A method and system for discriminating defects present on a frontside from defects present on a backside of a transparent substrate

Abstract

This invention relates to an apparatus and method for distinguishing between defects present on the front side and defects present on the back side of a substrate made of a material transparent at the detection wavelength. It includes placing the substrate in an inspection system comprising at least one light source coupled to an optical system, the substrate being positioned on a support and relative to the optical system such that a first beam and a second beam intersect at measurement points on the front side of the substrate, and controlling relative movement of the support and the optical system to scan measurement points along a measurement path on the front side of the substrate, the relative movement being controlled to keep a reference plane tangent to the measurement path. The method further includes identifying a first pattern in the signal corresponding to light scattered by particles on the back side of the substrate and exhibiting two intensity peaks spaced apart by a defined separation interval corresponding to the time required for a defect (P) to move over a distance (d) between the two illumination points (S1, S2).

Description

Technical Field

[0001] This invention relates to a system and method for optically inspecting the surface of a substrate, detecting potential defects such as particles, or more generally, describing the surface condition of the substrate. The substrate may be a wafer used to fabricate devices in the fields of microelectronics, optics, microsystems, or optoelectronics. Prior Technology

[0002] Substrates, such as those for electronic, optical, or optoelectronic applications, need to be inspected to detect, identify, and / or classify any defects that may be present on their surfaces. These defects can include impurity particles, crystal defects, scratches, or surface roughness.

[0003] This type of inspection is typically intended to provide qualitative or quantitative information, such as the location, size, and / or nature of defects. This information about the substrate surface condition can represent the quality of the substrate manufacturing process or the quality of the production steps using the substrate.

[0004] Documents WO2017167573, US2021215617A1, US20180231370, and WO0239099 disclose dark field systems for inspecting the front side of such substrates, i.e., the substrate surface to be processed in subsequent fabrication steps. These inspection systems are based on Laser Doppler Velocimetry (LDV) and include a light source emitting first and second beams, which are oriented relative to each other to form a measurement volume containing a plurality of parallel interference fringes at their intersection. The inspection system includes a support to receive the substrate and position it relative to the light source such that the measurement volume intersects with the front side of the substrate, defining a measurement spot on this front surface. The support is rotated so that the measurement spot is scanned along concentric measurement paths on the wafer front side. The scattered light is collected by a condenser lens and guided to a sensor, which receives the collected light and provides an electrical signal representing the intensity change of the collected light as the measurement point is scanned along the measurement path.

[0005] When a defect exists on the front side of the substrate, and this defect passes through the measurement point with a size smaller than the period of the interference fringes, it causes scattering of the Doppler pulse measured by the sensor. The Doppler pulse is a signal with two frequency components: a low-frequency component that forms the signal envelope, corresponding to the average light intensity scattered by the defect, and a high-frequency component corresponding to the Doppler frequency, which contains information about the velocity of the defect passing through the measurement point.

[0006] If a relatively large defect exists (its size is comparable to or larger than the period of the interference fringes), when this defect passes through the measurement point, it will cause pulse scattering without high-frequency components, or pulse scattering of high-frequency components with extremely low intensity.

[0007] When the wafer is made of a transparent material (at the detection wavelength of the two interfering beams), the two inspection beams also illuminate the back side of the substrate at their respective two illumination points. When a defect present on the back side passes through one of the illumination points, the light scattered back to the measurement point is also collected by the lens, and the sensor generates a pulse, although its intensity is small, similar to the pulse generated by a relatively large defect on the front side of the wafer.

[0008] When conventional inspection systems are used on wafers that are transparent at the detection wavelength, the light scattered by defects on the back side can contaminate the light scattered by defects on the front side, making it impossible to clearly determine the condition of the front side of the substrate.

[0009] The presence of high-frequency components in a pulse may help to associate the pulse with defects on the front side of the wafer, but a pulse without high-frequency components can be associated with defects on both the front and back sides. Summary of the Invention

[0010] The object of this invention is to at least partially solve the above-mentioned problems and to provide a system and method suitable for inspecting substrates made of a material transparent at the detection wavelength. More particularly, the object of this invention is to provide an inspection system and method for distinguishing between defects present on the front side of a substrate and defects present on the back side of a substrate made of a material transparent at the detection wavelength.

[0011] Therefore, the present invention relates to a method for distinguishing defects on the front side and defects on the back side of a substrate, the substrate being made of a material that is transparent at the detection wavelength, the method comprising: The substrate is placed in an inspection system comprising at least one light source coupled to an optical system for emitting first and second beams having detection wavelengths and being contained within a reference plane and intersecting. The substrate is placed on a support and positioned relative to the optical system such that the first and second beams intersect at a measurement point on the front side of the substrate, and also illuminate the back side of the substrate at two illumination points spaced apart. The relative movement of the support and the optical system is controlled to scan measurement points along the measurement path on the front of the substrate. The relative movement is controlled so that the reference plane remains tangent to the measurement path. Collect at least a portion of the light scattered by defects on the front and / or back of the substrate and establish a signal representing the intensity change of the light collected when scanning the measurement point along the measurement path.

[0012] The method further includes identifying a first pattern in the signal that corresponds to light scattered by a defect on the back of the substrate and exhibits two intensity peaks separated from each other by a defined separation interval, which corresponds to the time required for the defect to move over the distance between the two illumination points.

[0013] Further non-limiting features of the invention, whether alone or in any technically feasible combination: The method further includes filtering out a first pattern from the signal to prepare a front signal, which represents defects that exist only on the front side of the substrate; This method further includes preparing a backside signal, which represents defects that exist only on the back of the substrate; Identifying the first pattern involves using at least one of the following information to determine the separation interval: the intersection angle of the first and second beams in the substrate, the thickness of the substrate, and the velocity of the substrate relative to the beam group; Identifying the first pattern includes generating a reference signal or masking signal using information relative to the separation interval; This method further includes The signal is sampled at a given sampling rate to prepare raw measurement samples. Based on the intersection angle of the first and second beams, the thickness of the substrate, the relative moving speed, and the sampling rate, the separation interval is established to determine the number of samples. Furthermore, identifying the first pattern in the signal includes identifying two intensity peaks separated by a determined number of samples in at least some of the original measurement samples; The identification of the first pattern includes: The first step involves applying a transformation function to at least some of the original measurement samples to provide transformed samples. The transformation function is configured to map a first pattern in the original measurement samples to a second pattern of greater intensity in the transformed samples. The second step in detecting the second pattern in the converted sample; The second pattern includes a peak that indicates the location of the first pattern in the original measurement sample; The detection step further provides the average intensity of the two peaks of the first pattern and the background intensity; The substrate is disc-shaped with a symmetrical axis, and the relative motion of the support and optical system is a rotation around the symmetrical axis of the substrate.

[0014] According to another aspect, the present invention relates to an inspection system for distinguishing front and back defects of a substrate, the substrate being made of a material that is transparent at the detection wavelength. The system includes: At least one light source is coupled to an optical system for emitting first and second beams, the first and second beams having detection wavelengths and being contained within a reference plane and intersecting each other; A support member is used to receive the substrate and position the substrate relative to the optical system so that the first and second beams intersect at a measurement point on the front side of the substrate. The support member and the light source are movable relative to each other. A controller is used to control the relative movement of the support and the optical system to scan measurement points along the measurement path on the front of the substrate. The controller is configured to control the relative movement so that the reference plane remains tangent to the measurement path. A collector for collecting at least a portion of the light scattered by defects on the front and / or back of the substrate; The sensor associated with the collector is used to receive the collected light and provide a signal representing the intensity change of the collected light as the measurement points are scanned along the measurement path.

[0015] The inspection system also includes a processing unit connected to the sensor, which is configured to process signals to identify a first pattern. The first pattern corresponds to the light scattered by particles on the back of the substrate and exhibits two peaks, which are separated from each other by a defined separation interval, which corresponds to the time required for a defect to move over the distance between the two illumination points.

[0016] Further non-limiting features of this aspect of the invention, whether alone or in any technically feasible combination: The processing unit is also assembled as follows: The first pattern is filtered out from the signal to prepare the front signal, which represents defects that exist only on the front of the substrate, and / or; Prepare a backside signal, which represents defects that exist only on the back of the substrate; The processing device includes a converter to sample the signal at a given sampling rate and prepare raw measurement samples; The processing device includes an FPGA and / or a processor to identify the first pattern at a given sampling rate. Simple Explanation of the Diagram

[0017] Many other features and advantages of the present invention will become apparent from the following detailed description and with reference to the accompanying drawings, in which: FIG1 illustrates an inspection system according to an embodiment; FIG2 illustrates phenomena occurring on the front and back sides of a transparent substrate positioned on a support of the inspection system; FIG3a, 3b and 3c illustrate the benefits of the inspection system when used to inspect transparent substrates; FIG4a and 4b illustrate the application of conversion functions of signals provided by sensors of the inspection system; FIG5 illustrates a processing flow according to an embodiment of the present invention; FIG6a and 6b illustrate signals prepared for filtering from backside particle features. Implementation

[0018] [Inspection System]

[0019] Figure 1 illustrates an inspection system 1 according to one embodiment. The inspection system 1 includes at least one light source 2, for example, a laser source incident into an optical fiber. The light source 2 generates a main beam 3, preferably monochromatic, with a wavelength hereinafter referred to as the "detection wavelength," in the ultraviolet, visible, or infrared range. In this embodiment, the light source 2 is coupled to an optical system 2', which includes a beam splitter 4 that splits the light into two optical fiber paths 5, 5', such that the main beam 3 from the light source 2 generates two beams 6, 6'. The optical system may be an interferometric system. The beam splitter 4 and the optical fiber paths 5, 5' are configured such that the first and second beams 6, 6' are contained within a single plane called a "reference plane," or such a reference plane is defined by their optical axes. Two focusing elements 7, 7', such as collimating lenses, are respectively disposed in the propagation paths of the first beam 6 and the second beam 6', such that the two beams 6, 6' are oriented relative to each other to form a measurement volume at their intersection. In a preferred embodiment, the two beams 6, 6' are coherent and capable of interfering with each other, thereby causing the measuring volume to include a plurality of parallel interference fringes.

[0020] Other configurations of the light source 2 and the optical system 2' may differ from those presented in the embodiment of FIG1, which is naturally possible. Generally, in the inspection system 1 according to the invention, at least one light source 2 and optical system 2' are configured such that the first and second beams 6, 6' (preferably coherent beams) are contained in a reference plane and intersect in that plane.

[0021] Continuing with the embodiment shown in Figure 1, the inspection system 1 also includes a support member 8 to support the substrate S. The support member 8 is precisely positioned relative to the optical system 2' such that the intersection of the first and second beams, i.e., the measuring volume, intersects with the front surface of the substrate, and defines the measuring point M on this front surface.

[0022] The substrate S can be a wafer, i.e., a circular plate or disk of material having two opposing parallel surfaces, referred to as the back and front sides, respectively. Typically, the back side of the substrate corresponds to the surface that contacts or faces the support 8 of the inspection device 1. The wafer can be made of any material that is transparent at the detection wavelength, such as glass or synthetic sapphire, i.e., single-crystal alumina for visible light wavelengths, or semiconductor materials, such as silicon for infrared light detection wavelengths. However, the invention is not limited to wafer-shaped substrates; its principles can be applied to any substrate, especially substrates of any shape and construction that are transparent at the detection wavelength, as will be described in detail below.

[0023] The support 8 and the optical system 2' are movable relative to each other, and the inspection system 1 is designed such that the front side of the substrate S passes through the measuring body during movement. As is known in the art, the support 8 and / or the optical system can be coupled to an actuator capable of moving one of them relative to the other to achieve this relative movement.

[0024] The inspection system 1 includes a controller 12 for controlling the relative movement of the support 8 and the optical system 2' to scan measurement points M along a desired measurement path on the front side of the substrate S. The controller may include a microcontroller, data storage, input / output ports connected to the actuators and other components of the inspection system 1, and further computing resources, configured in hardware or software, to precisely control the displacement of the measurement points M along the desired measurement path. Advantageously, the controller 12 also controls the speed of the measurement points M along this path. The substrate S is precisely positioned on the support 8, and the controller 12 stores and controls the position (linear or polar coordinates) and speed (linear or rotational speed) of the measurement points M in a reference associated with the substrate S.

[0025] The inspection system 1 also includes a collector for collecting at least a portion of the light scattered by defects present on the front side of the substrate S. In the embodiment of FIG. 1, the collector includes a lens 9 for collecting a portion of the diffuse light radiation. In this embodiment, the lens 9 is centered on an axis passing through the measurement point M and perpendicular to the substrate S. A portion of the scattered light 12 reflected from the substrate S is collected by the lens 9 and guided to the optical conjugate point of the light spot M. The collector also includes a supplementary device 10, such as an optical fiber or fiber bundle, arranged at or near the conjugate point of the lens 9 to collect the scattered light. The inspection device 1 also includes a sensor 11, such as a photodetector or an array of photodetectors, which is connected to the collector to receive the collected light. When scanning the measurement point M along the measurement path, the sensor 11 provides a signal U(t), such as an electrical signal, representing the intensity change of the collected light.

[0026] Finally, the inspection system 1 shown in Figure 1 includes a processing unit 13 connected to the sensor 11 to receive a signal U(t) during measurement. The processing unit 13 is also connected to the controller 12 to receive the position XM(t) of the measurement point M from the controller during measurement, and may also receive the velocity VM(t) if this information is measured or determined by the controller 12. The velocity VM(t) may also be derived from the position XM(t), or may be a constant value, so that this information is not provided to the processing unit 13 via the controller 12. The processing unit may correspond to or include a microprocessor, microcontroller, FPGA, or any other computing resource (input / output ports, data storage, etc.) for processing the received information, especially the signal U(t), to detect particles present on the back and / or front of the substrate and to distinguish between the two, as will be explained below. This processing can be performed quickly simultaneously with the measurement, or it can be deferred until after the measurement is completed.

[0027] The processing device 13 (or sensor 11) may include a converter to sample the signal U(t) at a given sampling rate and prepare the raw measurement sample Uk.

[0028] Figure 2 illustrates the phenomena occurring at the front FS and back BS of the substrate S, located on the support 8 of the inspection system 1, when the substrate S is made of a material that is transparent at the detection wavelength. Figure 2 is a cross-sectional view of the substrate S in a reference plane containing the first and second beams 6, 6'. As previously described, the first and second beams 6, 6' interfere in the measuring volume V, which intersects with the front FS of the substrate S to form a measuring point M on the front FS. The substrate S is made of a material that is transparent at the detection wavelength of the beams 6, 6', which propagate into the thickness Th of the substrate and also illuminate the back BS of the substrate S at two separate illumination points S1, S2. The two illumination points S1, S2 are separated by a distance d, which depends on the thickness Th of the substrate and the angle f of the beams propagating into the substrate S. Therefore, the two beams 6, 6' are separated from each other on the back BS and illuminate the back BS at two separate illumination points S1, S2. The angle f depends on the angle of incidence, or the angle α between the first and second beams 6, 6' and the substrate surface, and the substrate refractive index which follows the Snell-Descartes law. If the two beams 6, 6' are at a certain angle f relative to each other inside the substrate S, then the distance d can be geometrically determined as d = 2*Th*tan(f / 2).

[0029] According to an important aspect, the controller 12 is configured to control the relative movement of the support 8 and the optical system 2', such that the reference plane remains tangent to the measurement path. This condition means that, because the point M is scanned along the measurement path, the velocity vector of the measurement point M relative to the substrate is contained in or parallel to the reference plane. Therefore, the velocity vector extends in the longitudinal direction x in Figures 1 and 2, without a component along the lateral direction y. It should be understood that the velocity vector should also be parallel to the substrate surface, or at least should not have a significant component along the height direction z, to ensure that the measured volume V remains intersecting the front surface FS of the substrate S. Therefore, when the velocity vector is located in or parallel to the reference plane, the velocity VM(t) of the measurement point M corresponds to the velocity of that point along the longitudinal direction x.

[0030] The relative motion operated by controller 12, tangential to the measurement path, can be rotational, or, for example, rotation of the support 8 (and substrate S) along a rotation axis r perpendicular to the plane of the support 8 / substrate S and positioned at the center of the substrate S. The measurement path, then described by the measurement point M, forms concentric circles on the front surface of the substrate S. If desired, the relative motion can be linear, along the longitudinal direction x, and the measurement path then forms a straight line on the front surface of the substrate S. After the support 8 undergoes relative displacement with respect to the optical system 2', the inspection system can be repeatedly operated to provide multiple measurement paths along parallel, radial, or concentric measurement paths to provide a complete inspection of the front surface of the substrate S.

[0031] Figures 3a and 3c illustrate the benefits of inspection system 1 for inspecting a substrate S made of a transparent material at the detection wavelength. The substrate S has particles P on its back side. In Figure 3a, at time t1 when the measurement point M scans the front side of the substrate S, particles P are precisely positioned at a first illumination point S1 formed on the back side by the propagation of the first beam 6. This beam is scattered by particles P, and a portion of the scattered light diffuses toward the measurement point M. This portion of the scattered light is collected and detected by the inspection system as if it were generated at the measurement point M itself. Therefore, the signal U(t) prepared by sensor 11 exhibits a first intensity peak at the first time t1, as shown in Figure 3c.

[0032] Figure 3b shows the substrate S at time t2, shortly after time t1 shown in Figure 3a. At this time t2, particle P is positioned at the second illumination point S2 formed on the back side by the propagation of the second beam 6'. This beam is scattered by particle P, and part of the scattered light diffuses toward the measurement point M. This part of the scattered light is collected and detected by the inspection system as if it were generated at the measurement point M itself. Therefore, the signal U(t) prepared by sensor 11 exhibits a second intensity peak at time t2, as shown in Figure 3c.

[0033] The relative motion is controlled such that the reference plane is tangent to the measurement path. A particle P appearing on the back side, passing through one of the illumination points S1 and S2, must also pass through the other of S2 and S1. Therefore, the light scattered by particle P on the back side of the substrate produces a first pattern in the signal U(t), exhibiting two intensity peaks U(t1) and U(t2), separated by a separation interval. This interval can be associated with a defined time period T = (t2-t1), corresponding to the time required for particle P to move over the distance d separating the two illumination points S1 and S2, and can be obtained by dividing this distance d by the velocity VM of the measurement point, i.e., T = d / VM.

[0034] When the signal provided by sensor 11 is expressed as a function of the position of the measurement point on the substrate or the measurement path, the separation interval corresponds to the distance d.

[0035] [Methods for distinguishing defects]

[0036] This invention proposes a method for distinguishing between defects present on the front side of a substrate S and defects present on the back side of a substrate S made of a material that is transparent at the detection wavelength, based on observations obtained from the aforementioned inspection system configuration.

[0037] The method includes placing a substrate S on a support 8 of an inspection system 1, the substrate S being positioned relative to an optical system 2' such that a first beam 6 and a second beam 6' intersect on the front surface of the substrate S. The method then includes moving the support and the optical system 2' relative to each other to scan measurement points M along a predetermined measurement path on the front surface of the substrate S, the relative movement being controlled by a controller 12 such that a reference plane remains tangent to the measurement path. When scanning measurement points M along the front surface of the substrate, at least a portion of the light scattered by the two surfaces of the substrate S, particularly light scattered by defects present on the front and / or back surfaces of the substrate S, is collected by a sensor 11, establishing a signal U(t) representing the change in the intensity of the collected light.

[0038] According to the invention, the method further includes identifying a first pattern in the signal U(t), the first pattern exhibiting two intensity peaks separated from each other by a defined separation interval. This has been illustrated with reference to Figures 3a to 3c, the first pattern corresponding to the light scattered by particles on the back side of the substrate S. Therefore, if such a first pattern exists in the signal U(t), it can be clearly determined that particle P exists on the back side of the substrate. The position of particle P (its coordinates in a plane parallel to the back side of the substrate) corresponds to the position of the measurement point M at the average time (t1+t2) / 2, i.e., at half the separation distance between the two peaks.

[0039] Conversely, the intensity peaks in signal U(t) that do not conform to the first pattern correspond to particles present on the front side of the substrate FS.

[0040] Since defects existing on the front and back sides of the substrate can be distinguished, the inspection device 1 can be configured to prepare a front signal Ufs(t), which represents defects existing only on the front side of the substrate, for example, by filtering out a first pattern from the electrical signal U(t). Additionally or alternatively, the inspection device 1 can be configured to prepare a back signal Ubs(t), which represents defects existing only on the back side of the substrate S.

[0041] Identifying the first pattern in the signal may require calibrating this pattern against the substrate being examined to determine the separation interval that separates the two intensity peaks. As previously stated, if the scanning speed remains constant or is known, this separation interval can be pre-calculated based on the known thickness of the substrate and the angular configuration of the first and second beams. More generally, the separation interval can be determined using at least one of the following information: the incident angles of the first and second beams on the substrate, the refractive index of the substrate, the intersection angle of the first and second beams on the substrate, the thickness of the substrate, and the velocity of the substrate relative to the beams.

[0042] The calibrated first pattern can be stored in the inspection device 1, for example in the processing device 13.

[0043] As an alternative, a set of first patterns, or some features of the first patterns, such as distance or time interval, each corresponding to a substrate, can be stored in the inspection device and read according to the properties of the substrate being tested before the inspection begins.

[0044] If the speed of the measurement points in the measurement path is not kept constant, the separation interval of the first pattern can be calculated or adjusted to the actual speed at each given moment, and the first pattern is dynamically calibrated.

[0045] When the signal is sampled by the converter of the processing device 13 to establish the original measurement sample Uk, the separation interval can be represented as a certain number of samples 2s, and identifying the first pattern in the signal includes identifying two intensity peaks, which are separated by the number of samples 2s determined in the continuous original measurement samples.

[0046] The identification of the first pattern can be performed by the processing device 13 in many different ways. For example, the processing device can implement any kind of pattern recognition algorithm. For example, a correlation method can be implemented to associate the signal with the template pattern signal. It can also be implemented by a pattern recognition algorithm trained by machine learning, in which an indicator signal is set if the first pattern is identified in the signal as measurement is performed.

[0047] Alternatively, the intensity peak can be established in the original measurement sample Uk (e.g., through numerical derivation). If the difference between the indices i and j corresponding to two consecutive peaks Ui and Uj is a predetermined number of samples, i.e., ji = 2s, then the first pattern can be identified in the original measurement sample Uk. This operation can be performed quickly and simultaneously when the original measurement sample is established by the processing device 13. The original information sample can be stored in the data storage included in (or associated with) the processing device 13, in which case the identification of the first pattern may be delayed until after the inspection is completed.

[0048] Whether identification is performed simultaneously or in a delayed manner, it advantageously provides the position of the first pattern in the original measurement sample, i.e., the index i+s or js. By combining this index information with the measurement point M position XM (received from controller 12) as the measurement progress, it is possible to precisely locate particle P on the back side of the wafer at the origin of the identified first pattern. The particle position can be stored as part of the inspection process and displayed as a substrate map, or processed in other ways at the end of the inspection process.

[0049] [Identification is performed using conversion functions]

[0050] The intensity peak of the electrical signal corresponding to the back particle P may have a relatively low amplitude. This is because only a portion of the light scattered by this particle P is guided to the measurement point M and collected and detected.

[0051] Therefore, it is advantageous in that the identification of the first pattern includes a first step of applying a transformation function to a signal U provided by sensor 11. The transformation function is selected to map this signal U to a transformed signal V, and the first pattern present in the original signal U is transformed into a second pattern with greater intensity, or more generally, greater detectability, in the transformation space. The identification then includes a second step of detecting the second pattern in the transformed signal V.

[0052] To illustrate this principle, and according to a specific embodiment, the conversion can be implemented using the following function (when processing raw digital measurement samples):

[0053] For ease of execution by processing device 13, the function can also be expressed as:

[0054] Wk represents a mask consisting of 6<s+1 elements, where only 4 are non-zero, and the remaining 6<s-3 elements are zero.

[0055] Figure 4a illustrates the behavior of the transformed sample Vk when the original sample Uk contains a first pattern corresponding to the backside particles. It contains a second pattern exhibiting a single peak at the midpoint between the two peaks of the first pattern; the single peak of the second pattern is precisely located at the backside defect. The amplitude A of this peak is twice the amplitude A / 2 of the peak of the first pattern in the original measured sample Uk. The transformed sample also contains negative values, which are not shown in the figure; negative values ​​are irrelevant and are simply set to zero.

[0056] As shown in Figure 4b, the transformed sample Vk exhibits a very different single-peak behavior in the signal generated by the frontal particle. In this case, the transformed sample presents two peaks, separated from each other by a length interval of 2 s. Specifically, if the frontal particle generates a peak centered at time k, the corresponding profile of the transformed sample Vk has two peaks at Vk+s and Vks. The amplitudes of the two peaks in the transformed signal are the same as the amplitudes of the peaks in the original signal.

[0057] It should be noted that the signal shown in Figure 4b is for a particle large enough to average out the high-frequency signal. If a smaller particle passes through the interference fringes of two interfering beams, it will lead to different results. Alternatively, the signal could be generated by any particle passing through two beams of different moduli, thus causing no interference.

[0058] In summary, the converted signal V or sample Vk according to this embodiment has the following characteristics: It is zero in the region of constant intensity; It is independent of the background signal level, i.e., the average signal value when there are no particles (sometimes referred to in the technical field as "haze"). It has a pulse at the back of the particles; It has two independent pulses around the frontal particles.

[0059] Transformation functions with these characteristics make the transformed signal or sample Vk extremely useful for detecting the location of back-side particles and for distinguishing between back-side and front-side particles. This transformed signal can be used alone. It can also be combined with the original measurement signal Uk to more easily distinguish between back-side and front-side particles by detecting, for example, a single peak in the original measurement signal that matches a double peak in the transformed signal and corresponds to a front-side particle, and vice versa.

[0060] Referring to FIG5, according to this preferred embodiment, the identification of the first pattern includes the application of a transformation function, which may include the following steps performed by a processing device.

[0061] First, in step 20, the transformed sample Vk is calculated for all original samples Uk, either in-situ as described above or with a delay. These samples may be affected by noise; to eliminate noise as needed, only the transformed samples Vk exceeding a first specified threshold can be retained for further analysis. Then, in step 22, the maximum values ​​are identified within the transformed samples. This can be done by derivation or by selecting samples exceeding a second specified threshold. In step 24, maximum values ​​with similar amplitudes and separated by 2s intervals are ignored because they correspond to front-facing particles. This detection can be enhanced by verifying that these paired maximum values ​​correspond to peaks in the original measurement signal. The remaining transformed samples Vk contain only single peaks, each attributed to a back-facing particle, thus revealing the location of these particles.

[0062] In step 26, the position of the back particles is used to process the original signal or sample to prepare a front signal and / or a back signal, wherein the front signal represents defects that exist only on the front of the substrate, and the back signal represents defects that exist only on the back of the substrate.

[0063] [Remove back-side particles from sensor signals].

[0064] According to one embodiment, the location of the backside particles can be used to process the raw signal or sample to prepare a frontside signal representing defects present only on the front side of the substrate. It should be understood that a similar method can be used to prepare a backside signal representing defects present only on the back side of the substrate.

[0065] As a preliminary explanation, the signal provided by sensor 11 (or the original measurement sample Uk) is the sum of the background portion b and the particle scattering portion. As mentioned above, the amplitude of the second pattern peak (A) is twice the amplitude of the corresponding peak in the first pattern (A / 2). The intensity distribution in the laser beam is perfectly approximated by a Gaussian distribution with a known standard deviation σ, hereinafter represented by the number of samples. For this reason, the bimodal signal generated for the number of samples i by the back-side particle located at the position of the number of samples m is simulated as follows: Where s is half of the separation interval, expressed as the number of samples.

[0066] According to this embodiment, a background level is determined for the original signal or original measurement sample. The method includes determining the amplitude A of a peak in a second pattern identified in the transformed sample, and subtracting half of this amplitude, A / 2, from the peak amplitude in the original measurement sample to establish a background level b. This effectively provides two background level estimates, each corresponding to a peak. The estimated background level can be selected to correspond to the average value: Where m corresponds to the number of samples at the detected background particle locations.

[0067] Then, based on the above model UBS, and based on the determined amplitude level A / 2, particle position m and background level b, the reference signal U'(i) of the light scattered by the back particle can be prepared.

[0068] The final step is to subtract the reference signal from the signal provided by the sensor. This operation removes the back-side particle features, leaving only the front-side particle features that were not captured in the signal.

[0069] Figures 6a and 6b illustrate this embodiment. Figure 6a illustrates the noise signal from the sensor, which has a single peak (front particle) and a double peak (back particle).

[0070] Figure 6b illustrates the signal after filtering out the features of the back particles, while the contributing factors from the front particles remain unchanged.

[0071] Alternatively, once the bimodal peaks corresponding to a backside particle are identified, any curve-fitting or least-squares method can be used to fit the mode function, such as those described above, to the original measurement signal and subtract this signal, retaining only the background noise.

[0072] By studying the drawings, specification and appended claims of this case, those skilled in the art will be able to understand and implement other variations of the disclosed embodiments when implementing the invention claimed in this case.

[0073] 1: Inspection System 2: Light source 2': Optical System 3: Main beam 4: Spectrometer 5,5': Fiber optic path 6: First beam 6': Second beam 7,7': Concentrating element 8: Support components 9: Collector / Lens 10: Supplementary device 11: Sensor 12: Controller 13: Processing device BS: Back d: distance f: Intersection angle FS: Front M: Measurement point P: Defects / Particles r: axis of rotation S: Substrate S1, S2: Illumination points t1, t2: Time points Th: thickness U(t): Signal UBS(t): Backside signal UFS(t): Positive signal U k: Original measurement sample V: Measuring volume Vk: Transformed sample 2s: Determine the sample size

[0074]

Claims

1. A method for distinguishing front and back defects of a substrate (S), the substrate (S) being made of a material that is transparent at a detection wavelength, the method comprising: The substrate (S) is placed in an inspection system (1), which includes at least one light source (2) coupled to an optical system (2') for emitting first and second beams (6, 6') having the detection wavelength and contained within a reference plane and intersecting. The substrate (S) is placed on a support (8) and positioned relative to the optical system (2') such that the first and second beams (6, 6') intersect at a measurement point (M) on the front side of the substrate (S) and each illuminates the back side (BS) of the substrate (S) at two illumination points (S1, S2) spaced apart by a distance (d). The relative movement of the support (8) and the optical system (2') is controlled to scan the measurement point (M) along a measurement path on the front side of the substrate (S), and the relative movement is controlled to keep the reference plane tangent to the measurement path. The method collects at least a portion of the light scattered by defects on the front and / or back of the substrate (S) and establishes a signal (U) representing the intensity change of the light collected when scanning the measurement point (M) along the measurement path; the method further includes identifying a first pattern in the signal (U) corresponding to the light scattered by a defect (P) on the back of the substrate (S) and exhibiting two intensity peaks separated from each other by a defined separation interval corresponding to the time required for the defect (P) to move over a distance (d) between the two illumination points (S1, S2).

2. The method of claim 1 further includes filtering the first pattern from the signal (U) to prepare a front signal (UFS) representing a defect that exists only on the front side of the substrate (S).

3. The method of claim 1 or 2 further includes preparing a backside signal (UBS) representing a defect that exists only on the backside of the substrate (S).

4. The method of claim 1, wherein identifying the first pattern includes determining the separation interval using at least one of the following information: the intersection angle (f) of the first and second beams (6, 6') in the substrate (S), the thickness (Th) of the substrate (S), and the speed of movement of the substrate (S) relative to the first and second beams.

5. The method of claim 1, wherein identifying the first pattern includes generating a reference signal or a masking signal using information relative to the separation interval.

6. The method of request item 1 further includes: The signal is sampled at a given sampling rate to prepare raw measurement samples; the separation interval is established as a defined sample number (2s) based on the intersection angle (f) of the first and second beams (6, 6'), the thickness (Th) of the substrate (S), a relative moving speed and the sampling rate; and wherein identifying the first pattern in the signal (U) includes identifying two intensity peaks separated by the defined sample number (2s) in at least some of the raw measurement samples.

7. The method of claim 6, wherein identifying the first pattern includes: A first step involves applying a transformation function to at least some of the original measurement samples to provide a transformed sample, the transformation function being configured to map a first pattern in the original measurement samples to a second pattern of greater intensity in the transformed sample; and a second step involves detecting the second pattern in the transformed sample.

8. The method of claim 7, wherein the second pattern includes a peak that provides the location of the first pattern in the original measurement sample.

9. The method of claim 7 or 8, wherein the detection step further provides the average intensity of the two peaks of the first pattern and a background intensity.

10. The method of claim 1, wherein the substrate (S) is disk-shaped and has an axis of symmetry, and the relative motion of the support (8) and the optical system (2') is a rotation about the axis of symmetry of the substrate (S).

11. An inspection system (1) for distinguishing front and back defects of a substrate (S), the substrate (S) being made of a material that is transparent at a detection wavelength, the system comprising: At least one light source (2) coupled to an optical system (2') for emitting first and second beams (6, 6') having the detection wavelength and contained within a reference plane and intersecting; a support (8) for receiving the substrate (S) and positioning the substrate (S) relative to the optical system (2') such that the first and second beams (6, 6') intersect at a measurement point (M) on the front side of the substrate (S) and illuminate the back side (BS) of the substrate (S) at two illumination points (S1, S2) spaced apart by a distance (d), the support (8) and the optical system (2') being movable relative to each other; a controller (12) for controlling the relative movement of the support (8) and the optical system (2') to scan the measurement point (M) along a measurement path on the front side of the substrate (S), the controller (12) being configured to control the relative movement so that the reference plane remains tangent to the measurement path; A collector (9) for collecting at least a portion of the light scattered by a front and / or back defect of the substrate (S); a sensor (11) associated with the collector (9) for receiving the collected light and providing a signal (U) representing the intensity change of the collected light as the measurement point (M) is scanned along the measurement path; the inspection system also includes a processing device (13) connected to the sensor (11), the processing device (13) being configured to process the signal (U) to identify a first pattern corresponding to the light scattered by a defect (P) on the back of the substrate (S) and exhibiting two intensity peaks separated from each other by a defined separation interval corresponding to the time required for the defect (P) to move over a distance (d) between the two illumination points (S1, S2).

12. The inspection system of claim 11, wherein the processing device (13) is also configured to: filter the first pattern from the signal (U) to prepare a front signal representing a defect present only on the front of the substrate (S), and / or; prepare a back signal representing a defect present only on the back of the substrate (S).

13. The inspection system of claim 11 or 12, wherein the processing device (13) includes a converter for sampling the signal (U) at a given sampling rate and preparing a raw measurement sample.

14. The inspection system of claim 13, wherein the processing device (13) includes an FPGA and / or a processor to identify the first pattern at the given sampling rate.

Images