Response Time for Optical Emission and Mass Spectrometric Signals During Etching of Heterostructures
- PDF / 427,038 Bytes
- 6 Pages / 414.72 x 648 pts Page_size
- 17 Downloads / 163 Views
grazing angle optical emission interferometry. 4 In the work by Collot et al.,5 AlGaAs layers as thin as 5 nm were resolved in GaAs/AIGaAs heterostructures. After detection and identification of the etch products, an effective endpoint detection scheme is needed to stop the etch once the underlying layer has been detected. An optimized endpoint detection system should include sensors with the best time response and signal-to-noise ratio. This is essential for precisely stopping an etch when the dimensions of the device are in the nanometer range. One such device is the GaInAs/AlInAs-based heterojunction bipolar transistor (HBT), whose base layer can be as thin as 60 nm.6 Using OES to monitor the Ga etch products, we have shown that an HBT emitter etch could be stopped with
W •"5
.-I
00 0
0
3~
0.1
1 10 FILTER FREQUENCY (Hz) Fig. 2 Dependence of Ga optical emission response time on cutoff frequency of lowpass filter for the amplifier. Also shown is the signal-to-noise ratio of the time derivative of the Ga signal. The etch conditions were the same as in Fig. 1. signal-to-noise decreases. The maximum value of the signal-to-noise was 20 and it occurred for a cutoff filter frequency of 0.3 Hz. The response time of the signal at this frequency was only 0.7 s. Therefore, for endpointing, a compromise is made between response time and sensitivity. At a filter frequency of 1 Hz, the signal-to-noise is 16 and the response time is 0.3 s. It is assumed that the response time will set a lower limit on the precision of the endpoint detection scheme. Figure 3 shows how the response time of the 145 AsC12+ signal depends on the update
29
2.0 2.0 r
70 70
mimmmi1 . .. I .
.
I . .
. I .
.
I .
. . I -
- -
•,
60 6-0
,
,, 1.5
VS50
"' +
" W
40
4-
1.0
Cl)
5 z
z o
30
-..
(.
"I
20
0
'''1' 1
0
"'
n-"-10
0
w
0'°
Ch 1 '1 ''I 3 2
'
1'l.li 4 5
0
. 6
FREQUENCY (Hz) Fig. 3 Response time and signal-to-noise for 145 ASC12+ signal as a function of the update frequency which is related to the integration time. The etch conditions were the same as in Fig. 1. frequency. For short integration times, the mass spectrometer responds quickly to changes in the 145 AsC12+ partial pressure and a minimum response time of 0.9 s is measured for an update frequency of 5 Hz. For processes that operate at higher pressure, the response time can be limited by transit times through differential pumping stages. Using in-situ MS sampling of polySi deposited at 5 Torr, a response time limited by viscous flow was measured to be -3 s.8 The drawback to using a short integration time is the increased sensitivity to noise which is not averaged out effectively during the short integration time. This can be seen by the low signal-tonoise ratio of only 6.4 measured at an update frequency of 5 Hz. For longer integration times, the signal-to-noise is increased significantly by lOx to 64 at an update frequency of 1 Hz. At a frequency of 1 Hz, the response time of the mass spectrometer increased to 1.8 s. For endpoint detection, a fre
Data Loading...
