JP4751298B2 - Height inspection method and height inspection apparatus for implementing the method - Google Patents

Description

  The present invention relates to a height inspection method and a height inspection apparatus for performing the method, and more particularly, a method for accurately and efficiently inspecting the height of a bump formed on a surface of a semiconductor chip and a height for implementing the method. The present invention relates to an inspection apparatus.

  A plurality of bump terminals as input / output terminals are arranged in a matrix on the surface of a semiconductor chip, and the semiconductor chip is bonded face-down to a substrate. Such a semiconductor chip is called a flip chip and is used for a multi-chip module or the like.

  This bump terminal is a spherical terminal usually formed on an electrode formed on the surface of a semiconductor substrate, and includes a solder bump formed by a solder dipping method and a metal bump formed by a plating method.

  A chip on which a plurality of bump terminals are formed is bonded face down on the substrate. In this case, each bump terminal must be normally connected to a corresponding electrode on the substrate. Therefore, it is necessary that the bumps have a uniform shape and a uniform height. If the height is lower than the other bumps, it may cause a connection failure with the electrode of the substrate, and if the height is high, the shape of the bump may be large, and the bump may be short-circuited with the adjacent bump in the bonding process.

Therefore, it is common to inspect the height of the bumps on the surface before bonding the flip chip. A general method of height inspection is a triangulation method in which light is irradiated and the reflected light is imaged on a two-part photodiode or PSD (Position Sensitive Light Detector) to detect the height. Therefore, the surface is scanned with laser light, and the height of each bump is inspected according to the position of the reflected light.
JP-A-10-009825 Japanese Patent Laid-Open No. 05-209719 JP 08-159719 A JP 08-159721 A

  However, when a spherical bump is scanned with a laser beam, the reflected light is greatly diffused in the vicinity of the bump, reflected at the top of the bump with a very high reflectance, and the reflectance outside the bump area is relatively constant. It is. Thus, since the reflectance of the surface changes greatly, the intensity of the reflected light greatly changes if the intensity of the irradiated light is made constant. For this reason, there is a problem that the amount of reflected light deviates from the dynamic range of the photodiode or PSD, and the correct height cannot be measured.

  In order to measure the height of a large number of bumps on the surface of the semiconductor chip, it is effective to directly measure the height from the position where the reflected light forms an image on the position detection element (PSD) by the triangulation method. is there. However, since the direction of the reflected light is not constant depending on the reflection position of the incident light on the bump due to the curved surface shape of the bump, an error occurs in the height detected by the position detection element. Further, an error occurs in the detected height due to the out-of-focus state of the optical system, and the direction of the error differs depending on whether the focus is the front focus or the back focus. Further, the presence of a deviation in the optical axis of the optical system that receives the reflected light causes the detected bump apex position to be displaced, and the detected height includes a deviation.

  As another problem, an acousto-optic deflector is used to scan an LSI chip having a large number of bumps with a laser beam at high speed. A frequency fluctuation signal supplied to the acousto-optic deflector is generated by, for example, a voltage oscillation circuit. However, due to the astigmatism problem of this acousto-optic deflector, the problem of dynamic astigmatism change due to the distorted output characteristics of the voltage oscillation circuit, and the tolerance of the optical system, etc. It is very difficult to scan the measurement object with astigmatism zero and at a constant speed.

  The above problems are all problems that arise when detecting the height of a large number of bumps on an LSI chip, and it is necessary for accurate height inspection to solve all of these problems.

  Accordingly, an object of the present invention is to provide a method and an inspection apparatus for inspecting the height of bump terminals formed on the surface of a semiconductor chip with high accuracy and efficiency.

  Another object of the present invention is to provide a method and an inspection apparatus for inspecting the height of a sample surface having regions with greatly different reflectivities, without being limited to the bumps of a semiconductor chip.

  Another object of the present invention is to provide a height inspection method and an inspection apparatus capable of correcting a detected height error caused by the curved shape of the bump.

  Another object of the present invention is to provide a height inspection method and an inspection apparatus capable of eliminating astigmatism, defocusing and scanning speed unevenness of scanning light due to the characteristics of an acousto-optic deflector and a voltage oscillation circuit for driving the acousto-optic deflector. Is to provide.

  According to the present invention, the above object is achieved by scanning the surface of an object to be inspected having regions with different reflectivities with a predetermined intensity and causing the reflected light from the surface to form an image on the light spot position detecting means. In the height inspection method for detecting the height of the surface by detecting the imaged light spot position, in the second region surrounding the first region having a high reflectance and having a low reflectance around the first region, Scanning with the first incident light intensity low enough not to saturate the light spot position detecting means even with the high reflectance, and detecting that the amount of imaged light of the reflected light exceeds a predetermined threshold It is achieved by providing a height inspection method comprising a step of scanning with an appropriate second incident light intensity obtained from the detected light quantity and the incident light intensity at the previous scanning. .

Further, according to the present invention, the second incident light intensity Pn + 1 (i) is calculated from the incident light intensity Pn (i) at the previous scanning and the detected light amount Bn (i) at the time, using the formula: Pn + 1 (i) = Pn (i) × Bt / Bn (i)
(Bt is obtained by calculation according to the predetermined value within the dynamic range of the light spot position inspection means).

  Furthermore, according to the present invention, scanning is performed with the incident light intensity set to zero in a region other than the first and second regions.

  Furthermore, according to the present invention, the threshold value is set to a value higher than a detected light amount with respect to the first incident light intensity in the second region.

  Further, according to the present invention, the object to be inspected has a bump terminal that is provided on the surface of the substrate and has a substantially spherical shape, and the first region is near the top of the bump terminal, The area is a surrounding area surrounding the vicinity of the top of the bump.

  In detecting the height of the bump terminal of the semiconductor chip, the height inspection method of the present invention is particularly effective.

  Further, according to the present invention, according to the present invention, incident light having a predetermined intensity is scanned on the surface of an object to be inspected having regions having different reflectivities, and the reflected light from the surface is connected to the light spot position detecting means. In a height inspection apparatus that detects the height of the surface by detecting the position of the focused light spot, the light emitting means for generating the incident light, the light emission driving means for driving the light emitting means, In the second region surrounding the first region having a high reflectance and having a low reflectance around the first region, the first incident light intensity is so low that the light spot position detecting means is not saturated even at the high reflectance. When the light emission driving means is controlled to detect that the light amount of the imaged light of the reflected light exceeds a predetermined threshold value, an appropriate second value obtained from the detected light amount and the incident light intensity at the previous scanning is obtained. Modulation means for controlling the light emission driving means so as to obtain an incident light intensity; Characterized in that it.

Further, according to the present invention, the modulation means calculates the second incident light intensity Pn + 1 (i) from the incident light intensity Pn (i) at the previous scanning and the detected light quantity Bn (i) at that time. , Pn + 1 (i) = Pn (i) × Bt / Bn (i)
(Bt is obtained by calculation according to the predetermined value within the dynamic range of the light spot position inspection means).

  In order to achieve the above object, the present invention irradiates the surface of an inspection object having a non-planar surface with incident light, forms an image of reflected light from the surface on a light spot position detecting means, In the height inspection method for detecting the height of the surface from the imaged light spot position, the height of the surface of the reference inspection object is first set by moving a stage on which the reference inspection object is mounted in the Z-axis direction. A first height detecting step for detecting the correction height value, and irradiating the surface of the reference inspection object with the incident light, and reflecting light from the surface to the light spot position detecting means. A second correction height detecting step of forming an image and detecting the height of the surface of the reference inspection object as a second correction height value from the imaged light spot position; and The surface is irradiated with the irradiation light, the reflected light from the surface is imaged on the light spot position detecting means, and the result is A measuring height detecting step for detecting the height of the surface of the object to be inspected as a measuring height value from the light spot position, and the measuring height value as the first correcting height value and the second correcting height value. And a correction step of generating a corrected height value by performing correction according to the correspondence relationship with the correction height value.

  According to the above-described invention, an error in the detected height value that occurs in the case of a measurement object having a curved surface such as a bump on an LSI chip is appropriately corrected. The first correction height detection process for obtaining the correction function or table includes a time-consuming process such as moving the stage in the Z-axis direction. After the correction function is set, the light spot position detection is performed. The height of a large number of bumps is detected by high-speed height measurement using an element.

  In order to achieve the above object, the present invention deflects incident light by an acousto-optic deflector, and scans the deflected incident light on the surface of an inspection object through a cylindrical lens and an incident optical system. In the height inspection method in which reflected light from the surface is imaged on a light spot position detecting means and the height of the surface is detected from the imaged light spot position, a scanning frequency signal is sent to the acousto-optic deflector. By scanning the incident light on a setting sample having a predetermined pattern on the stage on which the inspection object is placed, and observing the amount of reflected light while changing the input signal of the drive circuit that gives A setting step for obtaining a sweep curve of the input signal so that the incident light is focused on the setting sample without astigmatism and further scanned at a constant speed, and the input signal is driven according to the sweep curve. The incident light is scanned on the surface of the object to be measured, and the reflected light from the surface is imaged on the light spot position detecting means, and the height of the surface is measured from the imaged light spot position. And a height detection step for detecting the above.

  According to the above invention, the problem of astigmatism due to the cylindrical effect when using an acoustooptic deflector that enables high-speed scanning, the characteristic distortion of the circuit that drives the acoustooptic deflector, and the optical system It can solve problems caused by tolerances. The sweep curve of the input signal to the drive circuit can be obtained by scanning the sample with incident light and monitoring the amount of reflected light.

  Hereinafter, an example of an embodiment of the present invention will be described with reference to the drawings. However, such an embodiment does not limit the technical scope of the present invention.

  FIG. 1 is a perspective view showing bump terminals formed on the surface of a semiconductor chip. A plurality of bump terminals 2 are arranged in a matrix on the surface of the chip 1.

  FIG. 2 is a partial cross-sectional view of the semiconductor chip 1 on which the bump terminals are formed. A spherical bump terminal 2 is formed on the electrode 20 formed on the surface of the semiconductor chip 1. In the example shown in FIG. 2, the bumps B1, B3, and B5 have a normal size and are formed at a normal height h, but the bump B2 has a small shape and a normal height h. The bump B4 has a larger shape and a height higher than the normal value h. In general, since the height and size of the bumps have a relative relationship, whether or not bonding failure occurs is determined by inspecting whether or not the height of the bump is formed to a normal value h in advance. be able to.

  FIG. 3 is a plan view for explaining a method of detecting the height of the bump by scanning the bump with laser light. FIG. 4 is a side view thereof. In this example, in FIG. 3, the irradiation light 22 is irradiated from the front side toward the drawing, and the reflected light 24 reflected by the bumps and the chip surface is directed upward. And irradiation light is scanned in an order from the upper side of FIG.

  With reference to FIG. 4, the state of the irradiation light 22 and the reflected light 24 in the vicinity of the bump 2 can be understood. FIG. 4 shows scanning 1 to scanning 9. In the scan 1, the bump 2 is not irradiated but is irradiated and reflected on the surface of the chip 1. In the scan 2, in this example, the irradiation light 22 is irradiated on the top of the bump 2, and the reflected light 24 passes through the objective lens 26 of the inspection apparatus and enters the PSD. As shown by the broken line, when the bump 2 is low, the reflected light 24 is imaged at different positions on the PSD as the light spot position detecting means, and the bump 2 Height is detected.

  Furthermore, in the scans 3 to 8, the reflected light is irregularly reflected due to the shape of the sphere of the bump 2 and does not enter the objective lens 26, and PSD can be detected only by extremely dark light. Similarly, when scanning is performed in an area between scans 1 and 2 in FIG. 4, only dark light can be detected by the PSD as shown in the figure.

  The disturbance of the incident light position and the reflected light as described above is caused by the special shape of the bump 2, but it occurs not only in the surface of the semiconductor chip but also in the same special shape.

[Method of modulating incident light intensity]
FIG. 5 is a graph showing the relationship between the intensity of irradiated light and the amount of light detected by the PSD when the above-described bump 2 and the like are scanned. FIG. 5A shows a case where the intensity of irradiation light is constant with respect to the scanning position, and FIG. 5B shows an example of the detected light amount at that time. Both horizontal axes are time i. That is, in the example of FIG. 4, the amount of light detected between scans 1 and 2 is extremely small as shown in FIG. 4, an extremely large amount of light is detected in scan 2, and extremely large in scans 3 to 8. Only a small amount of light is detected. Alternatively, in scanning 2, the amount of light detected is small at portions other than the bumps and increases at the bump apexes. Therefore, the detected light amount varies greatly as shown in FIG. Bt in the figure is, for example, the median value of the dynamic range of a photoelectric conversion element such as a PSD. In order to operate with high accuracy, it is desirable that the detected light amount is located near the median value Bt, and at least the detected light amount is It must be within the dynamic range.

  FIG. 5C shows an example in which the intensity of the irradiated light is changed in response to the fluctuation of the detected light amount, and FIG. 5D shows the fluctuation of the detected light amount in that case. As described above, the detected light quantity is maintained in the vicinity of the median value Bt, thereby enabling highly accurate position detection.

In order to maintain the detected light amount as shown in FIGS. 5C and 5D, the irradiation light is generally modulated. That is, based on the detected light amount Bn (i) detected by the scan n, the detected light amount at the next scan n + 1 is Bt under the assumption that the reflected light is detected with the same reflectance in the next scan n + 1. In this modulation method, the intensity Pn + 1 (i) of the irradiation light is obtained so as to be a value. That is, the next irradiation light intensity Pn + 1 (i) is obtained by the following equation.
Pn + 1 (i) = Pn (i) × Bt / Bn (i)
That is, the irradiation intensity Pn + 1 (i) at the next scanning is obtained from the relationship between the irradiation intensity Pn (i) at the previous scanning n and the detected light quantity Bn (i).

  FIG. 6 is a diagram for explaining the modulation method. For example, if the detected light amount at the irradiation intensity P1 is B1 as in the point 31, the reflectance at that time can be expressed as R1. Therefore, the irradiation intensity in the next scan 2 is modulated to P2 according to the point 32 where the detected light quantity becomes Bt in the case of the same reflectance according to the above modulation formula. If the amount of light detected as indicated by a point 33 in the scan 2 is B2, the reflectance at that time is R2. Therefore, the irradiation intensity in the next scan 3 is P3 as indicated by a point 34.

  According to such a modulation method, it is not necessary to perform a pre-scan before the main scan to determine the irradiation intensity, and the next irradiation intensity is simply calculated using the result of the previous scan. Optimum modulation can be performed without reducing the overall scanning speed.

  However, when the above-described bumps or the like are formed on the surface, there are cases where this modulation method is inappropriate. When the irradiation light is reflected at the apex of the bump, the reflectance becomes extremely high. In addition, as described above, in the scanning before and after the reflection at the apex of the bump, the scanning light is irregularly reflected, and the PSD can detect only dark light. Therefore, as shown in FIG. 6, when the irradiation intensity in the next scan is set as P3, and the result is a high reflectance such as R3, the detected light amount B3 is in the dynamic range with respect to the irradiation intensity P3. It becomes logically much higher than the saturation level Bs. However, since the saturation level Bs is actually exceeded, the detected light amount B3 is indefinite. Therefore, the irradiation intensity in the next scan cannot be calculated.

  The above problem is that, in order to shorten the entire scanning time, in the modulation method that calculates the next irradiation intensity from the irradiation intensity and the detected light intensity in the previous scanning, the aluminum wiring other than the boundary part of the bump of the semiconductor chip It also occurs at the boundary.

  FIG. 7 is a diagram showing the distribution of the amount of reflected light in the vicinity of the bump 2 analyzed by the present inventors. As is apparent from the description made with reference to FIG. 4, the distribution of the amount of reflected light is such that the most reflected light amount of the bump apex 41 is slightly shifted upward from the actual bump 2 position. There is a large area, and the bump shadow 42 is dark before and after that area. Therefore, it is effective to limit the modulation area 40 to a rectangular area that surrounds the bump apex 41 and includes the periphery thereof in the scanning direction as shown in FIG.

  In the modulation region 40, modulation is performed using the above-described modulation calculation formula. It is also effective to change the modulation formula in the shadow portion 42 and the other vertex portion 41 in the modulation region 40. is there.

  FIG. 8 is an enlarged view of the modulation region. If the above-described modulation formula is simply applied in the modulation region 40 described above, the detected light amount is saturated at the boundary between the bump apex 41 and the bump shadow 42. Therefore, in the present invention, attention is paid to the fact that there is a very dark shadow portion 42 before and after the bump apex portion 42 is irradiated, and the region 45 corresponding to the shadow portion 42 in the modulation region 40 has a certain weakness. Scanning is performed using the irradiation intensity P0, and in the area of the bump apex 41, the irradiation intensity is modulated by the above-described modulation calculation as usual. Further, the above-described modulation calculation as usual is also applied to the modulation area 44 outside the bump shadow 42.

  FIG. 9 is a diagram showing the relationship between the irradiation intensity and the detected light amount for explaining the modulation method according to the present invention. In FIG. 8, it is assumed that scanning is performed from the upper side to the lower side. In this case, when entering the modulation region 40, the initial irradiation intensity is a value that does not exceed the saturation level Bs of the PSD that is the light spot position detection element even if the portion having the highest reflectance in the chip is irradiated. Set to P0. As shown in FIG. 9, P0 is selected that does not exceed the saturation level Bs even at the highest reflectance Rmax. By setting the irradiation intensity P0, even if the bump apex 41 is irradiated for the first time, the PSD is not saturated. If the amount of light detected in the previous scan is lower than the predetermined threshold B0, it is determined that the region is a bump shadow region 42, and the next irradiation intensity is also set to P0. Accordingly, the threshold value B0 is set to a value lower than the detected light amount of the reflected light from the portion of the shadow 42 of the bump.

  As a result, in the example of FIG. 8, in the region 44, the irradiation intensity P (i) is set by the above-described arithmetic expression as usual. Eventually, the bump shadow area 45 is reached. Then, the detected light amount B (i) becomes lower than the threshold value B0. As a result, the irradiation intensity P (i) at the next scanning is forcibly set to P0. Thereafter, when reaching the bump apex region 41, the detected light amount B (i) also increases due to the high reflectance. However, as described above, the detected light amount B (i) with respect to the irradiation intensity P0 never becomes higher than the saturation level Bs, so that Bn (i) in the above arithmetic expression does not become indefinite. In the bump apex area 41, the irradiation intensity of the next scan is modulated as it is according to the arithmetic expression, and the PSD can detect the position of the light spot with the detected light quantity within the dynamic range.

  The height of the bump apex is detected based on the light spot position when the light reflected from the bump apex is imaged on the PSD.

  Further, in the area other than the modulation area 40, the irradiation intensity P (i) is forcibly set to zero, so that it is not necessary to perform useless calculation, leading to shortening of the height inspection process.

  In order to set the above-described modulation area 40 to an appropriate position with respect to the position of the bump 2, in the present invention, the irradiation light is scanned and the CAD data storing the surface pattern data of the semiconductor chip is referred to. That is, since the scanning start point and the scanning pitch are set in advance, the position on the chip being scanned can be known on the height detection device side. Therefore, by referring to the CAD data, it is possible to determine whether or not the modulation area 40 which is a predetermined area near the bump 2 is scanned. Alternatively, the current scanning position can be determined by comparing an image at the scanning position with CAD data using an imaging device such as a CCD camera.

  Therefore, when scanning the area outside the modulation area 40, the irradiation intensity is set to zero. When the scanning position enters the modulation area 40, scanning is started at the initial value P0, and the next irradiation intensity is calculated by the above-described modulation calculation formula. When the detected light quantity falls below B0, the irradiation intensity is set to P0 again, and the irradiation intensity is maintained at P0 while the detected light quantity falls below B0. Eventually, when the scanning position moves to the bump apex area 41, the detected light quantity exceeds B0 due to its intense reflectance, and the next irradiation intensity is calculated again by the above modulation calculation formula. And when it becomes the shadow of a bump again, the amount of detected light falls below B0, and the irradiation intensity after that is fixed to P0.

  As described above, according to the present invention, the above modulation calculation formula is used only in the region 41 very close to the apex of the bump, so that the saturation of the light spot position detecting element can be suppressed as much as possible. Further, in the shadow portions 42 and 45 around the bump apex portion 41, the irradiation light quantity is scanned at P0. Therefore, the initial modulation value of the bump apex portion 41 is always stable at P0. The modulation by the above arithmetic expression can be normally performed.

  FIG. 10 is a structural diagram of the height inspection apparatus according to the embodiment of the present invention. In this example, the semiconductor chip 1 on which the bumps 2 are formed is placed on a stage having an X stage 8 and a Y stage 9 that can move in the X axis and Y directions. Then, irradiation light 22 from the laser light source 4 is irradiated onto the surface of the chip 1 through an acousto-optic deflector AOD5, lenses 10b and 10a, and an objective lens 6a. In FIG. 10, the laser beam 22 is applied to the apex of the bump 2.

  The reflected light 24 forms an image on the PSD 7 of the light spot position detecting element through the objective lens 6b and the imaging lens 11. In general, when detecting the position in the line direction, the PSD 7 detects the position of the light spot according to the difference between the photoelectrically converted current values A and B detected from the left and right electrodes. Further, the detected light amount is obtained from the sum of the current values A and B. That is, the height of the object to be inspected and the amount of reflected light are obtained by the height light amount calculation circuit 19 in FIG. 52 is data of the obtained height and light quantity. The modulation circuit 17 obtains the irradiation intensity P (i) of the next scanning by the modulation method described above according to the modulation command signal 53 from the control unit 50.

  The laser drive circuit 18 drives the laser light source 4 according to the irradiation intensity P (i) obtained by the modulation circuit 17. Reference numeral 12 denotes a CCD camera that captures an image of the surface of the chip 1 to be inspected during scanning through the imaging lens 13, the half mirror 15, and the objective lens 16. The captured image data 54 is given to the control unit 50.

  Now, the operation of the above-described height detection apparatus is as follows. First, the X and Y stages are driven by the stage drive signal 55 from the controller 50, and the chip 1 is moved to an appropriate scanning position. The surface image of the chip 1 at that time is captured by the CCD camera and given to the control unit 50. The control unit 50 inputs CAD data from a CAD data file 51 having a pattern on the chip surface, and detects where the scanning position is with respect to the bump 2.

  If it is outside the deflection region, the instruction signal 53 instructs the modulation circuit to force the irradiation intensity to zero. When the deflection area is reached, the command signal 53 instructs the irradiation intensity to be P0 as an initial value. Then, the laser drive circuit 18 drives the laser light source 4 with a drive signal according to the irradiation intensity P (i), and laser light is generated. Then, the deflector 5 is driven according to the scanning signal 56 from the control unit 50, the irradiation laser light 22 is deflected in the horizontal direction, and the chip surface is scanned.

Then, the reflected light 24 at that time is imaged on the PSD 7 and the height is detected. Therefore, the height data obtained by the height light quantity calculation circuit 19 and the detected light quantity data 52 are supplied to the control unit 50. On the other hand, in the modulation circuit, it is determined whether or not the detected light amount is lower than the threshold value B0. If the detected light amount is higher than the threshold value B0 or lower than the saturation level Bs, the irradiation at the next scanning is performed according to the above modulation calculation formula as usual. The intensity P (i) is obtained. If the detected light amount is lower than the threshold value B0 or higher than the saturation level Bs, the next irradiation intensity is forcibly set to P0. That is, Pn + 1 (i) = Pn (i) × Bt / Bn (i) (B0 ≦ Bn (i) ≦ Bs)
Or, Pn + 1 (i) = P0 (Bn (i) <B0, Bn (i)> Bs)
It is.

  With such a modulation method, the modulation circuit 17 gives the irradiation intensity P (i) of the next scanning to the laser driving circuit 18. Of course, the modulation circuit 17 may create a lookup table according to an arithmetic expression prepared in advance without referring to the arithmetic expression described above every time, and obtain the table by referring to the table value.

  FIG. 11 is a diagram illustrating an example of another modulation region. In this example, an elliptical area around the bump apex area 41 is used as the modulation area 40. Although the modulation area 40 is different, the above modulation method is applied as it is. The size of the modulation area 40 is such that when the bump size is small or large and the position of the vertex changes, the position of the scanning light reflected by the bump vertex changes. Even if it occurs, it is necessary that the bump apex 41 is within the modulation area 40. That is, as shown in FIG. 4, the scanning light reflected at the bump apex becomes scanning light at a different position as shown by the broken line when the size of the bump is reduced.

  In the embodiment described above, the method of modulating the irradiation light when mainly detecting the height of the bump terminal on the surface of the semiconductor chip has been described. However, the present invention can be similarly applied to objects to be detected that have other irregularities on the surface and in which regions having different reflectances are distributed. In particular, when a region having a very high reflectance and a region having a very low reflectance are adjacent to each other, the irradiation intensity is forcibly set to P0 in a region having a low reflectance so that the boundary between the region having a high reflectance is high. By the way, the detected light quantity is prevented from becoming indefinite beyond the saturation region of the light spot position detecting element such as PSD. In the region where the reflectance is high, the optimum height detection can be performed by determining the irradiation intensity at the next scanning based on the detected light quantity at the previous scanning in accordance with the above-described arithmetic expression and the like. .

[Correction of detected height]
In the above embodiment, the height inspection is performed by modulating the incident light intensity corresponding to the fluctuation of the reflected light intensity caused by the surface shape of the bump. Next, a method for correcting the detected height error caused by the bump surface shape being a curved surface will be described below.

  FIG. 12 is a diagram for explaining the principle of the triangulation method. The triangulation method is as outlined in FIG. As shown in FIG. 12, irradiation light 22 that is incident light is applied to the measured surfaces 60 and 61 through the objective lens 6 a from the left side. Then, the reflected light 24 reflected by the measured surfaces 60 and 61 forms an image on the surface of the light spot position detecting element PSD via the objective lens 6b and the imaging lens 11. For the measured surface 60, the light spot position is at position 62, while for the measured surface 61, the light spot position is at position 63. When the surface to be measured is located above 60, the light spot position is on the higher side (High) than 62. As described above, in the triangulation method, in principle, the optical path length is fixed by fixing the position in the Z-axis direction of the stage on which the object to be measured is placed, and on the light spot position detecting element PSD according to the difference in height. The height is detected from the difference in the position of the light spot that forms an image. Since the stage does not move in the Z-axis direction, the height of a large number of bumps can be inspected in a short time by simply scanning the irradiation light with an acousto-optic deflector or the like.

  However, since the bump surface shape is a curved surface, an error occurs in the detected height due to various factors. For example, when the height is inspected by the triangulation method, the optical system is first adjusted to the in-focus position with respect to the object to be inspected, and then the optical system is fixed and the height is measured according to the above principle. In that case, the focus position of the optical system of the inspection apparatus varies before and after the surface of the inspection object. As a result, the height of the bump apex cannot be detected accurately.

  FIG. 13 and FIG. 14 are diagrams for explaining the problem of the bump height measurement. FIG. 13 shows a state where the optical system is accurately focused on the light spot position detection element PSD when the irradiation light 22 is irradiated on the apex position of the bump 2 (FIG. 13A), the light spot position. State diagram focusing on position 64 behind detection element PSD (FIG. 13B), state diagram focusing on position 65 in front of light spot position detection element PSD (FIG. 13C) It shows.

  FIG. 13A shows a case where the optical system is set such that the light spot is reflected by the apex of the bump 2 and the light spot is focused by the light spot position detecting element PSD. As shown in the figure, the reflected light 24 reflected at the apex of the bump 2 is focused on the position 62 of the light spot position detecting element PSD. Therefore, the detected height is a height corresponding to the position 62.

  On the other hand, as shown in FIG. 13B, when the focus position of the optical system is the position 64 behind the light spot position detection element PSD, the reflected light 24 is not focused on the light spot position detection element PSD. . However, the light spot position detecting element PSD detects the center of gravity position of the light spot image by the calculation of (a−b) / (a + b) with respect to the signals a and b detected from the terminals at both ends. The detected height is a height corresponding to the position 62. Further, as shown in FIG. 13C, when the focus position of the optical system is the position 65 in front of the light spot position detection element PSD, the detected height is the height corresponding to the position 62.

  In this way, when the irradiation light 22 is reflected at the apex of the bump 2, since this light is reflected evenly, it is possible to accurately detect the height even if the focus position of the optical system shifts. . However, when the irradiation light 22 is reflected at a position shifted from the apex of the bump 2, an error occurs in the detected height as the focus position of the optical system shifts.

  FIG. 14 shows a case where the irradiation light 22 is reflected at a position 66 shifted from the apex of the bump 2. 14A shows the case where the optical system is set so as to focus on the light spot position detection element PSD, and FIG. 14B shows the back of the light spot position detection element PSD. FIG. 14 (c) shows a case where it is set to focus on the position 65 in front of the light spot position detecting element PSD. When the irradiation light 22 is reflected at the position 66 deviated from the apex of the bump 2, the characteristic feature is that the reflected light 24 is not uniform with respect to the optical axis 69 but is biased downward.

  In the case of FIG. 14A, the reflected light 24 is focused on the position 62 of the light spot position detecting element PSD. As described above, even if the reflected light 24 is biased downward, the position of the center of gravity of the light spot is the position 62, and the detected height is the height corresponding to the position 62. However, in the case of the back focus in FIG. 14B, the barycentric position of the light spot detected on the light spot position detecting element PSD is a position 67 in the figure. As a result, the detected height is a higher value corresponding to that position 67. Therefore, although the position 66 of the bump 2 is lower than the position of the apex, the detected height is higher than the apex of the bump 2. This means that when the irradiation light 22 is scanned and the highest position is determined as the apex position, an incorrect bump apex height is detected.

  In the case of the front focal point in FIG. 14C, the barycentric position of the light spot detected on the light spot position detecting element PSD is a position 68 in the figure. As a result, the detected height becomes a low value corresponding to the position 68.

  As described above, when the height of a large number of bumps is detected by the triangulation method, after adjusting the optical system of the height inspection device, the optical system is fixed and the light on the light spot position detecting element PSD is used. The height is detected by the point position. Therefore, as shown in FIG. 14B, when the optical system is fixed at the back focal position, the height at the position 66 shifted from the apex is always detected to be higher. Conversely, as shown in FIG. 14C, when the optical system is fixed at the front focal position, the height at the position 66 shifted from the apex is always detected to be lower.

  As a means for avoiding the problem shown in FIG. 14, when the light intensity (signal a + b) of the light spot output from the light spot position detecting element PSD is the highest, it is determined that the irradiation light is reflected at the bump apex. There is a method in which the height detection value of is used as the height of the vertex. This method utilizes the fact that the intensity of the reflected light captured by the optical system on the reflection side is the strongest because there is no inclination like the position 66 when reflected at the bump apex.

  FIG. 15 is a diagram for explaining a method for detecting the vertex position based on the light intensity of the reflected light. As shown in FIG. 15A, when the irradiation light 22 is reflected at the position 66 that is shifted from the apex of the bump 2, the reflected light 24 is biased to the lower side of the optical axis 69, so that the case of FIG. However, the light intensity (signal a + b) detected by the light spot position detecting element PSD is reduced. Accordingly, the position 66 is detected as other than the apex of the bump. However, when the optical axis of the optical system of the inspection apparatus is deviated, the light intensity detected by the light spot position detecting element PSD may be maximized when the irradiation light is reflected at a position other than the apex of the bump 2. . FIG. 15B shows a case where the optical axis 70 of the reflection-side optical system is shifted downward from the original optical axis 69. In this case, since the reflected light 24 reflected at the position 66 other than the apex of the bump 2 becomes uniform with respect to the optical axis 70, the light intensity (signal a + b) detected by the light spot position detecting element PSD is the maximum. Become. Accordingly, the position 66 is mistaken as a vertex, and the height of the position 66 is measured.

  As described above, when there is a deviation of the optical axis of the optical system on the reflection side, when the height inspection is performed with the position where the light intensity of the light spot detected by the light spot position detection element PSD is the maximum as the vertex, There is a tendency that the height of the position lower than the vertex is detected.

  Therefore, the height inspection apparatus according to the present embodiment adds a correction height inspection unit in addition to the high-speed height inspection apparatus based on the triangulation method. A height value X (second correction height value) detected for the reference inspection object by the height inspection device and a height value detected for the same reference inspection object by the correction height inspection unit A correction conversion function is obtained from the correlation with Y (first correction height value). When the actual bump height measurement is performed, a value Y (corrected height value) obtained by correcting the detection output X (measurement height value) from the high-speed height inspection apparatus by the triangulation method according to the conversion function. ). The correction height inspection unit employs a measurement method that does not cause an error in the detected height due to the curved shape of the bump. For example, there is a method of detecting the height of the bump from the optical path length at that time by moving the stage in the Z-axis direction and changing the optical path length to detect the in-focus position.

  FIG. 16 is a diagram illustrating a configuration of the height inspection apparatus according to the present embodiment. In this example, a height inspection unit 200 for correction is added to the height inspection apparatus 100 based on the triangulation method shown in FIG. If the bump 2 on the same chip 1 can be measured by the height inspection apparatus 100 by moving the Y stage 9 on the X stage 8 as indicated by an arrow 74, the height inspection unit 200 for correction can increase the height. You can also check the thickness. The correction height inspection unit 200 changes the vertical position (Z-axis direction) of the object to be measured and measures the height of the object to be measured from the optical path length at the in-focus position. To do. Accordingly, the Z stage 70 is movable in the vertical direction, and the Z stage 70 is provided with a scale 72 and an encoder 73 for reading the scale.

  The height inspection apparatus 100 by the triangulation method is the same as that shown in FIG. 10, and the corresponding reference numerals are assigned to the corresponding parts. That is, the irradiation light from the laser diode 4 is deflected and scanned by the acousto-optic deflector AOD, and is applied to the bump 2 of the object to be measured through the lenses 10b and 10a and the objective lens 6a. The reflected light forms an image on the light spot position detection element PSD via the objective lens 6b and the imaging lens 11. Reference numeral 19 denotes a height calculation unit that receives the signals a and b from the light spot position detection element PSD and calculates the height from the signal corresponding to the light spot position.

  On the other hand, the height inspection unit 200 for correction irradiates the illumination light from the light source 18 onto the bump 2 that is an object to be measured through the half mirror 82. The reflected light passes through the half mirror 82, passes through the imaging lens 83, and is picked up by the image pickup device 84 such as a CCD. The height calculation means 85 controls the surface of the inspection object to be in the in-focus position while moving the Z stage 70 in the vertical direction. Whether or not it is the in-focus position is determined by, for example, whether or not the light spot on the image sensor 84 is the smallest (sharp). Then, the scale 72 is read from the encoder 73 for the position of the Z stage when the in-focus position is reached. By comparing this Z value with the reference Z value, the height of the bump apex of the object to be measured can be measured.

  As described above, the correction height inspection unit 200 obtains the in-focus position by changing the optical path length and detects the height of the object to be measured from the optical path length, so that it does not depend on the curved surface shape of the bump. The height can be measured. However, since it is necessary to move the Z stage, the measurement time becomes long.

  FIG. 17 is a flowchart when measuring using the height measuring apparatus of FIG. First, the optical system on the height inspection apparatus 100 side by the triangulation method is adjusted to the in-focus position (S10). That is, the position of the lens of the optical system, the position of the Z stage 70, and the like are adjusted so that the reflected light reflected by the object to be measured is focused on the surface of the light spot position detecting element PSD. As a result, the optical system is set at the in-focus position or slightly in the front or back focus position.

  Therefore, the Y stage 9 is moved in the right direction of the arrow 74, and the bump 2 of the object to be measured is placed under the height inspection unit 200 for correction. Therefore, the position where the light spot image captured by the image sensor 84 is the sharpest or the smallest is found while moving the Z stage in the vertical direction. From the Z value at that time, the height Y of the bump 2 is calculated by the height calculating means 85 (S12). This height Y is obtained for a plurality of bumps. The obtained height Y is stored in the height result storage memory 88.

  Next, the Y stage 9 is moved to the left of the arrow 74, and the height X of the plurality of bumps 2 of the same object to be measured is measured by the height inspection apparatus 100 (S14). At this time, the focus position of the optical system remains adjusted in step S10. The measurement result X is stored in the height result storage memory 87.

  In FIG. 16, a broken line portion 86 is a height calculation unit configured by, for example, a microcomputer or a personal computer. The height calculation unit 86 reads the values of the heights Y and X detected by both the inspection apparatuses 100 and 200 from the memories 88 and 87, and obtains the relationship between the heights X and Y by the height data correction unit 89. It is done. From the distribution of the heights X and Y for a plurality of bumps, a relational expression of both values is obtained by, for example, the least square approximation method.

FIG. 18 is an example in which a relational expression of a linear function of Y = aX + b is obtained by the least square approximation method. The measurement value X of the measurement system and the measurement value Y of the correction system are plotted, and a relational expression approximated from the distribution is obtained. Figure 19 is a further Y by the least square approximation method ^{_{= a n X n + a n}} -1 X n-1 + ... + a 0 X + b
This is an example in which a multidimensional function is obtained. As described above, once the optical system on the height inspection apparatus 100 side is adjusted, the error of the height X detected in the optical system has a certain tendency. Accordingly, as shown in FIGS. 18 and 19, a fixed relational expression is obtained from the distribution of both values X and Y (S16). This relational expression is stored in the memory 90 as a conversion function or conversion table.

  Thereafter, the height of the bump 2 is measured on the plurality of chips 1 to be measured on the height inspection apparatus 100 side by the triangulation method described above (S18). The height value X calculated from the signal from the light spot position detection element PSD by the height calculating means 19 is converted into a value Y corrected according to the above conversion function or conversion table by the height correcting means 91. The corrected height value 92 is output (S20). The above steps S18 and S20 are performed until the measurement of the chips of all the objects to be measured is completed (S22).

  Once the optical system of the height inspection apparatus 100 is readjusted, the above steps S102 to S16 are performed again to obtain a new conversion function. Then, the subsequent measurement is corrected and calculated according to the newly obtained conversion function.

  In the above example, the conversion function is obtained from the relationship between the measurement values X and Y. However, a table of measurement values and correction values as shown in FIG. 20 may be generated and stored in the memory 90. In that case, the height correction means 91 needs to obtain a value Y corrected by linear interpolation or spline interpolation from the table value.

  FIG. 21 is a diagram illustrating another example of the height inspection apparatus according to the present embodiment. The same reference numerals are assigned to the same parts as those in the height inspection apparatus of FIG. In this example, the height inspection apparatus 100 of the measurement system is the same as FIG. The height measurement unit 200 of the correction system shown in FIG. 16 detects the position where the light spot of the reflected light from the object to be measured is the smallest or sharp as the in-focus position. On the other hand, in the height measuring unit 200 of the correction system in the example of FIG. 21, the reflected light from the object to be measured is reflected by the half mirror 96, and the amount of reflected light that has passed through the imaging lens 97 and the slit 95 is set as the light quantity. Monitoring is performed by the detecting means 94.

  When the reflected light reflected by the bump 2 is focused by the slit 95, the light amount detecting means 94 detects the maximum light amount. On the other hand, when the reflected light is focused in front of or behind the slit 95, the light amount of the reflected light passing through the slit 95 decreases, so the light amount detection means 94 detects a small amount of light. Accordingly, the Z value when the output of the light amount detecting means 94 reaches the peak value while the Z stage 70 is moved in the vertical direction is read by the encoder 73, and the Z value in the focused state can be obtained. . A height detection value Y for correction is obtained from the Z value.

  Also in the correction system height inspection unit 200 shown in FIG. 21, the Z stage is moved to change the optical path length, the in-focus position is detected, and the height of the object to be measured is detected. That is, in principle, it is the same as the height inspection unit 200 of the correction system in FIG.

  FIG. 22 is a diagram showing another example of the height inspection apparatus according to the present embodiment. The same reference numerals are assigned to the same parts as those in the height inspection apparatus of FIGS. In this example, the height inspection apparatus 100 of the measurement system is the same as that shown in FIGS. In the height inspection unit 200 of the correction system in the example of FIG. 22, a knife edge 98 is provided instead of the slit of FIG. 21, and a two-divided photodiode 99 is provided instead of the light amount detecting means. The knife edge 98 is provided at a position that blocks the upper half of the reflected light. Therefore, when the reflected light is focused at the position of the knife edge 98, almost all of the reflected light is sharply imaged in the middle of the two-divided photodiode 99. On the other hand, when the reflected light is focused before and after the knife edge 98, the upper half of the reflected light is blocked by the knife edge 98. Therefore, the center of gravity of the light spot moves to the photodiode below the two-divided photodiode 99.

  Thus, when the signal ab from the two-divided photodiode 99 is monitored, it is detected that the in-focus position is obtained when ab = 0. Therefore, by moving the Z stage in the vertical direction, the signal ab from the two-divided photodiode 99 is monitored, and the Z value at which the signal ab = 0 is detected to detect the Z value at the in-focus position. The Z value can be determined. Therefore, in the case of FIG. 22 as well, the principle of measuring the height by changing the optical path length and detecting the in-focus position is the same as in FIGS.

  FIG. 23 is a diagram showing another example of the height inspection apparatus according to the present embodiment. The same reference numerals are assigned to the same parts as those in the height inspection apparatus of FIGS. 16 and 21 and 22. In this example, the height inspection apparatus 100 of the measurement system is the same as FIG.

  The height inspection unit 200 of the correction system in this example is based on optical path length measurement using light interference. Light projected from the light source 81 such as laser light is reflected by the half mirror 82 and projected onto the bump 2 that is the object to be measured. The light transmitted through the half mirror 82 passes through the reference path lens 103 and is reflected by the reference mirror 102. The laser beams reflected by the object to be measured and the reference mirror are again synthesized by the half mirror 82. The laser beam reflected by the DUT 2 and the laser beam reflected by the reference mirror 102 have their phases at the position of the half mirror 82 due to the difference in optical path length. Accordingly, the laser beams interfere with each other by this synthesis, and a beat signal is generated. Interference fringes generated by this interference are detected by the image sensor 84, and the optical path length to the object to be measured can be detected based on the size of the interference fringes. Therefore, in this example, there is no need for an encoder or scale for detecting the Z value of the Z stage.

[Removing astigmatism from acousto-optic deflectors]
As described above, the problem of the intensity of irradiated light and the error of the detection height due to the bump surface being spherical have been solved. In the present invention, in order to detect the heights of a large number of bumps on the LSI chip, a triangulation method is used, the irradiation light scans on the LSI chip, and the reflected light is coupled onto the light spot position detecting element PSD. The height is obtained from the position of the imaged light spot. In that case, an acousto-optic deflector is used to scan the irradiation light at high speed without physical driving.

  FIG. 24 is a diagram showing such an acousto-optic deflector (hereinafter referred to as AOD). The AOD 122 is a crystal in which a change in refractive index is induced by the influence of distortion caused by the propagation of sound waves, and is used for high-speed scanning of light by utilizing the fact that the refractive index changes depending on the frequency of sound waves. . As shown in FIG. 24, a voltage Vin having a sawtooth waveform is applied to the voltage controlled oscillation circuit (VCO) 120, and the frequency of the generated signal is changed from f1 to fn. By applying the frequency signal to the AOD 122, the refractive index of the laser beam 110 changes with time, and the laser beam is scanned in the scanning direction SCAN.

  The incident laser beam 110 is parallel light, but the laser beam refracted from the AOD 122 becomes light having a certain spread. If the frequency is changed at a high speed in order to scan at high speed, the frequency of the applied sound wave is strictly different at both ends of the spot diameter of the laser beam 110. Therefore, the refractive index at the both ends of the spot diameter is different. It is. Moreover, the spread of the laser beam 110 occurs only in the scanning direction SCAN, and does not occur in the direction perpendicular to the paper surface of FIG. This is referred to herein as the AOD cylindrical lens effect.

  FIG. 25 is a diagram for explaining astigmatism of the acoustooptic deflector. As described above, the laser light refracted by the AOD 122 has the light 126 in the scanning direction spread due to the cylindrical lens effect, and the light 124 in the direction perpendicular to the scanning direction remains parallel. Therefore, when passing through the optical system 129, the focal point 125 of the light 124 perpendicular to the scanning direction and the focal point 127 of the light 126 in the scanning direction are in different positions. This difference 128 between the focal points 125 and 126 is called astigmatism. The presence of such astigmatism interferes with accurate height detection in a method in which the laser beam is focused on the surface of the bump, which is the object to be measured, and the height is detected from the position of the light spot of the reflected light. .

  FIG. 26 is a diagram showing a configuration example of an irradiation optical system for eliminating the AOD cylindrical lens effect. By inserting the cylindrical lens 130 on the refracted light side of the OAD 122, the refracted light spread in the scanning direction can be returned to parallel light. In the figure, 4 is a laser light source, 10a and 10b are relay lenses, and 6a is an objective lens. It is desired that the surface of the sample 131 to be measured is scanned in the X direction at a constant speed in a state where the laser beam is focused.

  By inserting the cylindrical lens 130 as shown in FIG. 26, the static cylindrical lens effect of the AOD accompanying the change of the scanning frequency can be somewhat removed. However, the AOD characteristic includes a dynamic cylindrical lens effect due to distortion of the frequency characteristic of the voltage oscillation circuit that generates the frequency signal supplied thereto.

  FIG. 27 is a diagram for explaining a dynamic cylindrical lens effect due to characteristic distortion of the voltage oscillation circuit. By applying a sawtooth voltage Vin to the voltage oscillation circuit 120, the laser light is refracted by the AOD 122, and the refracted light is passed through the cylindrical lens 130, so that the static cylindrical lens effect is temporarily eliminated. However, due to the characteristic distortion of the voltage oscillation circuit 120, the change of the output frequency is not linear but distorted with respect to the linear change of the input voltage Vin. Therefore, if the sample 131 passes through the optical system 129, the sample 131 may be focused on the sample 131 depending on the scanning position. FIG. 27 shows in-focus positions 134, 135, and 136 for light in the scanning direction. In this example, the focal point 134 and the front focal point are at the scanning start position of the sample 131. Furthermore, the focal point 135 just coincides with the sample surface at the scanning center position of the sample 131. Then, at the scanning end position of the sample 131, the focal point 136 and the front focal point are set again.

  FIG. 28 is a diagram illustrating an example of the frequency characteristic of the output with respect to the input voltage Vin of the voltage oscillation circuit 120. In this example, the solid line is the frequency characteristic, and the frequency change rate decreases in the region where the input voltage Vin is low and high, and the frequency change rate is high in the region where the input voltage Vin is near the center. . Ideally, it is desirable to have a broken line characteristic in which the frequency also changes linearly with respect to the linear change of the input voltage Vin. However, in general, the voltage oscillation circuit has a distorted characteristic as shown in FIG. This distortion characteristic is an example.

  As shown in FIG. 28, the distortion of the output frequency characteristics of the voltage oscillation circuit 120 causes a phenomenon that the focal point position differs depending on the scanning position as described in FIG. That is, the region where the input voltage Vin is high and the region where the input voltage Vin is low correspond to the region where the output frequency is low and the region where the output frequency is high in the characteristic diagram of FIG. At that time, the spread of light in the scanning direction described with reference to FIG. On the other hand, in the region where the input voltage Vin is in the center, the output frequency changes rapidly, and the spread of light in the scanning direction described with reference to FIG. 24 increases. Even if the spot diameter of the laser beam is the same, if the frequency change is slow, the difference in frequency at both ends is small, the spread of the spectrum is small, and if the frequency change is fast, the difference in frequency at both ends is large. This is because the spread of the spectrum becomes large. Accordingly, as shown in FIG. 27, assuming that the focal point 135 is on the sample 131 at the center of the laser beam scanning, the front focal points 134 and 136 are obtained at the start and end of the scanning.

  The dynamic cylindrical lens effect due to the distortion characteristics of the voltage oscillation circuit described above may be given by distorting the input voltage Vin so that the change of the output frequency becomes constant according to the characteristics of the circuit, for example. However, since the characteristics of the voltage oscillation circuit are different, it is necessary to generate a change curve of the input voltage Vin in accordance with the characteristics for each inspection device.

  Furthermore, since the cylindrical lens 130 itself has a tolerance, even if the cylindrical lens 130 is simply inserted behind the AOD 122, the completely static cylindrical lens effect cannot be removed. Moreover, since the tolerances of the optical system are different for each inspection apparatus, it is desirable to eliminate errors caused by the tolerances.

  Most ideally, the cylindrical lens 130 is inserted behind the AOD 122 and the input voltage Vin is set to the optimum sweep waveform, so that the laser light moves on the sample at a constant speed and on the sample in the entire scanning region. In other words, it is possible to achieve scanning with focusing at zero astigmatism. A method for this will be described below.

  FIG. 29 is a diagram showing the relationship between the scanning time of laser light and the input voltage, diffraction angle, astigmatism, scanning position, and scanning speed. When the sweep waveform of the input voltage Vin is linear as shown in FIG. 29A, the change in the diffraction angle due to AOD is as shown by the solid line in FIG. That is, the change in the diffraction angle is slow at the scan start time t0, the change in the diffraction angle is fast at the center time t1, and the change in the diffraction angle is slow at the scan end time t2. Accordingly, as shown in FIG. 29D, the scanning position is distorted as shown by a solid line, and the scanning speed is slow at the beginning and the end, and becomes fast at the center t1. Accordingly, astigmatism is small at both ends and large at the center as shown in FIG.

FIG. 30 is a diagram showing the relationship between the scanning speed and astigmatism. As described above, astigmatism increases when the scanning speed is high, and astigmatism decreases when the scanning speed is low. In FIG. 30, the vertical axis represents astigmatism when the cylindrical lens is not inserted and astigmatism when it is inserted. By inserting the cylindrical lens, the astigmatism of the light in the scanning direction becomes zero at a certain speed v ₀ . Therefore, if this speed v ₀ is realized in the entire scanning range, zero astigmatism should be realized in the entire scanning range.

FIG. 31 is a flowchart for detecting the height by setting an ideal sweep voltage curve of the input voltage Vin. In the flowchart of the present embodiment, the voltage applied to the voltage oscillation circuit can be arbitrarily controlled, and the light amount (light intensity) of the reflected light of the scanning light is detected, so that the optimum characteristic inherent to the optical system is obtained. A scanning voltage v ₀ and a sweep voltage curve of the input voltage Vin for scanning at a constant speed are obtained. In order to do so, two types of setting samples shown in FIG. 32 are used. In the setting sample 140, a pattern with a low reflectance (shaded portion) and a high region are formed in the Y-axis direction perpendicular to the X-axis, which is the scanning direction of the irradiation light. This pattern is composed of, for example, a silicon substrate (low reflectance, shaded portion) and an aluminum layer (high reflectance) formed on the surface thereof. In the setting sample 142, a pattern of a region with low reflectance (shaded portion) and a region with a high reflectance are formed in the scanning direction (X-axis direction) of irradiation light.

  As shown in FIG. 31, as a first step, the sample 140 is scanned with laser light in the non-curved surface direction of the cylindrical lens 130 to detect a focal point unique to the optical system (step S30). FIG. 33 is a diagram for explaining the method for detecting the focus position of the optical system in step S30. As shown in FIG. 33A, the input voltage Vin applied to the voltage oscillation circuit is made constant, and the state where the scanning of the laser beam is stopped at the center of the sample 140 is maintained. Then, the stage is moved at a constant speed in the Y-axis direction in the figure, and the amount of reflected light (light intensity) at that time is monitored by a PSD or a photodiode. The relationship between the amount of light observed at that time and time is as shown in FIG. 33B, and the amount of light rises at the pattern boundary portion of the sample 140. In that case, the optical path length is changed by changing the Z-axis direction of the stage or the lens position, and the position with the shortest rise time tr is detected. That position is the state where the sample is located on the focal point. The broken line in FIG. 33 (b) is a light amount waveform when not in the focal position, and the solid line is a waveform when in the focal position.

Next, as a second stage, a scanning speed at which the focal point of the light deflected by the AOD in the scanning direction matches the in-focus position obtained in step S30 is obtained. That is, this is a step of obtaining a scanning speed v ₀ (see FIG. 30) at which astigmatism is zero.

FIG. 34 is a diagram for explaining the process of detecting the scanning speed v ₀ with zero astigmatism in the second stage. First, the sample 142 is placed at the in-focus position obtained in step S30 above, and the input voltage Vin of the voltage oscillation circuit is linearly changed as shown in FIG. Scan in the direction. Further, as shown in FIG. 34 (b), the change rate of the input voltage Vin is changed as v ₁ , v ₂ , v ₃ , and the amount of reflected light detected at that time is changed as shown in FIG. 34 (c). It detects as follows. As indicated by the solid line v ₂ , the rate of change v ₂ of the voltage Vin when the rise time of the light amount becomes the shortest is obtained. It is understood that astigmatism is zero at the position of the pattern edge at the center of the sample 142 when the input voltage Vin is applied at this rate of change v ₂ .

  In the case of the input voltage Vin changing linearly as shown in FIG. 34B, the scanning speed of the laser light is large at the center t1 of the scanning due to the characteristic distortion of the voltage oscillation circuit as shown in FIG. It becomes smaller at t0 and last t2. Therefore, when the rate of change of the input voltage Vin is changed as shown in FIG. 34B, the scanning speed changes up and down as shown in FIG.

Therefore, the scanning speed of the solid line at the center time t1 of the scanning area corresponding to the edge of the pattern is an ideal scanning speed v ₀ at which astigmatism is zero. The above-described second stage process corresponds to steps S32, S34, and S36 in the flowchart of FIG. At this stage, the change rate v ₁ of the input voltage Vin that can realize the ideal scanning speed v ₀ at the position of the pattern edge has been found.

Next, as the third stage, an ideal scanning speed when the astigmatism is zero in the second stage is obtained. The sample 142 is moved by Δx in the scanning direction while scanning the laser beam while maintaining the change rate v ₁ of the input voltage Vin obtained in the second stage. At this time, the change Δt in the rise time of the amount (intensity) of the reflected light as shown in FIG. As a result, the scanning speed v ₀ with zero astigmatism is obtained as v ₀ = Δx / Δt (S38).

Then, as a fourth stage, a sweep waveform of the input voltage Vin is obtained so that the ideal scanning speed v ₀ is obtained at all positions on the scanning line (S40). Various specific methods are conceivable. FIG. 35 is a diagram for explaining an example thereof. As shown in FIG. 35B, when the input voltage Vin of the voltage oscillation circuit is linearly changed as indicated by the solid line 144, the scanning position is distorted as indicated by the solid line 146 in FIG. Is clear. This detection of the scanning position can be easily obtained by, for example, moving the edge position of the sample 152 to a plurality of known positions in the scanning direction (X-axis direction) and plotting the rise time of the reflected light amount. it can.

Therefore, by changing the scanning position of the solid line 146 shown in FIG. 35A linearly as shown by the broken line 147, the ideal scanning speed v ₀ can be realized in the entire scanning range. Therefore, the input voltage Vin is changed so that the scanning position of the solid line 146 can be changed to the scanning position of the broken line 147. When the input voltage vin for each changed time is plotted, an ideal sweep curve 145 shown by a broken line in FIG. 35B is obtained.

Therefore, by applying the input voltage Vin to the voltage oscillation circuit with an ideal sweep curve 145, it is possible to realize scanning at a scanning speed v ₀ where astigmatism is zero over the entire scanning region. The initial setting of the input voltage Vin is completed through the above four steps. The sweep curve of the input voltage Vin obtained by this initial setting is a curve in which the astigmatism is zero in the entire scanning region in consideration of all of the tolerance of the optical system, the tolerance of the cylindrical lens, and the inherent characteristics of the voltage oscillation circuit. is there. Moreover, it is a sweep curve in which scanning at an equal speed is realized. Therefore, as a height detection device, an ideal sweep curve of the input voltage Vin can be obtained by simply scanning the samples 140 and 142 and observing the change in the amount of reflected light at that time.

  Therefore, as shown in steps S42 to S46 of FIG. 31, the sample of the object to be measured is placed on the stage, and the input obtained in the fourth stage in the state of the optical system at the in-focus position obtained in the first stage. Laser light is scanned by a sweep curve of voltage Vin, and the height of the sample is detected by triangulation. This height detection is performed until the inspection of all the samples to be inspected is completed. In this height detection step, as described above, the laser element drive signal is optimally modulated, and the output of the detected light spot position detection element PSD is corrected, thereby performing more accurate height detection. be able to.

  FIG. 36 is a diagram illustrating a configuration example of an inspection apparatus for obtaining the above-described ideal input voltage Vin sweep curve. In this configuration example, the laser light from the laser light source 4 is deflected by the AOD 122 and irradiated onto the sample on the stage 152 via the cylindrical lens 130 and the scanning optical system 150. By sweeping the input voltage Vin applied to the voltage oscillation circuit 120, the laser beam is scanned over the sample on the stage 152. The amount of light reflected by the light receiving optical system 154 is detected by the optical sensor 156. The optical sensor 156 may be, for example, a light spot position detection element PSD, and the sum (a + b) of the output signals is the light amount (light intensity). A photodiode may also be used. The detected light amount is converted into a digital value by the A / D conversion circuit 158 and supplied to the CPU 160.

  The CPU 160 is connected to the memory 162 and the display 164. Further, the CPU 160 supplies a laser light source drive signal 175 to the laser driver 170 via the D / A conversion circuit 168. Further, the CPU 160 gives a signal 176 of the input voltage value of the voltage oscillation circuit 120. Further, a stage drive command signal 177 is given to the stage controller 172. A stage driver 174 is driven by the stage controller 172 and driven by the stage 152.

  With the inspection apparatus shown in FIG. 36, the input voltage Vin to the voltage oscillation circuit 120 can be arbitrarily changed by the CPU 160, and the amount of light detected at that time can be monitored. Further, the detected light amount can be monitored while moving the stage 152 in an arbitrary position and direction.

  Furthermore, as a correction pattern, a pattern having a uniform reflectance in the entire scanning region is added, the amount of reflected light when the uniform pattern is scanned is observed, and if there is unevenness in the amount of light, laser driving is performed. It is preferable to modulate the signal. As a result, it is possible to correct unevenness in the reflection intensity caused by the transmittance characteristics of the AOD, the visual field characteristics of the optical system, and the reflection characteristics of the sample.

  With the astigmatism generated due to the cylindrical lens effect of the acoustooptic deflector described above removed, the surface of the LSI chip with bumps of the material to be measured can be scanned at a constant speed. Therefore, a high-precision height inspection can be performed even when scanning is performed at high speed.

  As described above, according to the present invention, the irradiation intensity at the apex portion of the bump formed on the surface of the semiconductor chip or the like can be modulated at high speed and accurately. Therefore, the bump height inspection process can be performed at high speed and with high accuracy.

  Furthermore, even for an object to be inspected other than a semiconductor chip, the height inspection can be similarly performed at high speed and with high accuracy.

  Furthermore, according to the present invention, it is possible to appropriately correct an error in the height detection value due to the curved surface shape of the bump or the like, so that more accurate height detection can be performed.

  Furthermore, according to the present invention, the surface of the material to be inspected can be scanned with irradiation light at a constant speed without astigmatism.

It is a perspective view which shows the bump terminal formed in the surface of a semiconductor chip. It is a fragmentary sectional view of semiconductor chip 1 in which a bump terminal is formed. It is a top view explaining the method of scanning a laser beam with respect to a bump and detecting the height of a bump. It is a side view explaining the method of scanning a laser beam with respect to a bump and detecting the height of a bump. It is a graph which shows the relationship between the intensity | strength of irradiation light, and the light quantity which PSD detects. It is a figure for demonstrating the modulation method. It is a figure which shows distribution of the light quantity of the reflected light in the bump 2 vicinity. It is the figure which expanded the modulation area. It is a figure which shows the relationship between the irradiation intensity for demonstrating the modulation method concerning this invention, and a detected light quantity. 1 is a structural diagram of a height inspection apparatus according to an embodiment of the present invention. It is a figure which shows the example of another modulation area | region. It is a figure explaining the principle of triangulation It is a figure explaining the problem of the height measurement of a bump. It is a figure explaining the problem of the height measurement of a bump. It is a figure explaining the problem of the height measurement of a bump. It is a figure which shows the structure of the height inspection apparatus of the example of this embodiment. It is a flowchart figure at the time of measuring using the height measuring apparatus of FIG. It is a figure which shows the example of the conversion function for a correction | amendment. It is a figure which shows the example of the conversion function for a correction | amendment. It is a figure which shows the example of the table for correction | amendment. It is a figure which shows the other example of the height inspection apparatus of the example of this embodiment. It is a figure which shows the other example of the height inspection apparatus of the example of this embodiment. It is a figure which shows the other example of the height inspection apparatus of the example of this embodiment. It is a figure which shows an acousto-optic deflector. It is a figure explaining the astigmatism of an acousto-optic deflector. It is a figure which shows the structural example of the irradiation optical system for eliminating the cylindrical lens effect of AOD. It is a figure explaining the dynamic cylindrical lens effect by the characteristic distortion of a voltage oscillation circuit. It is a figure which shows the example of the frequency characteristic of the output with respect to the input voltage Vin of a voltage oscillation circuit. It is a figure which shows the relationship between the time in the scanning of a laser beam, input voltage, a diffraction angle, astigmatism, a scanning position, and a scanning speed. It is a figure which shows the relationship between a scanning speed and astigmatism. It is a flowchart figure which sets the sweep voltage curve of the ideal input voltage Vin, and performs height detection. It is a figure which shows two types of samples. It is a figure explaining the method to detect the focus position of an optical system. It is diagrams illustrating a process of detecting a scanning speed v ₀ of the astigmatism-zero. It is a figure for demonstrating the example of the method of calculating | requiring the sweep waveform of the input voltage Vin. It is a figure which shows the structural example of the test | inspection apparatus for calculating | requiring the sweep curve of the ideal input voltage Vin.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Semiconductor chip 2 Bump terminal 4 Laser light source 5, 122 Acoustooptic deflector, OAD
6 Objective lens 7 PSD, light spot position detection element 8, 9 Stage 12 CCD camera 17 Modulation circuit 18 Laser drive circuit 40 Modulation area 41 Bump apex area 42 Bump shadow area 50 Control unit 51 CAD data 120 Drive circuit, voltage oscillation circuit 130 Cylindrical lens

Claims (2)

The incident light is deflected by an acousto-optic deflector, and the deflected and spread incident light is returned to a parallel light beam by a cylindrical lens, and the incident light returned to the parallel light beam is incident on the object to be inspected via the incident optical system. In the height inspection method of scanning on the surface, imaging the reflected light from the surface on the light spot position detecting means, and detecting the height of the surface from the imaged light spot position,
A first region and a second region having different reflectivities are formed on a stage on which the inspection object is placed while changing an input signal of a driving circuit that supplies a scanning frequency signal to the acousto-optic deflector. By scanning the incident light on a setting sample having a patterned pattern and observing the amount of reflected light, the incident light is focused on the setting sample without astigmatism and further scanned at a constant speed. A setting step for obtaining a sweep curve of the input signal as
The input signal is applied to the drive circuit according to the sweep curve, the incident light is scanned on the surface of the object to be measured, and the reflected light from the surface is imaged on the light spot position detecting means, and the image is formed. A height detection step of detecting the height of the surface from the light spot position made ,
In the setting step, the first setting sample in which the focus position of incident light in a direction perpendicular to the scanning direction of the acousto-optic deflector is formed in the first region and the second region in the vertical direction. In the first step of setting the position of the incident optical system and the stage so as to be positioned above, and in the state of the position of the incident optical system and the stage set in the first step, the acousto-optic deflector While changing the input signal of the drive circuit that gives the scanning frequency signal, the incident light is scanned onto the second setting sample in which the first region and the second region are formed in the scanning direction, A second step of detecting a scanning speed at which an in-focus position of incident light in the scanning direction is located on the setting sample by observing the amount of reflected light; and the incident light is on the setting sample. The scanning speed within the scanning area Height inspection method characterized by having a third step of obtaining a sweeping curve of the input signal can be kept. The incident light is deflected by an acousto-optic deflector, and the deflected and spread incident light is returned to a parallel light beam by a cylindrical lens, and the incident light returned to the parallel light beam is incident on the object to be inspected via the incident optical system. In a height inspection apparatus that scans on the surface, forms an image of reflected light from the surface on the light spot position detection means, and detects the height of the surface from the light spot position formed on the surface,
A first region and a second region having different reflectivities are formed on a stage on which the inspection object is placed while changing an input signal of a driving circuit that supplies a scanning frequency signal to the acousto-optic deflector. By scanning the incident light on a setting sample having a patterned pattern and observing the amount of reflected light, the incident light is focused on the setting sample without astigmatism and further scanned at a constant speed. Setting means for obtaining a sweep curve of the input signal,
The input signal is applied to the drive circuit according to the sweep curve, the incident light is scanned on the surface of the object to be measured, and the reflected light from the surface is imaged on the light spot position detecting means, and the image is formed. A height detecting means for detecting the height of the surface from the light spot position ,
The setting means includes a first setting sample in which a focus position of incident light in a direction perpendicular to a scanning direction of the acousto-optic deflector is formed in the first region and the second region in the vertical direction. A first means for setting the position of the incident optical system and the stage so as to be positioned above, and the position of the incident optical system and the stage set by the first means in the acousto-optic deflector. While changing the input signal of the drive circuit that gives the scanning frequency signal, the incident light is scanned onto the second setting sample in which the first region and the second region are formed in the scanning direction, A second means for detecting a scanning speed at which a focus position of the incident light in the scanning direction is located on the setting sample by observing a light amount of the reflected light; and the incident light on the setting sample. The scanning speed within the scanning area Height inspection apparatus characterized by having a third means for determining the sweep curve of the input signal can be kept.

Images