Thin Solid Films, 200 (1991) 263-274 METALLURGICAL
ADHESION, MICROSTRUCTURE AND COMPOSITION OF THERMALLY EVAPORATED ALUMINIUM THIN LAYERS POLYETHYLENE TEREPHTHALATE FILMS P. PHUKU, UnivrrsitP (Bcl~ium) (Received
P. BERTRAND Curholique
AND Y. DE PUYDT
July 4, 1990; accepted
I Plrrc~ Croi.v du Sud. B-1348
Industrial polyethylene terephthalate (PET) films have been metallized by aluminium evaporation in two different sets of experimental conditions. In the first set, aluminium layers of 100 nm thickness were deposited at a constant deposition rate (10 8, s-r) for different residual pressures varying from I Pa to 10m4 Pa and, in the second set, the residual pressure was kept constant (2.6 x 10e3 Pa), while the deposition rate was varied from 5 A s- ’ to 40 A s-r. The adherence between the aluminium layers and the PET film was measured by means of scratch and peel tests. The critical load and the peel strength exhibit a maximum at about IO-’ Pa when the deposition rate is kept constant. The microstructure of the aluminium layers, mainly the mean grain size, was studied by transmission electron microscopy (TEM), while secondary ion mass spectrometry (SIMS) depth profiles through the aluminium layers were performed in order to provide the chemical information, mainly aluminium layer oxidation. Concerning the TEM results, the grain size increases when the residual pressure is decreased and also when the deposition rate is increased. The SIMS depth profiles show different levels for aluminium oxidation at the surface, in the bulk of the layers and at the interface, all increasing for high residual gas pressure and for low deposition rates. From these results, it appears that the oxygen content at the AI-PET interface plays a critical role in the microstructure owing to its influence on the nucleation and on the growth of the aluminium layers. It also influences the adhesion between aluminium and PET for which an optimum oxygen amount seems to be required.
Thin vacuum-deposited aluminium films on polyethylene terephthalate (PET) substrates are intensively used in the semiconductor industry, packaging applications and magnetic tapes. The structure of thin metal films may strongly influence their adhesion to a polymer substrate and hence the durability of thin film devices. In this paper, we study the influence of the experimental preparation conditions ofaluminium layers on their adhesion to PET, and we try to relate it to the grain size 0040-6090/91/$3.50
tc Elsevier Sequoia/Printed
in The Netherlands
Y. DE PUYDT
and to the chemical composition of the aluminium coatings. The adherence between the aluminium layers and the PET film is measured by two different methods widely described in the literature: the scratch testtp4 and the peel test5,6. The adhesion of the aluminium layers is found to vary with the oxygen content at the AI-PET interface. The grain size of the deposited aluminium layers is studied by transmission electron microscopy (TEM) and the oxidation level of the aluminium layers is determined by secondary ion mass spectrometry (SIMS) analysis. 2.
Industrial PET films were metallized by aluminium evaporation in two different sets of experimental conditions. In the first set. for a constant deposition rate, the residual pressure was varied from I Pa to 10m4 Pa. In the second set, the residual pressure was kept constant and the deposition rate was varied from 5 As- ’ to 4OA s- ‘. The aluminium layer thickness of all samples was 100 nm and was determined with a quartz crystal monitor (XTC-INFICON; model 75 1-001-G 1, Leybold-Heraeus). Evaporations were performed using a tungsten filament covered by aluminium wires (99.99’:/, purity). Depositions were carried out without any surface pretreatment of the PET films. The adherence between the aluminium layers and the PET film was measured by scratch and peel tests. Concerning the scratch test, the adherence measuring device was a Heavens’ balance7 (Fig. l(a)). The sample was set on a mobile support whereas an indentation stylus was fixed at the extremity of one balance arm. On its middle, a plate was suspended on which the load was placed. A series of scratches were performed for different loads. The scratch speed was constant and equal to I cmmin~‘. As the holder was flexible, the sample was stuck on a glass plate. We used a Rockwell C stylus indentation (r = 200 pm). The critical load, i.e. the load for which the aluminium layer is removed from the polymer substrate, was determined by observing the scratch indentation profile with an optical microscope. When the critical load was reached, the scratch was characterized by many periodic folds on its side (Fig. l(b))‘. This procedure gave reproductible results within 10%. For the peel test, the PET side of the Al/PET samples was attached to a metallic
(b) I. (a) Schematic
Al ON PET
support with a double-sided tape (Permacel P-94). To grip the aluminium layers, an ethylene-acrylic acid (EAA) copolymer, which presents a strong adhesion to aluminium, was laminated on the aluminium side of the sample at 105 “C under a pressure of 7 x lo3 Pa for 15 s. After this treatment, the sample was cooled down for 2 h and then the EAA/Al was peeled from the PET substrate at 180” in a traction tensile tester (Instron 6021) at a peel rate of 5 cm min- ‘. The force (newtons per metre) required to separate the aluminium layers from the PET was recorded when peeling took place in the sealed area (Fig. 2).
I Al film
Fig. 2. Schematic diagram of (a) the 180“ peel test (1, metallic support; 2, double-sided substrate: 4, aluminium layer; 5, EAA film) and (b) the sample after peeling.
tape; 3, PET
TEM analyses were performed with a Philips EM301 electron microscope used both in transmission and in diffraction modes with 100 keV electrons. In order to analyse the aluminium layers, the PET substrates were dissolved in trifluoroacetic acid and the aluminium layers were floated on TEM copper grids. The aluminium grain sizes were determined by Heyn’s method’. SIMS analyses were realized in an ultrahigh vacuum chamber where a low pressure of IO-‘Pa was maintained. The SIMS spectrometer (Riber Q156) consists of a 45” deflection energy selector and a quadrupole mass filter. The detection of
Y. DE PUYDT
secondary ions was carried out after deflection at 90” with respect to the spectrometer axis. The working conditions for the SIMS analysis were the following: a Xe+ (4 keV) ion beam of 200 urn diameter was rastered on a I mm2 surface area with an average target current equal to 30 nA. The incidence angle with respect to the sample normal was 30”. An electronic gate was used in order to accept the SIMS signal only from the central part of the crater (9(x of the total signal). SIMS depth profiles were determined automatically by a programmable multichannel analyser (Lecroy MCA 3500). To obtain the concentration depth profiles, the SIMS intensities of six selected m/q peaks corresponding to C’, O+, Al+, AlO+, Naf and Al*+ ions were successively accumulated for 10 s and stored as a function of the bombardment time. Since the PET substrates were electrical insulators. low energy electrons (IO eV) produced from a heated thoriated tungsten filament were directed onto the samples during the ion bombardment. This avoided uncontrolled ion beam displacement due to charging when the AI-PET interface was reached. 3.
The results of adherence measurements foliowthe same trends for both the scratch test (critical load) and the peel test (peel strength). At a constant deposition rate, the adhesion of aluminium layers to PET film exhibits a maximum for a residual pressure of 10 2 Pa (Fig. 3). The TEM results show that the mean grain size of the aluminium layers increases with decreasing residual pressure and also with increasing deposition rate (Fig. 4). The SIMS profiles give information about the chemical composition and oxidation of the aluminium layers. From the SIMS profiles, we can distinguish three distinct regions over the thickness of the aluminium layers (Fig. 5): the surface region where a native oxide is present, the aluminium bulk and the AI-PET interface. We evaluate the total ion dose needed to reach the Al-PET interface from the dose for which the carbon signal reaches 50’;;) of its saturation value obtained in PET. Indeed. the main constituent of the PET chain is carbon. The interface width is estimated from 167; to 84% of the increasing carbon signal. The surface region is estimated from the initial decrease in the aluminium signal; its width is evaluated for a 50’:;, decrease. The bulk aluminium, characterized by a constant value at the low level. is included between the surface and the interface regions. In order to study the oxidation of aluminium layers, SIMS profiles were also performed for pure alumina Al,O, (taken as reference) under the same experimental conditions as those used for the Al/PET samples. As a result of matrix effects in SIMS, the intensity of the aluminium signal increases with the oxygen amount of the aluminium layers; nevertheless, the relative yield of Al:0 can be considered as a measure of the oxidation level in the aluminium layers. The Al:0 ratios are calculated from the integrated intensities of the three different regions of the SIMS profiles. The SIMS results on the alumina reference show that the aluminium and oxygen signals are constant through the profile, so that a constant Al:0 ratio of 230 is found for surface and bulk. This ratio is very low compared with that of the aluminium layers and could be taken as the highest oxidation state of aluminium.
Al ON PET
40 I 0
30 0 8
I . . , . ....I , . . . ..
The SIMS Al:0
as a function
the aluminium layers is obtained that of pure alumina. A lower Al:0
ratio will correspond The SIMS results on Al/PET samples show that the ratio at AI-PET interface increases decreasing residual pressure and with increasing deposition (Fig. 6). TEM SIMS, we conclude that the size increases when the ratio at AI-PET interface increases (Fig. 7). The SIMS results show also that the Xe’ dose needed to reach the the profile increases when mean Al:0 8). With respect to quality the aluminium layers, not reveal any So, if present, alumina and would probably the grain boundaries.
Deposition Fig. 4. Variation dcposltion
in the aluminium
Y. DE PUYDT
mean grain size as a function of (a) the residual pressure and (b) the
rate during the aluminium
The experimental results show that, with the increase in residual pressure, the aluminium grain size decreases (Fig. 4) and the oxidation of the aluminium layer increases (Fig. 6) whereas the adhesion of the aluminium layers on the PET substrate goes through a maximum between lo- ’ and lo- ’ Pa (Fig. 3). From TEM results. it was observed that a lower residual pressure had the same effect as a higher deposition rate, both leading to an increase in the aluminium grain size (Fig. 4). This may be explained by considering the relative fluxes of aluminium adatoms and oxygen originating from the residual atmosphere (mainly water vapour) impinging the substrate during the metallization. For a low residual
pressure, the oxygen flux is low compared with the aluminium flux, and pure aluminium growth is favoured, giving rise to larger aluminium grains. The same situation occurs for higher deposition rates. For higher residual pressures, the oxygen flux becomes more important and, since aluminium has a great affinity for
; 3 j 2 1I ; ,-~_-_-__________________r_____~ , , I I I t I 5 , I !
+ : 0:o
1 : Al/O
Y. DE PUYDT
3 IX 1017 Xef/cm2)
Fig. 5. Aluminium. carbon and oxygen SIMS depth prolilca: (a) detinition of the three regions m the alumimum layers and (b). (c) examples of profiles on aluminium layers obtained for two difcrent metallization conditions.
oxygen, aluminium oxide growth is now favoured, which is known to disturb pure aluminium growth. In the latter case. small grains are observed’,“. The residual pressure influences also the oxygen content of the aluminium layers as determined by SIMS analyses. The oxygen content at the Al-PET interface decreases when the residual pressure decreases or the deposition rate increases and conversely (Fig. 6). The Xe + ion dose needed to reach the AI-PET interface decreases when the mean Al:0 SIMS ratio increases, i.c~.the aluminium grain size increases (Figs. 7 and 8). Similar results were obtained on industrial samples of the same type1”12. It has been shown that large grains are less oxidized and that aluminium oxide is more difficult to sputter than metallic aluminium because there are ionic bonds (Al-O: 512. I kJ mol ’ (ref. 13)) which are stronger than metallic bonds (AI-AI: 186.2 kJ molt ’ (ref. 13)). By considering the dependences both of the aluminium adhesion (Fig. 3) and of the AI:0 SIMS ratio at the interface (Fig. 6) on the residual pressure and on the deposition rate, we conclude that the adhesion of aluminium layers on PET substrates has a maximum as a function of the oxygen amount detected at the interface. The aluminium layers with small aluminium grains are more oxidized and are obtained for high residual pressure andior low deposition rate. The basic requirements for good adhesion are a good contact between the wetting phase and the substrate, an absence of weak boundary layers, and an avoidance of stress concentrations which could lead to disbanding’“. Owing to the high affinity of aluminium for oxygen, we might think that the aluminium grains are formed around oxygen atoms which are the preferential sites for nucleation. Two sources of oxygen have to be taken into account. the oxygen from the PET surface chains and the oxygen from the residual atmosphere. Concerning the oxygen from
Deposition rate (A/see)
Fig. 6. Variation in the Al:0 SIMS ratio at the AIL PET interface and(b) the deposition rate.
as a function
of (a) the residual pressure
PET, the carboxylic groups O=C-0 which are the principal sites for aluminium and the hydroxylic groups -OH of the chain ends are the most absorption’5T16 reactive surface entities. The adhesion of aluminium layers should be high if Al-O-C bonds are formed directly with the PET surface oxygen. That is indeed the case for metallic coating adhesion on polyimide (PI), where nickel and chromium adhesion is six times higher than for copper and silver because nickel and chromium react directly with the oxygen of PI ” . In fact, electron spectroscopy for chemical analysis studies I6 have shown that only the first three aluminium monolayers react with the PET surface atoms. The reaction between aluminium layers and PET surface atoms can take place only if aluminium atoms directly reach
Y. DE PUYDT
Al/O interface Fig. I Variation in the aluminium grain size as a function layers as determined by the Al:0 SIMS ratio
Y G x
of the mean oxidation
level of the aluminium
g ._ z w z
mean AI/O Fig. 8. Variation in the total ion dose needed to reach the Al-PET oxidation level as determined by the AI:0 SIMS ratio.
as a function
of the mean
the PET surface. This involves the requirement that evaporation should be performed in the best vacuum conditions in order to reduce the amount of oxygen coming from the residual atmosphere and to favour the reaction of aluminium with the PET surface. However, according to our results, if AL-O-C bond formation at the interface is necessary, we think that a good adhesion between aluminium and PET requires the formation of an intermediate interphase of several monolayers in which the aluminium is oxidized. This ensures the transition between PET and bulk aluminium. The reaction of oxygen from the residual atmosphere with the
Al ON PET
aluminium adatoms and with the PET surface chains via the benzenic rings or the -CH,groups is known to produce an interphase between the polymer and the aluminium film”~‘8~‘9. A critical amount of the oxygen from the residual atmosphere is then needed for the formation of this interphase. Indeed, a too small amount of oxygen could induce a depletion of oxygen near the polymer surface, leading to the formation of a weak boundary layer and hence to a decrease in the adhesion’ ‘. Otherwise, too much oxygen from the residual atmosphere will result in the formation of small aluminium grains which produces a great concentration of stress between the polymer and the aluminium layers and will lead to disbonding. About the stress, it has been shown by Kubovy and Jandra” that the stress of evaporated aluminium films is very sensitive to film contamination by oxygen from the residual atmosphere; a compressive stress occurs for a low contamination, whereas a high contamination leads to a tensile stress. According to this assumption, the adhesion of aluminium layers to PET will decrease when the residual pressure increases and when the deposition rate decreases. 5.
We found that the metallization conditions of aluminium deposited on PET influence the metal layers’ microstructure and composition as well as the adhesion to the polymer substrate. Indeed, both the residual atmosphere in the evaporation system and the deposition rate influence the amount of oxygen at the Al-PET interface as well as the aluminium mean grain size. The best adhesion is obtained for aluminium layers evaporated under a residual pressure of 10 ~’ Pa. At this pressure, an optimum oxygen amount is found at the Al-PET interface. ACKNOWLEDGMENTS
We would like to thank Dr. N. Terao for the TEM analyses, C. Poleunis for SIMS analysis and R. Bertrand (Du Pont de Nemours, Luxembourg SA) for the peel test measurements. This work was supported by the AGCD and BRITE organizations.
REFERENCES I 2 3 4 5 6 7 8 9 10 1I 12
L. F. Goldstein and T. J. Bertone, J. Vat. Sci. Techno/.. I2 (1975) 1423. A. Yu. Loginova, S. B. Aimbinder and A. M. Grinshten, (transl. from M&h. P&n.. 4 (1974) 641). P. A. Steinmann, Y. Tardy and H. E. Hintermann. Thin Solid Films. 154 (1987) 333. P. Laeng, P. A. Steinmann and H. E. Hintermann, Oberf?&che-Su$, 23 (4) (1982) 108. A. N. Gent,J. Adhes.,23(1987) 115;24((1987) 173. K. Bright, B. W. Malpass and D. E. Packham, Br. Po/ym. J., 3 (1971) 205. 0. S. Heavens, J. Phys. Radium, II (1950) 355. R. T. De Hoff and F. N. Rhines, Quantitative Microscopy. McGraw-Hill, New York, 1968, p.239. J. Weiss and C. Leppin, Thin Solid Films, 174 (1989) 155. M. J. Verkerk and W. A. Brankaert, Thin So/id Films, 139 (1986) 77. Y. De Puydt, P. Bertrand and P. Lutgen. Su$ Inre[face Anal.. 12 (1988) 486. Y. De Puydt and P. Bertrand, Nucl. Instrum.Methods Phys. Rex B. 39 (1989) 86.
Y. DE PUYDT
R. C. Weast and M. J. Astle (eds.). Hmdhooh of Chmisrr~~ cud Ph~tks. CRC Press, Boca Raton. FL, hlst edn.. 1980. Table F222. 14 D. M. Brewis and D. Brigs, I~I&SIY~LI/ At//w.\ro,z Proh/er?z.c, Orbital Press. Oxford, 1985. p. 4. 15 J. J. Pireaux, M. Vermeersch, N. Degosserie. C. Gri-goire. Y. Novis, M. Chtaiband R. Caudano, in M. Gruwe and H. 2. Kreuzer (eds.). Ad/wvio/r trnclfricrion. Springer Series on Surface Science, Vol. 17, Springer, Heidelberg, 1989. 16 Y. Jugnet. J. L. Droulas and T. M. Due. ACS Sjxlp. Sw. 440 (1990) 467. 17 N. J. Chou and C. H. Tang. J. I;w. SC,;. Twhrrol. d. Z ( 19X4) 75 I. 18 J. W. Barta. P. 0. Hahn, F. Legoues and P. S. Ho. J. vw. SC,;. Twhwl. A. 3 (19X5) I300 19 J. M. Burkstand, J. Appl. P/IJ,.\.. 5,7(19X1) 4795. 20 A. Kubovy and M. Jandra, ~I+I .SoliclFilr~~\. 4-7 (1977) 169. 13