JPH06194320A - Method and equipment for inspecting mirror face substrate in semiconductor manufacturing line and method for manufacturing - Google Patents

Abstract

PURPOSE:To allow to a mass production line of LSI to quickly rise, in early stage, with a yield by providing means for Performing photoelectric conversion with a pixel of specific mum or less, expressed in terms of the size on a mirror substrate, and detecting a signal corresponding to the number of scattering photons. CONSTITUTION:A mirror wafer 401 (provided with gentle irregularities having maximum height of 0.7nm or less at a pitch of 0.3mum or less) is mounted on a stage section 1560 and illuminated by an illumination optical system 1530. When a dust particle (microparticle of 0.05mum-0.109mum) is present at a detecting position 803 on the surface of the wafer, scattering light therefrom is detected by an objective lens 1541. Number of electrons arriving at a fluorescent surface 1523 varies depending on the number of photons being detected and thereby the intensity of a signal being detected by a linear sensor 1552, constituted of pixels having size of 1mum or less when expressed on the mirror surface wafer, varies. Consequently, a signal is detected by an appropriate threshold when the number of photons is large whereas a signal having a small value is detected when the number of photons is small.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a mass production start-up in a semiconductor manufacturing process and a foreign matter inspection method for a mass production line in which a foreign matter generated in a semiconductor production process and a mass production line is detected and analyzed to take countermeasures. And a semiconductor manufacturing method.

[0002]

2. Description of the Related Art In the manufacture of LSIs, foreign matter of 1/5 to 1/10 of the minimum pattern size is a problem, and 64M DRAM
Then, it is necessary to detect foreign matter of 0.05 μm on the wafer and take countermeasures. The LSI is manufactured by repeating processes such as exposure, etching, cleaning, and film formation. Fine particles are not allowed to be mixed in any process. These foreign substances have various causes such as those generated from the operating part of the transfer device, those generated from the human body, those generated by reaction in the processing device by the process gas, and those mixed with chemicals or materials. Are mixed in various states.

In the mass production start-up of LSI, it is necessary to investigate the cause of these foreign particles and take countermeasures. Since the foreign particle inspection on the wafer can inspect the foreign particles directly attached on the wafer, it is widely used in the LSI manufacturing. ing.

As a method of detecting particles on the wafer,
A scattered light detection method that enables high-speed nondestructive detection is excellent. A technique for detecting and analyzing minute objects on a wafer of this type has been developed in the past.
It is disclosed in the publication. This technique irradiates a laser on a wafer and detects scattered light generated when the foreign matter adheres to the wafer, and detects the detected foreign matter by laser photoluminescence or secondary X-ray analysis (XM).
It is analyzed by an analysis technique such as A).

[0005]

As the miniaturization of semiconductor elements progresses and foreign matter having a thickness of the above-mentioned insulating film becomes a problem, the foreign matter generated on the mass production start-up line of fine elements is detected and analyzed to take countermeasures. The conventional detection / analysis device is becoming insufficient in planning the above, and there has been a problem that it is necessary to detect and analyze finer foreign matter. The conventional technique has a problem in that minute particles having a thickness of the insulating film cannot be detected.

Further, in order to carry out the detection and analysis of the above-mentioned fine foreign matters, the detection and analysis equipment becomes extremely large, which requires cost and space, which is an obstacle to the reduction of the mass production line. This means that in order to detect a smaller foreign substance, it is necessary to more effectively scatter light by the foreign substance and collect the scattered light, and for analyzing a smaller foreign substance, Auger electron spectroscopy or secondary ion This is because an expensive and large-scale device such as a mass spectrometer (SIMS) is required, and this trend is expected to continue to progress in the future.
Further, since it takes a lot of time to detect and analyze such a minute foreign substance, there is a problem that it is necessary to use a necessary device as efficiently as possible to reduce the production cost. In addition, in order to reduce the mass production line, it was necessary to install necessary and sufficient monitors at necessary and sufficient locations.

An object of the present invention is to solve the above-mentioned problems of the prior art by providing a pitch of 0.5 μm or less and 0.05 μm existing on a mirror-like substrate having gradual fine irregularities with a height of 0.7 nm or less. It is an object of the present invention to provide a method and an apparatus for inspecting a mirror-finished substrate in a semiconductor manufacturing line, which can detect a foreign substance formed of the following fine particles so that an LSI mass production line can be quickly started up with a high yield.

Further, according to the present invention, a mirror-finished wafer is flown into an LSI mass production line, and 0.05 μm attached to the mirror-finished wafer.
It is possible to detect foreign matter consisting of the following fine particles
It is an object of the present invention to provide a semiconductor manufacturing method capable of realizing a high-yield high-speed mass production line of I.

[0009]

In order to achieve the above-mentioned object, the present invention provides a pitch of 0.5 μm or less and a thickness of 0.7 nm.
Laser light from an oblique direction is radiated to a mirror-finished substrate having the following gradual fine irregularities, and scattered light from the mirror-finished substrate is condensed by a condenser lens to form an image, which is converted by a photoelectric conversion means. It is characterized in that a signal corresponding to the number of scattered photons is photoelectrically converted by a pixel having a size of 1 μm or less converted on a mirror-finished substrate, and fine particles of 0.05 μm or less are discriminated from the minute irregularities by the signal and detected. This is a method of inspecting a mirror substrate.

Further, according to the present invention, laser light is obliquely irradiated to a mirror-finished substrate having a pitch of 0.5 .mu.m or less and fine undulations having a height of 0.7 nm or less, and the mirror-finished substrate is irradiated with the laser light. The scattered light of is condensed by a condensing lens to form an image, which is converted to a specular substrate by photoelectric conversion means and photoelectrically converted by pixels of 1 μm or less to detect a signal corresponding to the number of scattered photons. The method for inspecting a mirror-finished substrate is characterized in that fine particles of 0.05 μm or less are discriminated from the fine irregularities and detected to detect the generation state of the fine particles on the mirror-finished substrate.

Further, according to the present invention, there is provided a mounting means for mounting a mirror-finished substrate having a pitch of 0.5 μm or less and a fine unevenness of 0.7 nm or less in height, and the mounting means. The illumination means for illuminating the mirror surface substrate in an oblique direction, and the scattered light scattered from the mirror surface substrate is condensed by the detection optical system and converted into light on the mirror surface substrate and received by the photoelectric conversion means by the pixel of 1 μm or less. To detect foreign matter formed by fine particles of 0.05 μm or less on the mirror-like substrate on the basis of a signal detected from the photodetector of the detecting means. An inspection apparatus for a mirror-like substrate, which is characterized in that

Further, according to the present invention, a semiconductor mass production line is polished to a pitch of 0.5 .mu.m or less at a pitch of 0.5 .mu.m or less.
A step of flowing a mirror-finished wafer having gradual fine irregularities with a height of 7 nm or less, and irradiating a mirror-finished wafer obtained from a predetermined manufacturing apparatus with laser light from an oblique direction to collect scattered light from the mirror-finished substrate. A light lens collects the light to form an image, and the photoelectric conversion means converts it onto the mirror-finished substrate to perform photoelectric conversion at pixels of 1 μm or less to detect a signal corresponding to the number of scattered photons. Only a simple monitoring device based on the inspection step of detecting fine particles by discriminating them from the minute irregularities and feeding back the detected generation status of the fine particles to the mass production line, and in the mass production based on the generation status of the fine particles obtained in the inspection step. And a monitoring step of monitoring the semiconductor manufacturing method according to claim 1.

That is, the present invention uses a high-sensitivity detection optical system using a converging illumination optical system and an image intensifier. In the detection of fine particles by the laser scattered light detection method, scattered light noise from surface irregularities and fluctuations in the number of detected photons pose problems. In the present invention, in order to reduce this, Ar laser is irradiated, and photons are detected by the image forming optical system and the image intensifier. The detection pixel is 1 μm or less, and the wavelength of 500 nm is 100 mW. The wafer specifications that can be detected are:
The pitch is 0.3 μm and the maximum height is 0.7 nm.

Further, in order to achieve the above object, the method for mass production startup of a semiconductor manufacturing process and the foreign matter inspection method for a mass production line of the present invention, and the apparatus therefor use a sampling wafer at the time of mass production startup of a semiconductor manufacturing process. When evaluating foreign matter management conditions such as process, equipment, environment, etc., and the foreign matter detection / analysis / evaluation system at the time of mass production startup, the foreign matter on the sampling wafer is detected, and then the elemental species of the foreign matter are detected by a scanning tunneling microscope / Spectrometer (STM / STS)
And the STM / STS analysis data are stored in advance as a database and compared with the analysis target data to identify the foreign elemental species to be analyzed.

In the manufacture of LSI, the minimum pattern size is 1
Foreign matter of / 5 to 1/10 is a problem, and it is necessary to detect foreign matter of 0.05 μm on the wafer and take countermeasures in 64M DRAM. There is data that the concentration of foreign matter in the atmosphere also increases to around 0.05 μm. In a conventional semiconductor manufacturing process, the presence of foreign matter on a wafer causes defects such as poor wiring insulation and short circuits. With respect to these defects in the plane direction, the semiconductor element is further miniaturized, and defects in the vertical direction become a problem. When a minute foreign substance, specifically, a foreign substance having a thickness of a film such as a capacitor insulating film or a gate oxide film is present in a wafer, it causes damage to the capacitor insulating film or the gate oxide film. These new problems have been pointed out, and the need for particle detection is ever increasing. (See FIG. 18) As shown in FIG. 19, an LSI is manufactured by repeating processes such as exposure, etching, cleaning, and film formation. Fine particles are not allowed to be mixed in any process. These foreign substances have various causes such as those generated from the operating part of the transfer device, those generated from the human body, those generated by reaction in the processing device by the process gas, and those mixed with chemicals or materials. Are mixed in various states. In particular, minute foreign substances of heavy metals have recently been regarded as a problem. One of the main operations for mass production of LSIs is to investigate the cause of these foreign substances and take countermeasures, which involves detecting the foreign substances and analyzing the elemental species. It will be a great clue to the quest. In other words, dust generation in the apparatus of these manufacturing processes, dust generation in the atmosphere of the dust generation line of utilities such as gas, water, etc. is detected and managed using the in-liquid particle detection apparatus atmospheric particle detection apparatus wafer foreign matter detection apparatus. .
The detected foreign matter is SIMS (Secondery Ion Mass Spectroscopy).
scopy), XMA (X-ray Mass Analysis)
er) to analyze the cause, investigate the cause, and take countermeasures on the line. In this system, the foreign matter inspection on the wafer is capable of inspecting the foreign matter directly attached on the wafer, and since the foreign matter is sampled on the wafer, it has an advantage that it can be easily coupled with an analysis device and is widely used in LSI manufacturing. ing.

[0016]

As shown in FIG. 20, when a minute foreign substance is detected by this scattered light detection method, it is necessary to remove scattered light that becomes noise and detect weak scattered light. Specifically, it is necessary to solve the problem of fluctuations in the number of detected photons, the problem of scattering by air molecules (Rayleigh scattering), and the problem of scattered light from the wafer surface. This time, the problem of scattered light of air molecules can be solved by calculating by the Rayleigh scattering formula and miniaturizing the detection volume by the imaging optical system. For the scattered light from the surface irregularities, the scattering phenomenon can be modeled to quantitatively show the scattering mechanism. As for the problem of fluctuation, it is possible to quantitatively show the problems of the laser light intensity, the sampling time, and the false detection rate at the time of detection.

As shown in FIG. 21, in the optical system for illuminating the wafer and detecting the scattered light on the surface, the air molecules existing in the portion where the illumination light flux and the detection light flux overlap each other are also illuminated, and the light scattered by the air molecules is emitted. It becomes noise when detecting foreign matter. The scattered light intensity from the air molecules was calculated based on the Rayleigh scattering formula in which the scattered light intensity is proportional to the sixth power of the particle system. In the prior art, about 10 5 cubic m air molecules are illuminated,
The scattered light intensity cannot detect fine particles of 0.05 μm. By using an image forming optical system with an objective lens, this volume can be set to about 10.sup.2 cubic .mu.m, which is a level at which there is no problem in particle detection.

FIG. 22 shows a detected image when the mirror-finished wafer surface is irradiated with laser light. Light is scattered on the surface and appears bright. This scattered light noise poses a problem in detecting fine particles. Atomic force microscope (AF
The figure observed by M) is shown. The vertical axis is emphasized 100 times. The unevenness on the surface causes scattering. However, as shown in FIG. 23, the unevenness on the surface has a pitch of 10
Since the unevenness is about 0 to 1000 nm and the maximum height is about 0.1 to 5 nm, it is not possible to explain the phenomenon that the model using reflection does not enter the objective lens and can detect scattered light from the surface. .

Therefore, a diffraction phenomenon was introduced. As shown in FIG. 24, when there is a transparent object having a part where the phase difference is slightly different, the light is diffracted. The total intensity I of this diffracted light is calculated by the following formula.

I = | exp (-i (ωt + δ)) + exp (-i (ωt + π)) ** 2 = 2 (1-cosδ) = δ ** 2 (Equation 1) On this wafer surface It can be explained that when applied to the reflection of the light, the light is diffracted by the phase difference of the reflected light due to the unevenness and the light reaches the lens. In order to quantitatively calculate this diffracted light, the simulator based on the Fourier transform shown in FIG. 25 can be used. Pitch p, maximum height h on wafer surface
Assume that the unevenness is formed, and the case of illuminating with the light of wavelength λ at the incident angle θ is considered. The light intensity that can be detected in this case is expressed by the following equation using Fourier transform (F) and inverse Fourier transform (F-1) based on the data indicating the state of each surface.

I (x, y) = F-1 [F [F-1 [l (u, v)] * p (x, y)] * a (u, v)] F [f (x, y) )] = F (x, y) exp {-2πi (ux + vy)} dxdy F-1 [g (u, v)] = g (u, v) exp {2πi (ux + vy)} dudv (number 2) FIG. 26 shows the relationship between the uneven pitch p and the scattered light intensity.
When the unevenness pitch p is changed, the intensity of the light scattered from the unevenness having the pitch of the wavelength λ is maximized at any incident angle. This qualitatively explains the phenomenon in which a pitch of about the wavelength is emphasized and detected. FIG. 27 shows the relationship between the incident angle θ of illumination and the scattered light intensity. When the incident angle θ is changed, 0.
With a pitch of 5 μm, the scattered light intensity is proportional to the square of cos θ. In the case of other pitches, it is proportional to 2 · p / λ.
FIG. 28 shows the relationship between the maximum height h of the unevenness and the scattered light intensity. When the maximum height h is changed, the scattered light intensity is proportional to the square of h at any incident angle. FIG. 29 shows the results of simulation for various wavelengths λ while changing the uneven pitch p. For other wavelengths, the scattered light intensity is certainly maximized when the pitch is about the wavelength.

In order to verify this simulation result, scattered light was measured for several wafers while changing the incident angle, and the pitch p and the maximum height h measured by the AFM were measured.
The intensity is compared with the scattered light intensity calculated by simulation. Three types of wafers A, B, C and D are shown in FIG.
It shows in 0. The results are in good agreement with the simulation results. On the other hand, in the case of a mirror-finished wafer, it is possible to quantitatively measure the state of irregularities based on the intensity of scattered light.

Next, the scattered light from the 0.05 μm fine particles is
The experimental results of the scattered light level of fine particles on the mirror-finished wafer and the surface unevenness are shown under the detection condition of Ar laser of 100 mW and 1 μm pixel for about 20 minutes per inch wafer. Figure 3
1 shows the scattered light intensity by changing the incident angle. The scattered light from the wafer is a measurement of scattered light from a 1 μm square pixel. When the pixels are reduced to 1 μm in this way, the unevenness of the wafer and the 0.05 μm fine particles can be discriminated with a discrimination ratio of 5.

However, under this condition, the scattered light from the foreign matter of 0.05 μm is about 40 photons. The detection of 40 photons causes a new problem shown in FIG. It is assumed that an average of N background noises and 2N foreign object photons are detected at a specific illumination intensity and a specific sampling time Δt. That is, consider the case where the discrimination ratio is 2. In this case, photons are stochastically detected according to the Poisson distribution. For example, in the case of an average of 10, the distribution has about 1 to 2 to 30. As a result, the noise and the signal can be sufficiently discriminated when the illumination is sufficiently strong, while the fluctuation becomes large when the illumination is weak and the number of photons that can be detected is small, and the signal and the noise can be discriminated even with the same discrimination ratio 2. become unable.

Next, FIG. 33 quantitatively shows the relationship between the illumination light quantity and the detection rate. The horizontal axis represents the amount of light P irradiated onto the entire surface of the 8-inch wafer, and the vertical axis represents the number N of illumination photons from the fine particles.
When p is taken, the scattered light from the fine particles of 0.069 μm and 0.05 μm is shown by the dotted line in the figure, and increases as the irradiation light amount increases. On the other hand, the scattered light noise from the surface irregularities also increases with an increase in the amount of light and is associated with the horizontal axis of the figure. Here, a curve is plotted in consideration of the average number of photons Np that can be detected at a detection rate of 99% by setting a specific threshold value for an arbitrary noise photon Nn. Similarly, the average number of photons Np that can be detected at a detection rate of 50% for noise photons Nn is also shown. From this figure, for example, when an 8-inch wafer is irradiated with a 100 mW laser for 20 minutes, a pitch is about 100 to 1000 nm and a maximum height is about 0.1 to 0.7 nm on a mirror-finished wafer surface having very gentle unevenness. 10 photons are detected from the 1 μm square region, and 50 photons are detected from the 0.05 μm fine particles (foreign matter). The detection rate in this case is 9
It turns out that it becomes 9% or more.

As a result of the above examination, the pixel size is 1 μm square,
Light intensity of 100 mW and light irradiation time of 20 minutes
0.05 μm fine particles on an inch wafer can be detected with a detection rate of 95%. That is, when there is a certain surface irregularity, it means that only fine particles having a size equal to or larger than the value shown in the figure can be detected. In the detection of fine particles on a wafer, it means that the unevenness of the surface of the wafer greatly affects the detection performance.

FIG. 34 shows an example of device specifications. In detecting fine particles by the laser scattered light detection method by discriminating them from gentle unevenness, scattered light noise from the surface unevenness and fluctuations in the number of detected photons pose problems. In order to reduce this, in the present invention, as shown in FIG. 30, Ar laser light is irradiated to detect photons by the imaging optical system and the image intensifier, the detection pixel is 1 μm or less, and the wavelength is By using an Ar laser of 500 nm and 100 mW, fine particles can be discriminated from gentle unevenness and detected. As described above, the specular wafer specification that can detect fine particles (foreign matter) by discriminating them from gentle unevenness has a pitch p of about 0.3 μm or less (about 100 to 1000 nm) and a maximum height h of about 0.7 nm or less (0. 1 to 0.7
It is a gentle unevenness of about (nm).

However, in the case of detecting only gentle unevenness on a mirror-finished wafer, as shown in FIG. 30, wafers A, B, C and D (pitch p = about 0.2 μm to 0.7 μm, maximum height). It can be quantitatively detected to some extent by changing the incident angle θ for h = 0.5 nm to about 5 nm) and detecting the number of scattered photons.

[0029]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment in which the present invention is actually used in a semiconductor manufacturing line will be described below with reference to FIGS.
FIG. 1 is a configuration block diagram showing an embodiment of a method and apparatus for inspecting foreign matter in a mass production line and a mass production line in a semiconductor manufacturing process according to the present invention.

In FIG. 1, the exposure apparatus 10 is a foreign matter inspection apparatus for mass production startup and mass production line in the semiconductor manufacturing process.
1, an etching device 102, a cleaning device 103, an ion implantation device 104, a sputtering device 105, and a CVD device 106.
Semiconductor manufacturing apparatus group 100 including
1, a sensing unit 200 including a dust generating unit 202, a pressure sensor 203, a vacuum dust generating monitor 304 and the like, a sensing unit control system 205 thereof, and a utility group 3 including a gas supply unit 301 and a water supply unit 302.
00, a water quality sampling wafer 401, a gas sampling wafer 402, an in-apparatus sampling wafer 403, a device wafer 404, and an atmosphere sampling wafer 40.
5, a sampling unit 400 including a wafer foreign matter detecting unit 501 and a detecting unit 50 including a pattern defect detecting unit 502.
0, a scanning electron microscope (SEM), a secondary ion mass spectrometer (SIMS) 602, a scanning tunneling microscope / spectrometer (STM / STS) 603, and an infrared spectrometer 6
04, etc., a foreign matter fatality determination system 701, a minute foreign matter cause investigation system 702, and a pollution source countermeasure system 703. Further, these components are divided into an online foreign matter inspection system 1001 corresponding to a line and an offline foreign matter inspection system 1002 corresponding to a mass production start-up line. Make up.

FIGS. 2A to 2D are perspective views showing the construction of an embodiment of the sampling unit 400 shown in FIG. Figure 2
2 (a) to (d), the water quality sampling wafer 401 of FIG. 2 (a) is placed in the buffer chamber 408 in the unit including the pure pipe 406, the sampling faucet 407, the buffer chamber 408, and the drainage means. Foreign matter in the water supply unit 302 of FIG. 1 is sampled. Figure 2
Similarly, the gas sampling wafer 402 in (b) includes a gas pipe 410, a sampling valve 411, and a buffer chamber 4.
12, the rotary pump 413, and the exhaust means 414 are placed in the buffer chamber 412 in the unit, and the foreign substances in the gas of the gas supply unit 301 of FIG. 1 are sampled. In-apparatus sampling wafer 4 of the cross-sectional view of FIG.
03 is a processing device 415 (the etching device 102 of FIG. 1).
Etc.), the foreign matter generated in the processing device 415 is sampled by passing through the loader chamber 403, the processing chamber 417, and the unloader chamber 418. Either of these is possible. Further, the device wafer 404 is the processing device 4
15 is a wafer that is actually processed by the etching device 102 or the like. Atmosphere sampling wafer 4 of FIG.
05 is placed on the sampling table 420 in the processing environment 419, and the foreign matter in the processing environment 419 is sampled.

FIG. 3 is a configuration block diagram showing an embodiment of the detecting section 500 of FIG. In FIG. 3, the detection unit 500 includes a vacuum chamber 511, a high vacuum pump 512 such as an ion pump or a turbo molecular pump, and a valve 513.
And rough pump 514 such as rotary pump and window 51
5, 516, 517, a gate valve 518, a gate valve 520, a vacuum pump 521, a ha valve 522, a vacuum chamber system 510 including a gas spray nozzle 523, semiconductor lasers 531 and 532, condenser lenses 540 and 541, and mirrors 533 and 534. And a focusing objective lens 535, 536, a detection objective lens 537, a detection optical system 530 including a detector 538, and a cooler 539, a stage unit 560 including an XZY stage 561, a binarizing circuit 571, and a stage controller 572. A signal processing system 570 including a signal processing unit 573 and a coordinate data creating unit 574, an interface chamber 571, a load lock 582, and a wafer mounting unit 58.
3, a wafer transfer unit 584, a gas spray nozzle 585, a valve 586, a vacuum pump 587, and an interface unit 580 including a carriage 588.

FIG. 4 is a block diagram showing the configuration of an embodiment of the analysis unit 600 shown in FIG. In FIG. 4, the analysis unit 600 includes a vacuum chamber 611, a high vacuum pump 612 such as an ion pump or a turbo molecular pump, a roughing pump 614 such as a rotary pump, a gate valve 618, a preliminary vacuum chamber 619, a gate valve 620, and a vacuum pump. 621
And a vacuum chamber 610 including a valve 622 and a gas spray nozzle 623, and an atomic force microscope (AFM).
croscope) tip (AFM tip) 631, weak force reaction lever 632, lever fixing part 634, STM tip 633, STMXYZ fine movement unit 635, AFM bias power source 638, STM bias power source 637, current measuring means 639, 640, and sample mounting. STM unit 630 consisting of table 641, STM unit arm 636 and sample STM rough drive unit 642, STM XYZ fine movement unit controller 671, sample STM rough drive unit controller 672, analysis data creation unit 674, analysis data storage unit 675 and analysis data A signal processing unit 670 including a determination unit 676 and an information management unit 637, and the detection unit 5 in FIG.
No. 00 interface unit 580.

Next, the functions and operations of the sampling unit 400, the detection unit 500, and the analysis unit 600 of FIGS. 2, 3, and 4 which are the core of the offline foreign matter inspection system 1002 of FIG.

The sampling unit 40 shown in FIGS. 2 (a) to 2 (d).
0, for example, as an evaluation of pure water used in various manufacturing steps of the water supply unit 302 of FIG. 1 in the unit of FIG.
While pouring the pure water to be used on the sampling wafer 401, the foreign matter in the pure water is attached to the sampling wafer 401. Alternatively, the sampling wafer 403 is passed through a processing apparatus 415 (for example, the etching apparatus 102) having a vacuum processing chamber or the like of FIG. 2B to attach foreign matters generated in the processing apparatus 415. Further, the sampling wafer 405 is left at an arbitrary position in the clean room of the processing environment 419 of FIG. 2D, and foreign matter in the atmosphere is attached to the wafer 405. Details of the sampling wafers 401 to 405 used here will be described below with reference to FIGS.

FIG. 5 is a perspective view showing a mirror surface wafer of the sampling wafers 401 to 405 of FIGS. 1 and 2A to 2D. The mirror-finished wafer of FIG. 5 has a mirror-polished wafer surface, and has the advantage that it is most unlikely to be affected by the wafer surface when detecting and analyzing foreign matter.

FIGS. 6A, 6B and 6C show the sampling wafers 401 to 40 of FIGS. 1 and 2A to 2D.
6 is a perspective view showing a wafer on which Si3N4, Poly-Si, and Al films of No. 5 are formed, respectively. 6 (a), (b),
The wafer of (c) has the same material as that of the wafer to be cleaned when cleaning the cleaning tank (cleaning apparatus 103, etc.), for example, so that the adhered state of foreign matters at the time of cleaning is the same, so that the cleaning tank can be evaluated with higher accuracy. Further, these wafers can be passed through a film forming apparatus or the like in the next step of forming the respective formed films to evaluate the state of foreign matter generation in the film forming apparatus or the like.

FIGS. 7 (a), (b), (c) and (d) are
The sampling wafer 4 shown in FIGS. 1 and 2A to 2D.
FIG. 7A is a perspective view showing a wafer on which patterns 01 to 405 are formed, and the wafers in FIGS.
The patterns shown in the partially enlarged views of (c) and (d) are formed, and the pattern shapes of these wafers are models of patterns of actual devices. Figure 7
The wafers of (a) to (d) take into consideration that the attachment of foreign matter depends on the pattern shape, and these wafers can accurately reproduce the state of foreign matter attachment on an actual device. . Further, in the case of foreign matter detection, in the case of regularly formed patterns such as these wafers, the diffracted light from the pattern can be blocked with high precision by a wafer spatial filter or the like, so that highly precise foreign matter detection can be performed. Specifically, in the case of the wafer patterns of FIGS. 7A and 7C, if the wafer 401 is placed with the longitudinal direction of the pattern facing the illustrated y direction of the XYZ stage 561 of the detection unit 500 of FIG. Diffracted light from the pattern does not enter the detection objective lens 537 of the detection optical system. See also FIG.
In the case of the wafer patterns of (b) and (d), the direction of the pattern is set to the XYZ stage 56 of the detection unit 500 of FIG.
1 may be placed in a direction rotated by 45 ° with respect to the y direction shown in FIG. 1. In this case, since there is scattered light from the intersection 406 of the pattern of FIG. While it is not possible to detect foreign objects with a higher degree of accuracy than with patterns, Fig. 7
Since the actual device is modeled more faithfully than the pattern of (c), higher sampling accuracy can be achieved when evaluating the cleaning tank or the like.

FIGS. 8A, 8B, 8C and 8D show the sampling wafer 40 shown in FIGS. 1 and 2A to 2D.
It is a perspective view which shows the lapping direction at the time of manufacture of 1-405, and the scanning direction at the time of inspection. Normally, when the wafer is manufactured, the mirror surface is polished as the final finish, and the polishing direction at this time is the rotation direction around the central axis of the wafer as shown by the arrow in FIG. Case and FIG. 8 (b)
It is considered that there are a case in the y direction of the wafer as indicated by an arrow in FIG. 8 and a case of combining these FIGS. 8A and 8B. Therefore, a large number of minute flaws parallel to the polishing direction of the wafer of FIGS. 8A and 8B are formed on the wafer surface.
The wafers of (a) and (b) having the polishing direction combined also have flaws in the direction mainly formed by these synthesis, but these flaws hinder the detection of minute particles of foreign matter on the wafer. Becomes Therefore, as shown by the arrows in FIGS. 8C and 8D, the foreign matter inspection of the wafer surface is performed by scanning in the r and θ directions and the xy direction, respectively. The direction of the flaw can be kept constant with respect to 802, and the light mainly diffracted can be cut by this direction of the flaw.

The sampling wafers 401 to 405 to which the foreign matter is attached by the sampling unit 400 shown in FIGS. 1 and 2A to 2D are sent to the detection unit 500. If it is not desired to expose the wafer to the atmosphere with the vacuum processing apparatus or the like at this time, the sampling wafers 401 to 405 can be loaded into the vacuum chamber 511 by using the interface unit 580 of FIG. Further, the sampling wafer 401
An alignment mark may be attached to the reference marks 405 to 405 as a coordinate standard when connecting the detection unit 500 and the analysis unit 600. The alignment mark may be a cross mark, a # mark, or the like. For x,
Since y and θ are required, it is necessary to attach them at least at two or more places.

In the detection unit 500 of FIG. 3, the sampling wafer 401 carried in by the wafer foreign matter detection unit 501 is moved to XY.
The semiconductor lasers 531 and 5 are mounted on the Z stage 561.
Converging objective lenses 535 and 5 of an optical system for irradiating light from 32
At 36, the measurement point 803 on the wafer 401 is illuminated. The scattered light from the foreign matter on the measurement point 803 is imaged on the detector 538 by the detection objective lens 537 of the detection optical system. The signal photoelectrically converted by the detector 538 is binarized by the binarization circuit 571 and sent to the signal processing unit 573. One XY
The Z stage 561 drives the Z stage during the inspection to move the measurement point 80 to the focus position of the detection objective lens 537 of the detection optical system.
3 is controlled so that at the same time, the entire surface of the wafer 401 is inspected by scanning in the XY directions by the XY stage. If there is a foreign substance, the signal processing unit 573 takes in the coordinates of the XY stage from the stage controller 572, creates coordinate data in the coordinate data creation unit 574, and sends it to the analysis unit 600.

In the detector 600 of FIG. 4, the STM / STS is used.
The analysis using 603 will be described below. STM (Sc
Analysis using anning Tonneling Microscope (ST
S: Scanning Tonneling Spectroscopy) is {surface] Vo
l. 26 No. 6 (1988) PP. 384-391 “Scanning Tunneling Microscope / Spectroscopy (STM / STS) Application to Catalyst Surface Research”
Etc. in detail. In this document STM / S
It is described that TS can measure with high spatial resolution, but has a drawback that elemental species cannot be identified. According to this document, the information that can be collected by STM is bias voltage V
And the tunnel current i and the change ΔZ in the distance between the needle tip and the sample, and the work function φ of the sample surface can be calculated by calculating di / dz from these information.
The above-mentioned document does not clarify the reason why the elemental species cannot be identified, but it is considered that the elemental species cannot be identified for the following reason. That is, since the work function φ is a function of the bond state between the element species and the element, many element species and bond states with one work function φ are conceivable. Therefore, even if the work function φ can be determined, the element species and the element The binding state of cannot be determined. However, the present inventors have focused on the fact that there is a limit to the elements that may be mixed in the semiconductor manufacturing line, and based on the work function φ obtained by STM / STS, it is limited as to which element is to be measured. We paid attention to the fact that it is possible to select from the elements with the following.

The tunnel current iT that can be detected by the ammeter 640 of the analysis unit 600 shown in FIG.
Volume 9 PP. According to 1126-1137 “scanning tunnel microscope”, it is calculated by the following equation (Equation 3).

IT = k · JT = (e2 / h) · VT · (1 / (2π)) · (2π√2mφ / h) × (1 / Z) · exp (−2Z (2π√2mφ / h) ) (Equation 3) where JTf tunnel current density, e is electron burden, h is Planck's constant, m is electron mass, and φ is work function.
In this formula (Equation 3), VT is the bias voltage of the bias power supply 638, and the tunnel current CT can be measured.
Since only .phi. And .phi. Are unknowns, the work function .phi. Can be calculated by measuring the tunnel current CT with the distance between the needle tip and the sample being Z and Z + .DELTA.Z.

FIG. 9 is a block diagram of the analysis unit 600 of FIG.
It is an xy distribution diagram of the work function of 01. Analysis unit 60 of FIG.
When the tunnel current iT of the sample 401 is measured at 0, the work function φ is calculated from the equation (Equation 3), and the work function φ is distributed on the xy plane as shown in FIG. Can be understood in atomic order.

FIG. 10 shows the sample 40 of the analysis unit 600 of FIG.
2 is a sectional view of FIG. Since the value of the work function φ of the sample 401 of the analysis unit 600 of FIG. 4 is the work function of all three systems formed by the thin film 821 and the foreign matter 822 of the sample 401 as shown in FIG. Since the database of the xy distribution of the work function φ differs depending on the material of the base, it is necessary to measure various samples of the base. On the contrary, if these data are obtained for various substrates, the material of the sample 401 can be estimated.

FIG. 11 shows the sample 40 of the analysis unit 600 of FIG.
1 is the relationship between the distance Z between 1 and the AFM tip 631 and the atomic force f. When the Z direction of the STM chip 633 is controlled while keeping the current iA constant in the analysis unit 600 of FIG.
Sample 401 and AFM tip 631 by M tip 631
The distribution of the force between can be measured, and the relationship between the atomic force f between the sample 401 and the AFM tip 631 and the distance Z between the sample 401 and the AFM tip 631 can be obtained as shown in FIG. . This fZ waveform is described in, for example, Kittel, "4th Report: Introduction to Solid State Physics, Vol. According to 114-122, it is the sum of Coulomb force and repulsive force energy, and therefore has two parameters, and these two parameters determine the fZ waveform of FIG. If these two parameters are α and β, the following equation holds between the force f between the sample 401 and the AFM tip 631 and the distance Z between the sample 401 and the AFM tip 631.

F = k · exp (− (Z / α)) − (β / Z) (Equation 4) Further, from the equation (Equation 4), the force f at the distance Z of two places, that is, the Z of the chip drive unit 635. The parameters α and β can be calculated by measuring the coordinates.

FIG. 12 shows the sample 40 of the analysis unit 600 of FIG.
3 is a relationship diagram of parameters α and β of the relational expression (Equation 4) between the distance Z between the No. 1 and the AFM tip 631 and the atomic force f and the work function φ of the sample 401. FIG. 12 shows the result of calculating α and β from the f measured values at two Z positions by the above formula (Formula 4) and the value of the work function φ calculated from the measured value of iT by the above formula (Formula 3). When it is plotted in three dimensions like
The three-dimensional plot position 811 or position 812 is determined depending on the material of the sample 401 and the type of foreign matter. That is, if the data of α, β, and φ are measured for known fine particles (foreign matter), the type of fine particles (foreign matter) can be identified from the three-dimensional plot position of α, β, and φ of the measurement target.

FIG. 13 shows the sample 40 of the analysis unit 600 of FIG.
In the relationship diagram of the tunnel current i and the atomic force f of FIG. 1, the tunnel current iT and the atomic force f of the above formula (Equation 4) are changed while the bias voltage V of FIG. It is a measurement and can be a database. At the current STM-related research level, the tunnel current i
Is not fully understood about the exact correlation between the atomic force f and the atomic species, but the tunneling current i
-Atomic force f spectrum can be used for identifying the measurement target by measuring the material of the sample 401 and many substances related to fine particles (foreign matter). The spatial distribution of the work function φ is also a powerful clue for element identification.

The sampling unit 40 of the off-line inspection system 1002 for mass production start-up in the semiconductor manufacturing process of FIG.
0, the detection unit 500, and the analysis unit 600 obtain the measurement of the element of the foreign matter that may be mixed in advance and accumulate it as a database, and by comparing with the measurement result of the measurement target, Classification is possible. This concept ignores the precise meaning of the work function and other reasons why the phenomenon occurs, and its lack of accuracy is sufficient for the purpose of "estimating" the source by identifying the element of the foreign substance. Be useful.

The sampling unit 400 of the inspection system
By connecting each unit of the detection unit 500 and the analysis unit 600 by coordinate management, each unit can be operated at all times. Therefore, each unit disclosed in Japanese Patent Laid-Open No. 60-218845 is connected as a mechanism. The operating rate of each unit can be increased more than the technology used.
Furthermore, the performance of each unit alone can be easily improved, but this is because the mechanical coupling of the conventional units causes the balance of the vibrations to be lost and the vibration of the entire system to increase, or the electromagnetic waves of the entire system to increase. This is because it is possible to eliminate a situation where the field is out of balance and electrical noise increases.

Corresponding system 700 of off-line inspection system 1002 for mass production start-up in the semiconductor manufacturing process of FIG.
Then, the foreign matter generation source is estimated based on the information of the foreign matter on the sample detected and analyzed by the sampling unit 400, the detection unit 500, and the analysis unit 600, and one of the objects in the mass production line which is considered to be the generation source is considered. Measures are taken to eliminate dust generation, and this measure is evaluated by comparing the number of foreign particles generated before and after implementing the measures. On the other hand, the sensing unit 20 which can easily manage the process, the apparatus, the material, and the atmosphere of the online foreign matter inspection system 1001 corresponding to the mass production line which has been found to be a dust source for foreign matter generation in real time.
By installing a monitor of 0, this work is performed at the time of mass production startup of LSI. The sensing unit 200 also corresponds to a sampling wafer, but normally monitors the product wafer 406. During mass production, the installation, process, etc. are constantly monitored by the monitor of the sensing unit 200 installed in the online foreign matter inspection system 1001, and when there is an abnormality, the cause is investigated by the offline inspection system 1002.

The sensing section 200 of the online inspection system 1001 for mass production startup of the semiconductor manufacturing process of FIG.
Next, the in-vacuum dust generation monitor 204 will be described as an example of the above sensor. The air dust monitor cannot be used because there is no medium for carrying dust in the vacuum. However, the embodiment of the present invention uses the fact that there is no medium for carrying dust in the vacuum. That is, if there is no medium for carrying dust in a vacuum, the dust will fall by gravity, be attracted by electrostatic force, or randomly move by Brownian motion. Irregularly devises a technology that uses these two forces to count the number of dust particles in a vacuum.

FIG. 14 is a block diagram showing an embodiment of the in-vacuum dust generation monitor 204 of the sensing section 200 of FIG. In FIG. 14, this in-vacuum dust generation monitor 204
Is a place 1 that can be a foreign matter generation source in the vacuum processing apparatus 107.
08, the port 221, the negative grit electrode 223, the positive grit electrode 223, and the positive plate 22.
5, negative plate electrode 224, applied power supply 229, and ammeter 2
26 and 227 and a current counter 228, a DC voltage is applied between the negative grid electrode 222 and the positive plate electrode 225 and between the positive grid electrode 223 and the negative plate electrode 224 by an application power source 229, and an ammeter 226, Each 227 is made of a highly sensitive material that can measure even one electric charge.

With the above structure, the foreign matter 811 or the foreign matter 81
The operation will be described by taking as an example the case where 2 occurs and jumps to the positive grid electrode 223 or the negative grid electrode 222.
Now, when the foreign matter 811 passes through the positive grid electrode 223, if the foreign matter 811 has electrons excited, the positive grid electrode 223 receives the electron, and at this time, the foreign matter 811 is positively charged to the negative plate electrode 224. Reach As a result, a current flows from the power source 229 between the negative plate electrode 224 and the positive grid electrode 223, and this current is measured by the ammeter 2
26, and the current counter 228 can count the number of times this current has flowed, so that the number of foreign particles 811 that have come in can be counted. When the foreign matter 812 passes through the negative grid electrode 222 and easily emits electrons, the foreign matter 812 receives electrons from the negative grid electrode 222 and is negatively charged and reaches the positive plate electrode 225. At this time, current flows and the ammeter 22
The current counter 2 detects the number of times this current has flown as detected by 7.
By counting with 28, it is possible to count the number of foreign objects 812 that have come in.

Although the in-vacuum dust generation monitor 204 shown in FIG. 14 has both the negative grid electrode type and the positive grid electrode type, the one having only one of them is sufficiently useful depending on the application. In addition, foreign matter 81
For convenience of explanation, 1 and 812 exemplify those that easily receive electrons and those that easily release electrons, but this embodiment is not necessarily limited to this, and any particles are forcibly applied with a voltage. Even you can count. However, when the foreign matter described above comes in, it is considered that it is possible to count reliably because it is not disturbed by each voltage.

FIG. 15 is a structural perspective view showing another embodiment of the in-vacuum dust generation monitor 204 of the sensing section 200 of FIG. 15, the in-vacuum dust generation monitor 204 is installed in the piping system of the vacuum processing apparatus instead of the inside of the vacuum processing apparatus 107 of FIG. In the in-vacuum dust monitor 204, the ammeter 227 detects the current flowing between the negative grid electrode 222 and the positive grid electrode 225 arranged in the pipe 107 to detect the foreign matter moving on the gas flowing in the pipe 109. The number can be counted by the current counter 228.

A second specific example of the wafer foreign matter detecting section 501 will be described below with reference to FIGS. 16 and 17. That is, in the second embodiment, the XY stage 1561, the stage controller 1562, the z stage 1563, the automatic focus detection system 1564, and the z stage controller 156.
5, a stage unit 1560 composed of 5 and an Ar laser 1
531, a beam expander 1532, a cylindrical lens 1533, a polarization filter 1534, and an illumination optical system 1530 that is configured to rotate the illumination optical system 1530 to change the irradiation angle θ. The photon multiplication section 1 including a mechanism 1535, a detection optical system 1540 including an objective lens 1541, a photoelectric conversion surface 1521, a multi-channel plate 1522, a fluorescent plate 1523, and an applied voltage controller 1524.
520, imaging lens 1551, linear sensor 155
Detection unit 1550 composed of 2 and binarization circuit 157
1 and a signal processing unit 1570 including a microcomputer 1572.

In the stage section 1560, a mirror surface wafer (mirror wafer, sampling wafer) 401 (402, 4)
03, 404, 405), and the stage controller 156 according to the instruction of the microcomputer 1572.
The xy stage 1561 is driven via 2. At the same time, the focus position is detected by the automatic focus detection system 1564, and the Z stage 1563 is controlled via the Z stage controller 1565. Here, the automatic focus detection system 1564 may be a system that detects a fringe pattern as is obvious to those skilled in the art or a system that detects a change in the reflection position of laser light without any problem.

In the illumination optical system 1530, the Ar laser 15
The light emitted from 31 is expanded by the beam expander 1531, and the detection position 803 on the mirror-finished wafer 401 is illuminated at the incident angle θ through the cylindrical lens 1533 and the polarization filter 1534. Here, the detection unit 15
A cylindrical lens 1533 is used to focus the light in a direction perpendicular to the paper surface of FIG. 16 in order to make it linear in conformity with the shape of the linear sensor 50. Further, in order to correct the focal point of the front end of the beam (right side of the paper surface) and the front side of the beam (left side of the paper surface) when illuminating at the incident angle θ, the cylindrical lens 1533 is formed on a surface vertical to the irradiation optical axis of the beam. On the other hand, it is inclined by an angle α.

The Ar laser 1531 is used for A
This is because the r laser can easily oscillate a high-power laser with a short wavelength of 500 nm (100 mW that is 20 mW or more). Therefore, it does not have to be an Ar laser, but other lasers having a relatively short wavelength, such as a nitrogen laser,
It may be the second harmonic of a YAG laser, a helium cadmium laser, or the like.

Further, the illumination optical system 1530 may set the cylindrical lens 1533 so that the shape of the illumination area at the detection position 801 is circular. In this case, it is necessary to use a two-dimensional type detector (two-dimensional linear sensor) instead of the linear sensor 1552. Further, it is obvious to those skilled in the art that the shape of the illumination area as described here is formed by a cylindrical lens or an anamorphic prism. In the detection optical system 1540, the light scattered from the mirror surface wafer 401 (402, 403, 404, 405) is condensed by the objective lens 1541, and the photon multiplication unit 1
An image is formed on the photoelectric conversion surface 1521 in 520.

In the photon multiplication section 1520, the photoelectric conversion surface 15
At 21, the incident photons emit electrons, which are multiplied by the multi-channel plate 1522 and
23, and light is emitted from the fluorescent screen 1523. The DC high voltage applied to the multi-channel plate 1522 is controlled by the applied voltage controller 1524. By lowering the applied voltage, the foreign matter detection sensitivity can be lowered. Here, as the photoelectric conversion material,
A bialkali having a sensitivity in the band from 300 nm to 650 nm is suitable. This bi-alkali significantly reduces thermal noise by cutting the sensitivity on the long wavelength side (red). Thermal noise is approximately several pieces / second / square cm
Is. On the other hand, the photoelectric conversion material used in ordinary image intensifiers is called multi-alkali, which has a thermal noise of about 1000 to 20,000 pieces / sec · cm 2, and a thermal noise of several thousand during the inspection time. Therefore, tens of thousands of false alarms are produced, which cannot be used in the foreign matter inspection apparatus of the present invention. The ratio of the number of emitted electrons to the number of incident photons (quantum efficiency) is approximately 10 to 15%. Therefore, the number of detected electrons is a value obtained by multiplying the number of incident photons by this quantum efficiency. From now on, the number of detected photons will be referred to as the number of emitted electrons.

FIG. 16 shows a cross section of the image intensifier used in the present invention. The image intensifier is a photoelectron generating unit 1521 that converts light into electrons. The electrons are multiplied by accelerating the electrons in an electric field and colliding with a plurality of secondary electrons generated by colliding with the secondary electron generating surface. The photoelectron multiplying section 1522, a fluorescent surface 1523 that produces fluorescence by collision of electrons, and a power source and a controller section 1524 for applying a voltage across the photoelectron multiplying section 1522. Here, as the photoelectron multiplication section 1524, what is called a multi-channel plate in which thin tubular members having a secondary electron generation surface inside are arranged in a honeycomb shape is often used. The image intensifier used in the present invention has a dynamic range of about 100. Usually, in order to widen this dynamic range, the applied voltage at the time of accelerating the electron multiplying portion, that is, the electron is controlled. This applied voltage is 100V to 1000V
The total dynamic range can be changed to about 10 ⁶ by controlling within a range of about 10 ⁶ . By utilizing this, the dynamic range of the size of the detected foreign matter can be increased. That is, a large foreign substance to a small foreign substance can be inspected in one inspection.
Specifically, for example, an optical branching means such as a half mirror 1554 is provided between the objective lens 1541 and the first image intensifier 1520 shown in FIG. By using the electron acceleration voltage applied to the photoelectron multiplying section 1522 of each image intensifier 1520 in a stepwise manner by the power supply and the controller 1524, each image intensifier 1520 has a stepwise different size of foreign matter. Is configured to detect. With such a configuration, it is possible to detect from a large foreign matter to a small foreign matter, specifically, while measuring the size of the foreign matter from about 0.3 μm to about 0.01 μm. The size of the foreign matter is measured by the Mie theory that the scattered light intensity is proportional to the sixth power of the size of the foreign matter.

Here, the number of electrons incident on the fluorescent plate 1523 is several to several tens in the time for detecting one point.
These number of electrons are emitted from the fluorescent plate 15 at irregular time intervals.
It will be incident on 23. However, during the detection time of this one point, fluorescence is emitted from the fluorescent plate 1523 by the action of each electron for a certain period (longer than the time interval of electron incidence).
It continues to occur. Therefore, the fluorescence generated from the fluorescent plate 1523 does not have a pulsed shape due to the incidence of electrons, and fluorescence having a light intensity proportional to the number of incident electrons is continuously generated. That is, the number of electrons is detected as the intensity of fluorescence.

In the detector 1550, the light emitted from the fluorescent screen 1523 is imaged on the linear sensor 1552 by the imaging lens 1551. The signal processing unit 1570 converts the signal from the linear sensor 1552 into a binarization circuit 1571.
The foreign matter signal converted into a binary signal and detected by is stored in the foreign matter memory 1573 based on the position coordinate signal of the mirror-finished wafer from the stage controller 1562. It goes without saying that this data can be displayed and confirmed by the microcomputer 1572 in correspondence with the position of the mirror-finished wafer.

Next, the operation of the inspection apparatus will be described. A specular wafer (mirror wafer, sampling wafer) 401 shown in FIG. 1 before or after flowing through a predetermined process of the semiconductor mass production line or a manufacturing apparatus.
(402 to 405) (the pitch p is 0.3 μm or less, and the maximum height h is 0.7 nm or less, and there are gentle irregularities).
Is mounted on the stage unit 1560, and the illumination optical system 153
Illuminated by 0. When a foreign matter (fine particles of 0.05 μm to 0.109 μm) is present at the detection position 803 on the wafer surface, the scattered light from the foreign matter is used as the objective lens 1541.
Detected by. Since the foreign matter is minute, the scattered light is very weak and is at the level of 10 photons.

The number of electrons reaching the fluorescent screen 1523 changes according to the number of photons to be detected (10 to 10 ⁴ ), and the brightness of the fluorescent screen 1523 changes.
The intensity of the signal detected by the linear sensor 1552 configured by arranging with a pixel size of 1 μm or less changes.
As a result, when the number of photons is large, as shown in FIG. 17, the signal has a large value like a signal 851 and is detected as a signal by an appropriate threshold value. To be detected.

Here, the number of detected photons is set to the fluorescent screen 1.
Although the method of detecting the fluorescence intensity of 523 is used,
It goes without saying that a method of counting the number of photons may be used.

The detected result is the microcomputer 1
It is stored in the foreign substance memory 1573 by 572 and is displayed on the display at the same time as the foreign substance coordinates, the number of foreign substances, and the like.

In setting the pixel size, it is necessary to consider the surface irregularities of the mirror wafer to be inspected, the size of fine particles to be detected, the surface irregularities of the mirror wafer, the required detection rate, the permissible false alarm rate, and the like. It is preferable to make the pixel size and the stage scanning speed variable. That is, the setting of the pixel size has a great influence on the size of the fine particles that can be detected, the surface unevenness of the mirror wafer that can be detected, the required detection rate, the permissible false alarm rate, and the like.

Consider, for example, a case where the detection rate of foreign matters (fine particles) of 0.05 μm on the mirror-finished wafer surface is 95% and the number of false alarms on the entire mirror-finished wafer is about 100. When the amount of irradiation light P on the entire surface of the 8-inch wafer is plotted on the horizontal axis of FIG. 33 and the number of illumination photons Np from fine particles is plotted on the vertical axis, it is 0.069 μ.
The scattered light from the m and 0.05 μm fine particles is shown by the dotted line in the figure, and increases with an increase in the irradiation light amount. On the other hand, the scattered light noise from the surface irregularities also increases with an increase in the amount of light and is associated with the horizontal axis of the figure. Here, the curve in the figure is plotted by considering the average number of photons Np that can be detected with a detection rate of 99% by setting a specific threshold value for an arbitrary noise photon Nn. Here, the number of noise photons exceeding the threshold value is detected as a false alarm. Therefore, this threshold value must be set to a small value such that the detection rate becomes sufficiently large, and at the same time, set to a sufficiently large value so that the false alarm falls below the allowable range. Specifically, let Nk be the number of permissible false alarms on the entire surface of the specular wafer, Na be the number of pixels on the entire surface of the specular wafer, and if the probability p exceeds the threshold value by the Poisson distribution, the threshold value that satisfies Nk <Na · p. You need to set the value. Here, Na = 3.1E10, Nk =
The threshold value was set so that p would satisfy 100. Similarly, the average number of photons Np that can be detected with a detection rate of 50% for noise photons Nn is also shown. From this figure, for example, 2 lasers of 100 mW on an 8-inch mirror surface wafer
When irradiated for 0 minutes, about 10 photons are detected from the 1 μm square area due to surface irregularities (gradient minute irregularities with a pitch p of 0.3 μm and a maximum height h of 0.7 nm).
50 photons are detected from the foreign matter (fine particles) of 0.05 μm. The detection rate in this case is shown to be about 95%. It can be seen that the detection rate becomes 99% or more when the optical power is doubled. In the detection of fine particles on a mirror-finished wafer, it means that the surface irregularities of the mirror-finished wafer have a great influence on the detection performance. From the above examination, "when the false detection rate can be increased, the detection rate can be increased. When the detected foreign matter size is set large, the detection rate and the false detection can be reduced. When the inspection time is set large, the detection rate and the false detection are set. Can be made smaller. ”And other results are produced. Further, the specifications and concept of the foreign substance inspection device such as the detected foreign substance size and inspection time are determined, and then the device configuration such as the pixel size and the illumination light intensity is quantitatively and theoretically determined.

FIG. 35 shows the effect of miniaturizing pixels and reducing optical noise due to a high incident angle. In order to reduce the optical noise, if the area on the wafer taken in by the detector at one time, that is, the pixel size is reduced, the optical noise can be reduced as shown in FIG. However, with this alone, it is not possible to detect fine particles of 30 nm separately from noise.
As shown in (b), by increasing the incident angle, the optical noise was reduced, and it was possible to detect the signal from the fine particles of 38 nm with respect to the optical noise. By thus setting the incident angle to, for example, 20 degrees to 80 degrees, optical noise is reduced, 61 nm fine particles are detected when the pixel is 1 μm, and further 38 nm fine particles are obtained by changing the pixel from 1 μm to 0.3 μm. Is detected without the influence of noise.

FIG. 36 shows an output result obtained by detecting standard particles of each size under the conditions of an incident angle of illumination of 80 degrees and a pixel size of 0.3 μm. 38nm standard particles against optical noise,
It was confirmed that the discrimination ratios of 3 to 4 were detected. In this experiment, p-polarized light is used. Also, the calibration line is Mi
It is proportional to the sixth power of the particle diameter in the theory of e. The results also show that particles of 38 nm or less can be detected by further downsizing the pixels or by sampling on a wafer having a good surface condition.

Next, the detection rate will be shown. As described above, according to the present invention, it is possible to reduce optical noise and detect fine particles. Hereinafter, measures against the influence of fluctuations in the number of detected photons, which is a problem when implemented as an inspection device, will be described. FIG. 37 shows how the detection rate decreases due to the fluctuation of photons. As shown in FIG. 37, when the intensity of the detected light is sufficiently high, when a certain foreign matter is considered, the signal is detected stably, and if the S / N is about 2 with respect to the optical noise, the threshold value is set here. By setting a value, fine particles can be detected with a sufficient detection rate. However, when the detection light becomes weak and the number of detected photons becomes several tens, the number of photons detected from one fine particle has a fluctuation. As a result, if the threshold value is set by sufficiently distinguishing the signal from the noise, the detection rate is lowered in this way.

It is known that the distribution of the number of detected photons becomes a Poisson distribution. Taking the case where the ratio of the particle signal to the optical noise and the discrimination ratio = 4 as an example, the detection rates of 200 and 20 when the average number of detected photons of the particle signal becomes weak are compared. View. When a threshold value is set between the detection signal and the optical noise, a particle signal larger than this threshold value is detected at a specific detection rate. On the other hand, optical noise larger than this threshold is also detected. The noise detection rate (false alarm rate) must be sufficiently small.

FIG. 38 shows the position of the threshold value at which the detection rate becomes 95% with respect to an arbitrary average number of detected photons. At the same time, a threshold value is calculated and shown so that the false alarm rate becomes 10 ppt and the false alarm number becomes several in the entire 8-inch wafer. Here, when the discrimination ratio is 4, it can be seen that when the threshold value is set to about 120, the particle signal of about 200 average detection signals is detected at the detection rate of 95%. However, in the detection of 1/10, that is, about 20 weak lights, the fluctuation of the particle signal becomes large as shown here, and in the case of the discrimination ratio 4, the detection rate decreases to 10%. It can be seen from FIG. 38 that the detection rate sharply decreases in the detection of weak light.

FIG. 39 is a diagram showing the device specifications and the detection rate. 38 is a diagram showing the relationship between the detector speed and the inspection time, the diagram showing the inspection time and the laser power, and the diagram showing the relationship between the irradiation light intensity and the detection signal intensity in association with the detection rate diagram of FIG. FIG. 39. For example, if an inspection specification is made for an 8-inch wafer in 60 seconds, 6 pixels are required to use 1 μm pixels.
4MHz 8-channel detector is required, and 1W laser is used.
It shows that 0 nm fine particles can be detected.

Based on FIG. 39, the relationship between the detected pixel size and the detection rate is calculated and shown in FIGS. 40 (a) and 40 (b).
The vertical axis indicates the size of a foreign substance that can detect an 8-inch wafer at a detection rate of 95%, and the necessary irradiation energy at that time, that is, the product of the inspection time and the illumination intensity is well indicated on the axis. From the calculations and examinations shown above, it was found that the detection sensitivity of fine particles can be improved only by using a large irradiation energy and by reducing the pixel size to reduce optical noise. On the contrary, when the laser power is small, it was also found that the effect of improving the detection rate does not appear conspicuously only by miniaturizing the pixels. That is, in order to maintain the detection rate in the detection of fine particles, it is necessary not only to make the pixels finer to improve the discrimination ratio and to increase the illumination incident angle, but also to increase the illumination energy. I understood. Furthermore, the values are quantitatively shown in FIGS. 39 and 40.

FIG. 40B shows the size of foreign matter that can be detected with a detection rate of 10%. From this figure, it can be seen that fine particles of 0.1 μm or less can be detected even if the laser power is reduced when the detection rate is low. If this phenomenon is used, foreign matter is detected under the condition that the detection rate is small as shown in FIG. 40B, the detection rate is calculated in advance from FIGS. 39 and 40, and the detected number of foreign matter is divided by the detection rate. It is possible to predict the actual number of foreign substances. Of course, such a method is a statistical method, so there is a problem in detection accuracy, but it works well in the case where the number of detected foreign particles is sufficiently large or inspection that does not require high accuracy in the number of detections. Exert. That is, an inexpensive size inspection device of 0.07 μm level using a laser of about 20 mW with a detection pixel of several μm,
It is possible to realize the cleanliness control of the manufacturing line by a plurality of inspection systems of a high-sensitivity inspection apparatus using a 1 W laser with 0.3 μm pixels. By sharing the roles of these devices, it is possible to significantly reduce the capital investment in the manufacturing line and maintain the inspection performance at a sufficient level.

As described above, according to the present invention, 38 on a mirror-finished wafer is used for the purpose of cleanliness control during LSI production.
nm foreign matter inspection technology. According to the invention,
Since the optical noise caused by the surface roughness of the wafer can be analyzed by the diffraction model, the surface unevenness can be generally known by measuring the optical noise. Further, since the optical noise can be reduced by making the pixels finer and by the side scattered light detection method, it is possible to detect fine particles of 38 nm. Further, it can be verified by SEM that 38 nm fine particles are detected. In addition, we examined the decrease in the detection rate due to fluctuations in the number of detected photons when detecting weak light, and quantitatively showed the conditions for improving the detection rate, and found conditions under which even low-cost inspection equipment can perform inspection with sufficient performance. He also showed how to find out.

Further, as described above, in order to improve the detection rate of detecting ultrafine particles on the wafer, it is necessary to irradiate the wafer with a large amount of light energy. As a result, the wafer crystal may be damaged by crystal defects, movement of impurities, recrystallization due to melting, and the like. Therefore, when a high detection rate is required, it is necessary to use a sampling wafer and perform inspection using light of high energy, and when damage is a problem, it is necessary to perform inspection using energy that does not damage the wafer. is there. In this case, the detection rate decreases, so it is desirable to use it as a monitor in a semiconductor mass production line.

As described above, it is desirable that the illumination optical system 1530 shown in FIG. 16 can also control the energy of the irradiation laser beam.

Another embodiment of the present invention will be described below. In the above-described embodiment, the technique for detecting the particles adhering to the mirror-finished wafer has been described, but the present invention can be used not only for detecting the particles on the wafer but also for detecting the particles in water. FIG. 41 shows an embodiment of an apparatus for detecting foreign matter in water. In this embodiment, the Xθ stage 2101,
A stage controller 1562, a z stage 2102,
Automatic focus detection system 1564, z stage controller 15
A stage unit 2100 composed of 65, an Ar laser 1531, a beam expander 1532, a cylindrical lens 1533, an illumination optical system 2200 composed of a polarizing filter 1534, and objective lenses 2301 and 2303.
06, objective lens exchange means 2306, relay lens 23
03, an imaging lens 1551, and a variable magnification mechanism 2304, a detection optical system 2300 and a u-electric conversion surface 156.
1, multi-channel plate 1562, fluorescent plate 156
3, optical fiber 2305, applied voltage controller 15
64, a photon multiplication detection unit 2350 including a linear sensor 1552, a signal processing unit 2400 including a binarization circuit 1571, a microcomputer 1572, and an output unit 1573.

In the stage section 2100, the mirror surface wafer 80
3 is mounted, and the xy stage is driven via the stage controller 1562 according to a command from the microcomputer 1572. At the same time, the focus position is detected by the automatic focus detection system 1564, and the Z stage controller 156
The Z stage 2102 is controlled via 5.
Here, the automatic focus detection system 1564 may be a system that detects a fringe pattern or a system that detects a change in the reflection position of laser light, as will be apparent to those skilled in the art.

The illumination optical system 2200 is equivalent to the illumination optical system 1530 shown in FIG. 16, and is set to have the incident angle θ shown in FIG. In the detection optical system 2300, the image of the fine particles on the mirror surface wafer 803 is converted into the objective lens 23.
An image is formed on the photoelectric conversion surface of the photon image magnification detector through 01, the relay lens 2303, and the imaging lens 1551. Here, in the present invention, the pixel size on the sample of the detector can be changed by changing the total image forming magnification by the objective lens exchanging means 2306 and the magnification changing mechanism 2304. If the pixel size is increased, high-speed inspection can be realized, but the detection sensitivity does not improve. On the contrary, if the pixel size is reduced, the inspection speed cannot be increased, but the detection sensitivity can be improved. Therefore, it is desirable to use a detection optical system including a zoom lens and an illumination optical system including a zoom lens that illuminates an area having a size corresponding to the field size of the detection optical system.

In the photon multiplication detection section 2350, the photon image multiplication detection section uses a photoelectric conversion surface, a multi-channel plate, and a fluorescent surface. However, in order to satisfy the object of the present invention, The present invention is not limited to this, and a substance that directly detects an electric current without using a fluorescent screen or a substance that is used by cooling a solid state element may be used. However, the type shown in FIG. 41 is most suitable for realizing high-speed inspection. Here, in the embodiment shown in FIG. 16, a charge transfer type solid-state imaging device is used as the linear sensor 1552, but in the present embodiment, high-speed solid-state devices such as pin silicon photodiodes are arranged one-dimensionally. However, the signals can be detected in parallel. The charge transfer type solid-state imaging device can gain detection sensitivity but it is difficult to realize a high response speed.However, when pin silicon photodiodes are used in parallel, a high response speed is obtained if sufficient light intensity is obtained. Can be realized. In the structure of this embodiment, the pin silicon photodiode can be used because sufficient light intensity is realized by the multi-channel plate having an electron image multiplying function after photoelectric conversion.

In the signal processing unit 2400, the signal from the linear sensor 1552 is binarized by the binarizing circuit 1571, and the detected foreign matter signal is converted into the stage controller 1
Foreign object memory 1573 together with the position coordinate signal from 562
To store. This data is for microcomputer 1
It goes without saying that it can be displayed and confirmed by 572.

FIG. 42 shows a stage section 2100 for inspecting foreign matter in water. The stage unit 2100 includes an Xθ stage 2101, a z stage 2101, a vacuum chamber 2
103, sampling board 2501, board replacement means 21
02.

As the sampling substrate 2501, it is preferable to use a material such as sapphire which is easy to polish in order to reduce optical noise during illumination. However, this is not always necessary, and the wafer may be used as it is. When using sapphire or the like, since the substrate itself is transparent, the light transmitted through the substrate becomes stray light, which is an obstacle to detection. Therefore, in this embodiment, the back surface of the substrate is polished so that the light noise on the back surface does not become a problem, or the position where the light noise occurs on the back surface of the substrate is detected by the detection optical system 2.
The thickness of the substrate is set to a sufficient thickness so as not to be directly below 300. Further, the Xθ stage 2101 is provided on the substrate 251.
The structure is such that the detection position is hollowed out so that the light emitted from the back surface of 0 is not reflected.

A predetermined amount of water for detecting foreign matter is dropped on the sampling substrate 2501 and the substrate exchange means 2102 is used.
It is placed in the vacuum chamber 2103. After this, the vacuum chamber 2103 is evacuated to evaporate the water.
After that, the vacuum chamber 2103 is moved to inspect foreign matters on the substrate as described above. If the vacuum chamber is made of a material such as glass or quartz, the vacuum chamber can be inspected without moving. The vacuum here is for evaporating the water on the substrate and does not require a highly accurate object. Further, since only water is evaporated, as shown in FIG. 42B, gas such as dry nitrogen is supplied to the nozzle 2104 by the gas introducing means 2105.
It may be a means for spraying onto the sampling substrate 2501 through. In this embodiment, it is advisable to conduct an inspection before and after dropping water and use the difference between the inspection results as the number of foreign matters.

According to the present invention, unlike the conventional inspection apparatus, since the detection pixel size is sufficiently small, it is possible to distinguish and detect each foreign matter even when the foreign matter adhesion density is high. However, it is possible to detect the foreign matter with high accuracy. In the present invention, in order to improve the detection rate, it is necessary to increase the power of illumination or increase the inspection time so that the energy of light radiated on the entire surface of the wafer is sufficiently large. Specifically, to detect 38 nm particles on an 8-inch wafer with a detection rate of 95%,
180J of energy is needed. That is, it is necessary to irradiate a 1 W laser for 3 minutes. This value is not easy to implement. Hereinafter, a method for improving the detection rate with a low irradiation energy will be described.

FIGS. 43 (a) and 43 (b) show an embodiment of an illumination system capable of reducing the irradiation energy while maintaining the detection rate. As shown in FIG. 44 (a), when light is applied to a sufficiently large spherical particle, for example, the luminous flux 2 that strikes the particle
Most of 502 is reflected by particles or absorbed by particles. Here, a value obtained by multiplying the cross-sectional area of the particle by the ratio of the power of the light flux that is reflected (scattered) to the power of the light flux with which the particle is irradiated is called the scattering cross-sectional area. The scattering cross section of large particles is comparable to the actual particle size. This means that most of the light flux illuminating the particles is consumed by scattering. FIG. 44B shows an example when the particles are small. It is known that when the particles are small, the scattering cross section decreases in proportion to the sixth power of the particle diameter.
As a result, the scattered light beam 2503 becomes smaller, which means that the light beam irradiated on the particles travels straight without being consumed by the particles. That is, in the detection of ultrafine particles, which is the object of the present invention, almost no light is consumed. Therefore, we paid attention to the fact that the irradiation energy to be used can be reduced practically by repeatedly using this light flux.

Specifically, FIG. 43A shows a side view of the illumination system, and FIG. 43B shows a plan view. The wafer 803 is illuminated with a parallel beam 2203. This beam 2203 has a mirror 2201 arranged in a plane as shown in FIG.
The surface of the wafer 803 having a mirror-like surface is also used by 2202 as a mirror surface, and as a result, the region 2504 having a large area is illuminated. Here, by forming the light beam 2203 into a shape having a slightly large cross-sectional area, the region 2504 can be uniformly illuminated. However, since the light intensity is attenuated by repeating the reflection on the wafer or the mirror surface, it is necessary to correct it at the time of light detection. Further, when using such illumination, it is preferable to use a two-dimensional detector as the detector instead of using the linear sensor 1552. Further, the area 25
In order to improve the uniformity in 04, the light beam 2203 may be scanned about 1 to 2 degrees in the plane direction of the wafer during the light accumulation time of the detector.

FIG. 45 shows an embodiment of the illumination optical system for the purpose of reducing the irradiation energy. This embodiment is the illumination system shown in FIG.
By illuminating as in (c), the illuminating range from above, the illuminating range becomes larger than in FIG. 45 (b). With such a configuration, the illumination range is increased by W2 / W1. Therefore, the energy of the illumination light can be reduced by W1 / W2. In the embodiment described above, it is necessary to make the depth of focus of the luminous flux of illumination sufficiently large. However, in order to increase the depth of focus, the N.V. A. (Numerical Apertur
e) needs to be small. Specific numerical values are shown below by calculating using the equations (5) and (6).

W = λ / N. A. (Equation 5) W2 <λ / (NA) ^ 2 (Equation 6) As an example, when a pin-silicon photodiode array with 50 pixels in parallel is used, the detection area on the sample is 0.3 μm when the detection pixel is 0.3 μm. Has a length W2 of 15 μm. Therefore, the depth of focus needs to be approximately 15 μm or more. Such an illumination optical system is described in N. A. Is 0.15 and the beam width is W
Can only be squeezed to about 4 μm. When the detection pixel is 1 μm, the beam width W is 8 μm and the focal depth W2 is 50 μm. When the detection pixel is 10 μm, the beam width W is 20 μm, the focal depth W2 is 500 μm, and the beam width W is 5 when the detection pixel is 100 μm.
0 μm and the depth of focus W2 becomes 5000 μm. In FIG. A. Figure 6 shows a diagram showing depth of focus and resolution for. Therefore, an illumination optical system with higher efficiency can be realized as the pixel size can be increased. Further, from this figure, it is efficient to use a pixel size close to the resolution of the illumination and arrange the number of pixels corresponding to the length corresponding to the depth of focus at that time.
For example, when using a pixel size of 3 μm, N. A.
Is about 0.2, and 15 μm, that is, 5 pixels can be arranged in parallel. If more pixels are to be arranged in parallel, focus should be taken by another method, for example, by using a cylindrical lens to make the beam elliptical and tilting this lens so that the illumination is most focused on the wafer surface. It is necessary to increase the depth.

There is also a need to observe a wafer inspected for foreign matter. In this case, the position of the foreign matter is inspected in advance, the wafer is cut into small pieces including the vicinity thereof, and the cut wafer is placed on the xθ stage 2101 and further inspected. Since this embodiment has a mechanism capable of measuring the coordinates of the stage, it is possible to measure the coordinates of the edge portion of the small piece and measure the coordinates from this coordinate to the foreign matter. This information can be used to bring the small piece into a high-precision electron microscope or the like for observation after the foreign matter inspection. Of course, at this time, if the position where the foreign matter is detected can be cut while recognizing the coordinates with high accuracy, the cut small piece may be brought in as it is. Needless to say, the above cutting is not necessary if an electron microscope (SEM) that can be placed without cutting the wafer is used.
Even in this case, it is necessary to know the coordinates of the detected foreign matter. Further, in the present invention, the accuracy of the coordinates that can be reproduced is 1 μm.
m to at most 0.3 μm. Therefore,
SEM observation is not possible in high magnification mode. Therefore, even if the foreign matter cannot be confirmed at a low magnification, the coordinate position of the foreign matter can be detected by the device of the present invention.
It is possible to take a photograph by EM, observe foreign matter in the photograph, increase the magnification, and observe at a high magnification.

At this time, it is preferable to use a two-dimensional image pickup device instead of the linear sensor 1552. Of course, in the case of a manual inspection, a two-dimensional image pickup device is cheap and is not an essential one. Therefore, there is no problem even if the linear sensor 1552 is used. Further, according to the present invention, not only fine particles but also scratches and defects to the same extent as fine particles formed on the surface can be detected and their distribution can be known. The scratches include linear scratches during wafer polishing, crystal defects due to foreign substances mixed in during crystal pulling, and the like. These scratches and defects are inspected before processing to the wafer, and the gate breakdown voltage and the like are increased by using the device in the state where there are sufficiently few scratches and defects to manufacture the device, and not only can the reliability of the device be improved, A thinner gate oxide film can be realized, and a device with higher speed and higher integration can be realized.

Another method of using the present invention will be described. In the present invention, as described above, the detected scattered light intensity is related to the unevenness of the wafer surface. Therefore, by detecting the scattered light intensity, the pitch p of the surface irregularities of the wafer and the maximum height h can be known. As a result, it is possible to inspect the uneven defects and the wafer polishing. According to FIG. 27, when the scattered light from the surface unevenness is measured while changing the incident angle of illumination, the scattered light intensity is (c
osθ) ** (2 · p / λ). Therefore, the pitch p of the surface irregularities can be known by measuring the scattered light intensity while changing the incident angle and calculating the rate of change of the slope of the curve. The scattered light intensity is proportional to the square of the maximum height h. Therefore, the maximum height h can be known from the scattered light intensity.

FIG. 1 shows a configuration for achieving the above object.
It is necessary to add incident angle setting means 1535 to the illumination optical system of the sixth embodiment. As a usage method, the measurement position 80
There is a method of fixing the value to 3 and measuring the scattered light intensity with respect to the illumination angle. This method has a pitch p and a maximum height h.
Can be measured at the same time, but has the disadvantage that the measurement time is long. Further, there is a method in which the above-mentioned measuring method is used only at one specific measuring position and the incident angle is fixed at other points to scan the stage. According to this method, the pitch p and the maximum height h can be measured simultaneously at one point, and then the changes in p and h can be measured at high speed, which is considered sufficient for inspecting the surface unevenness of the wafer.

Further, in this case, since it is desired to positively detect the light reflected from the wafer surface, it is better to use S-polarized light having a large reflectance. It has been known that when S-polarized light is used, a stable detection signal can be obtained even when the incident angle θ of the illumination is changed, since a phenomenon such as the existence of a Blister angle does not occur.

Further, in the detection of minute foreign matter, it is known that the scattering from the minute foreign matter follows the Mie theory, so it is desirable to use P-polarized light so that more light is scattered in the detection direction. According to the Mie theory, when light is irradiated to fine particles, the light is scattered uniformly in the plane parallel to the polarization direction of the light, and the light decreases in the direction perpendicular to the incident direction in the plane perpendicular to the polarization direction. Scatter with a wide distribution. Therefore, in the device of the present invention, when P-polarized light is used, more scattered light from the foreign matter is detected. Of course, it is also possible to irradiate S-polarized light and place the detector at a position where the emission angle is large within the emission surface perpendicular to the incident surface. In this case, for the convenience of assembling the imaging optical system, it is preferable to form the light beam in an elongated shape on the straight line where the incident surface and the wafer intersect.

Similar effects can be obtained by detecting polarized light which is incident from above (incident angle θ = 0 °), is parallel to the polarization plane of the illumination light, and has a large emission angle. That is, more scattered light from foreign matter can be detected by the Mie theory, while scattered light from the wafer surface isotropically emitted in each direction, so scattered light from foreign matter can be selected more selectively.
That is, it can be detected with a high discrimination ratio.

As described above, the larger the incident angle of illumination, the smaller the scattered light from the surface irregularities. The data compressed as described above is displayed, for example, as the uneven distribution on the wafer. At this time, in some cases, it is better to divide the unevenness into several gradations and display the colors in several stages. Also, by displaying the locations (positional coordinates) determined as minute foreign matter exceeding the threshold value like a map, the defects of surface irregularities and the distribution of minute foreign matter can be similarly understood. In some cases, it is better to display the data of the minute foreign matter or the surface unevenness separately, not simultaneously.

According to FIG. 31, the incident angle θ of the illumination is about 60 to 70 degrees, and the discrimination ratio becomes maximum, and it can be said that it is optimum for carrying out the foreign matter inspection. Further, in the case of a foreign substance that is not spherical, this is not the only case, and the incident angle is preferably larger than that in FIG.
Further, by illuminating the illumination light with p-polarized light (light whose magnetic field vector is parallel to the incident surface) and illuminating it with a Blewster angle (angle at which the reflectance is 0), scattered light from the surface irregularities is minimized. be able to.

"The phase of the reflected light changes depending on the surface irregularities and the fine particles. Here, the height of the fine particles is usually 10 times larger than the surface irregularities. Therefore, the phase of the scattered light photon is measured. By doing so, it is possible to know whether the detected photons are scattered from the surface of the wafer or scattered from fine particles.This is not a stochastic number of photons but a certain phase change. Therefore, in the method of detecting the number of scattered light photons described above, about 10 or more photons are required as the number of detected photons in order to increase the detection rate, whereas in the method of detecting the phase change,
Even with one photon, it is possible to discriminate between scattered light from fine particles and scattered light from surface irregularities. Is rejected by the uncertainty principle based on quantum mechanics. According to the uncertainty principle, "It is impossible to measure the momentum and position of fine particles accurately at the same time. How can we not accurately measure the photon number and phase of light at the same time? (Yamamoto; Quantum Optics and New Technology [I]; The Institute of Electronics, Information and Communication Engineers, 7
Volume 2, Issue 6, pp. 669-675) ”Therefore, the number of photons of a specific number or more is required for the detection described in the foreign matter inspection apparatus of the present invention.

The above-mentioned method for mass production startup of a semiconductor process and a foreign matter inspection method for a mass production line and its apparatus are evaluated and improved (debugging) in materials, processes, devices, designs, etc. at the time of mass production startup.
To evaluate each process, equipment, etc. with expensive and high-performance evaluation equipment, reduce the production line equipment as much as possible during mass production, especially reduce the items of inspection and evaluation, and require equipment cost and inspection and evaluation Try to save time. For that purpose, the foreign matter detection analysis system that devised the sampling wafer such as polishing the surface with high accuracy so that the evaluation at the time of mass production startup can proceed smoothly and quickly can be used to investigate the cause of the foreign matter generation and inspect it at the time of material acquisition. Or changing the dust source of the equipment and feeding back the result to each material, process, equipment, etc. to make the dust-prone process specifications the design specifications of the element that is strong against dust generation. At the same time, it is used to make specifications for mass production line inspections and evaluations, and if necessary, install a foreign matter (dust) monitor at the location where foreign matter is likely to occur, or monitor the increase or decrease of specific foreign matter at a specific location. Or

By separating the mass production line from the mass production start-up as described above, foreign matter detection, analysis, and
The evaluation device can be operated efficiently, mass production can be started up quickly, and foreign materials (dust generation) used in the mass production line can be inspected, and the evaluation equipment can be used as a simple monitoring device with the minimum necessary to make the mass production line lightweight. Be promoted. In addition, by devising the sampling wafer at the time of mass production startup, the sampling interval can be shortened and the sampling time can be shortened to collect more accurate foreign particle generation data. The raising period can be shortened.

Further, the STM / STS technique used for the analysis of the foreign elemental species at the time of starting up the mass production has existed in the past, but it is said that this technique cannot determine the elemental species of the sample, and in the manufacture of the LSI. Although not used, the present inventors focused on the fact that foreign substances generated in LSI manufacturing are limited, and noticed that they can also be applied to the conventional technology of STM / STS, and dust can be generated on the production line. A system that accumulates STM / STS spectra of active elements in a database and enables analysis of dust by comparing with the data of the inspection target. It can be used for evaluation and countermeasures.

[0111]

According to the present invention, a mirror-finished substrate (mirror-finished wafer) having a minute unevenness with a height of 0.7 nm or less is flown in a semiconductor mass-production line, and 0.05 μm or less existing on the mirror-finished substrate. By enabling the detection of foreign matter consisting of fine particles and maximizing the functions of detecting, analyzing, and evaluating foreign matter necessary at the time of mass production startup, the feedback to the mass production line can be smoothly advanced, and the LSI mass production line This has the effect of enabling a high yield to be achieved early.

Further, according to the present invention, it is possible to evaluate the minute unevenness of the mirror-finished substrate which flows in the semiconductor mass production line.

[Brief description of drawings]

FIG. 1 is a configuration block diagram showing an embodiment of a method for initiating mass production in a semiconductor manufacturing process and a foreign matter inspection method for a mass production line and an apparatus therefor according to the present invention.

FIG. 2 is a configuration perspective view showing an embodiment of the sampling unit of FIG.

FIG. 3 is a configuration block diagram showing an embodiment of a detection unit in FIG.

FIG. 4 is a block diagram showing the configuration of an embodiment of the analysis unit of FIG.

5 is a perspective view showing a mirror surface wafer of the sampling wafer of FIG. 1. FIG.

6] Si3N4, Pol of the sampling wafer of FIG.
It is a perspective view which shows a y-Si and Al film formation wafer.

FIG. 7 is a perspective view showing a patterned wafer of the sampling wafer of FIG.

8 is a perspective view showing a lapping direction and a scanning direction of the sampling wafer of FIG.

9 is an xy distribution diagram of the work function of the sample of FIG.

10 is a cross-sectional view of the sample of FIG.

11 is a diagram showing the relationship between the distance between AFM tip samples and the force between base papers in FIG.

12 is a relationship diagram of α, β, and φ in FIG.

FIG. 13 is a diagram showing the relationship between the tunnel current and the force between base papers in FIG.

FIG. 14 is a configuration block diagram showing an embodiment of an in-vacuum dust generation monitor of the sensing unit of FIG.

15 is a configuration perspective view showing another embodiment of the in-vacuum dust generation monitor of the sensing unit of FIG. 1. FIG.

FIG. 16 is a side view showing the first embodiment of the present invention.

FIG. 17 is a diagram showing an example of a detection signal according to the present invention.

FIG. 18 is a diagram showing the necessity of fine particle detection according to the present invention.

FIG. 19 is a diagram showing the positioning of a foreign matter inspection on a mirror surface wafer according to the present invention.

FIG. 20 is a diagram showing a technical problem of ultrafine particle detection according to the present invention.

FIG. 21 is a diagram showing calculation results of scattered light from air molecules according to the present invention.

FIG. 22 is a diagram showing scattered light noise from the unevenness of the wafer surface according to the present invention.

FIG. 23 is a diagram showing a surface scattering model using reflection according to the present invention.

FIG. 24 is a diagram showing a light diffraction model of a surface uneven shell according to the present invention.

FIG. 25 is a diagram showing a simulator based on a diffraction model according to the present invention.

FIG. 26 is a diagram showing the relationship between the uneven pitch and the scattered light intensity according to the present invention.

FIG. 27 is a diagram showing a relationship between an incident angle and scattered light intensity according to the present invention.

FIG. 28 is a diagram showing the relationship between the maximum height and the scattered light intensity according to the present invention.

FIG. 29 is a diagram showing a relationship between irradiation light wavelength and scattered light intensity according to the present invention.

FIG. 30 is a diagram showing a scattered light level from the wafer surface according to the present invention.

FIG. 31 is a diagram showing a discrimination ratio with standard fine particles according to the present invention.

FIG. 32 is a diagram showing a problem of fluctuation at the time of detecting weak light according to the present invention.

FIG. 33 is a diagram showing a detection rate and a necessary irradiation light amount according to the present invention.

FIG. 34 is a diagram showing a schematic configuration of an apparatus according to the present invention.

FIG. 35 is a diagram showing a pixel miniaturization and an optical noise reduction effect by a high incident angle according to the present invention.

FIG. 36 is a diagram showing detection output of fine particles according to the present invention.

FIG. 37 is a diagram showing a decrease in detection rate due to fluctuations in photons according to the present invention.

FIG. 38 is a diagram showing calculation results of detection rates according to the present invention.

FIG. 39 is a diagram showing device specifications and detection rates according to the present invention.

FIG. 40 is a diagram showing a relationship between a pixel size and a detection rate according to the present invention.

FIG. 41 is a block diagram showing another embodiment of the present invention.

FIG. 42 is a block diagram showing an embodiment for implementing the sampling method according to the present invention.

43A and 43B are a side view and a plan view showing an embodiment for carrying out an illumination method according to the present invention.

FIG. 44 is a diagram for explaining the interaction between fine particles and light according to the present invention.

FIG. 45 is a diagram showing an illumination method according to the present invention.

46 is an N.V. of the imaging optical system according to the present invention. A. 3 is a diagram showing the depth of focus and the resolution for FIG.

[Explanation of symbols]

100 ... Semiconductor manufacturing equipment group, 200 ... Sensing unit 204 ... Vacuum dust monitor, 300 ... Utility group,
400 ... Sampling unit 401-405 ... Sampling wafer, 500 ... Detection unit, 600
… Analysis unit 603… STM / STS, 700… Compatible system 1000… Mass production start-up and mass production line particle inspection system for semiconductor manufacturing process 1001… Online particle inspection system, 1002… Offline particle inspection system 531,532… Semiconductor laser, 535 , 536… Focusing objective lens 537… Detection objective lens, 538… Detector, 571… Binarization circuit 572… Stage controller, 581… Interface room 631… AFM chip, 633… STM chip, 635… STMXYZ
Fine movement unit 642… Sample STM coarse drive unit, 1530… Illumination optical system,
1531 ... Ar laser 1532 ... Beam expander, 1533 ... Cylindrical lens 1534 ... Polarization filter, 1540 ... Detection optical system, 1520 ... Photon multiplier tube 1521 ... Photoelectric conversion surface, 1522 ... Multi-channel plate, 1523 ... Fluorescent plate 1550 ... Detection Section, 1551 ... Imaging lens, 1552 ... Linear sensor 1560 ... Stage section, 1570 ... Signal processing section, 1524 ... Applied voltage controller 1573 ... Foreign matter memory

Claims (16)

[Claims] 1. 0.5 is provided at a predetermined position of a semiconductor manufacturing line.
A mirror substrate having gradual fine irregularities having a height of 0.7 nm or less at a pitch of μm or less is flown, laser light is obliquely irradiated to the mirror substrate, and scattered light from the mirror substrate is irradiated. The light is condensed by a condenser lens to form an image, and the photoelectric conversion means converts it to a specular substrate and photoelectrically converts it in pixels of 1 μm or less to detect a signal corresponding to the number of scattered photons. A method for inspecting a mirror-like substrate in a semiconductor manufacturing line, which comprises detecting fine particles of 0.05 μm or less attached to the top by discriminating them from minute irregularities on the mirror-like substrate. 2. A method for inspecting a mirror surface substrate in a semiconductor manufacturing line according to claim 1, wherein the power of the laser light is 20 mW or more. 3. The method for inspecting a mirror-finished substrate in a semiconductor manufacturing line according to claim 1, wherein the photoelectric conversion means comprises a photomultiplier tube. 4. The method for inspecting a mirror-finished substrate in a semiconductor manufacturing line according to claim 1, wherein the oblique direction is 40 ° or more with respect to the vertical direction. 5. 0.5 is provided at a predetermined position of a semiconductor manufacturing line.
A mirror substrate having gradual fine irregularities having a height of 0.7 nm or less at a pitch of μm or less is flown, laser light is obliquely irradiated to the mirror substrate, and scattered light from the mirror substrate is irradiated. The light is condensed by a condenser lens to form an image, and the photoelectric conversion means converts it to a specular substrate and photoelectrically converts it in pixels of 1 μm or less to detect a signal corresponding to the number of scattered photons. The fine particles of 0.05 μm or less attached on the top are discriminated from the fine irregularities on the mirror substrate to be detected,
A method for inspecting a mirror surface substrate in a semiconductor manufacturing line, characterized in that a generation state of fine particles on the mirror surface substrate is detected. 6. The method for inspecting a mirror-like substrate in a semiconductor manufacturing line according to claim 5, wherein the generation state of fine particles with respect to the detected position of the mirror-like substrate is displayed. 7. A laser beam is emitted from a mirror substrate with a pitch of 0.7 μm or less and having a minute unevenness with a height of 5 nm or less, and the laser beam is emitted obliquely from the mirror substrate. The scattered light is condensed by a condenser lens to form an image, which is converted to 1 μm on the mirror substrate by the photoelectric conversion means.
The following pixels are photoelectrically converted to detect a signal corresponding to the number of scattered photons, the state of the fine irregularities is detected by the signal to evaluate the mirror surface substrate, and the evaluated mirror surface substrate is flowed to the semiconductor manufacturing line. A method for inspecting a mirror substrate in a semiconductor manufacturing line, which is characterized by the above. 8. The method of inspecting a mirror-finished substrate in a semiconductor manufacturing line according to claim 7, wherein the laser light has a power of 20 mW or more. 9. The method for inspecting a mirror-finished substrate in a semiconductor manufacturing line according to claim 7, wherein the photoelectric conversion means comprises a photomultiplier tube. 10. The oblique direction is 40 ° with respect to the vertical direction.
The above is the method for inspecting a mirror-finished substrate in a semiconductor manufacturing line according to claim 7. 11. The method of inspecting a mirror-finished substrate in a semiconductor manufacturing line according to claim 7, wherein the state of occurrence of minute irregularities with respect to the detected position of the mirror-finished substrate is displayed. 12. A mounting means for mounting a mirror-finished substrate having gentle minute unevenness of 0.7 nm or less at a pitch of 0.5 μm or less, which is flown to a predetermined portion of a semiconductor manufacturing line, and Illuminating means for illuminating the mirror surface substrate mounted by the mounting means from an oblique direction, and scattered light scattered from the mirror surface substrate is condensed by the detection optical system and converted into 1 μm on the mirror surface substrate.
foreign matter formed by fine particles of 0.05 μm or less on a mirror substrate based on a signal detected by a photodetector of the detecting means. The inspection device for a mirror surface substrate in a semiconductor manufacturing line, wherein the inspection is performed by discriminating from the minute irregularities. 13. A photoelectric conversion means of the detection means, a photoelectric conversion surface, a multi-channel plate that multiplies photoelectrically converted electrons, a fluorescent plate that emits fluorescence by the multiplied electrons, and detects the generated fluorescence. 13. The inspection device for a mirror-like substrate in a semiconductor manufacturing line according to claim 12, wherein the inspection device is a solid-state image pickup element. 14. A step of polishing and polishing, and flowing a mirror-finished wafer having gradual fine irregularities with a height of 0.7 nm or less at a pitch of 0.5 μm or less on a semiconductor mass production line.
Laser light from an oblique direction is irradiated to a mirror surface wafer obtained from a predetermined manufacturing apparatus, the scattered light from the mirror surface substrate is condensed by a condenser lens to form an image, and the light is converted onto the mirror surface substrate by photoelectric conversion means. A signal corresponding to the number of scattered photons is photoelectrically converted by a pixel having a size of 1 μm or less, fine particles having a size of 0.05 μm or less are discriminated from the fine irregularities by the signal, and the detected fine particles are generated. And a monitoring step of performing feedback with a simple monitoring device based on the generation state of the fine particles obtained in the inspection step in mass production. 15. The semiconductor manufacturing method according to claim 14, wherein in the inspection step, the element species of the fine particles are analyzed by STM / STS to determine the cause of the fine particle generation. 16. The semiconductor manufacturing according to claim 14, wherein in the inspection step, the elemental species of the fine particles are analyzed by STM / STS to determine the factors causing the fine particles due to the material, process, apparatus, environment and the like. Method.