Patent Application
20240044039
The present invention is directed toward systems for monitoring and controlling the quality of plating processes including bath composition, state of hardware readiness and fingerprint of the processing of the product (substrate) using electrolytic or electroless deposition, or electrochemical etching, techniques.
The prior art for the process control of electrochemical deposition or etching processes involves evaluating each process parameter individually with respect to its control limits. For example, bath analysis tools are used for monitoring the concentrations of each of the constituents of an electrolyte and the deposition tool is used to monitor the deposition time, the bath temperature and such process parameters. This complex industrial use is a univariate approach which leads to high rates of false alarms.
Many products are produced which involve coatings that are deposited on some part of the product using electrochemical deposition techniques. These techniques, as described below, are used to deposit layers of materials such as metals and alloys on a substrate part. The substrate part may comprise a metallic part, a plastic part, a printed circuit board, a wafer for fabricating microelectronic devices, or some other such substrate.
Microelectronic devices are manufactured by depositing and removing multiple layers of materials on a substrate such as a silicon wafer to produce a large number of individual devices. For example, layers of photoresist, conductive materials, and dielectric materials are deposited, patterned, etched, planarized, and so forth to form features in and/or on a substrate. The features are arranged to form integrated circuits (ICs), MicroElectroMechanical Systems (MEMS), and other microelectronic structures.
Wet chemical processes are commonly used to form features on microelectronic substrates. Wet chemical processes are generally performed in wet chemical processing tools that have multiple processing chambers for combinations of cleaning, etching, electrochemical deposition, and rinsing. Electrochemical deposition processes include electrolytic deposition, in which current is applied to the substrate, and electroless deposition, where no external current is supplied to the substrate. In these processes the substrate is immersed in a chamber that includes an electrolyte that supplies the material to be deposited on the substrate. It is also possible to remove material from the substrate by reversing the current through the electrolyte and substrate, or by supplying an appropriate etching solution. These processes can occur in wafer fabrication equipment such as the Raider® system offered by Applied Materials, the Sabre® system offered by Lam Research, the Stratus® system offered by Tel/Nexx, or similar processing tools.
It is customary to include the ability to monitor the chemical constituents or chemical activity of an electrolyte that is being used in one of the aforementioned processes. This is done in order to control the concentrations of the chemical constituents of the electrolyte in such a way as to maintain consistent processing of multiple substrates as the electrolyte is used throughout its lifetime. Commonly, analytical techniques must be developed specifically for each component of an electrolyte in the particular combination and concentrations of constituents that are present in that particular electrolyte. It can be time-consuming and expensive to design the analysis techniques or methods using electroanalytical, HPLC, and titration techniques for each of the constituents of an electrolyte. Furthermore, some commercial additives have two or more components and it may not be possible to separate them without prior knowledge of the chemical species. Monitoring of each of the chemical constituents of these electrolytes can be complex due to the multiple interactions that can occur between the constituents as their respective concentrations change and as the electrolyte ages. Aging of the electrolyte may result in the breakdown of one or more of its constituents due to oxidation, reduction, or catalytic activity, or it may involve interactions with the hardware or substrates that come into contact with the electrolyte over its lifetime, or other contamination.
Several methods of monitoring the constituents of an electrolyte involve electroanalytical methods. One limitation of using these electroanalytical methods in a classical approach is that they require some prior understanding of the appropriate operating range of the chemistry and system before they can be used to determine if the chemical constituent concentrations are “within specification” or not. Commonly, these concentration ranges are used to produce a “training set” of chemical solutions by varying the concentrations of multiple constituents of the electrolyte and by analyzing each of the training electrolytes to determine the impact of changing the concentrations on the analytical result. Such a method can very quickly result in an inordinate number of samples to be used as the training set, due to the number of constituents in the electrolyte and the multiple interactions between these constituents and the responses being generated to represent the concentrations of other constituents. Classically, a separate electroanalytical method is utilized to represent the concentration of each known constituent within the electrolyte. These electrochemical analyses can occur within chemical analysis equipment such as the Real Time Analyzer (RTA) offered by Technic, Inc., the Quali-Line® system offered by ECI Technology, the Ancolyzer® system offered by Ancosys, or similar analysis tools.
One approach that is commonly used to simplify electroanalytical methods and use them to represent the concentrations of chemical constituents is to choose one particular aspect of the electrochemical response and to correlate that aspect to the constituent concentration. The aspect of the response that is utilized may be the height of a particular response peak, the area under a particular response curve, or the ratio of the heights of two response peaks, for instance. While this approach makes it simpler to correlate the specified parameter to the changing concentration of one or more of the electrolyte constituents, it results in ignoring a large proportion of the data that is generated by an electroanalytical technique. This means that the opportunity to evaluate changes in the response in the regions not being considered is lost. These regions may contain information related to breakdown products, chemical contaminants, and such.
A further limitation of the common approach of using a training set of solutions with varying constituent concentrations is that they are commonly made by mixing fresh constituents together in order to produce multiple solutions with concentrations that vary in understood ways. This approach neglects any impacts of electrolyte aging on the analytical results. Electrolyte aging can result in oxidation and/or reduction of one or more of the electrolyte constituents as they come in contact with the anode and/or cathode of an electrolytic system. It can also include the build-up of components that were not intentionally added to the electrolyte due to leaching from the system components or from the product substrates (or holders) that are processed by the system. There are also other ways that contaminants could be inadvertently introduced to the electrolyte. Electrolyte aging is an important aspect of most industrial deposition processes and can be responsible for many of the problems that occur on products and materials that are produced utilizing these wet deposition methods. The importance of maintaining chemical constituent concentrations despite these bath aging effects can be appreciated by understanding the use of “bleed and feed” techniques that essentially throw away a proportion of the aged electrolyte and replace it with fresh electrolyte in an attempt to keep the concentrations of compounds associated with bath aging below acceptable limits. These techniques are common in the microelectronics industry, where they are used to maintain the electrolytes of deposition processes at steady-state conditions with low levels of breakdown products and/or bath contaminants. Of course, using bleed-and-feed operation of an electrolyte increases the chemical usage, environmental impact, and associated costs.
The classical methods of electroanalytical analysis involve the extraction of a sample of the electrolyte from the system for analysis. This entails flowing a portion of electrolyte through fluid sampling lines to the point of extraction in order to ensure that the extracted sample has a composition consistent with the electrolyte being used in the processing of product. Some of the electrolyte is also commonly used to rinse the analysis equipment in order to prevent analyzer contamination or “memory” from previous samples. If the analysis frequency is high, or if the volume of electrolyte required for flushing the system and analysis is large, the operation of the analysis system may effectively produce a bleed-and-feed type of operation due to the volume of electrolyte being consumed for analysis purposes, which must be replenished with fresh electrolyte.
The aforementioned electroanalytical methods may be complicated due to changes in reference electrodes or working electrodes used during the analysis. Reference electrodes must be chosen so as to be compatible with the electrolytes they are in contact with, and they must be maintained over their lifetime in order to maintain consistent results. Working electrodes are commonly comprised of rotating disk electrodes which involve electromechanical rotating mechanisms, and are subject to changes in surface roughness or electrode integrity over time.
When a rotating disk electrode is used as the working electrode in an electroanalytical method the electrode is typically cleaned and conditioned in order to maintain consistent results from one measurement to the next. Even so, it is important to monitor the results and frequently calibrate and/or to eliminate electrode drift effects and to know when it is time to implement a more rigorous maintenance procedure or to replace the electrode. Changes in electrode surface area, for instance, can cause changes in the measured response which are not due to changes in the chemistry.
U.S. Pat. No. 5,223,118 teaches a direct method of analyzing the quantity of brighteners and levelers in an electroplating bath. According to this method, a plurality of plating baths where each bath has a known and different quantity of brighteners and levelers are produced. In the following, for each bath, a counter electrode, a cleaned working electrode an a reference electrode is provided and immersed in the bath and for each bath the measured energy output value is correlated with the quantity of brightener, the initial energy output and the change in energy output. Based on these values the quantity of leveler and brightener is determined in a bath of unknown quantities of brightener and leveler. This method uses different working electrodes, namely one working electrode for each of the plurality of plating baths. The working electrodes are suitable metal disks.
EP 0 760 473 A1 relates to a method for evaluation the metal surface state after plating and deposition and during corroding. A metal like aluminium which is already deposited on a substrate is contacted with a solution containing ions to corrode the metal. The electrode potential of the Aluminium alloy layer which is a metallic thin layer for wiring on a semiconductor wafer serving as a semiconductor substrate is measured in the solution containing ions to corrode the metal.
It is the object of the present invention to provide a method for monitoring and controlling the quality of the electrochemical deposition and/or plating processes including bath composition, process chamber and/or the film formation process on the substrate.
The invention relates to an electrochemical analysis method that utilizes one or more substrates as working electrode(s) and analyzes the potential between the one or more working electrodes and one or more reference electrodes during the electrochemical deposition an/or plating process to provide an output signal fingerprint which is represented as potential difference as a function of time. The electrochemical analysis method uses a fingerprinting analysis method to have an indicator of whether the chemistry and/or process is operating in the normally expected range. The method utilizes one or more substrates of the deposition and/or plating process itself as working electrode(s) of the electrochemical analysis method.
Preferably the one or more substrates in this first embodiment comprises one or more substrate wafers or printed circuit boards.
Furthermore, a source of input power can be connected between the one or more substrates and one or more counter-electrodes. The additional source of input power is used to augment the process analysis by supplying an additional source of current and/or potential variation which is used to aid in analysis by generating a waveform that provides information beyond that which would be provided by the process alone.
It is also possible to use the current or potential between one or more substrates and one or more counter-electrodes as an additional input signal.
The invention also relates to an electrochemical analysis method that utilizes one or more substrates as the working electrode(s) and the input power of a process power supply to provide input energy in the form of current and/or potential between the working electrode(s) and a counter-electrode. The method utilizes the potential between the one or more working electrode(s) and at least one of: one or more reference electrodes; or one or more counter-electrodes; to provide an output signal. In this embodiment preferable the one or more substrates comprises one or more substrate wafers.
The electrochemical analysis system may use a fingerprinting analysis method to have an indicator of whether the chemistry and/or process is operating in the normally expected range.
Preferable, an additional waveform of input power is superimposed with the process waveform. The additional waveform is used to augment the process waveform by supplying an additional source of current and/or potential variation which is used to aid in analysis by generating a waveform that provides information beyond that which would be provided by the process waveform alone.
The output(s) of the electrochemical analysis method can be fed into a model that is used to produce a quantitative quality predictor which provides information about whether the chemistry and/or process is operating within the expected range or if it is approaching or has entered an abnormal range of operation.
The present invention is a method of using trace data (electrolyte flow, temperature, substrate rotation or agitation speed) as well as electroanalytical data (plating or etch current, voltage at the substrate, cell voltage) to fingerprint an electrolyte and/or a system for electrochemical deposition or etching to be used as quality monitors for a) readiness of the process chamber including all connections to the substrate, to anodes and such; b) fingerprinting of the film formation process performed on substrates such as silicon wafers or printed circuit boards (PCBs) used to fabricate microelectronic or electronic devices.
The present invention includes a fingerprint method according to the definition of the International Sematech Manufacturing Initiative (ISMI) after which “Fingerprinting” is “A set of data variables associated with the component being fingerprinted sampled at some rate over some period of time, transformed and analyzed using a set of mathematical techniques to generate a result representing the state of the unit during that timeframe.” Key characteristics of a fingerprint, therefore, are: a) A well-defined fingerprint points directly to the failing (or drifting) component when its values are out of the normal operating range. b) In contrast with most Fault Detection Classification (FDC) systems, the drill-down process can begin before a fault has been detected. C) Fingerprints often use data that is not generally available to the fab customer and models that require very specific equipment domain knowledge.
The invention preferably utilizes the substrate as the working electrode for the analytical technique. The invention allows the working electrode to be replaced for every measurement by utilizing the substrate itself as the working electrode for each electroanalytical measurement. This technique provides the advantage of always knowing that the working electrode is exactly representative of the product due to the fact that the product is actually used as the working electrode.
A further aspect of the invention provides for a parallel circuit to allow an additional signal to be superimposed upon the waveform that is applied to the substrate during the deposition process. In this form of the invention a signal such as a small amplitude current or voltage sweep can be added to the waveform used for deposition. This superimposed signal is used to modify the input energy used to provide an output signal which is used to monitor the electrochemical deposition system.
In yet a further aspect of this invention, the output(s) of the electrochemical analysis method are fed into a model that is used to produce a quantitative quality predictor which provides information about whether the chemistry (bath health factor) and/or process (equipment health factor) is operating within the expected range or if it is approaching or has entered an abnormal range of operation. The system uses a quality predictor in feedback with a deposition system in order to provide control of the deposition process by adjustment of the process power supply, therefore, the output signal can be used to provide feedback which is used to control the power supply used for the deposition or etch process, as well as the teach points for substrate entry into the electrolyte. (e.g. “Hot entry,” where potential is applied to the substrate before it is brought into contact with the electrolyte) Power supplies and substrate handling devices are key components for achieving the desired process results. In this embodiment, the output of the power supply can be varied in order to compensate for a measured characteristic of the analysis in order to compensate for changes that occur through the life of the electrolyte once a correlation to the metrology of the product and/or test substrate is achieved. Alternatively, the feedback may be used to suspend the operation of the process if an output limit associated with poor process performance is approached. In this manner the invention may advantageously be used to extend the life of an electrolyte used in a process or to inhibit the production of inferior products with a process that has drifted away from its ideal operating conditions.
In a final embodiment of this invention, the method may use a quality predictor in feedback with a deposition system in order to provide control of the deposition process by adjustment of the process power supply. The model and quality predictor are designed to learn over time in order to distinguish between normal and abnormal behaviour. Product quality parameters are fed back into the model to help it refine the differentiation between good and bad process results, and to allow it to narrow down the differences between good and bad process results as more data is fed back to the model.
The information collected is used to produce a quality factor or equipment health factor. Factors could comprise any combination of trace data in the way of multivariate statistical approaches like one class support vector machines (OC-SVM), robust principle component analysis (ROBPCA), and so on. The quality factor may be created from electroanalytical results, classical chemistry analysis results, comparisons to expected electrical waveforms, and the like. For electrochemical deposition or electrochemical etching processes the trace data may be grouped to allow the following classes of analyses: a) readiness of the process chamber including connections, anode quality, flow rates or agitation, and so on; b) behavior/readiness of the power supply; c) quality of the substrate (especially seed layer, photoresist, open area); d) behavior of the electrolyte determined via electroanalytical analysis results. The quality factor is then monitored in order to ensure that the products being produced are of sufficient quality. If the quality factor/health factor is seen to be trending toward an unacceptable limit, action can be taken to modify the process in order to improve the quality factor. Correlation between fingerprint data and yield data of the product are mandatory in order to optimize troubleshooting and/or root cause analysis.
One further aspect of this invention is the use of a high speed signal processing device to handle data in the range of milliseconds to nanoseconds. This is particularly helpful for switching on or switching over of a process parameter (for example, deposition current). Required data rates depend on the process parameter being monitored, but the capability of the high speed signal processing device should span a data acquisition range from nanoseconds to minutes.
The present invention is a method of using electroanalytical techniques to fingerprint an electrolyte and/or a system for electrochemical deposition or etching and creating a quality monitor, or health factor, that represents: a) readiness of the process chamber including all connections to the substrate, to anodes and so on and/or b) fingerprinting of film deposition processes performed on microelectronic substrates such as silicon wafers or printed circuit boards or interposers or any other form of electronic substrates. The invention preferably utilizes the substrates as the working electrode 11 for the analytical technique. A schematic representation of one embodiment of the present invention is shown in
In the regions above the upper line and below the lower line the expected quality of coated substrates are not good.
After having defined the quality of different substrates and the current Ip and/or the voltage Ucell or UR for several substrates, the quality of a further substrate can be predicted by measurement of Ip, or the voltage Ucell or UR of the further substrate.
Each new determination of the quality of the substrate and the voltage or current will lead to a revised model.
The current Ip, the voltages Ucell or UR are measured with the measuring instrument 16 and the voltmeters 17, 18 in
When the substrate is being used as the working electrode for an electrochemical measurement some limitations and complications arise which must be taken into account. It is common for a manufacturer of microelectronic devices to use the same facility and equipment to manufacture multiple product types. Each product type is likely to have a unique surface area and potentially different operating waveform due to its particular product requirements. Therefore, each product type will need to be compared separately, perhaps through the use of a lookup table or such, in order to have consistent results from the method of this invention. The parameters used to separate the product into categories may include product type, process chamber(s) utilized, previous process equipment parameters, and the like. The data may be considered with each value of each parameter separated out, or in an aggregated form considering all values of all parameters in a single data set. Additionally, there is likely to be variation within a product type due to normal manufacturing variations in the lithographic processes, for instance, which will need to be accounted for in the methods described here.
One advantage of using the product substrate as the working electrode in the gathering of electrochemical response data is that there is no need to fabricate or design a method that represents the process being performed on the product, because it is exactly this process which is providing the data used to analyze the process. There is no danger of the applied current (or voltage) being too low or too high to capture some characteristic performance of the chemistry that occurs only within a narrow range used on the product substrate, because it is that very process that is used to generate the data being used for analysis.
The example shown in
In order to utilize the product substrate as a working electrode in this method it is desirable to incorporate a reference electrode into the deposition system in a manner such that it is in electrochemical communication with the working electrode(s). In the foregoing discussion the product substrate was described as the working electrode. It is also within the scope of this invention to use the cell voltage and/or cell current (between the product substrate electrode and the anode(s), or counter-electrode(s)) as yet another set of trace data that indicates the performance of the system as a whole. In this respect, it would be feasible to have a relatively simple electrochemical processing system such as those shown in
As will be appreciated by one of ordinary skill in the art, a system as described in
When the system is used in a configuration that includes one or more ion-selective membranes 21 there may be an electrolyte in one of the fluid systems that comprises an acid solution. Acid solutions based on acids such as sulfuric acid or methane sulfonic acid are commonly utilized. When this is the case, it may be advantageous to monitor this solution in order to identify contamination or changes of the electrolyte. In this respect, an embodiment of this invention would be to use a fingerprinting analysis technique to identify contaminants in this electrolyte. This fingerprinting analysis technique may utilize an electrode in the deposition system as a working electrode for the analysis, and may optionally include a parallel electronic circuit for superimposing an additional analysis waveform, as described elsewhere in this application. Such contaminants may be introduced, for example, through migration of materials through one or more membranes that are part the system. This could be due to normal operation, the generation of a contaminant due to reactions in the system, or through a manufacturing problem with one or more of the electrolyte components that results in contaminated feedstock. This invention provides for a capability of identifying the presence of a contaminant such as an organic compound in an electrolyte that comprises an acid such as sulfuric acid or methane sulfonic acid. Examples of other acid electrolytes that may be tested with this method include formic acid, ethanoic acid, propanoic acid, butanoic acid, and sulfonic acids as aliphatic or nonbenzene alicyclic compounds of the general formula (I) wherein R is a C1-5 alkyl radical or carbon ring structure,
(X1)n—R—SO3H (I)
X1 is a halogen atom or hydroxyl, aryl, alkylaryl, carboxyl, or sulfonyl radical which may be in any optional position of the alkyl radical, and n is an integer of 0 to 3. Examples of these organic sulfonic acids are methanesulfonic, ethanesulfonic, propanesulfonic, 2-propanesulfonic, butanesulfonic, 2-butanesulfonic, pentanesulfonic, chloropropanesulfonic, 2-hydroxyethane-l-sulfonic, 2-hydroxypropane-l-sulfonic, 2-hydroxybutane-l-sulfonic, 2-hydroxypentanesulfonic, allylsulfonic, 2-sulfoacetic, 2- or 3-sulfopropionic, sulfosuccinic, sulfo-maleic, sulfofumaric, benzenesulfonic, toluenesulfonic, xylenesulfonic, nitrobenzenesulfonic, sulfobenzoic, sulfosalicylic, and benzaldehydesulfonic acids. Furthermore, as described elsewhere in this disclosure, the present method is capable of generating a training set of information during normal operation, and it is capable of refining its ability to identify contaminants over time by “learning” as additional data are generated and fed back to the system. A training set of contaminated and uncontaminated electrolytes can be generated as the method is used to analyze electrolyte samples. The training set may include at least one aged electrolyte. Aged electrolyte is produced by using the electrolyte in an electrochemical process. As described elsewhere in this application, it would also be apparent that a quantitative quality factor, or bath health factor, could be generated from the data produced by this method which could represent the quality of the electrolyte chemistry.
An additional aspect of this invention is that the model and quality factor are designed to learn over time in order to distinguish between contaminated and uncontaminated electrolytes. Electrolyte quality parameters are fed back into the model to help it refine the differentiation between contaminated and uncontaminated electrolytes, and to allow it to narrow down the differences between contaminated and uncontaminated electrolytes as more data is fed back to the model. The model can be created which represents the quality of the process output, and the system can learn the process behavior over time as the process continues to run. This allows a process to be monitored by this method without prior knowledge of the failure modes of the process. It is advantageous to provide feedback to the model when any problem occurs with the products being produced by the process. In this manner, the responses that were captured when the defective product was produced can be associated with the defect that is observed, and can be used by the system to set an appropriate limit that can be used in the future to flag the manufacturer when such a limit is approached. In this manner, the system can be used to prevent the re-occurrence of a failure once the failure information has been fed back into the model.
A model can be created from an unspecified number of process vectors by simply collecting the information for each vector associated with a particular product substrate as it is being processed. It is not necessary to know a prioriwhat values are desired for a particular vector. The information is captured, and the quality attributes of goodness or badness may be associated with the particular substrate as the information becomes available. In this manner, the system effectively “learns” which values are associated with good and bad product results over time.
In order to ascertain proper performance of the process trace data is monitored and analyzed. Collecting trace data is carried out through sampling or measuring at regular or irregular time instances and calculation of resulting quantities. In the first step feasibility sets for each point of the trace data are defined by means of test runs and engineers' expertise. This provides simple intervals for some data points and more complex sets for others, for example, compact ranges described by envelopes for time based trace data. A process is ranked among performing normal if each point of the trace data collection falls in its corresponding feasibility set. Equivalently, an outlier or abnormal behaving process, respectively, is identified by one or more points of the trace data being located outside of their region of feasibility.
In a next step, the occurrence of outliers is analyzed by identifying (local) critical ranges and quantities derived from the trace data which reflect the deviation from normally behaving processes most closely, involving the definition of key performance indicators (KPI). Based on those KPIs, a methodology is established which provides assistance for deciding a priori whether a prospective process is feasible or not. There is a large variety of well-established deterministic as well as stochastic classification models, ranging from linear classifiers like Fisher's linear discriminant and principal component analysis (PCA) to support vector machines (SVM), neural networks and vector quantization, which can be used as basis for the definition of the decision model.
Over time, the decision model can be refined and improved by both adapting decision thresholds or feasibility sets due to growing trace data collection, or training sets, and adding new KPIs because of a growing diversity of outliers, leading to an increased dimension of the model. An advantage of this method of operation is that a training set can be produced during normal operation. This allows the training set, or feasibility set, to be augmented and improved over time as the process runs, and it reduces the cost of process startup by not requiring an extensive training set of data to be generated before running product through the process.
Having to monitor processes in a number of chambers, a decision model is established for each chamber independently. In order to control and being able to compare the different processes in various chambers, the core of the individual decision models should be identical. Considering that processes for different products worked on in different or even same chambers are similar in some sense and, thus, are assessed by the same trace data, the same KPIs can be utilized in each chamber, regardless of the specific occurrence of outliers. That provided, the same decision model can be used throughout where only the feasibility sets, e.g., decision thresholds and envelopes, differ from chamber to chamber. By designing chamber specific transfer functions, the variation in the shape of the feasibility sets can be adjusted, which provides a tool for chamber matching and controlling processes for different products.
The foregoing discussion related mainly to the use of the disclosed systems and techniques for the analysis of electrolytic deposition processes, where an imposed current (or potential) is used to supply the energy needed to deposit material onto a substrate. In the class of processes known as electroless deposition and immersion deposition, the driving force for the deposition comes from electrochemical potentials within the chemistry or between the chemistry and the substrate. In these processes there is typically no external electrical circuit, and the substrate is electrically floating, or in open circuit with the electrolyte. In such a system, the foregoing invention can still be applied by installing a reference electrode in fluid communication with the substrate as it is being processed, and by monitoring the potential between the substrate and the reference electrode. The addition of a counter-electrode to the system also provides the ability to superimpose a low amplitude current or potential waveform on the zero current conditions typically associated with electroless or immersion deposition, as will be discussed below. Other aspects of this invention may advantageously be applied to electroless deposition or immersion deposition processes by the addition of an appropriate reference electrode to the system.
One of the main characteristics that differentiates this invention from any prior art is that the product substrate, such as a wafer, is utilized as the working electrode for an electroanalytical method. This means that data generated from the process itself is used to “fingerprint” or monitor the process responses and make sure they are within a normal operating range or feasibility set. Additionally, the electroanalytical results may be combined with other information collected from or about the system as inputs to a model that predicts the quality of the process output. This method of capturing electrochemical data from the product as it is being processed and effectively using the product itself as a single-use working electrode for electrochemical analysis is unique and has not been used in the electrochemical deposition industry to our knowledge.
In another embodiment, the main differentiator of this method of analysis is that characterization of a large number of solutions as a training set is not required in order to provide feedback on a process or electrolyte. The information regarding a chemistry or process can be gathered over time and compared to the previously acquired set of data. In this manner, it is possible to determine the normal range of the output data, and also to correlate abnormal data to undesirable chemical or process characteristics as these data are produced and identified over time.
The advantages that are obtained using this invention include the ability to begin running a process sooner, without waiting for the completion of extensive (and expensive) characterization of electrolyte analysis techniques to have bath component analyses available to be used for bath regulation or replenishment. They also entail the ability to get quick quality feedback while the process is running, on a per-substrate basis. The learning ability of the process also allows the control system to get better as the process is run for a longer period of time and the feasibility set is increased. The ability to monitor the electrolyte as it is being used, without extracting a sample for analysis, also provides the opportunity to reduce the consumption of chemistry and possibly to increase the bath life.
A further aspect of the invention provides for a parallel circuit to allow an additional signal to be superimposed upon the waveform that is applied to the substrate during the deposition process. An example of this is shown schematically in
The signal that is superimposed with the process waveform is designed such that it will not adversely impact the process results when it is added to the normal process waveform. Such considerations may lead to current or potential changes of up to 10% of the normal process current or potential. More preferably, the change in current or potential may be less than 1% or 2% of the normal process value. The process impact on the product will need to be determined on a case by case basis and will be dependent on the particular process being performed, the sensitivity of the product to variations in the waveform utilized during the process, and the particular waveform being superimposed on the process waveform. The superimposed waveform is to be determined in a way that provides additional information in the electrochemical analysis results without adversely impacting the product that undergoes the electrochemical deposition process in most cases.
The superimposed waveform will normally be chosen such that it provides a small change, relative to the normal process waveform. Exemplary waveforms that could be superimposed with the process waveform include, but are not limited to, low amplitude voltage or current sweeps with a triangular time-dependent shape, low amplitude voltage or current sweeps with a sinusoidal time-dependent shape, low amplitude voltage or current steps, low amplitude current modulation with varying frequency, or similar variations that provide current and/or potential outputs that can be used to provide electrochemical information about the process or chemistry being used.
The result of providing an additional waveform that is superimposed with the process waveform can be to provide the ability to perform additional electrochemical analysis on the process and/or on one or more of the electrolytes used in an electrochemical deposition or etching process. This could provide analysis results such as electrochemical impedance spectroscopy, cyclic voltammetry, step voltammetry, or the like. A suitable waveform or waveforms can be chosen so as to provide a set of electrochemical analysis results that are likely to be responsive to changes in the electrolyte or electrolytes being monitored, or to changes in the process or hardware that is used to perform the process. It is also expected that the analysis results would be responsive to variations in the incoming product as the electrochemical deposition process is being performed on such product.
An experiment has been performed in which a portion of a semiconductor wafer substrate was subjected to an electrochemical deposition process representative of a portion of an electrochemical deposition process used for depositing tin silver solder alloy, similar to that shown in
When this experiment was performed using chemistry that provided good deposits, the deposit was as expected, even after adding the small amplitude current variations to the normal DC current. Therefore, this same approach could be used with real product substrates.
As seen in
As suggested above, the electrochemical analysis methods described herein may be sensitive to variations in parameters such as: electrical contact resistance at the cathode or anode, seed layer thickness or resistivity, power supply output, product contamination or residues, chemical contaminants in the electrolyte, surface area subject to the process, or electrolyte constituent concentrations. The methods described herein may therefore be very powerful in monitoring changes in the electrochemical deposition process or in the electrolytes used for the process.
In yet a further aspect of this invention, the output signal can be used to provide feedback which is used to control the power supply used for the deposition process. In this embodiment, the output of the power supply can be varied in order to compensate for a measured characteristic of the analysis in order to compensate for changes (such as current efficiency) that occur throughout the life of the electrolyte, and/or as anodes are used or consumed. This mode of operation allows the process to be adjusted such that its useful life can be extended beyond what it would have been without such an operational scheme. The output of the analysis or a model that is based on the analysis data can be used to adjust the operating parameters of the process in order to keep the product within its specifications. For instance, if the electrolyte changes as it ages, the power supply output could be adjusted in a way that keeps the process producing the desired deposit properties on the product.
An example of this embodiment is a case where an electrolyte used for a plating process changes over time in such a way as to shift the potential required for depositing a desired metal. This behavior can be seen in
In yet another embodiment of this invention, the information collected is used to produce a quality factor that represents the quality of the deposition process, and also therefore of the deposit on the microelectronic substrate. The quality factor may be created from electroanalytical results, classical chemistry analysis results, comparisons to expected electrical waveforms, and the like. The quality factor is then monitored in order to ensure that the products being produced are of sufficient quality. If the quality factor is seen to be trending toward an unacceptable limit, action can be taken to modify the process in order to improve the quality factor or action can be taken to stop production until the process can be modified in order to produce product that has a more preferable quality factor.
It is further preferred that pH-electrodes are used as reference electrodes, as will be disclosed in DE 10 2012 106 831 A1. As pH-electrodes preferable glass electrodes or email electrodes are used, whereby a glass electrode according to this invention is a simple glass electrode that includes a leakage system 30, an electrode 31, an inner buffer or electrolyte 32 and a glass membrane 33, as shown in
The pH-electrodes are preferred as they do not contaminate the bath, as they do not have to be serviced or refilled, as they are longtime stable and do not change the potential after small currents applied to the electrode as the common Ag/AgCl-electrodes or Calomel electrodes do. Furthermore, pH-electrodes do not comprise a diaphragm, which often becomes defective due to depositions.
The invention relates to an electrochemical analysis system that utilizes one or more substrates as working electrode(s) and the input power of a process power supply to provide input energy in the form of current and/or potential between the working electrode(s) and one or more counter-electrodes. The system has the capability to analyze the potential between the one or more working electrodes and at least one of: one or more reference electrodes; or one or more counter-electrodes; to provide an output signal. Preferable one or more substrates comprise one or more substrate wafers or printed circuit boards. The electrochemical analysis system uses a fingerprinting analysis method to have an indicator of whether the chemistry and/or process is operating in the normally expected range.
An additional source of input power can be connected in parallel with the process power supply. The additional source of input power is used to augment the process power supply by supplying an additional source of current and/or potential variation which is used to aid in analysis by generating a waveform that provides information beyond that which would be provided by the process waveform alone. The output(s) of the electrochemical analysis system can be fed into a model that is used to produce a quantitative quality predictor which provides information about whether the chemistry (bath health factor) and/or process (equipment health factor) is operating within the expected range or if it is approaching or has entered an abnormal range of operation. The system can use a quality predictor in feedback with a deposition system in order to provide control of the deposition process by adjustment of the process power supply. The model and quality predictor can be designed to learn over time in order to distinguish between normal and abnormal behaviour. Product quality parameters are fed back into the model to help it refine the differentiation between good and bad process results, and to allow it to narrow down the differences between good and bad process results as more data is fed back to the model.
The invention also relates to an electrochemical analysis system that utilizes one or more substrates as working electrode(s) and has the capability to analyze the potential between the one or more working electrodes and one or more reference electrodes to provide an output signal fingerprint which is represented as one parameter as a function of a second parameter, said parameters being selected from the group of: potential difference, time, temperature, current, real component of impedance, imaginary component of impedance, frequency. Preferably the one or more substrates comprise one or more substrate wafers.
The electrochemical analysis system may use a fingerprinting analysis method to have an indicator of whether the chemistry and/or process is operating in the normally expected range.
A source of input power can be connected between the one or more substrates and one or more counter-electrodes. The additional source of input power is used to augment the process analysis by supplying an additional source of current and/or potential variation which is used to aid in analysis by generating a waveform that provides information beyond that which would be provided by the process alone.
Preferably the output(s) of the electrochemical analysis system are fed into a model that is used to produce a quantitative quality predictor which provides information about whether the chemistry (bath health factor) and/or process (equipment health factor) is operating within the expected range or if it is approaching or has entered an abnormal range of operation.
The system may use a quality predictor in feedback with a deposition system in order to provide control of the deposition process by adjustment of the process power supply. The model and quality predictor can be designed to learn over time in order to distinguish between normal and abnormal behaviour. Product quality parameters are fed back into the model to help it refine the differentiation between good and bad process results, and to allow it to narrow down the differences between good and bad process results as more data is fed back to the model.
The model and quality predictor are designed to learn over time in order to distinguish between normal and abnormal behaviour. Product quality parameters are fed back into the model to help it refine the differentiation between good and bad process results, and to allow it to narrow down the differences between good and bad process results as more data is fed back to the model.
An additional waveform of input power can be superimposed with the process waveform. The additional waveform is used to augment the process waveform by supplying an additional source of current and/or potential variation which is used to aid in analysis by generating a waveform that provides information beyond that which would be provided by the process waveform alone.
The invention also relates to a method for low-cost process start-up comprising the use of electrochemical fingerprinting of a process and/or electrolyte used in the process to provide an output that represents the process performance. The low cost aspect of the method is derived from eliminating the necessity of characterizing the process performance as a function of changing the electrolyte composition of at least one electrolyte used in the process. The method utilizes the electrochemical fingerprint of the process and/or electrolyte to represent a normally operating process.
In this embodiment the electrochemical fingerprint is produced by using at least one of the electrodes of an electrochemical deposition apparatus as the working electrode to produce the electrochemical fingerprint. The method may further comprise using the electrochemical fingerprint as at least part of an input to a model that provides a quantitative quality predictor as its output. The quality predictor is compared to the results of the process, which are fed back into the model after the process is completed, in order to be able to predict the expected relative quality of the product produced as the process continues to run.
Furthermore, the model and quality predictor may be designed to learn over time in order to compare the quality predictor to a feasibility set and/or to product quality. Product quality parameters are fed back into the model to help it refine the correlation between product quality and the model-generated quality predictor, and to allow it to improve the correlation as more data is fed back to the model.
Additionally, a decision model can be used to evaluate the input parameters and determine which ones are useful for creating a quality measurement.
The invention also provides a training set of data for an electrochemical analysis method that includes aged solutions. The training set of data for the electrochemical analysis method can be generated over a course of time while a process is running.
The invention also relates to an analysis method for an electrochemical process which identifies fault mechanisms based on the output of the analysis method. The analysis method comprising: utilizing one or more substrates as working electrode(s) and one or more reference electrodes to provide an output signal.
The analysis method can also be capable of learning fault mechanisms of the process through data feedback as the process runs.
Additionally, the system can be capable of identifying fault mechanisms associated with the substrates or system hardware.
The invention also relates to an analysis method using a high speed signal processing device to process data with acquisition rates within the range of minutes to nanoseconds, the data being used to make decisions regarding the switching on, or modification of, a process parameter which is an input parameter to an electrochemical process.
| Number | Date | Country | Kind |
|---|---|---|---|
| 13174725.5 | Jul 2013 | EP | regional |
The present application is a divisional of U.S. patent application Ser. No. 14/321,910, inventors Juerg Stahl et al., filed Jul. 2, 2014, the disclosure of which is incorporated herein by reference.
| Number | Date | Country | |
|---|---|---|---|
| Parent | 17015821 | Sep 2020 | US |
| Child | 18346769 | US | |
| Parent | 14321910 | Jul 2014 | US |
| Child | 17015821 | US |
