Results and Discussions
6.1 I-Line Stepper
One of the major issues with gate fabrication with an I-line stepper is scattering effect. As shown in Figure 6-1, during 2nd exposure, when light hits the edge of the nitride layer, light will be scattered in random directions.
These scattered beams of light will hit the photoresist in all directions even sideways. The outcome is the amount of exposure at different places will be different, and the result is shown in Figure 6-2b where the side of the photoresist is not smooth. The surfaces of the walls are wave-like.
Furthermore, when the light hit the plane surface of either nitride or substrate, it will be reflected. Due to the height difference between nitride and the wafer, the reflected light beams caused by the plane surfaces are different in intensity. With reflected light beams of different intensities, the amount of exposure experienced by the resist over the nitride and substrate are different.
Since the behavior of light is very complex and is not the focus of this study, there will be no further discussion about it. However, because of the complex behavior of light, carefully tuned dosage becomes very crucial.
Figure 6-2a is a SEM picture of a successful 0.2µm shifted sample. The
the photoresist of the 2nd layer. Originally, the trench was 0.4 micron, but after the shifting technique, the trench shrunk to 0.2 micron, half the original size, see Figure 6-2b. As mentioned in the previous paragraph, in Figure 6-2b, the walls of the photoresist are not even. This is due to the scattering of the light beams. However, with the correct dosage from I-line stepper, this problem can be minimized. When the optimum dosage is used the result, will be like that shown in Figure 6-3. In Figure 6-3, there is almost no sign of the uneven walls caused by the scattering of light.
In this study, the required dosage for 1st layer pattern was around 1650 J/m2 for all different shifts. For 0.2µm shift, the dosage for the second layer pattern was approximately 2500J/m2. This condition was only suitable for samples with 0.2µm shift. For other shifts, the required dosage of the 2nd layer pattern would be different. Different shift would require different exposure dosage because, as mentioned before, the height difference between nitride and substrate would cause the amount of exposure experienced by the resist to be different. Therefore, every different shift requires an optimum dosage of its own.
As mentioned previously, the advantages of using I-line stepper for gate fabrication include high throughput, high accuracy, high reliability and lower cost than electron beam lithography. Another major advantage is that the amount of shifting can be changed to any value desired. In another word, any gate length within the machine’s capability can be achieved. The result of 0.2 micron shift has already been shown previously. In Figure 6-4, a sample with 0.1 µm shift had been demonstrated.
After the desired trench had been made, Ti/Pt/Au was deposited.
Immediately after metal deposit on was metal lift-off. Figure 6-5 is a SEM picture of a metal gate for a 0.2 micron shift sample after metal lift-off.
6.1.1 Statistic Analysis
To achieve high throughput, reliability is very important. To prove that this method is reliable and accurate, we used 18 dies all with 0.2 micron shift.
Analyzing these dies by statistical analysis, we were able to calculate the amount of error and the accuracy of this study.
According to Canon, the FPA 3000 i5+ has a 3σ of 45nm. In statistics, this means that 95% of the errors are within 45nm. Refer to Figure 6-6, there are 18 bars each represents one reticle and its final gate length. The dash red line indicates the ideal gate length, 0.2µm. Table 6-1 shows the actual gate lengths and errors. The average feature size error is ±23nm. In the 18 dies we analyzed, 16 dies have errors smaller than 45nm. This means that 89%
are within 45nm of error, which is not as Canon suggested, 95%. However, not all errors were caused by the stepper alone. The dry etcher and SEM measurement could also be other sources of error.
The final line width was determined by SEM. In the SEM computer there is a built-in virtual ruler, which is used for measuring distances between two points of a SEM image. User simply clicks on a computer mouse to control the ruler and measure the desired distance. Hence, human error is introduced using this method of measuring line width. In another word, if the distance between two points were measured 10 times, the results may not be
consistent.
Figure 6-7 is a simple schematic drawing of a SEM stage. As shown in the figure, when trying to take SEM cross-section picture of a sample, the sample is taped inside the slot. However, as shown in Figure 6-7b, if the sample was taped at an angle to the horizontal, the resulting dimension would be inaccurate. This is the reason why SEM could be a source of error.
Other than stepper and SEM, another source of error was the TCP plasma etcher. This etcher was used for nitride etching; in another word, pattern was transferred to the nitride film. After etching, throughout the wafer, the widths of the trenches were not consistent; instead, some would be wider while others would be narrower. This may be caused by two reasons. First of all, the plasma etcher itself might cause the uneven trench dimension. Secondly, the problem might be caused by uneven thickness of the nitride film. Because of the dimension difference between the trenches of the first pattern, when the second pattern was defined, the resulting foot-print dimensions would be inconsistent. The outcome of the statistical analysis was hence affected.
So far, we had explained that not all the errors were due to the I-line stepper. SEM and TCP plasma etcher were other two major sources of error.
If we could limit the source of error to I-line stepper only, the outcome should be closer to the 3σ value suggested by Canon. In order to achieve this goal, we offer several solutions. For one thing, instead of PECVD, HDPCVD (High Density Plasma CVD) could be used for growing nitride to achieve higher uniformity of the film. Combine the more uniform film with a better plasma dry etcher could ensure higher consistency of the dimension of the
trenches. Furthermore, if a SEM with higher definition is used, the result should be more accurate.
6.1.2 Future Study of Gate Fabrication Using I-line Stepper
In this study, the new method of gate fabrication using an I-line stepper had been demonstrated on a Si wafer only. Later on, it will be demonstrated on a real III-V semiconductor high frequency device. At this point, the effect of a γ-shaped gate with uneven head on a real device is still unknown.