499_1.pdf

Microstructure Analysis in As-deposited and Annealed
Damascene Cu Interconnects using OIM
Kabir Mirpuri, Jerzy Szpunar and Kris Kozaczek*
Department of Metals and Materials Engineering, McGill University,
3610 University Street, Montreal, Quebec, Canada H3A 2B2
*HyperNex Inc., 3006 Research Drive, State College, PA, 16801, USA
Abstract. Microstructure variation with linewidth was studied for both as-deposited and annealed damascene Cu
interconnect lines. Ten specimens with different linewidth and pitch distances were investigated using
Orientation Imaging Microscopy (OIM). The microstructures, examined both before and after annealing,
displayed high percentage of E3 boundaries. The microstructure was characterized by measuring the mean grain
size, grain size distribution, grain boundary misorientation distribution and coincident site lattice (CSL)
boundary distribution. The mean grain size increased proportionally until the linewidth was equal to line depth
and slowly stabilized thereafter. The role of linewidth to pitch distance ratio was identified on influencing some
of the microstructural features. The grain shape was analyzed using the grain aspect ratio parameter. The strain
distribution in the line was studied using image quality (IQ) parameter.
change occurring as a function of linewidth and also
microstructure evolution upon annealing.
INTRODUCTION
After the replacement of Al by Cu as an interconnect
material, much research has been done to understand
the behaviour of Cu in narrow trenches. There have
been many reports which detail the microstructure
evolution using texture as characterization tool but not
much data has been published with regard to the grain
size, shape and distribution especially for Cu
damascene lines. These parameters have an important
role in deciding the stress voiding and electromigration
behaviour of Cu. The investigation of these parameters
is even more significant taking into account the
metastable nature of Cu film upon electrodeposition
wherein the phenomenon of self-annealing is observed
at room temperature [1]. Hence, annealing is a very
important step in controlling the final texture and
microstructure of the Cu lines. As a result, it is of great
interest to study the microstructure in the Cu lines in
as-deposited condition itself and its evolution upon
annealing. The present paper describes the
investigations done on both as-deposited and annealed
Cu lines, which gives more clear illustration of overall
EXPERIMENTAL
Ten specimens with different linewidth and pitch
distances were investigated using OIM apparatus
attached with Philips XL-30 field emission gun
scanning electron microscope. The linewidth/pitch
distances were 0.35/0.35, 0.4/3.6, 0.5/0.5, 1/1, 3.6/0.4,
4/36, 10/10, 36/4, 40/360 and 100/100 |im. The
interconnects were deposited in an area spanning 2.5 x
1.75 mm2 and were tested in as-deposited condition
(however, over a period of time they may have
undergone recrystallization at room temperature) and
after annealing at 400°C for 30 min. Annealing was
performed after CMP and deposition of top passivation
layer. Measurements were done at 20 kV, specimen tilt
of 70° and working distance of 15 mm. The underlying
Cu lines were exposed by removing the top passivation
layer by etching in 15% HF for 5 minutes. The number
of grains investigated via EBSD varied from 500 to
32,000 for all the lines giving a good statistics of
CP683, Characterization and Metrology for VLSI Technology: 2003 International Conference,
edited by D. G. Seiler, A. C. Diebold, T. J. Shaffner, R. McDonald, S. Zollner, R. P. Khosla, and E. M. Secula
© 2003 American Institute of Physics 0-7354-0152-7/03/$20.00
499
(a)
collected data except 0.4 and 1 |im lines where only 75
and 270 grains could be examined.
RESULTS AND DISCUSSION
Figure 1 reveals the grayscale OIM maps for the
narrowest 0.35 |im lines in both as-deposited and
annealed condition. Only small portion of the total
scanned area has been displayed. Since the lines were
highly twinned, as we show later in this paper, the
maps were computed both before and after neglecting
the presence of the £3 boundaries.
The OIM maps distinguish each grain from its
neighbouring grains with a different grayscale color
and give a more clear view of the grain shape and
structure. As can be seen in figure 1, the grain structure
is almost bamboo shaped at many places in both asdeposited and annealed lines. Also, larger grains in the
annealed lines compared to as-deposited lines indicate
grain growth having occurred during annealing.
Figure 2 depicts the mean grain size as a function of
linewidth for both as-deposited and annealed states.
The hypothetical straight line in the graphs shows the
dimensions where the mean grain size equals the
linewidth. The mean grain size was calculated by two
methods - firstly, before neglecting the presence of £3
boundaries (considering twins as separate grains) and
secondly, after neglecting the £3 boundaries
(considering twins as part of grains). The mean grain
size increased proportionally with linewidth for the
narrower lines until the linewidth of 1 |im, which is
close to the trench depth of 1 jum. This is mainly due to
the influence of trench sidewalls and bamboo grain
structure in the narrower lines. Beyond 1 |im the mean
grain size slowly stabilizes due to the constraints
(a)
i*S mil
(b)
(c)
tun
0.1
1
10
100
Line-width in microns (log.)
(b)
C
8
.2
:
: — linewidth
1 ^10 :
.E 0>
o) o
:
:
'-0) ~ 1' -E
.E
<C
:
ri n 1
0.1
.„/""
° with twins
* w/o twins
A
D^ -^
A
^
,/-/
n
•
D
:
fr
n
*
D
/*?
1
10
100
Line-width in microns (log.)
FIGURE 2. Variation of mean grain size with linewidth in
(a) as-deposited and (b) annealed damascene Cu interconnect
lines before and after neglecting the presence of £3
boundaries
imposed by neighbouring grains. The same observation
was made by Jiang et al. [2].
An interesting observation that could be made from
figure 2 is that though the 3.6 and 4 |im lines have
almost the same width there is significant difference in
their mean grain size. Similar is the case for higher
linewidths of 36 and 40 Jim and narrower linewidths of
0.35 and 0.4 |j,m. However, these lines had different
pitch distances indicating the role of linewidth to pitch
distance ratio on influencing the mean grain size.
Lower grain size in the 0.4 |im lines may also be
attributed to the lower statistics of collected data since
only single line was examined.
The grain size distribution was computed for all the
specimens but has been represented for the narrowest
0.35 and one of the widest 10 |im lines for comparison
in figure 3 in both as-deposited and annealed condition,
considering twins as separate grains. The distribution is
log-normal. Also, the distribution, upon annealing,
shifts to the larger grain size which is relatively more
significant for higher linewidths. The absence of this
shift in the lower linewidths may be attributed to the
recrystallization that may have occurred in these lines
at room temperature due to lower pitch distance. The
recrystallization led to the increase in the grain size in
narrower lines in as-deposited condition. The role of
pitch distances in inducing early recrystallization has
(d)
tJ ilia
FIGURE 1. Grayscale OIM maps depicting 0.35 {4,m Cu
damascene interconnect lines in (a) as-deposited condition
(b) as-deposited condition after neglecting presence of £3 (c)
annealed condition (d) annealed condition after neglecting
presence of £3 CSL boundaries
500
(a)
(a)
Grain Size (diameter)
100
0.25
c
0.20
£ 0.15
u_
o.i
2 o.oi
| 0.10
u
i: I I I I I I I I I I I
Ii 11111111 11 Ii
o
0.05
0.00
0.01
0.1
1
" j &
10
Grain Size (Diameter) [microns]
— • —• •
o.35m as-depo.
Misorientation (degrees)
0.35m anneal with twins
(b)
Grain Size (diameter)
1
10
Misorientation (degrees)
100
Grain Size (Diameter) [microns]
—« — - -
10m as-depo. with twins
10m anneal with twins
(c)
100
FIGURE 3. Grain size distribution for as-deposited and
annealed (a) 0.35 and (b) 10 |iim damascene Cu interconnect
lines
G>
o
10
B As-depo.
• Anneal
o
been demonstrated by other authors [3].
Figure 4 reveals the grain boundary misorientation
distribution for the 0.35, 0.5 and 36 |im lines. In all the
cases we find the grain boundaries with misorientation
between 55 to 60° to be present in significant fraction
followed by the boundaries between 35 to 40°. The
former are the L3 boundaries between the grains,
which are rotated by 60° around the common <111>
axis. The latter are the Z9 boundaries between the
grains, which are rotated by 38.5° around the common
<110> axis. Also we find that the fraction of high angle
grain boundaries with misorienation between 15 to 45°,
which have high energy, has decreased upon annealing
for most of the specimens, which indicates that some
grain boundary migration has occurred during
annealing. This effect was not seen in 36 \\m lines
indicating the possible role of linewidth to pitch
distance ratio on affecting the grain boundary mobility.
Figure 5 compares the sum of the fraction of twin
and twin variant boundaries in as-deposited and
annealed lines. We find that the fraction increased
4=
O
CO
0.1
O'
Misorientation (degrees)
FIGURE 4. Grain boundary misorientation distribution for
as-deposited and annealed (a) 0.35 (b) 0.5 and (c) 36 Jim
damascene Cu interconnect lines
upon annealing in most of the lines except the 3.6 and
36 |im lines, which also had the smallest pitch distance
indicating the role of pitch distance in influencing the
CSL distribution. Increase in the fraction of CSL
boundaries upon annealing was also observed by Field
[4].
The fraction of CSL boundaries was computed for all
the lines (not shown here). Predominant presence of L3
boundaries was observed followed by £9, £27a, Z27b
and L7. Also as observed from figure 5 the fraction of
501
(C)
100 T——
80
c
.2
60
O
40
£
20
I
III! I
III
Mil
1111111
0.318
II,.2
CIS
04
0.5
III M i l l
03
CN
O.S
0.8
0,283
6,191
0,114
OJ323
Linewidth (microns)
FIGURE 5. Variation of the sum of the fraction of twin and
twin variant boundaries as a function of linewidth in asdeposited and annealed damascene Cu interconnect lines
I pill
FIGURE 6. (a) Defining grain aspect ratio and
measurements (b) before and (c) after neglecting the
presence of 23 CSL boundaries in 0.35 |im damascene Cu
inteconnect lines
CSL boundaries was higher in the narrower linewidths
compared to the higher linewidths which is a good sign
since the role of these special boundaries in providing
better electromigration resistance in the Al
interconnects has been demonstrated [5].
It has been shown that the grain boundary planes
normal to the trench sidewalls offer more resistance to
electromigration compared to case when they are
parallel [1]. In that aspect the measurement of grain
shape may give more insight in predicting the
electromigration reliability of the Cu lines. We
measured the grain aspect ratio in the lines assuming
the grains to be ellipsoids before and after neglecting
Z3. Figure 6 reveals the measurement scheme for the
narrowest 0.35 |im lines wherein the imaginary ellipses
have been superimposed on the grains.
Moreover the grains have been classified based on
their aspect ratio with grayscale colors as shown in the
legends besides the figures 6b and 6c. We define the
grain aspect ratio as the ratio of the minor axis of the
ellipse to its major axis.
Figure 1 shows the grain aspect ratio distribution in
both as-deposited and annealed lines before and after
neglecting £3.
(a)
Aspect ratio = b/a
(b)
Grain aspect ratio
(b)
Grain aspect ratio
502
(b)
c
o
32
01
, ?> ,
"V
tx
,
<b
•
A
.^.'
200
300
400
Image Quality
Grain aspect ratio
FIGURE 8. IQ plot for as-deposited and annealed (a) 0.5 and
(b) 10 Jim damascene Cu interconnect lines
FIGURE 7. Grain aspect ratio distribution for (a) 0.35 (b)
0.5 and (c) 36 juim as-deposited and annealed damascene Cu
interconnect lines before and after neglecting the presence of
L3 CSL boundaries
CONCLUSIONS
The microstructure has been investigated in asdeposited and annealed damascene Cu interconnect
lines. Microstructure was characterized by measuring
the mean grain size, grain size distribution, grain
boundary misorientation distribution, CSL boundary
distribution, grain aspect ratio distribution and IQ
plots. Additionally most of the parameters were
analyzed before and after neglecting the presence of £3
CSL boundaries. Some important observations were
noted and variations were observed upon annealing and
also with change in linewidth. Role of linewidth to
pitch distance ratio was identified on some of the
parameters.
One can see that the distribution is symmetric with
higher fraction of the grains having aspect ratio
between 0.4-0.6. The distribution shifts towards lower
values upon neglecting the presence of £3 boundaries,
which is more prominent for the narrower linewidths
due to elongation of the grains along the trench length
and increased influence of trench sidewalls.
The image quality (IQ) graphs were computed for all
the lines but have been represented for only 0.5 and 10
(Lim lines (Fig. 8) since most of the lines showed the
same behaviour. The IQ parameter is directly related to
the strain distribution in the specimen though it also
depends on other factors like orientation of the grains
and sample surface condition. The higher grayness of
IQ maps shows larger strains in the specimen (while
higher IQ value shows less strain).
The higher IQ for the as-deposited lines indicates
lesser strain distribution in these lines compared to
annealed lines. This could be attributed either to higher
stresses induced in the annealed lines due to different
coefficient of thermal expansion between the Cu and
surrounding barrier material on the chip or to improper
surface condition.
ACKNOWLEDGMENTS
One of the authors K. Mirpuri would like to
acknowledge the support of NSERC-IPS, AESF,
Joseph Stauffer and Horace G. Young graduate awards.
REFERENCES
1. Harper, J.M.E., Cabral, Jr. C., Andricacos, P.C., Gignac,
L., Noyan, I.C., Rodbell, K.P., and Hu, C.K,, J. Appl.
Phys. 86(5), 2516-2525 (1999).
2. Jiang, Q.T., Nowell, M., Foran, B., Frank A., Havemann,
R.H., Parihar, V., Augur, R.A., and Luttemer, J.D., J. Elec.
Mat. 31(1), 10-15, (2002).
3. Lingk, C., and Gross M.E., J. Appl Phys. 84(10), 5547 5554(1998).
4. Field, D.P., "Texture evolution in Cu damascene lines"
edited by J.A. Szpunar, Proceedings of the Twelfth
International Conference on Textures of Materials
(ICOTOM -12) 2, Montreal, Canada, (1999), pp. 13091314.
5. Lee, K.T., Szpunar, J.A., and Knorr, D.B., Mat. Sci. Forum
204-206,423-430(1996).
(a)
0.15
I0'1
o
IJL0.05
200
•
'
•
'. '.
-.. ».t
<
cy o- o- cy o- o- o- o
100
'As-depo.
• Anneal
•
•
0.05 -
100
<b
....
* ,
'.
n
V
•
300
400
Image Quality
503