The Electromigration and Failure Behaviour in Layered Tungsten Via Structures

  • PDF / 1,685,427 Bytes
  • 7 Pages / 414.72 x 648 pts Page_size
  • 16 Downloads / 234 Views

DOWNLOAD

REPORT


459 Mat. Res. Soc. Symp. Proc. Vol. 391 01995 Materials Research Society

Our test structure is a chain of eight Ml to M2 tungsten vias. The width of the Ml and M2 links are a nominal 1.25pm. Table 1 outlines the thicknesses of Ml. Links in the chain are 42.5gam between contacts. All results in this study are with the electrons alternately flowing down through the tungsten plug from M2 to Ml and up through the plug from Ml to M2. Experiments have been initiated with reverse polarity but have not been completed at this time. All splits were tested at a nominal temperature of 200'C and I=0.004A; a 10%increase in initial resistance was considered a failure. Although some structures did increase by over 100% in resistance, we found no totally open structures. The second part of our study pursued a continuation of process improvements for these vias. Titanium was added to some of the ARC splits, clean strips were also added, as well as additional time to the Ti seed layer in the contact. Table 2 outlines these splits. Table 2: Process Improvements split

M1 ARC

additional process

1

200ATi/600ATiN

2

200A TiN/600A Ti/200ATiN

3

200A TiN/800A Ti/200A TiN

4 5

800o TiN 8ooA TiN

CO 2 sparge, solvent strips PSR2 rinse, solvent strips

6

800o TiN

l5sec Ti Flash

RESULTS AND DISCUSSION Figure 1 shows the initial resistance distribution of splits from Table 1. Figures 2 and 3 highlight the distribution of scaled t 50 s and sigmas from these splits. These figures clearly show that the 1200A splits have the tightest resistance distribution with the lowest mean values, resulting in high t 50 s and low sigmas. Only the 1200K TiN splits consistently met the reliability requirements. 80C 70R 60i

B

50401200A

800A

500A

splits

Figure 1: Distribution of Initial Resistance Ml to M2 vias @25°C

460

28002:606

2600-

24002200-

2080 2000-

1956 1832

18001600-

t

1567

154'

1531

5 1400-

o0

13

.302

1 520

1267

01200-

116n ý1091094

I4R

lOOO-

385

83 2

BOG-

-

740

0cLEI

6oo40 20

1200A

500A

800A splits

Figure 2: Comparison of scaled t 50 s 2.3

1.9 1.81.6-

1.41 s i

1.27

9 1.2am

Ko:M 1.25

.2'i

1i.D+

-=O0.91

0. B+

0.6

0.6%__

0.6

0.64

+41

iq

41

0.29 0.20.2f

0.2

0.1

O

1200A

7_:

ZZOý

I

I

0.

800A splits

500A

Figure 3: Comparison of scaled sigmas Resistance versus Time plots indicated two main types of failures: a steady climb in resistance through 10% and beyond (Figure 4), and a stable resistance with a sudden, rapid increase (Figure 5). The 1200K TiN ARC samples did not show this second type of failure plot.

461

J 26o

320 2o

46o

2oo

66o

760

Figure 5: R vs. T 800A TiN ARC

Figure 4: R vs. T 800K TiN ARC

Although effort was spent cross-sectioning failed contact sites, we could not find any distinguishing characteristics between the two failure modes. Failure analysis was performed on several via chains. Using liquid crystal, the main failure sites were identified. FIB cross-sections of these sites typically revealed large voids in the Ml line directly under the tungsten plug.