Professional Documents
Culture Documents
2634 J. Mater. Res., Vol. 16, No. 9, Sep 2001 © 2001 Materials Research Society
http://journals.cambridge.org Downloaded: 09 Apr 2015 IP address: 138.251.14.35
M. Pang et al.: Thin-film fracture during nanoindentation of a titanium oxide film–titanium system
FIG. 2. Load–depth curves showing permanent deformation prior to FIG. 4. Load–depth curves for a titanium with a 128-nm-thick oxide
pop in. The titanium oxide thickness is approximately 90 nm and was film and pure titanium without an oxide film, respectively.
formed by anodizing at 8.4 V (versus Ag/AgCl) in the sulfuric acid
solution.
and without surface oxides follow the same geometri-
cally necessary curves, indicating that the function of the
fractured oxide is very limited in determining the plastic
properties of the deforming system. It should be noted
that these indentation were carried out in two separate
grains, which accounts for any small differences in be-
havior after film fracture.
IV. MODEL
The titanium oxide/titanium system can be modeled as
a hard elastic plate on a soft yielding substrate under
contact as it is characterized by a thin stiffer ceramic
coating (Eoxide ⳱ 300 GPa25) on a more compliant and
ductile metal substrate (ETi ⳱ 110 GPa). The plate is
assumed to undergo elastic deformation until brittle frac-
ture. The substrate is displaced elastically, elasto-
plastically, and finally fully plastically depending on the
compression pressure. In particular, the substrate is as-
FIG. 3. AFM deflection image of an indentation of 2-mN load into sumed to be elastoplastically deformed prior to the yield
anodized Ti. The arrow denotes cracks in the oxide. discontinuity due to the constraints of surface film.
As shown in Fig. 5, the critical depth prior to the yield
Figure 4 explicitly demonstrates the considerable role point, i.e., the maximum elastic deflection of the film, is
of the surface film in supporting the contact pressure on the order of its thickness despite data scatter due to
prior to its fracture. The load carried by the 120-nm-thick effects of grain orientation and microstructural inhomo-
surface film accounts for up to 1⁄3 of the total applied load geneities. The error bars in this figure are the actual data
at the depth at which film fracture occurs. The effect of range, with the symbols denoting the average value of at
the surface film diminishes at relatively low loads and least 10 indentations in each condition. In such a condi-
becomes negligible at the initial loading, as the separa- tion, the midplane stretching effect must be considered
tion of the two curves is indistinguishable. Furthermore, owing to the change in plate geometry.26 The plate un-
the examination of the loading segments after the yield dergoes nonlinear deformation comprising of both bend-
discontinuity reveals that the loading curves of both with ing and membrane stretching solutions. Consequently,
c= 冑3Ps
2f
, (1)
formidable, a loading configuration with an analytical the elastoplastic deformation prior the yield point, the
solution available needs to be employed to simplify the plastic deformation after the yield point, and that of
situation. The appropriate selection depends on the rela- the typical elastic deformation in the condition of unload-
tive size of the plastic zone and the contact area, repre- ing. Similar to the power law approach of load versus
sented by c/a. The value of c/a is approximately depth dealt with by Oliver and Pharr,20 Hainsworth,
estimated as 2.7, which is neither close to the ideal cen- Chandler, and Page,31 defined the unit of Ke to be Pa m2−n
tral nor uniform loading conditions. (The details are to maintain the dimensionality. It is also worth noting that
elaborated in Appendix A). Bearing in mind that the the choice of allowing n to vary is an artifact made solely
central loading gives an upper bound and the uniform for the convenience of the curve-fitting process. Such an
loading the lower bound in the plate deflection, a better approach is rather robust in its ability to describe these
estimation of the plate bending in the current model is experimental data as discussed in detail later.
expected to give an intermediate plate deflection. By superposition of film deflection solution and sub-
The load versus center deflection for an elastic film strate yielding solution, the resultant relationship of load
prior to its fracture can be simplified as30 and depth for the coated system is
P= 再 冋
22f 16E f hf3
␦+
191
E h ␦3
648 f f 册冎 1Ⲑ2
, (2) P = Ke␦n + 再 冋
22y
3
16E f h f3
12共1 − f2)
␦
册冎
3 12共1 − f 兲
2
1Ⲑ2
191
where P is again the applied load, f is the characteristic + E h ␦3 . (6)
flow stress of the substrate, ␦ is the deflection depth, hf is 648 f f
the film thickness, Ef is the elastic modulus of the film By curve fitting of the upper loading part prior to the
(Ef ⳱ 300 GPa in the literature25), and f is the Poisson yield discontinuity using Eq. (6), a value of n can be
ratio of the film (f is taken as 0.3). The detailed develop- obtained. Figure 7 presents a typical curve-fitting proce-
ment of the plate deflection solution which incorporates dure to determine the value of n for loading prior to film
both the corrected plastic zone size and the compromised fracture in which plate bending dominates (nplate).
loading condition is discussed in Appendix B.
The load and depth relationship of the pure substrate V. DISCUSSION
metal can be described by the Hertzian elastic contact
A. Model analysis
mechanics when a spherical tip is loaded onto a flat
surface:24 Additional fits of the other regions of the indentation
include an offset in the load or depth to account for the
4 loading support of the film or the shape of the surface
P ⳱ √REs␦3/2 ,
3 (3)
where R is the tip radius, the value of which is 1.5 m for
this Berkvoich tip. Es is the reduced elastic modulus of
the substrate that is taken to be 110 GPa.
Furthermore, Eq. (3) can be rewritten as
P ⳱ Ke␦3/2 , (4)
where Ke is defined as a characteristic constant repre-
senting the elastic deformation behavior of the substrate,
the value of which equals to 4/3 √REs.
The elastoplastic deformation behavior of the substrate
while the oxide film is still intact is proposed to be de-
scribed by a similar equation, where the 3/2 power for the
elastic contact is replaced by a variable n which repre-
sents the characteristics of the elasto-plastically deforma-
tion of the substrate:31
P ⳱ Ke␦n . (5)
In the above equation, the same elastic Hertzian con-
FIG. 7. Representative curve-fitting procedure showing various re-
stant Ke is intentionally kept in an attempt to reveal gions of the indentation and the model fits to the plate-bending com-
the change in the deformation mechanism by sensing the ponent, the plastic deformation component, and the elastic unloading.
subtle alternation of the slope of the loading curve among The raw data are slightly offset for clarity.
after yield, respectively, for the region in which plastic addition, the value of nelastic is slightly greater than would
deformation of the substrate is the primary deformation be expected for a Hertzian spherical contact. This is as
mechanism (nplastic) and the elastic unloading of the tip would be expected from the work of Goodman and
(nelastic). Equation (5) is utilized for the curve fitting. At Keer32 for the elastic contact of spheres in cavities
first glance, the results may seem unreasonable since the (where in this case the cavity is the result of plastic
values of nplastic and nelastic are higher than that of a deformation). In thinner films, i.e., hf < 40 nm, the value
purely elastic contact (n ⳱ 1.5) which is not logical since of nplate falls into the range of nelastic instead of that of
conventionally a power-law relationship for the load– nplastic, which may indicate that the initial deformation is
depth coefficient of plastic flow should have a lower more elastically characteristic that is in good agreement
coefficient than that of elastic deformation (i.e., n prior to with the experimental results shown in Fig. 4. In such
the excursion is greater than the 1.5 value expected in a cases, the model to quantify loading using plate deflec-
purely elastic contact). However, in further examination tion becomes inappropriate.
of Eq. (5), the magnitude of the constant for elastic con-
tact, i.e., Ke, was employed regardless of elastic or plastic B. Effect of plate bending and stretching
loading; in reality the constant should be decreased to As described in the model, the applied load is balanced
account for plastic deformation. Therefore, it should be by two independent factors, the resistance of plate de-
emphasized that Eq. (5) is used as a convenient repre- flection and that of the substrate. The effect of the plate
sentation of the real load–depth behavior and that the bending and stretching is illustrated in Fig. 9. The pure
choice of fitting the coefficient was bound to lead to plate solution locates below that of load–depth curve and
increased values of n since the constant was not changed. reaches almost zero in the case of very thin films, indi-
Figure 8 summarizes the curve-fitting results of n as a cating that the plate-bending effect diminishes as the film
function of film thickness, showing the average values gets thinner. Consequently, it is stated that the present
(solid symbols) as well as the entire range of values ob- model works in the conditions of relatively thicker film
served (open symbols). Obviously, when the film is rela- that can be reasonably simulated as an integrated plate
tively thick, i.e., hf > 40 nm, the value of nplate is closer to with a certain mechanical property. It would be inappro-
that of nplastic than that of nelastic, indicating that the plas- priate to discuss the strengthening effect of the surface
tic deformation occurs in the substrate prior to the film film if the film is less than 40 nm for this tip.
fracture. The slightly lower nplate may be the result of
the elastoplastic-like deformation which is caused by the C. Estimation of film strength prior to fracture
surface constraint of the film prior to its fracture. In
The stress distribution for a film/substrate layered
system can be predicted numerically by finite element
simulations. Weppelman and Swain have already demon-
strated that radial tensile stress around the contact edge is
responsible for the film through-thickness cracking in
vapor deposited films.17 Accordingly, the strength of
the film prior to fracture can be roughly estimated by
taking into account the radial stress of both the plate
bending and membrane solution for an oxide film.
The radial stress for the condition of plate bending is
given by30
␦ hf
rB = r E f . (7)
c2
Similarly, the radial stress in the case of pure membrane
flexure is30
␦2
rM = ␣ r E f , (8)
c2
where ␣r and r are constants taken from Timoshenko
et al.,30 the value of ␣r is 0.357, and that of r is 2.198.
FIG. 8. Comparison of nplate, nplastic, and nelastic in the range of film
thickness used in this study. Total range of measurements is shown; Therefore, the stresses and strains in the film can be
solid symbols represent averages, and the open symbols are the maxi- calculated as a function of indenter displacement, and the
mum and minimum of each sample set. stress at which a “pop in” occurs can be calculated. For
a first-order estimate, the total radial stress is the super- properties. This titanium oxide film is microporous with
position of the above two effects as the strains, and sub- a pore size about 100 nm, as shown in Fig. 10, and the
sequent stresses would be additive. Table I lists the strength is therefore expected to be lower. Futhermore,
results in the conditions for relatively thick films. Note stresses that develop during film growth could lead to
that the load carried by the substrate instead of the total either an increase or decrease of film strength depending
applied load was used to calculate c, which is accom- on the particular stress state. Therefore, the calculated
plished by incorporating an empirical constant discussed film fracture strength based on Eqs. (7) and (8), the av-
in Appendix B. erage value of which is about 15 GPa, appears to be a
As a rule of thumb, the brittle fracture strength can reasonable estimation.
be estimated as E/10, which would be about 30 GPa
for the film. In practice, the film strength can deviate
VI. SUMMARY
from this estimation on the basis of specific film
(1) On the basis of experimental nanoindentation re-
sults, which demonstrate permanent deformation prior to
a yield point and circumferential cracking on the contact
perimeter for electrochemically anodized titanium, the
mechanism of oxide film fracture rather than a threshold
for dislocation motion is presented to explain the pop-in
phenomenon observed during indentation of an anodized
titanium alloy.
(2) Assuming that the titanium oxide/titanium system
acts as a hard coating/soft substrate under contact, a
model is developed to predict the mechanical response of
the titanium oxide/titanium coated system. The compos-
ite load–depth solution is the superposition of both the
large deflection of a hard coating and the plastic flow of 12. M.R. McGurk and T.F. Page, Surf. Coat. Technol. 92, 87 (1997).
the underlying substrate. The highlight of the model is 13. W.W. Gerberich, A. Strojny, K. Yoder, and L-S. Cheng, J. Mater.
Res. 14, 2211 (1999).
that the Hertzian-like solution is employed to present the 14. H. Djabella and R.D. Arnell, Thin Solid Films 213, 205 (1992).
deformation behavior of substrate by incorporating a co- 15. P.K. Gupta, J.A. Walowit, and E.F. Finkin, J. Lubr. Technol. 95,
efficient obtained by curve fitting experimental load– 427 (1973).
depth curves prior to film fracture. 16. W.T. Chen, Int. J. Eng. Sci. 9, 775 (1971).
(3) The model works well in the condition of rela- 17. E. Weppelman and M.V. Swain, Thin Solid Films 286, 111
(1996).
tively thick oxide films (greater than 20 nm). The film 18. M. Pang, D.P. Eakins, M.G. Norton, and D.F. Bahr, Corrosion
bending action gradually vanishes as the film gets (2001, in press).
thinner. 19. J.B. Pethica and W.C. Oliver, Phys. Scr. T19, 61 (1987).
(4) To a first-order approximation, the film fracture 20. W.C. Oliver and G.M. Pharr, J. Mater. Res. 7, 1564 (1992).
strength is estimated as 15 GPa, which is in good agree- 21. A.B. Mann and J.B. Pethica, Appl. Phys. Lett. 69, 907 (1996).
22. S.A. Syed Asif and J.B. Pethica, Philos. Mag. A 76, 1105 (1997).
ment with theoretical estimation (E/10) when also con- 23. J.D. Kiely and J.E. Houston, Phys. Rev. B 57, 12588 (1998).
sidering the effects of microscopic defects and stresses 24. K.L. Johnson, Contact Mechanics (Cambridge University Press,
in the film. New York, 1985).
25. M. Barsoum, Fundamentals of Ceramics (McGraw-Hill, New
York, 1997).
ACKNOWLEDGMENTS 26. A.C. Ugural, Stresses in Plates and Shells (McGraw-Hill, New
York, 1981).
The XPS characterization described in this paper was 27. P.K. Gupta and J.A. Walowit, J. Lubr. Technol. 96, 250 (1974).
performed by M.H. Engelhard in the Environmental Mo- 28. D. Kramer, H. Huang, M. Kriese, J. Robach, J. Nelson, A. Wright,
lecular Sciences Laboratory, a national scientific user D. Bahr, and W.W. Gerberich, Acta Mater. 47, 333 (1998).
29. J.C. Hay and G.M. Pharr, in Thin Films—Stresses and Mechanical
facility sponsored by the Department of Energy’s Office
Properties VII, edited by R.C. Cammarata, E.P. Busso, M. Nastasi,
of Biological and Environmental Research and located at and W.C. Oliver (Mater. Res. Soc. Symp. Proc. 505, Warrendale,
Pacific Northwest National Laboratory. The financial PA, 1998), pp. 65–70.
support through a grant from the donors of the Petroleum 30. S. Timoshenko and S. Woinowshy-Krieger, Theory of Plates and
Research Foundation, administered through the Ameri- Shells (McGraw-Hill, New York, 1959).
31. S.V. Hainsworth, H.W. Chandler, and T.F. Page, J. Mater. Res. 8,
can Chemical Society, is gratefully acknowledged. Ad-
1987 (1996).
ditional support was provided through Sandia National 32. L.W. Goodman and L.M. Keer, Int. J. Solids Struct. 1, 407 (1965).
Laboratories, under a contract from the United States
Department of Energy (No. DE-AC04-94AL85000). The
experimental assistance of C.L. Woodcock, D.E. Eakins, APPENDIX
and M.G. Norton of Washington State University and Dr.
A. Evaluation of the actual loading conditions
N.R. Moody of Sandia National Laboratory is greatly
under contact by estimating the ratio of the
appreciated.
plastic radius and the contact radius (c/a)
A series of ex situ indentations were made on a pas-
REFERENCES sivated titanium with a 128-nm-thick surface oxide film.
1. R.M. Johnson and R.J. Block, Acta Metall. 16, 834 (1968). According to the occurrence of loads at excursion, sev-
2. Chr. Panagopoulos and Hel. Badekas, Mater. Lett. 7, 201 (1988). eral indentations were made by intentionally controlling
3. T.F. Page and S.V. Hainsworth, Surf. Coat. Technol. 67, 305 the peak loads to be slightly lower than the average ex-
(1993). cursion load. The resultant continuous load–depth curves
4. N.G. Chechenin, J. Bottiger, and J.P. Krog, Thin Solid Films 261, prior to the film fracture provide a better estimation of
228 (1995).
5. M. Pang, D.E. Wilson, and D.F. Bahr, in Thin Films-Stress and the contact depth and thus the contact radius a at the film
Mechanical Properties VIII, edited by R. Vinci, O. Kraft, fracture. Refer to Fig. 2 in the text for a typical load–
N. Moody, P. Besser, and E. Shaffer II (Mater. Res. Soc. Symp. depth curve.
Proc. 594, Warrendale, PA, 1999), p. 501–506. For instance, by curve fitting the upper part of the
6. T.Y. Tsui, J. Vlassak, and W.D. Nix, J. Mater. Res. 14, 2204 unloading load–depth segment prior to film fracture in
(1999).
7. D.F. Bahr, D.E. Kramer, and W.W. Gerberich, Acta Mater. 46, Fig. 2, the contact depth h is estimated around 60 nm.
3605 (1998). Accordingly, the contact area, A, calculated on the basis
8. T.F. Page, W.C. Oliver, and C.J. McHargue, J. Mater. Res. 7, 450 of the tip area function,20 is around 296,800 nm2.
(1992). The contact radius a can be roughly calculated as
9. P.M. Ramsey, H.W. Chandler, and T.F. Page, Surf. Coat. Technol. a ⳱ √A/ ⳱ 307 nm, assuming a circular contact.
49, 504 (1991).
10. M.R. McGurk, H.W. Chandler, and T.F. Page, Surf. Coat. Technol. Theoretically, the plastic zone radius c can be evalu-
68/69, 576 (1994). ated by Eq. (1) in the text. As mentioned in the text,
11. A.J. Whitehead and T.F. Page, Thin Solid Films 220, 277 (1992). Ps represents the load solely carried by the substrate.
However, the experimentally recorded load is the overall where Df is the flexural rigidity of the film which is
load distributed over both the surface film and the un- given by30
derlying substrate. A better estimation of the plastic zone
can be achieved if the partial load carried by the surface E f hf3
film can be excluded from the total applied load. A series Df = . (B3)
12共1 − f2兲
of nanoindentation curves were compared among the ma-
terials with and without the surface oxide film, as shown Pf is defined as the partial load carried by the film, and
in Fig. 4. By examination of the loading section prior to c is the radius of the clamped plate which is identical to
the yield discontinuity in Fig. 4, it is obvious that the the plastic zone radius within the substrate in practice and
total load carried by the film and substrate composite is given by Eq. (B4):
higher than that without the film at a given displacement.
The actual load carried by the substrate Ps at film fracture 3Ps
c2 = . (B4)
can be estimated as the applied load for the pure substrate 2f
deformation corresponding to the penetration depth at the
For the convenience of term collection, it is assumed that
film fracture for the film/substrate system if assuming
Ps ⳱ fPf , where f is a constant, given that Ptotal ⳱ Ps + Pf.
only elastic deformation occurs in the film. Data taken
Upon the application of the empirical constant f, the
from indents of which Fig. 4 is representative give an
above equations can be further simplified as
average value of Ps at the yield point for the titanium
with a 128-nm oxide film of 1847 N. Therefore, the
radius of the plastic zone can be calculated as Pf = f − 0.5冋 2f
3 冉 ␦3
64Df ␦ + 0.488 2
hf
冊册 0.5
, (B5)
c= 冑 3Ps
= 857 nm . (A1) for the uniform loading condition, and
冋 冉 冊册
2y
2f 191 0.5
冉 冊
With substitution of the value of f into above equa-
␦3 tions, the solution for both ideal uniform and central
Pf c2 = 64Df ␦ + 0.488 . (B1)
h f2 loading conditions can be given by
191
Pf = 0.632 冋 2f
3 冉 ␦3
64Df ␦ + 0.488 2
hf
冊册 0.5
, (B7)
Pf c2 = 16Df ␦ + E f h f ␦3 , (B2)
648 for the ideal uniform loading. And
FIG. B1. Representative approach for determining the empirical con- FIG. B2. Plate deflection solutions for uniform loading [Eq. (B6)],
stant f for a titanium with a 124-nm-thick oxide film. central point loading [Eq. (B7)], and Eq. (2) compared to experimental
data.
Pf = 0.632 冋 冉
2f
3
16Df ␦ +
191
648
E f hf ␦3 冊册 0.5
, (B8) experimental data are approximately 1⁄2 the difference
between Eqs. (B6) and (B7). Therefore, Eq. (2) in the
for the ideal central loading. text appears appropriate for estimating Pf.
Figure B2 illustrates the theoretical plate bending It should be emphasized that Eq. (2) is not the direct
solutions represented by Eqs. (B6) and (B7). To first or- employment of the central loading solution as discussed
der, it will be assumed that the actual loading solution previously13 but is a simplification of the actual condi-
lies midway between the ideal uniform (upper bound) tion that incorporates both the corrected plastic zone size
and central (lower bound) solutions. Coincidentally, the for deformation in the substrate and a plate-bending ap-
central loading solution without the 0.632 coefficient proximation. Such an approach is materials-dependent,
[Eq. (2)] is midway between the solutions of Eqs. (B6) and the empirical constant f may differ for other coated
and (B7). Since Pf ⳱ Ptotal − Psubstrate, the experimental materials. However, the overall idea of achieving the
data from Fig. 4 are plotted with Eqs. (B6) and (B7). plate deflection solution is universal; current work is un-
At depths between half the film thickness and the film der way to examine the range of the constant f for a
fracture event, the data are within a factor of 2 of variety of materials characterized by hard films and soft
Eqs. (B6) and (B7). At the actual film fracture event, the substrates.