EP0904558A2

EP0904558A2 - Confocal microscopic device

Info

Publication number: EP0904558A2
Application number: EP98912431A
Authority: EP
Inventors: Thomas SCHERÜBL; Norbert Czarnetzki
Original assignee: Carl Zeiss Jena GmbH
Current assignee: Jenoptik AG
Priority date: 1997-03-29
Filing date: 1998-03-03
Publication date: 1999-03-31
Also published as: US6674572B1; WO1998044375A2; JP2000512401A; DE19713362A1; WO1998044375A3

Abstract

An autofocus for a confocal microscope is realized by means of a confocal microscopic device with scanned illumination of an object (5), means to produce a first wavelength selective division (e.g. chromatic color-length defect) of the illuminating light, means (6) to produce a second wavelength selective division of the light coming from the object, and means of detection (10) to detect the light distribution produced by the two means above-mentioned. Spectral division and detection of a wavelength selectively illuminated object image occur, and a control signal to adjust focal position is generated on the basis of determined deviations in frequency and/or intensity from a predefined reference value corresponding to the position of an object. The invention also relates to a method for detecting deviations between one first vertical profile and one second vertical profile, preferably to detect and/or control defects in semiconductor structures, wherein wavelength-selective illumination of a first object is carried out, the light arising from the first object is detected and electronically compared with a second object.

Description

Confocal microscopic arrangement -

From GB 2144537 A a device for profile measurement of surfaces is known which has a polychromatic light source.

The illuminating light is split longitudinally spectrally and focused on an object, with each focus point corresponding to a certain wavelength. The light reflected from the object reaches a dispersive element via a beam splitter and is focused by this onto a photodiode array. The strongest signal is determined by reading the photodiode array and related to the object surface.

WO 88/10406 describes a device for measuring distances between an optical element with high chromatic aberration and an object, also for profile measurement, with a structured light source and a spectrally dispersive apparatus and a CCD camera.

US Pat. No. 4,965,441 describes a scanning, ie point-by-point confocal arrangement with increased depth resolution, dispersive elements being arranged in the evaluation beam path for wavelength separation.

The principle of a scanning imaging method with an objective lens of strong chromatic aberration and spectral decomposition of the light backscattered from the object is also the subject of DE 4419940 A1.

WO 92/01965 describes an arrangement for simultaneous image generation with a moving pinhole in the illumination beam path, the objective having a high chromatic aberration and the arrangement being intended to be used as a profile sensor.

WO 95/00871 also describes an arrangement with a moving pinhole and a focusing element with axial chromatism, wherein in

Evaluation beam path two cameras upstream spectral filters are provided.

A signal division for determining the wavelength occurs pixel by pixel. Such arrangements are also known from SCANNING, Vol.14, 1992, pp. 145-153.

A system for generating a color height image is also implemented, for example, with the CSM additive for the applicant's axiotron microscope.

The height profile is inspected by visual evaluation of the color image.

A special application is wafer inspection, i.e. the detection of defects on wafers (e.g. particles on top, irregularities in the structure). Defects become visible as areas with different colors. In this way, height differences <0.1 μm can be differentiated in color. The accessible height range, the color spread, depends on the lens used and is, for example, 4 μm for a 50x lens.

With automatic wafer inspection, rapid detection of defects is necessary. Common methods used in practice work on the basis of laser scattering or digital image processing. Currently Defects in the 0.2 μm range can be detected with a typical throughput of 20 wafers (0 = 200 mm) per hour. Laser scattering methods are limited to the detection of particles (dirt, dust), while digital image processing can also detect other types of defects, such as structural defects, so-called "pattern defects".

With increasing integration density of electronic circuits, lower detection limits are required. With a 1 GB DRAM, for example, the detection of a defect size of 0.1 μm is considered necessary. In processes with digital image processing, the minimum detectable defect size is determined by the resolution, the inspection speed by the computer capacity. Considerable electronic computational effort has to be made to process the information at the appropriate speed. If, for example, a wafer with a diameter of 300 mm is scanned with a 0.3 μm grid, a total of 10 ¹² pixels must be processed with digital image processing. The object of the invention is to enable a rapid yet highly accurate detection of wafer defects.

The object is achieved by the independent claims. Preferred developments are described in the dependent ones.

The method of integral spectral decomposition and analysis of the color image information shown has the advantage that defects whose height expansion is less than 0.1 μm can be detected.

The optical preprocessing by means of spectral decomposition also enables faster processing of the present height profile, since the height profile of the entire image field is compressed in the spectrum. An arrangement according to the invention can also be used to implement a highly precise autofocus, which will be discussed in more detail below.

The invention and its effects and advantages are explained below with reference to the schematic representations. Show it:

Fig. 1: The measuring and evaluation principle according to the invention

Fig. 2: Different height profiles and corresponding color - height spectra.

Fig. 3: The shift of a spectral line due to a height difference

Conductor surface Fig.4: The mapping of the generated intermediate image on one

Spectrometer entry slit Fig. 5: A first version with cross-section converter Fig. 6: A second version with cross-section converter Fig. 7: A version with a color camera for image evaluation Fig. 8: A version with several lasers of different wavelengths Fig. 9: Spectra for different height profiles Focusing on two levels with laser wavelengths 11 and 12 Table 1: Color spreads for different lenses 1 with a light source 1, which can be a white light source, but can also consist of several lasers of different wavelengths or multi-line lasers, collector lens 2

, Mirror 3 and an optical divider element 4, a hole pattern 5, which is located in the intermediate image plane of a microscope objective 7, is illuminated.

The hole pattern 5 is preferably a rotating one

Nipkow disc or a double Nipkow disc with

Microlens arrangement, for example described in EP 727 684 A2.

The grid arrangement is moved appropriately for scanning the image and generates a parallel confocal image.

By means of a chromate 6, a targeted longitudinal color error is introduced into the beam path in such a way that after the beam has passed through the color corrected one

Objective 7, the focus points 8 of the different color components of the light source 1 are in different planes.

This gives a parallel confocal image of object 0, in which the

Height information is optically coded by a corresponding color representation.

This color representation is now by a dispersive element 9, which can be, for example, a prism or holographic grating on a diode row or

CCD - line 10 shown, which is connected to an evaluation and processing unit 14.

The color-coded height profiles are thus broken down into an equivalent color spectrum. This is shown in Fig. 2 a - d. The spectral positions of the maxima correspond to the corresponding contour lines, while the number of

Events (area below the maximum) the area share of the respective

Height levels is proportional.

The object is located on an xy shift table, not shown, and is scanned either in a continuous movement or in a step-and-go procedure.

The adjustment of the sliding table is synchronized with the reading of the receiver 10, so that the read-out is clearly assigned

Height information about the location on object 0 is given. The height setting in the z direction is controlled by an autofocus system or by a special height control algorithm described below. 2 a - d shows a height profile H and a defect point F to be detected as well as the spectrum S AD corresponding to the profile and measured on the receiver 10 and the difference spectra S AB.A-C, AD formed during the evaluation.

It is clear that by comparing a spectrum SA corresponding to the ideal height profile with the spectra SB - SD actually measured during the wafer inspection, information about the type and scope of the existing defect location F can be obtained. Some error types F are shown in FIG. 2 by way of example and without restriction of generality. Case A in Fig. 2a indicates an ideal, error-free height profile. In Fig. 2b a particle F is outlined on the top layer, in Fig. 2c a particle between two height structures H. 2d shows a faulty connection V between the two height structures H. It can be seen that the corresponding spectra SA-SD differ depending on the type of defect. Likewise, the corresponding difference spectra A-B, A-C, A-D differ depending on the type of error present with regard to the position, width and intensity of the corresponding maxima. These differences allow conclusions to be drawn about the type and size of the error.

Various methods can advantageously be used to detect defects in applications, for example in wafer inspection: 1.

Two adjacent spectra are recorded and subtracted from one another, as shown schematically in FIG. 2. With a routine, for example peak search routine, as is usually used in optical spectroscopy, the maxima in the difference spectrum are determined which are above a predetermined noise level. If such maxima are present, there is a defect in the area examined. The position and the corresponding spectra or the reference spectrum are saved. The spectra are evaluated in more detail later in a classification unit, since the position and half-width range contain further information about the type of defects that are used for a defect classification. 2. The comparison of spectra is carried out as described under 1., but an ideal stored spectrum is used as the comparison spectrum.

3. First several spectra are recorded and a reference spectrum is formed from these by averaging. This averaged reference spectrum is then compared with a current spectrum as described under point 1.

In the case of the specified defect detection algorithm, a height control is necessary, since a difference in height leads to a shift of the spectra to one another. In principle, this height control can be carried out with a conventional autofocus system or with the height control algorithm described below, which uses the spectral height information.

In Fig. 3, the spectral deviation Δλ from a wavelength λo, which corresponds to a height deviation Δz, is shown schematically.

The height of the object to be examined relative to the imaging optics is determined by a suitable z setting, e.g. B. z-table chosen so that a certain spectral line corresponds to a certain height level of the object. This line λ ₀ (target maximum) ideally characterizes an excellent level, for example the conductor track surfaces of a wafer.

Fluctuations in height (z direction) by an amount Δz then correspond to a shift in this main line by a corresponding Δλ, depending on the color spread set and chromates 6 used (see, for example, Table 1).

For any recorded spectrum, the exact position of the maximum is now determined in the range of λ ₀ .

If the maximum deviates from a predetermined value of λ ₀ , the Z table position is readjusted using an adjusting element, preferably a piezo adjusting element. If the deviation is less than this previously defined value, the entire spectrum is shifted accordingly by electronic means and then one of the defect detection algorithms described above is carried out. In addition, the above-mentioned method for height control is suitable - advantageously as an autofocus method for a confocal microscope. The advantages of confocal microscopy are known to be that a defined object plane, the focal plane, is worked out in the image. The confocal principle suppresses the optical image from another level. Therefore, only the focus plane is visible in the image. In the case of wafer inspection, a certain wafer level can be focused using the confocal principle. All other layers appear dark in this image. In applications in which only a certain level is to be examined, the confocal method is therefore advantageous. This simplifies digital image processing for defect detection.

However, the use of confocal microscopy for the analysis of special levels, for example a wafer, requires a highly accurate autofocus system that focuses precisely on the level of interest. Usual autofocus methods (such as the triangulation method) only measure the altitude of a certain object point. Depending on the structure at hand, this does not necessarily focus on the level of interest. Averaging processes over several object points also focus only on any plane. However, the present invention permits the recording of a height histogram, ie the height distribution over a specific object area, by means of the spectral analysis. An evaluation of this distribution according to the height control algorithm given above then enables focusing on a specific plane, which corresponds, for example, to the wavelength λ ₀ in FIG. 3.

In FIG. 4, a microscope objective 7 with a tube lens (not shown) generates an intermediate image Z, which lies in the plane of the pinhole arrangement 5. This intermediate image is imaged, for example, on the camera output KA of the microscope with the aid of imaging optics 9.

Part of the field of this camera output KA is imaged through a gap 11, which is preferably located in the image plane of the camera output or a plane optically conjugate to it, into a diode array spectrometer, which consists of a grating 12, here a holographic grating, a diode row 13 and an evaluation unit 14, consisting here of memory, display unit and comparator. The object is scanned by a step-and-go mode with an xy- table, not shown.

Known interferometric displacement measuring systems are provided for detecting the table position, for example, the control and detection of the table position in the x, y and Z direction being coupled to the spectral evaluation by a connection to the evaluation unit 14. In this embodiment, therefore, only a part of the intermediate image field is transferred to the spectrometer. The evaluation or defect detection takes place in the evaluation unit by means of a comparison with a stored ideal height professional (die-to-database comparison) or with one or more previous height profiles (die-to-die comparison) in accordance with the defect detection algorithm shown above with the height control algorithm shown.

In a further advantageous embodiment according to FIG. 5, part of the intermediate image field imaged on the camera output KA is mapped with a disordered glass fiber bundle 15, which serves as a cross-sectional converter, in a diode array spectrometer consisting of a grating 13, a diode array 13 and an evaluation unit 14, by the fibers are arranged in the intermediate image plane as a light entrance bundle and in front of the entrance slit 11. As a result, a larger area of the intermediate image field can be spectrally analyzed at the same time.

The evaluation and defect detection as well as the grid movement and z-control is carried out as already described.

For an integral spectral analysis of the entire intermediate image field, an anamorphic (cylinder-optical) cross-sectional converter 16 is used to image the intermediate image in the spectrometer slit 11 in a further advantageous embodiment according to FIG. 6, ie the imaging scales in the slit direction and perpendicular to the slit direction are different. This embodiment has the advantage that the entire intermediate image field or a larger part of it can be spectrally analyzed. The evaluation and defect detection as well as the raster movement and z control - are carried out as already described.

In a further embodiment according to FIG. 7, a camera K, which is mounted on the camera output KA of the microscope, is used for the spectral analysis of the entire intermediate image.

When using a black-and-white camera, the analysis is carried out pixel by pixel by comparing the gray values, when using a color camera by pixel-by-pixel color comparison. The gray values or colors are counted pixel by pixel and displayed spectrally according to color or gray value in a histogram. Alternatively, the spectral color information can also be obtained by using two black and white cameras in combination with different filters, as described in WO 95/00871 and SCANNING, Vol. 14, 1992, pp. 145-153. The height control is carried out by an autofocus system, preferably by the autofocus method shown above, and the defect detection by the defect detection algorithm shown.

FIG. 8 shows a further embodiment, the white light source being replaced by illumination with 3 differently colored lasers L1-L3. By using lasers, different image planes can now be highlighted. The resulting spectrum consists only of the lines of the laser wavelengths λi, λ ₂ and λ _{3 used} . The integral number of events of the respective lines is proportional to the area share of the corresponding focus level. If there are defects in this plane, the number of photons detected decreases accordingly, as shown schematically in FIG. 9 for two wavelengths.

Defects therefore only manifest themselves in the integral number of detected events, but not, as in the embodiments with a white light source, also in different wavelengths. In this embodiment, the cross-sectional converters shown in the described embodiments are used as cross-sectional converters 14. The anamorphic figure 16 is shown.

In the case of defect detection, the specified defect detection algorithm is simplified. After subtracting the spectra only in the range of laser wavelengths used events are counted above a noise threshold. The number of these events is then proportional to the area share of the defect on the corresponding contour line.

Table 1 shows the typical color spreads for different lenses.

Table 1 Color spread

Claims

Patent claims

1.

Confocal microscopic arrangement consisting of an illumination arrangement for the grid-shaped illumination of an object, first means for generating a first wavelength-selective splitting of the

Illumination light and second means for generating a second wavelength-selective

Splitting the light coming from the object, parallel for several points of the

Object and detection means for detecting the objects generated by the second means

Light distribution.

2.

Confocal microscopic arrangement according to claim 1, wherein the first means are at least one optical element for generating a longitudinal chromatic chromatic aberration.

3.

Confocal microscopic arrangement according to at least one of the claims

1 or 2, whereby the object is illuminated via a perforated or slotted grid disk.

4.

Confocal microscopic arrangement, according to at least one of claims 1 - 3, for determining deviations of at least a first object image from at least one simultaneously or previously detected and / or stored second object image by electronically comparing the detected spectral distribution of the first sample image with that of the second sample image.

5.

Confocal microscopic arrangement, according to at least one of the claims 1 - 4, whereby the illuminated object is imaged onto the entrance slit of a spectrometer.

6.

Confocal microscopic arrangement according to claim 5, wherein the entrance slit is arranged in an intermediate image plane into which the illuminated

object is depicted.

7.

Confocal microscopic arrangement according to at least one of claims 1-6, wherein means are provided for adapting an intermediate image generated by the illuminated object to the entrance slit of a spectrometer.

8th.

Confocal microscopic arrangement, according to at least one of claims 1 - 7, with means for cross-sectional conversion between an intermediate image plane arranged in the direction of detection and the entrance slit of a spectrometer.

9.

Arrangement according to claim 8, wherein fiber optics are provided for cross-sectional conversion.

10.

Arrangement according to claim 8, wherein anamorphic optics for

Cross-sectional conversion is provided.

11.

Confocal microscopic arrangement, according to at least one of claims 1-

10 whereby the object is illuminated with white light.

12.

Confocal microscopic arrangement, according to at least one of claims 1 -

10, whereby the object is illuminated with at least two light sources, preferably lasers of different wavelengths and/or at least one multi-line laser.

13.

Confocal microscopic arrangement, according to at least one of claims 1 -

12, with a moving grid disk in the illumination beam path.

14.

Confocal microscopic arrangement according to at least one of the claims

1 - 12, whereby the illuminated object is moved to generate a scanning process.

15.

Confocal microscopic arrangement according to at least one of the claims

1 - 14, with a color or black and white camera in the detection beam path.

16.

Confocal microscopic arrangement according to at least one of claims 1 - 15, wherein at least one prism and / or grating is provided for spectral splitting.

17.

Confocal microscopic arrangement according to at least one of the claims

1 - 16, with CCD lines or diode lines being provided as detection means.

18.

Autofocus for a confocal microscope, whereby a spectral splitting and detection of a wavelength-selective illuminated object image takes place at least point by point and a control signal for adjusting the

Focus position is generated by means of the vertical object position and / or the imaging system of the microscope.

19.

Method for determining deviations of at least a first height profile from at least a simultaneously or previously detected second height profile, preferably for detecting and / or controlling defects

Semiconductor structures, preferably by means of a confocal microscope according to one of claims 1 - 18, wherein a first object is illuminated with a light source in a wavelength-selective manner and the light coming from the first object is detected and compared electronically with ^a previously or simultaneously detected second object.

20.

Method according to claim 19, wherein a further wavelength-selective splitting takes place before the detection.

21.

Method according to at least one of claims 19 or 20, wherein a comparison of neighboring areas of one and the same object takes place.

22.

Method according to at least one of claims 19 - 21, wherein a comparison is made with a stored object.

23.

Method according to at least one of claims 19 - 22, wherein a comparison is carried out by averaging the detection of several object areas or objects and forming a reference value.

24.

Method according to at least one of claims 19 - 23, wherein a comparison is made by subtracting two recorded and / or stored

Images are taken.