JPH06318571A - End point monitor - Google Patents

Abstract

PURPOSE:To let the equipment hardly influenced by noise from a high-frequency power source, etc., and then to make it fully cope with complication of end point detection by constituting the equipment with a polychromator to be connected to a processor which conducts a specified processing and a programmable computer. CONSTITUTION:A polychromator 2 is connected to a processor 1 through a condensing lens 41 and an optical fiber 4. A computer 3 is connected to the polychromator 2 through a cable 5. To the computer 3, a display 6 and a keyboard 7 are connected. The computer 3 has a start signal input terminal 8 and an end point signal output terminal 9. Due to this structure, lights of a plurality of wavelengths can be simultaneously measured and a very precise end point detection can be conducted based on the mutual relations between the lights of a plurality of wavelengths. Furthermore, the equipment can cope with complication of end point detection in many kinds of etching processes.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an end point monitor for detecting the end of a reaction process in a semiconductor manufacturing apparatus or the like.

[0002]

2. Description of the Related Art A conventional endpoint monitor is a monochromator that allows only a desired spectrum to pass from the light generated in a chemical reaction chamber, that is, a chamber that constitutes a semiconductor manufacturing apparatus, and an intensity change of an emission spectrum with time. It is composed of a photosensor for converting into a corresponding waveform signal and a computer for calculating the waveform signal to determine the end of the reaction process. Such end point monitors are disclosed, for example, in Japanese Patent Laid-Open Nos. 63-107026 and 4
-37127, Unexamined-Japanese-Patent No. 4-177833,
It is disclosed in Japanese Patent Laid-Open No. 4-186727. Depending on the type of reaction process, the spectral distribution of the plasma light generated in the chamber has a maximum intensity with the reaction time at a certain wavelength, for example, and then disappears. Therefore,
A monochromator that passes only specific wavelengths is attached to the glass window of the chamber directly or through an optical fiber,
A change in the intensity of the emission spectrum of this wavelength is converted into a corresponding waveform signal by a photosensor or the like. The computer detects the end point of the reaction process by arithmetically processing the waveform signal according to a previously constructed algorithm. For example,
The end of the reaction process is determined by measuring the level point and end point of the input waveform signal. Specifically, the minimum peak is detected after the waveform signal has dropped below a predetermined level point.

[0003]

However, in the conventional endpoint monitor, the waveform signal data captured from the monochromator, which is a variable type single wavelength optical measuring instrument, is processed by the fixed algorithm to detect the end point. Therefore, there are the following problems to be solved. First, in order to specify the wavelength specific to each reaction process, a spectral analyzer such as a spectrum analyzer is required in addition to the monochromator, which has led to an increase in the price of the endpoint monitor. Also, since the end point was detected only by the single emission spectrum separated by the monochromator,
It is difficult to obtain accurate discrimination results for reaction processes in which the spectrum distribution changes intricately. Second,
A computer that performs arithmetic processing of an input waveform signal has a problem that it does not necessarily give an optimum end point determination result to a wide variety of reaction processes because it performs processing by a fixed algorithm. Since there is no freedom in programming the algorithm, it was not always functionally satisfactory.

[0004]

Means for Solving the Problems In order to solve the above-mentioned problems of the conventional technique, the following means were taken. That is, the endpoint monitor according to the present invention includes a polychromator connected to a processing device that performs a predetermined process, and a programmable computer. The polychrometer is capable of receiving wide band light generated in the processing device during processing and separating it into a plurality of spectra, and outputs a corresponding waveform signal by measuring the intensity change of these spectra over time. The computer can optionally program a desired endpoint detection algorithm using a formalized and simplified programming language, wherein at least one specific waveform signal is arithmetically processed according to the endpoint detection algorithm to perform the processing. Detect the end point. This computer can also carry out arithmetic processing on a plurality of waveform signals corresponding to different spectra at the same time according to the end point detection algorithm to detect the end point. The computer can program the desired endpoint detection algorithm by a combination of formulated functions and operators.
The polychromator preferably receives light generated in the processing device through an optical fiber.

[0005]

According to the present invention, the end point of the reaction process is detected by incorporating a polychromator which is a broadband spectroscope. For example, the emission wavelength can be measured over a wide range from the near-ultraviolet region to the near-infrared region, and the spectrum required for end point detection can be specified. Therefore, unlike an endpoint monitor using a conventional monochromator, a spectrum scope function can be obtained without separately installing a spectrum analyzer. Further, since multiple wavelengths can be measured simultaneously, it is possible to perform precise and sophisticated end point detection by combining a plurality of specific spectra. For example, by performing background processing with reference light, it is possible to reduce the drift and noise of the light emission amount and extract the signal required for the end point detection calculation. Since the polychromator detects the light received through the optical fiber, it is less susceptible to noise from the high frequency power source and the like built in the processing device. Due to the accumulation effect of the photoelectric sensor incorporated in the polychromator, a high-level waveform signal can be obtained, and there is also a low-pass filter effect, so the detection error can be reduced.

On the other hand, the computer used in the present invention is programmable, and the operator can freely construct a desired end point detection algorithm. This makes it possible to sufficiently deal with the complicated end point detection in a wide variety of reaction processes. In addition, the waveform signal once sampled can be recorded and saved. By reading the recorded data and performing a simulation,
Verification by different judgment algorithms is possible. Therefore, in the process of setting the optimal judgment algorithm,
The operation of the processing equipment can be minimized without the need to repeat the actual reaction process.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described in detail below with reference to the drawings. FIG. 1 is a block diagram showing the basic configuration of an endpoint monitor according to the present invention. As shown in the figure, the endpoint monitor comprises a polychromator 2 connected to a processing device 1 for performing a predetermined process, and a programmable computer 3. The polychrometer 2 is connected to the processing device 1 via a condenser lens 41 and an optical fiber 4. The condenser lens does not necessarily have to be used. The computer 3 is connected to the polychrometer 2 via a cable 5.
The computer 6 has a display 6 and a keyboard 7
Is also connected. It also has a start signal input terminal 8 and an end signal output terminal 9.

In this example, the processing device 1 which is the end point detection target
Is a plasma etching apparatus. Polychromator 2
Can receive wide band light generated in the plasma processing apparatus 1 during plasma etching and separate it into a plurality of spectra. The intensity change of these spectra is measured with time and a corresponding waveform signal is output. The computer 3 can arbitrarily program a desired end point detection algorithm using a formalized and simplified programming language, and performs plasma etching processing by arithmetically processing at least one specific waveform signal according to the end point detection algorithm. The end point of is detected. The computer 3 can also perform arithmetic processing on a plurality of waveform signals corresponding to different spectra at the same time according to the endpoint detection algorithm to detect the endpoint.
The computer 3 also programs a desired end point detection algorithm with a combination of formulated functions and operators.

The operation of the endpoint monitor will be described with reference to FIG. When a start signal is applied to the input terminal 8 in response to the start of the plasma etching process, the computer 3 sends a data collection command to the polychrometer 2 via the cable 5. On the other hand, the plasma light generated in the processing apparatus 1 is input to the polychrometer 2 via the condenser lens 41 and the optical fiber 4.
The broadband plasma emission is converted into a corresponding waveform signal by the polychromator 2 and transmitted to the computer 3.
The computer 3 collects data based on an algorithm created by the operator in advance. Further, the arithmetic processing of the waveform signal data is performed based on the end point detection algorithm created by the operator in advance. This processing includes a logical operation, a numerical operation, and a time operation, and performs a process corresponding to a complicated temporal change of plasma light. As a result, when the etching end point is detected, an end signal is sent from the output terminal 9 to give a process end command to the plasma etching processing apparatus 1. The desired algorithm is written in the computer 3 via the keyboard 7. The input waveform signal and the like are displayed on the display 6 as needed.

A detailed description will be given below of each of the processing device 1, which is an end point detection target, the polychrometer 2 and the computer 3 which constitute the end point monitor. First, the processing apparatus 1 will be described with reference to FIG. 1 again. As described above, the processing apparatus 1 is a plasma etching apparatus,
This example has a planar structure. This is an anode-coupled parallel plate type device, and a ground electrode 11 is provided inside a chamber 10 made of quartz or the like. The semiconductor wafer 12 is placed on the ground electrode 11 and etching is performed. After evacuating the chamber 10, the gas inlet 13
While introducing a predetermined reaction gas from the RF electrode 14
Then, a predetermined high frequency power is supplied to generate a plasma 15, and the semiconductor wafer 12 is dry-etched. Examples of the type of etching treatment include SiO ₂ dry etching. It is mainly used in the contact window opening process. Since it is necessary to accurately form fine contact holes in the SiO ₂ film formed on the surface of the silicon wafer, sufficient selectivity and anisotropy are required. Usually, in order to increase the selection ratio, CF ₄ + H ₂ , C ₂
A reaction gas such as F ₆ , C ₃ F ₈ or CHF ₃ is used.
Since the surface of the silicon wafer is not damaged in this plasma etching process, accurate end point detection is necessary. Therefore, for example, the spectrum from fluorine atoms, which increases with the progress of etching, is selectively monitored. Besides, S
In some cases, i ₃ N ₄ dry etching is performed. Si
₃ N ₄ is used as an oxidation resistant mask when performing selective oxidation on a field region or an insulating region of an LSI, and is usually deposited so as to be overlaid on a SiO ₂ film on a silicon wafer. This can be etched using, for example, CF ₄ gas. It is possible to detect the end point by monitoring the emission spectrum from nitrogen atoms which decreases with the progress of etching. Although the end point of the dry etching process is detected in the above example, the applicable range of the end point monitor according to the present invention is not limited to this. For example, it is also possible to use it for detecting the end point of the film forming process with a CVD device, a sputtering device or the like.

Next, the polychrometer 2 will be described with reference to FIG. An optical fiber 4 is connected to the polychromator 2 via an adapter 20, and a processing device (not shown)
Receives the light generated at. An entrance slit 21 is installed in the adapter 20. A spectroscopic optical system 22 is housed inside the polychromator 2 to disperse plasma emission to obtain a desired spectrum. A multi-channel detector 23 is arranged on the emission side of the spectroscopic optical system 22. This detector 23 is a linear image sensor 2
4 is provided, and it is possible to simultaneously observe the spectrum from the near-ultraviolet region to the near-infrared region. For example, the linear image sensor 24 is composed of an array of 512 elements, and the light receiving area of each element is 50 μm × 2.5 mm. The detection wavelength range is 200 to 800 nm. Linear image sensor 2
The output from 4 is relayed through the data processing unit 25 and transmitted to the computer (not shown) via the cable 5.
Due to the storage effect of the linear image sensor 24, a high-level waveform signal can be obtained, and a low-pass filter effect can also be obtained. This polychromator 2 has a function of an optical spectrum analyzer and can specify a desired spectrum wavelength.

Next, the programming function of the computer will be described. The endpoint monitor according to the present invention uses a uniquely formulated and simplified control program language to perform logic operation and analog data processing to execute optimum endpoint detection. This control program language was originally devised for the endpoint monitor, and can be handled relatively easily as compared with general program creation. As a program creation procedure, first, a control flow is created and a desired end point detection algorithm or determination algorithm is diagrammed. Next, a program list is created by combining predetermined functions and operators according to this control flowchart. After that, the editing program installed in the computer is activated, and the program list created for the desired recipe number is entered using the keyboard and the display such as the CRT. Next, the decryption program installed in the computer is activated and executed. If there is a grammatical error in the created program, it will be displayed on the display. When the program is decrypted,
Store on disk. These series of tasks can be carried out interactively by a support program installed in the computer. This is the end of the program creation work. When using this on an actual machine, it can be used repeatedly just by specifying the recipe number.

The specifications of the programming language adopted in the present invention will be described in detail below with reference to Tables 1 to 10 and FIGS. First, the programming environment will be described. Table 1 below shows the user information area. The user can freely handle two types of information areas, bit information and register information. Up to 64 pieces of bit information can be set, and the data range is 0 or 1. Bit information is represented by b, and particularly when specifying individual bit information,
For example, the 25th bit information is represented by b24. Up to 64 pieces of register information can be set. The data range can take integer values from -32767 to +32767. The register information is represented by r. When individually specified, for example, the 13th register information is displayed by r12.

[Table 1]

Next, the input / output specifications of the computer 3 will be described with reference to FIG. External information surrounding the computer 3 includes input and output data, and a desired function is achieved by monitoring and controlling these. As input, up to 4 waveform signals can be accepted at the same time. This is register information and is represented by wavelength data 1 to wavelength data 4. This wavelength data represents the intensity or energy of a specific spectrum, and the larger the value, the stronger the energy. On the output side, a maximum of four pieces of output data 1 to output data 4 can be obtained as register information. This output data represents a predetermined calculation result. In addition to this, a pair of end signal and alarm signal are obtained as bit information. The end signal is a signal indicating the time point at which the end point is detected, and the alarm signal indicates the occurrence of an abnormal state.

Table 2 below shows the wavelength data, output data, number of end signals, alarm signals, data range, and corresponding information type as a summary. As mentioned above, up to 4 wavelength data can be input, and 0 as register information.
It can take an integer value up to 32767. Output data is 4
It is possible to send up to a maximum of -3, and register information of -3276
It can take an integer value from 7 to +32767. The end signal can take 0 or 1 as bit information. A value of 1 puts the end signal into a valid state. Similarly, the alarm signal can take a value of 0 or 1 as bit information. Enabled with a value of 1.

[Table 2]

Next, the system state transition of the endpoint monitor will be described with reference to FIG. As shown in the figure, the system state includes a time zone in which the program is actually executed and a time zone other than that. Each of these has a meaning as a system state (status) and affects the overall function. The timing and contents of these statuses will be described.
RF On represents the generation timing of a system activation signal input from the outside. RF Off indicates the generation timing of the system end signal (end signal) sent to the outside. The status Dead time is a time zone provided for not performing the end point determination during the initial unstable period. It is a time period from the input of the system activation signal to the elapse of a set time (Dead time). The status Exec time is a time zone during which the program is being executed. This is the time from the end of Dead time to the generation of the end signal End or the alarm signal Alarm. The status Over time is a spare time zone provided to compensate for the insufficient etching portion. After the end signal End is generated, the set time (Ove
r time) is the time period until the elapse. In addition,
When the alarm signal Alarm is generated, this status does not exist and the system end signal is output immediately. Finally, Scan interval represents the execution cycle of the program and is set to a predetermined value. The dotted line portion in each execution cycle represents the time zone during which the program is actually operating.

Next, the system setting will be described. This is to set the basic value of the system necessary for the judgment operation. Table 3 below shows the types of setting statements and description methods.

[Table 3]

When the Scan interval is set, as shown in Table 3 above, for example, scaniv
Write 2.5. Note that #set is added to the head of the description to clarify that it is a system setting. This sets the interval at which the program will run,
It can be set in 0.1 second increments in the range of 0.1 second to 10.0 seconds. Especially, if not specified, it is automatically set to 1.0 second. Regarding Dead time, for example, deadt
Describe as m 30. This is set so that the end point determination is not performed in the initial unstable period. The control operation starts after the elapse of this time from the input of the activation signal from the outside. It can be set in 1 second increments from 0 to 99 seconds. Unless otherwise specified, it is set to 0 seconds. Ove
The r time is described as, for example, overtm 12. This is to set a preliminary time for compensating for the insufficient etching portion. Actually, it is the time from the end point determination (End) until the RF power is turned off, and it has no meaning when an alarm signal is generated.
It can be set in 1 second increments from 0 to 99 seconds. Unless otherwise specified, it is automatically set to 0 seconds. Storage
time represents the accumulation time of light energy in the polychrometer. The light energy increases proportionally as the time increases. For example, it is set as strgtm 1.0. It can be set in 0.1 second increments in the range of 0.1 second to 12.0 seconds. Unless otherwise specified, it is automatically set to 1.0 second. Input gain represents the input amplification factor of light energy in the polychromator. In order to set this, it is described as, for example, igain 4. This amplification factor can be set in increments of 1 to 8, and an amplification factor of 1 is automatically set unless otherwise specified. Input Waves 1 to 4 represent input wavelengths, and a maximum of 4 types of wavelengths can be selected. As a description method, for example, it is expressed as iwave1 420.3. It can be set in steps of 0.1 nm from 200.0 nm to 800.0 nm.

Next, the input / output setting will be described. This is for allocating input and output information required by the control program. The types and description methods are shown in Table 4 below. To clarify that it is an input / output setting statement, #
Following def, a predetermined description is given.

[Table 4]

First, #def inputs 1-4 are #
Input W set in set iwaves 1 to 4
It is used to assign register information to ave1 to ave4. #Def outputs 1 to 4 perform register information allocation for outputting the calculation result to the outside. #Def end allocates bit information for outputting as an end signal. #Def alarm
Assigns bit information for outputting as an alarm signal.

Next, the programming language specifications will be described. This programming language includes expressions that represent a single value by combining certain functions and operators. The operators used are shown in Table 5 below.

[Table 5]

In this programming language specification, a statement is described in the following form. [Return value] [=] Function name ([argument], [argument], [...]) When adding a comment to each sentence, delimit with (;) and display.

Next, a list of functions used in this programming language specification is shown in Table 6 below, and a brief description of each function will be added below.

[Table 6]

The logical sum function or is expressed as follows. or (B, B1, B2, B3, B4, B5, B6, B
7, B8) Here, B can take a bit value or an immediate value 1. The arguments B1 to B8 are bit information. The return value of this logical sum function is given by bit information. As a function, when the argument B has the bit value 1 or the immediate value 1, the logical sum operation is executed. A logical sum of up to eight bit information is calculated and assigned to the output bit. B1 to B8 are variable in number and require at least one bit information. When the argument B is 0, the output bit does not change. For example, when the logical sum of the bit information b2 to b6 is obtained and substituted for b10 under the condition of the bit information b1, the following expression is used.

B10 = or (b1, b2, b3, b4, b5, b6) Alternatively, when the logical sum of b2 to b6 is always taken and substituted into b10, it is expressed as follows. b10 = or (1, b2, b3, b4, b5, b6)

The logical product function is expressed as follows. and (B, B1, B2, B3, B4, B5, B6, B
7, B8) Here, the argument B can take a bit value or an immediate value 1. The other arguments B1 to B8 are bit information. The return value of this logical product function becomes bit information. As a function, B is bit value 1
Alternatively, if the immediate value is 1, the logical product is executed. A logical product of up to eight pieces of bit information is taken and assigned to the output bit. B1-B
8 is a variable number and requires at least one bit information.
When the argument B is 0, the output bit does not change. As an example of use, when a logical product between b2 and b5 is obtained and substituted into b10 under the condition of b1, the following expression is used. b10 = and (b1, b2, b3, b4, b5) Alternatively, when the logical product between b2 and b5 is taken and substituted for b10 with b1 as a negative condition, it is expressed as follows. b10 = and (! b1, b2, b3, b4, b5)

The exclusive OR function is expressed as follows. xor = (B, B1, B2, B3, B4, B5, B6
B7, B8) Here, the first argument B can take a bit value or an immediate value 1.
The remaining arguments B1 to B8 can each take bit information. The return value of this exclusive OR function is bit information. As a function, an exclusive OR of up to 8 pieces of bit information is taken and assigned to the output bit. B1 to B8 are variable numbers and at least 1
Bit information for each piece is required. When the argument B is 0, the output bit does not change. Here, the exclusive OR means that the output is 0 when all the bits are the same, and the output is 1 when even one bit is different. As a usage example,
The exclusive OR of b2 and b3 is obtained on condition that b1 is b.
When substituting into 10, it is expressed as follows. b10 = xor (b1, b2, b3)

The transfer function is represented as mov (B, R). The first argument B can be a bit value or an immediate value 1. The next argument R is register information or an immediate value (-32767).
~ + 32767). The return value becomes register information. As a function, when the argument B has the bit value 1 or the immediate value 1, the movement process is executed. That is, the argument R is assigned to the output register. When the argument B is 0, the output register does not change. As an example of use, r10 is replaced by r1 with b1 as a condition.
When moving to 11, it is written as follows. r11 = mov (b1, r10) Also, when always moving r10 to r11, it is expressed as follows. r11 = mov (1, r10) Alternatively, when it is desired to reset r11 to 0 only when b1 is 0, it is expressed as follows. r11 = mov (! b1,0)

The addition / subtraction function is expressed as add (B, R1, R2). The first argument B can be a bit value or an immediate value 1. The argument R1 is register information. Argument R
2 is register information or an immediate value (-32767 to +3276)
7) can be taken. The return value of this addition / subtraction function is register information. As a function, the argument B is a bit value 1 or an immediate value 1
In case of, addition / subtraction is executed. That is, the argument R1 is replaced by the argument R2
And add it to the output register. When the argument B is 0, the output register does not change. As an example of use, when r10 and r11 are added under the condition of b1 and substituted into r12, it is expressed as follows. r12 = add (b1, r10, r11) Further, r1 is subtracted from r10 under the condition of b1 to obtain r1.
When substituting into 2, it is expressed as follows. r12 = add (b1, r10, -r11) Alternatively, when b1 is used as a condition and immediate value 150 is subtracted from r10 and substituted into r12, the following expression is made. r12 = add (b1, r10, -150)

The multiplication / division function is multi (B, R1, R2,
Notated as R3). Argument B is bit value or immediate value 1
Can be taken. The argument R1 can take register information as a numerator. The argument R2 also takes register information or an immediate value (-32767 to +32767) that is also a numerator. The final argument R3 is the denominator register information or the immediate value (-32767).
~ +32767). The return value of this multiplication / division function is register information. As a function, when the argument B has the bit value 1 or the immediate value 1, the multiplication / division is executed. That is, R1 is multiplied by the numerator R2 and the denominator R3, and the result is substituted into the output register. B
When is 0, the output register does not change. As an example of use, r10 × r11 / r12, for example, under the condition of b1
When is substituted for r13, it is expressed as follows. r13 = multi (b1, r10, r11, r12) Further, when r10 × r11 is substituted for r13 under the condition of b1, it is expressed as follows. r13 = multi (b1, r10, r11, 1) Alternatively, when substituting r10 / 300 for r13 with b1 as the condition, it is expressed as follows. r13 = multi (b1, r10, 1,300)

The integral function is represented by intg (B, R). The argument B can be a bit value or an immediate value 1. The other argument R is register information, and the return value of this integration function is also register information. As a function, when the argument B has the bit value 1 or the immediate value 1, the integral operation is executed. That is, the integrated value of the argument R is assigned to the output register. Actually, the argument R obtained in the current scan operation is added to the output register value obtained in the previous scan operation and substituted in the output register. When the argument B is 0, the output register does not change.
As an example of use, when r10 is stored in r50 under the condition of b20, it is expressed as follows. r50 = intg (b20, r10) When r10 is always stored in r50 for each scanning operation, it is expressed as follows. r50 = intg (1, r10) or r1 when b20 is set to bit value 1
If 0 is stored in r50 and r50 is cleared when b20 is set to the bit value 0, it is expressed as follows. r50 = intg (b20, r10) r50 = mov (! b20,0)

The differential function is expressed as diff (B, R). The first argument B takes a bit value or an immediate value 1. The next argument R is register information, and the return value is also register information. As a function, when the argument B has the bit value 1 or the immediate value 1, the differential operation is executed. That is, the differential value of the argument R is assigned to the output register. Actually, the value obtained by subtracting R obtained in the previous scan operation from the argument R obtained in the current scan operation is substituted in the output register. However, in the first scan operation after the program is started, the value of R before that is undefined, so that the value 0 is unconditionally assigned to the output register. When B is 0, the output register value does not change. As an example of use, for example, b20
When r10 is differentiated to r50 under the condition of, the notation is as follows. r50 = diff (b20, r10) Alternatively, when r10 is always differentiated to r50 for each scanning operation, it is expressed as follows. r50 = diff (1, r10)

The comparator function is cmp (B1, B2
Notated as R1, R2). The argument B1 can take a bit value or an immediate value 1. The next argument B2 can be an immediate value 0 or 1. When the immediate value is set to 0, it means that the comparison determination is performed in the forward direction. On the contrary, when the immediate value is set to 1, it means that the comparison determination is performed in the negative direction. Argument R1
Is register information, and the argument R2 can take register information or an immediate value (-32767 to +32767). The return value of this comparator function is given as bit information. As a function, the comparison operation is executed when the argument B1 has a bit value 1 or an immediate value 1. Specifically, when the argument B2 indicates the determination in the forward direction, the output bit is 1 when R1 ≧ R2.
Becomes On the contrary, when the argument B2 indicates the determination in the negative direction, the output bit becomes 1 when R1 ≦ R2. In other cases, the output bit is 0. When B1 is 0, the output bit value does not change. As an example of use, when r1 is larger than r2 under the condition of b1, b2
When is set to 1, the notation is as follows. b2 = cmp (b1,0, r1, r2) Alternatively, when b1 is set to 1 when r1 is smaller than r2 under the condition of b1, it is expressed as follows. b2 = cmp (b1,1, r1, r2) Further, when b1 is set to 1 when r1 is larger than the immediate value 1500 under the condition of b1, it is expressed as follows. b2 = cmp (b1,1, r1,1500)

The absolute value function is expressed as abs (B, R). The argument B takes a bit value or an immediate value 1. Other argument R
Is register information, and the return value of this absolute value function is also register information. As a function, when the argument B has the bit value 1 or the immediate value 1, the absolute value operation is executed. That is, the absolute value of the argument R is assigned to the output register. The argument B is 0
In case of, the output register does not change. As an example of use, the absolute value of r10 is set to r11 under the condition of b30.
When it is taken, the notation is as follows. r11 = abs (b30, r10)

The square root function is expressed as sqr (B, R). The argument B is a bit value or an immediate value 1, and the argument R is register information. The return value of this square root function also becomes register information. As a function, the square root operation is executed when the argument B has a bit value 1 or an immediate value 1. That is, the square root of the argument R is substituted into the output register. When B is 0, the output register does not change. When the argument R has a negative value, the output register is set to 0. As a usage example,
When the square root of r10 is taken as r11 under the condition of b30, it is expressed as follows. r11 = sqr (b30, r10)

The timer function is timer (B, B1, R)
It is written like. The argument B takes a bit value or an immediate value 1, the next argument B1 takes a bit value, and the last argument R takes an immediate value (0 to +32767). The return value of this timer function is given as bit information. As a function, when B has a bit value 1 or an immediate value 1, a timer operation is executed. Specifically, after the argument B1 is set to 1, the immediate value R × (S
The output bit becomes 1 after a predetermined time given by (can interval). When the argument B1 is 0, the output bit immediately becomes 0. As an example of use, when the output bit b6 is set to 1 after 60 seconds under the condition of b5, it is expressed as follows. However, Scan int
It is assumed that the interval is set to 2.0. b6 = timer (1, b5,30)

The start function is written like start () and has no argument or return value. The function of this start function is to show the beginning of the program description. Similarly, the end function is written like end () and does not include an argument and a return value. The function of this end function is to indicate the end of the program description.

It is necessary to perform a series of programming operations in order to input the desired end point detection algorithm into the computer. Of these tasks, other than the design task can be executed on the system. The work flow is shown in FIG. First, in step S1, a program is designed. This is to create a desired program list based on the above-mentioned programming language specifications. Next, the computer is started and the program is typed in by an editor in step S2. In step S3, the input program is translated. Specifically, it performs error checking and conversion into executable files. In step S4, it is determined whether or not an error is included as a result of the translation. If there is no error, the process proceeds to step S5 to check the operation by the actual machine. As a result, it is determined in step S6 whether or not the operation as designed can be performed. In addition, when the operation as designed is not obtained,
Return to program design in step S1 or program type in in step S2.

For reference, a list of translation errors is shown in Table 7.

[Table 7]

Finally, the end point detection algorithm will be described with reference to a plurality of concrete examples. FIG. 6 is a waveform showing changes over time in the nitrogen atom spectrum observed when the nitride film on the silicon wafer is etched by using the plasma etching apparatus. An algorithm programmed to perform end point detection based on this waveform will be described. Figure 6
As shown in the graph, when the etching is started, the spectrum intensity (count value) once rises, and then decreases with the progress of the etching. When the nitride film is almost removed, an interference state is entered and small undulations are repeated, and then a steady state is reached. In this specific example, the end point End is set at a time point shortly after entering the interference state in order to prevent etching residue. That is, it is a portion slightly increased from the weakest point (MIN value). Here, as the end point detection algorithm, the change in amplitude from the strongest point (MAX value) to the weakest point (MIN value) is 100%, and the point returned by 10% is the end point End. However, since the end point detection at an early stage is inconvenient, the end point detection range is set to +15000 counts or less. In addition, a predetermined dead time is provided in order to exclude the time point at the initial stage of etching from the detection area.

Table 8 shows a program list of the algorithm created on the basis of the end point detection conditions described above.

[Table 8]

As is clear from the system setting statement (#set) in Table 8, in this specific example, Scan interval
(Period of program execution) is set to 1.0 second. Also, the dead time is set to 20 seconds. The Over time (preliminary time for compensating for the insufficient etching portion) is set to 10 seconds. Furthermore, S
The transfer time (accumulation time of light energy in the polychrometer) is set to 1 second. in addition,
Input gain (input gain of light energy in the polychrometer) is set to 4. Finally In
put Wave1 is set to 336.0 nm in order to target the nitrogen atom spectrum. In this example, the end point is detected with a single spectrum. However, the present invention is not limited to this, and it is also possible to simultaneously process two or more different spectra to detect the end point. Also, as understood from the input / output setting statement (#def) in Table 8, Input
The register r0 is allocated to 1, and the bit b10 is allocated to the End signal. Following the above setting statements, the content statements of this program are described from start () to end ().

The end point detection algorithm of Table 8 will be described in detail with reference to the flowchart of FIG. When the present algorithm is started, all bits and registers are first initialized to 0. Next, in step S71, it is determined whether or not the bit b0 is 0. Initially this bit b
Since 0 is 0, the flow advances to the next step S72 to set a predetermined initial value 32767 in the register r11. Then, in step S73, the bit b0 is rewritten to 1. next,
In step S74, the sizes of the registers r0 and r10 are compared. The current count value of the input waveform data is sequentially written in the register r0, while the maximum value MAX of the waveform data up to the previous time is stored in the register r10. If it is determined in step S74 that r0 ≧ r10, the content of r0 is written to r10 in the next step S75, and the value of MAX is updated. Then, step S76.
In, the magnitude comparison of the registers r0 and r11 is performed. The register r11 stores the minimum value MIN of the waveform data input up to the previous time. When it is determined in step S76 that r0 ≦ r11, the next step S7
In step 7, the value of r0 is transferred to r11 and the value of MIN is updated. In this way, step S75 and step S77
Therefore, the MAX value and the MIN value are constantly monitored.

In step S78, it is determined whether the current waveform data r0 is 15000 counts or less. If the result of the judgment is negative, the subsequent functions are not executed. If the determination result of step S78 is affirmative, it means that the count value of the input waveform data has entered the end point detection area. Therefore, in the next step S79, the difference (Full) between the maximum value MAX and the minimum value MIN is calculated. Next, in step S80, a count value corresponding to 10% of Full is calculated and stored in the register r13. Then, in step S81, the difference between the current waveform data count value r0 and the MIN value is calculated and stored in the register r14. Next, in step S82, it is determined whether the count value of the register r14 is larger than the content of the register r13. If the result of the determination is negative, the end point has not yet been reached, so the processing in this scan ends. On the other hand, if the result of the determination in step S82 is affirmative, it means that an end point has been reached, and therefore bit b1 is determined in step S83.
The end signal is output by adding 1 to 0.

Next, a second specific example will be described with reference to the graph of FIG. In this example, an oxide film formed on a silicon wafer is etched using a plasma etching apparatus. A fluorine-based gas is used as the plasma reaction gas, and the fluorine atom spectrum is selected here for end point detection. As shown in the input waveform of FIG.
The count value of the fluorine atom spectrum increases with the start of etching and becomes flat when the etching is completed. However, since the change of the input waveform is extremely gentle in the end point region, it is difficult to detect the end point accurately as it is. Therefore, in this specific example, the input waveform is second-order differentiated, and the time when the value is near the 0 level and continued for a predetermined time is defined as the end point. Since the input waveform becomes unstable at the initial stage of etching, a desired dead time is provided so that the end point is not detected.

Table 9 below shows a programming list of algorithms created based on the above-mentioned end point detection conditions.

[Table 9]

As described in the system setting statement (#set) of the programming list in Table 9, Scan is used in this example.
The internalval is set to 1.0 second. That is, this program is executed at intervals of 1 second. Also, Dead
The time is set to 50 seconds. Over ti
me is set to 30 seconds. Storage ti
me is set to 1.5 seconds. Input kai
n is set to 3 and specifies the input amplification factor of light energy in the polychromator. InputWave
1 is set to 660.0 nm, and the fluorine atom spectrum is selected as described above. In this specific example, the end point detection is performed based on the selected one type of spectrum. Input / output setting statement (#d
According to ef), register r0 is assigned to Input1. Bit r3 is assigned to the End signal. Further, bit b7 is assigned because of the Alarm signal. The content statement of this programming list is described between start () and end ().

The end point detecting operation based on the algorithm shown in Table 9 will be described in detail with reference to the flowchart of FIG. First, after starting the computer, step S91
Register information r indicating the count value of the input waveform data
0 is first-order differentiated and stored in another register r1. Then, in step S92, the register information r1 is differentiated again and the result is stored in the register r2. Therefore, r2 represents the secondary differential value of the input waveform data. Next, in step S93, r
The value of 2 is multiplied by 10 to expand the scale, and the result is stored in the register r6. Next, in step S94, the register r1
Increment 2 by the number 1. This register r1
Reference numeral 2 functions as a counter for measuring the number of times the scanning operation is repeated, and is used for monitoring a predetermined duration. Next, in step S95, the absolute value of r6 representing the second derivative is stored in the register r9.

Succeedingly, in a step S96, it is determined whether or not the value of r9 is smaller than the set value 80. If the determination result is affirmative, it is possible that the second-order differential value of the input waveform has moved toward a convergent state. At this time, the process proceeds to step S97,
The value of r12 is moved to another register r10. on the other hand,
If the determination result in step S96 is negative, the register r10 is reset to 0 at any time in step S98. Succeedingly, in a step S99, it is determined whether or not the value of the register r10 is 100 or more. If the result of this determination is affirmative, it means that the secondary differential value has become less than or equal to the set level and a predetermined time has elapsed. Therefore, in step S100, the value b is set to 1 and the End signal is output. On the other hand, if the determination result of step S99 is negative, the process proceeds to step S101, and it is determined whether the timer has passed 200 seconds. If the determination result is affirmative, it is considered that an abnormal state has occurred, and the bit b7 is set to the numerical value 1 in step S102 and Al is set.
Output the arm signal. On the other hand, if the determination result in step S101 is negative, the process in this scan ends.

Finally, the third specific example will be described in detail with reference to the graph of FIG. In this example, a plasma ion etching apparatus is used to introduce a fluorine-based reaction gas to etch the nitride film formed on the silicon wafer. In this example, two different spectra are selected to detect the end point. The input waveform A represents the intensity change of the nitrogen atom spectrum over time, and the count value decreases with the progress of the etching process to reach a steady state. On the other hand, the input waveform B represents the intensity change of the fluorine atom spectrum over time, and the count value increases with the progress of the etching process and reaches a steady level. Here, when both input waveforms reach a steady state and the input waveform B is larger than the input waveform A, it is determined as the end point. Similar to the above-mentioned example, since both waveforms show a gentle change in the end point region of the etching process, the end point cannot be precisely detected as it is. Therefore, both waveforms are subjected to a second derivative, and it is determined that the etching is completed when a predetermined duration elapses from the time when the value becomes less than or equal to a preset set value. Similar to the above-mentioned example, a predetermined Deadtime is provided because the end point is not detected in an unstable state at the initial stage of etching.

Table 10 below shows a programming list created based on the above-mentioned end point detection conditions.

[Table 10]

As shown in the system setting statement (#set) of the programming list in Table 10, in this example, Scan
The interval is set to 1.0 second. Dead
The time is set to 30 seconds. Over tim
e is also set to 30 seconds. Storage t
The time is set to 1.5 seconds and the accumulation time of the light energy in the polychromator is specified. Input g
The ain is set to 3, and the input amplification factor of the light energy in the polychromator is designated. InputWave
336.0 nm is set as 1, and the input waveform A shown in FIG. 10 is selected. Similarly, Input Wave2
Is set to 660.0 nm and the input waveform B shown in FIG. 10 is selected. In this way, in this specific example, two different input waveforms are simultaneously processed to detect the end point. Next, as is clear from the input / output setting statement (#def) described in the programming list in Table 10, the register r0 is assigned to Input1, the register r1 is also assigned to Input2, and the bit is assigned to the End signal. b0 is allocated. The content sentence of this end point detection algorithm is described between start () and end ().

Referring to the flowchart of FIG. 11, Table 1
The end point detection algorithm shown in 0 will be described in detail. First, after starting the computer, in step S111, the first derivative of r0 representing the count value of the input waveform A is performed, and the result is set as the register information r10. Then, in step S112, the register information r10 is differentiated and the result is set as the register information r11. Therefore, the register information r11 represents the secondary differential data of the input waveform A. In step S113, the register information r11 is multiplied by 10 to expand the scale, and the result is set as the register information r12. Next, in step S114, the absolute value of the register information r12 is taken and used as the register information r13. Then step S11
5, the register information r13 is set to a preset value 100
Determine if: If this determination result is affirmative, there is a possibility that the input waveform A is heading toward a convergent state. In this case, it is determined in the next step S116 whether or not the preset timer has elapsed for 3 seconds. This 3 second time period represents a predetermined duration. If the determination result of step S116 is affirmative, it is considered that the input waveform A has reached a steady state, and the bit information b11 is set to the numerical value 1 in the next step S117.
Is set to. On the other hand, if the determination results in both steps S115 and S116 described above are negative, the end point has not yet been reached, and the process proceeds to the next step.

Here, the processing shifts to the processing of the other input waveform B, and the primary differentiation of the register information r1 is performed in step S118, and the result is set as the register information r20. The register information r1 represents the count value of the input waveform B. Subsequently, in step S119, the register information r20 is differentiated, and the result is set as the register information r21.
As a result, r21 represents the secondary differential data of the count value of the input waveform B. In step S120, the secondary differential register information r21 is multiplied by 10 to expand the scale, and the result is set as register information r22. Step S121
Takes the absolute value of the register information r22 and sets the result as the register information r23. Next, in step S122, it is determined whether the register information r23 is less than or equal to a predetermined set value 60. If the result of this judgment is positive,
In step S123, it is determined whether a predetermined timer has passed for 5 seconds. If the determination result in step S123 is affirmative, it is estimated that the input waveform B has reached a steady state, and the bit information b21 is set to the numerical value 1 in the next step S124.

Succeedingly, in a step S125, it is determined whether or not the register information r0 is smaller than the register information r1. If the determination result is affirmative, it means that the count value of the input waveform A is lower than the count value of the input waveform B in the steady state near the end point region as shown in the graph of FIG. Therefore, in the next step S126, the logical product process of the bit information b11 and b21 is performed. The processing result is 1
Output bit b in the final step S127
Add 1 to 0. As a result, the End signal is output.
On the other hand, if the determination result in step S125 is negative and the processing result in step S126 indicates the numerical value 0, the processing in this scan ends.

[0056]

As described above, since the endpoint monitor according to the present invention uses the optical fiber and the polychromator in the optical sensor section, it simultaneously measures the light of a plurality of wavelengths and determines the mutual relationship. The effect is that precise end point detection can be performed. Since a polychromator is used, it is possible to measure a wide range of wavelengths without using equipment such as a spectrum analyzer as in the past, and it is possible to freely select the specific wavelength required for end point detection. is there. For example, it is possible to reduce noise and extract only signals necessary for calculation by selecting two spectra of reference light and signal light and performing background processing with reference light. In addition, simultaneous measurement of multiple wavelengths enables measurement and calculation of light that changes with time difference, and for example, the end point of multilayer film etching can be precisely and effectively detected.
Alternatively, light that changes rapidly near the end point and light that changes gradually from the start of etching can be simultaneously processed. In this way, it is possible to make a determination according to various etching processing contents from changes in a plurality of lights. In the present invention, since the polychromator is connected to the processing device such as the plasma etching device through the optical fiber, there is an effect that it is less likely to be interfered with by the high frequency power supply. In addition, a high-level signal can be obtained and a filter effect can be obtained due to the accumulation effect of the photoelectric sensor incorporated in the polychromator.

The endpoint monitor according to the present invention is controlled using a programmable computer. The end point detection algorithm can be freely constructed by using a simple and formalized programming language, and has an effect of being extremely versatile. This makes it possible to deal with the complicated end point detection in various etching processes. A recipe number is attached to each of the end point detection algorithms constructed as needed, and the algorithms can be read out at any time. Further, by recording and saving the input waveform once sampled and simulating the data again, it is possible to verify by a determination algorithm of a different recipe number, and there is an effect that a practical simulation function is achieved. Therefore, the number of wafers used in the experiment and the operation of the processing apparatus can be minimized until the final determination algorithm is determined.

[Brief description of drawings]

FIG. 1 is a block diagram showing a basic configuration of an endpoint monitor according to the present invention.

FIG. 2 is a block diagram showing a specific example of a polychromator incorporated in an endpoint monitor.

FIG. 3 is a schematic diagram showing an input / output relationship of a computer incorporated in an endpoint monitor.

FIG. 4 is a timing chart showing a transition of a processing system state of the endpoint monitor.

FIG. 5 is a flowchart showing the work of creating an end point detection program.

FIG. 6 is a graph showing an example of an input waveform.

7 is a flowchart showing an endpoint detection algorithm for processing the input waveform shown in FIG.

FIG. 8 is a graph showing another example of the input waveform.

9 is a flowchart showing an end point detection algorithm programmed to perform the processing of the input waveform shown in FIG.

FIG. 10 is a graph showing still another example of the input waveform.

11 is a flowchart showing an end point detection algorithm programmed to perform the processing of the input waveform shown in FIG.

[Explanation of reference numerals] 1 processing device 2 polychromator 3 computer 4 optical fiber 5 cable 6 display 7 keyboard 8 input terminal 9 output terminal

Claims (3)

[Claims] 1. An endpoint monitor comprising a computer and a polychromator connected to a processing device for processing a semiconductor wafer, wherein the polychrometer receives broadband light generated in the processing device during processing. Then, it is possible to separate into a plurality of spectra, the intensity change of these spectra is measured over time, and the corresponding waveform signal is output, and the computer uses a formalized and simplified programming language to obtain the desired waveform. An end-point monitor, which can be programmed with an end-point detection algorithm, and which performs at least one specific waveform signal according to the end-point detection algorithm to detect the end point of the machining. 2. The endpoint monitor according to claim 1, wherein the computer simultaneously performs arithmetic processing on a plurality of waveform signals corresponding to different spectra according to the endpoint detection algorithm to detect the endpoint. 3. The endpoint monitor of claim 1, wherein the computer programs a desired endpoint detection algorithm with a combination of predetermined functions and operators.