Method and device for monitoring a plasma chamber

Abstract

The present invention relates to a plasma chamber for the treatment of substrates with plasma, in particular for the cyclic treatment of substrates, in particular substrates of the same type. Furthermore, the present invention relates to a method for monitoring the operation of a plasma chamber for treating substrates.

Description

[0001] EV Group E. Thallner GmbH

[0002] EVG 104030-WO

[0003] Description

[0004] Method and device for monitoring a plasma chamber

[0005] The present invention relates to a plasma chamber for the, in particular cyclic, treatment of, especially similar, substrates of the semiconductor industry with plasma. Furthermore, the present invention relates to a method for monitoring the operation of a plasma chamber for the treatment of substrates, in particular wafers or electronic components.

[0006] In current technology, multiple sensors are regularly used to detect physical or chemical processes. Very often, the measurement signals are plotted as a function of another physical quantity, resulting in a curve. This curve can be assigned a minimum and / or maximum limit. The purpose of these limits is to define an acceptable or unacceptable measurement signal. However, measuring multiple, generally different, curves makes defining such limit curves complex and prone to errors. Furthermore, the definition is often subjective and not comparable. EV Group E. Thallner GmbH

[0007] EVG 104030-WO

[0008] - 2 -

[0009] Experiments conducted under identical initial and boundary conditions yield the same results, apart from measurement accuracy. However, in engineering, systems exist that are so complex that the initial and / or boundary conditions change even with minimal modifications to the processing equipment, the substrates being processed, or the process parameters. These changes can be static, dynamic, and especially cyclical. Particularly in the semiconductor industry, such changes are only detectable on the finished workpiece after complete processing. Furthermore, these processes are lengthy and complex, making the production of defective products especially critical, particularly since numerous previous steps often have to be repeated or components have to be painstakingly remanufactured.

[0010] A classic example of such a system is a plasma chamber. Capacitive plasma chambers have at least two electrodes. The electrodes are usually insulated and are brought to a potential by a generator. It is conceivable that the generator can be controlled solely by its power output. It is also conceivable that the generator can be controlled directly by its current and / or voltage. Inductive plasma chambers have at least one coil. A remote plasma chamber generates plasma in a first chamber and then directs it into the second, actual plasma chamber. The plasma chamber contains a gas that is ionized by the alternating field between the electrodes and / or a coil, or is generated in a remote plasma chamber and introduced into the plasma chamber.The plasma thus represents a conductive fluid that fills the entire plasma chamber or reaction chamber and can penetrate even the smallest corners to treat substrates with the plasma and trigger the desired reactions. EV Group E. Thallner GmbH.

[0011] EVG 104030-WO

[0012] - 3 -

[0013] Any additional screw, edge, or plate that is added or modified can cause the impedance of the entire system to change, and thus the behavior of the plasma. For example, if the plasma chamber is repaired, screws are replaced or forgotten, or screws made of a different material are used, the boundary conditions have already changed.

[0014] It is therefore very difficult to keep the initial and / or boundary conditions of a plasma chamber constant, since the smallest change causes the impedance and thus the behavior of the entire plasma chamber to change.

[0015] A static change in the initial and / or boundary conditions could, for example, be unintentionally caused by a technician who forgets to insert a screw after repairing the plasma chamber.

[0016] A dynamic change in the initial and / or boundary conditions could, for example, be unintentionally caused by an additional harmonic in the power supply.

[0017] Such problems are difficult to detect. Theoretically, relevant process parameters could be evaluated and analyzed by a qualified expert; however, the analysis of corresponding measurement curves by a human is highly subjective and time-consuming. Furthermore, such errors would only be detected after treatment or weeks later, especially after a large number of incorrectly performed treatments. EV Group E. Thallner GmbH

[0018] EVG 104030-WO

[0019] - 4 -

[0020] It is therefore an object of the present invention to eliminate at least some, and in particular to eliminate completely, the disadvantages listed in the prior art. In particular, it is an object of the invention to provide an improved plasma chamber for treating substrates in the semiconductor industry. Furthermore, it is an object of the invention to provide a method for detecting errors in the operation of a plasma chamber, in particular during or before the faulty treatment of the substrates.

[0021] The present problem is solved by the features of the dependent claims. Advantageous embodiments of the invention are specified in the sub-claims. The scope of the invention also includes all combinations of at least two features specified in the description, the claims, and / or the drawings. Where specified value ranges are given, values ​​lying within the stated limits are also considered disclosed as limit values ​​and may be claimed in any combination. Furthermore, all features, advantages, and disadvantages of the device in connection with the monitoring method are disclosed, and vice versa.

[0022] It has been surprisingly discovered that, based on the continuous monitoring of a single sensor during plasma treatment, errors can be detected early and automatically. Furthermore, it has been surprisingly found that, particularly in the repeated or cyclical processing of similar substrates, especially those treated consecutively, errors and, in particular, changes in initial and / or boundary conditions can be detected directly and efficiently. Moreover, the process of monitoring a plasma chamber can be carried out over extended periods, even when different treatments are performed or different substrates, especially wafers or electronic components, are being processed. EV Group E. Thallner GmbH

[0023] EVG 104030-WO

[0024] - 5 -

[0025] A particular advantage is that models for specific treatments and substrates can be created relatively easily, based solely on the sensor data from a single sensor. This allows for the detection of static changes before the start of treatment. Furthermore, errors, especially dynamic changes, can be automatically detected during treatment in the plasma chamber, enabling the operator to receive a corresponding signal, adjust the treatment, or even stop it.

[0026] Accordingly, the invention relates to a plasma chamber for treating substrates, in particular substrates of the semiconductor industry, with plasma, comprising at least: i) a reaction chamber with, in particular at least two electrodes, for receiving at least one substrate, ii) a sensor for continuously determining the intensity of the plasma in the reaction chamber within a specific wavelength range during the treatment, iii) a signal processing device with a data storage device for processing sensor data and storing the sensor data in the data storage device.wherein the signal processing device is directly configured to determine a temporal intensity profile for at least one specific wavelength during the treatment of the at least one substrate and to continuously compare it with at least one temporal intensity profile for the at least one specific wavelength of at least one previously treated substrate stored on the data storage device. EV Group E. Thallner GmbH,

[0027] EVG 104030-WO

[0028] - 6 -

[0029] In other words, the integrated signal processing device ensures that the intensity profile of a specific wavelength can be directly compared with the process data of a comparable, previously performed treatment of a similar substrate in the same plasma chamber. Advantageously, only one sensor is required. Based on this comparison, errors can be detected before initialization or completion of the treatment. The intensity, measured continuously or at short intervals, is determined based on the number of photons per unit time and area emitted by the plasma. The sensor can advantageously be located outside the reaction chamber and perform measurements even during the treatment without interfering with the error-prone process within the plasma chamber.

[0030] In this process, the ionized gas is preferably selected in a targeted manner, and the plasma chamber is evacuated beforehand. The substrates are, in particular, one or more wafers or one or more substrates equipped with electronic components that are to be subjected to plasma treatment.

[0031] The sensor data is evaluated by the signal processing device, in particular by an integrated or directly implemented computer or microcontroller. A data storage device is also available to save the corresponding sensor data. Furthermore, calculated trends or other related data, such as substrate material, process duration, atmospheric composition, etc., are made available on the data storage device for processing by the signal processing device. In this way, comparison can be automated particularly quickly and easily. A corresponding display for providing information to an operator is also conceivable. EV Group E. Thallner GmbH

[0032] EVG 104030-WO

[0033] - 7 -

[0034] The intensity profiles are continuously determined and evaluated for specific wavelengths. This allows the intensity of several or all wavelengths detectable by the sensor within the wavelength range to be determined during and before the start of treatment.

[0035] In a particular and preferred embodiment of the plasma chamber, the sensor measures an intensity ratio profile. Thus, it does not measure an intensity profile, but rather an intensity ratio profile, i.e., the ratio of two intensities. This provides information about the intensity distribution and offers known advantages during evaluation. Since measuring the ratio of two intensities has certain advantages, the intensity profile will henceforth always be described as an intensity ratio profile. For the sake of simplicity, the following text will only describe the measurement of intensities.

[0036] The previously treated substrate is preferably a similar substrate. Furthermore, all other parameters preferably remain identical. The same plasma chamber is used. After a period of inactivity or another treatment or process in the plasma chamber, changes may have occurred, which can, however, be advantageously detected by comparing the intensity profiles. Corresponding changes after a repair or changes due to, for example, incorrect loading of the plasma chamber can also be advantageously detected early and automatically.

[0037] The signal processing device is designed to perform all steps independently. Interaction with the data storage and the sensor is automatic and ensured by appropriate hardware and software. Furthermore, the advantageous direct comparability is guaranteed by directly comparing the sensor data, or at least one temporal intensity profile, with a previous treatment. EV Group E. Thallner GmbH

[0038] EVG 104030-WO

[0039] - 8 -

[0040] It is conceivable that during a single treatment, multiple intensity profiles for several wavelengths could be measured and compared in parallel, thus further increasing the accuracy of the comparison. Advantageously, the initial sensor data from multiple intensities at different wavelengths can be compared, enabling reliable early error detection.

[0041] In an advantageous embodiment of the plasma chamber, the signal processing device is further equipped to detect a deviation in the temporal intensity profile for the at least one specific wavelength during the treatment of the at least one substrate based on a comparison of both intensity profiles and to perform an action, in particular to issue a warning signal or to switch to a model stored on the data storage device.

[0042] In other words, the signal processing device directly detects any deviation from the previous measurement, particularly within a typical tolerance range, and can immediately initiate a corresponding action. This could include passive warning signals of a visual or audible nature, as well as direct measures, such as halting the initialization of the treatment at the start of the process.

[0043] Furthermore, direct comparisons with a larger dataset in the form of a model stored on the data storage device are already possible. In this model, multiple treatment data sets can be combined and optimized for comparison. In particular, several intensity profiles of different specific wavelengths from a single treatment process are combined, analyzed, and evaluated. EV Group E. Thallner GmbH

[0044] EVG 104030-WO

[0045] - 9 -

[0046] The more comprehensive the dataset of an assigned model, the more accurate the error detection by the signal processing device. In particular, the temporal evolution of the intensity is automatically corrected or adjusted so that the curves can be used for comparison.

[0047] In addition to intensity profiles, the model preferably provides further data on the substrates being treated, the process parameters, or the atmosphere, which can be taken into account by the signal processing device for assignment. This makes it particularly advantageous to switch to a different model regardless of the order in which the substrates are treated. Furthermore, the accuracy of the predictions is increased, and faster error detection is possible. Based on the models stored in the data storage, assignment in the event of a deviation is advantageously performed automatically, provided a known model is recognized by the signal processing device. This advantageously ensures smooth operation without, for example, issuing an error message in the form of a warning signal indicating that the boundary conditions have changed during processing.

[0048] In an advantageous embodiment of the plasma chamber, the signal processing device is further equipped to directly compare the temporal intensity profile for the at least one specific wavelength of the at least one substrate during treatment with a previously treated other substrate.

[0049] In other words, a different substrate, treated immediately prior in time and space, is used as a direct reference for comparison. This allows for a particularly direct and immediate comparison. Further sources of error and measurement inaccuracies are thus reduced, and the accuracy of the results is further increased. EV Group E. Thallner GmbH

[0050] EVG 104030-WO

[0051] - 10 -

[0052] The substrate(s) treated immediately beforehand are preferably identical in construction to the substrate(s) currently being treated in the plasma chamber. Identical treatment with the same process parameters and a temporally close relationship are also preferred. In this way, it is particularly advantageous that the signal processing device can confirm the fault-free operation of the process or the plasma chamber based on only one prior treatment in the plasma chamber, or, conversely, detect a fault. This fact can also be taken into account by the signal processing device, for example, by considering only the intensity profiles of the immediately preceding treatment in the comparison or by weighting them more heavily.In this way, a particularly reliable statement about the state of the boundary conditions of the plasma chamber is possible.

[0053] In an advantageous embodiment of the plasma chamber, the signal processing device is further equipped to directly assign the temporal intensity profile for the at least one specific wavelength of the at least one substrate to a stored model during the treatment and to take it into account when detecting the deviation.

[0054] In other words, the previously described advantages related to model assignment are ensured. The signal processing device can thus automatically and quickly make statements about the boundary conditions in the plasma chamber based on the stored models. Furthermore, the models are continuously improved and adapted to further increase the accuracy of the statements or to enable a reliable comparison in the first place. EV Group E. Thallner GmbH

[0055] EVG 104030-WO

[0056] - 11 -

[0057] It is also conceivable to control the results by assigning them to models in the data storage, all of which are based on previously performed treatments in the plasma chamber.

[0058] In an advantageous embodiment of the plasma chamber, limit curves are assigned to the temporal intensity profiles of at least one previously treated substrate, and the two temporal intensity profiles are compared based on these limit curves. In other words, the signal processing device not only performs a comparison but also processes the data according to defined limit curves. This advantageously ensures that the signal processing device compares within a certain tolerable error range and identifies any deviations for the action. In this way, upper and lower limit curves can be assigned to the intensity profiles and, if necessary, further refined within an existing model.Preferably, all curves are considered for determining a limit band or limit curve, thus further improving the accuracy of the results. Furthermore, detecting a deviation based on only one intensity curve is more feasible than attempting to make predictions about changed boundary conditions in the plasma chamber within the error tolerance. The signal processing device is configured to use technically reasonable distances to the at least one intensity curve of the currently treated substrate for at least one specific wavelength, based on the other process parameters. The limit curves form the basis for the initial classification models during treatment.

[0059] It is also conceivable that the operator can manually define the distances for the limit curves during a new treatment process. However, the assignment and determination of deviations always occur automatically by the process monitoring system, even without manual input. EV Group E. Thallner GmbH

[0060] EVG 104030-WO

[0061] - 12 -

[0062] In this way, it is advantageously possible to determine changed boundary conditions in the operation of the plasma chamber using the signal processing device.

[0063] In an advantageous embodiment of the plasma chamber, the signal processing device is further equipped to measure an emission spectrum in the reaction chamber immediately before the treatment of the at least one substrate and to determine at least one intensity maximum of a specific wavelength; in particular, the temporal intensity profile for exactly this wavelength is determined.

[0064] In other words, at the start of treatment, all intensities of the plasma's emission spectrum are determined as a function of all wavelengths, and corresponding maxima or a single maximum are defined. Surprisingly, it has been found that the accuracy of monitoring and comparison is particularly high for these wavelengths. Furthermore, the temporal intensity profiles can be monitored especially well and easily at these wavelengths. In particular, a comparison at these wavelengths, or at least at one wavelength, is especially informative and thus provides the signal processing device with a way to make a statement about the state in the plasma chamber, and especially during treatment, based on only one or a few profiles.

[0065] Furthermore, this initial emission spectrum measurement allows for direct assignment to a model by the single sensor.

[0066] The emission spectra are also preferentially stored on the model's data storage, allowing for an early assessment of whether a suitable treatment model already exists. EV Group E. Thallner GmbH

[0067] EVG 104030-WO

[0068] - 13 -

[0069] The signal processing device can also, if necessary, perform the assignment to a few models or limit the number of assignments.

[0070] In an advantageous embodiment of the plasma chamber, the sensor is a spectrometer. Using a spectrometer to measure the plasma treatment has proven advantageous because multiple wavelengths can be measured simultaneously. This allows for a simpler design. While the use of a spectroscope is conceivable, it is only feasible for specific applications.

[0071] In an advantageous embodiment of the plasma chamber, the wavelength range is provided to lie between 190 nm and 1100 nm, preferably between 400 nm and 1000 nm. Within these wavelengths, particularly reliable and accurate data can be obtained for monitoring the plasma chamber. Furthermore, suitable spectrometers can be easily integrated into the operation of such a plasma chamber.

[0072] Furthermore, the invention relates to a method for monitoring the cyclic operation of a plasma chamber during the treatment of several, in particular similar, substrates of the semiconductor industry, comprising at least the following steps: a) provision of the plasma chamber, b) performance of a plurality of similar treatments, in particular immediately one after the other, of the several substrates, wherein at least one temporal intensity profile of at least a certain wavelength is continuously measured by the sensor during the plurality of treatments, d) in a modeling phase, determination of an upper limit curve and a lower limit curve for at least one temporal

[0073] Intensity profile of at least one specific wavelength based on sensor data from at least one initial treatment, EV Group E. Thallner GmbH

[0074] EVG 104030-WO

[0075] - 14 -

[0076] II) in a monitoring phase, comparison of the at least one temporal intensity profile for at least one specific wavelength with the limit curves from the modeling phase based on the sensor data of at least one further treatment, wherein an action is performed if the limit curves are exceeded during the at least one further treatment, wherein after the at least one first treatment or after a predetermined number of treatments, the system switches from the modeling phase to the monitoring phase.

[0077] In other words, monitoring of a multitude of identical treatments of different, similar substrates is achieved by evaluating the intensity profile measured in a modeling phase at a specific wavelength of a first treatment against each other in the monitoring phase using boundary bands, and triggering an action accordingly if a minimum deviation is detected. The switch from the modeling phase to the monitoring phase occurs after at least one initial treatment or after a specific and technically feasible number of treatments. Performing the treatments in the plasma chamber according to the invention, as described above, is particularly preferred. The determination of the boundary curves and the comparison are carried out by the signal processing device and stored in the associated data memory, particularly in models.Furthermore, the switch from the modeling phase to the monitoring phase and, if necessary, back again is also carried out by the signal processing device.

[0078] Treatments with several substrates are monitored sequentially in the plasma chamber, with the substrates preferably being identical in every respect to ensure particularly good monitoring. However, parallel monitoring of several different treatments with different substrates performed consecutively is also conceivable. EV Group E. Thallner GmbH

[0079] EVG 104030-WO

[0080] - 15 -

[0081] In a multi-treatment process, the sensor data from a single sensor, preferably a spectrometer, is used to first measure at least one intensity profile at at least one specific wavelength. Based on this measurement, upper and lower limit curves are then defined. In the subsequent monitoring phase, the method compares at least one further intensity profile at the specified wavelength with these limit curves. If the limit curves fall below the limit or exceed the limit, an action is taken. This allows for targeted and automated monitoring, particularly of treatments performed in quick succession. The method specifically describes the procedure for obtaining a reproducible, objective, fast, efficient, and cost-effective definition of the limit curves for a measurement curve or intensity profile based on a single sensor.The advantages and specifications mentioned in connection with the plasma chamber also apply analogously to the procedure for monitoring the plasma chamber.

[0082] In a preferred embodiment of the method, it is provided that the at least one temporal intensity profile of the at least one further treatment is taken into account to adapt the limit curves if no exceedance has been detected.

[0083] In other words, the limit curves are improved or adjusted using statistically and technically sound algorithms, enabling even more precise identification of errors or changed boundary conditions. Further sensor data or time-dependent intensity profiles are only considered for adjustment if no exceedance has occurred previously, particularly within a defined and technically sound limit. EV Group E. Thallner GmbH

[0084] EVG 104030-WO

[0085] - 16 -

[0086] During the modeling phase, several intensity profiles may already have been considered to adjust the threshold curves, depending on the switching conditions, particularly if a specific number of initial treatments is predetermined. The "fall below" threshold encompasses the detection of an intensity value below the lower threshold curve as well as the detection of an intensity value above the upper threshold curve and is used accordingly in this document. This approach can advantageously increase the predictive accuracy of the method and allow for targeted implementation of appropriate actions.

[0087] In a preferred embodiment of the method, the action is either the output of a warning signal, in particular an acoustic signal, or the stopping of the treatment. The output of a warning signal also includes the display of a visual signal on a terminal or the screen of a corresponding control unit. Stopping the treatment may additionally require manual intervention by an operator. This allows for the direct comparison of treatments, enabling precise and rapid monitoring of the plasma chamber's operation.

[0088] In a preferred embodiment of the method, the action is the switch from the monitoring phase to the modeling phase. This switch is preferably performed based on the determination that the exceedance can be attributed to a known model stored in the data storage. Furthermore, the limit curves are redefined. In this way, a new model can also be advantageously implemented with this action or switch. EV Group E. Thallner GmbH

[0089] EVG 104030-WO

[0090] - 17 -

[0091] In a preferred embodiment of the method, it is provided that intensity profiles are measured for at least two or more specific wavelengths during each treatment, wherein limit curves are defined for each wavelength in the modeling phase and these are taken into account in the monitoring phase during the comparison.

[0092] In other words, the intensity profiles for multiple wavelengths are determined in parallel, so that the accuracy of the results can be further increased by means of the multiple defined limit curves. The intensity profiles are also simultaneously read out from the single sensor and processed by the signal processing device, and stored on the data storage for future consideration and model assignment.

[0093] In a preferred embodiment of the method, it is provided that all intensity profiles for all wavelengths of each of the multitude of treatments are stored and summarized in models.

[0094] In other words, all data is stored and assigned to models, particularly the same model, to further improve accuracy. The boundary curves are grouped within a single model to enable particularly precise monitoring. These models allow for the early detection of errors or changes in boundary conditions within the plasma chamber.

[0095] In a preferred embodiment of the method, the models are considered in the modeling phase for defining the limit curves for comparison in the monitoring phase. EV Group E. Thallner GmbH

[0096] EVG 104030-WO

[0097] - 18 -

[0098] In other words, the stored and, where applicable, aggregated models are already taken into account during the modeling phase. This way, a complete dataset is available for monitoring, instead of just a single directly measured intensity profile. The currently measured intensity profile can then be given greater weight, or used to define the limit curve for the monitoring phase, in order to reflect the current state of the plasma chamber. This is particularly useful when dealing with a treatment for which little data is available or when different treatments have been performed in the plasma chamber in the interim.

[0099] In a preferred embodiment of the method, the method includes the following additional step, which is performed before the at least one first treatment: a) determination of an emission spectrum with the sensor and identification of at least one intensity maximum, wherein the at least one intensity maximum defines the at least one specific wavelength for which the respective intensity profile is measured, wherein the emission spectrum is additionally stored in a new model or the plurality of treatments are assigned to an already stored model.

[0100] In other words, an emission spectrum of all wavelengths within the wavelength range is measured beforehand to identify a corresponding maximum or maxima that define the at least one wavelength or several specific wavelengths for continuous measurement. This advantageously increases the accuracy of the results and allows for early error detection. Furthermore, the emission spectrum, preferably along with other data, is stored in the data memory. The number of models thus increases with the processing of different treatment types or substrate types. EV Group E. Thallner GmbH

[0101] EVG 104030-WO

[0102] - 19 -

[0103] The assignment process continues to be automated, ensuring targeted monitoring. Assignment to a new model can also be advantageously achieved by determining the emission spectrum. Furthermore, changes in the boundary conditions within the plasma chamber can be detected very early based on the emission spectrum, should the treatment data actually require a corresponding existing model based on the associated emission spectrum.

[0104] An important aspect of the invention is to propose a method and a device for the automatic and optimized monitoring of measurement curves. In particular, it is a method for the automatic monitoring of intensity spectra of a plasma system, based on exactly one sensor, especially a spectrometer. Direct comparison of immediately successive treatments is preferred. Furthermore, the intensity profiles and, where applicable, emission spectra are advantageously used to create models for improving the accuracy of the results. In this method, more optimal monitoring of the measurement curves is achieved through the automatic adjustment of upper and / or lower limit curves. Each measurement curve is assigned to exactly one model according to its association with the limit curves.The limit curves are therefore automatically adjusted and improved for each additional recorded measurement curve or intensity profile. No human intervention is required.

[0105] A measured value is a data value determined by a measuring device. The measured value, or intensity profile, is specifically an averaged measured value, meaning a value calculated by averaging, for example, an arithmetic mean, from several individual measurements. EV Group E. Thallner GmbH

[0106] EVG 104030-WO

[0107] - 20 -

[0108] A measured value is, in particular, an intensity value at a specific point in time. For example, with a time resolution of 1 second and a measurement frequency of 100 Hertz, 100 intensities per second can be measured and averaged to obtain a measured value for a given time. By measuring several individual values, it is possible to calculate a standard deviation and a confidence interval for the measured value, especially for a specific wavelength.

[0109] A measurement curve is a set of measured values. Mathematically speaking, it is a function with at least one parameter.

[0110] In other words, a measurement curve is an intensity-time curve, i.e., the intensity of a maximum in a spectrum, especially an emission spectrum of a plasma, that has been measured as a function of time.

[0111] A limit value is a value that a measured value should not exceed or fall below. A limit value is preferably the maximum / upper or minimum / lower value of the confidence interval. A limit curve is a set of limit values ​​that the measured curve should not exceed or fall below. A limit band is the area bounded by two limit curves.

[0112] A model is, in particular, a set of measurement curves and the averaged boundary curves derived from them. The measurement curves, or time-dependent intensity profiles, assigned to a model have at least similar, preferably identical, initial and boundary conditions. Different models, however, differ in their initial and / or boundary conditions. Partitioning all measurement curves into models allows for easier identification of changed initial and / or boundary conditions. EV Group E. Thallner GmbH

[0113] EVG 104030-WO

[0114] - 21 -

[0115] Ideally, all initial and boundary conditions are assigned to each model, so that a comparison of the models reveals which initial and / or boundary conditions have what influence on the measurement curves.

[0116] In this process, several measurement curves and their limit curves are partitioned into different models.

[0117] Furthermore, the method improves the determination of threshold curves for multiple processes or treatments performed at different times. The intervals between processes can be hours, days, weeks, or months. In particular, the method enables increasingly accurate determination of threshold curves as the number of processes performed grows. This allows for the appropriate training of the corresponding software, hardware, and / or firmware, thereby increasing the accuracy of the predictions.

[0118] Preferably, the method is carried out over several processes or treatments, while the initial and boundary conditions of the processes remain largely the same. This enables the statistical determination of the limit values ​​for the specific type of process. Two examples are given.

[0119] In a first exemplary process, several ordinary, untreated silicon wafers could be treated in a plasma chamber. The plasma acts on the substrate surfaces of the silicon wafers and generates a characteristic, first emission spectrum, the measurement curves of which are recorded, evaluated and used to determine the first limit curves.

[0120] In a second exemplary procedure, several ordinary, untreated galium nitride wafers could be treated in a plasma chamber.

[0121] The plasma interacts with the substrate surfaces of the galium nitride wafers, generating a characteristic second emission spectrum. The measurement curves of this spectrum are recorded, analyzed, and used to determine the second limit curves. EV Group E. Thallner GmbH

[0122] EVG 104030-WO

[0123] - 22 -

[0124] In general, the determined limit curves will differ. The use of silicon wafers in the first process and gallium nitride wafers in the second results in different boundary conditions due to the different substrate materials. It is therefore conceivable that the initial and / or boundary conditions may change, for example, because one of the following applies: different substrates are used, and / or the electrodes have been replaced, and / or different gases or gas mixtures are used, and / or the plasma chamber material has changed, and / or additional components have been installed in the plasma chamber, and / or the current, voltage, and / or frequency of the plasma system have been changed.

[0125] In other words, a third exemplary monitoring procedure includes at least the following procedural steps.

[0126] In a first step of the third exemplary process, an emission spectrum, in particular the emission spectrum of a plasma, is recorded. In an emission spectrum, the intensity is recorded as a function of the wavelength or frequency. The intensity is plotted on the ordinate, and the wavelength or frequency on the abscissa. Preferably, the abscissa and ordinate are automatically scaled so that the majority, preferably all, of the intensity maxima are optimally represented.

[0127] In a second process step of a special process according to the invention, the abscissa values ​​of the intensity maxima are determined, preferably fully automatically.

[0128] In a third step of the third exemplary method, the intensity of at least one intensity maximum, but preferably of n intensity maxima simultaneously, is measured as a function of time. According to the invention, this measurement yields n measurement curves. EV Group E. Thallner GmbH

[0129] EVG 104030-WO

[0130] - 23 -

[0131] In a fourth step of the third exemplary procedure, the upper and lower limits of the measurement curve are calculated. It is conceivable that the upper and / or lower limit curve maintains a constant distance from the measurement curve at every point.

[0132] Process steps one to three are repeated at least once more, particularly under the same or at least similar initial and boundary conditions. Preferably, the repetition occurs at more than 2, more preferably at more than 10, most preferably at more than 50, most preferably at more than 100, and most preferably at more than 500 further times. In general, process steps one to three are repeated n times.

[0133] In a fifth step of the third exemplary procedure, it is now checked whether the newly measured curve lies within the previous limit band.

[0134] If the newly measured curve lies within the previous limit band, the upper and lower limit curves of the limit band are adjusted by an algorithm so that as many previous measurement curves as possible fall into the newly calculated limit band and are thus included by the newly calculated limit curves.

[0135] If the newly measured curve is not within the previous limit band, it can be assumed that initial and / or boundary conditions have changed to such an extent that the curve must be recorded in a new model. It could also be a one-time deviation that does not occur again in subsequent measurements from further runs. It is also conceivable that the curve is partially or completely outside the limit band, but the deviation is very small. Should the signal processing device or a user determine that the curve does not belong to the current model, a new model is created and the curve is added to it. It is also conceivable that software is able to determine which previously known and stored model the curve belongs to. EV Group E. Thallner GmbH

[0136] EVG 104030-WO

[0137] - 24 -

[0138] The software then automatically selects the correct model and continues working with that model.

[0139] In other words, the plasma system comprises at least one sensor for measuring values ​​as a function of time, a readout unit for the acquired measurements, and a processing unit or signal processing device for the measurements. The processing unit can include a microcontroller with firmware or a computer with appropriate software.

[0140] The sensor is preferably a spectrometer. A spectrometer is a measuring instrument used to measure the intensity of a particle as a function of a physical quantity of that particle, such as wavelength, frequency, energy, or mass. Intensity is defined as energy per unit area and time. Intensity is therefore also related to the number of particles. By measuring a spectrum, one can thus determine how many particles possess a certain physical quantity per unit area and time.

[0141] The wavelength range of the plasma's emitted photons extends from the infrared through the visible and into the ultraviolet range. This wavelength range thus extends from approximately 1000 nm to 250 nm.

[0142] Further advantages, features, and details of the invention will become apparent from the following description of preferred embodiments and from the figures. In the figures, identical components or components with the same function are identified by the same reference numerals. These are shown schematically in:

[0143] Figure aa shows an exemplary measured emission spectrum of a plasma system, with several intensity maxima. EV Group E. Thallner GmbH

[0144] EVG 104030-WO

[0145] - 25 -

[0146] The physical cause of the intensity maxima is not discussed in detail here, as it is of minor importance for understanding the plasma system and the method for monitoring it. The emission spectrum is measured for a specific system under well-defined initial and boundary conditions. The three intensity maxima 1a, 1b, and 1c have been marked as examples. For the plasma system and the method for monitoring it, several measurement curves per model are preferably stored, in particular several measurement curves for one wavelength per model. The measurement curves for 1a are shown as examples.

[0147] However, it is conceivable and preferred that several measurement curves of multiple wavelengths are stored per model, for example, measurement curves 1a, 1b, and 1c. The averaging of the measurement curves and the adjustment of the limit curves must then, of course, be performed separately for the measurement curves of the different wavelengths, for which the signal processing device is directly equipped. In the case shown, a model would practically be a set of three sets of measurement curves. The subsequent description of the figure explains that each model always stores only the measurement curves for one wavelength.

[0148] Figure 1b shows the time course of the intensity maxima a1a, b1b and c1c. The time course of the intensity maxima is only in the millisecond range.

[0149] Figures 2a-2f have a number marked with a diamond in the upper left corner. This number indicates the measurement curve currently being measured. The number following the letter M indicates the model number to which the measurement curve is assigned.

[0150] Figure 2a shows an exemplary first measurement curve la. The shape of the measurement curve la depends on the selected intensity maximum, the process or treatment, and the corresponding initial and boundary conditions. The measurement curve is assigned to a first model. EV Group E. Thallner GmbH

[0151] EVG 104030-WO

[0152] - 26 -

[0153] Figure 2b shows the exemplary first measurement curve la of the first model, which is enclosed by an upper limit curve 2o and a lower limit curve 2u. Since measurement curve la is the first measurement curve of the first model, the limit curves 2o and 2u are determined by initial conditions. One possibility would be to define a constant distance d between measurement curve la and the limit curves 2o and 2u.

[0154] Figure 2c shows an exemplary second measurement curve la', which lies entirely within the two limit curves 2o and 2u. The second measurement curve la' is therefore also assigned to the first model. The two measurement curves 1 and 1' are then used to recalculate the upper limit curve 2o and the lower limit curve 2u.

[0155] Figure 2d shows the newly adapted upper limit curve 2o' and the newly adapted lower limit curve 2u', which were calculated using the measurement curves la, la' and the limit curves 2o, 2u.

[0156] Figure 2e shows another recorded measurement curve, or rather, a temporal intensity profile α," which no longer lies within the limit curves 2u' and 2o'. The procedure is then continued such that the new measurement curve α" is added not to the first model, but to a new, second model. The reason for the significant deviation of measurement curve α" from the two preceding measurement curves α and α' could lie in any kind of environmental change. For example, components of the system could have been modified. The use of different process parameters is also conceivable. Therefore, an action is taken.

[0157] Figure 2f shows the generation of new upper and lower boundary curves 2o" and 2u", which can also be assigned to the second model. From this point on, each subsequent measurement curve is compared with the first and second models to determine which model it belongs to. If it does not belong to either of the existing models, a new model is generated. EV Group E. Thallner GmbH

[0158] EVG 104030-WO

[0159] - 27 -

[0160] The monitoring process, or rather the plasma system, collects all measurement curves and either assigns them to a new model and simultaneously adjusts the limit curves of this model, or generates a new model. This modeling is important to ensure and compare work results, especially successful plasma treatments, so that any changes in the plasma chamber can be detected.

[0161] EV Group E. Thallner GmbH EVG 104030-WO

[0162] -28 -

[0163] Reference symbol list la, lb, 1c, la', la“ Intensity maxima at a wavelength or measurement curve of the intensity as a function of time, temporal intensity profile o, 2o' upper limit curves u, 2u' lower limit curves

Claims

EV Group E. Thallner GmbH EVG 104030-WO - 29 - P a t e n t a n s p r ü c h e 1) Plasma chamber for treating substrates with plasma, comprising at least: i) a reaction chamber for receiving at least one substrate, ii) a sensor for continuously determining the intensity of the plasma in the reaction chamber within a specific wavelength range during treatment, iii) a signal processing device with a data storage device for processing sensor data and storing the sensor data in the data storage device, wherein the signal processing device is directly configured to determine a temporal intensity profile (α') for at least one specific wavelength during the treatment of the at least one substrate and to continuously compare it with at least one temporal intensity profile (α) stored in the data storage device for the at least one specific wavelength of at least one previously treated other substrate. EV Group E. Thallner GmbH EVG 104030-WO - 30 - 2) Plasma chamber according to claim 1, wherein the signal processing device is furthermore directly configured to detect a deviation in the temporal intensity profile for the at least one specific wavelength during the treatment of the at least one substrate based on the comparison of both intensity profiles and to perform an action, in particular to issue a warning signal or to switch to a model stored on the data storage device. 3) Plasma chamber according to at least one of the preceding claims, wherein the signal processing device is furthermore directly configured to compare the temporal intensity profile for the at least one specific wavelength of the at least one substrate during treatment with a previously treated other substrate. 4) Plasma chamber according to at least one of the preceding claims and at least claim 2, wherein the signal processing device is furthermore directly configured to assign the temporal intensity profile for the at least one specific wavelength of the at least one substrate to a stored model during treatment and to take it into account when detecting the deviation. 5) Plasma chamber according to at least one of the preceding claims, wherein the signal processing device is furthermore directly configured to assign limit curves to the temporal intensity profiles of the at least one previously treated other substrate and to compare the two temporal intensity profiles on the basis of the limit curves. EV Group E. Thallner GmbH EVG 104030-WO - 31 - 6) Plasma chamber according to at least one of the preceding claims, wherein the signal processing device is furthermore directly configured to measure an emission spectrum in the reaction chamber immediately before the treatment of the at least one substrate and to determine at least one intensity maximum of a specific wavelength, in particular the temporal intensity profile for exactly this wavelength is determined. 7) Plasma chamber according to at least one of the preceding claims, wherein the sensor is a spectrometer. 8) Plasma chamber according to at least one of the preceding claims, wherein the wavelength range is between 190 nm and 1100 nm, preferably between 400 nm and 1000 nm. EV Group E. Thallner GmbH EVG 104030-WO - 32 - 9) Method for monitoring the operation of a plasma chamber during the treatment of several, in particular similar, substrates, comprising at least the following steps: a) provision of a plasma chamber, in particular according to at least one of the preceding claims 1 to 8, b) performance of a plurality of similar treatments, in particular immediately one after the other, of the several substrates, wherein at least one temporal intensity profile of at least a certain wavelength is continuously measured by the sensor during the plurality of treatments, I) in a modeling phase, determination of an upper limit curve (2o) and a lower limit curve (2u) for at least one temporal intensity profile of at least one specific wavelength based on the sensor data of at least one first treatment, II) in a monitoring phase, comparison of the at least one temporal intensity profile for at least one specific wavelength with the limit curves (2o, 2u) from the modeling phase based on the sensor data of at least one further treatment, wherein an action is performed if the limit curves (2o, 2u) are exceeded during the at least one further treatment, wherein after the at least one first treatment or after a predetermined number of treatments, the system switches from the modeling phase to the monitoring phase. EV Group E. Thallner GmbH EVG 104030-WO - 33 - 10) Method according to claim 9, wherein the at least one temporal intensity profile of the at least one further treatment is taken into account to adapt the limit curves (2o, 2u) if no exceedance has been detected. 1 1 ) Method according to at least one of the preceding claims, wherein the action is the output of a warning signal, in particular comprising an acoustic signal, or the action is a cessation of the treatment. 12) Method according to at least one of the preceding claims, wherein intensity profiles are measured for at least two or more specific wavelengths in each treatment, wherein limit curves are defined for each wavelength in the modeling phase and these are taken into account in the monitoring phase during the comparison. 13) Method according to at least one of the preceding claims, wherein all intensity profiles for all wavelengths of each of the plurality of treatments are stored and summarized in models. 14) Method according to claim 13, wherein the models are taken into account in the modeling phase for determining the limit curves for comparison in the monitoring phase. EV Group E. Thallner GmbH EVG 104030-WO - 34 - 15) Method according to at least one of the preceding claims and at least claim 13 or 14, wherein the method comprises the following additional step, which is performed before the at least one first treatment: a) Determining an emission spectrum with the sensor and identifying at least one intensity maximum, wherein the at least one intensity maximum defines the at least one specific wavelength for which the respective intensity profile is measured, wherein the emission spectrum is additionally stored in a new model or the plurality of treatments are assigned to an already stored model.

Images