Patent Application
20080142046
Low-pressure chemical vapor deposition (LPCVD) plays a critical role as part of sequence of steps in the fabrication of complementary metal oxide semiconductor (CMOS) integrated circuits. Silicon nitride is typically deposited in LPCVD by reacting dichloro-silane (DCS) and ammonia in a hot-wall reactor. The primary driving force for the reaction is the thermal energy from the reactor that can operate at temperatures between 650-850° C. However, as technology nodes proceed from 130 nm to 32 nm, the thermal budget becomes a serious problem. To circumvent this problem, new LPCVD precursors that require less thermal energy to react are being introduced. For example, as an alternative to using DCS and ammonia to deposit silicon nitride films at 700° C., organic amine substituted silanes such as Bis(Tertiary Butyl Amino) silane (BTBAS), diisoprpyl amine silane (DIPAS), and diethyl amine silane (DEAS) or organic substituted silane such as tetra allyl silane, and trivinyl silane with bis-tertiary butyl amino silane reacted with ammonia at 600° C. or a lower temperature is gaining favor because it significantly reduces thermal budgets. Silicon nitride films produced by this process not only deposit on the wafer, but also on the walls of the quartz reactor and reactor components. Moreover, the nitride films produced by this process have the high tensile stresses which can lead to high levels of particle contamination of device wafers from film spalling due to cumulative deposits in the reactor. In order to prevent wafer contamination, nitride furnaces must be cleaned before the cumulative deposits start spalling. Higher stress films require more frequent cleaning of the LPCVD reactor and wafer holder due to particle generation caused by the film spalling.
The current practice is to manually clean the LPCVD reactor of silicon nitride deposits by cooling the quartz tubes to room temperature, then removing them and placing them into a wet HF etch. In all, the wet clean procedure requires 8 to 6 hours of equipment downtime. Compared to a DCS nitride furnace that only requires cleaning after 60 to 90 days of production, BTBAS nitride furnaces typically require cleaning after every two days of operation. Thus, for the organic amine substituted silanes or the organic substituted silanes nitride process to be practical in volume semiconductor manufacturing, it will require an in-situ clean with a 2-3 hour cycle time. Moreover, depositions of organic amine substituted silanes or organic substituted silanes require temperatures in the range of 500-600° C. Therefore, clean time which is a function of reactor temperatures which are around range of 500° C. to 600° C. becomes increasingly important.
D. Foster, J. Ellenberger, R. B. Herring, A. D. Johnson, and C. L. Hartz, “In-situ process for periodic cleaning of low temperature nitride furnaces,” in Proceedings of the 204th Meeting of the Electrochemical Society, Orlando Fla. (The Electrochemical Society, Inc., Pennington, N.J., October 2003) p. 285-293 (the subject matter of which is incorporated by reference) disclosed 20% NF3 has been used to etch SiNx in this temperature range, but the etch is relatively slow (0.02 μm/min at 550° C. and 30 Torr). For the organic amine substituted silanes or organic substituted silanes nitride process to be practical in volume semiconductor manufacturing, a more efficient cleaning method with lower thermal activation temperatures around 500° C. to 600° C. is needed.
One embodiment of a method according to the current invention comprises flowing pre-diluted fluorine in an inert gas through the chamber and maintaining the chamber at an elevated temperature of 230° C. to 565° C. to thermally disassociated the fluorine, thereby cleaning the CVD chambers by removing the volatile reaction products SiF4 formed by the chemical reaction of thermally activated fluorine with the undesired silicon nitride. According to another embodiment of the invention, the elevated temperature of 450° C. to 550° C. is used.
A schematic diagram of an experimental embodiment of the current invention is shown in
Silicon nitride films deposited with DCS were approximately 1 micron thick silicon nitride on a layer of SiO2. The silicon nitride wafers were cleaved into small coupons 3 about 2 cm by 2 cm. Prior to placing a coupon into the reactor, the samples were first cleaned in an ammonia-peroxide solution (RCA-1 clean) at 70° C. for 10 minutes to remove any organic contamination. The coupons were then placed into a 0.5% HF solution for 5-10 seconds to remove any surface oxides that may have built up. The samples were then rinsed, dried and placed into the reactor.
Silicon nitride samples are placed in the middle of the thermal reactor on a tray with one end slightly elevated (˜5 mm) off of the tray and the face of the coupon parallel to the gas flow. The end of the reactor tube is then sealed by replacing the conflat vacuum flange. Several pump/vacuum cycles are performed to remove atmospheric gases from the reactor. The reactor is then purged with 100 sccm of nitrogen. Once the chamber is purged, the furnace 4 is turned on and is programmed to reach operating temperature in 2 hours. Once the furnace reaches the set temperature, two additional hours are used to ensure the internal temperature reaches the target. The internal temperature is monitored with a thermal couple well that sticks into the chamber. Once the internal temperature of the reactor is at its target, the nitrogen valve is closed and the system is allowed to pump down to <100 mtorr. After the base vacuum level is reached, fluorine is introduced into the reactor by opening the fluorine valve. The wafer is then etched anywhere from 1 minute for the more aggressive etches (higher temperatures and pressures) to over 10 minutes for the less aggressive etches (lower temperatures and pressures). The etch is stopped by closing the fluorine valve and immediately opening gate valve 7 to fully evacuate the chamber and allowing the volatile reaction products to be completely pumped out from the chamber. The etch time is determined by the length of time that the silicon nitride is exposed to the fluorine gas. The reactor is then allowed to cool before the sample coupon is removed. In the real operation environment, the etching and the cleaning are performed in the typical operation conditions.
Silicon nitride samples were analyzed by reflectometry before and after etching to determine etch rate through change in film thickness. The etch rate is then calculated by dividing the change in thickness of material in nanometers by the etch time.
Dilute (no greater than 20%) molecular fluorine is used because DOT regulations restrict pure fluorine to be shipped in cylinders with pressures no greater than 400 psig. Using fluorine diluted with nitrogen or another inert gas decreases the hazards of fluorine, while maximizing the quantity of fluorine that can be shipped. This allows for the use of large quantities of fluorine for chamber cleaning without having the need for an onsite fluorine generator.
In the first experiment, the reactor tube was maintained at 4000 C temperature and with 30 torr pressure. The different concentrations of pre-diluted F2 in N2 were introduced into the reactor tube. The results of thermal etch rate measurements for silicon nitride (SiNx) etched with pre-diluted F2 in N2 as a function of F2 concentration are given in Table I.
The results show that the dilute (no greater than 20%) molecular fluorine has a low thermal activation temperature. The F2 reacts with the silicon nitride to form SiF4 that can be pumped from the chamber. The etch rates are 17 nm/min for 2.5% F2, 29 nm/min for 5% F2 and 155 nm/min for 20% F2. The results show that the etch rate increases as the F2 concentration increases at the fixed temperature and pressure. The results further show that even at a very low concentration of 2.5% F2, the etch rate is 0.017 μm/min at 4000 C, which is comparable with the etch rate of 0.02 μm/min from 20% NF3 at 5500 C.
To further determine the thermal etch rate of silicon nitride by pre-diluted 20% F2 in N2, a design of experiment (DOE) study was carried out. The parameter space of the DOE study covered a temperature range of 230° C. to 511° C. and a pressure range of 10 to 103 torr. A total of 12 silicon nitride etch rates were determined at various temperatures and pressures.
The results of thermal etch rate measurements for silicon nitride etched with pre-diluted 20% F2 at various temperatures and pressures are given in Table II. The results again show that the dilute (no greater than 20%) molecular fluorine has the low thermal activation temperature, etching of silicon nitride occurs even at a low temperature such as 2300 C.
The data shows that the etch rate is strongly dependent on temperature and, to a lesser extent, pressure. The etch rate is relatively low below 3000 C and increases rapidly to >600 nm/min at 5000 C and 100 torr.
Based on the second experiment, the etch rate with pre-diluted 20% F2 at various temperatures with a fixed pressure was further investigated. In this experimental set, a series of thermal etch rate experiments were carried out with 20% F2 in N2 at 30 torr with temperatures ranging from 300° C. to 550° C. The experimental data of the etch rate versus temperature is plotted in
The etch rates for 20% F2 are 53 nm/min at 3000 C, 139 nm/min at 4000 C, and increasing rapidly to 965 nm/min at 5500 C. The data shows that the etch rate increases exponentially as the temperature increases, as evidenced by the solid line of exponential fitting in
While the silicon nitride etch rate for 20% NF3 also shows an exponential increase, the window of this increase for 20% NF3 (˜5800 C) is approximately 200 degrees higher than for 20% F2 (˜3700 C). This is significant since the next generation silicon nitride deposition processes will take place at temperatures considerably below 5800 C.
As indicated by the graph in
The experimental results show that the dilute (no greater than 20%) molecular fluorine has much lower thermal activation temperature and higher etching rates. Therefore, the dilute (no greater than 20%) molecular fluorine provides more efficient cleaning of equipment surfaces of undesired silicon nitride in semiconductor processing chamber with lower thermal activation temperatures around 300° C. to 600° C.
To assess the potential damage that etching can cause to quartz reactors, experiments were carried out examining the effects of 20% F2 and 100% NF3 on quartz under the conditions of Example II. Weight loss and surface degradation of flame polished quartz (SiO2) were measured following thermal F2 and NF3 exposure. These measurements provide an estimate of nitride selectivity and illustrate the potential for damage to the quartz reactors by 20% F2 and 100% NF3.
A summary of the results is given in the Table IV. All quartz samples were etched for 20 minutes except for the sample etched with fluorine at 550° C. which was etched for 10 minutes. Upon visual inspection, quartz pieces etched with 20% fluorine exhibited appearances ranging from smooth and slightly hazy for the sample etched at 400° C. to very hazy for the sample etched at 550° C. Alternatively, etching with 100% NF3 causes discoloration of the quartz pieces (leaving them with a brownish appearance) in addition to causing them to become hazy. This quartz etch rate data coupled with the silicon nitride etch rate data above indicate that the non-desired etching of quartz is similar for 20% F2 at 4000 C as it is for 100% NF3 at 5500 C, while the etch rate for silicon nitride is much higher for 20% F2.
While specific embodiments have been described in details, those with ordinary skill in the art will appreciate that various modifications and alternatives to those details could be developed in light of the overall teaching of the disclosure. Accordingly, the particular arrangements disclosed are meant to be illustrative only and not limiting to the scope of the invention, which is to be given the full breath of the appended claims and any all equivalents thereof.
