Thin Solid Films, I29 (1985) 71-78 METALLURGICAL
THE INTERNAL FILMS R. ABERMANN
AND R. KOCH
Institute of Physical Chemistry, University of Innsbruck, Innrain 52 a, A-6020 Innsbruck (Austria) (Received
July 26. 1984; accepted
The internal stress of silver, copper and gold films was measured as a function of film thickness during deposition and as a function of time after deposition under ultrahigh vacuum conditions. A model for the origin of the internal stress was used to interpret each stress curve in terms of the corresponding film structure. During the deposition of the three metals onto MgF, substrate films tensile as well as compressive stresses are found as the film thickness increases and the maximum tensile stress occurs at the thickness at which the metal films become completely continuous. A compressive strain, built up at the metal substrate interface during the coalescence stage, is assumed to be the origin of the compressive stress in the continuous metal film. The tensile stress change after deposition of these metals is attributed to recrystallization and annealing processes that occur in the films. The compressive interface strain, which is maintained during this recrystallization, again determines the compressive stress measured during the deposition of a second metal layer. The growth of this second metal film is a continuation of the growth of the first film without renewed nucleation.
In earlier paperslp3 we have reported internal stress measurements on silver, gold and copper films deposited under high vacuum conditions. Since the unambiguous separation of the effect of certain parameters determining film growth was not always possible we have recently developed a stress measuring apparatus compatible with ultrahigh vacuum requirements. First experiments with this device have shown that under ultrahigh vacuum conditions the film stress differs significantly from that measured under high vacuum conditions. These experiments have also shown that perfect reproducibility of the film stress as a function of thickness is only achieved at total pressures in the lo-’ Pa range during the metal evaporation4. (Only in this pressure range does the effect of small variations in the residual gas composition become smaller than the experimental error of the technique.) In the first part of the present work we therefore reinvestigated the stress versus thickness curves during the deposition of the above metals onto MgF, 0040-6090/85/$3.30
0 Elsevier Sequoia/Printed
in The Netherlands
substrate films. Furthermore, the significantly improved reproducibility and the more defined vacuum conditions made it possible to investigate the stress changes after the metal deposition as a function of time. Finally we compare stress versus thickness curves measured during the deposition of the metals onto MgF, with those measured during the deposition of the metals onto metal in order to detect possible differences in the film growth on different substrate films. In these experiments a film 100 nm thick was evaporated first and the film stress was allowed to equilibrate until the rate of change was negligible before the additional metal film was deposited. 2.
The stress measuring apparatus was mounted in a stainless steel ultrahigh vacuum system with two diffusion pumps and two titanium sublimation pumps cooled with liquid nitrogen. A liquid nitrogen trap between the diffusion pump and the vacuum chamber reduced oil backstreaming. The base pressure was better than 4 x lo-’ Pa and with both evaporators (metal and MgF,) in operation the pressure was about (3-5) x lo-’ Pa. The substrate temperature was kept at 300 K by water from a thermostat running through the various radiation shields around the bending beam apparatus and the evaporators. The residual gas composition was analysed with a quadrupole mass analyser (Balzers QMG 3 1I ). The mean thickness of the films was measured with a quartz crystal oscillator (Balzers QSG 201). The deposition rate was 0.1 kO.01 nm s- r. The details of the apparatus will be given elsewhere4. After the ultrahigh vacuum system had been pumped down the MgF, substrate film 7 nm thick and the first metal film were evaporated in immediate succession. The evaporation of the MgF, film onto the bending glass beam eliminates the influence of contamination layers on the glass surface. In comparison with earlier experiments the MgF, film thickness was increased in order to reduce further the influence of thermal effects during the metal deposition. The stress curves reported in this paper have been corrected for these residual thermal effects as discussed in a separate paper”. After the deposition of a metal film 100nm thick the stress changes were measured until the rate of change was negligible. After 100 nm of the same metal had again been deposited and the film stress had again equilibrated, an MgF, layer was deposited and the corresponding stress was measured to allow a comparison of the stress curves for metal on metal with those for MgF, on metal. For the investigation of the microstructure of the evaporated metal films the MgFJmetal thin film system was deposited under identical vacuum conditions onto carbon-coated 400 mesh copper grids in separate experiments. Electron micrographs were taken at a magnification of 20000x-40000x using a Siemens Elmiscope 1A at an accelerating voltage of 80 kV. 3.
I shows the stress wrsus
curves for silver, copper and gold on
Ag, CU AND Au FILMS
Fig. 1. Internal stress measured (a) during and (b) after the evaporation (- - -) and gold (. .) onto 7 nm of MgF,.
of 100 nm of silver (-
MgF, substrates and the stress changes occurring after the metal deposition as a function of time. A small compressive stress is initially measured for silver, copper and gold. Above a mean film thickness of about 2 nm the film stress is tensile in each case. The onset of electrical conductivity is found at 6, 7 and 11 nm in the case of copper, gold and silver respectively at an evaporation rate of 0.1 nm s- ‘. The coalescence of the previously isolated metal clusters manifests itself as an increasing incremental stress above this thickness range. When the coalescence stage is completed the tensile stress reaches its maximum value at a mean thickness of 14.5, 16 and 23 nm for copper, gold and silver respectively. The internal stress of the continuous films is compressive. While the incremental stress of copper is only slightly higher than that of silver it is significantly greater than that of gold. According to our stress model this compressive stress arises from the fact that the isolated metal clusters grow with a lattice spacing slightly smaller than the bulk spacing because of the surface tension of the metal. As the clusters grow the lattice spacing tends towards the bulk spacing. Consequently, owing to the adhesion between the metal and the substrate, a compressive strain is built up at the metalsubstrate interface which reaches its maximum value at the end of the coalescence stage. The size of this interface strain is influenced by the surface tension of the metal and the size and adhesion of the metal particles on the substrate. The average surface tension of gold, copper and silver has been measured by Vermaak and KuhlmannWilsdorf’ and Wassermann and Vermaak’** respectively. These researchers reported surface stress values of 1.175 kO.093 N m-r (50 “C) for gold, O&O.450 N m-l (48 “C) for copper and 1.415 kO.300 N m-r (55 “C) for silver. Since gold and copper films become completely continuous at about the same mean thickness, which indicates comparable average grain sizes (Fig. 2), the respective incremental compressive stress in both cases agrees well with that expected from the corresponding surface stress value reported by these researchers. The incremental compressive stress of silver, however, should only be slightly lower than that of gold. The large
0.2 pm (c) Fig. 2. Electron micrographs onto 7 nm of MgF2.
of (a) 24 nm of silver,(b)
15 nm of copper and (c) 13.5 nm of gold evaporated
deviation from the expected value in this case can only be understood by assuming that the interface strain is released by slipping of the crystallites along the interface as soon as the strain reaches a critical value. From this interpretation it follows that the adhesion between the metal and the substrate film is good enough to retain the interface strain built up in the coalescence stage in the case of gold and copper while it is insufficient in the case of silver. As the film thickness of the continuous film increases, the initial incremental stress gradually decreases in each case. In terms of our stress model this indicates increasing recrystallization in the metal film with increasing film thickness. This effect is most obvious in the case of gold. This recrystallization is then the primary reason for the tensile stress change at the end of the film deposition, which continues for a fairly long time as can be seen from Fig. l(b). The magnitude and time to completion of this process are least in the case of silver and comparatively large in the case of gold. The degree of recrystallization occurring during the film deposition, deduced from the changes in the incremental film stress, agrees with that resulting from the stress change during a comparable time period after the film evaporation. The recrystallization was studied with films of different thickness. The result of these experiments was an increasing stress change in thicker films, and as expected from the lower annealing rate in thicker films an increase in the time to complete recrystallization. The extended annealing period in the case of gold films makes it possible to follow the corresponding changes in the film structure in the transmission electron microscope. Figure 3 shows three electron micrographs of the same gold film 100 nm thick taken at different times after the film deposition which clearly confirm the interpretation of the stress changes given above. The tensile stress change at the end of the metal evaporation could in principle also originate from a relaxation of the compressive strain built up at the metalsubstrate interface during the film deposition. To settle this question we have measured the stress changes as a function of thickness during the deposition of a second metal layer. The previously evaporated metal film was allowed to recrystallize until the stress changes were negligible. The results of these experiments are shown in Fig. 4. The high incremental compressive stress at the beginning of this second evaporation indicates that the compressive interface strain was not relaxed during the preceding recrystallization of the substrate film. The dominant contribution of the strain mechanism at the start is counteracted as soon as recrystallization again starts to play a role, which reduces the initial incremental stress until it eventually reaches a value which is identical with that found at higher film thicknesses during the first metal evaporation. The growth ofthe second layer is thus more or less a continuation of the growth of the first metal layer without new nucleation on the surface of the metal substrate. Since it has been argued in the literature *- l1 that the compressive stress measured for silver, gold and copper originates from the thermal load from the evaporation source owing to the large difference in the thermal expansion coefficient between the metal and the bending beam, we deposited an MgF, film 20 nm thick onto the various metal films. The temperature of the MgF, source and thus the thermal load to the bending beam was comparable with that for the abovementioned metals. The results of these measurements are shown in Fig. 5. In all three
Fig. 3. Electron micrographs of a gold evaporation onto 7 nm of MgF2.
film 100 nm thick
I h, (b) 3 h and Ic) 20 h after
STRESS IN 4s
( n m)
Fig. 4. Internal stress measured during the evaporation of 100 nm of silver onto 100 nm of silver (-), 100 nm of copper onto 100 nm of copper (- - -) and 100 nm of gold onto 100 nm of gold (. .).
C ,“lO ‘E
Fig. 5. Internal stress measured (a) during and(b) after the evaporation silver (-), copper (- - -) and gold (. .).
of 20 nm of silver onto 200 nm of
cases the MgF, film exhibits a tensile intrinsic stress which varies only slightly with the substrate metal film and agrees well with stress data reported for MgF, films (ref. 12 and references cited therein). The stress changes after the deposition are negligible in comparison with the large stress changes after the deposition of silver, gold and copper films discussed above. This confirms the practical experience that MgF, layers are rather immobile at room temperature while silver, gold and copper films at this temperature already recrystallize noticeably. 4. CONCLUSIONS In silver, gold and copper films deposited under ultrahigh vacuum conditions at room temperature onto MgF, substrate films a slight compressive stress develops during the initial growth stage. Then the film stress becomes tensile. When the films are completely continuous at 23 nm (silver), 16 nm (gold) and 14.5 nm (copper) the tensile stress reaches its maximum value. The internal stress of the continuous film is
compressive, with silver and copper showing comparable incremental stress which is lower than that of gold. In terms of our stress model this incremental compressive stress is due to a compressive strain built up at the metal-substrate interface. The size of this interface strain depends on the surface stress of the metal and the shape and adhesion of the metal clusters on the substrate. The incremental compressive stress of gold and copper can be correlated with the surface stress values reported in the literature’ ‘. The deviation of this incremental stress in the case of silver from that expected from the reported surface stress values is assumed to indicate that the shear stress developing at the metal-substrate interface exceeds the film adhesion and therefore, owing to slipping of the silver film along the interface, only part of the interface strain is maintained. After the metal deposition, part of the internal film stress built up during the film evaporation is again relaxed as a result of recrystallization and annealing processes that occur in the films; the compressive interface strain, however, is maintained. This recrystallization and the corresponding structural changes are most obvious in the case of gold and are less marked in copper and silver films. During the deposition of metal onto metal the interface strain determines the compressive stress measured during the early stages of the deposition until annealing processes give rise to a tensile stress contribution which reduces the initial incremental compressive stress. From the shape of the stress curve it is concluded that the growth of the second metal film is more or less a continuation of the growth of the first film without renewed nucleation on the surface of the metal substrate. While compressive stresses are invariably measured during the deposition of metal onto metal, only tensile stresses are found during the deposition of MgF, films onto the metal films. The stress change after the end of the MgF, deposition is negligible. confirming the practical experience that an MgF, film is a rather immobile layer at room temperature, in contrast with the comparatively mobile silver, copper and gold films. AC’KNOWLEDGMENT
We gratefully acknowledge Forderung der wissenschaftlichen
support for this work from the Fonds zur Forschung of Austria. Projects 4647 and 5718.
I R. Abermann. R. Kramer and J. MBser. Thin SolidFiln~s. 52 (1978) 215. 2 R. Abermann and R. Koch. Thin Solid Films, 62 (1979) 195. 3 R. Abermann and R. Koch. Thin Solid Films, 66 (I 980) 2 17. 4 R. Koch. H. Leonhard and R. Abermann. to be published. 5 R. Koch and R. Abermann, Thin SolidFilms. /2Y (1985) 63. 6 J. S. Vermaak and D. Kuhlmann-Wilsdorf. J. Phy.5. Chew.. 72 (1968) 4150. 7 H. J. Wassermann and J. S. Vermaak, Surf. Sci., 22 (1970) 164. 8 H. J. Wassermann and J. S. Vermaak, Sw/. Sci., 32 (1972) 168. 9 K. Kinosita, K. Maki, K. Nakamizo and K. Takeuchi, Jpn. J. A&. Ph_ys.. 6 (1967) 42. IO J. D. Wilcock, D. S. Campbell and J. C. Anderson, Thin Solid Films, 3 (1969) 13. II M. Laugier, ThinSolidFilms. 75(1981)213. 12 H. K.Pulker. ThinSolidFi/m,s.RY(l9R2) 191.