- Email: [email protected]
A study on resistance of PECVD silicon nitride thin film to thermal stress-induced cracking
Applied Surface Science 140 Ž1999. 12–18
A study on resistance of PECVD silicon nitride thin film to thermal stress-induced cracking Myung-Chan Jo a
a,)
, Sang-Kwon Park b, Sang-Joon Park
c
Department of Chemical Engineering, Dongseo UniÕersity, San 69-1, Churye-Dong, Sasang-gu, Pusan, 617-716, South Korea b Department of Chemical Engineering, Dongguk UniÕersity, Seoul, 100-272, South Korea c Department of Chemical Engineering, Kyungwon UniÕersity, Seongnam, 461-701, South Korea Received 3 March 1998; accepted 17 July 1998
Abstract A potentiostatic method was used to investigate the ability of Plasma-Enhanced Chemical Vapor Deposited ŽPECVD. silicon nitride thin films to protect aluminum bonding areas of integrated circuit chips. Four different films in thicknesses of ˚ were deposited on 0.9-mm-thick stainless steel substrate and CERDIP test chips by varying deposition variables. about 5 kA Ultimate strain of the films was measured by controlled strain apparatus and was constant at about 0.2%. Thermal shock and cycling test was conducted to investigate the resistance of the films to thermal stress-induced cracking.. After the test, the number of cracked bondpads on the CERDIP test chips was counted and corrosion current through cracks of the films on the stainless steel substrate was measured. The results showed that more tensile films were more susceptible to crack-induced failure. q 1999 Elsevier Science B.V. All rights reserved. PACS: 52.75. Rx; 68.60. yp; 81.15. Gh Keywords: PECVD; Silicon nitride; Thin film; Ultimate strain; Residual stress
1. Introduction Corrosion of aluminum metallization is by far the leading failure mechanism of integrated circuits in both molded plastic and hermetic packages w1–5x. Microscopic amounts of chloride, phosphoric acid or organic acids, together with water, can interrupt the electrical continuity of interconnect lines and bondpads to cause the loss in the functionality of the ) Corresponding author. Tel.: q82-51-315-3951, q82-0513 2 0 - 1 5 9 3 ; F a x : q 8 2 - 0 5 1 - 3 1 2 - 2 3 8 9 ; E - m a il: [email protected]
integrated circuit ŽIC. chips. Corrosive contaminants are ubiquitous in the manufacturing and operating environment, so their presence on an integrated circuit is almost impossible to avoid w6,7x. Over the coming years, ICs will become much more sensitive to contamination-induced failures than they are now because of the increase in the number of interconnect and the decrease in the wall thickness of packages. There is currently no effective passivation method for bondpads, bondwires, and bond interfaces in IC chip packaging w7,8x. As a postbond layer, polymeric films such as silicones and polyimides exhibit good mechanical properties, but are permeable to moisture
0169-4332r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 9 - 4 3 3 2 Ž 9 8 . 0 0 3 6 6 - 3
M.-C. Jo et al.r Applied Surface Science 140 (1999) 12–18
and intrusive contaminants. Ceramic films such as silicon nitride are excellent diffusion barrier against water and aggressive contaminants which might corrode aluminum metallization in packaged microelectronic assemblies. However, based on the bulk properties, films are prone to cracking during thermal cycling when used to coat bulk aluminum metallization, because of its low ultimate strain and wide mismatch of coefficient of thermal expansion ŽCTE. between metals and ceramics. Recent study indicated that PECVD silicon nitride thin films have higher ultimate strain than bulk samples w8x. In that study, silicon nitride thin films formed over aluminum substrate did not crack under thermal shock and cycling conditions. This may be attributed to the fact that the physical properties of thin films are often very different from those of bulk materials because of anisotropic atomic density distribution. Therefore, PECVD silicon nitride thin films have potential as an effective passivation layer over aluminum bonding areas in IC chip packaging. The purpose of this research was to investigate how resistance of PECVD silicon nitride thin films to thermal stress-induced cracking was associated with film’s mechanical properties which is dependent upon deposition variables. In order to do this, thermal shock and cycling test was employed to produce intentional cracks on the films by thermal stress. After the test, the amount of corrosion current through the cracks of the films under different deposition conditions was measured using a potentiostatic method and the number of cracked bondpads on CERDIP test chips was counted by a microscope. The results were correlated by the film’s residual stresses.
13
Table 1 Surface roughness of polished stainless steel strip with 0.05 mm slurry Longitudinal Scan lengthŽmm. Scan speed Stylus forceŽmg.
1st Scan 2nd Scan 3rd Scan 4th Scan Average
15 15 2
˚. Roughness ŽA 288 287 309 356 310
Transverse 15 15 2
˚. Roughness ŽA 765 542 397 432 534
using Dektak 3030 surface profilometer indicated an average roughnesses of about 0.03 and 0.05 mm, respectively. Each measurement was made by scanning different track on the surface. These values were about 10 times less than the thickness of the film used in this experiment, and thus, was considered to be smooth enough not to produce defective films by occasional peak on its surface. Aluminum-metallized silicon squares with 5 bondwires ultrasonically bonded in the shape of loops to simulate bonding areas in integrated circuit chips were used. These included an area of smooth, flat metallization representing bondpads, 10 stitch bonds, and five bondwires which is the most important features of a microelectronic assembly. The dimensions of the bondloop chips are as follows: 15 mm = 25 mm, 1-mm-thick Al metallization, 8 mil Al bondwires on Si wafer chip. 2.2. Deposition of silicon nitride thin film
2. Experimental 2.1. Test articles ATC-01 CERDIP test chip was used to simulate IC chip in commercial usage. Stainless steel 304 strip Ž90 mm = 20 mm = 0.9 mm. was used as a substrate for the silicon nitride thin film. To remove pinhole-induced defective films, the surface was polished with 0.05 micron slurry. As shown in Table 1, longitudinal and transverse scanning measurements
Silicon nitride films were deposited using Texas Instrument A-24 Reinberg-style, radial flow, parallel plate reactor. Silane ŽSiH 4 . and ammonia ŽNH 3 . gases were flowed into the reactor and reacted to form silicon nitride under RF electric field. Helium was used to dilute the silane gas for safety. Deposition variables are RF power, substrate temperature, pressure, input gas flowrate, and gas ratio ŽNH 3rSiH 4 .. Film thickness and refractive index were measured using a Gaertner L116B Auto Gain Ellipsometer.
14
M.-C. Jo et al.r Applied Surface Science 140 (1999) 12–18
2.3. Thermal shock and cycling test PECVD silicon nitride thin films over ATC-01 CERDIP chip, bondloop chip, and stainless steel strip were subjected to temperature cycling and shock in order to simulate typical industrial temperature stress testing. First, test articles were immersed in y658C solution Žacetone with dry ice. for 10 min. Then, they were transferred to 1508C solution Žethylene glycol. and immersed there for 10 min. This procedure was repeated 10 times and transfer time was less than 10 s. After the test, CERDIP and bondloop chips were immersed in 0.5 M NaCl solution for 30 min and were examined with a microscope in order to find any cracking that might have occurred. To determine the degree of cracking, the amount of corrosion current through the cracks in the film on stainless steel strip was measured using a designed potentiostatic apparatus as shown in Fig. 1. Pt cathode electrode was immersed in 0.5 M NaCl solution. The silicon nitride thin film surface was placed over two parallel knife-edge supports and contacted by a sponge of which bottom part is immersed in NaCl solution and thus, soaked up to the silicon nitride surface. The stainless steel strip was made the anode. While potentiostat was applied to the strip, corrosion current through the cracks on the film was measured with time using an EG and G Princeton Applied Research 363 Potentiometer.
Fig. 2. Measured current with potentiostatic voltage sweep at a rate of 1 mV sy1 .
To select optimum potentiostatic voltage, a bare stainless steel strip was tested by varying potentiostat from 0 to 0.8 V with a rate of 1 mV sy1 . Corrosion current under this condition was measured using the potentiometer and is shown in Fig. 2. From the results, 0.5 V was chosen as potentiostatic voltage throughout the experiment because it gave an measurable current of about 200 mA and was not high enough to break down the conductive solution. 2.4. Residual stress measurement To measure the residual stress of the silicon nitride film, 2-in.-diameter silicon wafer was used. Curvature difference of the silicon wafer before and after deposition was measured using a Dektak 3030 surface profilometer. This measured value of curvature difference was used to calculate the residual stress of the film by the following equation:
sf Žsi. s
Fig. 1. Potentiostatic apparatus for measuring ultimate strain and corrosion current.
2 EsŽsi. tsŽsi.
3 Ž 1 y nsŽsi. . t f
d
ž / r2
,
Ž 1.
where: sfŽsi. s residual stress of the film on silicon wafer Ždyn cmy2 .; EsŽsi.rŽ1 y nsŽsi. . for silicon wafer Ždyn cmy2 . s 2.31 = 10 12 ; tsŽsi. s thickness of silicon wafer Žcm.; t f s film thickness Žcm.; r s radial distance from the center of silicon wafer Žcm.; and d s curvature difference before and after deposition Žcm..
M.-C. Jo et al.r Applied Surface Science 140 (1999) 12–18
Residual stress and strain of the film on the stainless steel strip was calculated by the following equations:
sf Žss. s sf Žsi. q ´ f Žss. s
Ef 1ynf
sf Žss. Ž 1 y Õf . Ef
Ž a ss y a si . Ž Tr y Td . ,
Ž 2.
,
Ž 3.
where: sfŽss. s residual stress of the film on stainless steel strip Ždyn cmy2 .; ´ fŽss. s residual strain of the film on stainless steel strip; EfrŽ1 y n f . for silicon nitride Ždyn cmy2 . s 1.1 = 10 12 ; a ss s CTE of stainless steel Žppm 8Cy1 . s 17.3; a si s CTE of silicon wafer Žppm 8Cy1 . s 2.3; Tr s room temperature Ž8C.; and Td s substrate temperature Ž8C.. In Eq. Ž2., intrinsic stress of the film on the stainless steel strip was assumed to be equal to that on the silicon wafer. Intrinsic stress of the film was calculated by the following equation:
s iŽss. s sf Žss. y
Ef Ž a ss y a f . 1 y Õf
Ž Tr y Td . ,
Ž 4.
where: s iŽss. s intrinsic stress of the film on stainless steel strip Ždyn cmy2 .; and ´ fŽss. s CTE of silicon nitride Žppm 8Cy1 . s 1.5.The resolution of the residual stress calculated using the above equation was on the order of 10 6 dyn cmy2 . 2.5. Ultimate strain measurement
15
the other side of the strip. Forces were applied to the strip by adding weight over the weight table. Gradual increase in weight finally made crack occur on the silicon nitride thin film because of tensile stress. As shown in Fig. 1, Pt cathode electrode was immersed in 0.5 M NaCl solution. The silicon nitride thin film surface was contacted by a sponge of which bottom part is immersed in NaCl solution and thus soaked up to the silicon nitride surface. The stainless steel strip was made the anode. Therefore, applied controlled potentiostatic voltage drop across the electrodes lead to current flow through the crack to the stainless steel substrate when crack occurred on the film. When current started to flow, the forces applied was recorded at the instant and used for the calculation of the ultimate strain of the film. The ultimate strain of the thin film can be calculated from the following equation:
´aŽss. s
3mga EsŽss. wts2
,
Ž 5.
where: m s mass on the weight table Žg.; g s 980 cm sy2 ; a s distance between supporting knife-edge and near loading knife-edge Žcm. s 2; ts s thickness of stainless steel strip Žcm. s 0.09; w s width of stainless steel strip Žcm. s 2; and Es s Young’s modulus of stainless steel Ždyn cmy2 . s 1.93 = 10 12 .
3. Results and discussion
The potentiostatic apparatus shown in Fig. 1 was also used for measuring ultimate strain of the silicon nitride thin films. One side of the stainless steel strip with film was placed over two parallel knife-edge supports and two other parallel knife-edges which was connected to the weight table was placed over
3.1. Ultimate strain and residual stress of silicon nitride thin film Silicon nitride thin films were prepared under four different deposition conditions. The deposition con-
Table 2 Deposition conditions for PECVD silicon nitride thin films Condition a1
Condition a2
Condition a3
Condition a4
Temperature Ž8C. RF power ŽWatts. Ratio of SiH 4 rNH 3 SiH 4 flowrate Žsccm. NH 3 flowrate Žsccm. Pressure ŽTorr.
300 100 1r4 24 96 2.0
300 100 1r4 24 96 0.5
300 100 1r3 24 72 2.0
300 100 1r3 24 72 0.5
˚ miny1 . Deposition rate ŽA
262
52
213
52
16
M.-C. Jo et al.r Applied Surface Science 140 (1999) 12–18
Table 3 Characteristics for PECVD silicon nitride thin films Film on condition a1 Film on condition a2 Film on condition a3 Film on condition a4 Thickness of silicon wafer, tsŽsi. Žcm. 0.03048 Film thickness, tf Žcm. 0.00005632 Average, d r r 2 Žcmy1 . 0.00037 Density Žg cmy3 . 2.05 Refractive index 1.731 y0.00038 Residual strain, ´ fŽss. Ultimate strain, ´aŽs s. 0.0022
0.02794 0.00005122 0.000195 2.43 1.821 y0.00159 0.0022
0.032512 0.00005116 0.000355 2.28 1.748 y0.00252 0.0022
0.032258 0.00005761 y0.00009 2.65 1.850 y0.00636 0.0024
y Means compressive.
ditions are summarized in Table 2. As pressure increased, deposition rate increased. Table 3 shows the film’s characteristics obtained under deposition conditions described in Table 2. Density of the films with higher pressure was lower than that with lower pressure. This might be due to the fact that high pressure conditions cause rapid reaction of gases and thus, form less arranged molecular structure. Average drr 2 was obtained by measuring curvature difference of silicon wafer before and after deposition using surface profilometer and was used for the calculation of residual stress of the film. Residual strains of the films were all compressive and ultimate strains of the films were constant at about 0.2% for all films regardless of residual strain. As shown in Fig. 3, residual stress of the film on condition a4 was the most compressive among oth-
Fig. 3. Residual and intrinsic stresses of silicon nitride thin film.
ers. Intrinsic stresses of the films except that on condition a4 were tensile. However, residual stresses of the films were all compressive. This is because residual stress is the stress of the film at room temperature. 3.2. Resistance to cracking due to thermal stress CERDIP chips without thermal shock and cycling test and bare CERDIP chips without PECVD thin films were immersed in 0.5 M NaCl solution for 30 min and were examined by a microscope. Examination on bare CERDIP chips without films showed some corroded bondpads. Consequently, as mentioned above, there exists a possible contaminationinduced corrosion of metal bonding areas in microelectronic assembly and thus, passivation is necessary to protect those areas. Without thermal shock and cycling test, almost no corroded bondpads were found for all deposition conditions. The results indicated that PECVD silicon nitride thin films could be considered to be a potential protective films on the metal bonding areas. Thermal shock and cycling test was conducted to investigate the resistance to cracking according to film’s intrinsic and residual stress. While the film side of the stainless steel substrate was contacted with 0.5 M NaCl solution, 0.5 V potentiostat was applied across the substrate. Then, corrosion current through the cracks to the substrate was measured as a function of time. Before thermal shock and cycling test, all the films on the substrate was tested to check initial current flow and defective films were not used in this experiment. Fig. 4 shows that the current increased until 5 min of the potentiostatic experiment and then reached an
M.-C. Jo et al.r Applied Surface Science 140 (1999) 12–18
equilibrium state. The increase in the current is due to the corrosion of metal that is readily accessible to the solution under and around the cracks. As the corrosion front advances under the protective films, a steady state current is developed because the equilibrium state is reached between the dissolution of metal under the film and the rate that pieces of the unsupported film break off as a result of the evolution of corrosion products. As shown in Fig. 4, amount of corrosion current was maximum with bare stainless steel strip without film. For the film on condition a4, almost no corrosion current flowed. When these results were compared with the residual stress of the films in Fig. 3, it was found that more corrosion current was detected and thus, severer cracking occurred with more tensile films. Consequently, more tensile films were more susceptible to cracking due to thermal stress. CERDIP chips after thermal shock and cycling test were immersed in 0.5 M NaCl solution for 30 min and number of cracked bondpads of the chips were counted using a microscope. Fig. 5 shows that the number of cracked bondpads decreased with increased compressive stress. This result also indicated that more tensile films were subjected to severer cracking during thermal shock and cycling test. From the results, it was found that PECVD silicon nitride thin films have potential as a passivation layer over metal bonding areas in integrated circuit
17
Fig. 5. Number of cracked bondpads on CERDIP after thermal shock and cycling test.
chips and compressive films are thought to be resistant to thermal stress-induced cracking. Bondloop chips were also used in this experiment to study film’s resistance to cracking. However, visual examination could only confirm that severer cracking occurred on the films after thermal shock and cycling test and could not examine the film’s quality because of its inherent geometry such as occluded shape of stitch bonds. 4. Conclusion Ultimate strain of the silicon nitride thin film could be successfully measured using the designed apparatus. The ultimate strain of the silicon nitride films was constant at about 0.2% regardless of their residual stress. However, thermal shock and cycling test results indicated that film’s resistance to cracking due to thermal stress increased as the film’s residual stress became compressive. Consequently, compressive PECVD silicon nitride thin film can be considered to be a potential protective film over the aluminum bonding areas in microelectronic assemblies. Acknowledgements
Fig. 4. Corrosion current through cracks after thermal shock and cycling test.
This research was, in part, financially supported by a grant from the Advanced Materials and Applied
18
M.-C. Jo et al.r Applied Surface Science 140 (1999) 12–18
Technology Research Center ŽKyonggi Regional Research Center. at Kyungwon University.
References w1x R. Olberg, J. Bozarth, Microelectron. Reliab. 15 Ž1976. 601. w2x H. Inayoshi et al., Annu. Proc. of the Reliability Physics Symp., IEEE, New York 17 Ž1979. 113.
w3x K. Ritz et al., Annu. Proc. of the Reliability Physics Symp., IEEE, New York 25 Ž1987. 28. w4x R.K. Ulrich et al., Thin Solid Films 209 Ž1992. 73. w5x R.K. Ulrich et al., Thin Solid Films 224 Ž1993. 63. w6x K. Mittal, Treatise on Clean Surface Technology, Plenum, New York, Vol. 1, 1987, p. 211. w7x W. Paulson, R. Lorigan, Annu. Proc. of the Reliability Symp., IEEE, New York 14 Ž1976. 42. w8x D.H. Yi, PhD Dissertation, Dept. of Chemical Engineering, U. of Arkansas, 1991.
A study on resistance of PECVD silicon nitride thin film to thermal stress-induced cracking Myung-Chan Jo a
a,)
, Sang-Kwon Park b, Sang-Joon Park
c
Department of Chemical Engineering, Dongseo UniÕersity, San 69-1, Churye-Dong, Sasang-gu, Pusan, 617-716, South Korea b Department of Chemical Engineering, Dongguk UniÕersity, Seoul, 100-272, South Korea c Department of Chemical Engineering, Kyungwon UniÕersity, Seongnam, 461-701, South Korea Received 3 March 1998; accepted 17 July 1998
Abstract A potentiostatic method was used to investigate the ability of Plasma-Enhanced Chemical Vapor Deposited ŽPECVD. silicon nitride thin films to protect aluminum bonding areas of integrated circuit chips. Four different films in thicknesses of ˚ were deposited on 0.9-mm-thick stainless steel substrate and CERDIP test chips by varying deposition variables. about 5 kA Ultimate strain of the films was measured by controlled strain apparatus and was constant at about 0.2%. Thermal shock and cycling test was conducted to investigate the resistance of the films to thermal stress-induced cracking.. After the test, the number of cracked bondpads on the CERDIP test chips was counted and corrosion current through cracks of the films on the stainless steel substrate was measured. The results showed that more tensile films were more susceptible to crack-induced failure. q 1999 Elsevier Science B.V. All rights reserved. PACS: 52.75. Rx; 68.60. yp; 81.15. Gh Keywords: PECVD; Silicon nitride; Thin film; Ultimate strain; Residual stress
1. Introduction Corrosion of aluminum metallization is by far the leading failure mechanism of integrated circuits in both molded plastic and hermetic packages w1–5x. Microscopic amounts of chloride, phosphoric acid or organic acids, together with water, can interrupt the electrical continuity of interconnect lines and bondpads to cause the loss in the functionality of the ) Corresponding author. Tel.: q82-51-315-3951, q82-0513 2 0 - 1 5 9 3 ; F a x : q 8 2 - 0 5 1 - 3 1 2 - 2 3 8 9 ; E - m a il: [email protected]
integrated circuit ŽIC. chips. Corrosive contaminants are ubiquitous in the manufacturing and operating environment, so their presence on an integrated circuit is almost impossible to avoid w6,7x. Over the coming years, ICs will become much more sensitive to contamination-induced failures than they are now because of the increase in the number of interconnect and the decrease in the wall thickness of packages. There is currently no effective passivation method for bondpads, bondwires, and bond interfaces in IC chip packaging w7,8x. As a postbond layer, polymeric films such as silicones and polyimides exhibit good mechanical properties, but are permeable to moisture
0169-4332r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 9 - 4 3 3 2 Ž 9 8 . 0 0 3 6 6 - 3
M.-C. Jo et al.r Applied Surface Science 140 (1999) 12–18
and intrusive contaminants. Ceramic films such as silicon nitride are excellent diffusion barrier against water and aggressive contaminants which might corrode aluminum metallization in packaged microelectronic assemblies. However, based on the bulk properties, films are prone to cracking during thermal cycling when used to coat bulk aluminum metallization, because of its low ultimate strain and wide mismatch of coefficient of thermal expansion ŽCTE. between metals and ceramics. Recent study indicated that PECVD silicon nitride thin films have higher ultimate strain than bulk samples w8x. In that study, silicon nitride thin films formed over aluminum substrate did not crack under thermal shock and cycling conditions. This may be attributed to the fact that the physical properties of thin films are often very different from those of bulk materials because of anisotropic atomic density distribution. Therefore, PECVD silicon nitride thin films have potential as an effective passivation layer over aluminum bonding areas in IC chip packaging. The purpose of this research was to investigate how resistance of PECVD silicon nitride thin films to thermal stress-induced cracking was associated with film’s mechanical properties which is dependent upon deposition variables. In order to do this, thermal shock and cycling test was employed to produce intentional cracks on the films by thermal stress. After the test, the amount of corrosion current through the cracks of the films under different deposition conditions was measured using a potentiostatic method and the number of cracked bondpads on CERDIP test chips was counted by a microscope. The results were correlated by the film’s residual stresses.
13
Table 1 Surface roughness of polished stainless steel strip with 0.05 mm slurry Longitudinal Scan lengthŽmm. Scan speed Stylus forceŽmg.
1st Scan 2nd Scan 3rd Scan 4th Scan Average
15 15 2
˚. Roughness ŽA 288 287 309 356 310
Transverse 15 15 2
˚. Roughness ŽA 765 542 397 432 534
using Dektak 3030 surface profilometer indicated an average roughnesses of about 0.03 and 0.05 mm, respectively. Each measurement was made by scanning different track on the surface. These values were about 10 times less than the thickness of the film used in this experiment, and thus, was considered to be smooth enough not to produce defective films by occasional peak on its surface. Aluminum-metallized silicon squares with 5 bondwires ultrasonically bonded in the shape of loops to simulate bonding areas in integrated circuit chips were used. These included an area of smooth, flat metallization representing bondpads, 10 stitch bonds, and five bondwires which is the most important features of a microelectronic assembly. The dimensions of the bondloop chips are as follows: 15 mm = 25 mm, 1-mm-thick Al metallization, 8 mil Al bondwires on Si wafer chip. 2.2. Deposition of silicon nitride thin film
2. Experimental 2.1. Test articles ATC-01 CERDIP test chip was used to simulate IC chip in commercial usage. Stainless steel 304 strip Ž90 mm = 20 mm = 0.9 mm. was used as a substrate for the silicon nitride thin film. To remove pinhole-induced defective films, the surface was polished with 0.05 micron slurry. As shown in Table 1, longitudinal and transverse scanning measurements
Silicon nitride films were deposited using Texas Instrument A-24 Reinberg-style, radial flow, parallel plate reactor. Silane ŽSiH 4 . and ammonia ŽNH 3 . gases were flowed into the reactor and reacted to form silicon nitride under RF electric field. Helium was used to dilute the silane gas for safety. Deposition variables are RF power, substrate temperature, pressure, input gas flowrate, and gas ratio ŽNH 3rSiH 4 .. Film thickness and refractive index were measured using a Gaertner L116B Auto Gain Ellipsometer.
14
M.-C. Jo et al.r Applied Surface Science 140 (1999) 12–18
2.3. Thermal shock and cycling test PECVD silicon nitride thin films over ATC-01 CERDIP chip, bondloop chip, and stainless steel strip were subjected to temperature cycling and shock in order to simulate typical industrial temperature stress testing. First, test articles were immersed in y658C solution Žacetone with dry ice. for 10 min. Then, they were transferred to 1508C solution Žethylene glycol. and immersed there for 10 min. This procedure was repeated 10 times and transfer time was less than 10 s. After the test, CERDIP and bondloop chips were immersed in 0.5 M NaCl solution for 30 min and were examined with a microscope in order to find any cracking that might have occurred. To determine the degree of cracking, the amount of corrosion current through the cracks in the film on stainless steel strip was measured using a designed potentiostatic apparatus as shown in Fig. 1. Pt cathode electrode was immersed in 0.5 M NaCl solution. The silicon nitride thin film surface was placed over two parallel knife-edge supports and contacted by a sponge of which bottom part is immersed in NaCl solution and thus, soaked up to the silicon nitride surface. The stainless steel strip was made the anode. While potentiostat was applied to the strip, corrosion current through the cracks on the film was measured with time using an EG and G Princeton Applied Research 363 Potentiometer.
Fig. 2. Measured current with potentiostatic voltage sweep at a rate of 1 mV sy1 .
To select optimum potentiostatic voltage, a bare stainless steel strip was tested by varying potentiostat from 0 to 0.8 V with a rate of 1 mV sy1 . Corrosion current under this condition was measured using the potentiometer and is shown in Fig. 2. From the results, 0.5 V was chosen as potentiostatic voltage throughout the experiment because it gave an measurable current of about 200 mA and was not high enough to break down the conductive solution. 2.4. Residual stress measurement To measure the residual stress of the silicon nitride film, 2-in.-diameter silicon wafer was used. Curvature difference of the silicon wafer before and after deposition was measured using a Dektak 3030 surface profilometer. This measured value of curvature difference was used to calculate the residual stress of the film by the following equation:
sf Žsi. s
Fig. 1. Potentiostatic apparatus for measuring ultimate strain and corrosion current.
2 EsŽsi. tsŽsi.
3 Ž 1 y nsŽsi. . t f
d
ž / r2
,
Ž 1.
where: sfŽsi. s residual stress of the film on silicon wafer Ždyn cmy2 .; EsŽsi.rŽ1 y nsŽsi. . for silicon wafer Ždyn cmy2 . s 2.31 = 10 12 ; tsŽsi. s thickness of silicon wafer Žcm.; t f s film thickness Žcm.; r s radial distance from the center of silicon wafer Žcm.; and d s curvature difference before and after deposition Žcm..
M.-C. Jo et al.r Applied Surface Science 140 (1999) 12–18
Residual stress and strain of the film on the stainless steel strip was calculated by the following equations:
sf Žss. s sf Žsi. q ´ f Žss. s
Ef 1ynf
sf Žss. Ž 1 y Õf . Ef
Ž a ss y a si . Ž Tr y Td . ,
Ž 2.
,
Ž 3.
where: sfŽss. s residual stress of the film on stainless steel strip Ždyn cmy2 .; ´ fŽss. s residual strain of the film on stainless steel strip; EfrŽ1 y n f . for silicon nitride Ždyn cmy2 . s 1.1 = 10 12 ; a ss s CTE of stainless steel Žppm 8Cy1 . s 17.3; a si s CTE of silicon wafer Žppm 8Cy1 . s 2.3; Tr s room temperature Ž8C.; and Td s substrate temperature Ž8C.. In Eq. Ž2., intrinsic stress of the film on the stainless steel strip was assumed to be equal to that on the silicon wafer. Intrinsic stress of the film was calculated by the following equation:
s iŽss. s sf Žss. y
Ef Ž a ss y a f . 1 y Õf
Ž Tr y Td . ,
Ž 4.
where: s iŽss. s intrinsic stress of the film on stainless steel strip Ždyn cmy2 .; and ´ fŽss. s CTE of silicon nitride Žppm 8Cy1 . s 1.5.The resolution of the residual stress calculated using the above equation was on the order of 10 6 dyn cmy2 . 2.5. Ultimate strain measurement
15
the other side of the strip. Forces were applied to the strip by adding weight over the weight table. Gradual increase in weight finally made crack occur on the silicon nitride thin film because of tensile stress. As shown in Fig. 1, Pt cathode electrode was immersed in 0.5 M NaCl solution. The silicon nitride thin film surface was contacted by a sponge of which bottom part is immersed in NaCl solution and thus soaked up to the silicon nitride surface. The stainless steel strip was made the anode. Therefore, applied controlled potentiostatic voltage drop across the electrodes lead to current flow through the crack to the stainless steel substrate when crack occurred on the film. When current started to flow, the forces applied was recorded at the instant and used for the calculation of the ultimate strain of the film. The ultimate strain of the thin film can be calculated from the following equation:
´aŽss. s
3mga EsŽss. wts2
,
Ž 5.
where: m s mass on the weight table Žg.; g s 980 cm sy2 ; a s distance between supporting knife-edge and near loading knife-edge Žcm. s 2; ts s thickness of stainless steel strip Žcm. s 0.09; w s width of stainless steel strip Žcm. s 2; and Es s Young’s modulus of stainless steel Ždyn cmy2 . s 1.93 = 10 12 .
3. Results and discussion
The potentiostatic apparatus shown in Fig. 1 was also used for measuring ultimate strain of the silicon nitride thin films. One side of the stainless steel strip with film was placed over two parallel knife-edge supports and two other parallel knife-edges which was connected to the weight table was placed over
3.1. Ultimate strain and residual stress of silicon nitride thin film Silicon nitride thin films were prepared under four different deposition conditions. The deposition con-
Table 2 Deposition conditions for PECVD silicon nitride thin films Condition a1
Condition a2
Condition a3
Condition a4
Temperature Ž8C. RF power ŽWatts. Ratio of SiH 4 rNH 3 SiH 4 flowrate Žsccm. NH 3 flowrate Žsccm. Pressure ŽTorr.
300 100 1r4 24 96 2.0
300 100 1r4 24 96 0.5
300 100 1r3 24 72 2.0
300 100 1r3 24 72 0.5
˚ miny1 . Deposition rate ŽA
262
52
213
52
16
M.-C. Jo et al.r Applied Surface Science 140 (1999) 12–18
Table 3 Characteristics for PECVD silicon nitride thin films Film on condition a1 Film on condition a2 Film on condition a3 Film on condition a4 Thickness of silicon wafer, tsŽsi. Žcm. 0.03048 Film thickness, tf Žcm. 0.00005632 Average, d r r 2 Žcmy1 . 0.00037 Density Žg cmy3 . 2.05 Refractive index 1.731 y0.00038 Residual strain, ´ fŽss. Ultimate strain, ´aŽs s. 0.0022
0.02794 0.00005122 0.000195 2.43 1.821 y0.00159 0.0022
0.032512 0.00005116 0.000355 2.28 1.748 y0.00252 0.0022
0.032258 0.00005761 y0.00009 2.65 1.850 y0.00636 0.0024
y Means compressive.
ditions are summarized in Table 2. As pressure increased, deposition rate increased. Table 3 shows the film’s characteristics obtained under deposition conditions described in Table 2. Density of the films with higher pressure was lower than that with lower pressure. This might be due to the fact that high pressure conditions cause rapid reaction of gases and thus, form less arranged molecular structure. Average drr 2 was obtained by measuring curvature difference of silicon wafer before and after deposition using surface profilometer and was used for the calculation of residual stress of the film. Residual strains of the films were all compressive and ultimate strains of the films were constant at about 0.2% for all films regardless of residual strain. As shown in Fig. 3, residual stress of the film on condition a4 was the most compressive among oth-
Fig. 3. Residual and intrinsic stresses of silicon nitride thin film.
ers. Intrinsic stresses of the films except that on condition a4 were tensile. However, residual stresses of the films were all compressive. This is because residual stress is the stress of the film at room temperature. 3.2. Resistance to cracking due to thermal stress CERDIP chips without thermal shock and cycling test and bare CERDIP chips without PECVD thin films were immersed in 0.5 M NaCl solution for 30 min and were examined by a microscope. Examination on bare CERDIP chips without films showed some corroded bondpads. Consequently, as mentioned above, there exists a possible contaminationinduced corrosion of metal bonding areas in microelectronic assembly and thus, passivation is necessary to protect those areas. Without thermal shock and cycling test, almost no corroded bondpads were found for all deposition conditions. The results indicated that PECVD silicon nitride thin films could be considered to be a potential protective films on the metal bonding areas. Thermal shock and cycling test was conducted to investigate the resistance to cracking according to film’s intrinsic and residual stress. While the film side of the stainless steel substrate was contacted with 0.5 M NaCl solution, 0.5 V potentiostat was applied across the substrate. Then, corrosion current through the cracks to the substrate was measured as a function of time. Before thermal shock and cycling test, all the films on the substrate was tested to check initial current flow and defective films were not used in this experiment. Fig. 4 shows that the current increased until 5 min of the potentiostatic experiment and then reached an
M.-C. Jo et al.r Applied Surface Science 140 (1999) 12–18
equilibrium state. The increase in the current is due to the corrosion of metal that is readily accessible to the solution under and around the cracks. As the corrosion front advances under the protective films, a steady state current is developed because the equilibrium state is reached between the dissolution of metal under the film and the rate that pieces of the unsupported film break off as a result of the evolution of corrosion products. As shown in Fig. 4, amount of corrosion current was maximum with bare stainless steel strip without film. For the film on condition a4, almost no corrosion current flowed. When these results were compared with the residual stress of the films in Fig. 3, it was found that more corrosion current was detected and thus, severer cracking occurred with more tensile films. Consequently, more tensile films were more susceptible to cracking due to thermal stress. CERDIP chips after thermal shock and cycling test were immersed in 0.5 M NaCl solution for 30 min and number of cracked bondpads of the chips were counted using a microscope. Fig. 5 shows that the number of cracked bondpads decreased with increased compressive stress. This result also indicated that more tensile films were subjected to severer cracking during thermal shock and cycling test. From the results, it was found that PECVD silicon nitride thin films have potential as a passivation layer over metal bonding areas in integrated circuit
17
Fig. 5. Number of cracked bondpads on CERDIP after thermal shock and cycling test.
chips and compressive films are thought to be resistant to thermal stress-induced cracking. Bondloop chips were also used in this experiment to study film’s resistance to cracking. However, visual examination could only confirm that severer cracking occurred on the films after thermal shock and cycling test and could not examine the film’s quality because of its inherent geometry such as occluded shape of stitch bonds. 4. Conclusion Ultimate strain of the silicon nitride thin film could be successfully measured using the designed apparatus. The ultimate strain of the silicon nitride films was constant at about 0.2% regardless of their residual stress. However, thermal shock and cycling test results indicated that film’s resistance to cracking due to thermal stress increased as the film’s residual stress became compressive. Consequently, compressive PECVD silicon nitride thin film can be considered to be a potential protective film over the aluminum bonding areas in microelectronic assemblies. Acknowledgements
Fig. 4. Corrosion current through cracks after thermal shock and cycling test.
This research was, in part, financially supported by a grant from the Advanced Materials and Applied
18
M.-C. Jo et al.r Applied Surface Science 140 (1999) 12–18
Technology Research Center ŽKyonggi Regional Research Center. at Kyungwon University.
References w1x R. Olberg, J. Bozarth, Microelectron. Reliab. 15 Ž1976. 601. w2x H. Inayoshi et al., Annu. Proc. of the Reliability Physics Symp., IEEE, New York 17 Ž1979. 113.
w3x K. Ritz et al., Annu. Proc. of the Reliability Physics Symp., IEEE, New York 25 Ž1987. 28. w4x R.K. Ulrich et al., Thin Solid Films 209 Ž1992. 73. w5x R.K. Ulrich et al., Thin Solid Films 224 Ž1993. 63. w6x K. Mittal, Treatise on Clean Surface Technology, Plenum, New York, Vol. 1, 1987, p. 211. w7x W. Paulson, R. Lorigan, Annu. Proc. of the Reliability Symp., IEEE, New York 14 Ž1976. 42. w8x D.H. Yi, PhD Dissertation, Dept. of Chemical Engineering, U. of Arkansas, 1991.