Process diagnostic device and plasma exchange timing determination method

Abstract

A process diagnostic device for diagnosing the state of a semiconductor process device comprises: a sample holder supporting a sample processed in the semiconductor process device; an illumination optical system that irradiates illumination light onto a sample placed on the sample holder; a plurality of detection optical systems that collect light from the sample, convert it into an electrical signal, and output a detection signal; and a signal processing device that processes the detection signals of the plurality of detection optical systems. The signal processing device scans the sample to extract a haze signal of the sample, compares the haze signal of the sample with a reference haze signal, and outputs an alarm when the difference exceeds a set value.

Description

Technology Field

[0001] The present invention relates to a process diagnostic device for diagnosing semiconductor process devices such as plasma etching devices, and a method for determining plasma exchange timing. Background Technology

[0002] Since the degradation of components or a decrease in cleanliness of process equipment used to process semiconductor wafers leads to a decrease in semiconductor yield, it is important to manage the condition of the process equipment. For example, in the case of plasma etching equipment, consumables that degrade over time, such as plasma light sources, are regularly replaced or cleaned. However, there are individual differences in the lifespan of plasma light sources, so the occurrence of abnormalities in plasma light sources cannot necessarily be prevented in advance through regular replacement or cleaning.

[0003] However, Patent Document 1 discloses a technology for an optical inspection device used for inspecting defects in semiconductor samples such as wafers, which not only detects main defect information from the scattered light intensity obtained from the sample but also specifies information regarding wafer characteristics other than defects, such as the surface roughness or surface film thickness of the sample. In an optical defect inspection device, the entire surface of the sample is typically inspected. In Patent Document 1, a correlation is predetermined between characteristic values, such as the surface roughness or surface film thickness of the sample measured at several points on the sample using a measuring device other than an atomic force microscope (AFM) or a spectroscopic ellipsometer, and the scattered light intensity of the entire surface of the sample obtained when the defect is inspected by the optical defect inspection device. Under this correlation, the characteristic values ​​of the entire surface of the sample are calculated from the scattered light intensity measured during the defect inspection. With the characteristic values ​​of the sample calculated in this way, the surface roughness or surface film thickness can be referenced across the entire surface. Prior art literature

[0004] Japanese Publication No. 2011-013058 The problem to be solved

[0005] However, in the technology of the above-mentioned Patent Document 1, in determining the correlation between scattered light intensity and the characteristic value of a sample, it is necessary to measure the characteristic value of the sample using other measuring devices such as an AFM. Due to the circumstances of using measuring devices such as an AFM, the actual measurement of the characteristic value is limited to a part of the entire surface of the sample. The correlation between scattered light intensity and the characteristic value determined in the technology of Patent Document 1 is based only on a part of the actual measurement value. Therefore, for most of the remaining sample surface, it is obtained through supplementation. Accordingly, there is room for improvement in the technology of the said document regarding the precision of calculating the characteristic value of the sample and, furthermore, evaluating the state of the process device.

[0006] The objective of the present invention is to provide a process diagnostic device and a method for determining plasma exchange timing that can precisely diagnose the state of a semiconductor process device from data obtained from defect inspection of a sample after a process, suppress the occurrence of abnormalities in the semiconductor process device, and improve the yield of the semiconductor. means of solving the problem

[0007] To achieve the above objective, the present invention provides a process diagnostic device for diagnosing the state of a semiconductor process device, comprising: a sample stand supporting a sample processed in the semiconductor process device; an illumination optical system for irradiating illumination light onto a sample placed on the sample stand; a plurality of detection optical systems for collecting light from the sample and converting it into an electrical signal to output a detection signal; and a signal processing device for processing the detection signals of the plurality of detection optical systems. The signal processing device scans the sample to extract a haze signal of the sample, compares the haze signal of the sample with a reference haze signal, and outputs an alarm when the difference exceeds a set value. Effects of the invention

[0008] According to the present invention, the condition of a semiconductor process device can be precisely diagnosed from data obtained from defect inspection of a sample after a process, and the occurrence of abnormalities in the semiconductor process device can be suppressed to improve the yield of the semiconductor. Brief explanation of the drawing

[0009] FIG. 1 is a schematic diagram of one configuration example of a process diagnostic device according to a first embodiment of the present invention. FIG. 2 is a schematic diagram showing the scan trajectory of a sample. FIG. 3 is a schematic diagram showing the scan trajectory of a sample. FIG. 4 is a schematic diagram showing the attenuator being ejected. FIG. 5 is a schematic diagram showing the positional relationship between the optical axis of the illumination light guided to the surface of the sample from the inclined direction and the shape of the illumination intensity distribution. FIG. 6 is a schematic diagram showing the positional relationship between the optical axis of the illumination light guided to the surface of the sample from the inclined direction and the shape of the illumination intensity distribution. FIG. 7 is a diagram showing the area where the detection optical system collects scattered light when viewed from above. FIG. 8 is a schematic diagram showing the zenith angles of the low-angle and high-angle detection optical systems. FIG. 9 is a plan view showing the azimuth angle of the low-angle detection optical system. FIG. 10 is a plan view showing the azimuth angle of the high-angle detection optical system. FIG. 11 is a schematic diagram showing an example of the configuration diagram of the detection optical system being ejected. FIG. 12 is an example of a functional block diagram of a key part of a signal processing device equipped in a process diagnostic device according to a first embodiment of the present invention. FIG. 13 is a flowchart outlining an example of a series of diagnoses of a process device in a semiconductor manufacturing process. FIG. 14 is a schematic diagram showing a change in the state of a process device over time. FIG. 15 is a schematic diagram showing the effect appearing on a sample accompanying a change in the state of a process device over time, focusing on section XV of FIG. 14. FIG. 16 is a flowchart showing the sequence of diagnostic processing of a process device by a process diagnostic device according to a first embodiment of the present invention. FIG. 17 is a conceptual diagram of machine learning according to a second embodiment of the present invention. FIG. 18 is a schematic diagram showing a key part of a process diagnostic device according to a third embodiment of the present invention. FIG. 19 is a schematic diagram showing a key part of a process diagnostic device according to a fourth embodiment of the present invention. Specific details for implementing the invention

[0010] Embodiments of the present invention are described below using the drawings.

[0011] The process diagnostic device described in the following embodiments scans a sample, in this case a wafer (bare wafer, film-attached wafer, patterned wafer, etc.) during the semiconductor manufacturing process and performs a diagnosis of the process device based on the state of the sample processed by the process device. As an example of a process diagnostic device, there is an optical defect inspection device that inspects defects in a sample. The optical defect inspection device typically outputs the number, coordinates, and types of defects attached to or formed on the wafer based on a signal derived from reflected or scattered light obtained by scanning the sample. It is used to determine the state of the sample processed based on this information. The optical defect inspection device has a very fast scanning speed compared to diagnostic devices that use electron beams or X-rays as light sources. Therefore, while diagnostic devices using electron beams or X-rays as energy sources can only measure a fraction of the sample due to time constraints, the optical defect inspection device of this embodiment scans the entire surface of the sample and can diagnose the state of the entire surface of the sample. The light-based signal obtained by scanning the sample in this optical defect inspection device includes not only a defect signal used for defect detection but also a signal called a haze signal. In this embodiment, a method for diagnosing the condition of a process device primarily using this haze signal is described. If a signal with a fluctuation frequency (time variation of signal intensity) higher than a predetermined value is defined as high frequency and a signal with a lower value as low frequency, the defect signal corresponds to the high-frequency component among the signals based on light obtained from the sample. Conversely, the haze signal corresponds to the low-frequency component. For example, signals caused by relatively large irregularities attached to the sample, such as foreign matter, are easily detected as high-frequency components, while signals caused by characteristics of the wafer itself, such as the thickness of the film on the sample or minute irregularities (roughness) on the sample surface, are easily detected as low-frequency components.The process diagnostic device of the present embodiment utilizes a haze signal that is not generally used (removed) in defect inspection of a sample, and diagnoses the condition of the process device that processed the sample based on the haze signal obtained from the sample. Specifically, it monitors the change in the haze signal over time, and detects that there is some change (abnormality) in the sample processed by the process when, for example, a predetermined difference (relative change in the haze signal) occurs between the haze signal and a reference haze signal. Based on the change in wafer characteristics based on this haze signal, an abnormality or a warning sign thereof is estimated in the process device that processed the sample, such as a plasma process device for etching, and is notified to an operator, etc.

[0012] In the present specification, an abnormality of the process device refers to a state in which the processing quality of the sample (in the case of a plasma etching / film deposition device, the surface film thickness or surface roughness of the sample surface) has already fallen below an acceptable level, and the process device needs to be maintained promptly. An indication of an abnormality of the process device refers to a state in which, at the current stage, the processing quality of the sample is still above an acceptable level but falls below a set level, and it is predicted that it will transition to an abnormal state within a specified period.

[0013] (1st embodiment)

[0014] - Process Diagnostic Device -

[0015] FIG. 1 is a schematic diagram of one configuration example of a process diagnostic device (100) according to the first embodiment of the present invention. An XYZ orthogonal coordinate system extending the Z-axis in the vertical direction is defined as shown in FIG. 1. The process diagnostic device (100) takes a sample (W) as an inspection target and detects defects such as the attachment of foreign matter or abnormal film formation on the surface of the sample (W). The process diagnostic device (100) is a rotary scan type device that scans by rotating the sample (W) in the circumferential direction (θ direction) while moving it in the radial direction (r direction).

[0016] The process diagnostic device (100) includes a stage (ST), an illumination optical system (A), a plurality of detection optical systems (Bn (n=1, 2…)), sensors (Cn, Cn' (n=1, 2…)), a signal processing device (D), a memory device (DB), a control device (E1), an input device (E2), and a monitor (E3).

[0017] - Stage -

[0018] The stage (ST) is a device comprising a sample stand (ST1) and a scanning device (ST2). The sample stand (ST1) is a stand that supports a sample (W) processed in a process device such as a plasma etching device. The scanning device (ST2) is a device that drives the sample stand (ST1) to change the relative position between the sample (W) and the illumination optical system (A), and although detailed illustration is omitted, it is composed of a translation stage, a rotation stage, and a Z stage. The rotation stage is mounted on the translation stage via the Z stage, and the sample stand (ST1) is supported on the rotation stage. The translation stage moves in a horizontal translational direction together with the rotation stage. The rotation stage rotates (rotates) around a rotation axis that extends vertically. The Z stage functions to adjust the height of the surface of the sample (W).

[0019] FIG. 2 is a schematic diagram showing the scan trajectory of a sample (W) by a scanning device (ST2). As will be described later, the beam spot (BS), which is the incident area of ​​the illumination light emitted from the illumination optical system (A) on the surface of the sample (W), is a minute point having an illumination intensity distribution that is long in one direction as shown in the figure. The direction of the major axis of the beam spot (BS) is denoted as s2, and the direction intersecting the major axis (e.g., the minor axis direction perpendicular to the major axis) is denoted as s1. As the sample (W) rotates along with the rotation of the rotation stage, the beam spot (BS) is scanned in the s1 direction relative to the surface of the sample (W), and as the sample (W) moves in the horizontal direction along with the translation of the translation stage, the beam spot (BS) is scanned in the s2 direction relative to the surface of the sample (W). During one rotation of the sample (W), the beam spot (BS) moves in the s2 direction by a distance less than or equal to the length of the beam spot (BS) in the s2 direction. As the sample (W) rotates and translates by the operation of the scanning device (ST2), as shown in FIG. 2, the beam spot (BS) moves along a spiral trajectory from the center of the sample (W) to the outer edge or nearby, and the entire surface of the sample (W) is scanned.

[0020] Additionally, the stage (ST) may be configured to have another translational stage, which has a movement axis extended in a direction intersecting the movement axis of the translational stage within the horizontal plane, in place of (or in addition to) the rotational stage. In this case, as shown in FIG. 3, the beam spot (BS) scans the surface of the sample (W) by overlapping a straight trajectory rather than a spiral trajectory. In the example of the same figure, the first translational stage is translated at a constant speed in the s1 direction, and the second translational stage is driven in the s2 direction by a predetermined distance (for example, a distance less than or equal to the length of the beam spot (BS) in the s2 direction), and then the first translational stage is translated again by bending it in the opposite direction to the s1 direction. By doing so, the beam spot (BS) repeatedly scans the entire surface of the sample (W) by performing a straight scan in the s1 direction and moves in the s2 direction. Compared to this XY scanning method, the rotary scanning method of Fig. 2 does not involve reciprocating motion that repeatedly accelerates and decelerates, so the inspection time of the sample (W) can be shortened.

[0021] - Illumination Optical System -

[0022] The illumination optical system (A) shown in FIG. 1 is configured to include a group of optical elements to irradiate a desired illumination light onto a sample (W) placed on a sample stage (ST1). As shown in FIG. 1, this illumination optical system (A) is equipped with a laser light source (A1), an attenuator (A2), an emission light adjustment unit (A3), a beam expander (A4), a polarization control unit (A5), a light-gathering optical unit (A6), a reflective mirror (A7-A9), etc.

[0023] · Laser light source

[0024] A laser light source (A1) is a unit that emits a laser beam as illumination light. When a process diagnostic device (100) detects minute defects near the surface of a sample (W), a laser light source (A1) is used to emit a high-power laser beam of 2W or more using ultraviolet light of a short wavelength (wavelength 355nm or less) or vacuum ultraviolet light that is difficult to penetrate into the interior of the sample (W). When a process diagnostic device (100) detects defects inside the sample (W), a laser light source (A1) is used to emit a visible or infrared laser beam that has a long wavelength and is easy to penetrate into the interior of the sample (W).

[0025] · Attenuator

[0026] FIG. 4 is a schematic diagram showing an attenuator (A2). The attenuator (A2) is a unit that attenuates the light intensity of illumination light from a laser light source (A1), and in this embodiment, a configuration combining a first polarizer (A2a), a half-wave plate (A2b), and a second polarizer (A2c) is exemplified. The half-wave plate (A2b) is configured to be rotatable around the optical axis of the illumination light. The illumination light incident on the attenuator (A2) is converted into linear polarization at the first polarizer (A2a), and then passes through the second polarizer (A2c) with its polarization direction adjusted to the azimuth of the lag axis of the half-wave plate (A2b). By adjusting the azimuth of the half-wave plate (A2b), the light intensity of the illumination light is attenuated at an arbitrary rate. If the linear polarization degree of the illumination light incident on the attenuator (A2) is sufficiently high, the first polarizer (A2a) may be omitted. The attenuator (A2) is used such that the relationship between the incident illumination light and the light attenuation rate is corrected in advance. Furthermore, the attenuator (A2) is not limited to the configuration exemplified in FIG. 4, and may be configured using an ND filter having a gradient density distribution, or may be configured such that the attenuation effect can be adjusted by combining multiple ND filters with different densities.

[0027] · Light Emission Adjustment Unit

[0028] The light emission adjustment unit (A3) shown in FIG. 1 is a unit that adjusts the angle of the optical axis of the illumination light attenuated by the attenuator (A2), and in this embodiment, it is configured to include a plurality of reflective mirrors (A3a, A3b). Although the illumination light is sequentially reflected by the reflective mirrors (A3a, A3b), in this embodiment, the incident and exit surfaces of the illumination light with respect to the reflective mirror (A3a) are configured to be orthogonal to the incident and exit surfaces of the illumination light with respect to the reflective mirror (A3b). The incident and exit surfaces are surfaces that include the optical axis of the light incident on the reflective mirror and the optical axis of the light emitted from the reflective mirror. When the illumination light is configured to be incident on the reflective mirror (A3a) in the +X direction, although it differs from the schematic FIG. 1, for example, the illumination light changes its direction of travel in the +Y direction from the reflective mirror (A3a) and then in the +Z direction from the reflective mirror (A3b). This is an example where the incident and exit planes of the illumination light for the reflecting mirror (A3a) are the XY plane, and the incident and exit planes for the reflecting mirror (A3b) are the YZ plane. Additionally, the reflecting mirrors (A3a, A3b) are equipped with a mechanism (not shown) for translating the reflecting mirrors (A3a, A3b) and a mechanism (not shown) for tilting them, respectively. The reflecting mirrors (A3a, A3b) are, for example, moved parallel to the incident or exit direction of the illumination light relative to themselves, and are also tilted around the normal to the incident and exit planes. By doing so, the amount and angle of the offset within the XZ plane and the amount and angle of the offset within the YZ plane can be independently adjusted with respect to the optical axis of the illumination light emitted in the +Z direction from the exit light adjustment unit (A3), for example. In this embodiment, a configuration using two reflecting mirrors (A3a, A3b) is exemplified, but a configuration using three or more reflecting mirrors is also acceptable.

[0029] · Beam Expander

[0030] The beam expander (A4) is a unit that expands the diameter of the luminous flux of the incident illumination light and has a plurality of lenses (A4a, A4b). A Galileo type using a concave lens as lens (A4a) and a convex lens as lens (A4b) can be cited as an example of the beam expander (A4). The beam expander (A4) is equipped with a spacing adjustment mechanism (zoom mechanism) for the lenses (A4a, A4b), and the magnification rate of the luminous flux diameter changes by adjusting the spacing of the lenses (A4a, A4b). If the illumination light incident on the beam expander (A4) is not a parallel luminous flux, collimation (quasi-parallelization of the luminous flux) is also possible along with the luminous flux diameter by adjusting the spacing of the lenses (A4a, A4b). However, regarding the collimation of the light speed, it may be configured to be performed using a collimation lens installed upstream of the beam expander (A4) separately from the beam expander (A4).

[0031] In addition, the beam expander (A4) is installed on a translation stage with two or more axes (two degrees of freedom) and is configured to adjust its position so that it aligns with the center of the incident light. In addition, the beam expander (A4) is also equipped with a tilt angle adjustment function with two or more axes (two degrees of freedom) so that it aligns with the optical axis of the incident light.

[0032] In addition, although not specifically shown, the state of the illumination light incident on the beam expander (A4) during the light path of the illumination optical system (A) is measured by a beam monitor.

[0033] · Polarization control unit

[0034] The polarization control unit (A5) is an optical system that controls the polarization state of the illumination light and is configured to include a half-wave plate (A5a) and a quarter-wave plate (A5b). For example, when a reflective mirror (A7) described later is placed in the optical path and the sample (W) is illuminated obliquely, the illumination light is made P-polarized by the polarization control unit (A5), thereby increasing the amount of scattered light from the surface of the sample (W) compared to polarization other than P-polarization. If there is an oxide film on the surface of the sample (W), the amount of scattered light from the surface of the sample can be increased compared to P-polarization by using S-polarization depending on the material and thickness of the film. By selecting polarization according to the sample (W), conditions where haze light is likely to occur and conditions where it is unlikely to occur can be switched, thereby improving the sensitivity of defect inspection or improving the sensitivity of haze light to the sample characteristics. For example, when evaluating the condition of a sample (W) by the output of haze light, it is advantageous to make the illumination light S-polarized. It is also possible to make the illumination light circularly polarized or 45-degree polarized, which is intermediate between P-polarized and S-polarized light, by the polarization control unit (A5).

[0035] ·Reflective mirror

[0036] As shown in FIG. 1, the reflective mirror (A7) moves parallel in the direction of the arrow by means of a driving mechanism (not shown) and enters and exits the optical path of the illumination light directed toward the sample (W). By doing so, the incident path of the illumination light toward the sample (W) is changed. By inserting the reflective mirror (A7) into the optical path, the illumination light emitted from the polarization control unit (A5) as described above is reflected by the reflective mirror (A7) and incident obliquely on the sample (W) through the condensing optical unit (A6) and the reflective mirror (A8). In this way, illuminating the sample (W) from a direction inclined with respect to the normal of the surface of the sample (W) is described in this specification as "oblique incident illumination." Meanwhile, when the reflecting mirror (A7) is separated from the optical path, the illumination light emitted from the polarization control unit (A5) is incident perpendicularly on the sample (W) through the reflecting mirror (A9), the polarization beam splitter (B'3), the polarization control unit (B'2), the reflecting mirror (B'1), and the detection optical system (B3). In this specification, the illumination light being incident perpendicularly on the surface of the sample (W) is described as "vertical illumination."

[0037] FIGS. 5 and 6 are schematic diagrams showing the positional relationship between the optical axis of the illumination light induced from an inclined direction to the surface of a sample (W) by an illumination optical system (A) and the shape of the illumination intensity distribution. FIG. 5 schematically shows a cross-section of the sample (W) cut at the incident plane of the illumination light incident on the sample (W). FIG. 6 schematically shows a cross-section of the sample (W) cut at a plane that is orthogonal to the incident plane of the illumination light incident on the sample (W) and includes the normal to the surface of the sample (W). The incident plane is a plane that includes the optical axis (OA) of the illumination light incident on the sample (W) and the normal to the surface of the sample (W). Additionally, FIGS. 5 and 6 show a portion of the illumination optical system (A) and omit the illustration of, for example, the emission light adjustment unit (A3) and the reflection mirrors (A7, A8).

[0038] As described above, when a reflective mirror (A7) is inserted into the optical path, the illumination light emitted from the laser light source (A1) is concentrated by the concentrating optical unit (A6), reflected by the reflective mirror (A8), and incident obliquely on the sample (W). In this way, the illumination optical system (A) is configured to incident the illumination light obliquely on the surface of the sample (W). For this obliquely incident illumination, the light intensity is adjusted by the attenuator (A2), the beam diameter by the beam expander (A4), and the polarization by the polarization control unit (A5), respectively, so that the illumination intensity distribution within the incident plane is uniformized. As shown in the illumination intensity distribution (illumination profile) (LD1) in FIG. 5, the beam spot formed on the sample (W) has a light intensity distribution in the Gaussian distribution direction in the s2 direction.

[0039] In a plane orthogonal to the incident plane and the sample surface, the beam spot has a light intensity distribution with weaker intensity around the center of the optical axis (OA), as shown in the illumination intensity distribution (illumination profile) (LD2) in FIG. 6. This light intensity distribution is, for example, a Gaussian distribution reflecting the intensity distribution of light incident on the condensing optical unit (A6), or an intensity distribution similar to a first-order Bessel function of the first kind or a sinc function reflecting the aperture shape of the condensing optical unit (A6).

[0040] Additionally, the angle of incidence of the oblique incident light on the sample (W) (the angle of inclination of the incident optical axis with respect to the normal of the sample surface) is adjusted to an angle suitable for detecting minute defects by the position and angle of the reflecting mirrors (A7, A8). The angle of the reflecting mirror (A8) is adjusted by the adjustment mechanism (A8a). For example, the larger the angle of incidence of the illumination light on the sample (W) (the smaller the angle of elevation of the illumination formed by the sample surface and the incident optical axis), the weaker the haze light that becomes noise from the scattered light from minute defects on the sample surface.

[0041] - Detection Optical System -

[0042] The detection optical system (Bn (n=1, 2…)) is a unit that collects scattered light from the surface of a sample and is configured to include a plurality of optical elements including a collecting lens (objective lens). n of the detection optical system (Bn) is the number of detection optical systems, and in the process diagnostic device (100) of this embodiment, 13 detection optical systems are provided (n=13). However, the number of detection optical systems (Bn) is not limited to 13 and can be appropriately increased or decreased. In addition, the layout of the aperture (objective lens) of the detection optical system (Bn) can also be appropriately changed.

[0043] FIG. 7 is a diagram showing the area where the detection optical system (Bn) collects scattered light when viewed from above, and corresponds to the arrangement of each objective lens of the detection optical system (Bn). FIG. 8 is a diagram schematically showing the zenith angles of the low-angle and high-angle optical systems among the detection optical system (Bn), FIG. 9 is a plan view showing the azimuth angle of the low-angle detection optical system, and FIG. 10 is a plan view showing the azimuth angle of the high-angle detection optical system.

[0044] In the following description, with respect to the direction of incidence of oblique incident illumination on the sample (W), the direction of propagation of the incident light (right direction in FIG. 7) with respect to the beam spot (BS) on the surface of the sample (W) when viewed from above is defined as forward, and the opposite direction (left direction) is defined as backward. With respect to the beam spot (BS), the lower side in the figure is the right side, and the upper side is the left side. Additionally, with respect to the normal (N) (Fig. 8) of the sample (W) passing through the beam spot (BS), the angle φ2 (Fig. 8) formed by the incident optical axis (centerline of the aperture) of each detection optical system (Bn) is described as the zenith angle. Furthermore, when viewed from a plane, the angle φ1 (Fig. 9, Fig. 10) formed by the incident optical axis (centerline of the aperture) of each detection optical system (Bn) with respect to the incident plane of the oblique incident illumination is described as the azimuth angle.

[0045] As shown in FIGS. 7-10, the detection optical system (Bn) is arranged such that its orientation relative to the beam spot (BS) is different from one another. In this embodiment, each objective lens (aperture (α1-α6, β1-β6, γ)) of the detection optical system (Bn) is arranged along the upper half hemispherical surface of a sphere (celestial sphere) centered on the beam spot (BS) of the sample (W). Light incident on the aperture (α1-α6, β1-β6, γ) is focused at the corresponding detection optical system (Bn).

[0046] The aperture (γ) overlaps the ceiling (intersects the normal (N)) and is located directly above the beam spot (BS) formed on the surface of the sample (W).

[0047] The apertures (α1-α6) are opened to divide an annular area surrounding the beam spot (BS) at a low angle 360 ​​degrees. When viewed in a plane, the apertures (α1-α6) are arranged in the order of apertures (α1, α2, α3, α4, α5, α6) in a leftward rotation from the direction of incidence of the obliquely incident light. Additionally, the apertures (α1-α6) are laid out to avoid the incident light of the obliquely incident light and specular reflection paths. The apertures (α1-α3) are positioned to the right of the beam spot (BS), with aperture (α1) located to the right rear of the beam spot (BS), aperture (α2) to the right side, and aperture (α3) to the right front. Apertures (α4-α6) are positioned to the left of the beam spot (BS), with aperture (α4) located to the left front of the beam spot (BS), aperture (α5) to the left side, and aperture (α6) to the left rear. The arrangement of apertures (α4, α5, α6) is symmetrical to apertures (α3, α2, α1) with respect to the incident plane of oblique incident illumination.

[0048] Apertures (β1-β6) are opened to equally divide an annular area surrounding the beam spot (BS) at an angle (between apertures (α1-α6) and aperture (γ)). When viewed from a plane, apertures (β1-β6) are arranged in the order of apertures (β1, β2, β3, β4, β5, β6) in a leftward rotation from the direction of incidence of obliquely incident light. Among apertures (β1-β6), apertures (β1, β4) are laid out at positions intersecting the incident plane, with aperture (β1) located behind the beam spot (BS) and aperture (β4) located in front. Apertures (β2, β3) are arranged to the right of the beam spot (BS), with aperture (β2) located to the right rear of the beam spot (BS) and aperture (β3) located to the right front. The apertures (β5, β6) are positioned to the left of the beam spot (BS), with the aperture (β5) located to the left front of the beam spot (BS) and the aperture (β6) located to the left rear.

[0049] Scattered light scattered in various directions from the beam spot (BS) is incident on the aperture (α1-α6, β1-β6, γ), is collected by the detection optical system (Bn) for each, and guided to the corresponding sensors (Cn, Cn').

[0050] FIG. 11 is a schematic diagram showing an example of the configuration of a detection optical system. In the process diagnostic device of the present embodiment, each detection optical system (Bn) (or a part of the detection optical system) is configured as shown in FIG. 11, and the polarization direction of the scattered light transmitted can be controlled by a polarizer (Bb). Specifically, the detection optical system (Bn) is configured to include an objective lens (condensing lens) (Ba), a polarizer (Bb), a polarizing beam splitter (Bc), an imaging lens (tube lens) (Bd, Bd'), a field aperture (Be, Be'), and a sensor (Cn, Cn').

[0051] Scattered light incident from a sample (W) to a detection optical system (Bn) is collected and collimated by an objective lens (Ba), and its polarization direction is controlled by a polarizer (Bb). The polarizer (Bb) is a half-wave plate and can be rotated by a driving mechanism (not shown). By controlling the driving mechanism by a control device (E1) and adjusting the rotation angle of the polarizer (Bb), the polarization direction of the scattered light incident on the sensor is controlled.

[0052] The scattered light polarized by the polarization control of the polarizer (Bb) is split into an optical path by the polarization beam splitter (Bc) according to the polarization direction and incident on the imaging lenses (Bd, Bd'). By the combination of the polarizer (Bb) and the polarization beam splitter (Bc), a linear polarization component in any direction is cut. When cutting any polarization component including elliptical polarization, the polarizer (Bb) is configured with a quarter-wave plate and a half-wave plate that can rotate independently of each other.

[0053] Scattered illumination light focused through the imaging lens (Bd) is converted into photoelectric power at the sensor (Cn) via the field aperture (Be), and the detection signal is input to the signal processing device (D). Scattered illumination light focused through the imaging lens (Bd') is converted into photoelectric power at the sensor (Cn') via the field aperture (Be'), and the detection signal is input to the signal processing device (D). The field apertures (Be, Be') are installed so that their centers align with the optical axis of the detection optical system (Bn), and they cut off light generated from locations other than the inspection target location, such as light generated from a position away from the center of the beam spot (BS) of the sample (W) and stray light generated inside the detection optical system (Bn). Accordingly, they have the effect of suppressing noise that interferes with defect detection.

[0054] According to the above configuration, two polarization components that are orthogonal to each other can be simultaneously detected for scattered light generated at the same coordinate, and it is effective when detecting multiple types of defects or haze light with different polarization characteristics.

[0055] In constructing an objective lens (Ba) with multiple lenses arranged in close proximity, in order to reduce the loss of detection light due to the gap between the lenses, the outer periphery of the objective lens (Ba) may be cut off so as not to interfere with the sample (W) or other objective lenses, as shown in the example of FIG. 11.

[0056] - Sensor -

[0057] Sensors (Cn, Cn') are sensors that convert scattered light collected by a corresponding detection optical system into an electrical signal and output a detection signal. Sensors (C1(C1'), C2(C2'), C3(C3')…) correspond to detection optical systems (B1, B2, B3…). For these sensors (Cn, Cn'), single-pixel point sensors such as photomultiplier tubes and SiPMs (silicon photomultiplier tubes) that convert weak signals into photoelectric signals with high gain can be used. In addition, sensors that arrange multiple pixels in one or two dimensions, such as CCD sensors, CMOS sensors, and PSDs (position sensing detectors), may also be used for sensors (Cn, Cn'). The detection signal output from the sensors (Cn, Cn') is input to a signal processing device (D) at any time.

[0058] - Control Unit -

[0059] The control device (E1) is a computer that controls the process diagnostic device (100) in an overall manner and is configured to include a processing device (computation control device) such as a CPU, GPU, or FPGA, in addition to ROM, RAM, and other memory devices. The control device (E1) is connected to an input device (E2), a monitor (E3), and a signal processing device (D) via wired or wireless connections. The input device (E2) is a device through which a user inputs settings of inspection conditions, etc., to the control device (E1), and various input devices such as a keyboard, mouse, or touch panel may be appropriately employed. The output of an encoder of a rotary stage or a translation stage (the rθ coordinate of the beam spot (BS) on the sample) and inspection conditions, etc., input by an operator through the input device (E2) are input to the control device (E1). Inspection conditions include, in addition to the type, size, shape, material, lighting conditions, detection conditions, etc. of the sample (W), for example, sensitivity settings of each sensor (Cn, Cn'), gain values ​​or thresholds used for defect determination or process diagnosis (condition diagnosis of process device).

[0060] Additionally, the control device (E1) outputs a command signal that commands the operation of the stage (ST) or the illumination optical system (A) according to the inspection conditions, or outputs coordinate data of the beam spot (BS) synchronized with the defect detection signal to the signal processing device (D). The control device (E1) also displays the inspection condition setting screen or the sample inspection data (inspection image, etc.) on the monitor (E3). In addition to the final inspection result obtained by integrating the signals of each sensor (Cn, Cn'), the inspection data may also display individual inspection results by these sensors (Cn, Cn').

[0061] In addition, as shown in FIG. 1, a Review Scanning Electron Microscope (Review SEM) for defect inspection may be connected to the control device (E1). In this case, it is also possible to receive data of the defect inspection results from the Review SEM at the control device (E1) and transmit it to the signal processing device (D).

[0062] Additionally, this control device (E1) may be configured as a single computer that forms a unit with the device body (stage, lighting optical system, detection optical system, sensor, etc.) of the process diagnostic device (100), but it may also be configured as multiple computers connected via a network. For example, it may be configured such that inspection conditions are input into a computer connected via a network, and control of the device body or the signal processing device (D) is executed by a computer attached to the device body.

[0063] - Signal Processing Unit -

[0064] The signal processing device (D) is a computer that has the function of detecting defects in a sample (W) by processing detection signals input from sensors (Cn, Cn') of a detection optical system (Bn). Similar to the control device (E1), the signal processing device (D) is configured to include a memory (D1) (Fig. 12) that includes at least one of RAM, ROM, HDD, SSD, or other storage devices, as well as a processing device such as a CPU, GPU, or FPGA. This signal processing device (D) can be configured as a single computer that forms a unit with the main body of the process diagnostic device (100) (stage, lighting optical system, detection optical system, sensor, etc.), but it can also be configured as multiple computers connected via a network. For example, a computer attached to the main body of the device can acquire a defect detection signal from the main body of the device, process the detection data as needed and transmit it to a server, and perform processing such as defect detection or classification on the server. It is also conceivable to configure the signal processing device (D) and the control device (E1) to serve as a single computer.

[0065] FIG. 12 is an example of a functional block diagram of a key part of a signal processing device (D) provided in a process diagnostic device according to the first embodiment of the present invention. As shown in FIG. 12, the signal processing device (D) is equipped with a memory (D1), a fault determination circuit (D2), a low-pass filter circuit (D3), and a process diagnostic circuit (D4).

[0066] While the process diagnostic device (100) is scanning the sample (W), the signal processing device (D) receives a detection signal (scattered light intensity signal) from the sensors (Cn, Cn') and an encoder output of the stage (ST) (rθ coordinate of the beam spot (BS) on the sample) from the control device (E1). In the signal processing device (D), these detection signals and encoder outputs are matched and recorded in memory (D1).

[0067] The defect determination circuit (D2) reads detection signals input from sensors (Cn, Cn') from memory (D1) in chronological order, sequentially determines whether these detection signals are defect signals that have detected a defect, records the determination result in memory (D1) or storage device (DB), and also outputs it to the control device (E1). In the defect determination circuit (D2), for example, high-frequency components of the detection signals are extracted as defect signals related to defects such as foreign matter. High-frequency components are components with high fluctuation frequencies, specifically components in which the time fluctuation of the value exceeds a preset value. The control device (E1) displays and outputs the determination result to the monitor (E3) either automatically or in accordance with the operation signal from the input device (E2) that is input along with the operator's operation.

[0068] The low-pass filter circuit (D3) reads out detection signals from sensors (Cn, Cn') from memory (D1) in chronological order, extracts haze signals excluding defect signals for each region of the sample (W), and creates a haze map, which is a light intensity distribution of the entire surface, by adding coordinate information to the haze signals on the surface of the sample (W). The haze signal refers to the low-frequency components among the light signals obtained from the sample and is a signal mainly attributed to the characteristics of the sample. Here, for example, the low-frequency components of the detection signal, that is, the fluctuation frequency (time fluctuation of signal intensity), are extracted as the haze signal, which is lower than a preset value. Since the optical defect inspection device is capable of high-speed scanning, it is possible to extract the haze signal for the entire surface of the sample or the haze map based thereon. However, there may also be cases where one wants to diagnose the sample partially rather than the entire surface of the sample. In this case, the region for extracting the haze signal may be a sampling point, or the haze signal may be extracted by dividing the area into regions partitioned by a mesh with a mesh shape of arbitrary eye size.

[0069] When extracting haze signals by dividing the area into sections using a mesh-shaped grid, multiple haze signals are acquired for a single area. The statistical values ​​(average value, median value, etc.) of these multiple haze signals can be used as the haze signals for that area. One side of the mesh dividing the area can be set to, for example, 1 mm to several mm. Depending on the size of the mesh, for example, if the mesh is 1 mm, the sample surface is divided into more than 60,000 areas, and a fine haze map is generated. However, the mesh size is not limited to being smaller; by setting the mesh large within a necessary and sufficient range, the computational load of the signal processing device (D) can be suppressed as the number of areas on the sample surface is reduced.

[0070] The process diagnostic circuit (D4) diagnoses the condition of the process device that processed the sample (W) based on the haze signal extracted from the low-pass filter circuit (D3). When the process diagnostic circuit (D4) suspects an abnormality or a premonition thereof in the process device, it outputs an alarm related to the condition of the process device (e.g., a screen or warning sound suggesting an abnormal location or inspection of the process device) to the monitor (E3) via the control device (E1). An abnormality or a premonition thereof in the process device is suspected by comparing the haze signal of each entire surface or area of ​​the sample (W) with the reference haze signal of the corresponding area. An abnormality or premonition is suspected when the difference between the measured haze signal and the reference haze signal exceeds a preset value. In addition, the haze signal used in this comparison may be not only the haze signal itself but also other output forms such as a haze map.

[0071] The diagnostic result of the process device (including normal determination in addition to the above alarm) is recorded in memory (D1) or storage device (DB) and output to the control device (E1). The control device (E1) displays and outputs the diagnostic result to the monitor (E3) either automatically or in accordance with an operation signal from the input device (E2) that is input along with the operator's operation.

[0072] In addition, from the perspective of early detection of abnormalities or warning signs of the process device, it is desirable to sequentially process detection signals input from sensors (Cn, Cn') in conjunction with defect inspection of the sample (W) to perform process diagnosis. However, the system may also be configured to store detection signals acquired by scanning the entire surface of the sample (W) in a memory device (DB) and then perform process diagnosis by post-processing the stored data at a desired timing (for example, at the top of every day).

[0073] - Reference Haze Signal -

[0074] In this embodiment, the reference haze signal used for process diagnosis is defined for each sensor (Cn, Cn') for the entire surface of the sample (W) or for a predetermined area (rθ coordinate in this embodiment) and is stored, for example, in a memory device (DB). The reference haze signal defined for the same sensor may be different for each area of ​​the sample (W).

[0075] The reference haze signal is set, for example, by scanning a reference sample. The reference sample is a sample that meets the standard in quality inspection, and preferably is of the same species as the sample (W) and is also a sample of the same process as the sample (W). However, the reference haze signal may also be set by calculating statistical data (e.g., average value, median value) of the haze signal obtained from defect inspection in the semiconductor manufacturing process for samples judged to be good, rather than by measuring the reference sample. Alternatively, the reference haze signal may be set by simulating the haze signal obtainable for each detection optical system (Bn) based on the design data of the sample (W).

[0076] When setting a reference haze signal by actual measurement of a reference sample, for example, one or multiple reference samples are prepared and scanned by a process diagnostic device (100). This scan may be performed under the same conditions as the inspection of the sample (W), or adjustments may be made as necessary, such as performing it under lower sensitivity conditions than the inspection of the sample (W) when damage to the sensors (Cn, Cn') is a concern. From the detection signal obtained for the reference sample by this scan, a haze signal is extracted by the low-pass filter circuit (D3) of the signal processing device (D) as described in FIG. 12, and a reference haze is set based on this haze signal and stored in a memory device (DB). At that time, by using the statistical data of the reference haze signals obtained for multiple reference samples as the reference haze signal, the reliability of the reference haze signal as a reference value is increased.

[0077] - Correlation Data -

[0078] In addition, the memory device (DB) has a correlation between the detection optical system (Bn) (in other words, the direction of emission of haze light) and the variation factor of the haze signal stored in advance, along with the reference haze.

[0079] In semiconductor manufacturing processes, there are dozens of processes that perform plasma treatment. For each process, the materials (such as the film quality of the sample or the type of gas used in the chamber) and treatment conditions differ. In the sample (W) after etching treatment, there are processes where changes in surface roughness are likely to occur, and processes where changes in surface film thickness are likely to occur. Alternatively, there are cases where characteristic tendencies are likely to occur depending on the direction of gas flow within the chamber of the process device. The "variable parameters" differ for each process. Therefore, it is desirable for the process diagnostic device (100) to diagnose the process device using a detector signal that is easy to detect changes according to the variable parameters.

[0080] For example, when a sample (W) is processed in a plasma process device, the surface roughness of the sample (W) varies. If this variation is within a certain range, the change tends to appear in the intensity of scattered light incident on the apertures (α3, α4) located mainly in the direction of specular reflection of the illumination light relative to the beam spot (BS). Similarly, if the surface film thickness of the sample (W) varies within a certain range, it tends to appear as a change in the intensity of scattered light incident on the apertures (α1, α6) located on the opposite side from the direction of specular reflection relative to the beam spot (BS). Based on this relationship, for each process using the plasma process device, a variation factor and a detection optical system with high sensitivity to that variation factor are associated and stored in a memory device (DB) as correlation data. In advance, a specific detection optical system (Bn) is selected based on this correlation data by a signal processing device (D), and by monitoring changes in surface roughness or surface film thickness based on the haze signal of the selected detection optical system (Bn), abnormalities or pre-infections of related parts of the process device (e.g., plasma light source) are detected with high sensitivity.

[0081] In addition, the above correlation data is merely an example of the correlation between the fluctuation factor of the haze signal and the detection optical system (Bn). For example, if it is discovered that some abnormality or premonitory sign of the process device appears in the haze signal acquired from a detection optical system other than the detection optical system (Bn) corresponding to the aperture (α3, α4, α1, α6), correlation data may be established based on such discovery. Furthermore, it is not limited to the haze signal detected individually by the detection optical system (Bn), but there is also a possibility that some abnormality or premonitory sign of the process device appears in the difference or sum of haze signals detected by multiple detection optical systems (Bn). In this case, correlation data based on the said correlation is defined, and the difference or sum of the haze signals detected by multiple detection optical systems (Bn) is treated as a form of haze signal and compared with the difference or sum of a reference haze signal associated with the same set of detection optical systems (Bn) to be used for diagnosing the process device.

[0082] When diagnosing a process, the signal processing device (D) reads the correlation data from the memory device (DB) and automatically selects a detection optical system corresponding to the state of the process device to be diagnosed. However, the signal processing device (D) may also select a detection optical system according to a designation made by an operator through an input device (E2). Then, the signal processing device (D) detects a change in the state of the process device based on a haze signal output from the selected detection optical system. For example, according to the example described above, the signal processing device (D) selects a detection optical system (Bn) corresponding to an aperture (α3, α4, α1, α6) located in the direction of specular reflection of the illumination light with respect to the beam spot (BS). From the difference between the haze signal detected by the detection optical system (Bn) corresponding to the aperture (α3, α4) and the reference haze signal, a change in the surface roughness of the sample (W) within a predetermined range, and furthermore, a change in the state of the process device related to this change in the surface roughness within the predetermined range, is detected. From the difference between the haze signal of the detection optical system (Bn) corresponding to the aperture (α1, α6) and the reference haze signal, a change in the surface film thickness of the sample (W) within a predetermined range, and furthermore, a change in the state of the process device related to the change in the surface film thickness within a predetermined range, is detected.

[0083] In addition, even if haze light is incident on the same detection optical system (Bn), the manner in which variations occur according to the state of the process device may differ depending on the polarization direction. In this regard, in the present embodiment, the detection optical system (Bn) is equipped with a polarization beam splitter (Bc) that separates light according to the polarization direction, and a plurality of sensors (Cn, Cn') that each detect light with different polarization directions separated by the polarization beam splitter (Bc) (Fig. 11). Accordingly, in the present embodiment, two haze signals with different polarization directions can be acquired for the same coordinate on the sample in each detection optical system (Bn). Thus, by adding the polarization direction of the haze light as a parameter of the correlation data of the detection optical system (Bn) and the variation factor of the haze signal described above, a more precise relationship between the haze signal and the variation factor can be defined and stored in a memory device (DB). In this way, by increasing the parameters of the correlation data, changes in the state of the process device can be detected more precisely from the haze signal.

[0084] In addition, there may be a correlation between the deteriorated parts of the process device and the location on the sample. On the sample (W) processed by the plasma process device, the haze map can identify areas with significant variation (areas where there is a high probability of any finishing defects occurring) for the sample from which the reference haze signal was acquired. The memory device (DB) stores information on the process device's parts (location within the device, material, etc.) and maintenance information (e.g., maintenance cycle, date on which parts were actually replaced, etc.). The signal processing device (D) can analyze the correlation between the location of the area specified by the haze map on the sample and the parts of the process device, and store the data in advance in the memory device (DB). In this case, the signal processing device (D) can identify not only the presence or absence of an abnormality or a premonition of an abnormality in the process device, but also the parts of the process device or inspection points where the abnormality or premonition is suspected, and notify the operator along with an alarm. By comparing the diagnostic results of the process diagnostic device with component information or maintenance information on the process device side, the user can determine the optimal timing for maintenance.

[0085] In the memory device (DB), defect information, in addition to the haze map, is typically stored as information output by the process diagnostic device (100). The process diagnostic device (100), as a defect inspection device, can output defect information of the entire surface or a specific area (e.g., the coordinates or brightness of each defect, a defect map of the entire surface of the sample, etc.) by scanning the sample (W). The signal processing device (D) of the process diagnostic device (100) may also analyze the correlation with the process device by combining the haze map and the defect information.

[0086] - Semiconductor Manufacturing Process -

[0087] Here, FIG. 13 is a flowchart outlining an example of a series of diagnostics for a process device in a semiconductor manufacturing process. Here, a plasma etching device is used as an example for explanation.

[0088] In the drawing, first, a wafer is subjected to plasma etching treatment (step S10) in a plasma etching device (process device). The plasma etching device monitors fluctuations in its operating parameters during the plasma etching treatment. For example, the plasma etching device is equipped with optical emission spectrometry (hereinafter abbreviated as OES) to monitor the state of the plasma light source. In addition, the plasma etching device has sensors installed at various locations within the chamber to monitor the state inside the plasma chamber. Then, if an abnormality exceeding an allowable range in the behavior of the operating parameters is detected based on a predetermined algorithm, the plasma etching device outputs an alarm and stops.

[0089] The wafer processed in the plasma etching device has its CD (Critical Dimension) measured in-line, for example, using a CD-SEM or OCD (Optical Critical Dimension) (Step S20). "In-line" means as a process in semiconductor manufacturing or in the process of a semiconductor research and development / manufacturing line. The OES is mounted on the plasma etching device. In addition, the etching rate (surface film thickness) of the wafer surface is also inspected in-line using, for example, an AFM or a spectroscopic ellipsometer (Step S40). In addition, defects such as foreign matter on the wafer surface are inspected in-line using an optical inspection device (Step S60). The order of inspection for CD value, etching rate, and defects can be changed arbitrarily.

[0090] An abnormality in the plasma etching device affects the condition of the wafer processed by the plasma etching device. As a typical example, it affects the CD value, surface film thickness, and number of defects of the processed wafer. By combining wafer quality inspection and diagnosis of the plasma etching device, it is determined whether the inspection results of the CD value or etching rate fall within the reference range (Steps S30, S50). If the inspection results deviate from the reference range, the plasma etching device suspected of being abnormal is stopped, and the abnormality of the plasma etching device is resolved through processes such as identifying the cause, taking countermeasures, and performing operation tests. Once it is confirmed that the abnormality of the plasma etching device has been resolved, the plasma etching device is restarted (Step S80), and the process proceeds to the next wafer processing.

[0091] In this embodiment, in step S60, a process diagnostic device (100), which is also an optical defect inspection device, is used to perform a diagnosis of the process device in parallel with the defect inspection of the wafer, and it is determined whether the diagnosis result falls within a reference range (step S70). If the diagnosis result deviates from the reference range, the plasma etching device in which an abnormality or pre-existing condition is detected is stopped, and the abnormality of the plasma etching device is resolved through processes such as identifying the cause, taking countermeasures, and performing operation tests. Once it is confirmed that the abnormality of the plasma etching device has been resolved, the plasma etching device is restarted (step S80), and the process proceeds to the next wafer processing.

[0092] The sequence of process diagnosis of a process device by the process diagnosis device (100) according to steps S60 and S70 will be described later.

[0093] - Change in the state of the process device -

[0094] FIG. 14 is a schematic diagram showing the change in the state of the process device over time, and FIG. 15 is a schematic diagram showing the effect on the wafer accompanying the change in the state of the process device over time, focusing on section XV of FIG. 14. The horizontal axis of FIG. 14 represents time, and the vertical axis represents the state of the process device (in this case, the plasma discharge state of the plasma etching device).

[0095] As shown in Fig. 14, if the plasma state is in a normal state a, which is better than a certain level (the abnormal line in Fig. 14), many good wafers that satisfy standards such as CD value and surface film thickness are manufactured. However, the plasma state deteriorates with the passage of operating time of the plasma etching device, and as a result, the quality of the processed wafers also deteriorates. Therefore, the plasma etching device is regularly maintained (PM: Preventive Maintenance). However, the period until the next regular maintenance PM is not necessarily appropriate, and there are cases where the deterioration of the plasma light source, etc., exceeds a certain level and an abnormality occurs before the next maintenance period arrives. If the plasma state falls into an abnormal state c due to deterioration, the CD value and surface film thickness of the processed wafers, which are monitored daily by the user, do not meet the standards. The yield also decreases.

[0096] As described above, changes in the plasma state of a plasma etching device manifest as changes in the state of the wafer. Changes in the wafer state cannot be said to be linked one-to-one with specific physical properties such as the CD value, surface film thickness, or number of defects mentioned above. In the case of changes that are too minute to be measured by specific physical properties, or changes caused by complex factors, they cannot be identified by the CD value, etching rate, or number of foreign substances. To measure such changes, the process diagnostic device (100) of the present invention is used. In particular, abnormalities caused by plasma degradation may exhibit a characteristic distribution in the position of the plasma in the plasma etching device or in the direction of gas flow within the chamber. Such characteristic distribution tends to appear locally within the entire surface of the sample processed by the plasma etching device. Where on the entire surface of the sample the local area is located varies depending on the type of etching device. Even if the change cannot be directly linked to specific physical properties such as the sample phase, CD value, surface film thickness, or number of defects, the plasma diagnostic device of this embodiment can detect plasma degradation by measuring the entire surface of the sample and expressing the distribution. Furthermore, it can detect the degradation earlier than it appears in specific physical properties such as the CD value, surface film thickness, or number of defects.

[0097] In FIG. 15, examples of haze maps (haze distribution on the sample surface) obtained from a process diagnostic device (100) are shown, each using a wafer processed in a plasma etching device as a sample (W) when the plasma state is normal state a, transient state b, and abnormal state c of FIG. 14. Transient state b is a state in which the yield is not below the allowable value at the current time, but has deteriorated to the extent that it is expected to fall below the allowable value within a predetermined period. In FIG. 14, transient state b is a state in which the plasma discharge state has deteriorated to a level below a pre-treatment line (however, above the abnormal line) that is set higher by a predetermined margin than, for example, the abnormal line. Transient state b corresponds to a state in which the abnormal pre-treatment is visible.

[0098] When looking at the haze map obtained from the detection optical system (Bn) which has high sensitivity to changes in the haze signal, as shown in FIG. 15, for the sample (W) processed in normal state a, there is a difference in the intensity of the haze signal at edge X1 or local X2 for the sample (W) processed in transient state b. Since the haze map of the sample (W) processed in normal state a is approximated by the haze map of the reference haze signal, the difference from the haze map of the reference haze signal becomes larger than in transient state b when the plasma state deteriorates further and falls into abnormal state c. In abnormal state c, the yield falls below the allowable value. Therefore, the signal processing device (D) determines the transient state b in a timely manner during the inline inspection of the sample (W) by the process diagnostic device (100), and recommends to the operator to perform maintenance in the transient stage b before the plasma state falls into abnormal state c. That is, the signal processing device (D) detects a change in the state of the process device, which appears as a microscopic surface shape of the sample (W), based on the difference between the haze signal and the reference haze signal during the daily defect inspection.

[0099] - Process Diagnosis -

[0100] FIG. 16 is a flowchart showing the sequence of diagnostic processing of a process device by a process diagnostic device (100) according to the present embodiment. The processing of the said figure is executed in steps S60 and S70 of the flowchart of FIG. 13. The flow of FIG. 16 is executed in steps S60 and S70 of FIG. 13.

[0101] When performing a defect inspection on a sample (W) processed by a process device, the signal processing device (D) of the process diagnostic device (100) reads a reference haze from a memory device (DB) (Step S61). The signal processing device (D) extracts a haze signal from a detection signal detected during a defect inspection (full-surface scan) using a low-pass filter circuit (D3) and inputs the haze signal for an arbitrary area of ​​the sample (W) (Step S62). The arbitrary area may be the entire surface of the sample (W) or limited to a specific area within the sample (W). Alternatively, the sample (W) may be divided into multiple areas and the process may be performed sequentially for each area. For the area in question, the signal processing device (D) compares the haze signal input in Step S62 with the reference haze (Step S63) and determines whether the difference between the two is within a preset value (Step S64). The signal processing device (D) records the area as a normal area if the difference between the haze signal and the reference haze is within a set value (step S65), and records the area as a defective area if it exceeds the set value (step S66). Also, for convenience of explanation, the area with a large change in the sample from which the reference haze signal was acquired is designated as a “defective area.”

[0102] For example, if an arbitrary area is defined as the entire surface of the sample (W), the haze map can be perceived as a shape, and the shape in which the difference from the reference haze manifests can be used as a judgment criterion. For example, the surface of the sample (W) can be divided into a circular central part and an annular outer part surrounding it, and a judgment criterion can be applied in which a specific part among the central and outer parts—for example, the outer part—focuses on the change in the haze signal and determines it as a defective area when there is a change of a certain amount or more in the outer part. Alternatively, a judgment condition can be set where there is a change in the haze signal of a certain amount or more in both the central part and the outer part.

[0103] When the signal processing device (D) divides the sample (W) into multiple regions, it repeats the processing of steps S61-S66 for each region and performs the processing on the entire region of the sample (W), and determines whether the diagnosis result is within the reference range, specifically whether the number of defective regions is less than or equal to a preset allowable value (step S71).

[0104] Additionally, the processing of steps S62-S66 may employ an algorithm that creates a haze map of the sample (W) and then compares the haze map of the sample (W) with the intensity distribution of the reference haze signal (haze map related to the reference haze signal). For example, regarding the haze map of the sample (W), it is determined whether the luminance difference with the reference haze (in this case, referring to the haze map) exceeds a set value. Alternatively, it may be an algorithm that counts the area exceeding the set value.

[0105] The signal processing device (D) estimates the normal state of the process device if the difference from the reference haze across the entire surface of the sample (W) is less than or equal to a set value, or if the number of defective areas is less than or equal to an allowable value. In this case, the signal processing device (D) generates data indicating that the process device is in a normal state, records it, for example, in a memory device (DB), displays it on a monitor (E3) via a control device (E1), and terminates the procedure (Step S72). Conversely, if the difference from the reference haze exceeds the set value, or if the number of defective areas exceeds the allowable value, an abnormality or a premonitory sign is suspected to have occurred in the process device. In this case, the signal processing device (D) generates alarm data indicating that an abnormality or a premonitory sign has occurred in the process device, records it, for example, in a memory device (DB), and outputs it to a monitor (E3) to terminate the procedure (Step S72). By inviting the operator to perform maintenance of the process device through the alarm output to the monitor (E3), the procedure in FIG. 13 proceeds to step S80, and maintenance of the process device is performed.

[0106] In addition to messages via sound or text, the output to the monitor (E3) in steps S72 and S73 may also display a haze map of the sample (W). Furthermore, if characteristic changes appearing in the sample (W) (e.g., changes in surface roughness, changes in surface film thickness), deteriorated parts of the process device, or areas requiring inspection are identified by the signal processing device (D), these data may be output to the monitor (E3) along with an alarm as additional information.

[0107] - effect -

[0108] (1) According to the present embodiment, when inspecting defects in a sample (W), an abnormality or premonitory condition of the process device can be detected by the haze signal obtained by scanning the entire surface of the sample (W). It is also possible to perform high-speed process diagnosis in parallel with defect inspection with a single scan. The haze signal is not converted into a characteristic value of the sample (W), such as surface roughness or surface film thickness, and compared with a characteristic value measured in advance by an AFM, etc., but is evaluated as the difference from a reference haze, which is the same haze signal, that is, the relative change over time of the haze light. In order to set the value (reference haze) that serves as the judgment standard, it is not necessary to verify the correlation with a separate measurement device (ellipsometry, AFM, etc.) in addition to the process diagnosis device (100). In addition, since both the reference haze and the haze signal compared therewith are obtained by measuring the entire surface of the sample (W), changes appearing in the sample (W) according to the state of the semiconductor plasma device can be detected with high precision based on the intensity distribution of haze light (haze map) of the entire surface of the sample. In particular, since the plasma discharge abnormality of the plasma etching device appears locally within the surface of the sample (W), the process diagnosis of this embodiment, which can identify the distribution of haze light intensity fluctuations within the surface, is particularly effective.

[0109] By the process diagnostic device (100), a haze signal is acquired in conjunction with a defect inspection of a sample (W) performed daily in-line, and a relative change in the haze signal is detected to identify abnormalities or signs thereof in the process device. The diagnosis result of the semiconductor manufacturing process based on this haze light is naturally obtained in conjunction with the defect inspection, and does not force an increase in the workload or inspection costs on the operator. Since abnormalities in the process device are detected early in the defect inspection performed daily during the semiconductor manufacturing process, the occurrence of defects in the sample (W) is suppressed, and the occurrence of samples (W) provided for analysis of the cause of defects by TEM, etc. is also suppressed. This leads to avoidance of shutdowns caused by failures in the process device, and an improvement in the operating rate of the process device can be expected.

[0110] In addition, since the change in the intensity distribution of haze light in the sample (W) can be known, when an abnormality or an indication thereof is estimated from the correlation between the influence appearing according to the state of the semiconductor plasma device and the position on the sample, depending on the conditions, the defective location of the process device can also be estimated.

[0111] As described above, according to the present embodiment, the condition of the process device can be precisely diagnosed from data obtained from the defect inspection of the sample (W) after the process, and the occurrence of abnormalities in the process device can be suppressed to improve the yield of the semiconductor.

[0112] (2) Multiple detection optical systems (Bn) are arranged in different directions relative to the beam spot (BS), and one or more detection optical systems (Bn) effective for capturing changes in the intensity of haze light can be selected, and process diagnosis can be performed using only the selected detection optical systems (Bn). For example, if the configuration is such that the signals of multiple detection optical systems (Bn) are merged and output regardless of sensitivity to haze light, the change detected with high sensitivity by a specific detection optical system (Bn) is diluted, and the inspection sensitivity is reduced. In contrast, in this embodiment, process diagnosis can be performed with high sensitivity by utilizing a configuration having multiple detection optical systems (Bn) with different directions.

[0113] In this embodiment, in particular, the correlation between the haze signal and its variation factor is stored in a memory device (DB) for each detection optical system (Bn), and based on this correlation, the detection optical system (Bn) of a specific azimuth angle φ1 can be selectively used for process diagnosis. By doing so, abnormalities or signs thereof in the process device can be detected according to variation factors of haze light intensity, such as the surface film thickness or surface roughness of the sample surface. In this embodiment, an example was described of detecting a change in the state of the process device related to a change in the surface roughness of the sample (W) within a predetermined range from the difference between the haze signal incident on the apertures (α3, α4) located in the specular reflection direction of the illumination light with respect to the beam spot (BS) and the reference haze signal. In addition, an example was described of detecting a change in the state of a process device related to a change in the surface film thickness of a sample (W) within a predetermined range from the difference between a haze signal incident on an aperture (α1, α6) located opposite to the direction of specular reflection of the illumination light with respect to the beam spot (BS) and the reference haze signal.

[0114] In addition, in the present embodiment, the detection optical system (Bn) separates the haze light according to the polarization direction at the polarization beam splitter (Bc) and can detect two lights with different polarization directions for the haze light emitted from the same coordinates in the same direction on the sample (W). With this configuration, for each detection optical system (Bn) (i.e., for each direction of haze light emission), based on correlation data defined for the intensity of the haze signal, the polarization direction, and the variation factor of the haze signal, a more detailed process diagnosis can be performed in which the polarization direction of the haze light is also included as a parameter.

[0115] (2nd Example)

[0116] In the first embodiment, an example was described of diagnosing a process device by actually comparing the haze signal of each region of the sample (W) with a reference haze signal. In this way, by accumulating data regarding the diagnosis performed daily in the semiconductor manufacturing process and applying machine learning, process diagnosis can be performed from the haze map of the sample (W) that is frequently input, based on the completed model.

[0117] The completed learning model is an inference program in which completed learning parameters are introduced through machine learning of training data, and outputs a diagnostic result of the process device regarding data concerning the input haze signal. This completed learning model is created in a signal processing device (D) or a control device (E1) and is stored, for example, in a memory device (DB). The signal processing device (D) uses this completed learning model to diagnose the process device based on data regarding the haze signal acquired during defect inspection of the sample (W), and outputs an alarm as necessary.

[0118] Examples of training data include the haze map of the sample (W), the polarization direction of the haze light, the diagnostic results of the process device based on the sample (W), the actual maintenance history of the process device, and the disputes regarding the diagnostic results, etc., which are actual data accumulated in the daily semiconductor manufacturing process. The maintenance history and the disputes regarding the diagnostic results are a type of feedback data, and, for example, a person who has maintained the process device can input them via an input device (E2) according to a pre-prepared input screen. The disputes regarding the diagnostic results are, for example, the judgment of the person who has maintained the process device, and include matters such as whether the alarm notified by the process diagnostic device (100) was valid, or whether there was no notification of an alarm that the process device was maintained when it became necessary.

[0119] FIG. 17 is a conceptual diagram of machine learning. Here, an example is described in which machine learning is performed in a signal processing device (D) to generate a completed learning model. The signal processing device (D) retrieves performance data such as the haze map of the sample (W) described above, diagnostic results, maintenance history of the process device, and disputes regarding the diagnostic results, and reads and inputs this data from a memory device (DB) to generate training data. The signal processing device (D) reads this training data into a neural network (D9) and optimizes the weighting of connections between neurons in the input layer, intermediate layer, and output layer. In this way, a completed learning model for diagnosing the process device is generated from the data of the haze signal obtained for the sample (W), such as scattering direction, light intensity, polarization direction, and coordinates. Furthermore, the completed learning model is not limited to the signal processing device (D) and may be generated on another computer.

[0120] In addition, machine learning can be performed using a signal processing device (D) or a control device (E1) as input for the haze map and maintenance execution data, and the maintenance cycle for each part or section of the process device can be calculated and optimized. A learned model for estimating abnormal parts of the process device can be obtained, for example, from the correlation between the part (or section) of the process device and the variation area (haze map) of the haze signal in the sample (W).

[0121] In other respects, this embodiment is the same as the first embodiment. In this embodiment, a fully learned model is generated by adding feedback data, such as maintenance history and disputes over diagnostic results, along with the accumulation of diagnostic data based on the comparison of the difference between a haze signal and a reference haze signal, thereby allowing for an improvement in the diagnostic accuracy of the process device. Furthermore, as described above, optimization of the maintenance cycle of the process device can also be expected.

[0122] (3rd Example)

[0123] FIG. 18 is a schematic diagram showing the essential parts of a process diagnostic device according to a variant of the present invention. In FIG. 18, elements identical to or corresponding to the elements described in the first and second embodiments are given the same reference numerals as in the previously published drawings, and descriptions are omitted.

[0124] This embodiment is an example of including data obtained from a plurality of process diagnostic devices in the basic data of the reference haze signal (first embodiment) or the learned model (second embodiment) described above. In this embodiment, the process diagnostic device (100) is appropriately connected to a data server (DS) via a network (not shown). A process diagnostic device (100', 100") different from the process diagnostic device (100) is appropriately connected to the data server (DS) via a network. The process diagnostic devices (100, 100', 100") are preferably of the same type or equivalent type (same series, same manufacturer, etc.), but they may be of different types. Although FIG. 18 shows two different process diagnostic devices (100', 100"), the other process diagnostic devices connected to the data server (DS) may be one or three or more.

[0125] Diagnostic data, etc., from process diagnostic devices (100, 100', 100") are input into the data server (DS), and these data are accumulated. This accumulated data may include, for example, design data of the sample (W), maintenance data of the process device, the determination of diagnostic results, and inspection data of the sample (W), in addition to diagnostic data of the process device, such as haze signals and diagnostic results for each process diagnostic device. Furthermore, regarding defect inspection of the sample (W), inspection conditions (inspection recipes), defect review data, and defect material analysis data may also be accumulated in the data server (DS). Defect material analysis data includes, for example, information obtained through energy dispersive X-ray analysis. This may be a standalone device, or it may be mounted on a defect review device, or it may be acquired together with defect review information. Based on this accumulated data, the data server (DS) calculates a reference haze or a learned model that is compared with the sample (W) regarding process diagnosis. The calculation of the reference haze signal or the learned model is performed in the data server (DS). In this case, it may be executed at regular intervals, or it may be executed when new data accumulates above a certain level. Each process diagnostic device (100, 100', 100") receives a reference haze signal or a learned model that is frequently updated from a data server (DS), and performs a diagnosis of the process device based on the haze signal of the sample (W). The haze signal refers mainly to the low-frequency component among the optical signals obtained from the sample, and is a signal mainly attributed to the characteristics of the sample. Here, for example, the low-frequency component of the detection signal, i.e.

[0126] According to the present embodiment, in addition to the self-data of the process diagnostic device (100), a reference haze signal or a learned model is calculated using multiple data from other process diagnostic devices (100, 100') as basic data. As a result, more basic data is accumulated in a short period of time, and the diagnostic accuracy can be improved early.

[0127] (Fourth Example)

[0128] FIG. 19 is a schematic diagram showing the essential parts of a process diagnostic device according to the fourth embodiment of the present invention. In FIG. 19, elements identical to or corresponding to the elements described in the first to third embodiments are given the same reference numerals as in the previously published drawings, and their description is omitted.

[0129] This embodiment is a variation of the method for acquiring a haze signal. On the movement axis of the translational stage of the stage (ST), a sample transfer position (Pa) and an inspection start position (Pb) are set, and by driving the translational stage, the stage (ST) moves along a straight line passing through these positions. The inspection start position (Pb) is a position where an inspection of the sample (W) is started by irradiating the sample (W) with illumination light, and it is a position where the center of the sample (W) coincides with the beam spot (BS) of the illumination optical system (A). The sample transfer position (Pa) is a position where the sample (W) is loaded and unloaded by the arm (Am) relative to the stage (ST), and the stage (ST) that has received the sample (W) moves from the sample transfer position (Pa) to the inspection start position (Pb).

[0130] Recently, due to the demand for increasingly high-sensitivity inspection, the detection optical system (Bn) is positioned close to the sample (W). When the stage (ST) is located directly below the detection optical system (Bn), the gap (G) between the stage (ST) and the detection optical system (Bn) is approximately a few millimeters or less. Since it is difficult to insert the sample (W) into the gap (G) and place it on the stage (ST) using the arm (Am) at the inspection start position (Pb), a configuration is adopted in which the sample (W) is transferred at a sample transfer position (Pa) located away from the inspection start position (Pb).

[0131] In defect inspection, while the stage (ST) moves from the inspection start position (Pb), an illumination light of P-polarization is typically scanned onto the sample (W); however, in this embodiment, a preliminary scan is performed while the stage (ST) moves from the sample exchange position (Pa) to the inspection start position (Pb). In the preliminary scan, the illumination light is set to S-polarization, and the sample (W) is scanned in a spiral trajectory from the outer periphery toward the center. Then, diagnostic processing of the process device is performed based on the haze signal obtained from this preliminary scan.

[0132] In other respects, this embodiment is similar to the first embodiment, the second embodiment, or the third embodiment.

[0133] Here, the defect inspection of the sample (W) is set so that the generation of haze light, which is generally considered noise in defect inspection, is suppressed (for example, the illumination light is set to P-polarization). Therefore, under other conditions, it may be conceivable that the haze light cannot be sufficiently detected during the defect inspection of the sample (W), making it difficult to diagnose the process device based on the haze signal.

[0134] In this regard, in the present embodiment, a haze signal can be collected by performing a preliminary inspection under conditions different from the defect inspection by utilizing the opportunity for the sample (W) to move from the sample exchange position (Pa) to the inspection start position (Pb). By utilizing the sample (W) return operation for collecting the haze signal in this way, defect inspection and process diagnosis can be achieved simultaneously without changing the series of machine operations during the defect inspection.

[0135] In addition, although the present embodiment describes an example of acquiring a haze signal when the sample (W) moves from the sample exchange position (Pa) to the inspection start position (Pb) before the defect inspection, a configuration for acquiring a haze signal when the sample (W) moves to the sample exchange position (Pa) after the defect inspection can also be considered.

[0136] In addition, separate from the opportunity for the sample (W) to move from the sample exchange position (Pa), after the sample (W) arrives at the inspection start position (Pb), the sample (W) may be pre-scanned under conditions where haze light is likely to occur prior to defect inspection to collect a haze signal.

[0137] (Variation Example)

[0138] (1) Regarding the operation of the process diagnostic device (100)

[0139] As described above, the process diagnostic device (100) can diagnose the process device during defect inspection of the sample (W), but defect inspection and process diagnosis can be performed independently, and process diagnosis is not necessarily performed during defect inspection of all samples (W). Therefore, for example, when inspecting defects of samples (W) in one lot, it is possible to perform process diagnosis along with defect inspection only on the first sample (W) of the lot, and to perform only defect inspection without performing process diagnosis on samples (W) from the second sample onwards of the lot. Even by performing process diagnosis once for each inspection of one lot of samples (W), a sufficient opportunity for process diagnosis is obtained.

[0140] (2) Applications of the haze signal

[0141] In each embodiment, examples were described for estimating the state of the process device and defective parts or areas of the process device based on fluctuations in the haze signal and the polarization direction or region of the fluctuations in the haze signal; however, the scope of diagnosis can be broadened by understanding the haze signal from more multifaceted perspectives.

[0142] For example, as shown in FIG. 2, when scanning a sample (W) by rotating it, it is assumed that even with the same detection optical system (Bn), the intensity of the haze light may change according to the rotation angle of the sample (W). In this case, the selection of the detection optical system (Bn) for process diagnosis related to the same sample (W) may be configured to switch periodically according to the rotation angle of the sample (W). In each of the above embodiments, typically, a bare wafer without a pattern or a wafer with a film attached is the subject of inspection. For example, when rotating and scanning a patterned wafer as a sample (W), the scattering direction of the haze light may change regularly due to the influence of diffraction occurring in fine linear patterns formed periodically in the longitudinal and transverse directions. In such cases, data regarding the relationship between the rotation angle of the sample (W) and the selection of the detection optical system (Bn) may be defined in advance and stored, for example, in a memory device (DB), and a configuration in which the selection of the detection optical system (Bn) switches according to the rotation angle of the sample (W) may be effective.

[0143] In addition, regarding the same sample (W), it is also conceivable to analyze or machine learn the correlation with the state of the process device, etc., for the data set of haze signals and defect signals, in addition to the haze signal, and to diagnose the process based on the haze signal and defect signal. It is conceivable that defects in the sample (W) occur due to defects in the process device, or that such defects affect the haze light, and there is a possibility that the accuracy of the process diagnosis may be improved by monitoring the defect signal along with the haze signal.

[0144] In addition, the plasma etching device may be equipped with an OES for monitoring the plasma state. It is also conceivable to analyze or machine learn the monitoring data during plasma etching by this OES along with the haze signal in a signal processing device (D) or a server. If the correlation between the monitoring data of the plasma discharge state and the haze signal can be determined, further improvement in the precision of process diagnosis can be expected.

[0145] In addition, defect inspection is performed whenever one or several processes are passed during the semiconductor manufacturing process, and haze signals can be acquired during defect inspection before and after the process by the process device to be diagnosed. It is also conceivable to calculate the difference between the haze signals before and after the process for the same sample (W) and evaluate the degree of processing of the process device based on that difference. That is, calculating the difference between the haze signals before and after the process for a reference sample as a reference haze signal related to the degree of processing of the process device, and comparing the same difference related to the sample (W) with the reference haze signal, can also be considered as a form of process diagnosis.

[0146] In addition, in the fourth embodiment, an example was described in which a sample (W) is scanned under conditions where haze light is likely to occur separately from defect inspection to acquire a haze signal. In this case, haze light is acquired even during defect inspection, and by comparing the haze signals sampled under conditions where haze light is likely to occur and conditions where haze light is unlikely to occur, or by analyzing the difference, a new correlation between the haze signal and the process device can be identified.

[0147] In addition, in each embodiment, an example was described in which a haze signal incident on the apertures (α3, α4, α1, α6) is used for process diagnosis as described above. However, haze signals incident on other apertures may also be used for process diagnosis. For example, a correlation with the state of the process device can be found with respect to the haze signal incident on the apertures (α2, α5, β2, β3, β5, β6) located to the left and right of the beam spot (BS), and the sum or difference signal of the haze signal incident on the apertures (α3, α4, α1, α6). Explanation of the symbols

[0148] 100… Process diagnostic device, A… Illumination optical system, Bc… Polarizing beam splitter, Bn(n=1, 2…)… Detection optical system, BS… Beam spot, Cn(n=1, 2…)… Sensor, D… Signal processing device, DB… Memory device, ST1… Sample stage, ST2… Scan device, W… Sample

Claims

Claim 1 A process diagnostic device for diagnosing the state of a semiconductor process device, comprising: a sample stand supporting a sample processed in the semiconductor process device; an illumination optical system that irradiates illumination light onto a sample placed on the sample stand; a plurality of detection optical systems that collect light from the sample and convert it into an electrical signal to output a detection signal; and a signal processing device that processes the detection signals of the plurality of detection optical systems. The signal processing device is characterized by scanning the sample to extract a haze signal of the sample, comparing the haze signal of the sample with a reference haze signal, and outputting an alarm when the difference indicating a relative change over time exceeds a set value. Claim 2 In claim 1, the signal processing device is a process diagnostic device that detects a change in the state of the semiconductor process device, which appears locally in the distribution of the entire surface of the sample as a microscopic surface shape of the sample, based on the difference between the haze signal and the reference haze signal. Claim 3 A process diagnostic device according to claim 1, wherein the signal processing device selects the haze signal of a detection optical system having the strongest correlation with the variation factor among the plurality of detection optical systems based on correlation data regarding the variation factor of a plurality of detection optical systems and the haze signal stored in advance in a memory device, and performs the comparison between the haze signal and the reference haze signal. Claim 4 In paragraph 3, the process diagnostic device wherein the plurality of detection optical systems are arranged such that their directions with respect to the beam spot of the illumination light are each different from one another. Claim 5 In claim 4, the plurality of detection optical systems comprises a polarization beam splitter that separates light according to the polarization direction and a plurality of sensors that each detect light with different polarization directions separated by the polarization beam splitter, and the correlation is data representing the relationship between the intensity of the haze signal, the polarization direction, and the variation factor, which are each set for the plurality of detection optical systems. Claim 6 In claim 1, the signal processing device creates a haze map, which is an intensity distribution of the haze signal for the sample, compares the haze map with an intensity distribution of the reference haze signal, and diagnoses the state of the semiconductor process device based on whether the difference between the haze signal and the reference haze signal exceeds a set value. Claim 7 In claim 1, the signal processing device is a process diagnostic device that accumulates data regarding the diagnosis of the semiconductor process device and performs machine learning, and diagnoses the state of the semiconductor process device based on a learned model obtained through machine learning. Claim 8 A method for determining plasma exchange timing within a chamber equipped with a semiconductor process device, wherein plasma treatment is performed on a sample installed in the chamber, and the defect signal and haze signal are repeatedly acquired on multiple days by a process diagnostic device that irradiates light onto the plasma-treated sample to extract a defect signal from a high-frequency component and a haze signal from a low-frequency component of the light from the sample, and the process diagnostic device determines the plasma state of the semiconductor process device based on whether the difference between the haze signal and a reference haze signal exceeds a set value, and outputs the result when the difference exceeds the set value, and a user determines the next plasma exchange timing based on the result output from the process diagnostic device and maintenance data of the semiconductor process device. Claim 9 A method for determining plasma exchange timing according to claim 8, wherein the process diagnostic device has a plurality of detection optical systems with different azimuth angles for a beam spot that irradiates the light, and among the plurality of detection optical systems, the detection optical system that has the largest variation among the multiple detection optical systems is selected for the haze signal, and the plasma state of the semiconductor process device is determined based on whether the haze signal of the detection optical system and the reference haze signal exceed a set value.

Images