Method for classifying unknown particles on the surface of a semiconductor wafer
By applying known particles to a substrate and analyzing EDX spectra, the method addresses the challenge of classifying small particles on semiconductor wafers, providing accurate size and composition correlation for precise classification.
Patent Information
- Application Number
- JP2023527412
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2020-11-09
- Filing Date
- 2021-10-26
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2041-10-26
AI Technical Summary
Existing methods struggle to accurately classify particles smaller than 200 nm on semiconductor wafers due to limitations in SEM and EDX resolution, leading to unclear particle structure and origin, especially for SiO2 particles where native oxides interfere with EDX measurements.
Applying particles of known chemical composition and size to a substrate, using energy dispersive x-ray spectroscopy (EDX) to record spectra, analyzing these spectra to establish a mathematical relationship between size and composition, and comparing unknown particles against a best-fit curve to classify their origin.
Enables precise classification of particles by correlating size and chemical composition, distinguishing between known and unknown particles, even in the presence of substrate interference, thereby improving classification accuracy.
Smart Images

Figure 0007584649000001
Abstract
Description
[Technical field]
[0001] The present invention relates to a method for sorting small foreign particles on a semiconductor wafer. [Background technology]
[0002] Monocrystalline semiconductor wafers are the basis of modern electronics. During the manufacture of components on said semiconductor wafers, highly complex thermal manipulation and coating steps are nowadays carried out.
[0003] Semiconductor wafers, in particular silicon semiconductor wafers, are typically produced by first pulling a monocrystalline rod by the float zone method (FZ) or the Czochralski method (CZ). The rod thus produced is divided into crystal pieces by a saw suitable for this purpose, such as a wire saw, an internal diameter saw or a band saw, and these crystal pieces are then processed into semiconductor wafers, typically in a wire saw or an internal diameter saw.
[0004] After further mechanical, chemical-mechanical and / or chemical steps, an epitaxial layer may optionally be applied by CVD.
[0005] These semiconductor wafers thus produced are then made available for further component processing.
[0006] Due to the narrowing of design rules in component processing, particles are the main negative quality feature in wafer manufacturing, with associated developments in processes and cleaning. The shape, location, and chemical composition usually provide an indication of the origin of the contaminating particles and therefore the possibility of reducing their incidence in improved processes.
[0007] Particles constituting unwanted contamination of semiconductor wafers are typically imaged by SEM (Scanning Electron Microscope). The combination of SEM with EDX (Energy Dispersive X-ray Spectroscopy) provides in this context a technique that allows the simultaneous determination of the particle's location, size and shape, as well as its chemical composition.
[0008] This type of analysis is well suited for particles with a range above 200 nm. Particles with smaller extents are very difficult to classify since information about the structure of the particle is either absent or affected by large errors.
[0009] The closer the particle size is to the SEM resolution limit, the more difficult it becomes to classify the particles. For technical reasons, the resolution limit of EDX analysis is lower than that of the SEM resolution.
[0010] However, as a result of the continuing miniaturization in component processing, smaller particles are receiving more and more attention as they cause more and more problems in semiconductor component manufacturing, and therefore it is very important to avoid these particles in the manufacturing process.
[0011] A particle size range of particular interest at present is the range of about 15 nm to 200 nm. Particles of this size can no longer be classified by techniques commonly practiced in the prior art.
[0012] Prior Art / Problem Description Patent specification JP 2006 303 134 A2 describes an analytical technique for visualizing foreign particles on the surface of a semiconductor wafer by applying a weak acid in gas phase to the surface of the semiconductor wafer and detecting the adhering foreign particles with a particle counter.
[0013] The drawback of the method disclosed in this patent specification is that, although it can indeed make small particles "visible", in other words detectable or countable, it is not possible to derive any information regarding the origin of the particles.
[0014] Patent application US 2005 062 959AA describes a method in which small particles to be analysed are used as the nuclei for the polymerization of a monomer, after which the larger particles produced by polymerization are analysed in a particle counter.
[0015] This patent also concerns the detection of particles, but information about the structure of the particles remains hidden with this method. Moreover, this method is an indirect method and only the position of the particles can be measured.
[0016] Patent specification US 5 715 052 A describes a method in which small areas of a small particle are repeatedly analysed using additional light and using different optical conditions.
[0017] In fact, with the aid of this patent specification it is possible to classify particles according to their internal structure, however the method described quickly fails when the particles under investigation are smaller than the resolution of the microscope used. Summary of the Invention [Problem to be solved by the invention]
[0018] In the case of small particles for research purposes, the use of SEM (scanning electron microscope) does not achieve the objective, since it is no longer possible to reveal the internal structure of the particles.
[0019] Therefore, primarily, the only possible classification is that using energy dispersive x-ray spectroscopy (EDX) spectra.
[0020] The problem here is that the analysable EDX measurement signal becomes smaller the smaller the particle under investigation. The ambient conditions (e.g. the influence of increasing substrate surface / volume signal components) have an increasingly large influence on the measurement result, thus making a correct classification more difficult.
[0021] In the case of relatively small Si or SiO2 particles, for example, direct classification by EDX spectra is not possible, since in this case the native oxide of the particle and the surrounding native oxide significantly influence the measurement signal.
[0022] Another significant effect is the high signal component arising from the substrate volume (interaction volume), which suppresses the small fraction arising from the particle volume due to the unfavorable ratio of effective particle volume to substrate volume.
[0023] It is therefore an object of the present invention to provide a technique that does not suffer from the drawbacks of the techniques available in the prior art. [Means for solving the problem]
[0024] This object is achieved by the method described in the claims. The features specified in relation to the above embodiments of the method of the invention may be interchanged in the corresponding product of the invention. Conversely, the features specified in relation to the above embodiments of the product of the invention may be interchanged in the corresponding method of the invention. These and other features of the embodiments of the invention are set out in the description of the drawings and in the claims. The individual features may be realised separately or in combination as embodiments of the invention. Moreover, they may describe advantageous embodiments which can be protected in their own right. [Brief description of the drawings]
[0025] [Figure 1]FIG. 1 shows oxygen measurements obtained from energy dispersive x-ray spectroscopy spectra (EDX, photon count in units normalized to 1) as a function of the measured diameter of spherical particles located on a substrate (in this particular case a silicon semiconductor wafer). The data points indicated by open circles are values originating from intentionally applied SiO2 particles. The shaded areas indicate the confidence regions in which the SiO2 particles should be located with a 95% probability, under the assumption that for these particles the oxygen content shows a linear dependence on the diameter. The black circles relate to particles that do not show a significant dependence between oxygen content and particle size. The regions in which these particles are located in the figure are significantly different from the aforementioned confidence regions. The data points represented by open triangles clearly originate from particles that show a certain dependence between the oxygen content and their diameter, but they are nevertheless outside the confidence region defined above for the SiO2 particles. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0026] Detailed Description and Examples of the Invention In the present invention, particles of known chemical composition and different sizes are applied to the surface of a substrate (test wafer), which preferably comprises silicon.
[0027] The particles are preferably applied to the test wafer by means of a suspension containing the particles. The particles contained in the suspension preferably have a very limited size range between 15 nm and 3000 nm. It should also be ensured that the foreign matter content (undesirable impurities) is minimal.
[0028] The particle suspension may be applied similarly to the production of a calibration wafer for a scattering laser light system. The particle suspension in this case may be carried out over the entire area or preferably in a spot (partial area). In this case, the use of a particle counter may be useful.
[0029] A suitable basis for the correct classification of particles are standard particles of known size and chemical composition. Standardized particles are commercially available in a wide variety of different chemical compositions, such as aluminum, copper or yttrium.
[0030] The particles used preferably have the same chemical composition but different sizes, the size distribution of the particles used being very preferably between 20 nm and 100 nm.
[0031] After the particles are applied to the test wafer, the size of a certain number of particles is ascertained and an energy dispersive x-ray spectroscopy spectrum is recorded. Both items of information are stored.
[0032] Energy dispersive x-ray spectroscopy is performed using, for example, a 70 mm EDAX Elite Super EDX Detector. 2 The measurements may be performed using a Zeiss Auriga 60 SEM-EDX instrument with the following settings: 5 keV, working distance: 6.1 mm, orifice 30 μm, EDX integration time: 10 seconds, however, any other equivalent instrument may also be used.
[0033] The size of particles of known chemical composition is preferably measured using an electron microscope (scanning electron microscope).
[0034] "Size" in this context refers to the diameter determined for a particle, or equivalently, it may also refer to the area of a particle or a dimension derived directly therefrom.
[0035] The spectra obtained are preferably analyzed so that they yield a substantial content of the desired substance, which may be done by analyzing the spectra for peaks that correspond to the chemical species.
[0036] It is particularly preferred to use SiO2 particles, where the oxygen content and size are determined, however, other particles having other known chemical compositions may also be used for the same purpose.
[0037] The substantial content is preferably plotted as a function of size (e.g., diameter) and the plot is used to determine a best-fit curve, which generates a mathematical relationship between the substantial content and size in the form of a best-fit curve.
[0038] Both the size and chemical composition of particles of unknown chemical composition are ascertained on the surface of a semiconductor wafer, preferably comprising silicon.
[0039] In this case, it is preferred to use energy dispersive x-ray spectroscopy. The size of the unknown particles is preferably confirmed using scanning electron microscopy (SEM).
[0040] Figure 1, for example, shows the oxygen content of SiO2 particles as a function of diameter (open circles) normalized by energy dispersive x-ray spectroscopy. A linear best fit curve was fitted to show the proportional relationship between oxygen content and diameter. A confidence region can also be identified such that 95% of the data points lie within this region.
[0041] The locations of the measurement points on particles of unknown chemical composition (black circles, open triangles) shown in Figure 1 can provide information about the chemical composition without data on the particle's structure.
[0042] For example, the data represented by the filled circles show that the oxygen content of the particles is apparently largely independent of their size.
[0043] The inventors have now realised that SiO2 particles can with a high degree of certainty be excluded as the source of these particles, and therefore these particles will be classified as "non-SiO2 particles".
[0044] For data points that fall within the confidence region of SiO2 particles, it can be assumed that they are also likely to be SiO2 particles, and therefore the classification is read as "SiO2 particles".
[0045] For data points that are clearly not within the SiO2 confidence range and also have no oxygen content independent of size (e.g., open triangles in Figure 1), it can be assumed that these are particles with a specific oxygen content that are not spherical, or that they are particles that probably have an intrinsic oxide layer. This information can also be useful if the goal is to determine the origin of the particle's occurrence. Particles of this kind are also classified as "non-SiO2 particles".
[0046] The inventors have realized that the described procedure is particularly advantageous when it can be assumed that the semiconductor wafer on which the unknown particle is located carries a specific oxide layer, for example a native oxide layer. This oxide layer affects energy dispersive x-ray spectroscopy. Through additional information on the size of the particle and through comparison with a best fit curve for a known particle, this measurement error shows only a small negative effect.
[0047] A similar approach can be performed with particles of different chemical composition as well, for example to determine their likely origin.
[0048] The above description of the exemplary embodiments should be understood as an example. The disclosure made therewith, on the one hand, enables a person skilled in the art to understand the present invention and its attendant advantages, and, on the other hand, the understanding of the person skilled in the art extends to obvious modifications and variations of the structures and methods described. The present invention is therefore intended to cover within the scope of protection of the claims all such modifications and variations, as well as equivalents.
Claims
1. 1. A method for classifying unknown particles on a surface of a semiconductor wafer, comprising:
11. A method comprising applying particles of known chemical composition and different sizes to a test wafer, ascertaining the size of a plurality of particles of known chemical composition, and recording an energy dispersive x-ray spectroscopy spectrum of a plurality of particles of known chemical composition, followed by ascertaining a substantial content therefrom, and fitting a best fit curve to the size and substantial content of the particles of known chemical composition, as well as ascertaining a particle size of an unknown particle and recording an energy dispersive x-ray spectroscopy spectrum of the unknown particle, and therefrom determining the substantial content of the unknown particle on the semiconductor wafer, and classifying the unknown particle as a result of comparing the size and the substantial content of the unknown particle to the best fit curve.
2. The method of claim 1 , wherein the best fit curve is a straight line.
3. The method of claim 1 or claim 2, wherein the test wafer comprises silicon.
4. The method of claim 1 , wherein the semiconductor wafer comprises silicon.
5. 5. The method of claim 1, wherein the size of the unknown particle is determined by electron microscopy.
6. 6. The method of claim 1, wherein the size of the known particles is determined by electron microscopy.
7. Particles of known chemical composition are SiO 2 The method of any one of claims 1 to 6, comprising:
Citation Information
Patent Citations
JP1988100838U
Foreign object analytical device and method, and semiconductor manufacture control device and method
JP1996124982A
Automatic pattern classifying method based on defect- referencing system
JP2001068518A
