Texture and stress analysis in as-deposited and annealed
damascene Cu interconnects using XRD & 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. Texture variations were investigated as a function of linewidth in as-deposited and annealed damascene Cu
interconnect lines using x-ray diffraction (XRD) and orientation imaging microscopy (OIM). Texture was predominantly
{111} fiber in narrower lines in as-deposited condition with some contribution from {111}<110> component in narrower
linewidths and {111 }<112> component in the higher linewidths. Texture became sharper with decreasing linewidth in case
of as-deposited specimens and became more random upon annealing. The decrease in texture strength was attributed to the
increased twinning upon annealing. The role of linewidth to pitch distance ratio on influencing the texture strength was
identified. Electron back scatter diffraction (EBSD) investigations revealed variations in texture of the local grain
population. Presence of sidewall {111} component was identified in the 0.4 and 0.5 urn lines in as-deposited condition and
persisted upon annealing but was absent in the 0.35 um lines. Presence of {110} grains parallel to the surface became more
dominant upon annealing in the narrowest 0.35 um lines with a sidewall {100} component. Residual stresses measured in
the lines were tensile in all the three principal directions with very low values in the normal direction for higher linewidths.
INTRODUCTION
EXPERIMENTAL
After the replacement of Al by Cu as an interconnect
material in the modern 1C chip, lot of research has
undergone investigating reliability of Cu metallization.
Most of the studies have been carried out with respect
to the texture and microstructure of Cu since these two
parameters are indispensable in deciding the reliability
of Cu interconnects. Apart from this scenario, the new
damascene process which was developed to make the
adaptation of Cu on chip has also undergone major
developments with respect to new low-k dielectric and
barrier materials. The new schemes have led to
different reported textures in the Cu lines by different
authors. There have been contradictions in the
observations among different authors on some of the
aspects. As a result, the area of microelectronic
interconnects still needs more attention and research in
order to fully understand the role of new materials and
processing parameters on deciding the final texture and
microstructure in the Cu lines. In this paper we have
studied the variation of texture in the Cu lines in both
as-deposited and annealed condition and additionally
as a function of linewidth.
Ten different damascene Cu linewidths were
investigated using XRD and EBSD. The linewidth/
pitch distances of the Cu lines 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 jam. The interconnects were deposited in an
area spanning 2.5 x 1.75 mm2 and were tested in asdeposited 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. Texture was measured using
Siemens D500 diffractometer. Three pole figures {111}, {110} and {100} were measured and used to
compute the orientation distribution function (ODF).
The same pole figures were then recalculated back
from the ODF. EBSD measurements were carried out
using OIM apparatus attached with Philips XL-30
FEG-SEM at 20 kV, sample tilt of 70° and working
distance of 15 mm. The top passivation layer on the
specimens was removed prior to the EBSD
investigations by etching in 15% HF acid for 5
minutes. The number of grains investigated via EBSD
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
494
varied from 500 to 32,000 for all the lines except 0.4
and 1 Jim lines where only 75 and 270 grains could be
examined. Residual stresses were measured using
Rigaku rotating anode diffractometer. The stresses
were measured using sin2\j/ methodology and {311}
reflection. 20 scanning was done in the range 89-91°
with step interval of 0.02°. The samples were inclined
at various values of CO to obtain the \|/ tilts of 19°, 27°,
34° and 40°. Both texture and stress measurements
were done using Cu-Kcc radiation at 40 kW exposing
the entire interconnect area to x-ray beam.
RESULTS AND DISCUSSION
'(c)
.«!>
a={115}<051>
ODF plots were computed for all the specimens but
have been represented for the narrowest 0.35 jim and
widest 100 jim lines in figure 1. We find the texture to
be sharper in the narrower lines compared to the wider
lines in the as-deposited condition. Jiang et al. also
found the texture to be sharper in the narrower lines but
Vanasupa et al. reported sharper texture for the wider
lines [1,2]. One can see the presence of {111} fiber
texture together with some contribution from
{111}<110> component in narrower lines and
{111 }<112> component in the wider lines. This shows
the influence of linewidth on the alignment of {111}
planes. Also we notice the presence of {511}
component, which is generated by twinning on the
{111} planes during annealing.
Upon annealing, the intensities decrease indicating
loss of texture strength. This can be explained from
figure 2, which depicts the ratio of volume fraction of
{111} component to {511} twin component for both
as-deposited and annealed lines. We find the ratio to be
higher in the narrower linewidths in as-deposited
condition. The ratio decreases with increasing
linewidth indicating that recrystallization has already
begun in the wider lines. Other authors have also
demonstrated the recrystallization to occur first in the
wider lines [3]. Higher fraction of twin component led
to the decrease in strength of {111} texture in the
annealed Cu lines. The increase in the fraction of twin
density with progress of recrystallization has been
reported by other authors [4].
Additionally, we also find the ratio to be lower in the
0.4 jim lines amongst the narrower linewidths and
higher in the 36 jim lines amongst the wider lines. The
0.4 and 36 Jim lines also had the lowest and the highest
linewidth to pitch distance ratio respectively amongst
all the specimens. This denotes the probable role of
linewidth to pitch distance ratio on influencing the
texture strength. The role of line density on affecting
the texture has also been demonstrated by other authors
bl =
FIGURE 1. ODF plots depicted at (p2 = 45° section for asdeposited (a) 0.35 (b) 100 and annealed (c) 0.35 (d) 100 urn
damascene Cu interconnect lines (XRD)
o
i
c
I ^As-depo.
• Anneal
Line-width (microns)
FIGURE 2. Ratio of volume fraction of {111} component to
{511} twin component in as-deposited and annealed
damascene Cu interconnect lines (XRD)
[2].
Figure 3 reveals the inverse pole figure maps for the
as-deposited and annealed Cu lines. Only a small
portion of the total scanned area has been displayed for
clarity. The grains have been colored in grayscale
according to their orientation as shown in the legend
besides the figures. Maps have been displayed only for
submicron lines for representation.
Table I and II show the normalized inverse pole
figures computed from ODF, parallel to the normal
(ND), transverse (TD) and longitudinal (LD) direction
of the as-deposited and annealed submicron Cu lines.
We find the presence of sidewall {111} component in
the 0.4 and 0.5 Jim lines in the as-deposited condition.
This component originates from the grains, which have
nucleated on the sidewall during deposition process.
Such sidewall component was not identified by x-rays.
495
(a)
<W
0.35
0.4-0.5
1-4
10-100
Linewidth (microns)
(b)
|ifit
0,7
(0
0.35
0.4-0.5
1-4
10-100
Linewidth (microns)
0,7 pn
FIGURE 4. Crystallographic direction graphs computed
parallel to the (a) specimen normal direction and (b) trench
sidewall normal direction as a function of linewidth for all
the annealed damascene Cu interconnect lines (EBSD)
1pm
room temperature, thereby eliminating the sidewall
{111} component.
Figure 4 shows the Crystallographic direction graphs
computed parallel to the specimen normal direction
and parallel to trench sidewall normal direction for
annealed specimens with 10° angular tolerance using
EBSD data. These graphs show the variation of
average area fraction of {111}, {110} and {100} grains
as a function of linewidth. The linewidths have been
grouped together to make the visualization easier.
Similar plots computed for the as-deposited lines (not
shown here) had revealed higher fraction of {111}
grains in the 0.35 jim lines and {110} grains in the 0.4
and 0.5 jim lines in the specimen normal direction.
Upon annealing we see that the fraction of {110}
grains increases in all the submicron lines in the
normal direction (Fig. 4a) contrary to the XRD data
where we do not observe any {110} component. The
fraction of {111} grains was highest for the higher
linewidths of as-deposited and annealed specimens in
the normal direction.
Figure 4b depicts that from the sidewall direction, the
fraction of {100} grains increases upon annealing in
the 0.35 jim lines, while for the 0.4-100 jim linewidths,
sidewall {110} component develops. Thus, there is an
increase in fraction of sidewall {100} component and
at the same time the sidewall {111} component persists
upon annealing in the local grain population of the
submicron lines. Increase in the fraction of {110}
grains upon annealing in the normal direction may be
the consequence of this competition between {111}
FIGURE 3. Inverse pole figure maps for as-deposited (a)
0.35 (b) 0.4 (c) 0.5 and annealed (d) 0.35 (e) 0.4 (f) 0.5 jim
damascene Cu interconnect lines
In case of 0.35 jim lines we find the presence of
{111}<110> texture which is quite in agreement with
XRD data.
Upon annealing we find origin of Cube texture in the
narrowest 0.35 jim lines, which is weakened in the
normal direction by emergence of its first generation
annealing twin {221}. These components were not
identified by XRD. The sidewall {111} component as
identified in the local grain population of 0.4 and 0.5
jim lines in as-deposited condition persisted upon
annealing in the 0.5 jim lines. In case of 0.4 jim lines,
{100} component its twin {221} emerged from the
sidewall. Also the sidewall {111} component in the 0.4
Jim lines underwent an inclination of about 10°.
Surprisingly, we do not find the presence of sidewall
{111} component in the narrowest 0.35 jim lines in the
as-deposited condition. But it should be noted that
these lines also had the smallest pitch distance of 0.35
jim amongst all the specimens. Smaller pitch distance
probably gave rise to larger number of defects and
hence faster recrystallization in these lines. Presence of
defects in the upper corner of the trenches has been
demonstrated by other authors [5]. Also considerable
time had elapsed after these lines were deposited and
before the EBSD inspections were done. As a result,
these lines underwent complete recrystallization at
496
TABLE 1. Normalized inverse pole figures computed from the ODF in the normal (ND), transverse (TD) and longitudinal (LD)
direction of as-deposited submicron damascene Cu interconnect lines (EBSD)
111
&IH
lit
rtaot
,ii
\
TABLE 2. Normalized inverse pole figures computed from the ODF in the normal (ND), transverse (TD) and longitudinal (LD)
direction of annealed submicron damascene Cu interconnect lines (EBSD)
ND
I001J
• max* 1,878
1 1 .784
10611
(1001
11.008
|S,Si3
001
101
497
have reduced to very low values and are tensile both
along and across the lines. Tensile stresses originate in
Cu lines upon annealing due to higher coefficient of
thermal expansion (CTE) of Cu compared to
surrounding barrier and dielectric materials.
and {100}. Presence of {110} grains have been shown
to reduce the total energy of the system in the lower
linewidths according to results of other authors [6]. But
in our case the observed phenomenon is only
characteristic of the local grain population. Unlike
figure 4a, we do not observe the presence of {110}
component in the inverse pole figure for the annealed
0.35 urn lines in the normal direction (Table 2) due to
the tolerance angle of 10° used for the computation of
figure 4. Thus, variations were observed in the local
grain population of the Cu lines in both as-deposited
and annealed condition. Large variations in stress
across the polycrystalline Cu films with columnar
grains have been reported [7]. The variation observed
in the local grain population of our specimens could
also be the consequence of such stress variations.
Figure 5 reveals the stress measurements conducted
on the specimens using sin2\|/ methodology. In case of
narrower linewidths we find presence of compressive
stresses along the lines and tensile stresses normal to
the lines. For higher linewidths we find the stresses
both along and across the lines to be tensile while the
stresses in the normal direction are compressive. Some
authors have assumed zero stresses in the as-deposited
condition in their calculations [6]. But it should be
noted that in our case the stresses in the as-deposited
specimens originate from the deposition of top
passivation layer at high temperature.
Upon annealing we find the stresses to be tensile in
all the three directions for the narrower linewidths. For
higher linewidths the stresses in the normal direction
CONCLUSIONS
The texture evolution in as-deposited and annealed
Cu lines has been investigated as a function of
linewidth. The narrower linewidths had sharper texture
in the as-deposited condition compared to higher
linewidths. Texture was predominantly {111} fiber
with some contribution from {111}<110> component
in the narrower linewidths and {111}<112> component
in the higher linewidths, indicating the influence of
linewidth on alignment of {111} planes. Increased
twinning led to the decrease in strength of {111}
texture upon annealing. Recrystallization had already
begun in the higher linewidths in the as-deposited
condition. A possible role of linewidth to pitch distance
ratio was identified on influencing the texture strength.
Variations in texture were observed in the local grain
population of the Cu lines. Stresses became tensile in
all the three principal directions of the lines upon
annealing with very low values in the normal direction
of higher linewidths. This is due to higher value of
CTE of Cu compared to the surrounding dielectric and
barrier materials. The stresses observed in the asdeposited specimens were consequence of deposition
of top passivation layer at high temperature.
400 -i
REFERENCES
• o22
o33
1. Jiang, Q.T., Tsai, M. H., Frank, A., Parihar, V., Nowell,
ML, Auger, RA., Havemann, R.H. and Luttmer, J.D.,
"Annealing impact on damascene Cu resistivity and
microstructure", Proceedings of the IEEE International
Interconnect Technology Conference, 2001, pp. 400-404.
2. Vanasupa, L., Pinck, D., Joo, Y.C., Nogami, T.,
Pramanick, S., Lopatin, S. and Yang, K., Electrochem. and
Solid-State Letters 2 (6), 275-277 (1999).
3. Gross, M. E., Drese, R., Lingk, C, Brown, W. L., EvansLutterodt, K., Barr, D., Golovin, D., Ritzdorf, T., Turner,
J., and Graham, L., "Electroplated damascene Copper:
Process influence on recrystallization and texture", in
Polycrystalline Metal and MagneticThin Films, MRS
Symposium Proceedings 562, 1999, pp. 215-222.
4. Okayabashi, H., Ueno, K., Saitoh S. and Nomura, E.,
"Microstructure of Electroplated Cu films", Advanced
Metallization Conference Proceedings, 1999, pp. 93-99.
5. Lingk, C., and Gross, M.E., /. AppL Phys. 84 (10), 55475553 (1998).
6. Thompson, C.V., and Carel, R., /. Mech. Phys. Sol. 44,
657(1996).
7. Gudmundson, P. and Wikstrom, A., Microelectronic
Engg. 60, 17-29 (2000).
-400
Linewidth in microns (log.)
(b)
600
1
1
10
100
-200
Linewidth in microns (log.)
FIGURE 5. Stress measurements conducted on the
specimens along (au), across (a22) and normal (o33) to the
(a) as-deposited and (b) annealed damascene Cu lines
498