high temperature SACVD USG (HARP) does not require any pre-treatment process, and it does not allow any plasma attack to reduce the risk of devices damaged by high density plasma.
Figure 3.1 65nm node STI trench structure.
3.2 Evolution of HARP SACVD
3.2.1 HARP chamber configurations
The configurations of HARP chamber is much similarly to SACVD, but there is some different in both of block plates and showed heads. Because of the wafer bowling after thin film deposition by the low deposited rate, and vacuum chuck heater is the other characteristic hardware configuration for HARP SACVD chamber. Vacuum chuck heater is a heater flowing with Helium to attract wafer onto the heater surface by pressure servo during
- 72 -
process deposition. HARP systems mainly divide into process chamber body, liquid delivery controller and pump system.
Figure 3.2 Chamber design of SACVD for high aspect ratio process.
HARP is the twin chamber designed by sharing the same pressure control, while temperature control is independently for each heater during processing.
Due to HARP is the thermal process, there is no any RF or DC power to generate process gas or liquid into plasma, SACVD process is the decomposition process by using high concentration of ozone and high temperature to form dielectric film.
Remote plasma source RPS is used to clean chamber without any plasma damaged to chamber body and process kits. A larger unit of RPS provides a high efficiently power to break down high flowing clean gas of NF3.
- 73 -
Figure 3.3 Face plates comparison between standard SACVD and HARP SACVD.
It is apparently that the HARP faceplate with the slots is quite effective in eliminating the spots and streaks. On a standard faceplate, the gases flow through a 62 mils holes and then through a 29 mils holes prior to exit from the showerhead. This “nozzle” designed increases the exit velocity of the gases.
On the HARP faceplate, the gases initially flow through a 10mil hole, followed by a 25mils hole and then flow through an azimuth slot machined on the faceplate, the 25 mils hole and the 31 mils azimuth slot effectively reduce the downward vertical velocity of the gases. Reduction in downward velocity will reduce the intensity of the “spots”. The alignment of the holes can also be eliminated by the design of azimuth slots [51].
This model predicts that as the spacing reduces, the flow towards the edge of the wafer dramatically increases due to the large pressure drop from the center to edge of the blocker and faceplate. One method of addressing this problem is by increasing the pressure drop across the faceplate by reducing the diameter of the holes. A larger pressure drop across the faceplate will reduce
- 74 -
the effect of spacing overall uniformity by redistributing the flow between the blocker and faceplate.
3.2.2 Liquid delivery systems.
TEOS is the standard reactant for HARP process. Individual vapor line (IVL) high conductance GPLIS available on HARP chamber only, which supports high carrier gas flow (41slm) for high TEOS flow vaporization. A certain type of carrier gas, e.g., N2 or He, is utilized to deliver vaporized TEOS and liquid dopant sources to the process chamber to reduce gas-phase nucleation for particle control. Due to their distinct physical properties, such as heat capacity and thermal and mass diffusivity, different carrier gases may result in different film characteristics. In this study, we report a complete analysis using N2 as the carrier to develop a better understanding of the reaction mechanism and to achieve superior film quality. The N2 carrier process satisfies the technical requirements for STI. The main drawback for this process is the high cost per wafer due to its low deposition rate and high He cost in the Asia Pacific.
TEOS flow is controlled by digital liquid flow meter T4. The T4 LFM has achieved the strict requirements necessary for the HARP process which include faster response, minimal ambient temperature effects, and accurate ramp rate. The T4 LFM also resolves the device leakage issue caused by the slow shut off of the do-pants on BPSG processes [52]. The T4 LFM dimensions and hardware requirement is identical to the existing liquid flow meters and thus is a drop in solution to our existing platform.
- 75 -
Figure 3.4 Individual liquid line modules.
The response time of T4 LFM is about 5-7 seconds. On entire range of LFM flow rate is unparalleled by other liquid flow device in the market.
Further, the manufacturing factory will set PID values to allow the end user to install the hardware and without any changes or modifications to the hardware, making this LFM a truly a “plug and play” devices.
Vaporization test of liquid is for examining total amount of carrier gas flow during the new process development, exception of this, liquid line clogging can be used for troubleshooting by using vaporization test.
- 76 -
Figure 3.5 TEOS vaporization test.
3.2.3 Pump systems
There are a lot of by-products produced during high temperature process reaction. Most of the byproduct shall be pumped through gate valve and throttle valve to dry pump. Reliable pressure control likes high process pressure (600torr), and high gas/liquid flow rate (80slm) is well by using flapper-type throttle valve with D-Net control. The faster pressure servo and accurately pressure control are very important to HARP process. The slower pressure servo will induce particles and low throughput effect.
- 77 -
Figure 3.6 Pumping line systems.
Figure 3.7 Basic concepts of NDIR endpoint detector.
- 78 -
3.2.4 NDIR end point detector
The NDIR endpoint unit is configured on the Producer 300mm chamber with a bypass valve to provide a consistent measurement of SiF4 during the chamber clean process (fig-3.6). The endpoint times calculated is using slope of the endpoint signal, vary by approximately 10% and are very consistent within that range. Additionally, the endpoint times provided by the NDIR endpoint unit and slope algorithm result in a clean chamber over extended runs [53].
Optical methods that look at the effluent gas during the clean are the most used. Of the optical methods, there was a possibility of optical emission or optical absorption. Optical emission requires generating plasma in the gas stream and looking at that light through a “window”. Optical absorption uses two windows and its own light source. It shines light through the gas stream and looks at what comes out. In either case, the window problem exists and by using optical absorption, the problem of generating a consistently plasma will be eliminated. In fairness, the problem of providing a consistent light source is not an easy task. Another difference between the two technologies is that the optical absorption unit looks for the presence of fluorine, while the optical absorption unit looks for SiF4.
Currently, the NDIR endpoint unit is being developed to produce a signal to tell when the clean process is completed and it can perform this task well.
This simple endpoint function is good and offers significant advantage over a
“timed” chamber cleaning process. The real advantage of this product is its
- 79 -
potential to provide more data about the cleaning process. For example, there have been many questions about the “humps” in the SiF4 trace.
Figure 3.8 NDIR end point graphic monitoring.
What do they represent? Can anything be done to change these humps?
What happens when the chamber hardware is changed? During the time the NDIR unit has been developed for endpoint, it has also offered information on things like alternative vendor faceplates and information on different process kit parts. In fact, the NDIR endpoint is being included as an analysis tool for the development of a new chamber pumping ring that which is attempting to minimize the deposition in “slow cleaning locations”.
- 80 -