Sie sind auf Seite 1von 10

Thin Solid Films 301 1997.

4554

Analysis of thin film stress measurement techniques


S.G. Malhotra
a

a,),1

, Z.U. Rek b, S.M. Yalisove a , J.C. Bilello

Department of Materials Science and Engineering, Uniersity of Michigan, Ann Arbor, MI 48109-2136, USA b Stanford Synchrotron Radiation Laboratory, Stanford, CA 94395, USA Received 15 July 1996; accepted 27 November 1996

Abstract Residual stresses in several magnetron sputtered Mo thin films, with thicknesses from 100 nm to 1.60 mm, were determined using double-crystal diffraction topography DCDT., sin2c , and the high-resolution X-ray diffraction technique HRXRD.. The Mo films had a range of microstructures that included random and polycrystalline, textured out-of-plane, and textured in-plane. When the average biaxial stresses over the entire film thickness were determined for the films using the aforementioned techniques, the results were comparable in magnitude. However, the stresses determined with the substrate curvature technique, DCDT, were consistently smaller than those obtained with the sin2c and HRXRD techniques. The difference may arise for several reasons. For example, the HRXRD and sin2c measurements of a textured film may not be indicative of the mean film stress, and thus may differ from the curvature measurement. Also, substrate curvature techniques measure extrinsic stresses, or stresses that arise solely from the presence of the substrate. The techniques which analyze the film directly, such as sin2c and HRXRD, measure the extrinsic stresses and the intrinsic stresses that arise from defects or morphology changes within the film. The additional information that can be obtained from the depth-sensitive HRXRD technique concerning stress variations within thin films was also highlighted. q 1997 Elsevier Science S.A.
Keywords: Depth profiling; Molybdenum; Stress; X-ray diffraction

1. Introduction The quantification of stress and strain in thin films has been an important matter for many years w1x because of the adverse effects stress and strain can have on the desired properties of thin film structures w2x. Many experimental techniques have been developed to measure the amount of stress and strain in thin films. The two general classes of commonly used techniques include deflection techniques based on determining the radius of curvature of the substrate w312x, and strain measurement techniques based on the direct measurements of interplanar spacings in the film using X-ray diffraction w1320x. The substrate curvature techniques include optical interferometry w4x, laser scanning w5x, and double-crystal diffraction topography DCDT. w9x. The curvatures obtained from the above techniques before and after film deposition are linearly related to the

Corresponding author. 46 Sterling St., Beacon, NY 12508, USA. Present address: IBM Corporation, Semiconductor Research and Development Center, 1580 Route 52, Hopewell Junction, NY 12533, USA.
1

average stress over the entire film thickness using the modified Stoneys equation w11,12x . The X-ray scattering methods which measure the interplanar spacings in the film directly include the sin2c X-ray diffraction technique w1517x, grazing incidence X-ray scattering GIXS. w1820x, and the high-resolution X-ray diffraction technique HRXRD. w21x. The sin2c technique is popular because it can be used to quickly determine the average biaxial stress in a film from a linear fit of d hkl vs. sin2c data, with as few as two c tilts. The GIXS and HRXRD techniques have the advantage of depth sensitivity, unlike the sin2c and substrate curvature techniques, because they utilize the phenomenon of total external reflection. The residual stress in Mo films, which play an important role as a major component in a strong and tough MorW multiscalar multilayer being developed for hightemperature coatings w22,23x, MorSi X-ray mirrors w24x, and as a bottom electrode for solar cells w25x, has been studied with several of the aforementioned techniques. For example, the average residual stress has been determined using bending beam w2628x, optical interferometry w29 33x, double-crystal diffraction topography w34,35x, sin2c

0040-6090r97r$17.00 q 1997 Elsevier Science S.A. All rights reserved.

46

S.G. Malhotra et al.r Thin Solid Films 301 (1997) 4554

w25,26,36,37x, grazing incidence X-ray scattering w38x, and high-resolution X-ray diffraction w21,39x. However, few detailed comparisons of the results obtained from these methods exist in the literature w26,40x. The goal of this paper was thus to conduct a detailed analysis of the capabilities of several of these techniques by determining the residual stress in Mo films that have a range of thicknesses and microstructures using DCDT, sin2c , and HRXRD. Previous work w41x showed a correlation between Mo film thickness and degree of texture, so the thicknesses of the Mo films used for this work were chosen accordingly. Thus, the residual stresses in magnetron sputtered Mo thin films with thicknesses of 100 nm, 170 nm, 260 nm, 800 nm, and 1.60 mm were determined.

2.2. Measurements on wiggler beamline The high-resolution X-ray diffraction HRXRD. experiments for all of the films and the sin2c experiments for the 100, 170, and 260 nm films were conducted at the Stanford Synchrotron Radiation Laboratory SSRL. under standard conditions 3 GeV and 100 mA at fill. on the eight-pole focused wiggler station BL 7-2. A Si 111. double-crystal monochromator was used to select the incident X-ray wavelength of 0.124 nm 10 keV. from the continuous spectrum. The horizontal and vertical divergence of the beam on BL 7-2 is 3 mrad and 0.2 mrad, respectively. Slits 1 mm = 1 mm were used for the incoming beam, and 1 mrad Soller slits were used for the diffracted beam to limit vertical divergence. The signal was detected with a solid state detector. The samples were mounted on an automated Huber 5020 four-circle goniometer. The dedicated beam-line 7-2 computer was used to control the goniometer motions, the shutter, and the photon counting. The experiments were conducted in the dose mode by putting a scintillation counter in the path of the incident beam, because the current in the synchrotron ring decreased linearly with time. The HRXRD technique utilizes the phenomenon of total external reflection w42x for depth sensitivity and both the symmetric and asymmetric grazing incidence geometries w20x so that crystallographic planes with a variety of orientations can be probed. The X-ray penetration depths were varied by changing the angle of the incoming X-ray near the critical angle for total external reflection of 0.348 for Mo at 10 keV w42x. The diffraction peaks collected in the symmetric and asymmetric geometries, at each penetration depth, are shown in Table 1 for the 1.60 mm film, and in previous publications for the other films w21,39x. All experiments were conducted using the laboratory frame of reference shown in Fig. 1a.. The 3104 diffraction peaks 2u angle of ; 778. were used for the sin2c analysis of the 100, 170, and 260 nm Mo films. The laboratory and sample reference frames used for the sin2c experiments are defined in Fig. 1b.. The whole wafers were mounted on a vacuum chuck and aligned to ensure that the sample normal was coincident with the azimuthal w . rotation axis. The interplanar spac-

2. Experimental procedures 2.1. Sample preparation Molybdenum films with thicknesses of 100 " 5 nm, 170 " 8.5 nm, 260 " 13.5 nm, 800 " 40 nm, and 1.60 mm " 0.08 mm were deposited from a 99.95% pure Mo target onto 75 mm diameter Si 100. wafers by direct current planar magnetron sputtering. The total film thicknesses was determined for the 100, 170, and 260 nm films by Rutherford backscattering using a tandem accelerator with Heqq ions. The total film thickness for the 800 nm and 1.60 mm films was determined by scanning electron microscopy and surface profilometry. The Si wafers were in the as-received condition, with a native oxide coating. The sputtering occurred without significant heating of the substrates. The sputtering power was 308 W ; 8 nm miny1 deposition rate. for the thinner films and 1.75 W ; 34 nm miny1 deposition rate. for the 1.60 mm film. The chamber was pumped to a base pressure ; 6 = 10y6 Torr and was then back-filled with Ar and maintained at a pressure of ; 1 = 10y2 Torr during the deposition. The samples were mounted face-down 5 inches above the sputter source in a horizontal platen which rotated at 20 rpm. The target was presputtered onto the shutter for at least 3 min to prevent oxides or contaminants from being sputtered on to the wafers.

Table 1 List of diffraction peaks collected at each penetration depth from the free surface of the 1.60 mm thick Mo film. In the symmetric geometry, the diffraction condition could be satisfied for specific planes in the highly textured grains, whereas in the asymmetric geometry, the diffraction condition could be satisfied by randomly oriented grains Penetration depth nm. 5, 10, 50, 120, 180 105, 110, 130, 255, 260 Peaks collected in symmetric geometry 110., 002., 112., 220., 222., 004. Peaks collected in asymmetric geometry 1104, 2004, 2114, 3104 1104, 2004, 2114, 3104

S.G. Malhotra et al.r Thin Solid Films 301 (1997) 4554

47

Fig. 1. Reference frames used for a. HRXRD analysis and b. sin2c experiments. In the reference frame used for the HRXRD analysis, the sample normal, incident X-ray, k o , and diffracted X-ray, k, comprise the plane of diffraction. The sample normal forms the z axis, the perpendicular to the diffraction plane the x axis, and the cross product of the x and z axes forms the y axis. In the reference frame used for the sin2c experiments, the strain is measured in the direction L3 , which is related to the sample reference frame by the angles f and c .

ing was measured in the laboratory reference frame along the direction L 3 , which is related to the sample coordinate system by the angles w and c . The 3104 peak was collected for at least seven positive and negative c tilts for the w angles of 08, 458, and 908. The data is primarily located between sin2c values of 0.16 and 0.40, which corresponds to incident angles of 18 and 158, respectively. Above an incident angle of 158, the scattering volume in the films was too small to yield a good signal-to-noise ratio. To supplement the data, an additional measurement of the 3104 d-spacing was made in the symmetric grazing incidence geometry, which corresponds to a sin2c value of 1.0. 2.3. Measurements on a rotating anode The pole figures for all of the films and the sin2c experiments for the 800 nm and 1.6 mm Mo films were conducted on a Rigaku Rotaflex equipped with a 12 kW Cu rotating anode source 0.154 nm radiation. and diffracted-beam graphite monochromator. The voltage and current used for the experiments were 50 kV and 150 ma, respectively, and the signal was detected with a scintillation counter. The nature and degree of the texture in the Mo films was measured from reflection X-ray pole figures. The pole figures were obtained from a Rigaku texture diffractometer configured in the Schultz geometry w43x. Pole figures for the 110., 200., and 211. poles were collected for all of the films, as well as for a random polycrystalline Mo standard. The Mo standard was prepared by powder pressing and sintering, had an average grain size of ) 10 mm, and was thick enough to absorb ; 100% of the incident rays. The degree of the texture in the Mo films was quantified by a times random approach, which is elucidated elsewhere w41x. The 3104 diffraction peaks 2 u angle of ; 1018. were

used for the sin2c analysis of the 800 nm film, and 3214 diffraction peaks 2 u angle of ; 1328. were used for the 1.60 mm Mo film. A section ; 15 mm = 15 mm was cut from the center of the coated wafers and mounted on a glass slide for the experiments. The samples were aligned to ensure that the sample normal was coincident with the azimuthal w . rotation axis. For the 800 nm film, the 3104 peak was collected for seven qc tilts and four yc tilts for the w angles of 08, 458, and 908. The c angles ranged from 30.718 to 49.218 sin2c of 0.26 and 0.57, respectively.. Because the 1.60 mm film was highly textured, diffraction peaks could be only be obtained at fixed w and c tilts. The 3214 peaks illustrated in Fig. 2 were used for the analysis, which correspond to w tilts of 308 and 1508, and c tilts of 19.118 and 40.898 sin2c of 0.11 and 0.43, respectively..

Fig. 2. Stereogram for a bcc material with a 110. in-plane texture which indicates the 3214 planes used for the sin2c analysis of the 1.60 mm Mo film. The 3214 planes are located at f angles of 308 and 1508 and c angles of 19.118 and 40.898.

48

S.G. Malhotra et al.r Thin Solid Films 301 (1997) 4554

2.4. Substrate curature measurements The average biaxial stresses in the films were determined by a macroscopic substrate curvature technique, double crystal diffraction topography DCDT.. A detailed description of the DCDT apparatus used for this study is contained in a previous publication w10x. In summary, the technique is based the observation that, for a bent singlecrystal substrate Si for this work. which is illuminated with highly parallel and monochromatic X-rays, only that portion of the crystal lattice which is in the Bragg condition will create a topographic image in an X-ray photographic film detector w68x. By initially capturing the topographic image of the portion of the sample in the Bragg condition, and then rotating the sample about an axis perpendicular to the plane of diffraction and exposing the film again, a series of bands will form on the film. From this information, the substrate curvature can be calculated. If the curvature of the substrate is determined before and after the film deposition, the average biaxial stress throughout the entire film thickness can be determined using the modified Stoneys equation w9,11,12x.

3. Results The type and degree of texture in the 800 nm and 1.60 mm Mo films were determined from the 110. reflection pole figures shown in Fig. 3. This two-dimensional pole figure illustrates the location of 110. planes, in three-dimensional space, as a function of the inclination angle a and the azimuthal angle b . The presence of intensity concentrations at the a angles of 908 and 308, and symmetry of the diffraction intensity at all b angles, indicated a 110. out-of-plane texture in the 800 nm film. Although the pole figures are not shown, the maximum times random values for the 100 nm, 170 nm, and 260 nm films when compared to a polycrystalline solid standard were 1, 2, and 2, respectively. As indicated, the maximum times random value for the 800 nm film was 6. The discontinuous distribution of diffraction intensity about the substrate normal for the 1.60 mm film indicated an in-plane alignment as well. The positions of the intensity concentrations indicated a 110. in-plane texture, and the times random for this film was determined to be 69. The 200. and 211. pole figures supported the above conclusions, but were not shown for brevity. The sin2c data is shown in Fig. 4 for the 100 nm, 170 nm, and 260 nm films analyzed on the wiggler beamline at SSRL. The high brightness, good collimation, and low horizontal and vertical divergence available at a wiggler beamline was necessary for the analysis of films in this thickness regime. The data for only one of the three w tilts is shown, but this is sufficient because the stress in the films was isotropic in the plane. An analysis of the slopes

of the d h k l vs. sin2c plots in Fig. 4 yielded average biaxial stresses of y0.530 GPa, y0.695 GPa, and y0.440 GPa for the 100 nm, 170 nm and 260 nm films, respectively. Raw X-ray diffraction data used for sin2c stress analysis must be corrected polarization and absorption, in addition to errors dependent upon sample alignment and beam optics w17x. The data was corrected for polarization and absorption according to standard equations w13x. A strainfree lanthanum hexaboride powder standard NIST SRM 660. was used to detect any systematic offset from sample misalignment during the experiment. However, the horizontal and vertical divergence of the beam on BL 7-2 is 3 mrad and 0.2 mrad, respectively, so that a correction for these factors was not necessary. The sin2c data is shown in Fig. 5 for the 800 nm and 1.6 mm films analyzed on the 12 kW rotating anode at the University of Michigan. The data was corrected for absorption, polarization, sample misalignment, and horizontal beam divergence. The error in the biaxial stresses for all sin2c experiments was calculated using standard equations w17x, and was "0.070 GPa for all films. This is the upper limit for the error, because the standard deviation in the peak positions was conservatively set to equal to the increment size of 15 = 10y3 8. An analysis of the slopes of the d h k l vs. sin2c plots in Fig. 5a. and 5b. yielded average biaxial stresses of y0.415 GPa and q0.068 GPa for the 800 nm and 1.60 mm films, respectively. The average residual stress data obtained from the substrate curvature technique, double crystal diffraction topography, is shown in Fig. 6. The magnitude of the error associated with these measurements was calculated using the equation given by Noyan and Goldsmith w40x. The largest contribution to the error came from the uncertainty in the film thicknesses, which was conservatively set at "5% of each total film thickness. The three-dimensional stresses as a function of X-ray

Fig. 3. 110. pole figures for a. 800 nm and b. 1.60 mm Mo films. In all of the pole figures the sample normal is parallel to the page normal. Relative intensity refers to the percentage of the maximum "times random" diffraction intensity in each pole figure. The maximum "times random" for each film is indicated in the upper right. The intensity concentrations indicate a 110. out-of-plane texture in the 800 nm film and 110. in-plane texture in the 1.60 mm film.

S.G. Malhotra et al.r Thin Solid Films 301 (1997) 4554

49

Fig. 4. Results of sin2c analysis conducted on wiggler beamline for a. 100 nm film, b. 170 nm film, and c. 260 nm film. The plots are linear and show no c-splitting.

penetration depth for the 260 nm, 800 nm, and 1.60 mm film, obtained from the HRXRD technique, are shown in Fig. 7. The HRXRD technique is described in detail elsewhere w21,39x, and was used to determine the entire strain tensor in the reference frame shown in Fig. 1a., as a function of X-ray penetration depth, with a least-squares methodology w44x. The strain eigenvalues and eigenvectors

were then calculated for each symmetric strain tensor using the traditional approach w45x. The strain eigenvectors were converted to stresses using Hookes Law and rotated into the laboratory reference frame. The isotropic elastic modulus, E, was used for the thinner films, and the elastic compliance tensor, which was rotated into the films frame of reference, was used for the highly textured 1.60 mm

Fig. 5. Results of sin2c analysis conducted on rotating anode for a. 800 nm film, and b. 1.60 mm film. The plots are linear and show no c-splitting.

50

S.G. Malhotra et al.r Thin Solid Films 301 (1997) 4554

Fig. 6. Average biaxial stresses for 100 nm, 170 nm, 260 nm, 800 nm, and 1.60 mm films determined with the substrate curvature technique, double crystal diffraction topography DCDT.. The line was added as a guide for the eye.

film. The stresses were not determined for large penetration depths in the 800 nm and 1.60 mm films with the HRXRD technique because of difficulties associated with penetrating the large depths into Mo ) 400 nm. in the symmetric grazing incidence geometry. For the 260 nm

and 800 nm films, the in-plane stresses, sx x and sy y , are approximately equal at each penetration depth, and maintain a nominally constant value, except for near the free surface. The normal stress, sz z , is tensile near the free surface of the 170 and 800 nm films, and is compressive near the free surface of the 1.60 mm film. The normal stress decreases in magnitude to a nominally constant value in each case. Error bars are drawn on the data points shown in Fig. 7 for both the penetration depths and the stresses. The origin of the error in the penetration depth depends on the uncertainty in the incident angle and how close the incident angle is to the critical angle. Near the critical angle the uncertainty in the penetration depth is greatest ; 30 nm.. The uncertainty in the crystallographic stresses arose from both systematic and random errors that occurred during the X-ray diffraction experiments. Great care was exercised to minimize systematic errors from instrument misalignment, specimen displacement, and beam divergence. Also, the random counting errors were minimized by taking the following three precautions: 1. the experiments were operated in the count mode, and at least 1 = 10 5 counts were collected from the incident beam for each data point; 2. the data were collected in angular

Fig. 7. Stresses obtained from HRXRD as a function of X-ray penetration depth for a. 260 nm, b. 800 nm, and c. 1.60 mm Mo films. For each film, the azimuthal stresses, sx x and sy y , remain close in magnitude at each penetration depth. The normal stress, sz z , is large near the free surface of each film, and then decreases to a nominally constant value.

S.G. Malhotra et al.r Thin Solid Films 301 (1997) 4554

51

increments of 10 = 10y3 to 15 = 10y3 8; and 3. many of the diffraction peaks were collected multiple times. The magnitude of the systematic error was evaluated by a careful analysis of LaB 6 standard NIST SRM 660. data. The magnitudes of the errors in the stress eigenvectors resolved onto the laboratory frame were determined using the approach described by Witte et al. w46x which determines how counting statistical errors propagate from the strain tensor into the principal stresses and directions. This error is dependent only on the precision of the diffraction peak locations, and for this work, the resulting error in the principal stresses was "0.100 GPa.
Fig. 8. Schematic of the stress ellipsoid, which illustrates the basis of the sin2c technique.

4. Discussion The objective of this work was to analyze Mo films that ranged from random and polycrystalline to highly textured using several stress measurement techniques. The results from a previous publication w41x that described the texture evolution of sputtered Mo films as a function of film thickness were used to choose films with the desired degree of texture. Thus, Mo films with thicknesses of 100 nm, 170 nm, 260 nm, 800 nm, and 1.60 mm were chosen for this study. The 110. pole figures in Fig. 3 show the transition from an out-of-plane alignment in an 800 nm film to an in-plane alignment in a 1.60 mm film. The results of the sin2c analysis conducted on the wiggler beamline and rotating anode sources are shown in Figs. 4 and 5. The basis of the sin2c technique is that it determines the average strain tensor over the entire film thickness, if the extinction depth of the X-rays is less than the film thickness. When three orthogonal, elastic stresses are applied to a spherical element of an isotropic elastic solid, it will respond by deforming homogeneously and elastically. The sphere deforms into the ellipsoid, shown in Fig. 8. The strain, e , at any point in the ellipsoid is given by the equation below:
2 e s a1 e 1 q a 2 e 2 q a 2 e 3 2 3

ances must be used to calculate the stress, as shown in Eq. 4. for a bcc material w47x: m ss 4. s44 2 s11 y 2 s12 y s44 2 d0 q sin w 2 4 where m is the slope of the d h k l vs. sin2c plot at a particular w rotation, and s11 , s12 , s44 are the elastic compliances. The values given by Huntington were used for this work w48x. The unstressed interplanar spacing also can be calculated using the slopes of the d h k l vs. sin2c plot at w tilts of 08 and 908 w17x. Because the relationship between d h k l and sin2c is linear for the films shown in Figs. 4 and 5, Eqs. 3. and 4. were used to calculate the biaxial stresses. The linear relationship also indicated that the average strain value was equal in all of the different sets of grains sampled during the different c-tilts, i.e. the average strain tensor was homogeneous in all of the grains w17x. In addition, there was no c-splitting, which means that the shears e 31 and e 32 were not significant. The results of the DCDT analysis are shown in Fig. 6, which indicate that the average biaxial stress decreases slightly in compressive nature as the film thickness increases. Previous analysis of the stress in sputtered Mo films with DCDT showed that the stress state in the films changed drastically for thicknesses below 20 nm, and then reached a nominally constant stress value for film thicknesses over 80 nm w10,35x. The basis of the DCDT technique is that there is a linear relationship between the substrate curvature, before and after film deposition, and the average biaxial stress over the entire film thickness, as described by the modified Stoneys equation w11,12x:

1.

where a1 , a 2 , and a3 are the direction cosines and e 1 , e 2 , and e 3 are the principal strains. If the direction cosines are substituted into Eq. 1., then Eq. 2. results:

e s e 1 sin2c cos 2w q e 2 sin2w sin2c q e 3 1 y sin2c .

2.

If Hookes Law is used and a biaxial stress state is assumed, then the average biaxial stress in an untextured film can be determined from Eq. 3.:

ss

mE d0 1 q n .

3.

ss

2 Et 0 K y K 0 .

6 1 y n . t

5.

where m is the slope of the d h k l vs. sin2c plot, E is the Elastic modulus of the film, d 0 is the unstressed interplanar spacing, and y is the Poissons ratio for the film. If the film is highly textured, however, the actual elastic compli-

where E is the Youngs modulus of the substrate, t 0 is the thickness of the substrate, K is the curvature of the coated substrate, K 0 is the curvature of the uncoated substrate, y is the Poissons ratio for the substrate, and t is the film

52

S.G. Malhotra et al.r Thin Solid Films 301 (1997) 4554

thickness. This form of the Stoneys equation is commonly used, and was employed here as a figure of merit. This equation was derived for the following case: 1. the coating is single film with uniform thickness, much less than t 0 ; 2. the stress is a uniform, isotropic plane stress; 3. the temperature is uniform; 4. the maximum deflection due to bending is t or2; and 5. the composite plate substrate plus film. is mechanically free. This equation also assumes that the curvature change of the composite plate in reference to the bare substrate is solely due to the film, and that there is complete continuity across the substratefilm interface. The stresses as a function of X-ray penetration depth obtained from HRXRD are shown Fig. 7. This data represents the stress ellipsoid at several penetration depths from the free surface. The azimuthal stresses, sx x and sy y , remain close in magnitude at each penetration depth. The normal stress, sz z , is tensile near the free surface of the 260 and 800 nm films, and then decreases in magnitude as the penetration depth increases. On the other hand, the normal stress, sz z , is compressive near the free surface of the 1.60 mm film, and then decreases in magnitude as the penetration depth increases. The absolute value of the magnitude of the normal stress in the top 5 nm of each film is a function of total film thickness. This trend is illustrated in Fig. 9 for all of the films analyzed for this paper. The non-zero surface stress is most likely due to the development of highly oriented and faceted grains in the thicker films, which would cause the surface normal for the film to deviate from the z axis in the laboratory reference frame. This trend is discussed in greater detail in a forthcoming publication w49x. This variation of the normal stress as a function of penetration depth represents information that cannot be obtained from the sin2c and substrate curvature techniques discussed above. The threedimensional stress tensor can be determined with the sin2c

Fig. 10. Plot of average biaxial stresses in 100 nm, 170 nm, 260 nm, 800 nm, and 1.60 mm Mo films as determined with HRXRD, DCDT, and sin2c techniques. All techniques indicate that the films become slightly more tensile as the total thickness increases.

Fig. 9. The absolute value of the magnitude of the normal stress, sz z , in the top 5 nm of each film as a function of total film thickness. The normal stress increases in magnitude for films below 800 nm, and then saturates at a value of about 4 GPa. The line was added as a guide for the eye.

technique, but depth sensitive information is difficult to obtain from films less than 500 nm in total thickness. The modified Stoneys equation, used in conjunction with the substrate curvature techniques, can only yield the average biaxial stress over the entire film thickness. All of the aforementioned techniques sin2c , DCDT, and HRXRD can determine the average biaxial stress over an entire film thickness. A comparison of the average biaxial stresses for the 100 nm, 170 nm, 260 nm, 800 nm, and 1.60 mm Mo films is illustrated in Fig. 10. As previously mentioned, the average biaxial stresses for the two thicker films were not determined with the HRXRD technique. The general trend among all of the techniques is that the average biaxial stress becomes slightly more tensile as total film thickness increases. Also, the DCDT technique yielded stresses that were least compressive in each case. A recent study which compared average biaxial stresses in thin film metallizations which were determined with substrate curvature and the sin2c techniques also identified a difference in the resulting stresses w40x. In this work, the authors studied thin Cu films that were deposited on 100. Si wafers by electroplating, electroless plating, vapor deposition, or by a combination of the three methods. They found that the stresses determined from the substrate curvatures exceeded the stresses obtained from the sin2c method in most cases, which is the opposite of what is shown in Fig. 10. The authors concluded that they were not clear why the techniques did not agree, but made the argument that the disagreement may be due to intrinsic stresses, interactions between the film and substrate, or an error in the film thicknesses. The argument that the disagreement may be due to intrinsic stress components is also plausible for the present case. Another study of residual stress in hard TiN coatings by mechanical and X-ray methods w50x noted a differential in the stresses. The authors noted that the sin2c values obtained from a set of

S.G. Malhotra et al.r Thin Solid Films 301 (1997) 4554

53

crystallographic planes in a textured film may not be representative of the mean film stress. This as well is a plausible argument for this case because the high intensity hkl 4 planes from textured grains were used for the HRXRD and sin2c analysis, and thus the result may deviate from that obtained from the substrate curvature technique.

Acknowledgements This work was supported by the USARO and ARPA under contract DAAH04-95-1-0120. The synchrotron work was conducted at the Stanford Synchrotron Radiation Laboratory SSRL., funded by the U.S. DoE. The authors wish to thank G. Brooks from the University of Michigan Dept. of Materials Science and Engineering, J. Kulman at the University of Michigan Solid State Electronics Laboratory, and Dr. S. Brennan at SSRL for technical assistance.

5. Conclusions This study compared the stresses obtained with the sin2c , DCDT and HRXRD techniques for Mo films that ranged from 100 nm to 1.60 mm in total thickness. The degree of texture varied in the films studied, as the 100 nm film was random and polycrystalline, the 170, 260, and 800 nm films were textured out-of-plane, and the 1.60 mm film was textured in-plane. When each technique was used to determine the average biaxial stress in the films, the results were comparable in magnitude. However, the average biaxial stresses determined with the substrate curvature technique, DCDT, were consistently smaller in magnitude than the average biaxial stresses determined with the sin2c and HRXRD techniques. The differential may arise because the HRXRD and sin2c measurements of a textured film may not be indicative of the mean film stress, or because the substrate curvature techniques measure extrinsic stresses, or stresses that arise solely from the presence of the substrate. The sin2c technique is popular because it is relatively quick and easy to use. Films with a thickness greater than 500 nm can be analyzed with a laboratory X-ray source, and if the d h k l vs. sin2c plot is linear, the average biaxial stress can be determined in as few as two c tilts. Substrate curvature techniques, such as DCDT, also enjoy widespread use because of ease of use. The HRXRD technique requires more data reduction, but it has the added advantage of depth sensitivity. The HRXRD data illustrated differences in the stress states of the films that were not evident with the sin2c and DCDT techniques. For example, the HRXRD data revealed the trends below. There is a large variation in the normal stress, sz z , with X-ray penetration depth in each film. The normal stress is not equal to zero near the free surface of the films with highly faceted grains because the surface normal is no longer coaxial with the z axis in the laboratory reference frame. The variation of the normal stress in films that have an average biaxial stress which is compressive 170 nm and 800 nm films. is different than the normal stress variation for films that have an average biaxial stress which is tensile 1.60 mm film.. The normal stress is tensile near the free surface of the former films, and compressive for the latter film. The origin of this behavior is most likely microstructural in origin, and is the subject of ongoing work.

References
w1x w2x w3x w4x w5x w6x w7x w8x w9x w10x w11x w12x w13x w14x N.N. Davidenkov, So. Phys. Solid State, 2 1961. 2595. H. Windischmann, Crit. Re. Solid State Mater. Sci., 17 1992. 547. W.D. Nix, Metall. Trans. A, 20A 1989. 2217. J.D. Finegan and R.W. Hoffman, Transactions of the Eigth National Vacuum Symposium, Pergamon Press, New York, 1961. P.A. Flinn in John C. Bravman, William D. Nix, David M. Barnett and David A. Smith eds., Mater. Res. Symp. Proc., 130 1989. 41. M. Renninger, Phys. Lett., 1 1962. 104. M. Renninger, Z. Phys., 19 1965. 20. M. Renninger, Z. Naturforsch., 160 1961. 1110. C.L. Kuo, P.E. Vanier and J.C. Bilello, J. Appl. Phys., 55 1984. 375. J. Tao, L.H. Lee and J.C. Bilello, J. Electron. Mater., 20 1991. 819. G.G. Stoney, Proc. Roy. Soc. Ser A, 82 1909. 172. R.W. Hoffman, in G. Haas and R.E. Thun eds.., Physics of Thin Films, Academic Press, New York, Vol. 111, 1966, p. 211. H.P. Klug and L.E. Alexander, X-Ray Diffraction Procedures, Wiley, New York, 1974. M.R. James and J.B. Cohen in Herbert Herman ed.., Treatise on Materials Science and Technology, Academic Press, New York, Vol. 19A, 1980. I.C. Noyan and J.B. Cohen in Eric Kula and Volker Weiss eds.., Residual Stress and Stress Relaxation, Sagamore Army Materials Research Conference Proceedings, Plenum Press, New York, 1982, pp. 117. H. Dolle and V. Hauk, Z. Metal., 68 1977. 728. I.C. Noyan and J.B. Cohen, Residual Stress Measurement by Diffraction and Interpretation, Springer, New York, 1987, pp. 117 163. W.C. Marra, P. Eisenberger and A.Y. Cho, J. Appl. Phys., 50 1979. 6927. P. Eisenberger and W.C. Marra, Phys. Re. Lett., 46 1981. 1081. P.H. Fouss and S. Brennan, Ann. Re. Mater. Sci., 20 1990. 365. S.G. Malhotra, Z.U. Rek, S.M. Yalisove and J.C. Bilello, J. Appl. Phys., 79 1996. 6872. M. Vill, D.P. Adams, S.M. Yalisove and J.C. Bilello, Acta. Metall. Mater., 43 1995. 427. D.P. Adams, M. Vill, J. Tao, J.C. Bilello and S.M. Yalisove, J. Appl. Phys., 74 1993. 1015. C. Montcalm, B.T. Sullivan, H. Pepin, J.A. Dobrowolski and M. Sutton, Appl. Optics, 33 1994. 2057. John H. Scofield, A. Duda, D. Albin, B.L. Ballard and P.K. Predecki, Thin Solid Films, 260 1995. 26. T.J. Vink, M.A.J. Somers, J.L.C. Daams and A.G. Dirks, J. Appl. Phys., 70 1991. 4301. T.J. Vink and J.B.A.D. vanZon, J. Vac. Sci. Technol. A, 9 1991. 124. Brent C. Bell and David Glocker, J. Vac. Sci. Technol. A, 10 1992. 1442.

w15x

w16x w17x

w18x w19x w20x w21x w22x w23x w24x w25x w26x w27x w28x

54

S.G. Malhotra et al.r Thin Solid Films 301 (1997) 4554 w40x I.C. Noyan and C.C. Goldsmith, Adances in X-Ray Analysis, Plenum Press, New York, Vol. 33, 1990, p. 137. w41x O.P. Karpenko, J.C. Bilello and S.M. Yalisove, J. Appl. Phys., 76 1994. 4610. w42x L.G. Parratt, Phys. Re., 95 1954. 359. w43x L.G. Schultz, J. Appl. Phys., 20 1949. 1030. w44x T. Imura, S. Weissmann and J.J. Slade, Jr., Acta. Crystallogr., 15 1962. 786. w45x J.F. Nye, Physical Properties of Crystals, Oxford University Press, Oxford, 1985. w46x D.A. Witte, R.A. Winholtz and S.P. Neal, Adances in X-Ray Analysis, Plenum Press, New York, Vol. 37, 1994, p. 186. w47x James Alexander Bain, Ph.D. Thesis, Stanford University, 1993. w48x H.B. Huntington, Solid State Phys., 7 1958. 213. w49x S.G. Malhotra, Z.U. Rek, S.M. Yalisove and J.C. Bilello, J. Appl. Phys., submitted. w50x P. Weise, T. Hirsh and P. Mayr, Evaluation of residual stresses by mechanical and x-ray methods in hard coatings, in P. Mayr ed.. Surface Engineering, DGM Informationsgesellschaft Verlag, Oberursel, Germany, 1993, p. 335.

w29x John A. Thornton and D.W. Hoffman, J. Vac. Sci. Technol. A, 3 1985. 576. w30x R.E. Cuthrell, D.M. Mattox, C.R. Peeples, P.L. Dreike and K.P. Lamppa, J. Vac. Sci. Technol. A, 6 1988. 2914. w31x D.W. Hoffman and John A. Thornton, J. Vac. Sci. Technol., 16 1979. 134. w32x D.W. Hoffman and John A. Thornton, J. Vac. Sci. Technol., 20 1982. 355. w33x Arthur G. Blachman, Met. Trans., 2 1971. 699. w34x J. Tao, L.H. Lee and J.C. Bilello, J. Elect. Mater., 20 1991. 819. w35x D.P. Adams, L.J. Parfitt, J.C. Bilello, S.M. Yalisove and Z.U. Rek, Thin Solid Films, 266 1995. 52. w36x P. Gergaud, J.J. Bacmann and J.L. Lebrun, Proc. of the Fourth Int. Conf. on Residual Stresses, Society of Experimental Mechanics, Bethel, CT, 1994, p. 1122. w37x M. Zaouali, J.L. Lebrun and P. Gergaud, Surf. Coatings Technol., 50 1991. 5. w38x B. Ballard, X. Zhu, P. Predecki and D. Braski, Proc. of the Fourth Int. Conf. on Residual Stresses, Society of Experimental Mechanics, Bethel, CT, 1994, p. 1133. w39x S.G. Malhotra, Z.U. Rek, S.M. Yalisove and J.C. Bilello, J. Vac. Sci. Tech., 1996., submitted.

Das könnte Ihnen auch gefallen