JPH0868772A - Apparatus and method for automatic mask inspection by using electron beam microscopy - Google Patents

Abstract

PURPOSE: To automatically inspect a wafer, a X-ray mask, a substrate, etc., at a manufacturing site by providing an electronic beam generator to generate electronic beam. CONSTITUTION: An inspecting system 10 has two types of operation modes, a die-to-die comparison mode and a die-to-database comparison mode. In both modes, defectives are detected by comparing the image of electronic beam obtained by scanning a substrate 57 with a reference. The substrate 57 to be inspected is automatically placed on a X-Y stage 24 under an electronic beam generator 20 by a substrate handler 34. The electronic beam generator 20, an alignment optical system, an analog deflection circuit 30, a detector 32 are used to inject electronic beam to the substrate 57 and detect secondary electrons, backward scattering electrons and transmitted electrons. For such detecting operation and data collection, an electronic beam generator control computer 42, a video frame buffer 44, an image collecting preprocessor 48 are used.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an automatic inspection technique for various types of mask substrates used for manufacturing ultrafine semiconductor integrated circuits, and in particular, for phase shift masks using an electron beam.

[0002]

2. Description of the Related Art In order to produce an ultrafine semiconductor integrated circuit with an appropriate yield, a mask or a wafer used in a manufacturing process must be free from defects. Many systems have been developed and patented for automated inspection of optical masks and wafers over the last 12 years (eg, US Pat. No. 4,247,203, US Pat. No. 4,800).
5,123, U.S. Pat.No. 4,618,938, U.S. Pat.No. 4,845,55
(See No. 8). These systems compare two adjacent dies on a photomask or reticle or wafer to each other. Similarly, techniques have been developed for inspecting dies by comparing them to CAD (Computer Aided Design) databases (US Pat. No. 4,926,48).
(See No. 7).

However, since the defects of the X-ray mask cannot be detected in the visible or ultraviolet spectrum, the application of any of the above optical systems is limited to the optical mask. Also, optical inspection has limited resolution due to inherent light diffraction and interference, so linewidths below 0.35 microns cannot be achieved with simple optical lithography techniques. Linewidths of less than 0.35 microns are expected to be achievable with X-ray lithography technology.

Due to the progress of optical lithography in the manufacture of integrated circuits, the miniaturization thereof is becoming more and more advanced. For example, in the fabrication of semiconductor devices of 256 megabit DRAM wafer linewidth, the wafer linewidth is 0.25 to 0.35 microns. Phase shift mask technology is used in the manufacture of semiconductor devices of this line width.

Typically, these masks have a patterned chrome layer on a quartz substrate with wells that are selectively etched on the substrate to provide a phase shift for the light beam of exposure. ing. The phase shift is also caused by the optically transparent material pattern formed on the quartz substrate or the chrome layer of the phase shift mask.

Inspection of a mask having a fine pattern formed on the surface of the substrate, such as the type of mask referred to above, includes:
Whether or not it is necessary, it is necessary to detect and search for defects in the chromium layer pattern as well as to measure the depth of the trench or well formed in the quartz substrate. Furthermore, it is necessary to know the presence or absence of defects on quartz in relation to the technology of phase shift material. The reason for this is that quartz itself plays an important role in the technology of phase shift materials.

In the prior art, since the phase shift mask is typically transparent to light, it has been designed to be inspected using optical technology. Such optical inspection technology is often proved insufficient for the inspection of the latest mask having a fine pattern because the optical inspection apparatus used so far has a limit.

As has been made clear above, the present invention mainly relates to the structure and use of a set of apparatus for inspecting an X-ray mask, a phase shift mask and a wafer. In the past, the only deficiency of scanning an electron beam with such a device was the inability to measure the amount of optical phase shift or shift with depth of the well or trench on the known material of the mask or wafer. Is.

[0009]

As a preferred embodiment of the present invention, a method and apparatus for automatically inspecting a substrate by scanning charged particles will be described below. Particularly, for the purpose of inspecting the phase shift mask, an electron optical beam generation control tower, that is, a means relying on an electron beam generated in the column is used. The first embodiment is an apparatus and method for automatically inspecting a substrate, which includes a charged particle beam generator for radiating and scanning a charged particle beam on the substrate surface, and secondary electrons generated from the top or bottom surface of the substrate. The backscattered electrons are detected, their electric signal waveforms are obtained, and various analysis processing is performed on the signal waveforms to identify "what deal?" Is the pattern element on the mask. And die-to-
Check for defects by die or die-to-database method.

Further, as a preferred embodiment of the present invention, a method and apparatus for measuring the depth of the well of the phase shift mask will be described. As described above, in the phase shift mask, the well that gives a phase shift, that is, a shift, to the exposure light beam plays an important role. Of course, when measuring the depth of the well, it is needless to say that if there is a defect, it will hinder the depth measurement, so that the defect can be searched for even in this process.

Moreover, the present invention makes it possible not only to detect the presence or absence of the phase shift material, but also to know its thickness information.

[0012]

According to the present invention, it is possible to economically realize an inspection apparatus for automatically inspecting a wafer, an X-ray mask, a substrate or the like at a manufacturing site by using a charged particle beam. Hereinafter, the present invention will be described using an electron beam,
The scope of the invention is not limited to electron beams, as other types of charged particle beams can be used.

Although the present invention is mainly used for inspection of wafers, optical masks, X-ray masks, electron beam proximity masks, stencil masks, etc., it can be used for high speed electron beam imaging of any material. In addition, it can also be used for electron beam writing for exposing photoresist in the manufacture of masks and wafers.

Electron beams have two basic modes of operation, depending on whether the substrate is an insulator or a conductor. The "high voltage mode" is mainly used for inspection of X-ray masks, electron beam proximity masks, and wafer prints that are conductive or coated with a conductor. In this case, the substrate will not be charged using the high voltage scanning beam.

On the other hand, the "low voltage mode" is mainly used for inspection of wafers and optical masks having a non-conductive material layer during manufacturing. Charging and damage can be minimized by using a low voltage scanning beam. Except for the above differences, both modes achieve high speed defect detection and classification.

Current scanning electron microscopes have very slow scanning speeds, and require an operator with a high level of skill beyond the normal range. It is not possible to use a microscope.

A novel feature of the present invention is that not only can various types of defects be detected, but the types of defects can be identified. In the present invention, backscattered electrons in the "high voltage mode",
Since it is possible to simultaneously detect and identify transmitted electrons and secondary electrons, defects can be immediately classified. For example, defects detected only by the transmission detector on the X-ray mask are probably crevices in the absorbing material, and defects detected by the secondary electron detector but not by the backscattered electron detector are organic particles. Defects that are likely to be detected by the backscattered electron detector may be high atomic weight contaminants. Although some defects, such as organic contaminants on the X-ray mask, are not printed on the wafer, the ability to identify various types of defects is a significant advantage of the present invention.

Thus, the present invention is capable of not only detecting defects, but also identifying those defects. The present invention uses a number of techniques to make the system suitable for semiconductor manufacturing. For example, both the evacuation speed and the pressurization speed at which the vacuum is returned to normal pressure are both limited to keep the gas flow in a laminar state, thereby preventing the disturbance of contaminants. It also saves time by performing these operations concurrently with scanning other samples.

In addition to this, six field emission sources are provided in the turret to further reduce the wasted time. Finally, the major adjustments of the electron beam, which are usually performed by operator scanning, are computer-implemented, so that relatively unskilled persons can use the system of the present invention.

[0020]

1 is a block diagram of an overall inspection system 10 of the present invention. The inspection system 10 is an automatic inspection device for X-ray masks, wafers, and other substrates.
A scanning electron microscope is used as a sensor.

The inspection system 10 has two kinds of operation modes, that is, a die-to-die comparison mode and a die-to-database comparison mode. In any mode, the defect is detected by comparing the electron beam image obtained by scanning the substrate with a reference. That is, in die-to-die (die-to-die) comparison, signals from two dies on the same substrate are compared to each other. Die-to-database comparison inspection compares the signal from one die obtained from an electron microscope with the signal from the database used to make that die.

The substrate 57 to be inspected is held by the holder, and the holder is automatically placed on the XY stage 24 below the electron beam generator 20 by the substrate handler 34. This operation is achieved as follows. Instructions are sent from the system computer 36 to the substrate handler 34. The substrate handler 34 is a substrate 57 to be inspected.
Is removed from the cassette, and the flat portion or notch 59 (see FIGS. 2 to 3a) formed on the substrate 57 is automatically detected to properly orient the substrate 57, and To load.

Next, the operator uses the alignment optical system 2
While visually observing the substrate 57 via 2, the substrate alignment point is determined (the element pattern of the substrate is arbitrarily selected to be the alignment point), and the movement of the stage in the X-axis direction is the substrate pattern. Be substantially parallel to the X axis of the inspection area of. This completes the rough alignment work.

Following the rough alignment work, a precise alignment work is performed. The precise alignment operation is performed while the operator scans the substrate with the electron beam and observes the image displayed on the image display 46. All of the data related to the alignment are the alignment computer 2
Stored in 1. This alignment computer 21
Works in cooperation with system computer 36 to move the die
The actual composite X, Y needed to scan along both X, Y axes
Calculate motion. Therefore, it is not necessary for the operator to perform the alignment work for the same type of substrate thereafter.
When the precise alignment work of the substrate is completed, the inspection process is started.

The electron beam generator 20, the alignment optical system 22, the analog deflection circuit 30, and the detector 32 cause the electron beam to enter the substrate 57, as described in detail below.
Also, secondary electrons, backscattered electrons, and electrons transmitted through the substrate 57 are detected. This detection operation and data collection are
Electron beam generator control computer 42, video frame buffer 44, image acquisition pre-processor 48,
This is performed by the deflection controller 50 and the memory block 52. The VME bus, or VEM1 designated by the numeral 29, functions as a communication link between subsystems.

The position and movement of the XY stage 24 during the inspection of the substrate 57 is controlled by the stage servo 26 and the interferometer 28 under the control of the deflection controller 50, the memory block 52 and the alignment computer 21. To be done.

In the die-to-database compare mode, the database adapter 54 in communication with the memory block 52 is used as a source of signals representative of the intended die format.

The actual defect detection process is performed on the data in the memory block 52 by the post processor 58 and the defect processor 56. Communication between the post processor 58 and the defective processor 56 is done via the bus VME 2 shown at 31.

The scanning of the entire system is
While communicating with other blocks via the data bus 23, which is similar to the Ethernet bus,
It is executed by the computer 36, the user keyboard 40, and the computer display 38. Ethernet is a trademark of Xerox Corporation.

FIG. 2 shows the scanning trace of the present invention when the inspection is performed in the die-to-database comparison mode. Only one die 64 is shown on the substrate 57 in FIG. The die 64 has an inspection area 65 to be inspected. The inspection area 65 is an area in which important information is recorded on the substrate 57. When inspecting the die 64, the effective scanning movement in the X-axis direction moves. XY stage 2
4, the effective scanning movement in the Y-axis direction is performed by deflecting the electron beam with the same swing width as the width of the scanning region indicated by reference numeral 60 in the drawing. Electron beam is die 6
When it reaches the right side of 4, the XY stage 24 moves in the Y-axis direction by a distance less than the swing width of the electron beam. Board 57
The XY coordinate system is based on the XY stage 24 and the electron beam generator 2.
Since it does not exactly match each XY coordinate system of 0, X-
The actual movement of the Y stage 24 and the actual beam deflection of the electron beam generator 20 are controlled by the X-axis during scanning of the die 64.
And has a component of Y.

In order to sufficiently inspect the inspection area 65,
The inspection is performed by drawing a curved trajectory 62 as shown. Each trajectory in the X-axis direction of the bent locus 62 is a scanning area having the same width as the scanning area indicated by reference numeral 60, and each of them slightly overlaps the adjacent scanning area.

In the die-to-database compare mode, the signal corresponding to each scan area is compared with the simulated signal from the database adapter 54 for the corresponding scan area of the perfect die. This process is repeated for each scan area of the inspection area 65 of the die currently inspected before proceeding to the inspection of the next die.

FIG. 3 shows scanning trajectories when inspecting in the die-to-die comparison mode, and a substrate 57 having dies 68, 70, 66 from left to right is shown as an example. There is. In this inspection mode as well, the inspection is executed by drawing a curved locus 63 as in the example of FIG. However, since this inspection mode is a die-to-die comparison mode, the XY stage 24 continues to move in the X-axis direction until it crosses three dies for each scanning area, and only after it crosses three dies. Move in the Y-axis direction. In this comparison mode, the data obtained in the first scanning step of die 68 is stored in memory block 52 and this stored data is transferred to die 7
It is compared to the data obtained during the first scan step of zero.
While comparing die 68 and die 70, the data of die 70 is stored in memory block 52 for comparison with the data obtained in the first scanning step of die 66.

Next, the second scanning process is started. Since the second scan step is a return scan step, the order of passing through the die is reversed and the data obtained by the second scan step of die 66 is stored for comparison with the data obtained from die 70, The data obtained from die 70 is stored for comparison with the data obtained from the second scanning step of die 68. A series of operations of this scanning and comparison are repeated to inspect the entire inspection area of the substrate 57.

It may be necessary to obtain images with multiple scan integration techniques. In this case, each pixel must be exposed for a long time interval. Conventional scanning microscopes typically utilize slow scanning techniques where the beam has a long pixel dwell time before transitioning to the next pixel. However, in the substrate inspection system, the heating and charging of the substrate are not preferable, so that the slow pixel recording speed is not desirable.

It may be necessary to integrate multiple scans to obtain an image with sufficient contrast or to improve the signal-to-noise ratio of the image. In order to improve the signal-to-noise ratio, it is necessary to average the signal values obtained for each pixel by scanning the same position on the substrate several times. The image contrast in the "low voltage mode" (this mode will be discussed in detail in the "Electronic Optics" section below) also affects the return period when the electron beam returns to the position of a specific pixel on the substrate. It can be improved by scanning the substrate. Improving contrast in low voltage inspection of non-conductive substrates
It can also be achieved by replacing the electrons at a particular pixel location with secondary electrons that occur when a nearby area is scanned while the beam is returning, as described in the "Electronic Optics" section. . In addition, for heat-sensitive substrate materials, the pixel locations are beam-scanned in intervals to dissipate the heat accumulated by the beam.

FIG. 4 is a diagrammatic representation of an example of the scanning method adopted by the present invention. This figure shows a method of averaging the signals by scanning a series of rectangles of 512 columns by m (where m is a multiple of 4) rows of pixels by deflecting the beam four times. The center of each of the rectangles moves by m / 2 pixels along the moving direction of the stage.

FIG. 4 shows an example of the overlapping frame scanning technique that the present invention employs for signal averaging, contrast enhancement, and heat dissipation. In the example shown, each pixel is scanned 4 times. Each scan line has a length of 512 pixels in the Y-axis direction. A series of m side-by-side lines 1 ....... m for overlapping frame scanning
Are scanned on the substrate. The X-axis interval between lines is set equal to the pixel size, and the X coordinate of each line increases continuously.

FIG. 5 is a diagrammatic representation of the nominal deflection value of the beam in the X direction during the scan shown in FIG. 4 as a function of time. The horizontal direction is the time axis, and the vertical direction is
The position on the X axis. FIG. 5 shows the staircase output of the X-axis deflection system used to deflect the beam. After scanning m lines, the scan is retracted in the X-axis direction, as shown in FIG. The stage for moving the substrate under this deflection system is such that the position of the next scanning line corresponds to the number of the first m lines (m / 4 + 1) when the beam is retracted in the X-axis direction. , The speed in the X-axis direction is adjusted. Since the beam is scanned 4 times in this example, when the beam scans a rectangle of 512 pixels m pixels 4 times, the stage moves the substrate in the X-axis direction by a distance of m pixels.

FIG. 6 is a graphical representation of the X coordinate of the beam on the substrate during the scan shown in FIG. 4 as a function of time. The horizontal direction is the time axis, and the vertical direction is the position of the beam on the X axis. FIG. 6 shows the X coordinate of each of a series of scan lines on the substrate as a function of time. It has a stage that moves the substrate under the deflection system,
The beam is shown to scan the position of each line on the substrate four times in combination with a deflection system that moves the scan line back and forth along the X axis within the deflection region. By recording the image data in memory block 52 and averaging the data from the appropriate memory addresses, the averaged data can be provided to defect processor 56 and alignment computer 21. Although 4 is used as the number of averagings in this example, the number of scans actually combined and the number of lines per frame m are selected so as to produce the best combination of noise reduction, contrast enhancement, and inspection efficiency. To do.

Scanning in the Y-axis direction, which is perpendicular to the moving direction of the stage, is the same as scanning used for imaging by one pass.
Here the scan advances one pixel D every exposure interval t. For 512 pixels, to image the field scanning area in one pass, set the stage speed to (D / 512 / t) and advance the stage by one pixel for each scan. To do so. When imaging is repeated several times, square pixels cannot be recorded unless the scanning beam seen from the substrate also moves at a speed of (D / 512 / t) micron per second. To record an image with n pixels exposed per pass, the stage must be moved slowly at a speed less than (D / 512 / t) and the scan time (512 * t ), The beam must be advanced an extra (1-1 / n) * D micron so that the beam is scanned stepwise in the moving direction of the stage.

After the variable number of m steps, the scanning in the X-axis direction moves backward. Thus, the scanning orbit is 512 * (1-1 / n) *
It becomes the shape frame of m. When viewed from the surface of the substrate, the overlapping frame pattern shown in FIG. 4 is obtained. The interval between multiple exposure times of each image pixel is (512 * m * t). As long as m is set larger than n, both the number of pixel rescans and the repetition rate can be freely changed. By recording the image data in the memory block 52 and averaging the data from the appropriate addresses, the defective processor 56 as if recording the averaged data in one slow pass. Can be supplied to. The advantage of this technique is that the parameters can be adjusted to optimize the exposure time between pixels.

Returning to FIG. 3, the die-to-die comparison mode will be described in more detail. Electron beam is die 68 and die 7
When the 0 scan area is scanned, the signals 33 from the three types of detectors 32 shown in FIG. 1 are sent to the image acquisition pre-processor 48, where they are converted into digital signals and then stored in the memory.
Stored in block 52. Both data from die 68 and die 70 are simultaneously sent to the defect processor 56 where a significant mismatch between both data is designated as a defect. Next, the defect data from the defect processor 56 is accumulated,
This is sent to post processor 58 for integration. Post processor 58 determines the size and various characteristics of the defect and makes that information available to system computer 36 via bus 23.

In the die-to-database comparison test mode, system 10 operates in the same manner as above, except that memory block 52 receives data from one die, a reference for comparison at defective processor 56. The difference is that the data is provided by the database adapter 54.

When the entire board is inspected, a list of defects is displayed on the computer display 38 along with defect location information.
Is displayed in. The user keyboard 40 allows the operator to initiate a defect investigation. In response to this command,
System 10 scans around each defect and displays the image on display 46.

Scanning optics A combination of several main elements and a special design of the electron beam generator 20 can increase the image forming speed by about 100 times or more. Since the scanning speed is fundamentally limited due to the signal-to-noise ratio, it is essential to increase the beam current in order to increase the image forming speed. In the present invention, a high-intensity high-temperature radiation source is used to increase the beam intensity over a wide angle and increase the beam current value. However, when the electron density becomes high, a repulsive force is generated between the Coulombs, so a high electric field is applied near the cathode to rapidly expand the beam diameter. In the electron beam generator, the electron crossing that raises the charge density is prevented, and the numerical aperture is increased to reduce the problem of Coulomb repulsion.

Unless the substrate is scanned at a high rate of, for example, 100 megapixels per second, the detector will scan 2 consecutive times.
It is not possible to make a temporary identification of secondary (return) electrons originating from a single pixel. This means that the arrival time needs to be consistent compared to the dwell time of each pixel.

By accelerating the electrons as soon as they leave the target, it is possible to reduce variations in the arrival time of each pixel. By taking such measures, it is possible to maintain the variation in arrival time at the detector within about 1 nanosecond. If a reverse-biased high-frequency Schottky barrier detector is used for each type of electron to be detected, it is possible to further reduce variations in arrival time. The Schottky detector is shown only as an example, and other types of semiconductor detectors may be used.

Electron optics The electron optics subsystem is functionally similar to a scanning electron microscope and includes a scanning electron beam probe and detectors for secondary electrons, transmitted electrons, and backscattered electrons on the substrate surface. It has for imaging. During the inspection, the electron beam is scanned in one direction, and the stage is moved in a direction perpendicular to the scanning direction of the electron beam. Either low voltage secondary electrons or high energy transmitted or backscattered electrons are used to generate the video signal. The generated video signal is digitized and stored in the form of an elongated scan area image. This electro-optical subsystem not only can detect defects automatically at high resolution, but also can combine old and new technologies to quickly obtain a noise-free image at the resolution required for inspection.

The beam typically scans a field of 512 pixels (18-100 μm wide) using a very fast 5 microsecond period sawtooth sweep. The imaging characteristics are uniform in the scanning field because the deflection does not cause distortion and is almost perpendicular to the surface.

Due to the high detection efficiency, almost all of the secondary electrons generated by the electrons from the probe can be used for image formation. The bandwidth of the detection system is comparable to the pixel speed due to the short transit time. Since the secondary electrons are extracted coaxially, an accurate image of the edge shape can be obtained regardless of the direction of the edge of the observation object on the substrate.

FIG. 7 shows the elements of the optical system and the associated power supplies needed to understand their function. The electron gun includes a thermal field emission cathode 81 and an emission control electrode 8
3 and an anode 85 having an anode aperture 87
It consists of and. The cathode 81 has a power source 89 of 20 Ke.
It is held at the beam voltage of V 2. The amount of emission depending on the electric field strength on the surface of the cathode 81 is controlled by the voltage of the electrode 83 connected to the bias supply source 91. The voltage on the electrode 83 is negative with respect to the voltage on the cathode 81. The cathode 81 is heated by the current source 93.
A magnetic condenser lens 95 near the cathode 81 is used to collimate the electron beam. Upper deflector 97
Are used for alignment (position alignment), stigmation (astigmatism), and blanking. The optical system is further provided with a beam limiting aperture 99 consisting of several holes. The beam 100 is deflected by a pair of electrostatic deflectors 101 and 103 arranged in front of the objective lens 104, and oscillates around a point above the objective lens 104. The objective lens 104 is the lower lens pole piece 10.
6, the intermediate electrode 107, and the upper lens pole piece 105. In the high voltage operating mode, the probe is focused using only the upper lens pole piece 105 and the lower lens pole piece 106 of the objective lens 104. After all beam 1
00 is scanned over the substrate 57 with focus at a far distance. Therefore, an image of the beam generation source which is refocused by the almost parallel objective lens 104 and is magnified 1 × times, that is, is formed, and illuminates the substrate 57.

In the high voltage secondary electron imaging mode, the objective lens 104 extracts secondary electrons. XY stage 2
4, the substrate 57, and the lower lens pole piece 106 are floated to a negative potential of several hundred volts by the power supply 111. As a result, the secondary electrons are accelerated to this energy state and pass through the deflectors 112 and 113. Intermediate electrode 107
Are positively biased with respect to the XY stage 24 by a power supply 115. The intermediate electrode 107 is formed on the substrate 57.
Used to immediately accelerate the electrons away from the substrate and to efficiently collect the secondary electrons generated from the defective regions of the substrate. The combination of the XY stage 24 and the intermediate electrode 107 substantially eliminates the unevenness of the time for the secondary electrons to reach the secondary electron detector 117. The secondary electrons pass back through the lens 104 and back again, so that the returned secondary electrons are deflectors 113, 112 functioning as Wien filters.
Is deflected toward the secondary electron detector 117. Here, the return beam is the anode 11 of the secondary electron detector 117.
It is re-accelerated to a high energy state by the power supply 119 connected to the No. 8 power source, and the secondary electrons are made to collide with the secondary electron detector 117 which is a Schottky barrier solid state detector at an energy level sufficient for amplification. The anode 118 of the detector diode (secondary electron detector) 117 is reverse biased by the power supply 121. The amplified signal from the detector diode 117 is sent to a preamplifier 122 from which it is sent via a high voltage isolated fiber optic link 126 which is a secondary electronic component of the signal 33 of FIG. Sent to the circuit. This signal is a portion of the signal 33 in FIG. 1 which depends on secondary electrons.

In order to be able to inspect partially transparent substrates,
A transmission electron detector 129 is provided below the XY stage 24. Since the transmitted electrons pass through the substrate 57 with high energy, it is not necessary to re-accelerate the transmitted electrons. Upper electrode element 12
3, the transparent electron lens composed of the central electrode element 124 and the lower electrode element 127 spreads the transmitted electron beam to a diameter suitable for detection by the transmitted electron detector 129 which is a Schottky barrier solid state detector. The upper electrode element 123 is XY
The central electrode element 1 is held at the same potential as the stage 24.
24 is maintained at 0 to -3 KV by the power supply 114. The signal from transmission electron detector 129 is amplified by amplifier 133 and transmitted by fiber optic link 135 which is the transmission electron component of signal 33 of FIG.

The optical system is also designed to collect backscattered electrons that leave the substrate surface at about the same energy level as the primary electrons. The backscattered electron detector 160 has a detector 117 located beside the beam axis.
Is a Schottky barrier diode detector similar to. By slightly deflecting the electrostatic and magnetic settings of both deflectors 112, 113, which function as Wien filter deflectors, the beam is deflected to the left in the figure and is incident on the backscattered electron detector 160, which is a solid state detector. . The backscatter signal is amplified by the preamplifier 162 and
It is sent to the processor 48 (see FIG. 1).

To image with a low voltage beam in the range of 500-1,500 eV, the elements of the objective system must be biased very differently and two new elements must be used. The primary beam electrons are the power source 111 and the substrate 5
7, lower lens pole piece 106, intermediate electrode 107 about -19
It is decelerated in the objective lens by floating to Kv. With this technique, the electron beam is decelerated only near the end of the path, preventing aberrations that degrade the image and interaction effects in the electron beam generator when the entire path of the beam is operated at low beam energy. can do. With such a configuration, an excellent focusing effect can be obtained in the deceleration region between the upper lens pole piece 105 and the lower lens pole piece 106. A snorkel lens 125 having only one working pole piece is further provided under the substrate to form a magnetic focusing region near the substrate. The return bundle of this lens passes through the lower lens pole piece 106 to the outer shell of the snorkel lens. Since the magnetic field is strong in the vicinity of the substrate, in addition to the focusing effect, it helps in extracting low-energy secondary electrons from the deep element pattern, and the secondary electrons are re-accelerated to move within the aperture of the objective lens 104. Secondary electrons become parallel to each other as they rise.

In the low voltage imaging mode, the secondary electrons leaving the substrate at about 5 eV are accelerated to about 19 KeV in the objective lens. To minimize charging, it is desirable for the secondary electrons to pass through a short region near the field-free substrate. In the low voltage mode, if the objective lens is not at the voltage level of the intermediate electrode 107, an accelerating region is formed near the surface of the substrate 57 due to the magnetic field leaking from the objective lens. In low voltage mode,
The intermediate electrode 107 is negatively biased with respect to the lower lens pole piece 106, and the power supply 115 forms a magnetic field-free region that can be adjusted for optimum low-voltage imaging. After reacceleration, the secondary electrons pass through Wien filter deflectors 112 and 113, where they are deflected to the left and incident on the backscattered electron detector 160 used for backscatter imaging in high voltage mode. . The signal thus detected is amplified by the preamplifier 162 and becomes the most important image signal for low voltage inspection of the wafer. Detectors 117 and 12
Neither 9 is used for wafer inspection in this mode.

FIG. 8 is a schematic view of various electron beam paths inside the electron beam generator 20 and directly below the device board 57. The electrons are emitted radially from the thermal field emission cathode 81 and appear to originate from a very small bright spot source. The combined action of the accelerating field and the magnetic field of the condenser lens collimates the beam into a parallel beam. Electrons emitted at unusable angles are blocked by the anode aperture 87 of the electron gun, and only electrons emitted at unusable angles enter the beam limiting aperture 99 as a beam. By stigmatization and alignment by the upper deflector 97 in FIG. 7, the beam finally becomes round in cross-section and passes through the center of the objective lens consisting of the elements 105, 106, 107 in FIG. The magnetic condenser lens 95 of FIG. 7 is mechanically positioned so that its center is coincident with the axis defined by the thermal field emission cathode 81 and the beam limiting aperture 99. Due to the deflection, electrons are emitted from the objective lens 104 following the path shown in the drawing, and are scanned and converged to collide with the substrate at one point.

The diameter of the beam 100 to be scanned and the current value are
It depends on several factors. The beam flow is determined by the angular radiation from the radiation source (1.0 Ma / steradian) and the aperture angle defined by the beam limiting aperture 99. The beam diameter is determined by the aberrations of both lenses designed for high excitation (field width / focal length) to minimize spherical and chromatic aberrations. The size of the beam projected on the substrate 57 is determined almost half by the interaction effects of the beam (statistical blurring due to repulsion between the individual electrons that make up the beam), so that such high intensity beam systems are Then the beam interaction is important. By making the beam path as short as 40 cm and using relatively large half-angle lenses for the electron source and the substrate 57, respectively, to prevent electron crossover between the electron source and the substrate 57. , The effect of beam interaction can be minimized. A given beam spot is obtained by choosing an aperture diameter that maximizes the beam flow while balancing the aforementioned effects. The beam spot size can also be changed by changing the lens strength to expand or contract the beam from the beam source, but in such systems, adjustment is first made using the aperture. .

In high voltage mode, both deflectors 112 and 113 of FIG. 7 function as Wien filter deflectors.
Deflects the secondary electron beam 167 at about 100 eV with little effect on the high energy scanning beam 100. The Wien filter deflector consists of an electrostatic eight-pole deflector 112 and a four-pole magnetic deflector 113 which are arranged so that the electric and magnetic fields intersect (at right angles to each other). Return 2
The secondary electron is laterally deflected by both fields. However, since the beam 100 of primary scanning electrons is moving in opposite directions, even if the Wien filter deflector deflects the secondary electron beam 167 to a wide angle by properly selecting the strengths of both fields, The primary scanning beam 100 should not be affected in any way. The so-called "Vienna filter" is effectively used for coaxial extraction. The anode 118 of the secondary electron detector 117 is the secondary electron detector 1 where the secondary electron beam 167 is a solid state detector during reacceleration.
It is shaped so as to be collected and converged by 17 collectors.

FIG. 8 also shows detection paths of transmitted electrons and backscattered electrons. In order to detect the backscattered electrons in the high voltage operation and the secondary electrons in the low voltage operation, both Wien filter deflectors 112 and 113 perform different operations,
This allows the backscattered electrons and secondary electrons to travel backscatter detector 16 through paths shown to rise up the system.
Try to reach 0. When a partially transparent substrate is imaged, some electrons pass through the substrate 57 without losing any energy. Such transmitted electrons pass through the upper electrode element 123 and the central electrode element 124 of FIG.
Although entering the transmitted electron detector 129, both electrode elements function as lenses to spread the passing electrons 108, so that the transmitted electrons spread and then enter the transmitted electron detector 129. If you want to obtain the transmission signal in high voltage mode,
Lens 125 allows transmitted electrons to pass through the holes of snorkel lens 125 without substantially affecting the lens field required for low voltage secondary imaging.

In the low voltage mode operation in which the substrate 57, the lower lens pole piece 106 and the intermediate electrode 107 are floated to a high voltage, the beam paths are similar but the operation of the objective lens is quite different. A snorkel lens 125 generates a magnetic field that penetrates through the substrate 57 and reaches inside the lower lens pole piece 106. When the electrons are decelerated by the magnetic field around the upper lens pole piece 105, the intermediate electrode 107, and the lower lens pole piece 106, the effective refractive index becomes relatively short because the refractive index becomes large. This type of deceleration immersion lens has a remarkable effect of reducing aberrations. The immersion lens, unlike the cathode lens used in the conventional emission microscope technology, negatively biases its electrode 107 to form a short “electroless region” near the substrate 57. Even if the substrate is biased, the low-energy secondary electrons return to the substrate due to the action of the intermediate electrode 107, so that the undesired charging effect is neutralized.

Secondary electrons distant from the “non-electric field area” are re-accelerated in the area between the intermediate electrode 107 and the upper lens pole piece 105. Secondary electrons emitted from electron gun are 20 KeV
From the upper lens pole piece 105 with an energy that is approximately equal to the primary beam energy of ## EQU1 ## minus the energy impinging on the substrate. In the objective lens area, the subsequent path is similar to the primary beam, but the secondary electrons are widely dispersed and emitted, and the angle is very wide. This secondary electron beam is directed to the low voltage secondary electron detector 160. The secondary electrons go to the low voltage secondary electron detector 160. This detector can share the same as the backscattered electron detector 160. Since the energy of the returned secondary electron beam 104 is comparable to that of the primary beam energy, it has to be deflected by the very powerful Wien filter deflectors 112 and 113, at which time the path of the primary beam 100 is largely affected. You can choose not to.

Since the low voltage mode is often used to inspect partially insulated substrates, techniques to minimize charging are important. When the number of secondary electrons (low energy secondary electrons and backscattered electrons) and the number of primary beam electrons incident on the substrate are not equal, the insulating region is charged.
Whatever surface is to be imaged, the charge balance changes depending on the microstructure and material. The amount of energy emitted by the scattering of secondary electrons varies depending on the energy of the incident beam, but is higher than the range of 200 to 1,500 eV for many materials and smaller than this range for other materials. If the energy emission due to the scattering of secondary electrons is larger than the range of 200 to 1,500 eV, the surface is positively charged.

The secondary electrons leave the surface of the substrate 57 in the energy range of 0 to approximately 20 eV, but the most likely energy value is 2.5 eV. If the electric field near the surface of the substrate 57 can be controlled by, for example, the potential of the intermediate electrode 107, secondary electrons can be emitted from the substrate or returned to the substrate depending on the applied electric field and the energy with which the secondary electrons leave the substrate. can do. For example, when the suppression potential wall of about 10 eV is formed, only a small amount of secondary electrons emitted from one point of the substrate 57 cross the potential wall and enter the detector.

When the amount of emitted secondary electrons or back scattered electrons away from the magnetic field is larger than the amount of electrons of the primary beam, the substrate 57 is positively charged, and the suppression potential wall formed by the intermediate electrode 107 The size increases. Therefore, most of the secondary electrons with low energy stay. The surface potential remains positive until balanced. When the amount of secondary electrons or backscattered electrons emitted is smaller than the amount of primary electrons, the surface is negatively charged, and the suppression potential wall formed by the intermediate electrode 107 becomes low. Therefore, a large amount of low energy secondary electrons are emitted. The surface potential remains positive until balanced. Under such circumstances, the surface potential becomes stable after a certain period of time. In order to prevent a large potential difference from occurring in the region of the substrate, the intermediate electrode 107 is appropriately adjusted so that the equilibrium state (the state where the intensity of the primary electron beam is equal to the intensity of the secondary electron) is averaged. It is important to get it.

The equilibrium voltage changes in other regions of the substrate because differences in the microstructure and materials affect the energy emission of secondary electrons. However, in equilibrium, the amount of secondary electrons generated is the same in all regions. That is, no contrast occurs in the board image taken in the equilibrium state. To avoid this problem, the pixel-by-pixel dose is kept low and, if necessary, rescanned using the "multi-frame scan" technique described above to obtain good image statistics.

By controlling the time between scan trajectories, it is possible to neutralize the substrate by the electrons generated in adjacent regions between scans. The key elements of this technique are the intermediate electrode 107 that controls the electric field and the overlapping frame scan trajectory. In order to increase the reliability of the electron gun even when the life of the cathode is short, the electron gun has a hexagonal rotary turret 1 which is floated to a high voltage as shown in FIG.
The structure has six cathode control electrode assemblies provided on 37. Each assembly is rotated and moved over and fixed above the anode aperture 87 to make electrical contact with the appropriate power supplies 91 and 93 of FIG.

In an electrostatic deflection system consisting of electrostatic deflectors 101, 103 placed in front of the lens in FIG. 7, the electric field generated by the high speed sawtooth wave deflection voltage must be kept highly uniform. There is. The structure is a monolithic ceramic-metal monolithic construction that forms 20 deflection plates by etching. In order to match the coordinate system of the XY stage 24 with the coordinate system of the substrate 57 for scanning, four driving devices are required for each of the two stages.

From the standpoint of ease of operation, the system introduced an automatic adjustment mechanism and an automatic adjustment procedure at the start. The lens, deflection / stigmation element and total high voltage source are all under the control of a data acquisition control system interfaced to the electron beam generator control computer 42 shown in FIG. . The program for adjusting the deflection ratio and the electrostatic plate voltage, which have special values required for individual operations, is the electron beam generator control computer 42.
Built in. The control and adjustment of the electron gun is based on the rated values modified by an adjustment routine that uses analog-digital feedback to set the radiation dose, aperture pass-through and power supply.

The beam alignment is performed by using a known method of removing the deflection when the transmission current of the lens changes. This scan used specific test samples imaged by the bi-axial frame scan function, and also added the image analysis capabilities needed for alignment and inspection. There is an autofocus mechanism to compensate for changes in the height of the substrate, and aberration correction called stigmation is performed before inspection. These adjustment procedures are based on the analysis of image contrast and harmonic content by the image acquisition pre-processor 48 and associated electronics.

When the optical system is operating at the rated condition in the high voltage mode, the present invention provides a beam energy of 20 KeV and a beam spot size of 300 nA and 0.05 μm.
To 1,000 μA and changes to 0.2 μm. The scan rate is 5 microseconds using a 512 pixel scan field imaged at 100 megapixels / second. The diode current amplification factor of the secondary electron detector 117 is
It is about 1000 times at 5 KeV to 5000 times at 20 KeV. For samples with edge contrast greater than about 14% at 100 megapixels / second using a 0.05 micron spot, the entire system can be operated beyond this range of operating conditions. The acquisition electronics allows integration of multiple scan lines so that low contrast or high resolution images can be recorded with low bandwidth.

In the low voltage mode, the upper lens pole piece 105
The beam energy up to is 20 KeV and the beam energy at the substrate is 800 eV. The relationship between the beam intensity and the spot size is 0.05 μm and 15 at 25 nA.
It is 0.1 μm at 0 nA. The scan period and field size are the same as in high voltage mode. Backscattered electron detector 160
The amplification factor is 5000 times. For samples with edge contrast greater than about 20% at 100 megapixels / second using a 0.05 micron spot, the system can be operated beyond this range of operation.

Defect Processor The defect processor 56 compares the image data obtained from the die 68 with the image data obtained from the die 70 in the case of die-to-die comparison inspection, and in the case of die-to-database comparison inspection. To compare the image data obtained from die 64 with the data obtained from database adapter 54. The routine and basic structure of the defective processor 56 is almost the same as the routine and basic structure of the defective processor disclosed in US Pat. No. 4,644,172. U.S. Pat. No. 4,644,172 was issued February 17, 1987, and was handed over to the applicant of the present application. The inventor is Sandland and others, and the title of the invention is "automatic wafer inspection. Electronic control of the system ".
Whereas the US patent uses three parameters to determine defects, the present invention uses four parameters.

For both die-to-die inspection or die-to-database inspection, whether the data is obtained from memory block 52 (depending on how alignment correction is performed), or After registration, it is obtained from the registration computer 21. The data format is 6 bits per pixel for each detector type. The defect processor 56 determines the following four parameters for each pixel of each detector to which both data are input.

A. I is the grayscale value of the pixel b. G is the magnitude of the gradient of the grayscale pixel c. P is the phase of the gradient of the grayscale value, or the angle of orientation d. C is the local Slope Contour Curvature The grayscale value is simply the value in memory block 52 for a particular pixel. The magnitude of the tilt and the direction of the tilt are obtained as follows. First, of the Sobel operator
Compute the X and Y components.

[0077]

[Equation 1] Therefore, the size of the tilt is

[Equation 2] And its direction angle P is P = arctan (Sy / Sx).

The curvature is defined as follows.

[0079]

(Equation 3) Here, the coefficient aij is a set of parameters selected depending on the wave situation, and Rij is defined as follows.

[0080]

[Equation 4] Where Iij is the grayscale value of the pixel in the row at column i and column j of the image, and aij and bkl
Is an empirically obtained parameter.

Typical values in the preferred embodiment are as follows.

[0082]

(Equation 5) For each image, I, G, and
Find the values of P and C. These parameters for pixel A of die 68 are compared with the parameters of the corresponding pixel B 1 of die 70 and with the parameters of the eight pixels adjacent pixel B 1. If the corresponding parameter of pixel A differs from the corresponding parameter by more than a predetermined tolerance, a flag indicating a defect in both dies will be displayed in pixel B.
Attached to.

Similarly, the parameters of each pixel of die 70 are compared with the parameters of the corresponding adjacent pixels of die 68, and pixels that differ by more than a predetermined tolerance are flagged as defective.

This algorithm is described in US Pat.
It can be implemented with the pipeline logic disclosed in 644,172. Matrix operation is an application-oriented integrated circuit ASIC (Appl) connected to a pipeline calculation system capable of calculating defect data at a speed of 100 megapixels / second.
ication Specific Integrated Circuit (ASIC)).

Deflection Controller Deflection controller 50 positions electron beam 100 at equidistant grid points within each scan area 60 of die 68 in the die-to-die comparison mode. The outputs of the detectors 129, 160, 117 thus obtained are the same detectors 129, 160, 11 at corresponding positions of the die 70.
7 output is compared. Similarly, in the die-to-database comparison mode, the simulated image obtained from the database adapter 54 is compared with the output of the secondary electron detector 117 obtained from the die. The deflection controller 50 controls the positions of the XY stage 24 and the electron beam 100 to position the electron beam, as described below with reference to FIG.

When scanning the first die in the scan area, the output of the alignment computer 21 is set to zero. This is because no misalignment occurs during the scanning of the first scan area of the first die. Therefore, during scanning of the first scan area of the first die, deflection controller 50 receives commands only from electron beam generator control computer 42. The deflection controller 50 includes a command from the electron beam generator control computer 42 and the X-axis and Y-axis interferometers 28.
XY stage 2 based on the position data obtained from
4 calculates the desired movement amount, and sends a signal corresponding to this movement amount to the stage servo 26 to move the XY stage 24.

The deflection controller 50 similarly calculates a desired deflection amount of the beam 100 and sends the deflection amount data to the analog deflection circuit 30. When the XY stage 24 moves, its position is constantly monitored by both X-axis and Y-axis interferometers. If a discrepancy with the desired XY stage position is found, an error signal is generated based on this discrepancy. This error signal is fed back to the stage servo 26 by the deflection controller 50. Since inertial force acts on the XY stage 24, if an error frequently occurs, the position of the XY stage cannot be corrected by the error signal. The error that frequently occurs in both the X and Y axes is the electron beam 10
Corrected by a zero deflection. In this case, the deflection controller 50 calculates the deflection amount of the electron beam 100 and sends a signal corresponding to the deflection amount to the analog deflection circuit 30 in digital form.

As the beam 100 scans the die 68, grayscale values are stored in the memory block 52.
As the electron beam 100 begins to scan the die 70, the grayscale value of the die 70 is immediately transferred to the memory block 52.
And sent to the defect processor 56 and the alignment computer 21. The alignment computer 21 compares data from the die 68 and the die 70 for alignment (alignment). If the positions are not aligned, a position alignment correction signal is generated and sent to the deflection controller 50. This alignment signal is beam 10
It is used for fine adjustment to position 0 at an accurate position on the substrate 57.

In the die-to-database compare mode, the deflection controller 50 works much like in the die-to-die compare mode, but instead of the input image obtained from the first die in the scan area, The difference is that the output of the adapter 54 is used.

The deflection controller 50 is still in this mode.
Parameters relating to the moving amount, speed, direction of the XY stage 24 and deflection of the electron beam are calculated and defined.

Registration Computer The registration computer receives both digital images in the form of gray scale values and determines the misalignment of the registration between the images as a slight misalignment of pixels. A preferred embodiment of the calculation for registration is disclosed in US Pat. No. 4,805,123. This patent was issued in February 1989 and is entitled "Automatic Photomask and Reticl Inspector and Method with Improved Defect Detection and Alignment Subsystem."
e Inspection Method and Apparatus Including Improv
ed Defect Detector and Sub-System).

In the preferred embodiment, the alignment correction signal 51 is calculated continuously over the entire examination area. The position alignment correction signal calculated in this way is used by the position alignment computer for moving or moving the image from the memory block 52 and for interpolation (movement of sub-pixels). Alternatively, assuming that the positional misalignment does not suddenly occur during scanning, a small number of specific feature points on the substrate 57 are selected and the positional misalignment is calculated only for the selected feature points. good. In this case, force computer, Inc.
You can use a single board computer such as the model CPU30ZBE. The position alignment correction signal can also be used to reverse offset subsequent data collection positions to reduce position misalignment, and can also be used to determine misalignment between images sent from memory block 52 to defective processor 56. .

Analog Deflection The analog deflection circuit 30 produces an analog gradient function for the electrostatic deflectors 101 and 103 of FIG. 7 which are composed of 20 pole plates. The operation of the analog deflection circuit 30 is shown in FIG.
2 is shown. The digital signal from the deflection controller 50 is converted to an analog voltage by the gradient DAC 230 and then guided to the gradient generator 232. The size of the gradient is
Controlled by DAC 236. The sample and hold circuit 238 is used to define the start of the slope and the sample and hold circuit 240 is used to define the end of the slope.
A high voltage, low noise driver amplifies the waveform to produce a gradient with a dynamic range of ± 180 V, which is applied to electrostatic deflectors 101, 103.

Memory Block The memory block 52 consists of three identical modules, each module corresponding to one of the secondary electron detector 117, the transmitted electron detector 129 and the backscattered electron detector 160. There is.

As conceptually shown in FIG. 13, each module of the memory block 52 consists of two First In-First Out memories. The first first-in first-out memory stores the grayscale values of the entire scan area obtained from the die 68 by each detector, and the second first-in first-out memory is short, corresponding to only a few scans of the die 70. Store the grayscale value obtained by the detector. The outputs from both first-in first-out memories are the defect processor 56 and the alignment computer 21.
Sent to Each first-in first-out memory operates at a speed of 100 Mhz and stores the grayscale value of each pixel with an accuracy of 8 bits per detector.

The memory receives at the input register 302 the 8 bytes sent in parallel from the image acquisition pre-processor 48 for each detector. The input register 302 is
It behaves like a register, shifting 8 bytes to the right and then receiving another 8 bytes until the eight sections of input register 302 are full. When eight sections of input register 302 are full, 64 bytes are clocked into memory 303. DRAM 303 can be used for the memory block, usually 1
Cover with 28 MB.

[0097] pre-processor 48 for pre-processor image capturing for image capture (pre-processor), the 8-bit value at a speed of 100Mhz analog signals from the detectors 117,160,129 The output signal is digitally converted and reformatted for storage in memory block 52. The pre-processor 48 consists of three identical modules, one of which is shown in FIG. Each module receives the output from the corresponding detector, digitizes the received output into 8 bits, (AD
It is sent to the converter 9) and the multi-scan integrator 11. The purpose of the multi-scan integrator 11 is to average grayscale values from the same pixel to reduce noise. In some cases, the result obtained by scanning the same pixel several times, that is, the result obtained by sampling is the average value of the pixel. This value is sent to the shift register 13. The shift register 13 receives 8 bytes serially and sends the received 8 bytes to the memory block 52 in parallel.

The interferometer XY stage 24 has a teletrack position on the X and Y axes.
XY interferometer 28 such as TIPS V (Teletrac TIPS V)
To monitor. The position of the XY stage 24 is defined with a precision of 28 bits with the least significant bit corresponding to about 2.5 nanometers.

The overall control of the system computer inspection system 10 is implemented by the system computer 36. System computer 36
Performs a wide variety of steps in sequence, including other setup tasks. Each of the steps connected to each other is accomplished at a predetermined time according to the program.
When several kinds of series of processes do not contradict each other, the processing capacity of the system computer 36 is maximized so that several kinds of series of processes which do not contradict each other can be simultaneously executed.

The work performed by the system computer 36 is performed via the user keyboard 40 with a mouse or trackball pointer, or otherwise by data communication with a remote computer. In the case of local interaction, the computer display 38 displays graphics and text from the system computer 36.

The routines of system computer 36 are organized into four communication tasks:

1) Control computer 4 for electron beam generator
2. A master task that communicates with the post processor 58 and the substrate handler 34.

This task stores in the system computer a file recording lens operating settings, device operating parameters such as vacuum pressure and beam flow.

2) A user interface task that manages the display on the computer display 38 and handles input from the user keyboard 40 and mouse.

This task modifies the data file or transmits a message to other parts of the system to initiate processing in response to input from the user keyboard 40 or mouse.

3) The characteristics of the inspection area for image acquisition are controlled by the electron beam generator control computer 4 via the master task.
Inspection task to be transmitted to 2.

4) A command language interpreting task that enables command input from the user keyboard 40.

This task also manages a timer that allows automatic scheduling of repetitive actions. In addition, this task handles the creation and updating of a text log file that describes all of the device operations and the time that they occur. This task is typically used only by service engineers to control the equipment.

As an example of a system computer, a UNIX operating system (UNIX operating system)
system) Sun Microsystems Spark Processor (Sun Microsystems SPARC pro
cessor). UNIX is a registered trademark of AT & T.

Electron Beam Generator Control Computer The electron beam generator control computer, or column computer 42, comprises an autofocus computer, a vacuum control computer, and a deflection command computer. Functions and specific examples of the autofocus computer will be described in the section "Autofocus System", and functions and specific examples of the vacuum control computer will be described in the section "Vacuum system".

The column computer 42 is a system computer.
A command is received from the computer 36. The column computer 42 has a force computer (Force Comp
68030, such as the CPU 30ZBE manufactured by uter, Inc.)
A base single board computer can be used.

Post Processor The post processor 58, from the defective processor 56,
A map is received for each detector showing all defective pixels. The post-processor 58 combines these maps to determine the size and position of each defect and classify it according to the defect type. In this way, the data available to the system computer 36 is obtained. Post processor 58 may be a 68030-based single board computer, such as the CPU30ZBE manufactured by Force Computer.

Video Frame Buffer The video frame buffer 44 is a commercially available video buffer that has a storage capacity of 12 bits per pixel and can store 480 x 512 pixels.
It is a frame memory. An appropriate frame buffer is Image Technology I
nc. ) Model FG100V can be mentioned. video·
The frame buffer has 30 image displays per second
Refresh times.

Image Display Image display 46 is a Sony model PVM 1342Q.
Commercially available color monitors, such as
Pseudo-color technology is used to allow the operator to easily evaluate the image. The pseudo-color technique assigns different colors to gray shades of a black and white image.

Database Adapter The database adapter 54 is an image simulator that generates a gray scale corresponding to each pixel based on the computer-aided design data used for designing the pattern to be formed on the die. A typical input device for a database adapter is a digital magnetic tape in the format used to pattern integrated circuits. Digital data is
It is converted to a series of pixel data representing the scan area in the same format as the output of the image acquisition preprocessor 48. Such a database adapter is described in US Pat.
It has already been disclosed in 926,489. This patent is 199
Issued May 0, its name is "Reticle Inspection System",
(Reticle Inspection System).

Substrate Handler The substrate handler 34 has a function of automatically taking out the substrate 57 from the cassette and placing the taken-out substrate in the substrate holder in an appropriate direction. It is a robotic device similar to a wafer handler normally used for handling.
The substrate handler 34 first detects the flat notch 59 shown in FIGS. Substrate handler 34
Is a linear extension from the center of rotation of the substrate 57 in the radial direction.
The CDD sensor optically detects the flat 59 and its direction. When the board rotates, the output of the linear CDD sensor is converted into a digital value, and the force computer CP
Stored on a single board computer like the U 30ZBE. This computer determines the flat 59 and its direction and position. The substrate 57 rotates until it is in the proper direction,
After that, it is automatically placed on the substrate holder. Board 57
The board holder that holds the load elevator 2 of FIG.
Take 10 The operation of the substrate handler is entirely controlled by the system computer 36.

[0117] The function of the XY stage XY stage 24 is moving the substrate 57 under the electron beam 100 and the alignment optical system 22. In order to minimize the system complexity, the XY stage 24 has two degrees of freedom in the X-axis direction and the Y-axis direction. That is, the XY stage 24 cannot rotate or move in the direction perpendicular to the XY plane of the substrate 57. XY stage is in X-axis direction, Y
It can only move axially and diagonally. The rotation of the electron beam raster is achieved electronically by splitting the scan into two types of electrostatic deflection components of the beam and moving the XY stage in the X axis direction, Y axis direction, and diagonal direction by mechanical servo. To be done. Since the objective lens has a variable focus in a range sufficient to compensate for the height change of the substrate, movement in the Z-axis direction is unnecessary.

The XY stage 24 is a device that can control linear movement, right-angle movement, and repetition with extremely high precision. Use crossed roller bearings. Of course
The XY stage can be used even in a vacuum, and the electron beam 10
It is made of a non-magnetic material so as not to interfere with 0. The transmitted electron beam 108 has a transmitted electron detector 1 under the XY stage 24.
To reach 29, the XY stage has an open frame. The open frame is also effective for mounting the substrate 57 on the open frame from below in the mounting process.

By using two three-phase brushless linear motors (not shown) per axis to drive the XY stage 24, the best system function is achieved. A suitable linear motor is the Anoline Model L1 manufactured by Anorad Inc.
And L2 (Anoline mod el L1 nd L2).

[0120] the entire vacuum system vacuum system is under the control of the electron beam generator control computer 42. Although not shown, ordinary pressure sensors are arranged at various places in the system to measure the pressure and notify the electron beam generator control computer 42 of the measurement result. The electron beam generator control computer 42 sequentially controls various valves as needed at the start of inspection or during placement or removal of a substrate. The sequential valve control routine will be described in detail in the section "Placement Operation". The high voltage is automatically shut off to prevent damage to the thermal field emission cathode 81 when the vacuum is insufficient and the electron beam is unsuitable for operation. This operation is executed by a combination of the electron beam generator control computer 42, the system computer 36, and the pressure sensor. At the same time, the air sluice valve 145 (FIGS. 9 and 11) is actuated, and the ultra high vacuum region 140 of the electron beam generator 20 is activated.
Prevent pollution of. The operation of the vacuum system will be described below.

The vacuum containing the electron gun is achieved using a two-stage differential pump. If it is baked enough, it can be used for a long time without any maintenance. The ultrahigh vacuum region 140 of about 10 to the 9th power (10 **-9) torr is partitioned by the anode aperture 87 of the electron gun,
Exhausted by the pump 149. The intermediate vacuum region is about 10 −8 torr, and the main vacuum region 143 and the aperture assembly 9 are evacuated by the ion pump 149.
It is isolated from 9 by a pneumatic valve (to insulate the electron gun). The above vacuum-related elements provide an optimum environment for the thermal field emission of an electron beam.

The main vacuum region 143 is maintained in a vacuum state by the turbo pump 204, and the inspection chamber 206 is maintained in a vacuum state by the turbo pump 208. The inspection chamber 206 is separated from the main vacuum region 143 by a plate. This plate has a small aperture through which the electron beam passes. Since the inspection chamber 206 and the main vacuum region 143 are partitioned in this way, a high vacuum state can be maintained even if the substrate to be inspected is covered with a photoresist having a considerable vapor pressure. .

The vacuum system has two airlocks 224 and 226. One is used to place the substrate 57 in the inspection chamber 206, and the other is used to take out the substrate 57 after the inspection is completed. Both airlocks communicate with the vacuum pump 220 via valves 212 and 214 that are arranged in parallel. The valve 212 is for exhausting the air lock 224 at a low speed, and the valve 214
Has a large opening and can exhaust a large volume. A similar mechanism is provided on the airlock 226. Since this exhaust mechanism has the same structure, it is shown by the same reference numeral. The purpose of dually providing the exhaust mechanism of the same structure is to prevent charged particles from being disturbed by the exhaust process,
Moreover, this is to shorten the time required for exhausting and pressurizing the chamber.

As described in detail below, when the substrate 57 is placed on the airlock 224, first, the low speed exhaust valve 2 is placed.
Only 12 open. This keeps the flow velocity in the chamber low enough so as not to disturb the charged particles in the region of the airlock 224. When the pressure in the chamber is reduced and the air flow reaches the level of the free molecular flow region, that is, the region where charged particles are no longer disturbed, the large volume exhaust valve 214 is opened and remains in the airlock. Exhaust air rapidly. A similar two-step operation is used for the pressure treatment.
However, in the pressurization process, a further set of valves 228 and 230 were provided for both high speed and low speed ventilation for both airlocks 224, 226, respectively.

Mounting Operation As described before, the substrate 57 is mounted on the substrate handler 34.
Mounted on the substrate holder of the mounting elevator 210
To be installed on. At this time, the airlock 224 is in the atmospheric pressure state. Valve 212, which evacuates airlock 224 at low speed, opens. When the pressure in the airlock 224 reaches the pressure of the molecular flow, the large volume exhaust valve 214 opens and the remaining air is exhausted. Here, the gate valve 216 opens and the loading elevator 2
10 pushes the substrate 57 and substrate holder up to the inspection chamber 206 through the gate valve 216, and the stage 24
Place on. When the inspection of the board 57 is completed, the board 57 is stored in the board storing cassette in the reverse order.

Alternatively, the substrate cassette can be placed in the chamber in a similar manner. When the inspection of all the substrates stored in the cassette placed in the chamber is completed, the cassette is taken out of the chamber and replaced with another cassette.

Further, since the present invention has the double airlock structure, while the substrate in one chamber is inspected, the other chamber is simultaneously used to mount and pressurize another substrate, and to depressurize and pressurize. You can take out the cassette.

Autofocus System The electron beam 100 is focused by changing the current in the objective lens 104 of the system shown in FIG. Since the substrate is not necessarily flat and the surface of the XY stage 24 may not be perfectly perpendicular to the axis of the electron beam generator 20, the optimum focus current varies over the inspection area. However, this change is slow as a function of distance in both the X and Y axis directions, so that the optimum focus current can be determined at several specified points on the substrate 57.

As one step of the preparation and starting procedure of the inspection process, the optimum focus current is measured at the designated point. This focusing process consists of positioning the beam at a specified point and measuring the grayscale value along a straight line perpendicular to the edge of the feature on the substrate 57. For example, for 10 different values of focus current, the digitized gray scale values are convolved with a high pass filter, not shown. The best focus current is
It is the current corresponding to the maximum value in the output of the high pass filter. In the preferred embodiment, 2 with the following convolution factor:
It uses a second derivative filter.

-4 0 0 0 8 0 0 0 0 -4 The output of the high pass filter must be smoothed for best effect. The focus computer is part of the electron beam generator control computer 42. The focus calculation is performed by special purpose hardware consisting of a convolutional integrated circuit and a few DSP elements.

Alignment Optics Alignment optics 22 are used by the operator to visually perform a coarse die alignment after the die has entered the inspection chamber. The subsystem consists of a window facing the vacuum chamber and a lens that projects the image pattern onto a CCD camera for display on display 46. The operator can select one of the two lenses. According to the present invention, experience has shown that the magnification of one lens is 0.46 and the magnification of the other lens is 5.8.
Is set to. The lens is placed outside the vacuum region to prevent dirt from the substrate from adhering to the optical surface.

SEM Plasma Cleaner When the electron beam apparatus of the present invention operates, the target substance evaporates and is attracted to the high-pressure region by the proximity interaction (charging of particles near the surface), so that electron beam formation and deflection can be prevented. Organic materials are deposited on the various electrodes used. Since the insulator gradually deposited due to the charging of the surface adversely affects the electron beam formation and the deflection mechanism, the deposited insulator must be periodically removed. Periodic removal of the insulator is accomplished by forming an oxidizing plasma near the area where it is deposited. For the formation of oxidative plasma,
Oxygen is used as the main gas for the formation of the cleaning plasma.

Since plasma is required to perform the above cleaning function, a plasma source mainly containing oxygen will be used below. According to FIG. 11, the oxygen supplier 199 is connected to the upper part or the lower part of the chamber through the valve 193 and the mass flow controller 195. At this time, the electrostatic capacitance type pressure gauge 197 acts as a pressure sensor to adjust the pressure.
Oxygen regulates the pressure so that it is easy to combine with RF energy and confines the plasma. The corresponding excitation electrode is selected depending on the confined place, and the pressure is adjusted in consideration of the difference in mean free path of oxygen gas. By strictly controlling the plasma space density during discharge and maintaining the density at a level just below the sputtering potential of the electrode surface, only the organic substance can be selectively oxidized. This is achieved by suppressing self-biasing of the electrodes at high frequencies and by precisely controlling the RF power level and applied voltage.

Now, turning to FIG. 15, electrodes and the like that need to remove deposits are connected through relay 191 to multiplex relay 179 having RF characteristics. The high frequency power 173 is oscillated and passed through the power detector 175 and the voltage detector 178 to be smoothed. 177 is an automatic matching network (trademark example is auto match), and the smoothing output is the voltage,
A sufficient avalanche voltage is created by appropriately converting the current and phase values. Accordingly, the plasma discharge is started to match the impedance and maintain the plasma discharge.

Similarly, another surface or electrode as indicated by reference numeral 171 can be cleaned with plasma.
Although several kinds of operation modes and typical routines have been described above with reference to the apparatus of the embodiment, those skilled in the art can make various modifications based on the above description and the contents shown in the drawings. It goes without saying that the invention can be implemented.

Inspection of Phase Shift Mask FIG. 16 shows a cross section of a typical phase shift mask 500 having an opaque chrome layer on a transparent quartz substrate 504. The chrome layer can be deposited on a quartz substrate by vapor deposition or the like, and is typically about 0.1 μm thick, and here, it has a square pad shape and has a shape of 506, 508, 510.
Each number is attached. It is desirable to use the phase shift mask 500 in the semiconductor wafer manufacturing process. The phase shift is referred to as a shift and the description will be continued below. The phase shift is an appropriately sized well, also known as trench 502.
(Typically has a depth of 25 microns). The trench is formed by etching a necessary portion of the quartz substrate 504.

Historically, the phase shift mask 500
Has been inspected by an optical method. However,
The optical method has become quite difficult at present due to the fact that the pattern processing size of the semiconductor surface becomes finer, and as a result the size of the mask pattern becomes smaller. Since the present invention makes good use of backscattered electrons and secondary electrons in the electron beam at the time of inspection, even if various phase shift masks and their modified versions can be inspected by grasping the characteristics of their patterns. It is possible. In FIG. 16, the phase shift mask 500
In preparation for electron beam inspection, a thin metal conductive layer 512 such as aluminum or gold, or Nagase Lt
The material TQV501 manufactured by d. is deposited on the entire structure of the mask 500 (including wells described above, of course) by vacuum deposition or the like.

Next, the electrically conductive layer 512 is electrically grounded (501) to circulate and close the electronic circuit. This is because the backscattered electrons and secondary electrons cannot play such a role by themselves. Thus, at the time of inspection, the possibility that the inspection point of the substrate is charged by the electron beam is minimized. The phase shift mask 500 can naturally have various defects. Examples are extra chrome where it shouldn't, unnecessary etching imprints on the quartz substrate, and so on. Quartz columnar debris may fit into the well or may spread over the quartz surface 544. However, it should be no wonder that etching traces that should not be present occur anywhere on the surface of the quartz substrate 540, and it is possible that they exist under the chrome pads.

Referring now to FIG. Here, a simplified phase shift mask is shown, and the user decides "what is it?" Existing under the conductive layer 512 on which the electron beam hits. It becomes possible by using electrons. In the model of FIG. 17, in order to simplify the contents of the present invention and make it easier to understand, one beam of the substrate 500 moves from 100 to 100 'with the passage of time, and points 516 and 516' are respectively moved. , And irradiate only with an electron beam. To further clarify this point, the electron beam system of the present invention has one column and therefore one electron beam. In the figure, the two beams are drawn to show that the only electron beam moves to two positions 516 and 516 'on the surface of the substrate and irradiates them at a substantially right angle as time passes. Shows. Only electron beams 100 and 10
0'is of course the same, and since it is usually in a high energy state of 20 kV, it penetrates through the conductive layer 512, the chrome layer 510, the quartz substrate 504, and anything such as a rigid body. Since the molecular weight of chrome is much larger than that of quartz, beam 10
0 ′ cannot penetrate deeply inside the chrome layer 510 such that the beam 100 penetrates the quartz 504. Of course, in this case, the beam energies are the same in each case. In FIG. 17, the circumstances during this time are shown in 516 and 5 respectively.
At the 16 'point, the droplets are typically divided into two sizes and shown as the teardrops 524 and 526, and the difference in the theoretical beam range is shown by the teardrop size. Beam 10 at the two points
When 0 or 100 'collides with the conductive layer 512, it is usually "SE
Secondary electrons 528, which are referred to as "I electrons", are generated.

When the electron beams 100 and 100 ′ are shot into the chrome layer 510 or the quartz layer 504, respectively, some of the electrons that participate in the collision are scattered and backscattered electrons 538.
And 536. Backscattered electrons 536 and 538 leave secondary layer 540 and 542, respectively, behind conductive layer 512.
As a general name, "SE II Electronics" is produced.
I will call it. In FIG. 8 where the columns are referenced, each 2 detected by the secondary electron detector 117, as described above.
The backscattering detector 160 also outputs an electrical signal in the same manner as it outputs an electrical signal corresponding to the next electron.

To show the implantation effect on the optical mask 500 by the high energy electron beam 100, FIG.
Then, the mask cross section of FIG. 16 is arranged in the upper part of the figure, and the corresponding waveforms 545 and 546 are shown as being due to secondary electrons and backscattered electrons, respectively. Further, in FIG. 18, waveforms 545 and 546 due to the secondary electrons and the backscattered electrons are shown in Material 5 respectively.
It was drawn together with the 00 cross section to express the physical relationship between the waveform of each signal and the surface characteristics of the mask. The two signals depicted in FIG. 18 are representative of the signals obtained from the secondary electron and backscattered electron detectors 117 and 160, respectively, when the electron beam scans the material. Examination of the secondary electron waveform 545 reveals various physical properties of the mask 500. In the figure, the following four results appeared in the electric signal from the secondary electron.

The one signal level is the flat area of the mask 500, shown as 544. The signal level of the second is greater than the signal level of the first and is mask 5 shown as 504 etc. (506, 508 and 512).
00 is a region having a chrome layer structure. The signal level of 3 is smaller than the signal level of 1, and
Is the area of the widest well on the mask 500 indicated as. The four signal level is less than the three signal level and is the area of the small diameter well on the mask 500 shown as 514.

Although not illustrated in the figure, other signal level variations are possible. Examples are thin films of different thicknesses of chrome pads, wells with varying diameters, widths and depths, columns on a quartz surface or quartz columns extending up in the wells. Another feature of the secondary electron waveform 545 is the pulse waves 550 and 552 corresponding to the area of the height transition on the mask 500. Most notably, this occurs at the edges of wells 502 and 514. The large peaks seen in the secondary electron waveform are the result of increasing secondary electrons emanating from the vertical walls of the etched wells 502 and 514. Similarly, the small peak signal is due to secondary electron generation from the sidewalls of chrome pads 506, 508 and 510.

Similarly, the backscattered electron waveform 546 has
The following four results appear. The signal level of 1 is
The flat area of the mask 500 without the chrome structure shown as 544. The signal level of the second signal is higher than the signal level of the first signal, and is equal to 504 (506, 5).
The mask 500, designated as 08 and 512), is the area with the chrome layer structure. The signal level of the 3 is
The area of the widest well below the signal level of one and shown as 502 on the mask 500. Part 4
Signal level is lower than the signal level of 3
The area of the small diameter well on the mask 500, shown as 14.

Further, as is clear from the example of secondary electrons, it can be understood that if there is a physical deformation other than the mask 500, the level of the electric signal will be formed accordingly. Examples of these are thin films of chrome pads of different thicknesses, wells of different diameters, widths and depths, columns on the quartz surface, or quartz columns extending up in the wells. Is.

What has been solved above is that the backscattered electron waveform shows a peak only in the transition region seen between the well and the flat portion. It has been found that the transition regions of such wells have much smaller peaks at the ends of the wells 502 and 514 in the case of backscattering compared to secondary electrons. Further clarification was the chrome pad (506, 508 and 510)
No visible peaks are visible in the transition region between and the flat quartz surface 544. Therefore, the secondary electron and the backscattering waveform are compared with each other, and the position, size, and shape of the well can be definitely determined.

Since the number of backscattered electrons scattered and emitted from the well is determined by the depth and width of the well, it is possible to know the depth and width of the well by using the waveform 546. To do so accurately, it is necessary to calibrate the well depth and backscattered electron waveform with each other, using a sample of the well whose depth and width are known in advance. Now, the number of backscattered electrons can be theoretically calculated by assuming Lambertian radiation in a well sample whose depth and width are known in advance.
Similarly, the secondary electron waveform 545 emits a prominent peak at various transition points in the scan, whereas in the case of backscattered electrons, 546 peaks only at the ends of the well,
First look for the end of the well in the backscattered electron peak,
Then, it is known which of the many peaks of the secondary electron waveform 545 corresponds to the edge of the well. And the secondary electron waveform 54
The information on the other peaks in 5 can be included to determine the end of the well with high reliability. Moreover, regarding the level of the secondary electron waveform 545, comparing the first level representing the quartz substrate flat portion 544 and the third level corresponding to the well bottom 502, even if the same quartz is used, the first signal level is obtained. Pay attention to the bigger thing. It is clear that such a difference comes from the fact that the ratio of secondary electrons when they are scattered from the well is small compared to the flat portion. The depth of each well is determined by calibrating the secondary electron scattering from the well in the same manner as before, and using the waveform signal of the backscattered electron at the same place as well. You can do it.

As understood from the above, the chrome structure on the surface of the mask 500 causes backscattering and 2
It can be concluded that the signal level of both secondary electrons is high. As mentioned above, since the molecular weight of chrome is larger than that of quartz, the amount of electrons scattered back is naturally large, so it can be considered that more SE II electrons are generated than on the quartz surface. Moreover, both the secondary electron waveform 545 and the backscattered electron waveform 546 exhibit an upwardly convex curve (see 548) at the etched well. The characteristics of these waveforms are:
It can be used to distinguish between the 544 surface and the plane of the etched well bottom. Thus, in order to find a defect, since the physical characteristics of the inspection substrate are reflected, it suffices to investigate and compare the signal waveforms of the secondary electrons and the backscattered electrons for each region. This comparison operation may be performed by comparing two or more same patterns in one mask substrate with each other, or by comparing the same positions with other masks having the same design.

On the other hand, as described above, the target for investigating the signal waveforms of the secondary electrons and the backscattered electrons for each region and comparing them later is the image obtained from the data of the CADS database in which the mask is created. May be. To carry out the above process, first, the signal waveforms 545 of the secondary and backscattered electrons are
And 546 are each digitally processed to digitally perform various functional actions similar to the example described above. Furthermore,
After the inspection, when the conductive layer 512 is opaque to the light wavelength or other lithographic media, the 512 layer should be removed as it is unsuitable for production and the like. For example, when doing so, specifically when the conductive layer 512 is copper or aluminum, 0.3
4 normal KOH solution, gold for KOH + I solution (potassium iodide plus iodine) diluted to 300: 1 with water, and for polymers dissolved in typical organic solvent on quartz substrate It is possible to remove it without damaging it.

In the inspection mask, the energy of the electron beam 100 can be adjusted and optimized in order to maximize the signal-to-noise ratio when outputting the signal waveforms of secondary electrons and backscattered electrons. As seen above, the envelope or teardrop 524 of FIG. 17 is scattered within the chrome layer 510, resulting in a teardrop depth that approximately matches the thickness of the chrome layer 510. Further, as shown in FIG. 18, when the levels of the signal waveforms of the secondary electrons and the backscattered electrons are observed with respect to the chrome layer and the quartz surface, compared with the fluctuation of the secondary electron signals at the much smaller signal level, the backscattering The fluctuation of the electronic waveform is much larger.

When setting initial parameters for mask inspection, it is desirable to optimize the electron beam energy level. The reason is that the exact characteristics of the phase shift mask cannot be known in advance and cannot be predicted. Similarly, the location of the chrome is not very sensitive, but it must be taken into account for the depth of the well as the accuracy of the measurement changes at different energy levels of the beam.

There are various detectors for increasing the efficiency of capturing and detecting backscattered electrons. For example, in a semiconductor detector, a combination of a scintillator and a photomultiplier, or in a microchannel plate, it is important to arrange the main beam and its tail so that they are aligned. These types of detectors are described in detail by Ludwig Reimer in "Scanning Electron Microscopy", "Scanning Electron Microscopy", Springer-Verlag, pages 181-182 and 189-190.

FIG. 19 shows the electron beam shown in FIG.
An inspection substrate 50 which is an objective lens portion provided in the column.
An installation state of an annular backscattering detector 560 for capturing and detecting backscattered electrons generated from 0 is shown as an example in the present invention. As described above in connection with FIG. 7, the low-positioned pole 106 and the middle-positioned electrode 107 are drawn in FIG. 19, and at the same time, backscattered electrons pass through the pole 106, the electrode 107 and the central annular opening. A specific positional relationship with the annular backscattering detector 560 is shown as an example.

20 and 21 respectively illustrate other structural features found in certain phase shift masks. In the upper part of FIG. 20, a cross section of a mask 500 ′ having a quartz substrate 504 ′ is drawn, and a well 502 ′ expresses a pair of chrome pads 562 protruding from opposite sides at their openings. In other words, the pair of chrome pads 562 are “digged” under the word “protruding”. Hereinafter, “digging” will be referred to as undercut. Further discussion, quartz substrate 504 '
Together with the expected waveforms 54 of secondary and backscattered electrons
5'and 546 'are drawn downward and shown in relation to the mask cross section to show the relationship between the waveform and the pattern on the mask. Of course, the waveforms of the secondary electrons and the backscattered electrons in FIG. 18 shown above have already been displayed in correspondence with various other main patterns including the surface of the mask, for example, the well 502 '. There is. The difference between this example and the previous example is 2
The secondary and backscattered electron waveforms 546 'and the negative going pulses 566 and 568 seen therein, indicating the presence of the preceding undercut. Although not drawn, some electrons penetrate the chrome pad 562 and the well 50
This small negative pulse is generated due to the fact that it rushes into 2'and therefore produces a smaller number of secondary electrons and backscattered electrons than it would otherwise, ie without an undercut. Of course, when it is not so, that is, when there is no undercut, it means that the quartz substrate is in direct contact under the chrome pad. Thus, using the comparison method described above, we can deduce that there is a chrome pad that extends above the well and there is an undercut in this inspection mask due to the characteristic negative waveform added here. You can do it.

Similarly, a quartz substrate 50 is provided on the upper portion of FIG.
A mask 500 ″ having 4 ″ is represented and a pair of quartz pads 570 is placed on its surface to provide another example of a phase shift material. In the same manner as before, the mask 500 ''
At 54, expected waveforms of secondary electrons and backscattered electrons
5 ″ and 546 ″ are drawn below and shown in relation to the mask cross section to show the relationship between the waveform and the pattern on the mask. Expected Waveforms of Secondary and Backscattered Electrons 545 ''
The shapes of and 546 "present some interesting pulse shapes in the mask 500" combination. As expected overall, there are more 2 near the quartz pad 570 compared to the other locations labeled 544.
It is known that secondary electrons and backscattered electrons are generated. The basic feature described above is, of course, a result of the extra thickness of quartz in place.

Another basic feature related to the waveforms of the secondary electrons and the backscattered electrons is that the signal level on the quartz pad 570 gradually decreases as compared to the 544 region regardless of which direction the signal level approaches. is there. Similarly, the waveform of secondary electron 545 ″
With respect to, the signal level on the quartz pad 570 peaks at its outer edge and tapers off toward the center. Similarly, regarding the backscattered electron waveform 546 ″, the quartz pad 57
The signal level above 0 has a small peak at its corners, but apart from this it is generally relatively flat. In addition, in the case of the quartz pad 570, the secondary electron waveform 545 ″ is narrow and does not generate a high pulse.
Different from wells and chrome pads. Therefore, the comparative technique described above with reference to FIG. 18 can be further developed to automatically determine whether the phase shift material is a chrome pad or a well and whether there is an undercut. Of course, the presence or absence of the necessary phase shift material, its size, location, etc.
It can be known from the waveforms of secondary electrons and backscattered electrons.

As described above, the search for defects may be performed simply by comparing signals, ie, waveforms, and the object of comparison is a similar waveform obtained from another mask having the same design, or otherwise,
Use the waveform calculated in advance using the database. In these comparison operations, all different tolerances are applied depending on whether the source of the electronic signal is a chrome layer, quartz, a well, another material, or a foreign substance. Therefore, for example, if the waveform depicted in FIG. 18 is judged by an electronic system and various conditions and conditions are input, the structure and characteristics of the mask surface can be determined only from the above observed waveform. In addition to the spike heights shown at 550 and 552, what is taken into consideration when identifying defects and the like using a waveform is a curvature like 548, which recognizes trenches like 502 and 514.

In the automatic analysis of the phase shift mask as described above, the reference of the automatic waveform recorder which has similar contents and which has been extensively studied in the past can be referred to. For example, Stockman by GC Stockman: "A Problem-Reduct
ion Approach to Linguistic Analysis of Waveforms "
"Linguistic Vocal Problem Solving Method for Waveform Analysis", Univ.
of Maryland, Department of Computer Science, May
There is 1977. This type of pattern recognition is often found in the literature on syntactic pattern recognition, which analyzes samples based on features such as peaks and valleys. In applying the above to the phase shift mask, the bottom of the well, that is, the trench or the like can be identified from its waveform characteristics. The specific method is to observe an upward convex curvature in the waveforms of secondary electrons and backscattered electrons in the low signal value range, and to generate spikes of secondary electron signal waveforms seen at both ends of the trench. It depends on grasping the characteristics such as. Larry S. from the "Pattern Recognition and Image Processing Handbook".
Davis's work "Two-dimensional representation"("Two-Dimensional
Shape Representation "Handbook of Pattern Recognit
ion and Image Processing, Academic Press, Inc., Sa
n Diego, CA, 1986) pp. 233-245, examples of well bottoms, or other pattern shapes, such as quartz on quartz (Fig. 21) can also be searched by the method described in Chapter 10 above. .

In addition, when the methods of the above two documents are used for the comparison method, the allowable values corresponding to the conditions of different masks are applied, and it depends on the combination of various patterns searched for. The structure can be determined with high reliability.

The various operation modes of the present invention have been introduced in relation to typical routines and their apparatus. The true intent of the invention is that if a person who is familiar with the present technology reads this specification with reference to the drawings, it is a modification that is within the scope of the invention and is not described here. It seems to come up. For this reason, it should be construed that, even in the case where the application is not described in the scope of claims for the purpose of the present invention, something that conflicts with the true core of the invention is included in the scope of the claims.

In summary, one of the gist of the present invention is to make the mask surface electrically conductive, and in addition to that, secondary electrons and secondary electrons formed due to the difference in the topology of the mask surface and the molecular weight of the structural material are analyzed. The characteristic is that the waveform of the backscattered electrons in the low signal value range is used for the analysis. Although various explanations have been given above with reference to the drawings, there are other methods for making the mask surface electrically conductive besides coating the finished mask.

Another technique is, for example, a method of applying a conductive transparent film such as ITO (oxide of indium and tin) before patterning the chrome layer. Of course, if the above technique is used, it is not necessary to remove the conductive transparent film between the pads after forming the chrome pads. Since this film has electric conductivity and is transparent in line with the gist of the invention, it does not interfere with the exposure.

On the other hand, it is possible to make the quartz substrate conductive by ion implantation. The merits of making the quartz substrate semi-permanently conductive are that it facilitates the repair of the mask due to the ion beam, and that it can prevent damage to the structure due to static electricity.

Moreover, the inspection technique of the present invention is not limited to the phase shift mask. There is a definite reason for using the example of the phase shift mask to describe the method of using the waveforms of the secondary electrons and the backscattered electrons in the low signal value range in the description of the present invention. It is known that the phase shift mask is easy to provide various discussions for convenience, has a variety of surface elements, and has a high degree of completion of inspection. The phase shift mask is the most complicated and includes various surface elements, and universal inspection techniques can be used. It is no exaggeration to say that all of these can be applied to other mask inspections. It is self-evident that the technique of the present invention can be applied to the inspection of many optical masks.
A structure in which certain layers are superposed, for example, a glass film formed by solidifying an aqueous glass called spin-on glass on a substrate can be distinguished from the substrate or the chrome film by the method of the present invention. . Immediately, a method of generating secondary electrons and backscattered electrons and using their signal waveforms can be similarly used.
Furthermore, as an example of the light attenuating mask material, there is an ultrathin layer of chrome having electrical conductivity, and it should be noted that this can also be inspected and analyzed by the above method. Similarly, it is added that the present invention is effective in inspecting a proximity mask, which is a kind of optical mask.

[Brief description of drawings]

FIG. 1 is a block diagram of the entire system of the present invention.

FIG. 2 is a schematic diagram of a scanning pattern used for die-to-database comparison inspection.

FIG. 3 is a schematic view of a scanning pattern used for die-to-die comparison inspection.

FIG. 4 is a graphical illustration of a multi-layer frame scan integration technique for obtaining images averaged over several scan areas.

5 shows the normal X-ray of the electron beam during the scan shown in FIG.
The figure which plotted the deflection value of a direction as a function of time.

6 is a graph showing the X coordinate of the electron beam on the substrate in the scan shown in FIG. 4 as a function of time.

FIG. 7 is a schematic diagram showing an electro-optical beam generation control tower (column) and functional elements of a collection system therein.

FIG. 8 is a schematic diagram showing the paths of primary electrons, secondary electrons, backscattered electrons, and transmitted electrons passing through the inside of the electron optical beam generation control tower (column) shown in FIG. 7, particularly the collection system thereof.

FIG. 9 is a schematic diagram of a multi-head electron gun and a vacuum system.

FIG. 10 is a block diagram of a positioning control system of the present invention.

FIG. 11 is a schematic diagram of a vacuum system of the present invention.

FIG. 12 is a block diagram of an analog deflection system of the present invention.

13 is a block diagram of the memory of the present invention shown in FIG.

FIG. 14 is a block diagram of a preprocessor for image capture of the present invention.

FIG. 15 is a modified schematic of the electro-optical control tower of FIG. 7 to show the electrical components of the plasma oxidation subsystem.

FIG. 16 is a cross-sectional view of a phase shift mask formed by etching a well for generating a phase shift on a quartz substrate and patterning a chrome layer on the well.

FIG. 17 is a diagram showing how an electron beam penetrates and scatters at two typical places on a phase shift mask in a quartz layer and a chromium layer and is scattered by a tear drop.

FIG. 18 shows the physical structure of the mask, the secondary and backscattered electron waveforms, and the alignment with the mask cross-sectional view of FIG. 16 in order to match the physical structure on the mask and to understand the state of the corresponding waveform. FIG.

19 is a schematic diagram showing an objective lens portion of the annular backscattering electron detector in the electron beam control tower of FIG. 7 to capture the backscattering electrons emitted from the phase shift mask. Is.

FIG. 20 shows the physical structure of the mask and the secondary and backscattered electron waveforms, both aligned with the mask cross-section, in order to understand the state of the corresponding waveforms in agreement with the physical structure on the mask. FIG. Here, the physical structure is particularly the case where the well extends and the chrome layer above it is undercut.

FIG. 21 shows the physical structure of the mask and the secondary and backscattered electron waveforms, both aligned with the mask cross-section, to understand the state of the corresponding waveforms in agreement with the physical structure on the mask. FIG. Here, in particular, the physical structure is that of a patterned quartz layer.

[Explanation of symbols]

 10 ... Inspection system, 20 ... Electron beam column, 21 ... Positioning computer, 22 ... Positioning optical system, 23 ... Data bus, 24 ... XY stage, 26 ... Stage servo, 27 ... Interferometer, 29 ... VME 1 , 30 ... Analog deflection circuit, 31 ... VME 2, 32 ... Detector, 33 ... Signal, 34 ... Substrate handler, 36 ... System computer, 38 ... Computer display, 40 ... Keyboard, 42 ... Column control computer, 44 ... Video frame buffer, 46 ... Image display, 46 ... Image capture preprocessor, 50 ... Deflection controller and roller, 52 ... Memory block, 54 ... Database adapter, 56 ... Defect processor, 57 ... Substrate, 58 ... Post・ Processor, 70 ... Die, 100 ... Beam, 500 ... Phase shift mask, 501 ... Ground, 502 ... Well, 506 ... Chrome pad, 512 ... Conductive layer, 545 ... Secondary electron waveform, 546 ... Backscattered electron waveform, 562 ... Undercut chrome pad, 570 ... Quartz pad ..

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification number Internal reference number FI Technical indication location H01L 21/66 J 7514-4M (72) Inventor Dan Meisberger San Francisco, CA 95120, USA Jose, Montalban Drive 1507 (72) Inventor Alan Dee Brodie 94303, Palo Alto, Van Oken Circle 998 (72) Inventor Thuong-Wai Chen 95129, USA San Jose, Blanev Avenue 1561 (72) Inventor Jack Wye Joe, Burlington Terras 2721 (72) Inventor, Burlington Terras, California 94536, USA 94536 (72) Brian Jay Glenon, United States, Vermont 05446, Colchester, Dunlop Way 3

Claims (12)

[Claims] 1. A thin film forming apparatus for forming a conductive mask layer by coating the surface of an optical mask with a conductive thin film; and a ground connection means for electrically grounding the conductive thin film potential. An electron beam source for generating an electron beam; a charged particle beam control tube for sending the electron beam onto the surface of a conductive mask and allowing the beam to be swept and scanned; and a conductive mask A backscattered electron detector that detects backscattered electrons from the surface of the conductive mask and outputs a backscattered electron waveform while the electron beam scans the conductive surface; Detect and
And a secondary electron detector that outputs a secondary electron waveform while the secondary electron scans the conductive surface; 2. The system according to claim 1, wherein the optical mask is a phase shift mask. 3. The system according to claim 1, wherein the thin film forming apparatus is a vapor deposition apparatus. 4. The system according to claim 1, wherein the thin film forming apparatus is a sputtering apparatus. 5. The method according to claim 1, wherein a memory is introduced to connect the backscattered electron detector and the secondary electron detector to each other so that the backscattered waveform and the secondary electron waveform are stored in the memory. System that is deployed in. 6. The system according to claim 1, wherein a processor is installed to evaluate the backscatter waveform and the secondary electron waveform, respectively, and thereby to evaluate the characteristics of the mask thin film. 7. The surface of the optical mask is coated with a conductive thin film, (b) the conductive thin film is electrically grounded to the ground, and (c) the conductive film described in step (b) is then used. The electron mask is swept so that the electron beam scans on the surface, and (d) backscattered electrons are detected to obtain the backscattered electron waveform generated from the conductive mask described in step (c), (e) An automatic mask inspection method including a procedure of detecting secondary electrons in order to obtain a secondary electron waveform generated from the conductive mask described in step (c). 8. The automatic mask inspection method according to claim 7, wherein the conductive mask is a phase shift mask. 9. The automatic mask inspection method according to claim 7, wherein step (a) includes (f) a step of producing the coated thin film by a vapor deposition method. 10. The automatic mask inspection method according to claim 7, wherein step (a) includes (g) a step of producing the coated thin film by a sputtering method. 11. The automatic mask inspection according to claim 7, further comprising: (h) storing backscattered electron waveforms and secondary emission electron waveforms obtained in steps (d) and (e) in a memory, respectively. Method. 12. The method according to claim 7, (i)
An automatic mask inspection method in which the backscattered electron waveforms and the secondary electron waveforms obtained in (d) and (e) are evaluated, respectively, and (j) the coating thin film characteristics are evaluated.