DE102018216969A9 - Plasma treatment apparatus and method for outputting pulses of electrical power to at least one process chamber - Google Patents

Abstract

A plasma treatment device and a method for outputting pulses of electrical power to at least one process chamber are described. The plasma treatment device has a plasma generator for outputting pulses of electrical power to at least one process chamber; the plasma generator is configured to generate and output the pulses over a process time in accordance with a predetermined pulse sequence. The plasma generator has an arc unit which is suitable for recognizing a rollover or an imminent rollover, also called an arc event, and in response to this, to cancel, change and / or interrupt a pulse that is or is to be output to the process chamber to fail what is known as arc interference. Furthermore, at least one control unit is provided which is configured to carry out a method that has the following: determining or receiving a process time and pulse sequence for the process chamber in / on a control unit, activating a plasma generator in order to output the pulse sequence over the determined process time , the detection of an arc event that indicates a rollover or an impending rollover, in response to the detection of an arc event, the implementation of an arc intervention, the aborting, changing and / or omitting a pulse that is output to the process chamber or is to be output, the receipt of data about arc interventions and determination based on the data of a process time extension with a corresponding pulse sequence in order to at least partially compensate for a power reduction caused by the arc interventions, and the activation of the plasma generator in order to follow the determined process time over the duration of the best Output the corresponding pulse sequence whenever the process time is extended

Description

The present invention relates to a plasma treatment device and a method for outputting pulses of electrical power to at least one process chamber for the plasma-assisted treatment of substrates.

A wide variety of fields of application are known in technology, in which electrical power is usually provided with a frequency of at least 40 kHz in order to excite a plasma from a gas and to maintain it for certain processes. An example of this is plasma-assisted vapor deposition in semiconductor technology or the photovoltaic industry.

Here, for example, wafers are loaded in so-called wafer boats, some of which consist of electrically conductive plates, and which are placed in corresponding process chambers. Electrical power with a frequency of at least 40 kHz is then applied to the wafer boats in order to generate a plasma from a suitable process gas between the plates and between wafers held on the plates. An example of such a plasma treatment device is shown in FIG DE 10 2015 004 419 A1 shown by the applicant.

Such plasma treatment devices generally each consist of a single process chamber to which a single plasma generator is assigned. In the case of the adjacent arrangement of several process chambers, a shared use of gas cabinets and other peripheral devices was sometimes considered, but so far each process chamber has had an individual plasma generator. Such a plasma generator usually has a controllable power supply unit with an output which is suitable to output a direct current with a predetermined voltage and / or strength at its output, a converter which is suitable for converting a direct current at the input into an alternating current with a predetermined frequency of shape at least 40KHz as an output signal and apply the output signal to the connected process chamber

During the processes, the electrical power is usually provided periodically as a pulse by the respective plasma generator, with a duty cycle (quotient of pulse duration and period duration) of less than 0.1 being set here as a rule over long periods of time. Such a pulsed operation of a plasma in the process chamber has proven to be particularly advantageous. Typically 90% of the available continuous power of a plasma generator therefore remains unused over long stretches of the processes. There are also processes with significantly smaller duty cycles, but also processes or process sections with higher duty cycles. The power supply unit and the converter are usually controlled via a controller which, depending on a process recipe (the required power), determines a corresponding pulse sequence for the process chamber over a process duration and generates corresponding control signals. The process duration and the pulse sequence are usually specified in advance and no longer changed.

Plasma generators usually also have an arc suppression unit (also known as an arc suppression unit), which is suitable for receiving current process data about processes in the process chamber and current data from the power supply unit and / or the switch and evaluating it in real time to detect a flashover or an impending flashover ( also called Arc). Arcs can lead to locally burned, melted or unevenly coated wafers. Furthermore, the wafer carrier can also be damaged, which can lead to production downtimes or high service costs. Therefore arcs should be avoided whenever possible.

The arc suppression unit can, for example, be connected to the power supply unit and / or the switch in order to control them optionally in response to the detection of a rollover or an impending rollover, in order to avoid or quickly suppress flashovers in the process chamber. This happens, for example, that electrical power pulses are not built up in the first place or a corresponding pulse is diverted to an absorber, for example. This can also take place while a pulse has already been partially passed into the process chamber.

Plasma generator manufacturers usually specify suitable settings for arc suppression management (also called arc management), which can also be changed by the system operator if necessary. In addition, modern treatment systems can use arc suppression values to decide whether to terminate the process, for example if too many arc events have occurred within a period of time, or if power, current and / or voltage are outside the tolerance.

Plasma generator manufacturers offer arc management or arc suppression software solutions in particular that intervene in the process by reducing current / voltage and forcing process pauses. A reduction in current / voltage and process pauses can, however, if they are planned in advance, lengthen the process times accordingly and thus impair the throughput of a process plant. However, if in this way in the ongoing process is intervened, this leads to a low average output over the process time. This can lead to the deposited layer thicknesses being less than intended, since, for example, in PECVD processes, the deposition thickness is proportional to the applied power. Smaller layer thicknesses can lead to problems in later machining processes or even to a complete functional failure of the deposited layer.

Plant operators also define process termination criteria such as voltage, current, power or the number of arcs. This enables you to cancel the process before the workpiece carrier is damaged. To do this, the system operator must select the termination criteria in such a way that the Arc Management System hardly has a chance to react. Aborting the process followed by a search for the cause of the arcing is time-consuming, reduces system throughput and leads to material loss.

The system operator therefore wants no process interruptions, no damage to the wafer and / or workpiece carrier caused by arcs, but also no layers that are deposited too thinly.

The present invention is therefore based on the object of making the provision of electrical power more efficient in order to avoid or reduce at least one of the above-mentioned problems as far as possible. According to the invention, this object is achieved by a plasma treatment device according to claim 1 and a method for outputting pulses of electrical power to at least one process chamber according to claim 9. Further refinements of the invention emerge from the subclaims, among other things.

The plasma treatment device has a plasma generator which is configured to output pulses of electrical power to at least one process chamber, the plasma generator being configured to generate and output the pulses over a process time in accordance with a predetermined pulse sequence, and the plasma generator being an arc unit which is suitable to recognize a rollover or an imminent rollover, also called an arc event, and, in response, to cancel, change and / or fail what is called an arc Intervention is called. The plasma treatment device in particular also has at least one control unit which is configured to receive and evaluate process data about processes in the at least one process chamber and, based on the process data, at least one process time and one pulse sequence to be made available by the plasma generator over the process time to determine, or to receive a certain process time and pulse sequence, the pulse sequence configured over the process time to bring a predetermined power into the process chamber, to control the plasma generator, to output the pulse sequence over the process time. The at least one control unit is also configured to receive data on arc interventions and to use the data to determine a process time extension with a corresponding pulse sequence in order to at least partially compensate for a power reduction caused by the arc interventions. Such a plasma treatment device offers the possibility of at least partially compensating power losses that occur due to arc events, which improves the homogeneity of processes and enables more aggressive arc management.

The plasma treatment device preferably has a unit which is configured to determine the number of arc interventions and to transmit it to the at least one control unit. The number of interventions, which is easy to determine, alone can provide a basis for determining an approximate power loss for determining a process time extension with a corresponding pulse sequence. In particular, the unit can be configured to recognize at least two different types of arc interventions, to determine the number of each type of arc intervention and to transmit the number to the at least one control unit. By differentiating between the arc interventions, a better assessment of the power loss can be made. The unit can be part of the arc unit that performs the arc interventions. In one embodiment, the arc unit is configured to recognize at least two different types of arc events and to select different predetermined arc interventions corresponding to the arc event.

In particular, the arc unit can be configured to recognize at least two of the following arc events: an Imax event that indicates an overcurrent, ie a current above a threshold value, and thus the occurrence of an arc; a U * I event which, before an overcurrent is reached, indicates an excessive current increase at a given voltage (Uxl) in such a way that an overcurrent would be expected if the power continued to be supplied; a micro-arc, which is indicated by a current peak at a given voltage. These events have different influences on the process, with an Imax event having the greatest influence and should be avoided as far as possible, as it can damage the substrate or a substrate carrier. Corresponding interventions in micro-arc events and U * I events, which are often harbingers of an immanent Imax event, can possibly be prevented. The arc unit can in particular be configured to completely abort the current pulse when an Imax event is detected and then slowly increase the power so that the following pulse may not be fully output either; to interrupt the current pulse when a U * I event is detected and, as far as possible, to restart it within the planned pulse time with an increasing power ramp; and to interrupt the current pulse when a micro-arc is detected and to continue outputting it normally after a short waiting time.

In a further embodiment, the control unit is configured to control the plasma generator when a certain number of arc events are reached within the process time or a process section, to reduce the power output per pulse of the remaining pulse train and a corresponding power reduction when determining the process time extension with a corresponding pulse train to consider. By reducing the power of the remaining pulses, the probability of arcs can be reduced while the process is running, but a corresponding power loss can be at least partially compensated for in order to obtain a desired process result.

When determining the process time extension with a corresponding pulse sequence, the mean power of the pulses can be selected to be lower than the mean power of the pulses during the previous process time in order to reduce the probability of arcs during the process time extension.

The method for outputting pulses of electrical power to at least one process chamber has the following: determining or receiving a process time and pulse sequence for the process chamber in / on a control unit, activating a plasma generator in order to output the pulse sequence over the specific process time, recognizing an arc Events that indicate a rollover or an imminent rollover, in response to the detection of an arc event, carry out an arc intervention which has the cancellation, modification and / or omission of a pulse that is or is to be outputted to the process chamber of data about arc interventions and determining, based on the data, a process time extension with a corresponding pulse sequence in order to at least partially compensate for a power reduction caused by the arc intervention, and control of the plasma generator after the specific process time over the duration of the specific process time extension to output the corresponding pulse train away. The advantages already indicated above can be achieved by the method.

The data on arc interventions preferably have the number of arc interventions, which is easy to determine. It is preferably possible to carry out different types of arc interventions and the data on arc interventions then have at least the types of arc interventions and the respective number of arc interventions.

In one embodiment, the detection of arc events comprises the detection of at least two different types of arc events, the implementation of the arc intervention selecting a predetermined arc intervention from a plurality of predetermined arc interventions corresponding to the arc event and performing the selected arc procedure. In this way, specific events can be addressed in a more targeted manner. In particular, the detection of arc events comprises the detection of at least two of the following arc events: an Imax event that causes an overcurrent, i.e. a current above a threshold value, and thus indicates the occurrence of an arc, a U * I event which, before an overcurrent is reached, indicates an excessive current increase at a given voltage in such a way that an overcurrent would be expected with further power delivery, and / or a micro-arc, which is indicated by a current spike at a given voltage. Correspondingly, performing the arc intervention can have the following effects: when an Imax event is detected, the current pulse is completely interrupted and the power is then slowly increased so that the following pulse may not be fully output, when a U * I event is detected, the current one Interrupt the pulse and, if possible, restart it within the planned pulse time with an increasing power ramp, and / or interrupt the current pulse when a micro-arc is detected and continue to output it normally after a short waiting time. The advantages of the different arc interventions have already been explained above.

In one embodiment, when a certain number of arc events are reached within the process time and / or a process section, the plasma generator is triggered to reduce the power output per pulse of the remaining pulse train, a corresponding power reduction being taken into account when determining the process time extension with a corresponding pulse train. The power reduction for the remaining pulses is at least 10%, preferably at least 20%.

When determining the process time extension with a corresponding pulse sequence, the average power per pulse can be selected to be lower than that Average power per pulse during the process time to reduce the likelihood of arcs in the process time extension. In particular, the average power per pulse during the process time extension can be selected to be at least 10%, preferably at least 20% lower than the average power per pulse during the process time.

• 1 eine schematische Darstellung einer beispielhaften Ausführungsform einer Plasma-Behandlungsvorrichtung;
• 2 eine schematische Darstellungen einer Soll-Pulsfolge, wie sie über einen Zeitraum von einem Plasmagenerator abgegeben werden soll;
• 3 eine schematische Darstellungen einer beispielhaften Pulsfolge entsprechend 1, wie sie tatsächlich vom Plasmagenerator abgegeben wurde;
• 4 eine schematische Darstellungen eines Einzelpulses, bei dem ein erster Fehler detektiert wurde und eine dem Fehler entsprechende Korrektur vorgenommen wurde;
• 5 eine schematische Darstellungen eines Einzelpulses, bei dem ein zweiter Fehler detektiert wurde und eine dem Fehler entsprechende Korrektur vorgenommen wurde;
• 6 eine schematische Darstellungen von drei Pulsen der Soll-Pulsfolge, bei denen ein dritter Fehler detektiert wurde und eine dem Fehler entsprechende Korrektur vorgenommen wurde;
• 7 eine schematische Darstellungen einer beispielhaften Pulsfolge, wie sie von einem Plasmagenerator über einen vorbestimmten Zeitraum abgegeben wurde, mit einem Beispiel einer Verlängerung des Abgabezeitraums;
• 8 eine schematische Darstellungen einer beispielhaften Pulsfolge, wie sie von einem Plasmagenerator über einen vorbestimmten Zeitraum abgegeben wurde, mit einem alternativen Beispiel einer Verlängerung des Abgabezeitraums;
• 9 eine schematische Darstellungen einer beispielhaften Pulsfolge, wie sie von einem Plasmagenerator über einen vorbestimmten Zeitraum abgegeben wurde, mit einem weiteren Beispiel einer Verlängerung des Abgabezeitraums.

The invention is explained in more detail below with reference to the drawings; in the drawings shows:

• 1 a schematic representation of an exemplary embodiment of a plasma treatment device;
• 2 a schematic representation of a desired pulse sequence as it is to be delivered over a period of time by a plasma generator;
• 3 a schematic representation of an exemplary pulse train accordingly 1 how it was actually delivered by the plasma generator;
• 4th a schematic representation of a single pulse in which a first error was detected and a correction corresponding to the error was made;
• 5 a schematic representation of a single pulse in which a second error was detected and a correction corresponding to the error was made;
• 6th a schematic representation of three pulses of the target pulse sequence, in which a third error was detected and a correction corresponding to the error was made;
• 7th a schematic representation of an exemplary pulse sequence, as it was delivered by a plasma generator over a predetermined period of time, with an example of an extension of the delivery period;
• 8th a schematic representation of an exemplary pulse train as it was delivered by a plasma generator over a predetermined period of time, with an alternative example of an extension of the delivery period;
• 9 a schematic representation of an exemplary pulse sequence as it was delivered by a plasma generator over a predetermined period of time, with a further example of an extension of the delivery period.

1 shows a schematic representation of an exemplary embodiment of a plasma treatment device 1 with two process units 3a , 3b and a plasma generator 5 . Although two process units in the embodiment 3a , 3b with a plasma generator 5 are shown, the person skilled in the art will recognize that only a single process unit can be provided and the plasma generator can be simplified accordingly.

The process units 3a and 3b can each have the same structure and each have a process chamber for receiving one or more substrates, in particular semiconductor wafers or PV substrates. The process chambers and the process units can be sealed tightly 3a and 3b have different means, not shown, for setting a desired gas atmosphere within the respective process chambers, such as pumps and a gas cabinet. While each process chamber is usually assigned its own pump, a gas cabinet can supply several process chambers if necessary. Means for generating a plasma are also provided in the process chambers, these being able to be formed in part by a wafer boat which, for example, is introduced into the process chamber together with the substrates and is electrically contacted there, as is for example mentioned in the above DE 10 2015 004 419 A1 is described, which is made the subject of the present disclosure.

The process units 3a , 3b stand over power lines 7a or. 7b and data connections 8a or. 8b with the plasma generator 5 in connection. About the power lines 7a , 7b is supplied by the plasma generator 5 electrical power with a frequency greater than 40KHz to the respective process units 3a and 3b . About the data connections 8a or. 8b can transfer data between the process units 3a , 3b and the plasma generator 5 be replaced. The data connections 8a or. 8b can be wired or wireless. In particular, the process units 3a , 3b the plasma generator 5 deliver different process data about processes in the respective process chambers. In particular, actual data with regard to the actually incoming electrical power, the presence of a plasma, etc. can be transmitted, but also corresponding target data, as is the case with already known plasma treatment devices with a single process unit and a single plasma generator.

The plasma generator 5 has an adjustable power supply 10 , a switching unit 12th as well as a control unit 14th . The power supply unit has an input (not shown) and an output via a power line 15th with an input of the switching unit 12th connected is. The power supply 10 is suitably responsive to a control signal from the control unit 14th to output at its output a direct current having a predetermined voltage and / or magnitude, as is known in the art.

The switching unit 12th has the already mentioned input with the power line 15th is connected to receive DC power from the power supply 10 , as well as two separate outputs, one output to the power line 7a and the other output with the power lines 7b connected is. The switching unit 12th has a converter circuit which can form an alternating current with a predetermined frequency of at least 40KHz as an output signal from a direct current at the input. The converter circuit can, for example, have a bipolar transistor that is derived from the power supply unit 10 provided direct current forms a quasi sinusoidal signal (stepped signal). The switching unit also has 12th a switching part which generates the output signal thus shaped in response to a control signal from the control unit 14th applied to one or the other output, thus to the process unit 3a or the process unit 3b . The output signals (pulses) must be coordinated in such a way that the respective requirements of the process chamber are met. A corresponding pulse coordination is for example in the not previously published DE 10 2018 204 585 the same applicant explained in more detail, which is made part of the subject matter of the present application by reference. If there is only one process unit, a corresponding coordination is not necessary and the switch can, for example, be omitted or used to interrupt and / or divert the output signal into an absorber.

The control unit 14th again has two separate controllers for the display 16a , 16b , as well as an arc suppression unit 18th , also known as the arc unit. The regulator 16a , 16b stand over the data connections 8a or. 8b with the process units 3a or. 3b in connection. The regulator 16a , 16b are each designed in a known manner on the basis of actual data and target data about the respective processes in the process units 3a , 3b Performance requirements for the process units 3a , 3b to investigate. The controllers create from these 16a , 16b (or a downstream unit) then control data for the power supply unit 10 and the switching unit 12th . Although the regulator 16a , 16b are shown as separate units, they can also be designed as a single unit, which essentially processes data from the process units in parallel 3a and 3b enables. In particular, the controller 16a , 16b be designed as separate software routines that can be executed essentially in parallel and that are executed on a processor.

The performance requirements of the individual process units are usually specified for a particular process and consist of a specific pulse sequence that is to be output over a specific process time. The performance requirements can be made available from a memory of a corresponding system control, not shown, from a memory, but they can also be determined by the system control, for example, with a PECVD operator specifying a layer thickness to be achieved on substrates in the process chamber and the System control determines the corresponding parameters, including the performance requirements in the form of a specific pulse train that is to be output over a process time. The pulse sequence usually consists of a sequence of pulses of the same duration, the same distance between the pulses and the same power, even if this is not necessary and can vary. The total output over the process time essentially determines the process result achieved. With a PECVD, the deposited layer thickness is proportional to the output power. Corresponding pulse sequences can consist of a few hundred or fewer, up to several thousand or even tens of thousands of pulses, depending on the type and thickness of the slice. The number is of course also determined by the power and the pulse length. In order to achieve short process times, there is a tendency to provide as much power as possible per pulse, but this increases the risk of arcs or arcs.

In the embodiment shown, there are corresponding outputs of the controller with corresponding inputs of the arc unit 18th connected. The arc unit 18th can therefore be controlled by the controllers 16a , 16b receive the performance requirements of the individual process units, or the control data for the power supply unit created from them directly 10 and the switching unit 12th . In addition, the arc unit 18th but also unprocessed data directly from the respective process units 3a , 3b receive. Such data are, in particular, those which enable it to be evaluated in real time whether it is in one of the process units 3a and 3b a rollover has occurred or is imminent, which is hereinafter referred to as an arc event. The data required for this are known to the person skilled in the art, as are the corresponding recognition algorithms, which are therefore not initially explained in more detail. If an arc event is detected, ie a rollover or the risk of such a rollover, the arc unit can 18th the control signals to the power pack 10 and the switching unit 12th adapt and intervene in the provision of the service to the respective process unit by canceling, changing and / or omitting a pulse. Such an intervention is also known as an arc intervention.

The arc unit 18th can furthermore be designed in this embodiment in such a way that, when power requirement data of the individual process units are received therefrom, control data for the power supply unit 10 and the switching unit 12th created. This is of course not necessary if these control data are already from the controllers 16a , 16b to be created. In addition is the arc unit 18th trained so that you can create your own or those of the controllers 16a , 16b generated control data for the power supply unit 10 and the switching unit 12th coordinated. The control data for the power supply 10 and the switching unit 12th are to be coordinated so that the respective process units 3a , 3b performance is provided according to your performance requirements. In addition, the control data for the power supply 10 and the switching unit 12th but also to be coordinated in time, since power can only be made available at one of the outputs of the switching unit at any given time. As already explained above, the performance is transferred to the respective process units 3a , 3b provided as a pulse with a predetermined pulse duration. The respective pulses are therefore via the control data for the power supply unit 10 and the switching unit 12th to be coordinated so that the pulses do not overlap, but the process chambers can still be operated at the same time. Such pulse coordination is of course not required for a single process unit.

The arc unit 18th can, if necessary, differentiate between different types of arc events and perform different arc interventions accordingly. Furthermore, the arc unit is able to count the number of arc interventions as a whole, but also according to the respective type. As those skilled in the art will recognize, each intervention in a requested pulse causes a difference between the requested power and the power actually output.

However, since the performance is essential for the process result, the control unit is 14th (or the system control) configured to receive data about the arc interventions and to use the data to determine a process time extension with a corresponding pulse sequence in order to at least partially compensate for a power reduction caused by the arc intervention. Following the normal process time, a corresponding additional process time with a corresponding pulse sequence can then be added. Of course, the other process components, such as gas supply, pump, heating, etc. are informed about a corresponding process time extension and controlled accordingly for longer.

This is explained below using the 2 to 9 explained in more detail. It shows 2 an exemplary pulse train generated by a plasma generator 5 to be issued to a process chamber while 3 shows an example of the corresponding actually output pulse sequence. In 3 the actually desired pulses are shown in dashed lines.

The desired or requested pulse train according to 2 consists of a certain number of pulses that are to be output over a certain process time in order to bring the desired electrical power into a process. A real process usually has several hundred to several thousand or even tens of thousands of pulses. For the sake of simplicity, however, only 12 pulses are shown in the illustration, which are to be output over a process time. Each pulse is an alternating current pulse with a frequency of at least 40KHz and a predetermined power (P) that is to be output over the pulse duration. Then there is a pulse pause followed by another pulse. In the illustration, all pulses are the same, ie they each have the same pulse duration and the same amplitude (power). The pulse pauses are also shown to have the same length. In the illustration, the pulse pauses are approx. Twice as long as a respective pulse (ratio 2: 1). In real processes, however, the ratio is often much larger, such as 9: 1, which would result in a duty cycle of 0.1. In addition, the pulse duration, the pulse pauses and / or the amplitude of the pulses can vary over the process time, although the same pulses with the same pulse pauses are generally requested over a main section.

3 now shows the corresponding pulse sequence actually output by a plasma generator, the actually desired pulses being shown in dashed lines. As can be seen, most of the pulses were completely output, while some pulses were only partially output and one pulse was not at all. For the pulses that were only partially output, the average power of the pulse over the pulse duration is displayed.

The incomplete pulse output is due to arc interventions by the arc unit 18th . In particular, it is considered in the context of the application that the arc unit 18th is able to recognize at least two, preferably three or more, different arc events and carry out correspondingly different arc interventions. Examples of this are a micro-arc event, a U * I-arc event and an Imax-arc event, which are also shown below using the 4th to 6th are explained in more detail with their corresponding interventions. It shows 3 a micro-arc event with corresponding arc intervention, 4th a U * I-Arc event with a corresponding arc intervention and 6th an Imax arc event with a corresponding arc intervention. An Imax arc event occurs when an overcurrent occurs, ie a current above a threshold value, which indicates the occurrence of an arc. A micro-arc event is indicated by a current (power) peak at a given voltage, a U * I event by an excessive current (power) increase at given voltage, whereby the current is below the overcurrent, but with further power delivery an overcurrent would be expected.

Micro-arc events usually occur at the beginning of a pulse and are characterized by a small spike in power. But they can also occur at any point in time of the pulse. 4th shows schematically a single pulse with a micro-arc event with corresponding intervention. As can be seen, there is a peak in the power at the beginning of the pulse, whereby, for example, transients in the range from microseconds to nanoseconds can be taken into account. Although the power drops back to its normal level, the power output is interrupted for a short time shortly after the peak occurs and then continues to be output as normal. The delay after the peak is due to the response time of the system. The interruption is only of a short predetermined duration and is achieved, for example, by diverting the power into an absorber. In the illustration, the interruption is about 1/10 of the pulse duration, so that only about 10% of the power of the pulse is lost through the interruption. Of course, the actual lack of power depends on the duration of the interruption and the length of the pulse. Although not shown, another micro-arc event or another event can occur in the same pulse, which would lead to a further arc intervention within the same pulse. Although micro-arc events do not significantly affect the process, appropriate intervention can be advantageous here, as they often occur before another arc event and thus announce this. The intervention can possibly prevent a subsequent, possibly more serious arc event. There is no need to react to every Mirco-Arc event in the above way. Rather, the amplitude and / or length of the event can be taken into account when deciding whether or not to intervene.

U * I-Arc events can occur at any point in time of a pulse and are characterized by an excessive increase in the current (the power at constant voltage), whereby the current is below the threshold value for an overcurrent. However, the increase indicates that further power supply would likely result in an overcurrent. 5 shows schematically a single pulse with a U * I arc event with corresponding intervention. When a corresponding increase in power is detected (corresponding to an increase in current with constant voltage), the arc unit interrupts the power output within the pulse and, after a predetermined interruption period, begins again with the power output, but the power is increased in a ramp-like manner. The interruption duration is considerably longer than with the micro-arc procedure and is, for example, four times as long. The possibility of renewed power output naturally depends on the time of the interruption, the duration of the interruption and the remaining pulse length. The lack of power depends primarily on the duration of the interruption, the ramp and the length of the pulse. In addition, the time of the interruption would also have to be taken into account, but this would be too time-consuming. Rather, the missing power of the pulse can be averaged and in the embodiment shown it is assumed that it is 50%. Here, too, the intervention criteria for an intervention can be freely defined, with the rate of increase and / or the duration of the increase generally being taken into account.

Imax-Arc events can occur at any point in time of a pulse and are characterized by an overcurrent (excessive power with constant voltage), the current being above a threshold value that indicates the occurrence of a flashover or arc. 6th shows schematically three successive pulses of a conventional pulse train with an Imax-Arc event in the first pulse and a corresponding intervention. When a corresponding Imax event is detected, the arc unit interrupts the output of power within the pulse and, after a predetermined interruption period, begins again with the output, although the output is increased in a ramp-like manner. The interruption duration is considerably longer than with the other procedures and is usually longer than a pulse period consisting of pulse duration and pulse duration. In the representation, the interruption duration is approximately twice a pulse duration. The first pulse in the sequence of three is therefore only partially output, the second pulse in the sequence of three shown is always completely absent and the third pulse in the sequence of three is also only partially output. The missing power of the first and third pulse depends, among other things, on the time of the interruption in the first pulse. The total missing power over the three-way sequence depends essentially on the interruption duration, the pulse period and the ramp, which are essentially known to the system. The power not introduced by the Imax intervention can thus be averaged, and in the embodiment shown it is assumed that it is 200%, that is, the power of two full pulses is not introduced during an Imax intervention.

The above-mentioned values are of course only to be taken as examples and depend essentially on the set intervention parameters, such as interruption pause and, if applicable, ramps and parameters of the pulse sequence, such as pulse length and pulse period. In any case, the expert can recognize from the explanation above that different arc events can result in different arc interventions. These different interventions can, in turn, have different effects on the service provided. By determining the number of arc interventions (per type) and knowing the approximate effect on the performance introduced (defective performance), a corresponding control unit can determine a process time extension with a corresponding pulse sequence that can be appended to the normal process time in order to at least partially reduce the defective performance to compensate. This allows process results to be homogenized across several processes. The provision of a process time extension can also depend on whether a substantial number of arc interventions have taken place. In other words, if arc events occur below a threshold, the process can also be ended normally.

Of course, the control unit can also carry out a corresponding analysis of arc events and attacks for the attached process time and determine a further process time extension. It is also possible to output pulses with a lower power during the process time extension than during the normal process time. This can reduce the likelihood of further arc events, even if the attached process time is extended as a result.

The 7th and 8th show the above variants, the figures each showing an output pulse train (requested pulses are shown in dashed lines) within a process time (normal process end is shown by a broken line) to which a further pulse train (pulses with hatching) are appended within a process time extension. As can be seen, some pulses are incomplete due to an intervention by the arc unit and one pulse is not output at all). It is assumed that the interventions were carried out as described above or in a similar manner. The respective interventions are as Arc-Type 1 (corresponds to Micro-Arc-Event with corresponding intervention), Arc-Type2 (corresponds to U * I-Arc-Event with corresponding intervention) and Arc-Type3 (corresponds to Imax-Arc-Event with corresponding intervention ) and indicated for each pulse. Based on the number of interventions per type ((6 * Arc-Type1, 4 * Arc-Type2 and 1 * Arc-Type3) based on the above values, the control unit can determine that the output of approximately 4.6 pulses was too low .

In the 7th the process control has therefore determined the process time by a process time extension with a pulse sequence of 4 identical pulses corresponding to the pulses of the normal process time and added it directly to the normal process time. While 4.6 of this was rounded down to the last whole number, it can of course also be rounded off normally or always rounded up to the next higher whole number. It is also possible, for example, to output the last pulse with a power of 60% of the power of the other pulses. A precise assignment of the error to the appended with additional service is not necessary and, as described above, is not possible due to the lack of knowledge of the precise error occurring. Although the exact error could be determined more precisely by a correspondingly higher effort, this is usually not necessary, since the primary aim is to compensate for errors at least partially in order to be able to better comply with specified and possibly narrower tolerances than previously indicated. Even the distinction between arc types is not absolutely necessary if the controller can estimate an approximate loss of performance from a total number of arc events, for example if an approximate frequency of occurrence of the individual types is known, or for certain arc events, such as for example a micro-arc is not intervened at all. Therefore, there are a wide variety of options for estimating the defective performance and determining a corresponding process time extension with a pulse train.

In the 8th the process control has determined the process time by a process time extension with a pulse sequence of 9 identical pulses each with half the power of the pulses of the normal process time and added it directly to the normal process time. Choosing pulses with a lower power per pulse (with the same pulse duration) has the advantage that the probability is reduced that arcs occur during the process time extension. The resulting extension of the process time is usually not significant.

In the previous embodiments, the control unit has taken note of the arc events and only determined a process time extension. In one embodiment, however, it is also possible for the control unit to intervene in the normal pulse sequence via arc intervention. For example, it is known and possible that a control unit can completely abort a process if too many arc events occur, although in the past no distinction was made between arc types. As a result, the batch of substrates in the process was generally discarded completely or the substrates were reconditioned (the partially applied layer was removed, for example, by etching), inspected and re-coated).

According to the present embodiment, however, instead of aborting a process, when a certain number of arc events occur, the pulse sequence is changed in such a way that the power per pulse is reduced in order to increase the probability of arc events for the remaining pulse sequence to decrease. When deciding whether to reduce the performance, the total number of arc events in the previous process or the number of arc events per arc type can be used. In addition, a frequent occurrence of arc events within a shorter period of the process sequence can lead to a corresponding decision. In particular if several arc-type 3 arc events occur in quick succession, this can quickly lead to a corresponding decision to reduce the power of the pulses in the pulse train in order to protect the substrates and / or the substrate carriers. The pulse sequence is then continued until the end of the process time with lower power per pulse. On the basis of the number of arc events and the number of pulses with reduced power (and knowledge of the reduced power), a process time extension with a corresponding pulse sequence can then be determined and added to the normal process time.

A corresponding example is in 9 shown that again shows an output pulse sequence (requested pulses are shown with dashed lines) within a process time (normal process end is shown by a dashed line), to which a further pulse sequence (pulses with hatching) are attached within a process time extension. In the pulse train, arcs of types 2 and 3 appear in quick succession (four events within a pulse train of 10), whereupon the control for subsequent pulses reduces the power by half, which means that no further arc events occur, but A significantly lower performance than planned is also brought in within the process time. On the basis of the number of arc events and the number of pulses with reduced power (and knowledge of the reduced power), the controller determines a corresponding process time extension with a pulse sequence with pulses that each output a power reduced by 50% compared to a normal pulse .

The specified values with regard to the number of arc events that lead to an intervention beyond an arc intervention are of course only to be seen as examples. This also applies to the 50% power reduction. Rather, it is of course possible to first reduce the power by a lower value (e.g. 10%) in order to then see whether the power reduction is sufficient to sufficiently reduce the arc events. If this is not the case, there may be a further (or more) reduction (s) within the process sequence. The fact that the control unit is able to at least partially compensate for a malfunction resulting from this by extending the process time with a pulse sequence means that batches of substrate can possibly be processed successfully despite major problems in the process.

If an excessive number of arcs occurs, instead of reducing the power for subsequent pulses, the arc interventions could also be adapted by, for example, lengthening the respective interruption times and / or flattening the ramps for U * I arc intervention or Imax intervention . While these were previously designed in such a way that the intervention produced the least possible error, it can now be designed in such a way that it reduces the probability of subsequent arcs. The error that occurs in each case can be estimated and at least partially compensated for by extending the process time with a corresponding pulse sequence.

The possibility of extending the process time with a pulse train can increase the homogeneity of the process results and offers improved options for intervening in the process. In particular, arc interventions can primarily be designed in such a way that the risk of subsequent arcs is reduced (longer breaks, flatter ramps), since greater power losses can be compensated for. In addition, for example, the criteria for when an intervention takes place can also be adapted so that intervention may take place earlier or more often in order to avoid the Imax-Arc events as far as possible. This applies in particular to U * I events but also to micro-arc events.

The different approaches above can be combined with one another so that they are compatible with one another so that the person skilled in the art will find a wide variety of approaches depending on the process requirements. As stated, the power not fed in due to arcing can be fed in at the end of the normal pulse train. The process duration is therefore no longer stored as a fixed value, but is essentially controlled via the power fed in. This enables a much better integration of the plasma generator with the coating system as well as modified recipes.

New recipes would no longer have to define process duration and performance, but could count the amount of energy with a target performance. The target performance could be defined with wider tolerances than is usual with the currently common recipes. This means that the process controller could simply stop using an if query or a while loop as soon as the desired amount of energy was reported. This further change could be called alternatives to the process of querying the number of arcs at the end of the deposition and then calculating the additional time. It would be advantageous to calculate the correct active power (amount of energy delivered in the workpiece carrier). Modern plasma generators can determine the real power with the power fed in, the power taken away and the phase shift of the current. Inaccuracies should be kept as small as possible here, since the performance correction can otherwise lead to a larger deviation than if it is not made.

The invention has been described above on the basis of preferred embodiments of the invention, without being limited to the specific embodiments. In particular, the plasma generator can also only be used for one plasma unit. This can involve a wide variety of plasma units, in which the process result depends largely on the power applied. The number of plasma units to be supplied by a plasma generator can also differ from the number shown (here two).

QUOTES INCLUDED IN THE DESCRIPTION

This list of the documents listed by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent literature cited

• DE 102015004419 A1 [0003, 0025]
• DE 102018204585 [0028]

Claims (17)

A plasma treatment device comprising: a plasma generator that is configured to output pulses of electrical power to at least one process chamber, wherein the plasma generator is configured to generate and output the pulses over a process time according to a predetermined pulse sequence, and wherein the plasma generator has an arc unit that is suitable for a flashover or to recognize an imminent rollover, also known as an arc event, and, in response, to abort, change and / or fail a pulse that is or is to be output to the process chamber, which is referred to as arc intervention; and at least one control unit configured to To receive and evaluate process data about processes in the at least one process chamber and to use the process data to determine at least one process time and a pulse sequence to be provided by the plasma generator over the process time, or to receive a specific process time and pulse sequence, the pulse sequence being configured A predetermined power is to be introduced into the process chamber over the process time; - to control the plasma generator, to output the pulse sequence over the process time, - Obtaining data on arc interventions and using the data to determine a process time extension with a corresponding pulse sequence in order to at least partially compensate for a power reduction caused by the arc interventions. Plasma treatment device according to Claim 1 which has a unit which is configured to determine the number of arc interventions and to transmit the number to the at least one control unit. Plasma treatment device according to Claim 1 which has a unit which is configured to recognize at least two different types of arc interventions, to determine the number of each type of arc intervention and to transmit the number to the at least one control unit. Plasma treatment device according to Claim 2 or 3 , where the unit is part of the arc unit. The plasma treatment apparatus of any preceding claim, wherein the arc unit is configured to recognize at least two different types of arc events and to select different predetermined arc interventions corresponding to the arc event. Plasma treatment device according to Claim 5 , wherein the arc unit is configured to recognize at least two of the following arc events: an Imax event which indicates an overcurrent, ie a current above a threshold value, and thus the occurrence of an arc; - a U * I event that, before an overcurrent is reached, indicates an excessive current increase at a given voltage (Uxl) in such a way that an overcurrent would be expected if the power continued to be supplied; - a micro-arc, which is indicated by a current peak at a given voltage; and wherein the arc unit is further configured to completely abort the current pulse when an Imax event is detected and then slowly increase the power so that the following pulse may not be fully output either; - when a U * I event is recognized, to interrupt the current pulse and, as far as possible, to restart it within the planned pulse time with an increasing power ramp; and - when a micro-arc is detected, to interrupt the current pulse and to continue outputting it normally after a short waiting time. Plasma treatment device according to one of the preceding claims, wherein the control unit is configured to control the plasma generator when a certain number of arc events are reached within the process time, to reduce the power output per pulse of the remaining pulse sequence and a corresponding power reduction when determining the process time extension to take into account the corresponding pulse sequence. The plasma treatment device according to claim 1, wherein the control unit is configured to select the mean power of the pulses to be lower than the mean power of the pulses during the process time when determining the process time extension with a corresponding pulse sequence. Method for outputting pulses of electrical power to at least one process chamber, comprising: determining or receiving a process time and pulse sequence for the process chamber in / on a control unit; - Activation of a plasma generator in order to output the pulse sequence over the certain process time, - Detection of an arc event that indicates a rollover or an imminent rollover, - In response to the detection of an arc event, an arc intervention is carried out to cancel, change and / or omitting a pulse that is or is to be outputted to the process chamber, Obtaining data on arc interventions and using the data to determine a process time extension with a corresponding pulse sequence in order to at least partially compensate for a power reduction caused by the arc interventions; and - controlling the plasma generator in order to output the corresponding pulse sequence following the specific process time over the duration of the specific process time extension. Procedure according to Claim 9 , where the data on arc interventions includes the number of arc interventions. Procedure according to Claim 9 or 10 , wherein different types of arc interventions are performed, and wherein the data on arc interventions include at least the types of arc interventions and the respective number of arc interventions. Method according to one of the Claims 9 to 11 wherein the recognition of arc events, the recognition of at least two different types of arc events and the implementation of the arc intervention comprises the selection of a predetermined arc intervention from a plurality of predetermined arc interventions corresponding to the arc event and the implementation of the selected arc intervention. Procedure according to Claim 12 wherein the detection of arc events comprises the detection of at least two of the following arc events: an Imax event which indicates an overcurrent, ie a current above a threshold value, and thus the occurrence of an arc; - a U * I event that, before an overcurrent is reached, indicates an excessive current increase at a given voltage (Uxl) in such a way that an overcurrent would be expected if the power continued to be supplied; and / or - a micro-arc, which is indicated by a current peak at a given voltage; and wherein the execution of the arc intervention comprises the following: when an Imax event is detected, the current pulse is completely interrupted and the power is then slowly increased so that the following pulse may not be fully output either; - when a U * I event is recognized, to interrupt the current pulse and, as far as possible, to restart it within the planned pulse time with an increasing power ramp; and / or - when a micro-arc is detected, to interrupt the current pulse and to continue outputting it normally after a short waiting time. Method according to one of the Claims 9 to 13 When a certain number of arc events is reached within the process time, the plasma generator is activated to reduce the power output per pulse of the remaining pulse train, and a corresponding power reduction is taken into account when determining the process time extension with the corresponding pulse train. Procedure according to Claim 14 wherein the power reduction for the remaining pulses is at least 10%, preferably at least 20%. Method according to one of the Claims 9 to 15th , whereby when determining the process time extension with a corresponding pulse sequence, the average power per pulse is selected to be lower than the average power per pulse during the process time. Procedure according to Claim 16 , the mean power per pulse during the process time extension being selected to be at least 10%, preferably at least 20% lower than the mean power per pulse during the process time.

Images