TEXTURE AND RESISTIVITY OF Cu and DILUTE Cu ALLOY FILMS

K. Barmak, A. Gungor, A. D. Rollett, C. Cabral, Jr.1, J. M. E. Harper1

Department of Materials Science and Engineering, Carnegie Mellon University, Pittsburgh, PA
15213
1
IBM T. J. Watson Research Center, P. O. Box 218, Yorktown Heights, NY 10598

ABSTRACT

Annealing of dilute binary Cu(Ti), Cu(In), Cu(Al), Cu(Sn), Cu(Mg), Cu(Nb), Cu(B),
Cu(Co) and Cu(Ag) alloy films resulted in the strongest <111> fiber texture for Cu(Ti) and the
lowest resistivity for Cu(Ag). The behavior of the alloy films was compared and contrasted with
that for a pure evaporated Cu film. Electron beam evaporated films with compositions in the
range of 2.0-4.2 at% and thicknesses in the range of 420-560 nm were annealed at 400˚C for 5
hrs. Two different approaches were used to derive volume fractions of texture components,
namely fiber plots and orientation distributions. It is argued that for polytextured films such as
the copper alloys studied here, orientation distributions derived from pole figures provide the
most reliable basis for quantitative characterization.

INTRODUCTION

The need to reduce delays in integrated circuits prompted the replacement of Al(Cu) with
Cu in the 1990’s. However, as line widths approach 100 nm, the addition of alloying elements to
tailor the grain size and crystallographic orientation (i.e., texture), and thus the reliability and
functionality, of Cu interconnections may become necessary. In previous work we investigated
the impact of alloying elements on the decomposition and electrical resistivity of a group of
dilute binary, non-compound forming Cu alloy films annealed at a constant heating rate up to
950˚C [1].
In this work, we address the evolution of texture and resistivity in a series of nine Cu
alloy films that are annealed isothermally at 400˚C. By way of connection to our previous work,
five of the alloys have been chosen from among compound forming binary Cu systems and four
have been chosen from among non-compound forming systems. The five compound forming
systems are Cu(Ti), Cu(In), Cu(Al), Cu(Sn), Cu(Mg), and the four non-compound forming
systems are Cu(Nb) Cu(B), Cu(Co), and Cu(Ag). We compare and contrast the behavior of the
alloy films with that for pure Cu. In addition, we discuss the advantages and disadvantages of
two different x-ray based methods commonly used in the study of texture, i.e., in the
determination of the spatial distribution of the crystal axes in a polycrystalline aggregate, and
how to choose the best method for a given film type.

EXPERIMENTAL

Pure Cu and dilute binary Cu alloy films were electron beam evaporated onto thermally
oxidized silicon wafers. The composition and thickness of the films are listed in Table I. X-ray
diffraction experiments were conducted using Cu Kα radiation in a Rigaku θ-2θ powder
diffractometer with a curved graphite monochromator. These patterns were used to give a
qualitative indication of the texture in the films. For the quantitative analysis of film texture,
fiber plots and pole figures were used. Pole figures and fiber plots were obtained in a Philips
X’Pert Pro diffractometer, utilizing an X-ray lens and 0.27o Soller slits in the path of the incident
beam and a flat graphite monochromator in the path of the diffracted beam. The fiber plots were
measured in 0.1˚ intervals per second as is commonly done in texture studies of Al films [2],
while the pole figures were determined at 5˚ intervals every two seconds. Three pole figures -
{111}, {200} and {220} - were obtained for each sample. In order to ensure that the use of 5˚
intervals for the collection of pole figures did not affect the resolution of intensity data used for
the calculation of the orientation distribution functions (ODF), sections through the pole figures
were compared with the fiber plots. Three such sections for the annealed Cu film have been
overlaid on their respective fiber plots in Fig. 1 and show no discernable difference.

Table I. Composition and thickness of the films.

Sample Solute Solute Thickness

(at%) (wt%) (nm)
Cu - - 420
Compound Forming Alloys
Cu(Ti) 3.0 2.3 488
Cu(In) 2.6 4.6 520
Cu(Al) 2.0 0.9 480
Cu(Sn) 2.3 4.2 540
Cu(Mg) 4.0 1.6 520
Non-Compound Forming Alloys
Cu(Nb) 2.4 3.5 560
Cu(B) 3.0 0.5 500
Cu(Co) 2.8 2.6 535
Cu(Ag) 3.0 5.0 550

All fiber plots and pole figures were corrected for background, defocusing, and
absorption. Defocusing corrections were made using a thick polycrystalline, electroless plated
Cu film. For the absorption correction, the actual composition of each alloy film was used in
conjunction with tabulated mass absorption coefficients [3]. The fiber plot data was normalized
such that the intensity of the random component was set equal to one. The angular offset of the
fiber axis from sample normal, δ, and the volume percents of the various texture components
were determined from the normalized fiber plots[4].
The volume fraction of oriented grains was calculated from both the fiber plots and the
orientation distribution function (ODF). For the calculation of the volume percents of the
oriented grains from the <111> fiber plots, it was necessary to deconvolute the <100>, <110>
and the <111> peaks between 30 and 85o tilt; this was done by fitting the peaks to Gaussians.
The ODF was obtained from the three sets of pole figures using the Philip X’Pert software
package [5], and the volume fraction of each orientation, g, was calculated by integrating over
the appropriate subvolume in orientation space,

dV 1
= 2 f ( g )dg ,
V 8π
 dV
where is the volume fraction of grains with orientation g(ϕ1,Φ , ϕ2), f(g) is the orientation
V
distribution function, dg = sinΦdΦdϕ1dϕ2, and finally Φ , ϕ1, and ϕ 2 are Euler angles that
parameterize the orientation space[6]. Each point in Euler space corresponds to a particular
orientation. For example, g = (0°, 55°, 45°) corresponds to the {220} <111> orientation [6].
Similarly, orientations in the <111> fiber component have Φ = 55°, φ2 = 45°, with φ1 taking all
values over the range 0-360°. Discrete orientation distributions were used and the intensity was
integrated over all cells within 15˚ of a given orientation with due allowance for overlaps.

30000
<111>
25000
Intensity

20000

15000

10000
<100>
5000

<110>
0
0 20 40 60 80
Tilt Angle (Degree)

Figure 1. The sections through the {111} {200} and {220} pole figures (open symbols) for
400˚C, 5 hr annealed Cu are overlaid on the <111>, <100> and <110> fiber plots (solid lines),
respectively.

RESULTS AND DISCUSSION

The room-temperature electrical resistivities of as-deposited and annealed films are

summarized in Table II. As can be seen from the table, alloying clearly increases the resistivity
in the as-deposited state, with room-temperature resistivities of the as-deposited alloys ranging
from as high as 30.8 µΩ-cm for Cu(2.4Nb) to as low as 4.2 µΩ-cm for Cu(3.0Ag). Annealing at
400˚C for 5 hours resulted in a decrease of resistivity in all cases, with the lowest value, 2.6 µΩ-
cm, occurring for annealed Cu(3.0Ag) and the highest value, 6.4 µΩ-cm, being seen for both
Cu(2.3Sn) and Cu(2.8Co). The resistivity for the as-deposited pure Cu film was 2.3 µΩ-cm and
after annealing the resisitivity dropped to 2.0 µΩ-cm.
The XRD patterns for the as-deposited and annealed films in Fig. 2 revealed the presence
of Cu {111} and {200} peaks. The large {111} to {200} peak intensity ratios in these patterns
and the absence of peaks other than {111} and {200} qualitatively indicated that the films were
textured.
 Table II. Electrical resistivity of as-deposited and
400˚C, 5 hr annealed pure Cu and Cu-alloy films.

Sample As-deposited Annealed

Cu (at% solute) (µΩ-cm) (µΩ-cm)
Cu 2.3 2.0
Cu(3.0Ti) 22.5 3.9
Cu(2.6In) 6.2 3.9
Cu(2.0Al) 5.5 4.4
Cu(2.3Sn) 8.6 6.4
Cu(4.0Mg) 5.7 3.4
Cu(2.4Nb) 30.8 5.3
Cu(3.0B) 10.0 3.2
Cu(2.8Co) 10.3 6.4
Cu(3.0Ag) 4.2 2.4

140000 100000
Cu(111) As-deposited 90000 As-deposited
120000 400oC, 5 hr 80000
Cu(111) 400oC, 5 hr

100000 70000
Intensity

Cu(200)
Intensity

Cu(200) 60000
80000
50000
Cu(3.0Ti) Cu(2.4Nb)
60000 40000
Cu
Cu(2.0Al) 30000
40000 Cu(2.8Co)
Cu 20000
20000
Cu(2.6In) Cu(3.0Ag)
10000
Cu(2.3Sn) Cu(3.0B)
Cu(4.0Mg) 0
0
40 43 46 49 52 55 40 43 46 49 52 55

2-Theta 2-Theta

Figure 2. X-ray diffraction patterns for as-deposited and annealed, (a) compound forming and (b)
non-compound forming Cu-alloy films. The traces have been displaced for clarity.

The <111> fiber plots for the as-deposited and annealed films are given in Fig. 3. The
ODF for annealed Cu(3.0Ti) and Cu(2.0Al) films, chosen as representative examples, are shown
in Figs. 4a-d. The results of the analysis of the <111> fiber plots, stated as the angular offset of
the fiber axis from the sample normal, δ, and the volume fraction of a given texture component,
are given in Table III. The volume fractions of fiber components calculated from the ODF are
listed in Table IV.
 As evidenced by the fiber plots in Fig. 3 (and pole figures, not shown), all as-deposited
and annealed films exhibited “true” fiber texture, i.e., δ was zero in all cases. This is in contrast
to Al or Al(Cu) films where the texture is “near” fiber, and where the angular offset of the fiber
axis from sample normal can be several degrees [7,8]. Annealing resulted in the strengthening
of texture for all the films, seen as a decrease of the volume fraction of the random component
(Fig. 3, Tables III and IV). In the case of pure Cu, and the four alloys, Cu(3.0Ti), Cu(2.0Al),
Cu(2.4Nb) and Cu(2.8Co), it was the <111> fiber component that became the strongest
component upon annealing. By contrast, for the four alloys, Cu(2.6In), Cu(2.3Sn), Cu(4.0Mg)
and Cu(3.0B), <110> became the strongest component. For Cu(3.0Ag), <100> became the
strongest component (Table IV).

70 As-deposited
400oC, 5 hr
Normalized Intensity

60
<111>
<100> <111>
50
Cu(2.4Nb)
40
Cu
30
Cu(2.8Co)
20
Cu(3.0Ag)
10
Cu(3.0B)
0
0 20 40 60 80
Tilt Angle (Degree)
 85
80

Normalized Intensity
75
As-deposited
70 <111> 400oC, 5 hr
65
60 <100> <111>
55 Cu(3.0Ti)
50
45 Cu(2.0Al)
40
35 Cu
30
25 Cu(2.6In)
20
15 Cu(2.3Sn)
10
5 Cu(4.0Mg)
0
0 20 40 60 80
Tilt (Degree)

Figure 3. <111> fiber plots for as-deposited and annealed, (a) compound-forming and (b) non-
compound forming Cu-alloy films.
Figure 4. The two-dimensional sections and the three dimensional representation of the
orientation distribution function (ODF) for annealed Cu(3.0Ti) (a and b), and annealed
Cu(2.0Al) (c and d) films. The Cu(2.0Al) film shows a significant <100> component.

In drawing the above conclusions, it is worth noting that the calculated values for the
fraction of oriented grains are not the same for the <111> fiber plot versus the ODF method
(Tables III and IV). The reason for this discrepancy is that the extraction of volume fractions
from fiber plots requires an assumption about the presence and distribution of each component.
In fact, for the most weakly textured films that contain In, Sn, Mg, B, or Ag, the volume
fractions cannot be determined from the <111> fiber plots (Table III). By contrast, the
calculation of the volume fractions from the ODF makes no assumptions about the presence or
absence of any particular component.

Table III. Analysis of the fiber plots presented in Fig. 3. δ is the angular offset of the <111> axes
from the substrate normal. The volume percents are those of grains with the given axis normal to
the film plane.

Annealed Films
As-deposited Films
Alloy _ 400oC, 5 hr
(Degree) <111> <100> <110> random <111> <100> <110> random
Cu(Ti)
0 47.6 16.8 0.8 34.8 89.7 4.4 1.2 4.7
Cu(Al)
0 20.7 22.9 1.0 55.4 67.1 15.7 1.7 15.5
Cu(In)a
0 - - - - - - - -
Cu(Sn)a
0 - - - - - - - -
Cu(Mg)a
0 - - - - - - - -
Cu
0 59.8 20.5 2.2 17.5 68.7 17.9 1.8 11.6
Cu(Nb)
0 55.3 19.3 3.2 22.5 83.1 2.1 0.8 14.0
Cu(B)a
0 - - - - - - - -
Cu(Co)
0 51.0 24.6 2.2 22.2 56.5 23.8 1.5 18.2
Cu(Ag)a
0 - - - - - - - -
a
Volume percents cannot be calculated from the fiber plots.

Thus, we emphasize that in the case of polytextured films such as the Cu and Cu-alloy
films studied here, orientation distributions derived from pole figures provide the most reliable
basis for the quantitative characterization of the volume fraction of texture components. By
contrast for monotextured films such as pure Al or Al(Cu) films, it is more appropriate to collect
the texture data in the form of a high-resolution <111> fiber plot, as has been the practice [9-11].
In order to simultaneously assess the impact of the alloying elements on texture and
electrical resistivity upon annealing, the ODF-calculated <111> volume fractions are plotted
versus resistivity for the 400˚C, 5hr annealed films in Fig. 5. The only two alloy films that are
found to be significantly more strongly textured than pure Cu are Cu(Ti) and Cu(Nb). By
contrast, the low resistivity films containing Ag, B, Mg, or In have week <111> fiber texture.
The best combination of <111> texture (high) and resistivity (low) among the alloy films is
found for Cu(3.0Ti).
Table IV. Analysis of the orientation distribution function. The volume percents are those of
grains with the given axis normal to the film plane.

Annealed Films
As-deposited Films
Alloy 400oC, 5 hr
<111> <100> <110> random <111> <100> <110> random
Cu(Ti)
50.1 11.8 8.8 29.3 94.2 2.3 0.5 2.9
Cu(Al)
29.6 14.5 9.8 46.1 62.3 13.6 2.8 21.3
Cu(In)
10.5 11.9 16.7 60.9 13.8 20.8 28.8 36.6
Cu(Sn)
15.3 7.4 12.2 65.1 8.4 10.6 22.4 58.6
Cu(Mg)
7.9 13.7 29.6 48.8 7.8 15.5 43.2 33.5
Cu
52.4 17.4 2.3 27.9 59.8 19.2 2.5 18.5
Cu(Nb)
65.2 6.8 2.9 25.1 78.7 5.4 2.5 13.4
Cu(B)
5.7 11.8 24.0 58.5 21.6 16.8 24.6 37.0
Cu(Co)
43.6 20.3 1.7 34.4 47.0 24.7 2.1 26.2
Cu(Ag)
14.6 15.6 14.8 55.0 15.6 19.4 13.9 51.1
 100
Ti
90

<111> Volume % (ODF)

80 Nb
70 Al
Cu
60

50
Co

40

30
B
20 Ag In
Mg Sn
10

0
2 3 4 5 6 7
Electrical ResistivityµOhm-cm)
(

Figure 5. Volume % of <111> fibers calculated from the ODF is plotted versus resistivity for
400˚C, 5 hr annealed Cu-alloy films.
CONCLUSIONS

After annealing at 400oC for 5 hr, Cu(3.0Ag) showed the lowest electrical resistivity
among the Cu-alloy films, whereas Cu(2.3Sn) and Cu(2.8Co) showed the highest resistivities.
All the alloy films, whether annealed or in the as-deposited state, were found to be fiber textured
and showed “true” fiber texture, rather than “near” fiber texture typically seen in Al based films.
Annealing resulted in the strengthening of texture for all films, with the strongest <111> fiber
being found for Cu(3.0Ti) and the weakest for Cu(4.0Mg). The best combination of texture
(strong) and resistivity (low) was found for Cu(3.0Ti). It was argued that for Cu and Cu-alloy
films, with significant fractions of non-<111> fiber components and large angular distributions
of the fiber axes about the sample normal, the most reliable measure of the volume fractions of
texture components is obtained through the acquisition of three pole figures and the calculation
of the orientation distribution functions.

ACKNOWLEDGMENTS

KB and AG gratefully acknowledge the NSF for financial support under DMR-9996315.
ADR and KB acknowledge partial support from the MRSEC program of the National Science
Foundation under award number DMR-0079996. The authors thank R. A. Carruthers, A.
Kellock, K. P. Rodbell, C. Lavoie, and J. Wolf for assistance with various aspects of film
preparation and analysis.

REFERENCES

[1] K. Barmak, G. A. Lucadamo, C. Cabral, Jr., C. Lavoie, and J. M. E. Harper, J. Appl. Phys.
87, 2204 (2000).
[2] K. P. Rodbell, private communication.
[3] B. D. Cullity, Elements of X-Ray Diffraction (Addison-Wesley, Boston, 1978).
[4] D.B.Knorr, D.P.Tracy, and K.P. Rodbell, Appl. Phys. Lett. 59, 3241 (1991).
[5] Philip Analytical, http://www-eu.analytical.philips.com
[6] H. J. Bunge, Texture Analysis in Materials Science: Mathematical Methods (Butterworths,
London, 1982).
[7] D. B. Knorr, Mater. Res. Soc. Symp. Proc. 309, 75 (1993).
[8] D. P. Tracy, D. B. Knorr, and K. P. Rodbell, J. Appl. Phys. 76, 2671 (1994).
[9] K. P. Rodbell, D. B. Knorr, and J. D. Mis, J. Electron. Mater. 22, 597 (1993).
[10] P. L. Besser, J. E. Sanchez, Jr., D. P. Field, S. Pramanick, and K. Sahota, Mater. Res. Soc.
Symp. Proc. 472, 313 (1997).
[11] W. M. Kuschke, A. Kretschmann, R. M. Keller, R. P. Vinci, C. Kaufmann, and E. Arzt, J.
Mater. Res. 13, 2962 (1998).