Lam Research Corporation's patent: DYNAMIC PROCESS CONTROL IN SEMICONDUCTOR MANUFACTURING
This patent is that provides a method and system for dynamic process control in substrate processing, for example in semiconductor manufacturing applications. Some example systems and methods are provided for advanced monitoring and machine learning during atomic layer deposition (ALD). Some examples also involve dynamic process control and monitoring of chamber parameter matching and gas line charging time.
Currently, many parameters related to substrate processing chambers are monitored to run near component set working points. For example, the flow rate or chamber pressure of the mass flow controller (MFC) may include a certain error margin. Typically, the upper or lower limit of the parameter can be set to a range value or a specific percentage to satisfy this error. For example, in the atomic layer deposition process, the valve opening and closing times can be monitored and these times are counted into the monitoring parameters accordingly.
However, in general, current process monitoring methods are only suitable for detecting relatively widespread or severe failures of processing chambers or their components, such rough detection may be acceptable in steady state or single step situations, such as in chemical vapor deposition (CVD) or plasma enhanced chemical vapor deposition (PECVD) processes, but are limited in use or applications in multi-step processes, such as ALD, where chamber conditions can vary within a few milliseconds.
In an example embodiment of this patent, a system is provided for monitoring processing cycles during atomic layer deposition (ALD) semiconductor manufacturing process. An example system includes a processing chamber for the ALD manufacturing process; and one or more controllers configured to perform process monitoring operations, including: defining a reference time reference for the ALD cycle based on repeated actions in the manufacturing process; accessing a golden curve, including a series of parameter values of a series of data points based on periodic time increments of the time reference; accessing a variability or tolerance margin for each data point in the golden curve; collecting parameter data based on periodic time increments of a periodic period during the atomic layer deposition manufacturing process; dynamically monitoring whether the parameter values in the parameter data are within the variability or tolerance range of the data point; adjusting the manufacturing process to match the parameter values in subsequent cycles with the associated parameter values in the gold curve based on determining whether the parameter values exceed the variability or tolerance margin.
In some examples, the repeated actions forming the reference time reference base include opening or closing the designated valve of the supply processing chamber. In some examples, parameter data is collected periodically based on the collection frequency, which is in the range of 0-1 Hz, or 1-10 Hz, or 10-100 Hz, or 100-1000 Hz.
In some examples, the fixed interval is based on trigger points in the atomic layer deposition manufacturing process, each trigger point defining or based on one point in a certain step in the atomic layer deposition manufacturing process.
In some examples, the operation also includes comparing the parameter data collected at the trigger point with the corresponding parameter data set in the golden curve. In some examples, the parameter data includes parameter values associated with one or more precursor manifold pressure, purge pressure, conversion manifold pressure, chamber pressure, airflow, RF reflected power, and RF forward power.
Figure1
In the above figure, substrate support component 107 usually provides two or more RF frequencies. For example, in various embodiments, the RF frequency may be selected from at least one frequency of approximately 1 MHz, 2 MHz, 13.56 MHz, 27 MHz, 60 MHz, and other frequencies as needed. Coils that can be designed to block or partially block a particular RF frequency as needed.
Process parameters fluctuations will also make process control and monitoring difficult in other fields. For example, changes between substrates (or batch to batch) may be caused by accumulation of chamber heat during substrate processing. The differences between tools may be caused by differences in pump efficiency. Traditionally, the main work of controlling change has focused on monitoring the performance of a single device.The example device and its associated parameters may include the MFC stream, where the device error limit is set to 1% of the stream. During substrate processing, the MFC stream is monitored to operate within this limit. Other devices and parameters may include valve timing (eg, monitoring the timing of the ALD valve to run an opening time of 50ms and a closing time of 70ms).
In another example, the valve may be set to switch between the open and closed positions of 25 ms. In other examples, the base temperature can be controlled using thermocouple monitoring deviations within the set range. RF power control may include monitoring of forward and reflected power. These devices usually have inherent performance or response limitations that can lead to poor chamber control and random or fluctuating chamber conditions. Inadequate monitoring limits and multiple processes and equipment limitations can lead to this adverse effect.
Figure 6
In addition, referring to Figure 6, unlike most PECVD processes, parameters such as gas flow, pressure and RF power are usually kept constant throughout the deposition process, while certain parameters in each cycle change continuously during the ALD process (cycle). The table 600 in Figure 6 shows example steps and related parameters in the ALD loop. These steps may include dose, post-dose purging, RF power application and purging, etc.;
monitoring process (rather than equipment) parameters routine work includes monitoring chamber pressure around the set point of the set error band. The error band is usually set large enough to ignore the inherent fluctuations occurring during the ALD cycle. Therefore, these efforts do not really monitor more detailed aspects or chamber conditions in the atomic layer deposition cycle. Detailed and deep chamber controls are increasingly used to create high-side nano-size substrate formation and semiconductor devices. In addition, the precursor manifold and blasting pressure are also usually monitored around a frequency band. Conventional error bands are usually set too wide to capture smaller fluctuations during the atomic layer deposition period, resulting in similar challenges discussed above.
In terms of the provided RF power, a one-time check of RF power is just to check whether the RF is on or off after the RF impact. The voltage-current (VI) sensor monitors RF power during the plasma "on" step and simply monitors at a frequency higher than the RF power (e.g., 1 kHz). Therefore, in a broad sense, the current method is based on limit or error band settings. It is a passive or "stupid" method of monitoring, usually based on limited data. Generally, the monitoring frequency band is very wide, which does not solve or even solve the increasingly demanding process control problems in today's semiconductor manufacturing. Among other drawbacks, the same monitoring band is suitable for all tools and no tool-to-to-tool modifications or customizations are made. Usually, any customization is performed manually temporarily. There are few or no substrate-to-substrate or tool-to-to-to-to-to-to-to-to-to-to-to-to-to-to-to-to-to-to-to-to-to-to-to-to-to-to-access, after preventive maintenance of , or after hardware replacement, usually performed in conventional techniques.
Figure 14
As mentioned above, the atomic layer deposition process can be considered as a multi-step process. Referring to Figure 14, a typical ALD cycle 1400 includes four main steps: dose 1402, purge 1404, conversion 1406, and purge 1408. In some current examples, each step in the ALD loop and subsequent loops in a given ALD procedure are individually monitored for different variables. During each atomic layer deposition step and/or cycle, the monitoring variables are matched using the curve fitting or the error margin defined in the "intelligent" self-learning monitoring process. Each atomic layer deposition cycle is repeatable and can monitor the repeatability of parameters such as chamber pressure, precursor manifold pressure, temperature, etc. for each cycle or step (and make it repeatable).
