Dielectric Fracture Leading to Metallic Connections by Laser Heating
- PDF / 2,194,894 Bytes
- 5 Pages / 414.72 x 648 pts Page_size
- 80 Downloads / 244 Views
Via
Metal 1
Links
Cuts
12-
-o-
2.0ýtm
pitch
Figure 1: A continuous chain of metal links to allow 4-point measurements of one link in series with a short segment of metal 1 and two vias to metal 2. The system used for laser linking included a Spectra Physics diode-pumped Q-switched Nd:YLF laser (A = 1047 nm) shuttered mechanically to select single pulses, focussed to a 1/e' diameter of approximately 4yrm and could be expanded to 8 pm. The pulse repetition frequency (PRF) was 180 Hz, producing a pulse length of approximately 8 ns. A four-wire ohmmeter was connected to the packaged chip that was placed under the computer controlled laser table. A program was written to position each link under the
laser spot, pulse the laser and measure the resistance of each connection. If the link was successfully formed, the laser then cut away the link; however, if the link was not formed, the beam would be moved to the next link site without cutting away the link. The only variable parameters were focus and pump power. RESULTS The link yield was sensitive to laser pump power, which can be controlled very precisely. At 0.47 Watts, the yield was less than 80%, and it was clear that this was not a sufficient power to form the links. At 0.50 Watts, large craters formed in some cases and the yield was lower. Between 0.480 and 0.490, the yields were better than 99%, with 0.485 Watts giving a total yield of 99.8% after 1164 links were tested. Two failures were found to have been due to inaccuracies in laser positioning. A histogram of resistance values is shown in Fig. 2, where a 1.25il correction was made for the resistance of two vias and a 12yrm length of metal 1. The distribution of resistance values, R, fit a second order Hermite Gaussian 2 (1) R2e-R 2/2G Prob(R) = a3V where the standard deviation, a, was found to be 1.23Q by a non-linear fit that is the heavy dashed line in Fig. 2. The cumulative link probability is also plotted for comparison.
834
1.0
S0.5 I
/
I_
/ III
I II
0.0
0.0
/ /
-I
2.0
4.0
6.0
8.0
10.0
R (Ohms) Figure 2: A histogram of 4-point probed link resistances from 1164 link attempts, fit to a Hermite Gaussian distribution and a cumulative probability.
835
Figure 3: Optical micrograph of a typical formed link adjacent to an unformed link. In every case links were either formed, with a resistance less than 10M2, or were not formed and remained completely open. The laser could not be positioned to better than ±0.25pm accuracy across the chip, so the beam was expanded to the full 8/im diameter and the laser pump power was also increased. The optimum power was then determined to be 520 mW where perfect success was achieved in over 8,000 consecutive links. When the power was raised to 530 mW, the yield fell to approximately 1 failure in 2,000 links (99.95% yield). Between 500 and 510 mW, the yield degraded to between 99.5 and 99.8%. DISCUSSION An optical micrograph of a formed and an unformed link is shown in Fig. 3. The high reflectance material between the lines is metal 2 that flowed fro
Data Loading...
