Physical Results

4.1.1.1 Thickness Uniformity

First, we compare the thickness uniformity between disilane-based and silane-based nitride, including as-deposited and post-etching. Figure 16 shows the 49-point thickness map of nitride film. In Figure 16-(a), the silane-based nitride shows worse uniformity than disilane nitride shown in Figure 16-(b), and Figure 16-(a) shows a special area, which can be called a “cold spots” of the heater. But disilane does not show the cold spot area since it is not sensitive to temperature.

Hence, the uniformity of a blanket wafer can be improved by using disilane nitride instead of silane in the range of 1.6 nm to 0.5 nm. The thickness uniformity of the device wafer also shows that disilane-based nitride can be improved about 68% on measurement pads. Since disilane is non-sensitive to temperature, the within-wafer uniformity is independent of tool effects such as heater type and gas flow path.

Otherwise, post etch uniformity is more important than as-deposited uniformity since the device is related to post etch profile and uniformity directly. Hence, the dry etch rate and wet etch rate are present in the two following sections.

(a). Silane-based SIN (b). Disilane-based SIN

Figure 16. Based on the same thickness, the range comparison of disilane and silane nitride on blanket wafer

4.1.1.2 Dry and Wet Etch Rate

The dry etch rates of disilane nitride and silane nitride are 12.2nm/min and 15nm/min, respectively. For SPM, the wet etch rates of disilane nitride and silane nitride are 0.88nm/min and 1.5nm/min, respectively. The etch rate is shown in Figure 17. The lower etch rate for dry and wet etch of disilane nitride is beneficial to controlling the uniformity, and to increase the etching process window at the same time, especially once this kind of film is applied on the spacer layer. Furthermore, the map measured by SCD (spectrum CD) also showed the within-wafer range of disilane nitride was improved 32%.

4.1.1.3 Surface Roughness

Generally, lower activation energy means higher growth rate, which might lead to a rougher surface. For the accepted temperature, the growth rate is 10.7nm/min and 8.6nm/min for disilane-based and silane-based nitride, respectively. But the AFM results show disilane-based nitride can obtain comparable roughness to silane-based nitride. Figure 18 shows the roughness results.

15

Figure 17. Dry etch rate and wet etch rate

0.15 0.177

Figure 18. Roughness comparison

4.1.1.4 Pattern Loading Effect

Here micro-loading was defined as (maximum thickness - minimum thickness) / (minimum thickness) in the same die. Normally, a dense area results in minimum thickness, and an iso area results in maximum thickness. Naturally, decreasing deposition rate can improve micro-loading. In this work, we find that even decreasing deposition rate will not suffer the activation energy, which means within-wafer uniformity won’t suffer. Consequently, we can improve micro-loading effect and

uniformity at the same time. The micro-loading of disilane nitride is improved about 8% compared to silane nitride if the gas flow is fine tuned by disilane. On the contrary, if we decrease the deposition rate of silane nitride to improve micro-loading, throughput suffers seriously.

On the other hand, morphology is a concern because lower activation energy means higher growth rate, which might lead to a rougher surface [20]. But surface roughness was verified by AFM methodology. The results show that disilane-based nitride can obtain comparable roughness to silane-based nitride even though it has a higher growth rate.

The physical data show within-wafer uniformity with disilane nitride can be improved 68% over silane-based nitride. Furthermore, micro-loading was improved 8% by disilane gas.

4.1.1.5 Discussion

Figure 19 and Table 1 show that a disilane-based SiN process has low activation energy, which makes the reaction close to a diffusion-limited reaction. On the contrary, silane-based SIN showed a sharp slope in the same temperature period, which means it has larger activation energy. Disilane-based SiN has smaller activation energy than silane-based SiN in these temperature periods, meaning that disilane-based SiN comes closer to diffusion-limited reaction than silane-based SiN.

Figure 19. Temperature versus growth rate for disilane and silane-based nitride

Nitride Recipe Heater Temperature (Degree C)

Activation Energy (eV)

> 650 0.2592

< 650 0.6685

> 650 0.2896

< 650 0.6969

> 690 1.5024

< 690 1.9984

Disilane Nitride (recipe1)

Disilane Nitride (recipe2)

Silane Nitride

Table 1. Temperature versus activation energy for disilane and silane-based nitride

Whether the higher growth rate (disilane 1) or lower growth rate (disilane 2) is used, it always keeps the similar activation energy and slope. It reveals that growth rate does not affect the activation energy for the same precursor, and it is very important for process tuning because the micro-loading phenomena can be improved by decreasing the deposition rate.

Disilane-based nitride is non-sensitive to temperature since the reaction was

dominated by diffusion-limited reaction. We can conclude it has a slower slope compared with silane nitride in Figure 19. The growth rate and thermal budget are acceptable for the 65nm node or beyond.

Naturally, the non-sensitivity to temperature will prevent cold spots on the heater.

This means disilane-based nitride is heater independent.

Of course, we can improve the silane-based nitride uniformity if we increase the deposition temperature to approach a diffusion-limited reaction, but it will suffer the thermal budget for the same uniformity target. Decreasing the thermal budget is the future trend of advanced manufacturing. Consequently, it is necessary to search a low activation energy precursor. In our experiment, disilane recipe 1 and recipe 2 have different growth rates, but present the same curve in Figure 19 and similar activation energy in Table 1. The results are beneficial for micro-loading tuning since we can decrease micro-loading effect by tuning its growth rate.

In the results section, we understand that micro-loading can be improved 8%

once we modify the gas flow to lower down growth rate. Even so, the characteristic of non-sensitivity to temperature still exists. Normally, lower growth rate decrease pattern sensitivity due to increased gas phase diffusion length [21], which also explains the reason why the micro-loading is minimized.

Disilane is a better precursor for nitride film deposition than silane because it can be controlled on the specific area shown in the top-right corner in Figure 19.

Disilane has the characteristics of low thermal budget, comparable throughput, and is non-sensitive to temperature.