FMR study of surface‐tension‐related stress effects in
ultraclean Ni thin films
Citation for published version (APA):
Janssen, M. M. P. (1970). FMR study of surface‐tension‐related stress effects in ultraclean Ni thin films. Journal of Applied Physics, 41(1), 384-398. https://doi.org/10.1063/1.1658352
DOI:
10.1063/1.1658352
Document status and date: Published: 01/01/1970
Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page numbers.
Link to publication
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:
www.tue.nl/taverne Take down policy
If you believe that this document breaches copyright please contact us at: openaccess@tue.nl
FMR Study of Surface-Tension-Related Stress Effects in Ultraclean Ni Thin Films*
M. M. P. JANSSENt
Center for Surface and Coatings Research, Lehigh University, Bethlehem, Pennsylvania 18015 (Received 22 May 1969)
Ni thin films, ranging in thickness from 19 to 837 A, were deposited under ultrahigh vacuum conditions (1~ to 10-10 Torr) onto extremely well degassed substrates. Intrinsic isotropic stress values in the films
were determined by ferromagnetic resonance; these measurements were carried out while the films were in the vacuum system in which they had been prepared. Films deposited onto substrates at 25°-35°C were in a state of compressive stress when measured in UHV. The surface tension provides an explanation for these stresses; the stress levels are shown to obey a simple model. Mter adsorption of 0., N.O, or air, the stress levels in the films dropped to very low values, while adsorption of H., H20, CO, and pyridine resulted
in slightly smaller effects. Admission of N. to the films caused no stress release. A small degree of reversibility of the latter effect was found for H. and H20. Adsorption of gases had a marked effect on the ferromagnetic
resonance linewidths. Films thinner than 100 A showed deviations from the surface tension model and the possible discontinuity of such films is indicated as the reason. Experiments on annealed films and experi-ments with films deposited onto substrates at elevated temperatures were performed. The influence of vacuum conditions other than UHV during film deposition was investigated. The relevance of this study to the solution of the problems of stress corrosion cracking and hydrogen embrittlement is pointed out.
I. INTRODUCTION
The objective of this work is to determine stresses and stress changes in Ni thin films as a result of adsorp-tion of gases. A preliminary report has been given previously.l The stresses meant here are the planar compressive stresses in the bulk of the film. The postulation is made that the isotropic surface stress2•3
of the polycrystalline films is the stress source and is"identical with the surface tension (1') of the films,
b~sed on the encountered strong dependence of the
observed stress on adsorption of gases and on the quantitative agreement of the measured stress values with those of the adopted surface tension model. Thus this work can be interpreted as a study of the influence of'adsorbed gases on the surface tension of clean nickel surfaces- at room temperature.
From - a simple qualitative consideration it is clear that the mechanical effect of surface tension on meas-urable bulk properties such as lattice constant and bulk stress will be observable for very fine particles and very thin films only. Because ultraclean small particles are difficult to prepare and to define, bulk stresses in thin films were chosen for study.
* This research was supported by the Advanced Research Projects Agency, Department of Defense, through the Office of Naval Research under Contract No. Nonr. 610(09).
t Present address: Laboratory for Physical Chemistry, Uni-versity of Technology, Eindhoven, Netherlands.
1 M. M. P. Janssen, J. Appl. Phys. 40, 3055 (1969).
Surface tension effects are believed to playa vital role in the delayed failure processes of metals which are known as hydrogen embrittlement and stress cor-rosion cracking. A key to the understanding of the mechanism of the fai:ure processes is provided by the stress sorption theory, proposed by Petch and Stables.4
This theory applies the Griffith fracture mechanism of glasses to metals. The Griffith relationship shows that the fracture stress is proportional to the square root of the surface free energy5 of the newly produced surface. It is obvious that when the new surface can be covered by adsorbed gas atoms as it is formed, the fracture stress will be lower than when a clean surface is produced. From basic thermodynamics it follows that the surface free energy (and surface . tension) will be lowered by gas adsorption. Petch and Stables pro-posed this theory to explain hydrogen embrittlement in steels. Coleman et al.6 proposed the surface free
energy mechanism for stress corrosion cracking of metals and alloys by all stress corrosion media in general. The free energy of surface production, as con-sidered for metals and alloys, consists of two parts. One part involves irreversible deformation of the material, while the other can be related to a reversible process of surface creation and is identical with the surface tension. Quantitative data on the effect of gas adsorption on the surface tension, needed to support the stress sorption theory, are lacking.
It is believed that this study presents the first data 2 For a discussion of the distinction between surface stress
(elastic stretching) and surface tension, see J. W. Gibbs, The
Collected Work of J. W. Gibbs (Longmans, Green and Co., Inc., 4N. J. Petch and P. Stables, Nature 169, 842 (1952).
New York, 1928), p. 315, footnote. 5 Surface free energy means here free energy of surface
produc-3 A discussion of the mechanical aspects of surface tension can tion; for liquids this is identical to the surface tension on basis
be found in R. Defay, 1. Prigogine, A. Bellemans, and D. H. of therelation-y=(oF/oA) (T,V, reversible). .
Everett, Surface Tension and Adsorption (John Wiley & Sons, 6 E. G. Coleman, D. Weinstein, and W. Rostoker, Acta Met.
Inc., New York, 1966), Chap. 17. 9,491 (1961).
STRESS EFFECTS IN Ni FILMS 385 on the influence of gas adsorption on the surface
ten-sion of metals at room temperature. At 932°C, by the zero creep method, Buttner et aU found a marked decrease in the surface tension of silver wires due to oxygen adsorption. Rhead and Mykura8 obtained
quali-tatively the same result from thermal etching experi-ments at 900°C. Similar results were obtained for the Fe/O system by Moreau and Benard9 and for the
Cu/O system by Robertson and Shewmon.1o
Internal stresses in thin films have been excellently reviewed by Hoffman'!! A few studies in which com-pressive stresses were encountered will briefly be men-tioned here. Weiss and Smith12 found compressive
intrinsic stresses in permalloy (85 Ni, 15 Fe) films on glass, deposited at elevated substrate temperatures. (Intrinsic stresses are those that do not relate to thermal expansion effects.) At substrate temperatures near room temperature, the stress was reported to be tensile. Klokholm and Freedman13 found similar results
for Ni films; in this case the stresses were measured at room temperature, but the residual intrinsic stresses were reported. In both investigations use was made of the bending plate method for measuring stresses and the vacuum conditions under which the films were prepared and the stresses measured were relatively poor (lo-L 10-7 Torr). In their speculation on the mechanism for the observed compressive stress, Weiss and Smith postulated a relationship to the surface ten-sion. This interpretation is doubtful, since compressive stress generated by surface tension appears to be tensile when measured by the bending plate method. Some results and a brief discussion on compressive intrinsic stresses and stress release on adsorption of gases in ultraclean Ni films, related to surface effects, were reported recently.l Freedman!4 and Pomerantz et al.!6 reported compressive stresses in single crystal Ni films deposited onto NaCI substrates at elevated tempera-tures and analyzed at room temperature. These stresses were easily explained as resulting from thermal effects. The stresses were relieved when water was admitted to the film, causing the film to loosen from the substrate. Tensile stresses have been invariably found by various authors in investigations of Ni films deposited on substrates at or near room temperature and 'pre-pared in vacuum conditions ranging from poor to
7 F. H. Buttner, E. R. Funk, and H. Udin, J. Phys. Chern. 56,
657 (1952).
8 G. E. Rhead and H. Mykura, Acta Met. 10, 843 (1962).
9 J. Moreau and J. Benard, Acta Met. 10, 247 (1962).
10 W. M. Robertson and P. G. Shewmon, J. Chern. Phys. 39,
2330 (1963).
11 R. W. Hoffman, in-Physics of Thin Films (Academic Press
Inc., New York, 1966), Vol. 3; in the discussion of Eq. (3) the term surface free energy is used.
12 G. P. Weiss and D. O. Smith, J. Appl. Phys. 33,1166 (1962). 13 E. Klokholm and J. F. Freedman, J. Appl. Phys. 38, 1354
(1967) .
14 J. F. Freedman, J. Appl. Phys. 33, 1148 (1962).
16 M.'Pomerantz, J. F. Freedman, and J. C. Suits, J. Appl.
Phys. 33,1164 (1962).
excellent.J6-2B To explain these stresses many
qualita-tive models have been proposed.l l The results on
com-pressive stresses in Ni films reported in the present study can be explained by a simple quantitative model involving surface tension.
II. PRINCIPLES
Techniques of measuring stress in Ni thin films by the ferromagnetic resonance (FMR) method are well established. In this investigation the films were oriented with the film plane perpendicular to the static magnetic field which yielded an increase in the sensitivity of the stress measurements by a factor of 10 compared to orientation of the film plane parallel to the static field. With a perpendicular arrangement the resonance con-dition for uniform precession is as follows24:
wh=HRJ.-47rM.- (2K/M,). (1)
In Eq. (1), w is the angular frequency of the microwave field (w=27rv, v=9.3X109 sec! in this study) and'Y is
the magnetomechanical ratio, equal to 2.807rgNiX 108
sec! G-!. The value 2.18 was assigned to the spectro-scopic splitting factor gNi. HRJ. is the applied static magnetic field in gauss and M. is the saturation mag-netization of the film, taken equal to the value for bulk Ni, i.e. 490 emu/cm.s NeugebauerJ8 showed that
Ni films evaporated and measured under ultrahigh vacuum (UHV) conditions, as used in this study, had M. values not differing from the bulk value. Further-more, it may be concluded from Neugebauer's17 experi-ments that exposure of the film to the atmosphere does not change the M. value. The demagnetizing factor value 47r can only be applied when the film is
con-tinuous, and the experiments of Neugebauer give reason to expect that films thicker than 30
A
are indeed continuous.K (erg/ cmS) is the anisotropy constant for
magnetiza-tion perpendicular to the film plane (in K the shape anisotropy contribution is not included; the shape anisotropy constant 27rM,2 is treated as the demag-netizing field). Uniaxial anisotropy was neglected here
16 E. C. Crittenden and R. W. Hoffman,' Rev. Mod. Phys. 25,
310 (1953).
17 C. A. Neugebauer, General Electric Rep., Contr. No. AF-19
(604)-5566, Rep. No.2 (1961); Trans. 8th Vac. Symp., 1961, 924 (1962).
18 C. A. Neugebauer, Structure and Properties of Thin Films
(John' Wiley & Sons, Inc., New York, 1959), p. 358; Phys. Rev. 116, 1441 (1959).
19 M. Kuriyama, H. Yamanouchi, and S. Hosoya, J. Phys. Soc.
Japan 16, 701 (1961).
20 S. Fujiwara, T. Koikeda, and S. Chikazumi, J. Phys. Soc.
Japan 20,878 (1964).
21 H. Fujiwara, Y. Sugita, and N. Saito, J. Phys. Soc. Japan
20, 2088 (1965).
22 J. F. Freedman, ]. Appl. Phys. 36,964 (1965).
23 S. Usami, H. Nagashima, and H. Aoi, J. Phys. Soc. Japan
22,877 (1967).
which is valid when films are deposited in the absence of a magnetic field2S and at normal incidence.26 Both requirements were fulfilled in this case. In view of the strong magnetoelastic coupling in Ni and the relatively high stress levels involved, K could be considered to depend on the isotropic stress only by the relationship K = 3 SA/2, where S is the isotropic stress in dyn/ cm2 and A is the magnetostriction constant. For A the value of -37X 10-6 for bulk polycrystalline Ni was taken,
which value is very close to that found by magnetic resonance experiments on poly crystalline Ni films by Matsumoto et al.27 Evidently films deposited at room
temperature substrates, as investigated in this study, are polycrystalline. For a stress-free film, the calcu-lated resonance field under the conditions mentioned above is HRl.O= (wh) +4?rM.= 9200 G. Pomerantz
et al.Is measured a value of 9200 G for HRl.O on a
reportedly stress-free film (v= 9.023X 109 secl ) . The isotropic stress can then be expressed in terms of the resonance fields as
or
S= -4.4(HRl.-9200) X 106 dyn/cm2. (2)
Stress differences between consecutive measurements on the same film as small as 0.1 X 109 dyn/ cm2 can be detected easily. Absolute stress determinations were found to be reliable to ±0.3Xl09 dyn/cm2, the error mainly being due to inaccuracies in alignment of the film in the static magnetic field. In experiments in which standing spin wave modes were observed, the uniform precession mode was selected for stress de-termination. This method of stress measurement can be applied to films as thin as 20
A;
it is, however, limited to ferromagnetic materials with a large value of A. When HRl. is found to be greater than 9200 G,S is negative (compressive stress); when HRl. is ob-served to be smaller than 9200 G, S is positive (tensile stress).
The apparent stress caused by surface tension in a thin film can be expressed byu
(3) where 1'1 and 1'2 are the surface tension of the metal at the film/vacuum (or gas) and film/substrate inter-face, respectively, and t is the film thickness. As a result of the interaction between the substrate and the film 1'2 will be smaller than 1'1. The interaction between well' degassed substrates, as used in this work, and metal films is thought to be weak and absolute values are hard to estimate; therefore, 1'2 was assumed as a 25 C. D. Graham and J. M. Lommel, J. Phys. Soc. Japan 17,
570 (1962).
26 J. D. Finegan and R. W. Hoffman, J. Appl. Phys. 30, 597
(1959). .
27 G. Matsumoto, M. Kato, and A. Tasakl, J. Phys. Soc. Japan
21,882 (1966).
first approximation to be equal to 1'1. A combination of Eqs. (2) and (3) was used to calculate 1'1 and 1'2, or preferably 1'1+1'2, from the observed resonance fields and film thicknesses. A value of 2500 dyn/ em for the surface tension of clean polycrystalline Ni (I'Ni) at room temperature was adopted after consideration of the work of Kozakevitch and Urbain2s on the surface tension of liquid Ni in argon at 1550°C, and the work of Hayward and Greenough29 on the determination of I'Ni from zero creep measurements close to the melting point in argon atmosphere. Extrapolation of their data to room temperature involves many uncertainties, but the value of 0.6 dyn. em-I. °C-I for the temperature coefficient of I'Ni seems not unreasonable. Thus the value of 2500±300 dyn/ em was obtained. This value represents an average of I'Ni over all crystallographic planes, and is somewhat low in comparison to cal-culated values of I'Ni at room temperature. Combining Eqs. (2) and (3) and substituting I'1=I'2=I'Ni=2500 dyn/cm, yields the expression for the predicted reson-ance field as a function of the film thickness for a perfectly clean film:
(4) where t is in Angstrom units. In Eq. (4) the term (1. 13/t) X lOS represents the stress contribution to the resonance field; the term 9200 is stress independent. This treatment is relatively independent of the roughness factors of the films. The magnitude of the surface stress in the plane of the film is not influenced by protrusions; the stress component normal to the film plane will have negligible effects. Roughness means local variations in the film thickness, which gives rise to FMR line broadening. Finally, roughness might influence the shape anisotropy to a slight degree.
III. EXPERIMENTAL PROCEDURE The apparatus used in this study (see Fig. 1) was similar to the equipment used by Neugebauerl7
•1S
for magnetic torque measurements on Ni thin films prepared in a vacuum below 10-9 Torr. Lykken et al.30
and Usami et al.23 described systems for FMR studies of thin films in UHV, although in both systems the vacuum during film deposition was relatively poor. The vacuum system, built from stainless steel and glass (mostly 2 in. o.d. tubing), was mounted on a heat-resistant board. The board was fixed to a metal frame that could be rolled from the bakeout furnace to the microwave spectrometer (Varian V 4502). Roughing was done with a Varian Vacsorb pump. All
28 P. Kozakevitch and G. Urbain, J. Iron Steel lust. 186, 167
(1957).
29 E. R. Hayward and A. P. Greenough, J. lnst. Metals 88, 217
(1959--1960) .
30 G. 1. Lykken, W. L. Harman, and E. N. Mitchell, J. Appl.
STRESS EFFECTS IN Ni FILMS 387 parts of the system were baked at 450°C for at least
16 h, with a Varian 8 liters/sec Vacion pump in opera-tion, to attain UHV. The ionization gauge and filaments were degassed during cooling. The vacuum finally reached was better than 1.5 X 10-10 Torr. System prep-aration took at least one and a half days. The sample tube, fitting into the spectrometer dual sample cavity, was made from Vycor brand glass (11 mm o.d. tube), which has a far greater microwave transmission than Pyrex. The readings from a GE miniature ionization gauge, placed approximately 25 cm from the center of the magnet pole shoes, appeared to have lowered under the influence of the magnetic field, so pressure readings were made with the magnetic field switched off. For measurements of high pressures, a bakeable Pirani gauge was attached to the system.
Substrates used were soft glass (Corning 0211, alkali borosilicate, thickness 0.3 mm), Vycor glass (Corning 7900,96% silica, thickness 0.5 mm) and {l00l-oriented N aCI single crystals (thickness 1 mm). The length of the substrates was 60-100 mm. The width, limited by the size of the sample tube, was 8 mm. Cleaning of the glass substrates was done ultrasonically in detergent solution, water, concentrated acid solution, and distilled water successively. NaCI substrates were freshly cleaved and immediately sealed into the vacuum system. The substrate could be heated by a thoroughly degassed tungsten coil for additional degassing, and for heat treatments of the prepared films. Substrate temperatures were measured by a fine wire thermo-couple (Pt/Pt-13% Rh), the junction of which was sealed in glass and rested against the back of the substrate. The substrates had a small hole on one side so they could be attached to the sample holder which consisted of a glass rod with a piece of iron sealed in glass on top. After evaporation, the film was lowered into the sample tube by magnets operated outside the system, and aligned perpendicularly to the static magnetic field. All manipulations could be done while the vacuum system was attached to the spectrometer and while the pump was in operation. The films produced had an area of 50X8 mm. When in the measuring position, the films protruded from the cavity (height 40 mm) at both ends. In this arrangement the maximum possible area of the film was subjected to the microwave field, yielding maximum sensitivity. Further, this geometry assured good homogeneity of the microwave field inside the cavity. A magnetically operated shutter was used to shield the substrate from the Ni source when the latter was degassed. Two kinds of Ni sources, mounted 80--100 mm from the substrate and parallel to it, were used: (a) Ni wire, zone refined 99.995%, heated close to the melting point. This source yielded low evaporation rates (3-12 A/min), film thickness up to 200 A and vacuum better than 5X 10--10
Torr during evaporation. (b) W wire (99.999%) provided with a series of loops to contain Ni. This
o
k
c
d
FIG. 1. Schematic representation of the apparatus: (a) micro-wave cavity; (b) micromicro-wave bridge; (c) 12 in. magnet; (d) 8 liters/sec Vacion pump; (e), (f), and (g) bakeable valves; (h) gas reservoir; (i) ionization gauge; (j) stainless steel bellows; (k) arrangement for moving film into cavity and for alignment of the film in the field; (1) shutter; (m) substrate; (n) nickel source; (0) heating element; (p) thermocouple; (q) Pyrex-Vycor graded seal; (r) sample tube; (s) metal; (t) Pyrex glass.
wire was first degassed at a high temperature in UHV until the vacuum was better than IX 10--9 Torr. Then,
the system was opened and pieces of the Ni wire mentioned under (a) were carefully attached to the loops. When UHV was attained again, evaporation took place after degassing of the Ni. During the latter degassing procedure the nickel melted and covered the tungsten wire. The vacuum during this evaporation was in the 10-9 Torr range owing to the larger heat
production of the source. Evaporation rates were con-siderably faster (40--100 A/min) and film thicknesses up to 1000 A were produced. During evaporation the substrate was heated by radiation from the Ni source to no more than 35°C, eliminating the need for thermal expansion corrections in the stress levels. During evaporation there was practically no rise in pressure; the shutter was opened when the expected lowest pressure was reached. After evaporation the pressure in the system dropped to below 2X10--JO Torr. The
film was brought into the cavity as soon as it was prepared and FMR spectra were taken. In most cases the sweep time was 10 min for a sweep range of 7-12 kG. Gas admission to the system proceeded via a fine
-;a
'"
:J 0 01 .~ x. -a! 11 I 0 -' w ii: w 10 U z ~ &l w Ct: 9 HRl =9200+ 1~3 x10 5 gauss... t
0 _ _ _ _ 0 0 0 ---Fi'=;r~--=---:-~.-~:I!I~~-~--' ';...t
i
HR10 =9200 gauss 8 7 200 400 600 800---i> FILM THICKNESS (Angstrom) FIG. 2. Observed resonance fields in films deposited at high rates, compared to values according to theory. Soft glass sub-strates; substrate temperature during deposition maximum 35°C. 0, values measured in URV (HRJ.l); " final values after oxygen adsorption (HRJ.a); 8, in pure H2 (100 Torr) j 8., in pure H 20 (10 Torr) j (A) and (B) observations on films deposited at low rates (see Fig. 3 for more detail).
control valve. Gases were stored in break-seal flasks. H2 , N2, O2 (impurities less than 2 ppm) and CO (60
ppm active impurities) were dried via a liquid nitrogen trap. H20, N20, and pyridine were cleaned by repeated
freezing and pumping before the vapors were stored in break seal glass flasks. When air was admitted, H20 was removed with a liquid nitrogen trap. During gas admission the pump was operating but isolated from the system by a high conductance valve. For reversibility studies the gas was pumped off by opening the latter valve. After the experiment, the film was taken out of the system, exposed to air, and measured again after several days.
Film thicknesses were measured by x-ray fluorescence. Standards were calibrated by interferometry and careful weighing of the amount of deposited Ni. The accuracy in film thickness determination was estimated to be 3-5%. The uniformity of the film thickness was checked with an electron microprobe, showing a varia-tion of less than 5% over a 40-mm length of a typical film.
IV. RESULTS AND DISCUSSION
A. Resonance Fields Measured in UHV and after Adsorption of Oxygen
The results of the experiments in which high evapora-tion rates (40-100 A/min) and soft glass substrates were used are summarized in Table 1. The observed resonance fields in UHV and after oxygen adsorption, HSJ.l and HsJ..a, respectively, are plotted in Fig. 2.
~~
S
~~$S~~~:gSl~$~$t!t::~ ~~Me"")M('oI')~~C"":)C'f";IC'I")C"")C'f'")C'ljt'f')('f') "';000000000000000++
""""""""""""+1-+1,,
~~VVVVVVVVVVVVVV <"') o"
V <"') o"
V N If)...-t....-llf) CXJ 0\....-1 OON ~ ~ 0\ oo~ 0'\ u)~~~c--i~o"";o"";"";"";oooo I I I I I I I I I I I I I I I I ..-...-.l.I)If) lI)l.I')lJ"')1.I')U)It')l..I)lI)l.I)U')1.I')lJ"') OON....-t '1""""1....-1....-1....-1"""'....-1'1""""1....-1....-1,..."""'....-1 '1""""1 0000000000000000I I""""""""""""""
~~VVVVVVVVVVVVVV o"
V 0000"++,,
V V o"
V 0\lf)<",)>---1' Olf) \0--1'>-00 00lf)--1'--1'lf)NNN"";"";"";oooooooooo
++++++++++++++++
19 24 37 38 40 41 43 50 54 56 57 69 70 71 79 80 101 103 120 121 132 153 154 160 182 • Pyridine.
TABLE II. Results of experiments with low evaporation rates (see Table I, Footnote a).
R P PIR HRJ.I 8 4XIo-IO 3X1O-9 10 310 4 id 6X 10--9 10 920 4 id id 10 350 4 id id 11 030 4 id id 10 340 4 id id 10 440 10 5XIO-IO 3X1Q--9 10 200 3 4XIo-IO 6XIQ--9 10 140 3 id id 9 970 3 id id 10 800 4 id 4 id id id 10 070 10990 9 id 3 X lQ--9 10 010 4 id 6X lQ-9 10 850 12 id 2XIQ-9 10 940 5 id 6XIQ--9 10 340 6 id id 10 790 5 id id 11 020 8 id 3XI0--9 10430 6 id 6X lQ-9 10 580 8 id 3XIQ-9 10 780 12 id 2XIQ-9 10 200 8 id 3 X lQ--9 10 220 8 id 12 id 3XIQ-9 10 460 2XIQ--9 10 000 HR J.2 HRl.2 KI K2 Ks SI 6510 8 360 (H20) 6690 7280 8 470 (H2) 8370 8 110 (H20) 7860 8 280 (H20) 8060 8 180 (CO) 7830 10 140 (N2) 7900 8160 8190 10 07l (N2) 8100 8 590 (02) 8590 8 140 (N20) 8140 8 490 (02) 8490 8 610 (02) 8610 9 025 (CO) 8810 9170 9220 9 380 (HoO) 9060 9 180 (0.) 9180 9260 9 170 (N20) 9170 9250 +2.7 +4.2 +2.8 +4.5 +2.8 +3.0 +2.4 +2.3 +1.9 +3.9 +2.1 +4.4 +2.0 +4.0 +4.3 +2.8 +3.9 +4.5 +3.0 (-2.1) ( -1.8) (-2.7) ( -2.3) ( -2.5) +2.3 +2.1 ( -1.5) ( -2.6) (-1. 7) ( -1.4) ( -0.4) +0.4 ( -6.6) ( -6.1) (-4.7) (-2.0) ( -3.3) ( -2.8) ( -3.4) (-3.2) (-2.5) (-2.5) (-2.7) ( -1.5) (-2.6) (-1. 7) ( -1.4) (-0.9) <±0.15 <±0.15 -0.3 +3.4 <±0.15 <±0.15 +3.9 <±0.15 -4.8 -7.6 -5.0 -8.1 -5.0 -5.4 -4.3 -4.1 -3.4 -7.0 -3.8 -7.9 -3.6 -7.2 -7.7 -5.0 -7.0 -8.1 -5.4 -6.1 -7.0 +2.4 <±0.15 <±0.15 -4.3 +2.5 <±0.15 -4.5 9380 (H2) 9260a 9240 +3.1 +0.4 <±0.15 -5.6 9130 +2.0 <±0.15 -0.2 -3.6 S2 Ss t:.HI t:.H2 t:.H. II 12 I. (+3.8) (+3.2) (+4.9) (+4.1) (+4.5) -4.1 -3.8 (+2.7) (+4.7) (+3.1) ( +2.4) (+0.7) -0.7 (+11.9) 860 (+11.0) 480 (+8.5) 470 (+3.6) 490 (+5.9) 310 (+5.0) 350 (+6.1) 700 (+5.8) 280 (+4.5) 370 (+4.5) 330 (+4.9) 610 (+2.7) 160 (+4.7) 350 (+3.1) 200 (+2.5) 280 (+1.6) 280 <±0.3 150 <±0.3 200 +0.5 360 <±0.3 <±0.3 310 <±0.3 180 <±0.3 <±0.3 160 <±0.3 160 -0.7 <±0.3 <±0.3 r +0.4 280 300 300 300 240 220 360 220 610 160 350 200 240 240 280 220 180 160 160 1020 32 11 420 36 75 22 550 61 37 470 60 117 55 200 41 124 103 220 54 86 75 400 26 49 38 220 27 41 53 270 57 70 300 65 56 390 46 46 85 160 99 120 120 350 45 34 34 200 108 66 66 240 40 70 56 240 42 56 53 150 87 88 200 64 87 280 60 95 49 220 59 78 78 180 65 61 180 67 117 117 160 72 115 160 80 189 140 160 63 166 120 en ~ );d t<1 en en t<1 >"1j >"1j t<1 C") ~ en
....
Z Z >"1j....
t"' is: en c...:. 00 \0- - FILM THICKNESS (Angstrom)
FIG. 3. Obs6rved resonance fields in films deposited at low rates. Soft glass and Vycor glass substrates; substrate temperature during deposition 35°C. Upper band (A):
i",
v~lues measured in UHV (BBl.I); 0, pure N2 (100 Torr); 1, predIcted from theory CEq. (4)]. Lower band (B): after adsorption of reactive gases; 2, zero stress value B Bl.o=9200 G; El, pure H2 (100 Torr); 8,pure H20 (10 Torr); W, pure CO (100 Torr); 0, pure pyridine
(10 Torr); " final values after oxygen adsorption (B Bl.3), ad-mitted gases: pure 02 (100 Torr), N20 (100 Torr) or air (1 atm).
The resonance fields are shown in all the figures; stress and K values obtained with help of the equations given in Sec. II are listed in the tables. The areas (A)
and (B) in Fig. 2 relate to the upper and lower bands of Fig. 3. The data in Fig. 3 and in Table II are from an extensive study on films of 20-200 A thickness, deposited at low rates (4-12 A/min). The latter study was carried out because the shift in the resonance field on adsorption is the largest in this thickness range, and because a deviation from predicted behavior was found with films thinner than 100 A and was thought to be worth investigating. In the slow deposition study
TABLE III. Calculated vs measured total resonance field shifts on adsorption of oxygen; high evaporation rates.
Resonance field shift on
Film adsorption (gauss)
thickness in A Measured Calculated" 74 1530 1530 113 1160 1000 118 990 960 156 730 720 204 540 550 267 340 420 402 250 280 445 260 250 462 180 240 477 270 240 524 250 220 550 250 200 574 140 200 657 150 170 808 100 140 837 160 130
TABLE IV. Calculated vs measured total resonance field shifts on adsorption of oxygen; low evaporation rates.
Film thickness in A 19 24 37 38 40 41 43 50 54 56 57 69 70 71 79 80 101 103 120 121 132 153 154 160 182
Resonance field shift on adsorption (gauss) Measured Calculated" 3800 4230 3070 2660 2480 2380 2370 2240 1810 2610 1970 2400 1870 2360 2330 1530 1620 1800 1370 1400 1520 1030 970 1220 870 5950 4710 3050 2970 2830 2760 2630 2260 2090 2020 1980 1640 1610 1590 1430 1410 1120 1100 940 930 860 740 730 710 620
a From Eq. (4). assuming complete stress release on adsorption of oxygen.
soft glass and Vycor glass substrates were used without discrimination; no difference in the behavior of films deposited onto these glasses was found. The experi-mental results on films evaporated with high and low rates, respectively, overlap in the thickness range
74-182 A (see Fig. 2).
~ 200 400 600 BOO
§,
~ FILM THICKNESS (AAgstrom)J;1 6(;15 I ~ .... LL :1'4 til
i
3 \. .. :.i
2 \ : d •\; . 1.i3 x 105 gauss (Eq.4)
\'. I
.'-.~~.~ • • ' - - 0 -0-
~---FIG. 4. Measured total shifts in resonance field on adsorption of oxygen. Solid line: calculated shifts assuming complete stress
STRESS EFFECTS IN Ni FILMS 391 The overlap region shows that the film characteristics
are relatively independent of the evaporation rate, which is not surprising because the P /R ratios of both rates are equal (see Tables I and II).
It is evident from Figs. 2 and 3 that for films above 60
A
in thickness the URV resonance field values agree with the values calculated from Eq. (4), indicating that the films are quite clean and that the surface tension model for compressive stresses is obeyed. After admission of oxygen the resonance fields of films thicker than 100A
shift to a HRJ. value of 9200±60 G (= HRJ.o) ,corresponding to a final stress value lower than ±0.3X 109 dynl cm2• This will be called complete stress release.
The resonance fields in the upper band of Fig. 3 for films thicker than 60
A
are widely scattered around the predicted resonance fields and are in average somewhat too high. The scatter is large in comparison with that in Fig. 2 mostly because of the extended thickness scale; it is attributed to slight variations in experimental conditions.TABLE V. Average values for the saturation magnetization and demagnetizing factor of the films.
HRJ. [Jdem M, Neff ~100 9200 6150 490 411" 80 8700 5650 450 3.711" 60 8200 5150 410 3.311" 40 7700 4650 370 3.011" 20 7200 4150 330 2.711" =film thickness in A.
HRJ. =average resonance field from Fig. 3 (Table II) after oxygen
adsorp-tion (gauss).
H dem = demagnetizing field in gauss,
M8 =saturation magnetization of the film in emu/eml, assuming N =4?r.
Neff =effective demagnetizing factor of the film. assuming M, =490
emu/em'.
For films thinner than 60
A,
measured in URV, an appreciable deviation from the predicted behavior becomes evident. For example, a film of thickness 19A
should yield a HRJ. value of more than 15000 G, according to Eq. (4), while only 10 310 G was found. Failure of the model to predict absolute HRJ. values for very thin films is shown also in the HRJ. values after oxygen adsorption; films thinner than 100A
yieldHRJ. values lower than complete stress release can account for. In Tables I, II, IX, X, and XI, HRJ. values below 9200 G are formally related to tensile stress via Eq. (2), although it seems most unusual that the sign of the stress changes from highly compressive to highly tensile on adsorption of gases_ It is therefore postulated that for films thinner than 100
A
the compressive stress is completely released after oxygen adsorption, as for films above 100A
in thickness, and that the additional drop in resonance field must be attributed to a lowering of the demagnetizing field (shape anisotropy) from its ideal value 47rX490(=NXMs ). Thus, in the Eqs.FIG. 5_ Electron transmission micrographs of Ni thin films de-posited onto NaCI substrates.
(2) and (4),9200 G is no longer valid for HRJ.o [=NX
M.+
(w/y)J
for oxidized films thinner than 100A
andfor films in URV thinner than 60
A.
Assessments ofHRJ.o after oxidation as a function of thickness for films thinner than 100
A
can be made by averaging the resonance field values in the lower band of Fig. 3. Introduction of the postulation of a thickness de-pendent HRJ.o value for films thinner than 100A
enables all resonance data, those from the thinnest films included, to be explained by the surface free energy modeL This is shown by considering the total shifts in resonance field on oxygen adsorption. The calculated total shift is (l.13/t)X105 G [Eq. (4)J,independent of HRJ.o if it is assumed that HRJ.o does not change on adsorption. This assumption is not valid for films in the 60-100
A
thickness range; here an error of up to 1000 G is to be expected in the cal-culated shift value. The thus calcal-culated and measured thtal shifts are given in the Tables III and IV, for all film thicknesses. The shift values are plotted in Fig. 4.TABLE VI. Results of experiments with freshly cleaved {100)-NaCI substrates.
Film Resonance fields (gauss)
thickness
inA InUHV In air
48 9200 7000
67 9 200 7200
172 10 130 9220
TABLE VII. Pressures at which resonance fields shift at measurable rates; rates of shifts.
Environment Aa
UHV (1.SXlO-IOTorr) id
O2 SXlO-7 b N20 2X1()-t • air lO-ld N2 no effect H2 SX1()-te li20 1X1Q-8 f CO 7XIQ-t Pyridine 2X1Q-8
a A. lowest pressure In Torr at which resonance field shift was observed at measurable rate; B. rate of shift at pressure listed under A.
b For film thicker than 300 A approximately 1 Xl0-' Torr.
o NsO decomposes on metal parts of vacuum system; oxygen Is adsorbed
From these considerations it becomes clear that the compressive stresses in films thinner than 60
A
in UHV are actually much higher than listed in Table II. For a film of 19A
thickness the measured stress is-17X 109 dyn/cm2, in magnitude close to the yield
strength of Ni films.
The demagnetizing field in films thinner than 100
A
is lower because
N
orM
s have values lower· than those to be expected. Assuming M s to be lower than the bulk value is in contradication to experiments by Neuge-bauer18 and Hoffmann.31 Most likely the demagnetizing factor N is lower than 411" owing to magnetically isolated island formation. Values for Ms and N derived from the average HRJ.o values in the lower band of Fig. 3are summarized in Table V. Evidence of the lowering of
N
is that the thickness below which deviation from ideality occurs is approximately 60A
for films in UHV, while it is 100A
after oxidation. An oxide layer of approximately 15A
thickness is formed on a Ni film while being exposed to oxygenF Oxygen attack might occur preferentially at specific areas such as grain boundaries of the film as suggested by NeugebauerF This will have the result that after oxidation a film between 60 and 100 A in thickness, shown to behave close to ideal in UHV, may become discontinuous as far as ferromagnetic nickel is concerned, leading to nonideal behavior. From the above it follows that films thinner than 60A
are probably discontinuous as prepared in UHV. It was pointed out that these films show total resonance field shifts on adsorption of oxygen equal to those calculated from the model for continuous films. This means that the islands must have large dimensions in the film plane compared with the thickness. A shape in which all dimensions are equal would lead to considerable stresses perpendicular to the film plane, reducing the magnetic anisotropy field and thus the field shift.31 H. Hoffmann, Z, Angew. Phys. 13, 149 (1961),
Ba
in 18 h shift less than 100 G shift completed in 30 min shift completed within 10 min
shift completed within 10 min 80% completed in
!
h fully completedat higher pressures
90% completed in 10-20 min fully completed at higher pressures 90% complete in less than 15 min completed in
!
hbefore reaching the film. N. pressure measured In initial stages. dO. adsorbed 'before reaching the film. as under Footnote c. • For film thicker than 300 A approximately 5 X 10-'1 Torr. f For all film thicknesses.
Comparison by transmission electron microscopy of Ni films deposited at low rates in URV onto freshly cleaved N aCI substrates at room temperature provided more evidence of island formation. A 48-A thickness film appeared to consist of islands, a few hundred
A
in diameter, while a 187-A thickness film seemed to be continuous (see Fig. 5). How far this evidence is reliable and applicable to the films which were de-posited on glass is unknown. Films dede-posited on NaCl show different FMR behavior than those deposited on glass (see Table VI). Resonance fields of films 172 and18',
A
thick agree well with the fields obtained from films on glass; films of 48 and 67A
thickness show resonance fields in UHV and after oxidation that are approximately 1000 G too low for agreement. In addi-tion, what happens to the films during preparation for electron microscopy (coating with carbon and floating off the NaCI in water) has not been in-vestigated. Finally, heating of the films during observa-tion might enhance island formaobserva-tion. The observaobserva-tions by electron microscopy reported here are at variance with those by Neugebauer.18Resonance fields of over 9200 G for nickel thin films have never been found before. Usami et al.23 gave a
maximum value for HRJ. in URV of 8915 G for a 1050-A
thickness Ni film. After air admission the value dropped considerably; no relation to the surface tension was made. The' reason for: finding these low values is not clear.
B. Interaction of Clean Films with Various Gases As pointed out in Sec. IV.A, oxygen adsorption on clean films results in what was called complete stress release. In terms of surface tension this means that
'YNi drops from 2500 to less than 100 dyn/cm. From
the observations of the oxygen pressure at which this effect takes place and from the comparison of the interactions of clean films with other gases, it can be
STRESS EFFECTS IN Ni FILMS 393 TABLE VIII. Observed reversibility of resonance field shifts.
Film
thickness Pressure change at Resonance field shift
(A)
Environment pumpdown (Torr) (gauss)
O2 69 2XIo-a to 5Xl()-'8 not significant
71 5 X 1()-7 to 6X 1()-'8 not significant
N20 153 lXI0-1 to 5XlO-a not significant
70 100 to 2XIQ-8 not significant
H2 38 6X 10-4 to 6X 10-10 8530 to 9280 (8 h) 41 2X 10-7 to 8X 1()-'8 8510 to 8850 (1 h) 68 lXIQ-ito lXlo-a 8510 to 8730 (1 h) 160 6Xl()-7 to 2XIQ-8 9380 to 9440 (1 h) H2O 40 1 X 1()-5 to 6X 100D 8240 to 8600 (2 h) 120 5XIQ-8 to 2XIo-a 9260 to 9380 (2 h)
CO 80 4XI0-6 to 3XIO-s not significant
concluded that the presence of a monolayer of oxygen is sufficient to cause complete stress release. In films thicker than 400
A
on which a considerable NiO layer was built up, standing spin wave modes were observed in the resonance experiments, which effect will be reported elsewhere.32 NzO, admitted to clean films,had the same effect as pure O2 or air in all respects, indicating that NzO decomposes at the clean Ni as well as at the NiO surface into oxygen and nitrogen. All values of BRi3 listed in the tables were measured after
exposure of the films to air for a considerable time; it was observed that after stress release had taken place, the values of BRi3 remained constant even after long
exposure to air.
Admission of gases such as CO, H2, H20 and pyridine
to clean films did not yield as much stress release as O2 admission. When finally O2 was admitted to the
films a small additional stress release was observed, causing the stress level to drop below ±0.3 X 109
dyn/cm2
• This can be seen in the Figs. 2 and 3. The resonance field values in the presence of the various gases are listed as BB,i2 in Tables I and II. On the basis
of the experiments, no differentiation can be made between the action of the gases discussed in this paragraph; all adsorbates yield a drop of 'YNi from 2500 to roughly 200-500 dyn/ cm. The coverage on adsorp-tion at high pressures in these cases is most probably not more than a monolayer. Nitrogen showed no activity (see Fig. 3) but is reported in the literature to chemisorb on Ni at room temperature to a very small extent only.
The pressures at which the shifts in the resonance fields appear after admission of the various gases w clean films and the rates of the shifts at the pressures considered are listed in Table VII. These values are not very accurate, mainly because the apparatus was
32 M. M. P. Janssen, J. Appl. Phys. (to be published).
unsuited for quantitative kinetic adsorption studies. From the data in Table VII no reliable conclusions can be made about the coverage of the films at which stress release takes place, although the times listed seem longer than needed to form a monolayer on the film/gas interface when sticking coefficients range from 0.1 to 1. This leads to the conclusion that diffusion of gases along the grain boundaries and the film/ sub-strate interface plays a role in the achievement of stress release. The diffusion effect is a requirement from a theoretical point of view since 'Y2 entered in the surface tension model which was found to be valid. Table VIII shows that a small degree of reversibility of the resonance field shifts was observed for Hz and HzO. Brocker and Wedler33 found in their calorimetric
studies that a certain amount of adsorbed Hz could indeed be pumped off from films that had been made under conditions as used here. In the latter study the heat of desorption of H2"'on Ni was found to be as low as 19 kcaljmole. From the similarity in behavior of Hz and H20 reported here, it follows that the heat of
desorption of H20 must be close to that of Hz.
Examination of the FMR spectra (see Fig. 6 for examples) shows that, except for the very thinnest films, the linewidth decreases on adsorption of gases while the peak height increases. The values of the peak-to-peak linewidth (AB) and the total height of the peak (I) in derivative representation are listed in the Tables I and II. The decrease in linewidth is most probably due to the elimination of the linewidth contribution from local stress variations as a result of film roughness, as mentioned in the discussion of Eq. (4). Adsorption of gases reduces 'Yl and 'Yz, thus reducing the linewidth. Roughness factors of films as prepared in this study may be as large as 2.33,34
Re-33 F. J. Brocker and G. Wedler, Disc. Faraday Soc. 41, 87
(1966) .
34 R. Suhrmann, R. Gerdes, and G. Wedler, Z. Naturforsch.
film thickness 24 A
I-
~-1
7kG 12 kG (a) film thickness 160 A (a) UHV film thickness 41 A (a) UHV (b) 100 torr H2 (c) in air ~-1
12 kG (b) film thickness 181 A (b) annealed in UHV FIG. 6. Examples of ferromagnetic resonance spectra.~
I-7kG I-7kG ~-1
12kG (c)duction of the linewidth leads to an increase in signal height when the total microwave power adsorption remains constant. No determinations of the absolute absorptions by integration of the signals have been made in this work. Using integral FMR adsorption, magnetization-volume adsorption isotherms can be constructed as shown by Andreev and Selwood30 for
superparamagnetic nickel particles; probably this method can be extended to very thin Ni films as in-vestigated here. The linewidths found for very thin films (up to approximately 60
A)
are very large and3Ii A. A. Andreev and P. W. Selwood, J. Catalysis 11, 261 (1968).
~
-1
12 kG
(d)
close to the values found in studies of superparamag-netic Ni particles, which observation further supports the contention of Sec. IV.A that these films are dis-continuous.
C. Influence of Annealing on the Film Characteristics Two films evaporated at low rates in URV onto soft glass substrates at room temperature were annealed in URV for 20 min at 2500 and 270°C, respectively. During annealing the pressure in the system was in the high 10-10 Torr range. The resonance fields of the
'"
"-o 0 -tI+
v
o 00 V) ' " I I V)....
~ o 0 -tI I oc: N N+ +
~ ~
X X V) V)STRESS EFFECTS IN Ni FILMS
"
" :J g, 11 395 ~ (A) ~ii! 10 L - -_ _ _ _ _ - -r---
-]
~ Q - - - --? - - - --~-~--I=(o)======:;;-;(b)n [i: 0 0 w 8 \i ::f 7 ~ w '" 6i
5 4 (8)soft glass substrates
50 100 150
----+ FILM THICKNESS Oingstrom) FIG. 7. Effect of annealing on film characteristics. Substrate temperature during deposition~35°C; low deposition rates; anneal-ing time 20 min. Pressure duranneal-ing annealanneal-ing high 10-10 Torr range.
., as prepared in URV; 8, after annealing at (a) 250°C, (b) 270°C in URV; 0, after final exposure to air. (A) and (E): results from Fig. 3; 1 and 2: curves predicted from theory.
exposure to air are represented in Fig. 7 and Table IX. The resonance fields of the films as prepared agree very well with the values calculated from the surface tension model for compressive stresses. Annealing results in apparent stress release and the resonance fields measured at room temperature after annealing are in the range of stress-free films; however, admission of air gives an additional lowering of the resonance field, indicating that tensile stress is the final result. The tensile stress can partly be explained by thermal effects during cooling: The thermal expansion co-efficient of 0211 glass is 72X 10-7 aC-l, that of nickel
is 125 X 10-7 0C-J , so that assuming a value for the
modulus of elasticity for Ni films of 2.0X 1012 dyn/cm2
results in a stress level of approximately +3X109
dyn/cm2 in the films, while +4 to +5X109 dyn/cm2
was found. The reason for this devia ion might be the lowering of the demagnetizing field owing to island formation during annealing, which is very likely with films of this thickness. From these experiments it is clear that the intrinsic stresses investigated here cannot be removed by annealing, which is strong evidence of the surface tension model. The apparent stress is zero after annealing only because the tensile stress from thermal sources is approximately equal to the com-pressive intrinsic stress. The latter component is re-moved by adsorption of oxygen. The total field shifts on adsorption of oxygen expected for the films, if not annealed, are 1140 and 870 G with the 108-A and 181-A thickness films, respectively. The total shift values on adsorption of oxygen found on the annealed films are 1130 and 770 G with the 108-A and 181-A thickness films, respectively. This indicates that the resonance field shifts are independent of film treatment in UHV, which is in complete agreement with what is to be expected from the surface tension model.
4 35
soft glass substrates
145· 235 300 335
SUBSTRATE TEMPERATURE DURING EVAPORATION <OC)
,
FIG. 8. Effect of substrate temperature during deposition on resonance fields, measured at room temperature. Low deposition rates. Upper curve: as prepared in UHV; Lower curve: after air admission. (A) and (B): Results from Fig. 3 for films thicker than 100 A..In Fig. 6(d) the FMR spectra of one of the films are shown. The data on linewidths and intensities are given in Table IX. The linewidth appears to broaden on annealing, and on adsorption of oxygen the linewidth decreases (but not completely) to its original value. The broadening on annealing and the partial narrowing on adsorption of oxygen can be explained by an in-crease in the film roughness during annealing, similar to the linewidth effect discussed in Sec. IV.B. The fact that the linewidth does not reach its original value after adsorption of oxygen might be due to inhomogeneous thermal stress distribution.
D. Films Deposited onto Heated Substrates Experiments on films deposited onto heated soft glass substrates at low rates appear not to be as con-clusive as those described in Sec. IV.C. The vacuum during evaporation was relatively poor (see Table X), possibly leading to partial contamination of the films. The measured values of the resonance fields after cooling are plotted in Fig. 8. After adsorption of oxygen the stresses are tensile in all cases. Only ap-proximately one half of the stress levels of +4 to +7Xl09 dyn/cm2 can be accounted for by thermal
effects. This means considerable island formation or lowering of M.(internal contamination) of the films and island formation during deposition at high tempera-ture is expected with films in this thickness range. The total field shifts on adsorption of oxygen are lower than predicted by the surface tension model, except for the film deposited at 145°C. If the films are assumed to be clean as prepared, the explanation of the latter observation might be that the shape of the islands formed during deposition is such that considerable stresses exist normal to the film plane.
01) 0 0 0 ~ ~ ~ ~
+ + + +
~ ~ ~ ~ " ' 0 1 ) 0 1 ) 0 1 ) I I I I UU o 0 0 ' " 0 ' "..,,,,
.
""
STRESS EFFECTS IN Ni FILMS E. Influence of the Environment During
Film Deposition
The experiments reported and discussed In this
section were performed to obtain some information about the importance of the quality of the vacuum during film deposition and measurement. Films were evaporated at low rates onto soft glass substrates at
3SoC, conditions equal to those in the experiments that are represented in Fig. 3.
When evaporation was carried out in a system in which a pure N2 pressure was maintained after prep-aration to attain UHV, no deviation in the resonance fields from the predicted values for films deposited and measured in URV were found (see Fig. 9 and Table XI); in fact, the film evaporated in 2X 10-7
Torr N2 showed the highest resonance field ever obtained, viz. 11 420 G. These observations are in agreement with the lack of activity found when N2
was admitted to films prepared in UHV as reported in Sec. IV.B.
Investigation of films deposited in a system in which a pressure of pure O2 was maintained after preparation to attain URV showed that with a film deposited in
2X1O-9 Torr O
2 no deviation from ideality was found.
A film deposited in SX 10-7 Torr O
2 yielded lower
resonance fields than ideal ones. The low values of the resonance fields are most probably due to lowering of M. as a result of bulk film contamination, although in Table XI they are related to tensile stresses.
Films prepared in an unbaked system at pressures in the 10-7 Torr range showed large deviations from
ideality, while dependence on film thickness is apparent from Fig. 9. The components of the residual gas
atmos-~
&
11 g (A) :;;: -; ~ 10 L-____________________ ~~== ____ _=~~ I9
9 w G: w 8 ~ ~ 7 ~ w <r /5i
5 4 ---5---;:..0--:::::=f=====::::j
(B) 100 150---+ FILM THICKNESS (Angstrom)
FIG. 9. Effect of vacuum conditions during deposition on reso-nance fields. Low deposition rates; substrate temperature during deposition 35°C. (A), results from Fig. 3 obtained under ideal conditions as measured in UHV; (B), as measured after gas adsorption; 1 and 2, curves predicted from theory; 0, films pre-pared in unbaked system (pressure 10-7 Torr range), upper value
before air admission, lower value after air admission; A, films prepared in baked system in pure O2 at pressure indicated, (a)
after air admission. El, films prepared in baked system in pure N2 at pressure indicated, (a) after air admission.
<- ":
t-:
<") ~ ~ <") N N ...-4 \0 '1"""'4 '1""""1 0++++++-+1
"! N c:::: \Q ....~ ~ ~
+
I
V) ~ ~ ~ U') ~ \0 '1""""1 '1"""'4 ... N 00 \Ci ...0 0 I II'
I I I-+1
-.j< <") 00 0\ N-.j< V) ~ .... <- -0 ~ <")+ +
I'
I I I+
~ ~ ~ ~
$
~ ~
.... 0 -.j< V) \Q <- 0....
....
....
397phere are mainly CO and H20. Here the shifts in
resonance field on exposure to air are very small, in-dicating almost complete absence of compressive stresses owing to surface tension in the films as pre-pared.
Freedman22 reported a similar investigation, in which
only HRJ. was measured after exposure to air. His
experimental results as far as the influence of N2,
H20 and O2 are concerned, are comparable with those
reported here, but the absolute values of the resonance fields reported by Freedman are lower than those obtained in this study.
V. CONCLUSIONS
(1) The reported compressive stresses in Ni films as prepared and measured in UHV can be accounted for completely by surface tension. This conclusion is made on the basis of the quantitative agreement of the measured stress values with those predicted from a simple model, the strong influence of gas adsorption on the stress levels, and the fact that these stresses cannot be removed by annealing of the films in UHV. The surface tension contribution to the stress levels in films of technically important thicknesses (1000
A)
is found to be smaller than 1 X lO9 dyn/cm2; neglect
of this contribution does not lead to serious errors as expected by Hoffmann.ll Exposure of the films to low
pressures (1(J-6 Torr range) of active gases leads to
virtually complete stress release.
(2) Clean Ni films thinner than approximately 60
A,
deposited at 25-35°C substrates, are most probably built up of magnetically-partly-isolated islands. Afterexposure to oxygen the discontinuity thickness limit shifts to approximately 100
A.
(3) This study offers a method for the determination of the magnitude of the surface tension at room tem-perature of ferromagnetic materials with a large value of the magnetostriction constant. Studies of the effect of gas adsorption on the surface tension can be per-formed.
(4) The results of this study provide theoretical support for the stress sorption theory for stress cor-rosion cracking and hydrogen embrittlement. However, the results in themselves do not prove that this theory is a ubiquitous explanation for stress corrosion cracking phenomena. The generalization of the results to other materials and their interaction with specific active gases is likely to be valid. Thus, for example, the work of Williams and Nelson36 on hydrogen embrittlement
of steel is supported. The data from Hanna et al.,37 for example, on delayed failure of steel in an O2 , argon, and H20 environment can be explained by surface
tension considerations.
ACKNOWLEDGMENT
The author wishes to thank Dr. A. C. Zettlemoyer, former director of the Center for Surface and Coatings Research, and Dr. H. Leidheiser,
Jr.,
present director of that institute, for many helpful and stimulating discussions.36 D. P. Williams and H. G. Nelson (unpublished).
37 G. L. Hanna, A. R. Troiano, and E. A. Steigerwald, Trans.
