KR101324346B1 - Method and system for detecting arc in a semiconductor process - Google Patents

Abstract

A method and system for detecting arcs occurring in a process chamber in a semiconductor process is disclosed. The arc detection method includes normalizing process row data output from a process chamber to obtain normalization data, and processing the obtained normalization data to detect whether an arc has occurred in the process chamber.

Description

Arc detection method and system in semiconductor process {METHOD AND SYSTEM FOR DETECTING ARC IN A SEMICONDUCTOR PROCESS}

The present invention relates to a method and system for detecting arcs occurring in a process chamber in a semiconductor process.

1 is a view schematically showing a structure inside a general process chamber.

As shown in FIG. 1, a semiconductor process such as a deposition process, an etching process, or the like is performed in the process chamber 100, and a target 102 for the deposition process or the like is installed at an upper end of the process chamber 100. RF power may be applied to a support for supporting a wafer positioned at a lower end of the target 102 and the process chamber 100.

Generally, the deposition process, the etching process, and the like are performed using plasma, but arcs may occur at the start of the process and during the process due to various causes. The arc may affect the target 102, the wafer, and the like to reduce the yield, and may affect the process chamber 100.

Therefore, there is a need to detect an arc in the process chamber 100, and a technique using an optical emission spectrometer has emerged. Specifically, in order to detect an arc generated during the process, the OES is connected to a specific port of the process chamber 100, and a photo sensor is connected to the OES. However, this method of using the OES has increased the cost of building a semiconductor process system because additional devices such as the OES and the photo sensor have to be installed. In addition, since the method of using the OES detects the light emitted from the plasma reaction during the process, the arc generated at the time of applying the RF power could not be detected.

The present invention provides an arc detection method and system capable of detecting an arc in a process chamber without installing a separate device.

In addition, the present invention provides an arc detection method and system capable of detecting arcs in a process chamber when RF power is applied as well as arc detection during process progress.

In order to achieve the above object, the arc detection method in a semiconductor process according to an embodiment of the present invention comprises the steps of: normalizing the process row data output from the process chamber to obtain normalized data; And processing the obtained normalization data to detect whether an arc has occurred in the process chamber.

 In the semiconductor process according to another embodiment of the present invention, the arc detection method may include obtaining process row data output from a process chamber; And processing the obtained process row data to detect whether an arc has occurred in the process chamber.

According to another embodiment of the present invention, an arc detection method in a semiconductor process includes selecting power related data among process row data output from a process chamber; And processing the selected power related data to detect whether an arc has occurred in the process chamber.

An arc detector used in a semiconductor processing system according to an exemplary embodiment of the present invention may include a parameter selector configured to select parameter data related to a predetermined parameter among process row data output from a process chamber; An algorithm applying unit which applies at least one algorithm to the selected parameter data; And an arc detector for detecting whether an arc is generated in the process chamber through a result of applying the algorithm.

An arc detection system according to an embodiment of the present invention includes a data collector configured to collect process row data output from a process chamber; And an arc detector which processes the process row data provided from the data collector to detect whether an arc has occurred in the process chamber.

The arc detection method and system according to the present invention detects whether an arc has occurred in the process chamber by using process row data (log data) output from the process chamber, and thus does not additionally install an apparatus in the process chamber for arc detection. Therefore, the construction cost of the semiconductor process system is not increased, and it can be saved as compared with the prior art.

In addition, since the process row data output from the process chamber is analyzed, the arc detection system may detect not only arc generation during the process but also arc generation at the time when RF power is applied.

1 is a view schematically showing a structure inside a general process chamber.
2 is a diagram illustrating an arc detection system according to an embodiment of the present invention.
3 is a flowchart illustrating an arc detection process according to an embodiment of the present invention.
4 is a diagram illustrating an output graph of parameter data according to an actual experiment.
5 is a diagram illustrating an arc detection result graph according to an actual experiment.
6 is a block diagram showing the structure of an arc detector according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

2 is a diagram illustrating an arc detection system according to an embodiment of the present invention.

Referring to FIG. 2, the arc detection system of the present embodiment detects arcs that may occur in the process chamber 200 during a semiconductor process, particularly a deposition process using a plasma, an etching process, and the like, and the data collection unit 202. ) And an arc detector 204.

Unlike the prior art, which detects an arc by sensing light generated from the reaction of the plasma and the process material, the arc detection system of the present invention detects the arc by using process raw data (log data) output from the process chamber 200. do. Specifically, process row data is output from the chamber 200 in real time at the start of the process, i.e., from when the RF power is applied to when the process is completed, and the arc detection system comprises a portion of the output process row data, That is, the data related to the preset parameters are selected, and the arc is detected using the selected data. Hereinafter, the data selected in relation to the parameters will be referred to as parameter data. Detailed description of the arc detection process will be described later with reference to the accompanying drawings.

The process chamber 200 is a chamber used in various processes such as an etching process and a deposition process, and in particular, refers to a chamber for performing a semiconductor process using plasma. Although not shown in FIG. 2, a target is attached to the upper end of the chamber 200, and an upper RF power part for applying RF power input from the outside to the target is formed, and a lower end of the chamber 200 supports the wafer. A lower RF power unit to which RF power is applied may be formed. Of course, the internal structure of the process chamber 200 may vary depending on the process. However, basically, RF power is provided to the process chamber 200, and the arc detection system may use data related to the RF power for arc detection.

The data collector 202 collects process row data output from the process chamber 200 in real time, and provides the collected process row data to the arc detector 204. Of course, the data collector 202 may transfer the process row data to the arc detector 204, and may provide the selected parameter data to the arc detector 204 after selecting the parameter data from the process row data. However, in consideration of the current semiconductor process system, the data collection unit 202 preferably collects the process row data and provides the same to the arc detector 204 as it is.

The arc detector 204 detects an arc using the process row data provided from the data collector 202. Specifically, the arc detector 204 selects parameter data related to predetermined parameters among the process row data, and applies an algorithm to the selected parameter data to determine whether an arc has occurred in the process chamber 200. Detect. A detailed description thereof will be described later.

In summary, the arc detection system of the present invention may detect whether or not an arc occurs in real time by using process row data (log data) output from the process chamber 200. Therefore, the manager can immediately take appropriate measures in the event of arcing, and as a result, the wafer and the process chamber 200 can be prevented from being damaged.

In the related art, an optical emission spectrometer for detecting light generated by a plasma reaction and a photo sensor connected thereto are additionally installed in the process chamber. However, the arc detection system of the present invention includes the process chamber 200 during the process. Since the process raw data output in real time from the use of a separate device for the arc detection does not need to be installed in the process chamber 200. That is, the arc detection system of the present invention can be implemented in a simple and inexpensive manner as compared with the prior art, and the process for determining whether or not the arc occurs as described below is also simple. In addition, the arc detection system can detect not only arcs generated during process progress but also arcs generated at the time RF power is applied.

In the above, the data collector 202 collects process row data output from the process chamber 200, but process row data output from the process chamber 200 without the data collector 202 is transmitted to the arc detector 204 in real time. It may be provided as. Of course, the data collector 202 and the arc detector 204 may be implemented as one device.

In addition, although the data collection unit 202 and the arc detector 204 is located in the process chamber 200 is connected to the wired and wireless spaced apart from Figure 2, the data collection unit 202 and the arc detector 204 is a process chamber It may be formed as part of the 200 or installed in the process chamber 200 itself. In this case, the arc detector 204 will transmit the arc detection result to the external device through a wired or wireless manner.

Hereinafter, the arc detection process of this embodiment will be described in detail with reference to the accompanying drawings.

3 is a flowchart illustrating an arc detection process according to an embodiment of the present invention, FIG. 4 is a diagram illustrating an output graph of parameter data according to an actual experiment, and FIG. 5 is a graph of an arc detection result according to an actual experiment. Figure is shown. The experimental data are data obtained during the etching process using plasma, and 100 MHz power was provided to the upper power part and the lower power part.

Referring to FIG. 3, the arc detection system acquires reference row data (offline data), which is data when operating normally in an etching process, and stores the obtained reference row data in the arc detector 204 (S300). ).

Subsequently, the arc detection system acquires process row data which is log data output from the process chamber 200 during the etching process, and selects parameter data related to preset parameters from the obtained process row data (S302). According to an embodiment of the present invention, the parameters are reflected power of 100MHz RF power provided to the upper RF power unit inside the process chamber 200, reflected power of 100MHz RF power provided to the lower RF power unit, lower RF power 2MHz bias voltage input to negative, Capacitor position related to 100MHz RF power provided to the lower RF power section, Capacitor position related to 2MHz bias voltage provided to the lower RF power section, 100MHz provided to the upper RF power section Capacitor position associated with RF power may be included. Of course, the arc detection system may detect whether an arc has occurred by selecting data related to other parameters. Parameter data related to the above six parameters is shown in Table 1 below.

Time (seconds) Parameter 1 Parameter 2 Parameter 3 Parameter 4 Parameter 5 Parameter 6 487 126 7 2923 31.53 64.8 51.85 488 126 7 2945 31.48 64.8 51.75 489 126 8 2920 31.45 64.95 51.83 490 127 8 2928 31.45 64.92 51.73 491 127 8 2933 31.5 64.9 51.8 492 132 131 2828 31.5 64.88 51.75 493 57 326 2720 37.7 54.5 51.28 494 99 311 3040 35 50.58 52.5 495 138 8 2963 26.88 61.7 51.6 496 118 8 2930 32.13 64.33 51.65 497 118 8 2923 31.93 64.3 51.68 498 118 8 2925 31.93 64.3 51.65 499 119 8 2923 32.05 64.38 51.65

Here, the parameters of Table 1 above represent parameter data, and time represents time elapsed from the start of the process, and process row data is continuously output from the process chamber 200 in real time, although not shown in Table 1 above. Parameter 1 represents real time output data of the reflected power of 100 MHz RF power provided to the upper RF power unit, and parameter 2 represents real time output data of the reflected power of 100 MHz RF power provided to the lower RF power unit. Means real-time output data of the 2MHz bias voltage input to the lower RF power unit. Parameter 4 represents the real time output data of the capacitor position associated with the 100 MHz RF power provided to the lower RF power section, and parameter 5 represents the real time output data of the capacitor position associated with the 2 MHz bias voltage provided to the lower RF power section. 6 shows real time output data of the capacitor position with respect to 100 MHz RF power provided to the upper RF power section.

Subsequently, the arc detection system applies a first algorithm to the reference row data and the process row data to normalize the process data to obtain normalized data μ _b (x) (S304). According to an embodiment of the present invention, the first algorithm may be a fuzzy model as shown in Equation 1 below.

Where x is process row data and r is reference row data.

According to Equation 1, the arc detection system normalizes the process row data within the 2r range based on the reference row data. In other words, the normalized data is normalized to a value between 0 and 1 (y-axis) between 0 and 2r range (x-axis). Normalizing the data in Table 1 yields the results shown in Table 2 below.

Time (seconds) Parameter 1 Parameter 2 Parameter 3 Parameter 4 Parameter 5 Parameter 6 487 0 0 0 0 0 0 488 0 0 0 0 0 0 489 0 0 0 0 0 0 490 0 0 0 0 0 0 491 0 0 0 One 0 One 492 0 One One One One One 493 0 One One 0 0 One 494 0 0 One 0 0 0 495 One 0 0 0 0 0 496 0 0 0 0 0 0 497 0 0 0 0 0 0 498 0 0 0 0 0 0 499 0 0 0 0 0 0

As shown in Table 2 above, the parameter data of the process row data is changed to "0" or "1" by normalizing. Looking at the graph, the parameter data shown in Fig. 4A is normalized as shown in Fig. 4B by applying a fuzzy model. Hereinafter, normalized parameter data will be referred to as normalized data.

According to an embodiment of the present invention, the arc detection system detects an arc generation expected time point through normalization data. Specifically, the arc detection system detects a time point associated with normalization data having many "1s" as an anticipation time of arc generation. Of course, the detected arc generation expected time may be a time when the arc is actually generated, or may be a time when no actual arc is generated. In Table 2, 492 seconds and 493 seconds are detected as anticipated timing of arc generation.

According to one embodiment of the invention, the arc detection system is a second algorithm, for example Dempster Shaper (Dempster) as shown in Equation 2 below to detect the time when the actual arc generation of the expected arc generation time (S206) whether or not an arc is generated by applying a coupling law of -Shafer theory.

Here, m1 denotes a normal probability / arc occurrence probability at the last time point, and m2 denotes a normal probability / arc occurrence probability at the current time point.

In the following, we will look at the process of detecting the arc by applying the coupling law of the damper shaper while looking at the actual application example.

First, it is detected whether an arc has occurred at 491 seconds.

Probability of Normal from Last Time
m1 (normal) = 1 Probability of Arcing at Immediate Time
m1 (arc) = 0 Probability of Normal at Current Time
m2 (normal) = 4/6 normal, 4/6 0 Arc occurrence probability at the current time
m2 (arc) = 2/6 2/6 arc, 0

이 되고, m(A, arc)는

이 된다. 즉, 491초에서는 정상일 확률인 m(A, normal)가 아크 발생 확률인 m(A, arc)보다 크기 때문에, 상기 아크 검출 시스템은 491초에서는 아크가 발생되지 않은 정상 상태로 결정한다. As shown in Table 3, at the immediately preceding point (490 seconds), since all normalized data have "0", the probability m1 (normal) that is normal is 1 and the arc occurrence probability m1 (arc) is 0. At the present time (491 seconds), only two of the normalized data have "1" and the rest of the data have "0", so the probability m2 (normal) is normal is 4/6 and the probability of arc occurrence m2 (arc) is 2/6. . Applying this probability to Equation 2 above, m (A, normal) is

Where m (A, arc) is

. That is, since the probability m (A, normal) of 491 seconds is larger than the probability of arc generation m (A, arc), the arc detection system determines that the arc state does not occur in 491 seconds.

Hereinafter, an arc occurrence at 492 seconds will be described.

Probability of Normal from Last Time
m1 (normal) = 4/6 Probability of Arcing at Immediate Time
m1 (arc) = 2/6 Probability of Normal at Current Time
m2 (normal) = 1/6 normal, 4/36 2/36 Arc occurrence probability at the current time
m2 (arc) = 5/6 20/36 arc, 10/36

이 되고, m(A, arc)는

이 된다. 즉, 492초에서는 아크 발생 확률인 m(A, arc)가 정상일 확률인 m(A, normal)보다 크기 때문에, 상기 아크 검출 시스템은 492초에서 아크가 발생한 것으로 결정한다. As shown in Table 3, since 4 normalized data have "0" and 2 normalized data have "1" at the immediately preceding time point (491 second), the probability m1 (normal) is 4/6 and the probability of arc occurrence is normal. m1 (arc) is 2/6. At the present time (492 seconds), 5 normalized data are "1" and the other normalized data has "0", so the probability m2 (normal) is 1/6 and the probability of arc occurrence m2 (arc) is 5 / 6. Applying this probability to Equation 2 above, m (A, normal) is

Where m (A, arc) is

. That is, since the arc occurrence probability m (A, arc) is greater than the probability of normality m (A, normal) at 492 seconds, the arc detection system determines that an arc has occurred at 492 seconds.

In summary, the arc detection system of the present invention may normalize by applying a fuzzy model to the parameter data, and detect whether an arc is generated by applying a coupling rule of the dempster shaper to the normalized data. Using this method, the arc detection system uses a method such that the actual arc of the anticipated arc generation points as shown in FIG. The generated time point can be detected accurately. Although not shown in the drawings, it was confirmed through experiments that the results were the same as when the arc was detected by the OES method.

In the above, the fuzzy model is applied to the process row data, but various algorithms may be modified as long as the process row data can be normalized using the reference row data.

In addition, the above method detects whether an arc has occurred by applying the coupling rule of the dempster shaper to the normalized data. Can be applied.

6 is a block diagram showing the structure of an arc detector according to an embodiment of the present invention.

Referring to FIG. 6, the arc detector 204 of the present embodiment includes a controller 600, a transceiver 602, a parameter selector 604, an algorithm applier 606, an arc detector 608, and a storage 610. ).

The transceiver 602 receives process row data from the data collector 202 and transmits data to other devices.

The parameter selector 604 selects parameter data related to preset parameters for arc detection among the process row data. In this case, the number of parameters is more than one as long as the arc can be detected.

An algorithm applying unit 606 applies a first algorithm, eg, a fuzzy model, to the parameter data to normalize the parameters to obtain normalization data, and a second algorithm, eg, a dempster, on the obtained normalization data. Apply the shaper's law of coupling.

The arc detector 608 detects whether an actual arc has occurred through the result of applying the second algorithm to the normalized data.

The storage unit 610 stores various data such as process row data, reference row data, algorithm programs, parameters, normalization data, and the like.

The controller 600 controls the overall operation of the components of the arc detector 204.

Although not mentioned above, the arc detector 204 transmits an arc detection result to a control device (not shown) that controls the operation of the process chamber 200 or provides information on an appropriate countermeasure when an arc is detected. It may further include an arc counter to provide. Of course, it is preferable in the process that the arc detector 204 transmits only the arc detection result to the control device and the control device takes appropriate measures.

The embodiments of the present invention described above are disclosed for purposes of illustration, and those skilled in the art having ordinary knowledge of the present invention may make various modifications, changes, and additions within the spirit and scope of the present invention. Should be considered to be within the scope of the following claims.

100: process chamber 102: target
200: process chamber 202: data collection unit
204: arc detector 600: control unit
602: transceiver unit 604: parameter selection unit
606: algorithm application unit 608: arc detection unit
610: storage unit

Claims (14)

Normalizing the process row data output from the process chamber to obtain normalization data; And
Processing the obtained normalization data to detect whether an arc has occurred in the process chamber. The method of claim 1, wherein obtaining the normalization data comprises:
Selecting parameter data related to preset parameters from process row data output from the process chamber; And
Normalizing the parameter data by applying the selected parameter data and reference row data to a fuzzy model,
The reference row data is an arc detection method in a semiconductor process, characterized in that the data when the normal operation without the arc is generated in the process chamber. The method of claim 2, wherein the semiconductor process is an etching process, and the parameters are reflected power of 100 MHz RF power provided to an upper RF power part within the process chamber, reflected power of 100 MHz RF power provided to a lower RF power part, 2MHz bias voltage input to the bottom RF power section, Capacitor position related to 100MHz RF power provided to the bottom RF power section, Capacitor position related to the 2MHz bias voltage provided to the bottom RF power section, Provided to the top RF power section And a capacitor position associated with said 100 MHz RF power. The method of claim 1, wherein detecting whether the arc has occurred comprises:
Analyzing the obtained normalized data to detect arc generation prediction points; And
And detecting whether an arc has occurred by applying normalization data of a specific arc occurrence predicted time point and normalization data of a time immediately before the specific arc occurrence predicted time point to an arc detection algorithm. The method of claim 4, wherein detecting whether an arc occurs by applying the arc detection algorithm comprises:
Acquiring a new normal probability and an arc occurrence probability by combining the normal probability and the arc occurrence probability for the normalized data of the specific arc occurrence predicted time point with the normal probability and the arc occurrence probability for the normalized data of the last time point; And
And determining that the arc has occurred at the specific arc occurrence predicted time point when the obtained arc occurrence probability is greater than the normal probability. The method of claim 4, wherein the arc detection algorithm is a combination law of Dempster-Shafer theory. Obtaining process row data output from the process chamber; And
Processing the obtained process row data to detect whether an arc has occurred in the process chamber,
Detecting whether the arc has occurred,
Selecting parameter data related to preset parameters from process row data output from the process chamber;
Applying the selected parameter data and reference row data to a fuzzy model to normalize the parameter data to obtain normalized data;
Analyzing the obtained normalized data to detect arc generation prediction points; And
Detecting normality of arc by applying normalization data of a specific arc occurrence predicted time point and normalization data of a time immediately before the specific arc occurrence predicted time point to an arc detection algorithm;
The arc detection algorithm is a combination law of a dempster shaper, and the reference raw data is data when the arc operates in the process chamber without generating an arc. delete Selecting power related data among process row data output from the process chamber; And
Processing the selected power related data to detect whether an arc has occurred in the process chamber,
The power related data may include data related to the reflected power of the RF power provided to the upper RF power part within the process chamber, data related to the reflected power of the RF power provided to the lower RF power part, and a bias voltage input to the lower RF power part. Data relating to capacitor position associated with the RF power provided to the lower RF power unit, data relating to capacitor position associated with the bias voltage provided to the lower RF power unit, and the RF power provided to the upper RF power unit; And at least two of the data associated with the associated capacitor position. The arc detection method of claim 9, wherein the arc generation is detected by applying a combination law of a fuzzy model and a dempster shaper to the selected power-related data. A parameter selector configured to select parameter data related to a predetermined parameter among process row data output from the process chamber;
An algorithm applying unit which applies at least one algorithm to the selected parameter data; And
And an arc detector for detecting whether an arc is generated in the process chamber through the result of applying the algorithm. The method of claim 11, wherein the algorithm applying unit applies a fuzzy model to the selected parameter data to normalize the selected parameter data, and applies a combination rule of a dempster shaper to the normalized data that is a result of the normalization to generate a normal probability and an arc. Detect probabilities,
The semiconductor processing system is characterized in that the arc detection unit uses the reference raw data that is the data when the arc is not generated when the fuzzy model is applied, and the arc detection unit determines that the arc has occurred when the arc occurrence probability is greater than the normal probability. Arc detector used for. A data collector configured to collect process row data output from the process chamber; And
And an arc detector which processes the process row data provided from the data collection unit and detects whether an arc has occurred in the process chamber.
The arc detector selects parameter data related to preset parameters among the process row data, normalizes the selected parameter data, detects a normal probability and an arc occurrence probability by using normalization data resulting from the normalization, And determining that the arc has occurred when the arc occurrence probability is greater than the normal probability. The arc detector of claim 13, wherein the arc detector applies a fuzzy model to the selected parameter data to normalize the selected parameter data, and applies a combination rule of a dempster shaper to the normalized data to determine a normal probability and an arc occurrence probability. And detecting and using reference row data which is data when an arc is not generated when the fuzzy model is applied.

Images