GIXSW For Thin Films
GIXSW For Thin Films
net/publication/236011744
CITATIONS READS
32 64
5 authors, including:
Some of the authors of this publication are also working on these related projects:
Application of the PAMONO-Sensor for detection and characterization of individual biological nano-vesicles. View project
All content following this page was uploaded by Alex von Bohlen on 17 October 2022.
1136 | J. Anal. At. Spectrom., 2006, 21, 1136–1142 This journal is c The Royal Society of Chemistry 2006
reflecting interfaces and so-called marker atoms emitting
detectable fluorescence radiation. Distances in a GI-XSW
intensity pattern are in the range of nanometres and permit
characterization of structures of this order of magnitude.
An XSW measurement procedure and a computer simula-
tion program developed to calculate the XSW wavefield will be
described in detail below.
Theoretical background
Beams of monochromatic and sufficiently longitudinally co-
herent X-ray radiation impinging onto and reflected at a sharp
interface can interfere at their intersection point. Depending
on the position of these intersection points relative to the Fig. 1 Calculated XSW field above a Si substrate for E = 10 keV.
mirror, constructive or destructive interference will occur,
leading to a periodic pattern of nodes and antinodes of the critical angle. They average around 1 because less radiation
X-ray standing waves above the interface. Positions of minima is reflected at the interface leading to less intensity in the
and maxima only depend on angle of incidence and energy interference with the incident beam.
(wavelength) of the X-ray radiation. Marker atoms located at Fig. 2 displays the calculated XSW field for a 70 nm Si layer
the position of a maximum will emit fluorescence radiation on a thick Ge substrate. The X-ray energy is again 10 keV.
detectable by an energy dispersive detector. Positions of nodes Similar to the pure Si substrate no radiation penetrates the
and antinodes vary with varying angle of incidence and/or sample for angles below the critical angle of Si. Above this
wavelength of the radiation,28 thus changing the angle (as it angle and below the critical angle for Ge (i.e. 0.241) radiation
was done in measurements presented here) or energy of the in- enters the Si layer. However, it is totally reflected at the Ge
coming beam results in a measurement of fluorescence versus substrate. Partial internal reflection at the Si–vacuum interface
angle or wavelength. leads to multiple inner reflections inside the Si layer. XSW field
De Boer29 and Klockenkämper30 have outlined a method to intensities far above 4I0 are produced due to coherent multi-
calculate the XSW field above a surface or within a multilayer, reflections. Above the critical angle of Ge, radiation can
utilizing formalisms introduced by Abelès31 and Parrat.32 This penetrate the substrate, too, and the XSW field resembles
method has been expanded and implemented in a computer the high angle region in Fig. 1 again.
program that will be introduced in the following. Detailed Typically an XSW scan is performed by changing the angle
calculation procedures are presented elsewhere.1 of incidence by tilting the sample with the marker position
To calculate the XSW field in a multilayer system the angle kept constant. It can be extracted from the 3-dimensional
of incidence for each subsequent layer must be determined as a graphs in Fig. 1 or Fig. 2 by plotting the cross-section of
function of wavelength and refractive index of the layer intensity versus angle at a fixed height value. If the marker is
material. Further, absorption inside each layer has to be distributed throughout a layer of finite thickness integration of
considered. Dispersion and absorption coefficients for all the whole layer has to be performed. This procedure and other
elements and many compounds are tabulated or can be possibilities of XSW scans are described in more detail else-
calculated.33,34 Then interference of the electrical fields of where.1
reflected and incoming beams in each layer are calculated
under consideration of path length and phase for each beam.
Squaring of the electrical field amplitudes delivers the field Experimental procedure
intensity. Marker atoms can probe the XSW field as they emit XSW measurements presented here were performed at beam-
fluorescence radiation depending on their position inside the line BL 9 of the DELTA synchrotron facility35,36 in Dortmund
XSW field. Elements with absorption edge (i.e. binding en-
ergy) below the energy of the incident beam can serve as
marker elements. Finally, taking into account absorption of
emitted fluorescence radiation on its way to the detector one
obtains the theoretical XSW signal.
In the following, the discussion of two examples will de-
monstrate the methodology. Fig. 1 shows the calculation of an
XSW field above a flat silicon substrate for X-ray energy of
10 keV. Below the critical angle of Si (i.e. 0.181) virtually no
radiation penetrates the sample and interference of totally
reflected and incoming beam lead to maxima of up to four
times the intensity of the incident radiation I0. Above the
critical angle, the field enters the substrate, resulting in a field
intensity decreasing exponentially with depth. Oscillation am- Fig. 2 Calculated XSW field for a 70 nm Si layer on a Ge substrate
plitudes above the sample reduce to smaller values than below for E = 10 keV.
This journal is c The Royal Society of Chemistry 2006 J. Anal. At. Spectrom., 2006, 21, 1136–1142 | 1137
Fig. 3 Schematic set-up for XSW and reflectivity scans. Detailed
explanations are given in the text.
(http://www.delta.uni-dortmund.de/) with energies of the in- Fig. 4 XSW scans and fits for Ge layers on Si for different thicknesses
cident beam between 13 keV (wavelength 0.0954 nm) and (from top to bottom): 1010 nm, 309 nm, 76 nm, 29 nm. The curves are
15.2 keV (0.0816 nm) depending on the element utilized as a discussed in more detail in the text.
marker. Synchrotron radiation was generated by a super-
conducting asymmetric wiggler (SAW) and energy was mono-
each thickness, the measured signal, a simulation fit for the
chromatized using a silicon (311) double crystal
nominal thickness and a best fit with varied thickness is shown.
monochromator. Longitudinal coherence length ranged from
While the simulated curve for the thickness generally deviates
3.5 mm (15.2 keV) to 5.0 mm (13.0 keV). Detailed description of
from the other curves towards higher energies (with the
the beamline set-up can be found in Krywka et al.37
exception of the thinnest Ge layer where the best fit leads to
The experimental set-up is displayed in Fig. 3. The beam
the nominal thickness), best fit and measurement curves can be
was bent down and focused by a curved silicon mirror. The
distinguished by the noise affected appearance of the measure-
beam cross section (and thus the scanning region on the
ment curves. The thicknesses obtained by a best fit to the data
sample) was then defined by two slit systems in front of the
shown here are (from top to bottom): 93 nm, 106 nm, 59 nm
sample to a width of 1 mm and a height of 150 mm. If necessary
and 29 nm. It can be seen that deviations between the
(to prevent damage to the detector) the beam intensity could
measured and simulated curve for the nominal thickness get
be reduced by an absorber system. The angle of incidence on
very large for thicknesses above approximately 100 nm.
the sample was selected with an accuracy of 0.0021 by tilting
XSW oscillations are clearly visible for the thinnest film
the sample stage. Fluorescence radiation from the sample was
(29 nm) but virtually disappear for thicker layers because of
detected by an energy dispersive detector (XR-100CR, Amp-
high absorption inside germanium. Nevertheless, the shape of
tek, Bedford, MA, USA) 1 cm perpendicular above the sample
the XSW curve is still dependent on the thickness. A very good
and processed by a multi-channel analyzer (MCA).
accordance of measured and calculated curves (accuracy
To perform an XSW scan, first an energy spectrum of the
about 2 nm) could be achieved for the 29 nm sample. Only
total fluorescence radiation at one angle of incidence is
for greater angles the measured signal falls below the calcu-
recorded and fluorescence intensity peaks of the marker ele-
lated curve. This effect is even stronger for the 76 nm sample.
ment(s) of interest are located. Then regions of interest (ROIs)
Here the best fit turns out for a theoretical thickness of 57–
are allocated to channels comprising the peaks. Finally, an
60 nm. Dramatic deviations occur for the thicker samples. The
angle scan from 0.011 to approximately 0.61 is performed
theoretical intensity value approximately rises linearly for
measuring the integrated intensity of each ROI versus the
great angles as expected,30 while the experimental value con-
angle of incidence.
verges towards a horizontal asymptote. If the thickness in
With the same set-up a reflectivity scan is performed using a
simulation is varied widely enough to fit the measurements it
scintillation counter approximately 1 m behind the sample at
converges to a value of 102–110 nm for the 309 nm sample and
specular position (angle of incoming beam = angle of reflected
90–96 nm for the 1010 nm sample, respectively. Obviously
beam) that is defined by another two slit system. This scan
thickness determinations cannot be performed with XSW at
provides supplementary information on the sample and at the
this energy for layers thicker than approximately 100 nm.
same time serves as a cross-check.
This can be understood regarding Fig. 5. The calculated
XSW field inside a 100 nm Ge layer on Si for E = 15.2 keV (cf.
Analyzed samples and results Fig. 5a) shows clearly visible oscillations. The theoretical
determination of the angle resolved fluorescence intensity
Layer thickness determination
requires integration over the entire 100 nm thickness of the
Germanium layers on silicon (JENOPTIC, Jena, Germany) of Ge layer. Thus oscillations disappear in this averaging process
four different thicknesses (29 nm, 76 nm, 309 nm and 1010 nm) (cf. Fig. 5b). Consequently, the nanometre sensitivity of XSW
were characterized with XSW scans at 15.2 keV. Fig. 4 shows is lost, especially if the signal is additionally corrupted by noise
measured and calculated curves for all analyzed samples. For and other errors. Obviously this method is most powerful if
1138 | J. Anal. At. Spectrom., 2006, 21, 1136–1142 This journal is c The Royal Society of Chemistry 2006
Fig. 6 XSW scan and fit for a polystyrene layer on Si/SiO2 sample
covered with Au clusters. Given values for the sample were 80 nm
polystyrene thickness, 1 nm nominal gold thickness and silicon with a
native oxide layer of unknown thickness. The fit curve was calculated
for 1 nm gold on 80 nm polystyrene on 1 nm SiO2 on silicon.
This journal is c The Royal Society of Chemistry 2006 J. Anal. At. Spectrom., 2006, 21, 1136–1142 | 1139
Table 1 Calculated and measured layer thicknesses inside a multi-
layer laser mirror of alternating layers of TiO2 and a second material
(MgO, SiO2 or LiF)
Fig. 7 Oscillating XSW signals for a phospholipid mono- or double- could be detected. However, together with the angle infor-
layer covered with KCl buffer solution. From top to bottom: Cl, K, P. mation—that is complementary to the element information
Curves are shifted vertically for clarity. Oscillation periods are related obtained with TXRF—other materials can be characterized
to wavelength of the radiation and layer thickness by Dy = l/2d. indirectly as will be shown below.
Fitting of the measurement shown in Fig. 8 with the XSW
simulation program turned out to be too complicated because
the layout of the sample (i.e. materials and number of layers)
scattering from the buffer solution made measurements diffi-
was not known. To extract information from the strongly
cult, but a periodic oscillation could be recorded for all
noise-affected measurement a computer program was devel-
elements. Knowing the wavelength of the incident radiation
oped to calculate periodicities in the signal using fast Fourier
l = 0.0954 nm (E = 13.0 keV) one can calculate a typical
transform algorithms.40–44 Transforming the frequencies of
distance corresponding to the oscillation wavelengths that
interest back to the angle space leads to an oscillation wave-
amount to 8–11 nm. The reliability is best for the P scan
length spectrum shown in Fig. 9.
because for K and Cl the period could not be determined as
From these angular periods and the wavelength of the
precisely. With a typical phospholipid molecule chain length
synchrotron radiation used one can deduce the typical (multi)-
of 3 nm measurements suggest that the lipids form double or
layer thicknesses dexp,ges listed in the last column of Table 1. To
triple layers. However, the formation of a monolayer can be
obtain qualitative and quantitative information on the mirror
excluded. Alternatively, the measurements could suggest an
one has to assume a sample structure. Typically, a laser mirror
amorphous phospholipid layer of 8–11 nm thickness.
consists of several double layers of two materials. As TXRF
measurements have proven Ti in the sample, titanium dioxide
Multilayers (TiO2) is taken as one component; the second one must be a
A commercial multilayer laser mirror (Spectra-Physics, Moun- light element or compound, e.g. MgO, SiO2 or LiF. Further-
tain View, USA) was analyzed using TXRF and XSW tech- more, each layer thickness must be 1/4 of the centre laser
niques. TXRF showed Ti content in the mirror. However, wavelength (i.e. 570 nm) for the mirror used divided by the
light elements like C, Li, Si or Mg that are possibly incorpo- known index of refraction n of the respective compound.45
rated in different layers were not accessible by this technique. Table 1 lists the added thickness values of a periodic stack of
In XSW measurement (cf. Fig. 8) again only the Ti signal alternating TiO2 (52.8 nm) plus MgO (81.9 nm), SiO2 (92.1
nm) or LiF (102.2 nm). It can be seen that theoretical and
experimental values agree very well (only) for MgO as the
second component together with TiO2. The table shows the
summarized thickness of the sample region that a beam passes
through if it is reflected at the first, second to the seventh
Fig. 8 XSW scan of a multilayer laser mirror, scanning of the Ti Fig. 9 Oscillation wavelength spectrum obtained from FFT of XSW
fluorescence signal. scans on a multilayer laser mirror.
1140 | J. Anal. At. Spectrom., 2006, 21, 1136–1142 This journal is c The Royal Society of Chemistry 2006
Finally, a complicated periodic multilayer structure—in this
case a laser mirror—can be analyzed qualitatively and quanti-
tatively. Combined with a fast Fourier transform algorithm
and general information about laser multilayer mirrors, the
advantages of XSW (element sensitivity, periodicity and nm
resolution) can be exploited.
Limits of XSW methods have come visible during this work,
too. Very thick layers (e.g. Ge samples) or strongly scattering
materials (like buffer solution in biological samples) can cause
problems. Some might be reduced by changing experimental
parameters like the wavelength of photon flux or by sample
preparation. A marker positioned at a characteristic location
in the sample is essential for XSW measurements. Thus,
samples not fulfilling these conditions cannot be analyzed
until a marker is added to the system. Further, generating a
theoretical curve fitting to the measurements gets increasingly
difficult with the number of unknown parameters. On the
Fig. 10 Correlation of measured (ordinate) and calculated (abscissa)
other hand, well known system structures can be analyzed
layer thicknesses in the multilayer mirror.
quickly, easily and accurately if only one or few parameters
have to be adjusted.
In the future other scanning variables can be utilized (e.g.
interface. The very first layer is not visible because its small
varying the energy of the incoming beam with a monochro-
thickness corresponds to a long oscillation wavelength located
mator), which open new approaches and possibilities.
inside the very broad maximum beyond the great angle limit in
Due to the need for high photon fluxes and the tunability of
Fig. 9.
the energy, XSW is a typical synchrotron technique. None-
Fig. 10 displays the correlation of the expected and mea-
theless, XSW has been proven to be a versatile technique
sured thicknesses inside the multilayer mirror of TiO2 and
which can be employed in a wide variety of problems which
MgO. The coincidence of the values is obvious. Thus, XSW-
otherwise are difficult to tackle.
method has been proven to be capable of determining com-
position and thicknesses of an unknown multilayer sample in
just one scan. Acknowledgements
Conclusions The financial support by the Ministerium für Innovation,
Wissenschaft, Forschung und Technologie des Landes
It has been shown that X-ray standing waves are a versatile Nordrhein-Westfalen and by the Bundesministerium für Bil-
and powerful tool to analyze layered structures on a nano- dung und Forschung is gratefully acknowledged. We would
metre scale, both concerning distances/layer thicknesses and also like to thank Simone Streit, Experimental Physics I,
element determination and distribution. A necessary require- University of Dortmund, for providing the gold polymer
ment for these kinds of analysis is a flat, smooth surface; sample together with helpful data, and Jessica Irrgang, Biolo-
however, different types of sample ranging from liquid to gical–Chemical Microstructure Technology, University of
solid, non-conducting to conducting, monolayer to periodic Dortmund, for preparation of and information about the
multilayer systems can be analyzed. The versatility of the phospholipid layers. The authors thank the DELTA machine
method has been demonstrated with 4 different types of group for providing the synchrotron radiation and technical
samples. support. C. S. would like to acknowledge M. Tolan for
Germanium layer thicknesses on a silicon substrate could be supporting this work.
measured in a range below 100 nm. Increasing thickness of the
layer corrupted the signal more and more. However, as has
been shown, the accessible thickness range can be expanded by References
moving to longer wavelengths. 1 M. Krämer, A. v. Bohlen, C. Sternemann and R. Hergenröder,
A sample of gold clusters on a polystyrene film was char- Appl. Surf. Sci., 2006, DOI: 10.1016/j.apsusc.2006.07.076.
2 R. Weber, Dissertation, Christian-Albrechts-Universität, Kiel,
acterized qualitatively and quantitatively. Incorporating heavy 2002.
marker atoms, a weakly absorbing matrix and dimensions in 3 C. P. Wong, Polymers for Electronic and Photonic Applications,
the nm range, XSW measurements appear to be ideal for this Academic Press, Boston, 1993.
technologically important kind of sample. 4 R. Klockenkamper, H. W. Becker, H. Bubert, H. Jenett and A. von
Bohlen, Spectrochim. Acta, Part B, 2002, 57, 1593–1599.
Only few methods are sensitive to low contrast interfaces 5 M. Krämer, Diploma thesis, University of Dortmund, Dortmund,
especially in biological samples that are of increasing impor- 2003.
tance in life science. As has been shown, XSW can deliver—in 6 J. Stettner, Dissertation, Christian-Albrechts-Universität, Kiel,
certain limits—even quantitative thickness information using Germany, 1995.
7 M. Tolan, X-Ray Scattering from Soft Matter Thin Films –
marker elements already incorporated in the materials under Materials Science and Basic Research, Springer-Verlag, Berlin,
investigation. 1999.
This journal is c The Royal Society of Chemistry 2006 J. Anal. At. Spectrom., 2006, 21, 1136–1142 | 1141
8 M. Caffrey and J. Wang, Annu. Rev. Biophys. Biomol. Struct., 23 B. P. Tinkham, D. M. Goodner, D. A. Walko and M. J. Bedzyk,
1995, 24, 351–377. Phys. Rev. B: Condens. Matter Mater. Phys., 2003, 6703, 5404
9 L. W. Cheng, P. Fenter, N. C. Sturchio, Z. Zhong and M. J. 5404.
Bedzyk, Geochim. Cosmochim. Acta, 1999, 63, 3153–3157. 24 T. P. Trainor, A. S. Templeton, G. E. Brown and G. A. Parks,
10 V. R. Dhanak, A. G. Shard, S. D’Addato and A. Santoni, Chem. Langmuir, 2002, 18, 5782–5791.
Phys. Lett., 1999, 306, 341–344. 25 T. P. Trainor, A. S. Templeton and P. J. Eng, J. Electron Spectrosc.
11 M. Drakopoulos, J. Zegenhagen, T. L. Lee, A. Snigirev, I. Relat. Phenom., 2006, 150, 66–85.
Snigireva, V. Cimalla and O. Ambacher, J. Phys. D: Appl. Phys., 26 J. Wang, M. Caffrey, M. J. Bedzyk and T. L. Penner, J. Phys.
2003, 36, A214–A216. Chem., 1994, 98, 10957–10968.
12 M. Drakopoulos, J. Zegenhagen, A. Snigirev, I. Snigireva, M. 27 J. Wang, M. Caffrey, M. J. Bedzyk and T. L. Penner, Langmuir,
Hauser, K. Eberl, V. Aristov, L. Shabelnikov and V. Yunkin, Appl. 2001, 17, 3671–3681.
Phys. Lett., 2002, 81, 2279–2281. 28 R. T. Zhang, R. Itri and M. Caffrey, Biophys. J., 1998, 74, 1924–
13 S. K. Ghose and B. N. Dev, Phys. Rev. B: Condens. Matter Mater. 1936.
Phys., 2001, 6324, 5409. 29 D. K. G. de Boer, Phys. Rev. B: Condens. Matter, 1991, 44, 498–
14 T. Jach, Y. Zhang, R. Colella, M. de Boissieu, M. Boudard, A. I. 511.
Goldman, T. A. Lograsso, D. W. Delaney and S. Kycia, Phys. Rev. 30 R. Klockenkämper, Total-Reflection X-ray Fluorescence Analysis,
Lett., 1999, 82, 2904–2907. John Wiley & Sons, New York, 1996.
15 R. G. Jones, A. S. Y. Chan, M. G. Roper, M. P. Skegg, I. G. 31 F. Abelès, Ann. Phys., 1950, 12, 596–640, 706–782.
Shuttleworth, C. J. Fisher, G. J. Jackson, J. J. Lee, D. P. Woodruff, 32 L. G. Parrat, Phys. Rev., 1954, 95, 359–369.
N. K. Singh and B. C. C. Cowie, J. Phys.: Condens. Matter, 2002, 33 CXRO, http://www.cxro.lbl.gov/optical_constants/.
14, 4059–4074. 34 B. L. Henke, E. M. Gullikson and J. C. Davis, At. Data Nucl. Data
16 T. L. Lee and M. J. Bedzyk, Phys. Rev. B: Condens. Matter Mater. Tables, 1993, 54, 181–342.
Phys., 1998, 57, R15056–R15059. 35 Dortmunder Elektronen Speicherring Anlage, http://www.delta.
17 A. Lessmann, M. Schuster, H. Riechert, S. Brennan, A. Mun- uni-dortmund.de/.
kholm and G. Materlik, J. Phys. D: Appl. Phys., 1999, 32, A65– 36 M. Tolan, T. Weis, K. Wille and C. Westphal, Synchrotron Radiat.
A70. News, 2003, 16, 9–11.
18 D. L. Marasco, A. Kazimirov, M. J. Bedzyk, T. L. Lee, S. K. 37 C. Krywka, M. Paulus, C. Sternemann, M. Volmer, A. Remhof,
Streiffer, O. Auciello and G. R. Bai, Appl. Phys. Lett., 2001, 79, G. Nowak, A. Nefedov, B. Pöter, M. Spiegel and M. Tolan,
515–517. J. Synchrotron Radiat., 2006, 13, 8–13.
19 S. Nakatani, K. Sumitani, A. Nojima, T. Takahashi, K. Hirano, S. 38 S. Streit, private communication.
Koh, T. Irisawa and Y. Shiraki, Jpn. J. Appl. Phys., Part 1, 2003, 39 K. B. Blodgett and I. Langmuir, Phys. Rev., 1937, 51, 964–982.
42, 7050–7052. 40 J. W. Cooley and J. W. Tukey, Math. Comput., 1965, 19,
20 G. Scherb, A. Kazimirov, J. Zegenhagen, T. L. Lee, M. J. Bedzyk, 297–301.
H. Noguchi and K. Uosaki, Phys. Rev. B: Condens. Matter Mater. 41 P. Duhamel and M. Vetterli, Signal Process., 1990, 19, 259–299.
Phys., 1998, 58, 10800–10805. 42 M. Frigo and S. G. Johnson, Proceedings of International
21 F. Schreiber, K. A. Ritley, I. A. Vartanyants, H. Dosch, J. Conference on Acoustics Speech, and Signal Processing, 1998,
Zegenhagen and B. C. C. Cowie, http://www.esrf.fr/info/science/ 1381–1384.
highlights/2001/surfaces/SURF1.html. 43 A. V. Oppenheim and R. W. Schafer, Discrete-Time Signal Proces-
22 F. Schreiber, K. A. Ritley, I. A. Vartanyants, H. Dosch, sing, Prentice-Hall, Upper Saddle River, NJ, 1989, 611,619.
J. Zegenhagen and B. C. C. Cowie, Surf. Sci. Lett., 2001, 486, 44 C. M. Rader, Proc. IEEE, 1968.
L519. 45 LOT-Oriel Group Europe, http://www.lot-oriel.com.
1142 | J. Anal. At. Spectrom., 2006, 21, 1136–1142 This journal is c The Royal Society of Chemistry 2006