Sei sulla pagina 1di 6

Z. Kristallogr. Suppl.

23 (2006) 123-128 123


© by Oldenbourg Wissenschaftsverlag, München

X-ray diffraction from epitaxial thin


films: an analytical expression of the line
profiles accounting for microstructure
A. Boulle*, R. Guinebretière, A. Dauger
Science des Procédés Céramiques et de Traitements de Surface – CNRS UMR 6638, ENSCI,
47 à 73 avenue Albert Thomas 87065 Limoges Cedex
*
Contact author; e-mail: a_boulle@ensci.fr

Keywords: X-ray diffraction, thin films, microstructure, roughness, thickness fluctuations

Abstract. An analytical interference function describing the X-ray diffraction line profile of
epitaxial thin films is derived. This interference function describes the intensity distribution
in a direction perpendicular to the film surface and accounts for the combined effects of
different type of defects: interface and surface roughness and their state of correlation,
cumulative and non cumulative lattice disorder. The derived interference function is
analytical which allows easy implementation and can hence be used for routine analysis of
the microstructure of thin films.

Introduction
During the three past decades X-ray diffraction (XRD) has become essential for the non-
destructive characterization of epitaxial thin film structures. There are two theories that
describe the scattering from thin crystals. The rigorous dynamical theory [1] applies mainly
to perfect, or nearly perfect crystals. Takagi [2] and Taupin [3] extended this theory to the
case of non-ideal crystals containing a slowly varying strain profile. The numerical or
iterative resolution (within the lamellar approximation) [4, 5] of the Takagi-Taupin equations
enables one to recover depth-resolved information concerning strain and/or composition in
the films. Real structure effects, e.g. random lattice spacing fluctuations, remain very
difficult to incorporate into dynamical theory [6]. Besides, the more approximate kinematical
scattering theory has attracted much attention as it more easily deals with structural defects
commonly encountered in real crystals.
The description of the intensity distribution diffracted by a thin crystal of thickness t actually
goes back to Laue [7] who first derived the interference function known as the Laue
function: sin²Q(t/2)/sin²Q(d/2), where Q is the length of the scattering vector and d the
interplanar spacing of the diffracting planes. In practice however, this relation doesn't apply
to most samples as the presence of defects strongly alters the shape of the intensity
distribution, principally by suppressing the interference fringes. Several authors modified
this expression in order to account for the effect of defects: thickness fluctuations [8, 9],

Unauthenticated
Download Date | 12/3/17 5:20 PM
124 European Powder Diffraction Conference, EPDIC 9

roughness [10] and lattice disorder [9,11]. The combined effects of different defects are in
general handled using a numerical integration of the expression of the intensity distribution.
In this communication we derive a generalized interference function able to describe
analytically the XRD profile of a single epitaxial thin film with a microstructure made of
different type of defects: film thickness fluctuations, roughness, cumulative and non-
cumulative random lattice spacing fluctuations. We emphasize that the concepts used in the
present paper are not new when considered separately (see for instance refs [8-11]).
However, to our knowledge we present here the first derivation of an analytical expression
where all these defects are taken into account simultaneously. Each of these defects affect the
XRD profiles in a distinct fashion which enables one to recover these parameters by a direct
quantitative comparison of the generalized interference function with experimental data. This
task is greatly simplified by the analytical character of the proposed function which allows a
straightforward implementation in any existing computing program (or even any spreadsheet
software) and makes it attractive for routine use. However, when more complicated
structures are considered (multilayers for instance) then the intensity distribution taking into
account microstructural defects still has to be numerically evaluated (this can be for instance
performed using the freely available program SUPREX [9]).
The paper is organized as follows. In the first section we focus on the effects of a non-ideal
morphology (thickness fluctuation and roughness). In the second section, we consider the
effect of additional lattice spacing fluctuations.

Effects of a non-ideal morphology


Framework and assumptions
The derivation is carried out within the framework of kinematical scattering theory. Within
this theory, the amplitude diffracted along a rod perpendicular to the film surface is given by
[12]:
E q z Fh dx dydz x, y,z exp iq z z (1)
where Fh is the structure factor of the reflection with vector h. Ω(x,y,z) is the shape factor of
the crystal (equal to 1 when the position vector r = (x,y,z)T points within the crystal, and 0
outside), qz is the z component of the reduced scattering vector q = Q – h (Q being the
scattering vector and the z axis being chosen parallel to the outwards film surface normal).
In the following, we consider a film made of flat crystallites for which the shape factor can
be written Ω(x,y,z) = Ω(x,y).Ω(z), and Ω(z) = 1 if z ∈ [zL, zU], otherwise Ω(z) = 0 (with zU
and zL being the coordinates of the upper and lower surface of the film, respectively). In the
case of a perfectly flat film the intensity computed with the aid of equation (1) is
|Fh|2sin2(qzt/2) / (qz/2). If the film is not perfectly flat, zU and zL are functions of the in-plane
coordinates (x,y). We assume in-plane statistical homogeneity so that zU(x,y) and zL(x,y) can
be described by random variables of the in-plane coordinates such as zU,L(x,y) = <zU,L> +
∆zU,L(x,y). These variables are connected through the film thickness t = zU – zL (which is also
a random variable) and are therefore not independent. The subsequent averaging procedure is
greatly simplified if one is able to identify independent random variables [13]. It can hence
be noticed that the roughness of the substrate (σL) is in general induced by the different
preparation steps undergone by the substrate (cutting, polishing,…), whereas thickness

Unauthenticated
Download Date | 12/3/17 5:20 PM
Z. Kristallogr. Suppl. 23 (2006) 125

fluctuations (σt) are solely dependent on the deposition process. A convenient choice of
independent random variables is therefore zL and t = zU – zL.
Different parts of the film scatter coherently
We first consider the case where different parts of the sample (with different zL and t)
illuminated by the X-ray beam scatter coherently. The intensity is given by I = <E><E>*, and
we therefore average the scattered amplitude over all possible values of t and zL. In
performing this average we assume that all random variables can be described by continuous
probability density functions (PDFs), i.e. the film thickness is much greater that the smallest
height variation, that is to say one interplanar spacing. In the case of very thin films discrete
PDFs must be used [8,13]. Additionally we shall arbitrarily assume Gaussian PDFs.
Obviously, different PDFs will yield different diffraction profile shapes. For instance the
effect of the crystal lattice termination has been deeply studied in the field of crystal
truncation rod scattering on semi-infinite crystals [14-16] and, though to a lesser extent, in
thin film XRD [8, 17]. In fact, it turns out that equation (1) can be solved analytically for
several interesting cases. For brevity we focus on the Gaussian PDF, solutions for other
distributions will be given in a forthcoming paper. With these assumptions the intensity can
be written
2 2 2 2
I qz F h exp q z L qz
1 2
1 exp q 2z 2
L
2
U 2r L U 2 exp qz 2
L
2
U 2r L U cos q z t
. (2) 2
The interface correlation coefficient (r = < ∆zL ∆zU>/(<∆zL2><∆zU2>)1/2) describes how the
substrate/film and the film/air interfaces are correlated; r = ± 1 for completely correlated (+1)
and anti-correlated (-1) interfaces, and r = 0 for uncorrelated interfaces.

(d)
(d)

(c)
Intensity
Intensity

(c)

(b)
(b)

(a (a)

-0.05 -0.025 0 0.025 0.05 -0.05 -0.025 0 0.025 0.05


-1 -1
qz (Å ) qz (Å )

Figure 1. (left) and figure 2 (right): influence of σL, σU and r on the diffraction profile of a 100 nm-
thick film calculated with equation (2) (left) and (3) (right). (a) σL, = σU = 0; (b) σL, = 0, σU = 5 nm;
(c) σL, = σU = 5 nm; (d) σL, = 5 nm, σU = 10 nm. Circles: r = 1; light grey line on left side of dashed
line: r = 0; dark grey line on right side: r = -1.

Figure 1 represents diffraction profiles calculated for a 100 nm-thick film, with interplanar
spacing d = 5 Å, for different values of the σ’s and r. Inspection of equation (2) and figure 1
reveals that σL affects the decay of the intensity when moving away from the Bragg position
qz = 0 (i.e. the intensity decreases faster that 1 / qz2), whereas σt (= σL2+σU2–2rσLσU) affects
Unauthenticated
Download Date | 12/3/17 5:20 PM
126 European Powder Diffraction Conference, EPDIC 9

the damping of the thickness fringes. The effect of the interface correlation coefficient is
most easily understood by examination of figure 1(c) for which σL = σU. For r = 1 (circles)
we have σt = 0, and the only effect is an accentuated decay of the diffracted intensity. For r =
0 (light grey line) σt = 2 σL, the intensity decreases at the same rate as for r = 0 but in
addition the oscillations are damped. Finally for r = -1 (dark grey line) σt = 2σL, the damping
is more pronounced. It appears that σL and σt affect the XRD profile in a very distinct
fashion, these parameters can hence be determined by fitting equation (2) to an experimental
profile. Moreover, if the surface roughness σU is known from a complementary experiment
(e.g. atomic force microscopy) then the interface correlation coefficient can also be
determined.
Different parts of the film scatter incoherently
We now investigate the case where different parts of the sample with different zU and t
scatter incoherently, which happens for instance when the film exhibits a mosaic structure. In
such a case, the quantity being averaged is the intensity, i.e. we have to compute <EE*>.
With the same assumptions as above this average can be calculated analytically:
2
Fh 1 2 2
I qz 2 2 exp q z L U2 2r L U cos qz t
2
qz 2
. (3)
Inspection of equation (3) reveals that the sensitivity to interface roughness (i.e. the term
exp(-qz2σL2)) is lost. This is easily understood: since different parts of the sample scatter
incoherently, all information concerning the beam path difference is therefore lost in the
diffracted intensity, i.e. the total intensity is the sum of the intensities scattered by different
parts of the sample with different thickness. Figure 2 shows calculated profiles with the same
parameters as for figure 1. Figure 2 indeed shows that the only effect altering the line shape
is a damping of the fringes, the rate of decay is unaffected. This effect particularly visible in
figure 2(c): for r = 1 and σU = σL, σt = 0 and the profile is a Laue function, whereas for
decreasing r the fringes are damped. In addition we observe an excess intensity in the profile
tails in figure 2 as compared to figure 1 (this is most easily evidenced in the profiles (b)).
This additional intensity correspond to the diffuse scattering cov(E,E*), since <E.E*> =
<E>.<E*> + cov(E,E*). It should be noticed that the same parameters (roughness and
thickness fluctuations) can be evaluated by X-ray reflectivity. However, contrarily to X-ray
reflectivity, XRD (when performed at non grazing incidence) is not a surface-sensitive
technique. Therefore the roughness values probed by X-ray reflectivity are in general in the
range of a few angstroms [12], whereas the values probed by XRD are about one order
magnitude larger (as in the present case), or even larger. In that respect these two techniques
are complementary. The applicability of the equations derived have been tested using
diffraction profiles recorded from thin films of ZrO2 doped with 10 mol% Y2O3 epitaxially
grown on sapphire with thickness ranging between 40 and 70 nm [18,19]. The results derived
from a least-squares fitting procedure (not shown here) are identical to those obtained using
numerical integration of equation (1).

Effects of lattice disorder


The effect of lattice disorder is in general more complicated to handle because this effect is
strongly dependent (in terms of magnitude and direction) on the nature of the defect
Unauthenticated
Download Date | 12/3/17 5:20 PM
Z. Kristallogr. Suppl. 23 (2006) 127

involved. In the presence of lattice disorder the position vector r in equation (1) should be
replaced with r+u(r), where u(r) is a displacement vector and is a function of the position
vector. In the general case there is no analytical solution, in particular when u(r) is a non
linear function of r, then equation (1) must be numerically evaluated. Analogously when the
film contains randomly distributed strain (εij = dui/dxj , i,j = 1,2,3) then the diffracted
intensity can be described as the convolution of a profile due to strain effects and a profile
due to size effects [20]. In some cases however an analytical solution can be obtained. In
particular in case of randomly distributed displacements, the overall intensity is reduced by
the so-called Debye-Waller factor exp(-h2σu2), where σu is the standard deviation of the
displacement PDF (assumed to be Gaussian). Finally, when the lattice planes are randomly
distorted, but the displacements uz accumulate from one plane to another [9,11], this effect
can be accounted for by substituting the reduced scattering vector in equation (2),with [13] qz
– ih2σu(c)2/2d where σu(c) is the standard deviation of the displacement PDF (assumed to be
Gaussian), the superscript (c) stands for cumulative.
1.6 σu = 0.2 Å
FWHM (Å-1) 1.2

(d) 0.8 σu = 0.15 Å


Intensity (a.u.)

0.4 σu = 0.1 Å
(c)
(b) 0 σu = 0.05 Å
0 10 20 30 40
Q2 (Å-2)
(a)

-0.05 -0.025 0 0.025 0.05


-1
qz (Å )

Figure 3. Influence of cumulative lattice disorder on the diffraction profile of a 100 nm-thick film. (a)
to (d): σu(c) = 0 ; 0.1 ; 0.15 and 0.2 Å. Inset: evolution of the full width at half maximum (FWHM) with
the squared scattering vector

The influence of the latter effect on the diffraction profiles is depicted in figure 3. This type
of disorder broadens the main peak as well as the thickness fringes. In contrast to
morphological effects, it is dependent on the order of reflection (since it is function of Qz,
and not qz). This type of disorder mainly has a phenomenological interest in that it produces
peak and fringes broadening similarly to what is encountered with several type of defects.
However, it is unlikely that the actual state of disorder in thin films can be described in such
a simple way. Therefore the values of σu(c) cannot be straightforwardly interpreted on a
quantitative level (in terms of defects densities for instance). If such interpretations have to
be performed, other approaches should be used (see for instance ref. [12]).

Conclusions
We have derived a generalized interference function that takes into account the effects of
interface roughness, surface roughness, interface correlations and cumulative and non
cumulative lattice disorder. This interference function is completely analytical and can be
written, in the case where different parts of the film scatter coherently:

Unauthenticated
Download Date | 12/3/17 5:20 PM
128 European Powder Diffraction Conference, EPDIC 9

2
2 c2
2 2 2 2 2 2
h u
I qz F h exp h u
q z L
q z
2d

2 2 2 c2 t 1 2 2 2 c2 t
1 exp qz t
h u
2 exp q h cos q z t
d 2 z t u
d
and σt2 = σL2 + σU2 – 2r σLσU . When different parts of the film scatter incoherently a similar
equation is straightforwardly obtained from equation (1) as previously detailed.

References
1. Authier, A., 2001, Dynamical theory of X-ray diffraction, IUCr monograph on
crystallography 11. (Oxford University Press).
2. Takagi, S., 1962, Acta Cryst., 15, 1311, 1969, J. Phys. Soc. Jap., 26, 5, 1239.
3. Taupin, D., 1964, Bull. Soc. Franç. Minér. Crist., 87, 469.
4. Bartels, W. J., Hornstra, J., W. Lobeek, D. J., 1986, Acta Cryst. A, 42, 539.
5. Halliwell, M. A. G., Lyons, M. H., Hill, M. J., 1984, J. Cryst. Growth, 68, 523.
6. Pavlov, K. M., Punegov, V. I., 2000, Acta Cryst. A, 56, 227.
7. Friedrich, W., Knipping, P., Laue, M., 1913, Annalen der Physik, 14, 971.
8. Miceli, P. F., Palmstrøm, C. J., Moyers, K. W., 1992, Appl. Phys. Lett., 61, 17,
2060.
9. Fullerton, E. F., Schuller, I. K., Vanderstraeten, H., Bruynseraede, Y., 1992, Phys.
Rev. B, 45, 16, 9292.
10. Gibaud, A., Cowley, R. A., McMorrow, D. F., Ward, R. C. C., Wells, M. R., 1993,
Phys. Rev. B, 48, 19, 14463.
11. Zolotoyabko, E., 1998, J. Appl. Cryst., 31, 241.
12. Holý, V., Pietsch, U., Baumbach, T., 1999, High resolution X –ray scattering from
multilayers and thin films, Springer tracts in Modern Physics 149 (Berlin:
Springer), p63.
13. Boulle, A., Guinebretière, R., Dauger, A., 2004, J. Phys. IV (France) 118, 183.
14. Andrews, S. R., Cowley, R. A., 1985, J. Phys. C, 18, 6427.
15. Robinson, I. K., 1986, Phys. Rev. B, 33, 6, 3830.
16. Harada, J., 1992, Acta Cryst. A, 48, 764.
17. Ern, C., Donner, W., Dosch, H., Adams, B., Nowikow, D., 2000, Phys. Rev. Lett.,
85, 9, 1926.
18. Boulle, A., Pradier, L., Masson, O., Guinebretière, R., Dauger, A., 2002, Appl. Surf.
Sci., 188, 80.
19. Boulle, A., Masson, O., Guinebretière, R., Dauger, A., 2003, Thin Sol.Films, 434, 1.
20. Warren, B. E., X-ray diffraction, 1969 (New-York: Addison-Wesley) p268.

Unauthenticated
Download Date | 12/3/17 5:20 PM

Potrebbero piacerti anche