Fig. 3. SIMS Cu profiles: curve 1: after implantation; curve 2: plus 800 8C RTA for 30 s; curve 3: plus 1000 8C RTA for 30 s; curve 4: plus 1100 8C RTA for 30 s.
RTA time is so short, however, Cu diffusivity in silicon is so high. Thus, it is not surprising to see some of Cu in the Si substrate since the very high diffusivity of Cu in Si at 1100 8C (i.e. 4 106 cm2/s at 1100 8C, the diffusion length (DAt)0.5 for 30 s could be as large as 1:09 102 cm). 3. The retained 63 Cu dose after RTA are as follows: 2:0 1013 atoms/cm 2 after 800 8C RTA, 1:2 1013 atoms/cm2 after 1000 8C RTA and 8:6 1012 atoms/cm2 after 1100 8C RTA. It should be noted the above calculation is based on the assumption that all measured Cu is in the low-k film since the quantified concentration profiles shown in Fig. 3 are based on the sensitivity factor obtained from the low-k matrix. For curves 2–4 shown in Fig. 3, since there is a small percentage of trapped Cu atoms at the interface and in the silicon substrate, the measurement conditions used underestimate the concentration of Cu in the silicon substrate since we did
not use a Si RSF in the region beyond the interface. However the error is small since the majority of the Cu resides in the low-k film or at the interface. For the 1100 8C annealed sample, the error is larger since the silicon trapped near half of the remained Cu. For this sample, as shown in Fig. 4 after stripped the low-k with HF acid, Cu analysis on the silicon substrate only is a more accurate method for calculation of the Cu in Si. It is clear from the retained dose calculations that most of the implanted Cu has ‘‘evaporated’’ due to the RTA process. At this point we do not have an explanation for the phenomenon. It is interest to note that another possibility: because Cu is such a fast diffuser in Si it may be that the Cu did not ‘‘evaporate’’ but moved all the way to the back of the wafer. In summary, Cu migrates at a temperature below 800 8C in the F-doped low-k film and the surface region and the interface between the film and silicon
Y. Li et al. / Applied Surface Science 231–232 (2004) 791–795
795
1E+20
CONCENTRATION (atoms/cm3)
1E+19
Cu=4.0e12 atoms/cm2 in Silicon 1E+18
1E+17
1E+16
1E+15
1E+14 0
1000
2000
3000
4000
5000
6000
DEPTH (Angstroms)
Fig. 4. SIMS
63
Cu profile in silicon substrate for the 1100 8C RTA sample after removal of the low-k film.
substrate are sinks for migrating Cu during the annealing. Cu is a fast diffuser in the F-doped low-k film.
[2]
[3]
References
[4]
[1]
[5]
Does Advanced Technology Really Matter, Intel 2002 Annual Report. http://www.intel.com.
J.F. Ziegler, J.P. Biersack, U. Littmark, The Stopping and Ranges of Ions in Solids, Pergamon Press, New York, 1985. T.D. Townsend, J.C. Kelly, N.E.W. Hartley, Ion Implantation and their Applications, Academic Press, New York, 1976. W. Jost, Diffusion in Solids, Liquid, Gases, Academic Press, New York, 1960. H.F. Wolf (Ed.), Silicon Semiconductor Data, Pergamon Press, New York, 1969.