TW200408806A - Detection method and apparatus - Google Patents

Abstract

A method of quality control of cleaning processes for semiconductors such as silicon during device fabrication comprises taking an as-cleaned semiconductor structure following a critical cleaning process step; exposing the surface of the semiconductor structure to at least one high-intensity beam of light from a suitable light source, preferably a laser, and in particular a high-intensity laser, and collecting photoluminescence (PL) produced by excitation of the semiconductor structure by the light beam; making an analysis of the collected photoluminescence signal and using that analysis as the basis for a quality classification of the cleanliness of the semiconductor, in particular by: determining an average photoluminescence intensity emitted across the area of the structure; comparing the average with a predetermined acceptable specification range of photoluminescence; making a quality classification of the cleanliness of the semiconductor structure based thereon, and in particular rejecting or selecting for remedial action such as further cleaning semiconductor structures exhibiting a photoluminescence response outside the said predetermined acceptable specification range. In a preferred embodiment the method is applied as part of a quality control metric during device cleaning and processing. In a refinement of the method a spatially resolved PL map is also collected. An apparatus for performing the method is also described.

Description

200408806 (1) Description of the invention: [Technical field to which the invention belongs] The present invention relates to a non-destructive method and device for detecting surface layer metal pollution in semiconductors (such as silicon). The component manufacturing process performed on the semiconductor wafer or the wet cleaning of the semiconductor wafer performed during the process is introduced. In particular, the present invention provides improved quality control metrics for wet cleaning processes during component manufacturing. [Previous Technology] Over the past 50 years, the progress made in silicon technology has led to significant improvements in chip performance, and these advances have also led to the rapid growth of broken technologies. The success of the semiconductor industry is unparalleled because it has accelerated technological progress and thus impacted the world economy. In the 1990s, the semiconductor industry led high technology to become the main source of economic and job creation. Due to the ability to double the number of transistors on a wafer and increase the efficiency of the wafer without increasing costs, this trend continues. Thirty years ago, there were only a few thousand transistors on each wafer, but now they have reached 100 million. Making the chip size smaller means that it is faster and allows higher integration. To maintain this trend, stricter requirements have been imposed on the quality of raw materials and the cleanliness of component manufacturing processes. The surface metal concentration should be less than about. Lxio10 atoms / cm², otherwise yield loss may occur in the manufactured components (Guidelines for Semiconductor Materials Country Guide ff. Read ATECH, 3HH Industrial Terrace as 7 1 () 6 , As fine as τχ π%). Therefore, future me design will require a more effective cleaning process to remove metal ions and particles. Because these processes can modify the surface of Shi Xi and affect the IC performance

86768.DOC 200408806, this research area is receiving more and more attention. During wafer preparation for 1C manufacturing, surface metal contamination may be introduced onto the wafer surface. These surface metals diffuse into the wafer surface during the component processing. However, since metals can precipitate during the 1C process cooling process, it is common for the mixing to occur. These metals can then decorate defects and exist as precipitates. This can lead to loss of oxide quality, increased joint leakage and subsequent degradation of 1C performance. Although a wide range of methods have been developed to remove surface metals by cleaning with high-purity chemicals, the control of seal surface cleaning and wet etching processes is critical for any semiconductor processing step . The manufacture of prior art components involves approximately 50 critical cleaning steps. There is a great need for an accurate assessment of the effectiveness of such cleaning processes. Novel cleaning technology for large wafers uses a single wafer cleaning chamber. This feature allows faster cleaning times than conventional batch techniques, thereby reducing time and costs. However, when cleaning this large area wafer, it is also important to understand the efficiency of cleaning across the entire wafer surface. The most common technique used to analyze metal pollution is total reflection X-ray fluorescence analysis (TXRF). The detection limit of TXRF varies with different metals, and most of them are on the order of 101G atoms / cm². Only a small part of the wafer was analyzed (1 cm²), and the technique took about an hour to make a measurement. It can accurately measure the spatial variation of metal across a wafer surface. However, this process may be slower. Another more time-consuming method for monitoring surface metals is the body life method. After the surface is cleaned, the wafer must be heat treated to remove any residual

86768.DOC 200408806 Surface metal drives into the body. These wafers can then be analyzed for surface impurities. This requires the assumption that no additional pollution is introduced during the drive-off step. Because of the above reasons, the previous technology is usually limited to the observation of representative batch sampling 'and / or through-sample wafer area rather than the entire wafer, because the time scale cannot realize the full area test of each unit in a given batch. [Summary of the Invention] One object of the present invention is to provide a cleaning process for a half-thickness (cuts) in a component manufacturing process by detecting surface layer metal contamination. And the device reduces some or all of the above disadvantages. The specific purpose of the present invention is to provide a method and a device for the gasification of components & for the effectiveness of the cleaning process of the semiconductor (cutting) For quick evaluation. This system provides a method and equipment for effectiveness evaluation :: The device provides a higher output rate, and especially improves the quality control metric of the cleaning process. In this metric, The test of all the samples and / or the spatial analysis of the I: area in each key cleaning stage may now be performed. 7 w Therefore, according to the first aspect of the present invention-the No. 4 piano for semiconductor (Zhu Yao) ^ kind In the component manufacturing process, the following steps: ~ The method of quality control that includes self-improvement includes the semi-conductor / semi-complaint "Semiconductor structure undergoes critical cleaning treatment steps; the surface of the + conductor structure is exposed from At least a source (preferably a laser ray Udoh) - A high-intensity light beam at, and collected

86768.DOC 200408806 Photoluminescence (pL) generated by exciting the semiconductor structure with the light beam; analysis of the collected photoluminescence signal, and use this analysis as the basis for quality classification of the cleanliness of the semiconductor, In particular, this analysis step is performed by the following steps: determining the average photoluminescence intensity; comparing the average < a range of acceptable specifications with one predetermined photoluminescence; based on the comparison, the semiconductor The cleanliness of the structure is graded for quality, and certain structures are specifically excluded or selected to take corrective measures, such as further cleaning and demonstrating-semiconductor structures that exhibit photoluminescence that are outside of the acceptable specifications for this unit. ~ The related literature indicates that photoluminescence technology can be used as a method for processing impurities introduced into the semiconductor structure during the device manufacturing process. However, according to the present invention, it is surprisingly found that the photoluminescence technology will also be followed immediately: after cleaning and before further heat treatment, metal impurities in the semiconductor structure are detected in a clean state. Therefore, the technology was found to be surprisingly effective in terms of the efficiency of the ㈣-type chemistry in the characterization process and the efficiency of achieving quality control decisions in the key steps. The quality grading steps include: performing a number analysis of the collected photoluminescence signals, analyzing the results of the numerical value analysis with a predetermined acceptable photoluminescence, and good quality. A predetermined range of electroluminescence and a tritium comparison; and based on the comparison, a quality classification of the semiconductor structure is made. In a simple alternative method, the method includes the following steps: Sentence Photoluminescence Degree; Predetermined Acceptance of the Average and Radiation

86768.DOC -10- 200408806 compare the range of specifications; and make the comparison based on the quality classification of the semiconductor structure as described above. The average value can be the entire area average based on the average photoluminescence emitted by the entire structure area, or it can be the local area average, where the structure area is divided into sub-areas of a two-dimensional array, and The average photoluminescence intensity is compared with the average value of each sub-region with a predetermined acceptable photoluminescence specification, and the quality classification as described above is based on the comparison. This approach is advantageous because, in the entire regional average, a response to a single defect can be buried, even if the defect is severe enough to identify a defective product. With an appropriate sub-region size, it is ensured that the response can still be detected. The use of a predetermined photoluminescence specification based on the average value is only an example. In this alternative method, especially when using the sub-area method, other numerical parameters, such as standard deviation, local maximum and / or minimum, deviation from a pre-filled baseline, or other numerical analysis methods may be applied to the photoinduced The analysis of the luminescence signal is to determine the deviation between the photoluminescence response and predetermined parameters on the basis of a partial or the entire area. We know that these predetermined parameters are about semiconductor structures with good quality. In the following, reference is made to numerical analysis based on average luminescence, which should be understood as an example only, and the precise numerical parameters selected to compare the observed response with a predetermined acceptable response are not critical to the present invention. Photoluminescence technology can generate much faster response than TXRF technology in the prior art. In its preferred basic form described above, it takes samples throughout the entire wafer area and produces an average result based on the entire wafer, and because

86768.DOC -11-200408806 Che Xun "TMX, which is only concentrated on a specific, arbitrary sample area, is much more reliable. The speed and accuracy of this technology make it a more effective and practical quality control method than previous technology balances. Station ^ 疋 ^ "'can test all semiconductors after each of the key processing steps, and the characteristics of 带来 have brought about improvements in wet chemical cleaning and improved many of the rich industrial processes. Therefore, in another preferred aspect, the present invention includes a quality control metric for wet chemical cleaning of an attractive semiconductor structure :: j in a plutonium manufacturing process performed on a bulk structure, which is incorporated into the wet type Chemical cleaning: Structure: II :: The structure undergoes the test according to the former < first aspect " in the Demegger's complaint, and will show a photoluminescence response within the predetermined acceptable range of the photoluminescence The structure of the photoluminescence is passed to the next stage, and the structure of the photoluminescence response displayed outside the predetermined acceptable specifications of the photoluminescence is excluded from the next stage of manufacturing. Corrective treatment by further cleaning 'Subsequent inspection and acceptance / exclusion operations are performed as described above. According to a preferred aspect, all structures can be tested after each critical cleaning phase. Prior to the expensive manufacturing process, potential problems were identified earlier. Through hanging out, corrective measures can be taken and samples can be cleaned more thoroughly. Since the method of the present invention can accurately diagnose and correct insufficient cleaning operations at each stage of the process in a fast and convenient manner, the number of defective products unavoidable at the end of manufacturing should be greatly reduced. At a resolution determined by the characteristics of the high-intensity light beam, a 'photoluminescence' technique produces a spatially resolved PL map. Another preferred feature of this method

86768.DOC -12- 200408806 takes advantage of this feature, but for the basic purpose of the invention as a simple and fast quality test of a whole wafer, 'average PL intensity results over the entire wafer area: . This average pL intensity result can be linked to a research institute developed using a slower analytical method (such as TXRF or surface method)-a predetermined range of acceptable specifications. I was surprised to find that, as detailed below, a close correlation can be shown between the use of the near surface-based PL technology of the present invention and the contamination information obtained from the conventionally used prior art method. The beam is controlled in this way, and in particular the beam power and / or wavelength and / or spot size is controlled so as to identify defects and contamination in the semiconductor structure at a selected depth, from a suitable near surface depth (such as from the semiconductor structure). Above 12 microns) to collect PL information. For some materials and components, a smaller depth ' The present invention is a defect and pollution monitoring tool that can be used to monitor surface dirt and other surface structural defects, such as lamination defects and edge slipping. Because this technology measures the surface area, it will accurately detect near surface defects and defects ^ 3, which are very decisive in terms of their impact on component quality and performance. This further enhances the accuracy and reliability of the technology. According to a preferred basic embodiment of the present invention, the average photoluminescence < pre-accepted acceptable specification range is first determined and then used as a benchmark for the results of any given clean wafer for quality control purposes. The predetermined range & will include a minimum and / or maximum value. In particular, we know that depending on the specific chemical contained in the impurity, the photoluminescence signal can be affected in different ways. Therefore, the range of specifications will preferably include a minimum and a

86768.DOC -13- 200408806 Extreme photoluminescence value. Depending on whether the result of the measurement is within the predetermined specification range, a quality control decision is made. If it is within the range, the structure is accepted for component manufacturing, and if it is outside the range, it is excluded. Discarded products can be discarded or subjected to corrective actions such as additional cleaning. The predetermined acceptable range will vary depending on the specific materials and processes involved, and by linking the PL response generated according to the present invention to the < response generated according to existing measurement techniques in the prior art, initially The predetermined acceptable PL range can be determined from the existing quality specification range. Once the JL specifications are established, the present invention can provide a very high output compared to the prior art. For example, for a _12-inch (gang millimeter) wafer, the result can be obtained in about five minutes, while the existing method takes about one hour. Photoluminescence (PL) spectroscopy is a very sensitive technique for observing internal and external electron transitions on impurities and defects in semiconductors. Electron hole pairs are generated when silicon is excited at low temperatures with laser radiation above the band gap of the material. These vectors can be recombined in a variety of ways, and some recombination methods result in luminescence. The electrons and holes pairs formed at a low temperature can be trapped on the impurities' and the electron and hole pairs can emit photons with the characteristics of the interaction, thereby giving impurity specific information in the photoluminescence spectrum. There are a large number of PL spectroscopy techniques applied to Shi Xi, which include Shi Xi's characterization and component manufacturing (such as implantation, oxidation, plasma etching, detection of point defect complexes and the existence of misalignment) after different processing steps. ) Characteristics. One of the most important applications includes non-destructive measurement of shallow donors and recipients such as kun, rot and phosphorus.

86768.DOC -14- 200408806 In particular, this technology enables the measurement of the concentration of these shallow donors and receptors. However, in the conventional application, in order to obtain the spectral information of the optical center and clear chemical identification, the measurement needs to be performed at the temperature of liquid helium. Throughout the semiconductor industry, we know that at room temperature, the PL signal is significantly weakened and very little useful spectral information can be obtained. Therefore, room temperature technology is preferred, especially a non-destructive technology such as described in International Patent Application W098 / 11425, which enables the detection of electroactive defects in PL semiconductor structures based on room temperature. The patent application Mujie 7F has a pL technology with industrial applications. In its application, the technology can generate images within minutes, and the technology has another additional advantage I7 /, which can generate especially near wafers. Fortunately, the surface (on which the component is fabricated) is a microscopic image of a single defect. This technology uses information at a rate i suitable for industrial applications, and information on defects in semiconductors or broken structures, and in particular enables us to visually inspect defects in the upper region of the semiconductor cut structure, especially near the surface of the structure. This technology utilizes the enhanced non-trivial recombination of electron-hole pairs at the defects in a semiconductor or silicon structure to enhance the contrast in the semiconductor-structure-PL image to enhance the observation of defects in the image. The preferred PL technology in W098 / 11425 is hereby incorporated by reference. The invention complies with the fact that contamination can be detected at the surface even in samples in a clean state without further heat treatment. The success of the room temperature PL method disclosed above can be attributed in part to the smaller, better spatial resolution of 0.02 micron, ideally micron size, and a range of 10 to 1G9 watts per square centimeter. Peak or average power density

86768.DOC -15- 200408806 detection capacity, so local defects have a greater impact on the measured PL intensity, and I also believe that the success of this method is partly due to the injection of The carrier density is higher. This greatly increases the possibility of non-radiative recombination at the defect and therefore enhances the physical location of the defect. In some preferred embodiments described in more detail below, the present invention utilizes this method by preparing a spatial map of one of the defects of a representative pL response, and more preferably 2 images. The description of high-intensity lasers here includes (but is not limited to) high-power-density lasers, which means that in this high-power-density laser, the radiation is focused regardless of the laser power. In a preferred method of the present invention, a pulsed laser excitation source is used, and ideally, the emission data is measured and / or the emission image is collected as a function of time. The idea is that both depth and spatial resolution are improved and can be used to obtain information on the defect < carrier capture section. You can also use time-resolved measurements to measure the effective carrier life and obtain a life map. The PL technology of the present invention generates a spatially resolved pL map over the entire wafer area. In the main method of the present invention, the data map is subsequently processed to provide an average PL level across the wafer, and the average pL level is compared with a reference standard to make a quality control decision. If this method is used for simple acceptance / rejection and recleaning quality control judgments as a test before transferring the structure to the next manufacturing stage, only the average pL level need to be considered, and the mapping generated by this method Resolution is not important. A resolution of about 7 mm is sufficient. At child-level resolution, processing time is reduced and test output rates are maximized. For example, obtained from a 12-inch (3 mm) wafer

86768.DOC -16- 200408806 It takes only five minutes to get a satisfactory acceptance / exclusion result. However, a particular advantage of the preferred photoluminescence technology of the present invention is that: in addition, this technology can be used to generate a spatially resolved map of the PL signals running through the surface of the semiconductor structure under test, and in particular the space in which such signals can be generated Parse the image. This spatially resolved map is completely unachievable on actual time measures in the prior art. Therefore, in a preferred implementation, the method further includes the step of generating the map and / or the image. In such environments, it may be appropriate to work with a mapping / imaging resolution of 0.5 mm or less. 6 Conveniently, the method further comprises storing the spatially resolved pL map on an appropriate data storage component, and / or transmitting the digitized data obtained from the spatially resolved map for forward processing by an appropriate processing component, and / or The spatially resolved image is displayed on an appropriate display member. The basic technology identifies an average PL intensity, and the acceptance / rejection or recleaning decision is based on the average PL intensity. Spatial analysis information is particularly useful when it comes to excluded structures. A better and more advanced quality control strategy may therefore use this faster, more basic technology to process each unit and only generate spatially resolved data for excluded < structures. In one embodiment, in order to fully utilize the technology's ability to generate spatial analysis data on the PL response of the semiconductor structure under test, (as indicated) it may be appropriate to operate at a higher resolution. As a result, the output will be slower than just using this technology as the basis for accepting / excluding the basis of quality control decisions. A better and more advanced quality control strategy may therefore use the faster, more basic technology to process each unit, and use the additional functionality provided by collecting spatially parsed data for a higher resolution 86768.DOC -17- 200408806 analysis Reprocess the excluded structures. The ability to generate maps or images allows roughly identifying the location of the defect. This feature is used in one of the preferred embodiments of the present invention. The method described above is used to quickly screen semiconductor structures (a 300 mm wafer takes only 5 minutes) and identify samples for complete chemical analysis. Embodiments of the method include subjecting a semiconductor structure to the tests described above to generate a spatially resolved map of the PL signal across the surface of the semiconductor structure, and at least in the case of excluded structures' using a spatially resolved map to identify The approximate location of the contamination, and a specific analysis technique such as TXRF is used to further analyze the semiconductor structure at the identified location to identify impurities. In addition, the ability to generate spatially resolved structures can help diagnose and correct clean interrogations. After a given critical cleaning phase, multiple samples are tested consecutively, the test results are stored and compared as described above, and therefore the system inefficiencies during a particular cleaning process can be identified. This method is applicable to any basic semiconductor structure on which a component can be manufactured by heat treatment in a familiar manner. In particular, this method is applicable to the structure of silicon and silicon alloy-based wafers. These components can be made from a simple single-layer wafer or from a multi-layer wafer, such as in an epitaxial layer on a base silicon wafer. According to another aspect of the present invention, a device for performing quality control cleaning on a semiconductor structure during a component manufacturing process on the semiconductor structure (such as silicon) includes a cleaning device that subjects the semiconductor structure to wet chemistry. Cleaning, and a test device that detects foot impurities on the surface of the structure in the clean state, wherein the test device includes a high-intensity light source, the light source is preferably a transport and especially a radon intensity laser; High intensity light

86768.DOC -18- 2UU4U55U0 Beam-focusing soil is a semi-conductor component of the semiconductor structure under test. It collects semiconductors, components, and semiconductor structures by beams. ^% And numerical analysis collected ::: 致 发: Γ 分 一 一定 Acceptable rules ... Comparator, which compares the analysis result with the pre-determined value and is based on the photoluminescence signal. Within the P target garden, it was decided to accept or exclude the knot = τ born into the line-part, and including the transmission member that can transfer the structure to 1 = ;;, paragraph, according to the photoluminescence signal) Determine whether the transmission is within the range of acceptable specifications. ^ "The reading and writing component can transmit the photoluminescence response structure of the photoluminescence predetermined in the photoluminescence to the next component manufacturing stage 'and recorded in the photoluminescence The photoluminescence response structure outside the acceptable specification range is transmitted back to the cleaner for further cleaning and flocculation. The device that executes the basic method generates a Bayesian based on a numerical analysis of the signal response. For example, as described, the average PL signal or a collection of local data across the entire area. Compare this information with predetermined acceptable specifications. ^ 疋 In order to implement the improved alternative method of the above method, the device preferably needs to further include a component that parses the collected pL data into a spatially-resolved PL mapping that penetrates the semiconductor structure region, and additionally includes the parsed A component that converts data into a PL image, and / or stores the mapping / image, and in particular an image / data storage component that continuously maps / images for future comparison, and / or transmits the mapping / image to an appropriate remote Information Office

86768.DOC -19- A component of a controller, a male / female, or an image display component that displays an image and / or related information to a user's foot, such as a visual display screen. In another aspect of the present invention, a computer program is provided and / or a computer is appropriately programmed to perform some or all of the steps of the methods previously described herein, and in particular, to perform data processing steps on the collected PL data, For example, the average PL of the entire wafer area is measured, and / or the PL data is spatially resolved from the collected PL data-the PL mapping 'and / or the average value are compared with a predetermined acceptable specification range. [Embodiment] The device shown in FIG. 1 mainly includes a PL imaging microscope, which includes: a right-hand side, including a set of lasers (3_8); a bottom, including a sample stage, such as XY room or R hall; left-handed Surface, including a microprocessor (40) and a display π screen (39), the center of the JL chart shows various optical elements for detecting light passing through the system. In the embodiment shown in Figure 1, six laser beams are provided to detect different depths in the sample. However, it is within the scope of the invention to use only one laser, or to use a greater number of lasers. In any case, at least one of the lasers is a high-intensity laser, and ideally has a spot size between 0.01 mm and 0.05 μm, and between 104 and 109 watts / cm 2. Power density. A laser selector (丨 6) coupled to the set of lasers is provided to select one or more lasers for use, and the frequency and wavelength of these lasers are also selected. Use conventional optical devices, such as optical fibers (9) to direct light towards the collimator (10) and laser beam expander (11). Place all toe plates (12) between the laser beam expander 86768.DOC -20- 200408806 and the beam splitter (31). The beam splitter (31) directs a part of the light from the aforementioned laser toward the sample (2) by the objective lens (34). An auto focus controller (30) is provided and is coupled to a piezoelectrically driven focus stage (33). The microscope is equipped with a conventional turntable (36), which has at least one high numerical aperture objective lens for microscopic inspection and a low numerical aperture objective lens (34, 35) for macroscopic inspection. In addition, an optical displacement measurement system (38) coupled to the turntable (36) is also provided. Cable laying is provided to connect the autofocus controller (30) to the microprocessor (40), and also to connect a microscope objective indexing device (32) to the microprocessor (40) 〇—to the beam splitter (31) A filter wheel (13) with a laser notch filter downstream, and a folding mirror (14) swinging sideways downstream of the filter wheel (13), its function will be described below. A filter wheel (27) for wavelength selection is arranged in a row with the mirror (14), and a zoom lens connected to an appropriate CCD 2-D array detector (29) is arranged after the filter wheel (27). . In the front optical path of the cold mirror (17), there is an infinite system compensation lens (37), which reflects light to another filter wheel (23) and a focusing lens (24) for wavelength selection. The focusing lens is at The front of a UV and visible light detector (25). A detector (25) is coupled to the lock-in amplifier (26). It is used to obtain a reflection image of one of these surfaces. On the rear surface of the cold mirror (17), there is another filter wheel (18) for wavelength selection, and at the end of the filter wheel (18), there is a focusing lens (22) and another for pinhole selection. Aperture wheel (19), the aperture wheel (19) is at the forefront of the debt detector (21) for detecting light emission. 86768.DOC -21-200408806 Both UV and visible area detectors (25) and infrared detectors (21) are coupled to the lock-in amplifier (26). The operation of the system is explained below. A range of wavelengths is provided by several lasers (3-8) to detect different planes in the sample. The laser can be modulated by the frequency generator (16) so that the signals emitted by the detector and the detector can be separated from the background and the sample and emitted by the sample. The lock-in amplifier (26) can be used to make the detectors and Laser modulation frequency is synchronized. In another real case, this range of wavelengths can be generated by using a tunable laser and / or an optical parametric oscillator. Each beam of lasers is connected to a multi-branch optical fiber (9) and arranged in a row with ash so that any or all of these lasers can be made ..., shot w (2). The common ends of the multi-branched optical fibers terminate in an optical system of collimated: emerging light. The optical system is aligned with—beam expansion), the beam expander ⑴) matches the diameter of the laser beam to the required diameter of the microscope objective lens (34, 35) on the sample ⑺. The expanded beam then passes through a toe plate (12) which distributes optical energy evenly across the beam region. The extended and apodized beams are reflected by a beam splitter (31), and the beams are passed to the head micromirror objectives (34 and 35). Through a microscope objective (34 or 35), the Θ beam is focused on the sample. In the micro mode, the objective is selected to focus the Θ beam to a diffraction-limited spot size. The turntable (36), which is operated by an indexing mechanism, allows the objective lens to be transformed into a viewing mode &, which can be shot over a wider area of the sample. In another embodiment, the toe plate (12) can be removed, so that the spot size of the micro mode can be reduced to allow a higher injection volume. An optical displacement sensor (38) measures the distance of the sample, and

86768.DOC -22- 200408806 Pass the feedback loop of the anti-focus controller (30) to maintain the correct spacing generated by the piezoelectrically actuated focus hall (33). The photoluminescence signal from the sample is collected by the microscope objective (34) (which is in the micro mode), and transmitted back through the notch filter in the beam splitter (31) and the filter wheel (13). The filter wheel (13) Contains a notched calender that matches the laser wavelength range. The notch filter removes any reflected laser light ’and only passes the photoluminescence signal. Turn the folding mirror (14) out of the beam to allow the signal to pass to the barrel lens (37) and to the cold mirror (17). The barrel lens (37) can be incorporated to compensate for any broadcast Limited microscope objectives. The cold mirror (17) reflects these wavelengths below the selected cut-off point (about 700 nm) to a focusing lens (24) that focuses the signal into the detector (25). A filter wheel (23) in front of the detector focusing lens (24) contains a filter to separate the selected wavelength band. t-square; the photoluminescence signal in the wavelength range above the cut-off point partially passes through (17) and is similarly focused into the detector (21) by the lens (22). The signal also passes through a filter wheel (18) containing a filter to separate the selected wavelength band. A aperture wheel (19) placed in front of the detector (21) contains a series of different diameters. Pinhole. The aperture wheel can be moved axially by a piezoelectric actuator (20) so that the pinholes can be placed so that they are confocal with the desired image plane. In this way, planes at different depths in the sample (2) can be imaged to provide accurate depth information. The electric signal from the detector (21, 25) is supplied to the lock-in amplifier (26). In the lock-in amplifier (26), a reference signal is received from one of the frequency generators (15).

86768.DOC -23- 200408806 synchronize the signal with the modulation frequency of laser (3-8). This signal is then supplied to a central processing unit (40) for analysis. The raster scans the sample stage to obtain PL images. Alternatively, optical scanning using a galvanometer mirror can also be used. In another operating micro mode, the folding mirror (14) is turned into a photoluminescence beam. The steering signal passes through a filter wheel (2) and enters a zoom lens (28), where the filter wheel (27) includes a filter to separate the selected wavelength band. The zoom lens allows the sample (2 The illuminated spot on) is imaged on the CCD two-dimensional array (29). It allows the irradiated area of the sample (2) to be imaged at different resolutions. The electrical signal from the CCD array is supplied to the central processing unit ( 40) for analysis. The data processing is illustrated diagrammatically in Figure 2. The sample (101) that has undergone a standard wet chemical cleaning stage in a single wafer processing chamber (100) by a manipulator arm (102) The signal is transmitted to a sampling base (103) to be tested by the device in Figure 丨 to generate a PL signal. The signal is collected by the collection device in Figure 丨 (shown as device 105 in the form of a schematic diagram). The figure further displays the data In a first processing path, according to the main point of the present invention, the PL mapping data is transmitted to a processor (107), which processes the data to determine the average intensity across the entire area of the sample (101). The average value obtained Passed to a comparator (108), which associates the average PL intensity data with a predetermined storage specification range within the data memory (109), and transmits a quality control decision to the control unit based on the comparison ( 11〇) 上。 In this embodiment, the control unit (110) operates directly on the control arm (102) 'the control arm then transfers the sample (101) to a component manufacturing processing line or 86768.DOC- 24- 200408806 2 Exclude the line to take appropriate green positive measures. In place of the 'control unit' during operation, it can, for example, give a directive to an operator, who can then separate the control component by (Eg, to manipulate the arm 002) to make an acceptable / exclude selection to transfer the sample under test to a correction process, etc. The / line shows a second processing path, which reflects an alternative second aspect of the invention. In this optional aspect, the data corresponding to the PL intensity mapping across the surface of the sample (101) is also transmitted to a second processing unit (m) 'processing unit (ill), which can be parsed into a Digitized spatially resolved intensity mapping across the sample (ιοί) surface. The obtained mapping is transmitted to the data memory (112) and the visual display screen (113). Access to parsed data

Xinbei defect location. In this way, wafers can be quickly screened using this basic device. The location of the contamination can be identified from the wafer map, and the wafer can then be further analyzed using txrF technology to identify impurities. Figure 3 shows the contamination that can be applied to this technique after debt cleaning. The selected control wafers (2 inches in diameter) were first measured on the device in Figure 1, and then these wafers were intentionally contaminated with Cu, A1, and CoOl X 1012 atoms / cm2. These wafers are then measured. The measurement results are shown in Figure 3. After contamination, a significant difference in the average PL signal was obtained. In all cases, the average PL value increased, indicating a change in the surface defect density of the amount of surface contamination. Although these results are surprising, the experiment was performed at room temperature (to reproduce normal cleaning conditions). Although source materials such as WO97 / 09649 describe the use of such techniques to identify contamination after thermal treatment, such previous source materials indicate that in order to activate metal contamination 86768.DOC -25-200408806, wafers must be heat treated. However, it is determined according to the present invention that the average PL value can be used as a metric to determine the presence of surface metal contamination to monitor surface cleanliness. Further experiments were performed to observe the effect of Fe contamination (i.e., X 2 (atom / cm 2)). Figure 4a shows the wafer map before contamination (average PL value = 3.97), and Figure 4b shows the corresponding map after Fe contamination (average value)? [Value = 11.19), the average PL value increases significantly. This confirms that surface Fe contamination can also be detected using pL values. To calculate the effect of cleaning, a commercial single wafer chamber was used to clean Fe-contaminated wafers. Figure 4c shows the mapping of the cleaned wafers (average pL value = 3.50). It is worth noting that with the removal through the cleaning process, the average PL signal dropped to a level comparable to that before cleaning, thus confirming this point: the average PL level is an indication of surface cleanliness. In addition, after rigorous inspection of the wafer mapping, it was found that there was a discernable difference in the average PL level across the wafer. It was found that the level of the wafer edge (Figure 4d «4.0) is higher than the level of the wafer center (Figure 4 (^ 31). This clearly indicates the difference in the cleaning performance across the wafer surface. TXRF also showed that The level of surface contamination on the wafer is lower than i X 1010 atoms / cm 2. This clearly shows that the average PL value is a very sensitive measurement method of surface cleanliness. The PL is recorded and the average pl is calculated. Value, and then compare this value to a predetermined value. The specific range will be used to determine the quality in the cleaning state, and thus a pass-fail process control program will be generated, in which it is known that the sample is passed by the ideal clean Proceed to the next manufacturing stage and hand over the contaminated samples for further cleaning. Figure 5 shows this procedure in the form of a chart. 86768.DOC -26- 200408806 By setting the specific range, the range can then be made with any color or and The grayscale patterns are associated and a corresponding image is prepared. Figure 6 shows an example. In this example, the black areas in the wafer map are shown in areas outside the specific range. Use this simple code to show changes in wafer quality. After observing wafer quality using the average PL signal level, you can record further measurements on wafers that do not meet this specification with higher resolution. This will allow us Determining the type of pollution or other defects that exist, while allowing us to detect spatial and temporal changes in more detail, can give a clearer indication of the source of the pollution (such as a specific processing step or incomplete cleaning). To identify positives The source of contamination can be recorded on the back of the wafer at the same time. By comparing these wafer mappings, it can be clearly known whether the contamination originated from the back of the wafer. Then further measures can be taken to prevent cross-contamination of other metrology tool process instruments.% Therefore According to the present invention, a photoluminescence tool can be used as a rapid process suppression tool to determine wafer quality. Wafer mapping is obtained and measured by numerical analysis (PL response to provide a quality metric, in which photoluminescence responds The discrepancy of predetermined parameters related to a semiconductor structure with good quality is used as a basis for quality control judgment. In the basic example given, it can only be a comparison of the average response over the entire area. But—in general, contamination can be very local, with average PL signals (all measured in the wafer map) What is sought on a pixel may not necessarily be a reliable indication of the contamination. If a paste ... circle map is used, then average PL analysis can be used to detect very

86768.DOC -27 · 200408806 sex signals. After recording a wafer map, a virtual grid can be applied and displayed on top of the wafer map using appropriate software. You can use this virtual grid to perform analysis. At the same time, the local area analysis also allows for a better position for problem locations in the structure. For example, i can indicate a failed grid element; a micro-scan can be initiated at the same location as the grid element to allow more detailed inspection of the area of interest; and the failure in the wafer mapping analysis can be output in an appropriate form Record of components. The local grid method is not limited to numerical analysis based on average values. Any suitable pre-defined parameters (including average intensity, small value, maximum value, standard deviation, and baseline) can be used to determine the contaminated area. Since the signal variation across the wafer is not uniform, the PL signal baseline method can be a more useful parameter. This technique will be explained below. Figure 7a shows a typical wafer map, and Figure 7b shows the associated histogram of pL intensity. Contamination is detected in wafer mapping by deviation from the baseline value, and limits can be set. However, the average PL value in the wafer map can be corrected by contamination. The value used should represent the signal level of an uncontaminated wafer. The < peak value shown in Figure 8 represents the true PL value of an uncontaminated wafer. Modifying the baseline function to have a peak will allow the user to accurately track contamination with higher sensitivity. A suitable algorithm includes the following steps: 1. The peak of the bounding foot is the maximum value in the histogram, and it is used for the baseline. 2 · Search the data of 嶂 r value. 3. Then calculate ± 70% of the maximum value (which is defined by the user), and then re-

86768.DOC -28-200408806 Defines the maximum value of the sharpening center of the point. 4. The exact value of the peak maximum is then calculated and then the pL level is defined. 5 · It also allows the user to enter typical baseline values from uncontaminated wafers. It will be helpful if there are two peaks of equal intensity. The baseline of the wafer is the PL value corresponding to the maximum number of points. Baseline deviation is defined by the following relationship: Baseline deviation = PL value-Baseline PL value is the AVGpL of each element of the grid. Baseline deviation must be measured for each component. [Brief description of the drawings] The present invention is described above by way of example only with reference to the attached drawings 1 to 8, wherein: FIG. 1 is a diagrammatic illustration of a suitable device for obtaining PL data; FIG. 2 is a diagram showing how to process the data Schematic diagram; Figure 3 shows the PL signals of a wafer in a clean state and three intentionally contaminated wafers recorded according to the present invention; a solid display of the clean state, contaminated state, and recleaned state recorded according to the present invention PL signals of Fe contaminated wafers; Figure 5 shows the correlation between the PL materials and a basic quality control acceptance / exclusion determination; Figure 6 shows the spatially resolved pL images of different wafers; Figures 7 and 8 show One possible numerical analysis technique according to the present invention. [Illustration of Symbols in the Drawings] Sample Hall 2 Sample

86768.DOC -29- 200408806 3-8 Laser 9 Optical fiber 10 Collimator 11 Laser beam expander 12 Apodizer plate 13, 18, 23, 27 Filter wheel 14 Folding mirror 15 Frequency generator 16 Laser selector 17 Cold mirror 19 Aperture wheel 20 Piezo actuator 21 Infrared detector 22, 24 Focusing lens 25 UV and visible area detector 26 Lock-in amplifier 28 Zoom lens 29 CCD two-dimensional array 30 Automatic focusing controller 31 Beam splitter 32 Objective lens indexing device 33 Piezo driven focusing stage 34, 35 Microscope objective 36 Turntable -30-

86768.DOC 200408806 37 Barrel lens 38 Optical displacement measurement system 39 Display TF screen 40 Microprocessor 100 Chemical gas deposition epitaxial reactor 101 Sample 102 Manipulator 103 Sampling base 105 Collection device 107 Processor 108 Comparator 109 Information Memory 110 Control unit 111 Second processing unit 112 Data memory 113 Visual display 86768.DOC -31-

Claims (1)

200408806 Patent application park: 1. A method for quality control of the cleaning process of semiconductors such as silicon during component manufacturing, which includes the following steps: subjecting a semiconductor structure in a clean state to a critical cleaning process Step; exposing the surface of the semiconductor structure to at least one south-intensity light beam from a suitable light source, and collecting photoluminescence (PL) generated by exciting the semiconductor structure with the light beam; The light-emitting signal is analyzed, and β is used as the basis for one quality classification of the semiconductor cleanliness. 2. The method according to item 1 of the patent application range, wherein the quality grading step comprises: determining an average photoluminescence intensity emitted through the structural region or a subregion thereof; and prescribing the average value with one of the photoluminescence The acceptable specification range is compared; based on the comparison, a quality classification is made for the cleanliness of the semiconductor structure. 3. The method according to item 1 or 2 of the patent application scope, wherein the quality grading step includes-excluding or selecting a photoluminescence response semiconductor structure displayed outside a predetermined range of acceptable specifications to take corrective measures. 4. If the method of claim 1 or 2 is applied, the method is applied as a quality control measure for one of wet chemical cleaning of semiconductor structures such as silicon introduced into the semiconductor manufacturing process, and Incorporated as part of a component manufacturing process, the method includes the following steps: • Chemically cleans the structure in a conventional manner so that the knots in the clean state are subject to the same scope as patent applications The test of item 3 passes a structure that displays a photoluminescence response within the predetermined acceptable specification range of the photoluminescence to the lower-component manufacturing stage, and-displays-in the photoluminescence < pre- Photoluminescence response structures outside the range of tritium's acceptable specifications are excluded from this next component manufacturing stage. 5. The method according to the scope of the patent application, wherein a plurality of excluded structures are then passed for further cleaning, and then retesting is performed and the acceptance / exclusion steps and steps as described in the scope of the patent application are repeated. 6. The method according to item 1 or 2 of the scope of patent application, wherein the beam power and / or spot size of the light beam is controlled to collect near-surface PL information from the upper part of the semiconductor structure in micrometers. 7. The method of claim 6, wherein the beam is controlled to collect near-surface pL information from a point 1 micrometer above the semiconductor structure. 8. The method of applying for item No. 丨 or 2 of the patent application, wherein the PL response is obtained at about room temperature. 9. The method of claim 1 or 2, wherein the light source is a high-intensity laser. 10. The method according to item 9 of the patent application range, wherein the laser has a spot size of 0.1 to 20 microns, preferably 2 to 5 microns, and a peak or average power mountain range of 10 to 109 watts / Cm2 smaller detection capacity. 11. The method according to the scope of the patent application, wherein a pulsed laser excitation source is used, and the luminescence data is measured and / or the luminescence images are collected as a function. ^ 12. If the method of item No. 2 or 2 is requested, it additionally includes using the PL signals collected in 86768.DOC -2-200408806 to generate a spatial analysis of one of the PL signals across the semiconductor surface under test. A step of mapping and, in particular, generating a spatially resolved image of one of these signals. 13. If the method of claim 12 is applied, it further comprises storing the spatially resolved PL map on an appropriate data storage component, and / or transmitting the digitized data obtained from the spatially resolved map by an appropriate processing component for use in Steps for forward processing. 14. If the method of claim 13 is applied, it further includes the step of displaying any generated PL image on an appropriate display member. 15. The method of claim 13 or 14, further comprising the steps of: identifying the approximate location of the pollution using the spatial resolution map; and further analyzing the identified location using a specific analysis technique such as TXRF The semiconductor structure to identify the contamination. 16. · A method of identifying and / or characterizing cleaning contamination in a continuous clean state of a semiconductor structure such as silicon, which includes each of the continuous structures after a given critical cleaning phase The method of one of the items in the scope of the patent application is implemented on the above, and only on those structures that are excluded due to the photon response π that is outside the predetermined acceptable specification range. Ask for these additional steps in one of the 12th to 15th of the patent scope to produce only the excluded structures: spatially resolved data. η • -A method for assisting diagnosis and cleaning problems, which includes sequentially testing a plurality of structures after a given critical cleaning phase in accordance with any of the methods described in the aforementioned patent application scope, storing and comparing these sequential tests The result is 86768.DOC 18.200408806, and the comparison results are used to identify system errors during the cleaning process. An apparatus for quality control cleaning of a semiconductor structure such as Shi Xi during a component manufacturing process performed on a semiconductor structure, comprising: a cleaning device that subjects the semiconductor structure to wet chemical cleaning, and a test device, It detects impurities on the surface of the structure in the clean state, wherein the test device includes a high-intensity source; focusing a high-intensity beam from the light source onto a member on a surface of the semiconductor structure under test, a collection member 'It collects the photoluminescence data on the surface of the semiconductor structure in the Bayesian test generated by exciting the semiconductor structure with the light beam. The analysis component processes and numerically analyzes the collected data. The analysis results are compared with predetermined acceptable specifications, and the structure is accepted or excluded based on whether the photoluminescence signal is within the predetermined acceptable specifications. 19. If the device under the scope of patent application 18 is incorporated into a production line, and includes a transmission member suitable for transmitting the structure to a determined next processing stage, wherein according to whether the photoluminescent signal is in the Predetermined accessibility = within the range of specifications to decide whether to transmit. 20. 21. = The device of the scope of application for patent No. 19, wherein the transmission member is adapted to transmit the photon-response structure within the predetermined acceptable specification range of Wei photoluminescence to the next element At the manufacturing stage, the photoluminescence response structure outside the predetermined acceptable specification range of photoluminescence is transmitted back to the cleaner for further cleaning. For example, the device of claim 18, 19, or 2G of patent application range 11, wherein the beam power and / or wavelength and / or spot size of the beam is controlled so that the device = 86768.DOC -4- 200408806 is suitable for the semiconductor structure Near-surface pL information was generated and collected at 12 micrometers above. For example, the device of claim 21 is patented, wherein the beam power and / or wavelength and / or spot size of the beam is controlled so that the device is suitable for generating and collecting near surface PL information from 1 micron above the semiconductor structure. . 23. The device according to claim 18, 19 or 2 (), wherein the light source is a high-intensity laser. 24. For example, the method of claim No. 23 of the patent, wherein the laser has a spot size of 0.1 to 20 microns, preferably 2 to 5 microns, and the peak or average power poverty is 104 to 109. Smaller detection capacity between W / cm2. 25. The device according to claim 18, 19, or 19, further includes a component that parses the PL data collected by the child into a spatially resolved PL map that runs through the semiconductor structure area. % If the device of the scope of application for patent No. 25, it further includes parsing the collected information into-a space that runs through the area of the semiconductor structure: analyzing the components of the PL mapping, and optionally including converting the parsed data to- PL image "component, and / or the image / data misstore component that stores the mapping / image, especially the continuous mapping / image for future comparison, and / or transmits the image to the remote processing material 'And / or display the image and / or related information to an image display component such as a visual display screen. C 86768.DOC