You are on page 1of 5

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/264488034

Determining the diffusion coefficient of Ni in


Cu from a Ni/Cu thin film using linear heating,
scanning Auger microscopy and a numerical
solution of Fick's second law

Article in Surface and Interface Analysis · June 2010


DOI: 10.1002/sia.3377

CITATIONS READS

3 35

3 authors:

Heinrich Joubert Jacobus Terblans


Pannar Seed University of the Free State
8 PUBLICATIONS 18 CITATIONS 131 PUBLICATIONS 753 CITATIONS

SEE PROFILE SEE PROFILE

H. C. Swart
University of the Free State
549 PUBLICATIONS 3,398 CITATIONS

SEE PROFILE

Available from: H. C. Swart


Retrieved on: 08 September 2016
Research Article
Received: 27 August 2009 Revised: 18 February 2010 Accepted: 3 March 2010 Published online in Wiley Interscience: 11 May 2010

(www.interscience.wiley.com) DOI 10.1002/sia.3377

Determining the diffusion coefficient of Ni


in Cu from a Ni/Cu thin film using linear
heating, scanning Auger microscopy and a
numerical solution of Fick’s second law†
H. D. Joubert, J. J. Terblans∗ and H. C. Swart
In previous studies, linear temperature ramping was used to determine diffusion coefficients from bulk-to-surface segregation
experiments of a low concentration solute. Thin film diffusion studies usually employ a classical heating regime, where a
sample’s annealing time is taken as the time between insertion and removal from a furnace. The aforementioned study type
assumes that the time it takes to heat up a sample after insertion is instantaneous, while the sample cools down instantaneously
after removal from the furnace. This assumption is incorrect, as it does not compensate for the various mechanisms that govern
heat transfer. In order to eliminate the uncertainty, a linear ramping regime is used and samples were annealed inside an ultra
high vacuum (UHV) environment with a programmed linear heating scheme. After each anneal, a depth profile was obtained
by simultaneously bombarding the sample with Ar+ ions and monitoring the exposed surface with an electron beam which
excites Auger electrons, among others. The depth profiles were normalised and the time scale converted to depth. In order to
compare the diffusion profiles obtained from classical annealing studies to the linearly ramped studies, the diffusion coefficient
obtained for a classical study of Ni diffusing in Cu was inserted into a numerical solution of Fick’s second law, modified to
include linear temperature ramping. A diffusion coefficient was also obtained from linearly ramped samples. Comparison of the
diffusion profiles calculated with the diffusion coefficients obtained from classical heating and linear heating showed a large
discrepancy between the profiles. Copyright  c 2010 John Wiley & Sons, Ltd.

Keywords: diffusion; thin films; linear temperature ramping; Mixing–Roughness–Information model

Introduction ting three parameters to a measured diffusion profile:[4] (i) Atomic


mixing width, (ii) information depth and (iii) the surface/interface
Ni/Cu thin films are of great importance in the microelectronics roughness. The MRI model has proved its efficacy over the years
industry where thin film couples form an integral part in and has the added advantage of extracting diffusion coefficients
semiconductor chips.[1] Diffusion coefficients provide an insight (D) from experimental data using the relation[5]
into how much mixing between the elements of a thin film couple
takes place after a certain time and at an elevated temperature,[2] 2Dt = σ 2 (1)
similar to the conditions experienced by semiconductor chips
during its lifetime. The inherent problem with diffusion parameters where t is the annealing time and σ 2 is the change in the interface
available in literature for Ni/Cu thin films is the variation in the roughness between the reference sample and a particular sample
structure and composition of the thin film couple.[3] The difference subjected to annealing.
in microstructures results in diffusion parameters that have a very As mentioned previously, most diffusion studies rely on a
wide range of values, as indicated in Table 1. Of further concern constant, well-defined annealing time which ignores heat transfer
is the annealing methods used in certain studies of thin film mechanisms. The MRI model is no different in this respect. In
diffusion couples,[2] where the time of annealing is taken as the this study, a linear heating scheme for binary thin film diffusion
time between insertion of a sample into the furnace and removal couples described in Ref. [2] was expanded to bypass the effect
of the sample from the furnace (classical heating), ignoring the that heat transfer mechanisms have on classical heat treatments,
mechanisms that govern heat transfer.[2] by using a programmed heat treatment that involves linearly
Classical thin film diffusion studies rely on annealing of many ramping and cooling a sample under ultra high vacuum (UHV)
samples for different times and at different temperatures. Inter-
face broadening results from such a heat treatment and analysis of
this broadening yields the diffusion coefficient for the system be- ∗ Correspondence to: J. J. Terblans, Department of Physics, University of the
ing investigated. Superimposed on the diffusion-induced profile Free State, PO Box 339, ZA-9300 Bloemfontein, Republic of South Africa.
broadening is the broadening that results from the analysis tech- E-mail: TerblansJJ@ufs.ac.za
nique used to probe the sample, known as the depth resolution.[4]
One popular model that describes a depth resolution function † Paper published as part of the ECASIA 2009 special issue.
is the Mixing–Roughness–Information (MRI) model.[5] The MRI
1213

Department of Physics, University of the Free State, PO Box 339, ZA-9300


model attempts to describe the depth resolution function by fit- Bloemfontein, Republic of South Africa

Surf. Interface Anal. 2010, 42, 1213–1216 Copyright 


c 2010 John Wiley & Sons, Ltd.
H. D. Joubert, J. J. Terblans and H. C. Swart

Table 1. Diffusion parameters obtained in the present work and from


literature

References D0 (m2 s−1 ) Q (kJ mol−1 )

Present study 7.06 × 10−13 87.7


[1] 2.6 × 10−10 133.1
[3] 4.9 × 10−13 103.2
[6] 6.48 × 10−9±1 130.5 ± 13
[8] 2 × 10−5 228.7
[9] 6.93 × 10−7 90.4
[10] 1.4 × 10−14 106.1

Figure 2. Comparison of diffusion profiles calculated from linear ramping


and perfect classical heating.

heat treatment, then diffusion profiles obtained by the two heating


methods should also be the same.
The energy transferred to the ramped sample can be repre-
sented by the area underneath the curve shown in Fig. 1. If one
now uses a weighted temperature obtained from the area un-
der the curve (also shown in Fig. 1), a perfect classical diffusion
calculation can be performed with the weighted temperature by
simply changing the annealing time until the area underneath the
perfect classical curve matches the area underneath the ramped
Figure 1. Representation of the change in the diffusion coefficient during curve. The amount of energy transferred to a sample during a
a temperature ramp. perfect classical anneal will then equal the energy transferred to
the sample during a linear ramp anneal, and hence the diffusion
profiles will be the same for the two different annealing methods.
conditions. The diffusion coefficient was then obtained from This hypothesis was confirmed by performing a linear ramp
the ramped sample data, which is Ni/Cu in this study. The diffusion calculation and comparing it to a perfect classical
drawback in using temperature ramping for diffusion studies diffusion calculation performed with the weighted annealing
is that the diffusion coefficient is not constant during the heat temperature, with the perfect classical calculation accurately
treatment. One can numerically solve Fick’s second law[2] and following the linear ramp calculation, shown in Fig. 2. Using
incorporate linear ramping into the numerical calculation. The
the method described above, one can analyze ramped diffusion
constants of diffusion are determined from the calculations.
profiles with the well-established MRI model and extract diffusion
A recursive calculation scheme is required to obtain the pre-
parameters from the ramped data. A detailed description of the
exponential factor and activation energy, but this process is
MRI model is given in Ref. [7] and the method of determining the
a very time consuming method as the numerical solution of
diffusion parameters is given in Ref. [5].
Fick’s second law is a time intensive process. In order to speed
up the analysis of the linear ramped sample data, the profiles
were analyzed with the MRI model using a diffusion-coefficient-
weighted temperature to simulate a perfect classical annealing Experimental
scheme in which the temperature of the sample does increase
and decrease instantaneously after insertion and removal from a Physical vapor deposition was employed to deposit Cu/Ni thin
furnace. In other words, the linear ramp heat treatment equated films onto a passivated silicon (100) substrate. The nickel film had
to a perfect classical heat treatment. a thickness of 1000 Å, whereas the copper film had a thickness
The perfect classical annealing scheme is the preferred method of 6000 Å. Evaporation was performed in a vacuum with a base
of analysis as the classical annealing approach employs a direct pressure of <10−6 Torr.
error function solution of Fick’s second law. The difficulty in Auger electron spectroscopy combined with Ar+ ion bombard-
reconciling the two different methods arises from the varying ment was used to obtain depth profiles of the Ni/Cu samples. The
diffusion coefficient that is found when ramping the temperature Ar+ ion beam was set at 2 keV Ar++ , whereas the primary elec-
of a sample. An example of the varying diffusion coefficient is tron beam was set at 5 keV. Linear ramping was performed with a
shown in Fig. 1, which was constructed by calculating the diffusion custom heater inside the Auger apparatus. A Chromel/Alumel ther-
coefficient at each point during the temperature ramp using the mocouple measured the temperature at the back of the sample
diffusion parameters listed in Ref. [6]. while the heater’s temperature was computer controlled. Samples
If one assumes that the energy transferred to a sample in a were heated at 0.3 ◦ C s−1 and cooled at −0.1 ◦ C s−1 to maximum
1214

perfect classical heat treatment is the same as in a linearly ramped temperatures ranging from 350 to 500 ◦ C.

www.interscience.wiley.com/journal/sia Copyright 
c 2010 John Wiley & Sons, Ltd. Surf. Interface Anal. 2010, 42, 1213–1216
Diffusion coefficient of Ni in Cu from a linearly ramped Ni/Cu thin film

Figure 5. Arrhenius graph constructed from the MRI fit parameters.


Figure 3. Example of four linear ramp temperature profiles used in the Temperature values are the diffusion weighted temperatures.
study.

Figure 4. Experimental data and MRI fit for sample linearly ramped to
a maximum temperature of 400 ◦ C at a heating rate of 0.3 ◦ C s−1 and
a cooling rate of −0.1 ◦ C s−1 . The MRI parameters are also shown in Figure 6. Arrhenius plot of the diffusion parameters listed in Table 1.
the figure, with λ the information depth, w the mixing width and σ the
surface/interface roughness.
diffusion profile calculated with classical and linear ramp diffusion
parameters is shown in Fig. 7 to illustrate this point. The diffusion
Results and Discussion profile calculated using classically obtained diffusion parameters
predicts more diffusion than the linearly ramped profile. This
Four linear temperature ramp profiles used in this study are shown correlates well with the results obtained in Ref. [2], highlighting
in Fig. 3 (375 ◦ C was omitted for clarity). The effects of the heat the importance of controlling the heat treatment that is used
treatments on the diffusion profile of a Ni/Cu thin film are shown during diffusion studies.
in Fig. 4 where the sample was linearly ramped to a maximum of
400 ◦ C, repeated five times. For clarity, only the reference profile
along with profiles measured after two and five ramps are shown Conclusion
in Fig. 4. From the MRI fit parameters obtained from equation (1),
an Arrhenius graph was constructed and is shown in Figure 5. The effect of annealing method on the diffusion parameters
From the regression line the diffusion parameters were obtained extracted from thin film diffusion experiments was analyzed.
and are listed in Table 1. To aid the comparison of the diffusion Several samples were subjected to a controlled heat treatment
parameters obtained in this work with the parameters of other that involved linear heating and cooling in an UHV environment.
authors, an Arrhenius plot is constructed from the data in Table 1 To analyze the samples that underwent linear ramping, a perfect
and is shown in Fig. 6. classical heat treatment was proposed. The MRI model, in
The diffusion parameters obtained in this study are significantly conjunction with the proposed perfect classical heating method,
smaller than the diffusion parameters obtained by applying the was used to determine the diffusion parameters of Ni in Cu.
MRI model to a classical heat treatment. This implies that the These parameters were smaller than the parameters obtained
classical analysis employed in Ref. [6] yielded diffusion parameters in another study conducted by the authors which employed
1215

that overestimated the rate of diffusion. A comparison between a classical annealing of Ni/Cu thin films. The results indicate

Surf. Interface Anal. 2010, 42, 1213–1216 Copyright 


c 2010 John Wiley & Sons, Ltd. www.interscience.wiley.com/journal/sia
H. D. Joubert, J. J. Terblans and H. C. Swart

Acknowledgements
The authors would like to thank the National Research Foundation
for financial assistance for this study and the staff of the Physics
department of the University of the Free State for many informative
discussions. Special thanks to Mr Adriaan Hugo for his tireless
maintenance of the experimental apparatus.

References
[1] B. C. Johnson, C. L. Bauer, A. G. Jordan, J. Appl. Phys. 1986, 59(4),
1147.
[2] H. D. Joubert, J. J. Terblans, H. C. Swart, Nucl. Instrum. Methods B
2009, 267, 2575.
[3] N. N. Rammo, M. N. Makadsi, A. M. Abdul-Lettif, Phys. State Solids (A)
2004, 201(14), 3102. No.
[4] S. Hofmann, Thin Solid Films 2001, 398–399, 336.
Figure 7. Comparison of the diffusion profiles calculated with diffusion [5] J. Y. Wang, A. Zalar, Y. H. Zhao, E. J. Mittemeijer, Thin Solid Films
parameters obtained from a classical heat treatment and a linear ramp 2003, 433, 92.
heat treatment. [6] H. D. Joubert, J. J. Terblans, H. C. Swart, Proceedings of the 2009
European Conference on Surface and Interface Analysis, Antalya,
Turkey, to be published.
[7] S. Hoffman, Surf. Interface Anal. 1994, 21, 673.
an overestimation of diffusion when using the classical heat [8] T. Vandijk, E. J. Mittemeijer, Thin Solid Films 1977, 41, 173.
treatment. Since the preparation conditions of the samples [9] S. Divinski, J. Ribbe, G. Schmitz, C. Herzig, ActaMater. 2007, 51, 3337.
analyzed (linearly ramped and classically annealed) are the same, [10] R. Venos, W. Pamler, H. Hoffmann, Thin Solid Films 1988, 162, 155.
only the annealing method will result in different diffusion
parameters being extracted from the measured depth profiles.
One can therefore conclude that the method of annealing plays an
important part when attempting to extract diffusion parameters.
1216

www.interscience.wiley.com/journal/sia Copyright 
c 2010 John Wiley & Sons, Ltd. Surf. Interface Anal. 2010, 42, 1213–1216

You might also like