METHOD FOR EVALUATING CRYSTAL DEFECTS

Abstract

A method for evaluating crystal defects, by which a distribution of crystal defects present in a silicon wafer is assessed, comprising: forming an oxide film with a thickness equal to a crystal defect size to be assessed on the silicon wafer, measuring GOI features of the silicon wafer, and determining the distribution of crystal defects with the crystal defect size to be assessed in the silicon wafer from a measurement result of the GOI features under the assumption that the crystal defects whose size corresponds to the thickness of the oxide film are present in an area where the GOI features are impaired.

Description

TECHNICAL AREA

[0001] The present invention relates to a method for evaluating crystal defects. STATE OF THE ART

[0002] Although a silicon oxide film exhibits very good insulating properties, it can be formed in a simple process consisting of heat treatment in an oxidizing atmosphere, and is therefore widely used in device manufacturing. When a gate oxide film is applied with respect to thickness, high voltage is required, thus necessitating a high-quality film.

[0003] Furthermore, it is known that if a wafer contains crystal defects (which may hereinafter be referred to simply as defects), these are incorporated into a gate oxide film to form a defective oxide film which may be the cause of device defects.

[0004] One oxide breakdown voltage (GOI) evaluation method used to detect such a defective oxide film is the TDDB method. This method involves continuously applying a fixed voltage or current to an insulating film, recording the current or voltage at predetermined time intervals to determine changes over time, and evaluating the time to breakdown, its process, and other details.

[0005] A percolation model has been proposed as a mechanism for generating insulator film breakdown in the TDDB method. This is a probabilistic model that, for example, divides a silicon oxide film into a lattice structure to form small cells, estimates a probability that each cell will become defective, and assumes that breakdown will occur when defective cells form a line in one direction of the oxide film's thickness.

[0006] According to this mechanism, if a weak point is present in the oxide film, defects in the oxide film increase over time, the distance between the defects decreases, thus facilitating the movement of electrons, and finally a current path is formed, resulting in a breakdown.

[0007] This means that if a small weakness exists in the oxide film, it grows and leads to film penetration. Therefore, it was assumed that even a defect whose actual size is unknown, but which is small enough to account for the thickness of the oxide film, will become a defect that leads to oxide film penetration. LIST OF CAPLETS PATENT LITERATURE

[0008] Patent literature 1: Japanese unexamined patent specification (Kokai) No. 2007-191350 SUMMARY OF THE INVENTIONAL ENGINEERING TASK

[0009] According to the present inventor's investigation, a highly sensitive LST (Laser Scattering Tomography) was recently developed, enabling the direct observation of any void approximately 10 nm in size. It was thus found that when an oxide film thickness and a crystal defect size (hereinafter referred to simply as defect size) are similar, such as a conventional oxide film thickness of 25 nm and an LST sensitivity of 20 nm, the LST defect density and an oxide breakdown voltage show a very good correlation. However, a void density of 10 nm in size cannot achieve a correlation with a breakdown voltage of the oxide film 25 nm thick.

[0010] Table 1, Fig. 1 and the paragraphs 17 until 19Patent literature 1 shows that an oxide film thickness of 71 nm, 83 nm, and 108 nm is specified, and GOI yields, GOI defect densities, and GOI minimum defect sizes are achieved in such situations. However, a distribution of microdefects, which can hardly be detected by a measuring instrument with a defect size detection limit of approximately 10 nm, cannot be determined from a GOI measurement result at such oxide film thicknesses.

[0011] The cavity (COP), which is a target in patent literature 1, has a size of 40 nm or more, which is very easy to measure; however, the task is not to detect defects with a size of about 10 nm, which are hardly detectable by the measuring instrument.

[0012] Regarding how a defect size is determined in patent literature 1, defects of a predetermined size are not generated; however, a minimum GOI defect size is determined from the density of all COPs (cumulative defect density) and GOI defects (a GOI yield) detected by an OPP (optical precipitate profiler), under the assumption that relatively large COPs impair the GOI. That is, a relationship between a defect size and a GOI defect is not confirmed.

[0013] In the measurement of patent literature 1, a good chip yield (the GOI yield) of 99.1% at the level of a gate oxide film thickness of 108 nm actually means that the number of defects is only 2 at 229 measurement points, and an accuracy is inappropriate if the good chip yield is used to calculate the density (the GOI defect density).

[0014] As previously described, the crystal defect size affecting the oxide breakdown voltage is larger than previously considered and corresponds to the thickness of any oxide film. Small defects that were missed in conventional thick oxide films and correlate with GOI features of thin oxide films must be detected. However, a defect density meter using LST or similar technology has a detection limit, and the problem remains that detecting small defects of 10 nm or less is difficult.

[0015] Because of this problem, one object of the present invention is to provide a method for evaluating crystal defects by which a distribution of crystal defects can be determined even when the crystal defect size is 10 nm or less. TECHNICAL SOLUTION

[0016] To solve the problem, the present invention provides a method for evaluating crystal defects, by which a distribution of the crystal defects present in a silicon wafer is evaluated, comprising: Formation of an oxide film with a thickness equal to the size of a crystal defect to be evaluated on the silicon wafer, Measuring GOI characteristics of the silicon wafer, and Determining the distribution of crystal defects with the crystal defect size to be evaluated in the silicon wafer from a measurement result of the GOI features under the assumption that the crystal defects, whose size corresponds to the thickness of the oxide film, are present in an area where the GOI features are impaired.

[0017] Since defects whose size corresponds to half to twice the oxide film thickness impair oxide breakdown voltage, forming an oxide film with a thickness equal to the crystal defect size to be evaluated on the silicon wafer, measuring GOI features of the silicon wafer, and determining the distribution of crystal defects with the crystal defect size to be evaluated in the silicon wafer from a measurement result of the GOI features, assuming that the crystal defects whose size corresponds to the thickness of the oxide film are present in an area where the GOI features are impaired, as previously described, can eliminate the need to use a defect density measuring instrument with LST or similar and enable the testing of the density or distribution of small crystal defects that cannot be measured in prior art examples.

[0018] At this stage, it is preferable to define a majority of the crystal defect sizes to be evaluated and to determine the distribution of crystal defects according to each crystal defect size in the silicon wafer.

[0019] In this way, the crystal defect distribution can be easily determined for each crystal defect size.

[0020] At this stage, it is preferable that the crystal defect size to be evaluated is 10 nm or less.

[0021] If the crystal defect size to be evaluated is 10 nm or less, the cavity cannot be directly observed, even when the LST is used, and the defect density cannot be measured, and thus the method for evaluating crystal defects according to the present invention can preferably be used, which does not employ a defect density measuring instrument with LST or the like. ADVANTAGEOUS EFFECTS OF THE INVENTION

[0022] As previously described, when the method for evaluating crystal defects according to the present invention is used, an oxide film whose thickness is the same as a crystal defect size to be evaluated is formed on a silicon wafer, GOI features of the silicon wafer are measured, and the distribution of crystal defects with the crystal defect size to be evaluated in the silicon wafer is determined from a measurement result of the GOI features; and thus the density or distribution of crystal defects, whose sizes are 10 nm or less, which cannot be measured in prior art examples, can be checked without the need for a defect density measuring instrument with LST or similar. List of characters Fig. Figure 1 shows a flow diagram illustrating a method for evaluating crystal defects according to the present invention. Fig. Figure 2 shows a view illustrating GOI features at each oxide film thickness level. Fig. Figure 3 shows a view illustrating error size distributions in a boundary section and a central section of a sample according to the example. Fig. Figure 4 shows a view illustrating the defect size-specific radial distribution when defects of 20 nm or more and defects of 10 nm or more are actually measured by using an LST in the sample of the example. Fig. Figure 5 shows a view illustrating that an oxide breakdown voltage is not affected, even if any defect size is too large or too small for a thickness of an oxide film. DESCRIPTION OF THE EXECUTION FORMS

[0023] Although the present invention is described in detail below as one embodiment with reference to the drawings, the present invention is not limited thereto.

[0024] As previously described, a TDDB method exists for evaluating oxide breakdown voltage (GOI) to detect a defective oxide film, and a percolation model has been proposed as a mechanism that generates insulator film breakdown in the TDDB method. According to this method, if a small defect exists in an oxide film, it grows and leads to film breakdown. Thus, it was assumed that even a small defect, the actual size of which is unknown but small enough to be within the thickness of the oxide film, will become a defect that leads to oxide film breakdown.

[0025] However, it was discovered that when the highly sensitive LST, which allows direct observation of a cavity with a size of approximately 10 nm, is used, if an oxide film thickness and a defect size are similar, such that a conventional oxide film thickness of 25 nm and the LST sensitivity of 20 nm, the defect density measured by the LST and an oxide breakdown voltage show a very good correlation, but a cavity density (defect density) with a defect size of 10 nm and an oxide film breakdown voltage with a thickness of 25 nm cannot achieve a correlation.

[0026] As previously described, the crystal defect size affecting the oxide breakdown voltage is larger than previously considered and corresponds to the thickness of the oxide film. Small defects, which were missed in conventional thick oxide films and correlate with the GOI characteristics of thin oxide films, must be detected. However, a defect density measuring instrument using LST or similar technology has a detection limit, and the problem remains that detecting small defects of 10 nm or less is difficult.

[0027] Thus, the present inventor has comprehensively investigated a method for evaluating crystal defects by which a distribution of crystal defects can be determined even when a crystal defect size is 10 nm or less. The present inventor has thus discovered that forming an oxide film with a thickness equal to the crystal defect size to be evaluated on a silicon wafer, measuring GOI features of the silicon wafer, and determining the distribution of crystal defects with the crystal defect size to be evaluated in the silicon wafer from a measurement result of the GOI features, assuming that the crystal defects whose size corresponds to the thickness of the oxide film are present in an area where the GOI features are affected, eliminates the need to use a defect density measuring instrument with LST or similar.which can eliminate and enable the testing of the density or distribution of small crystal defects that cannot be measured in prior art examples, thereby carrying out the present invention.

[0028] The following is the method for evaluating crystal defects according to the present invention with regard to Fig. 1 described.

[0029] First, a silicon wafer is prepared (see S11 in Fig. 1).

[0030] Subsequently, an oxide film with a thickness equal to a crystal defect size to be evaluated is formed on the prepared silicon wafer (see S12 in Fig. 1) Here, a silicon oxide film can be used as an oxide film, and the formation of the silicon oxide film preferably takes place through thermal oxidation.

[0031] Subsequently, GOI characteristics of the silicon wafer on which the oxide film was formed are measured (see S13 in Fig. 1) The GOI features can be measured using the TDDB method described above. Furthermore, to determine an accurate defect density, the GOI features are preferably measured at as many measurement points as possible across the entire surface of the silicon wafer.

[0032] Subsequently, assuming that crystal defects whose size corresponds to the thickness of the oxide film are present in an area where the GOI features are impaired (i.e., at the measurement points where the GOI features are impaired), a crystal defect distribution of the crystal defect size to be evaluated in the silicon wafer is determined from a measurement result of the GOI features (see S14 in Fig. 1) Here, in the present invention, “correspond” means half to twice the thickness. It is assumed that the crystal defects, whose size corresponds to the thickness of the oxide film, are present in the area where the GOI features are impaired, because it is assumed that an oxide breakdown voltage is not affected, even if the defect size is too large or too small relative to the thickness of the oxide film. That is to say, if, as in Fig. As shown in Figure 5(a), where the size of the defect corresponds to the thickness of the oxide film, and since the thickness of the oxide film at a position of the defect becomes smaller than the thickness of the oxide film at its edge, it can be assumed that an electric field concentration occurs and the GOI features are affected. Furthermore, as shown in Fig. 5(b) shows that if the size of the defect is larger than the thickness of the oxide film, the thickness of the oxide film at the location of the defect is the same as the thickness of the oxide film at its edge, and thus it can be assumed that the GOI features are not affected. Furthermore, as shown in Fig. 5(c) shows that if the size of the defect is smaller than the thickness of the oxide film, the thickness of the oxide film at the location of the defect is the same as the thickness of the oxide film at its edge, and thus the GOI characteristics can be assumed to be unaffected. Therefore, any defect whose size is half to twice the thickness of the oxide film that affects the oxide breakdown voltage can be considered acceptable. This has been experimentally confirmed using data from an example described below.

[0033] As previously described, the oxide film with a thickness equal to the crystal defect size to be evaluated is formed on the silicon wafer, the GOI features of the silicon wafer are measured, and the distribution of the crystal defects with the crystal defect size to be evaluated in the silicon wafer is determined from a measurement result of the GOI features; thus, no defect density measuring instrument with LST or similar needs to be used, and the density and distribution of small crystal defects, which cannot be measured in examples according to the state of the art, can be checked.

[0034] In the procedure for evaluating crystal defects, it is preferable to define a majority of the crystal defect sizes to be evaluated and to determine the distribution of crystal defects according to each crystal defect size in the silicon wafer.

[0035] In particular, a majority of silicon wafers are prepared in S11 in Fig. 1. Oxide films with a plurality of thicknesses corresponding to a plurality of sizes of crystal defects to be evaluated are deposited on the plurality of prepared silicon wafers, each in S12 in Fig. 1 formed, GOI features of the majority of silicon wafers, on which the oxide films with the different film thicknesses are formed, are each in S13 in Fig. 1 measured, and the distribution of crystal defects according to each crystal defect size in the silicon wafers can be determined from a plurality of measurement results of the GOI features in S14 in Fig. 1 will be determined.

[0036] Here, the majority of silicon wafers are prepared in S11 in Fig. 1. The majority of silicon wafers to be prepared are preferably cut from a silicon single crystal of the same batch. This is done because defect sizes and defect distributions become essentially the same in this case.

[0037] As previously described, the distribution of crystal defects according to each crystal defect size can be easily determined in the present invention.

[0038] In the method for evaluating crystal defects according to the present invention, the size of the crystal defect to be evaluated is preferably 10 nm or less. When the crystal defect size to be evaluated is 10 nm or less, the voids cannot be directly observed, even when using LST, and the defect density cannot be measured. Therefore, the method for evaluating crystal defects according to the present invention can preferably be used without employing a defect density measuring instrument with LST or similar technology. Furthermore, a lower limit for the crystal defect size to be evaluated is not particularly restricted; however, any size greater than 0 nm can be specified. EXAMPLE

[0039] The present invention is described in more detail with reference to an example; however, the present invention is not limited to this example. (Example)

[0040] Oxide breakdown stresses were measured in three silicon wafers with a diameter of 300 mm, on which three types of oxide films with oxide film thicknesses of 5 nm, 10 nm, or 25 nm were formed and cut from a silicon single crystal of the same batch. It is assumed here that the silicon wafers cut from the silicon single crystal of the same batch exhibit essentially the same defect distributions.

[0041] Furthermore, since the oxide breakdown voltage is typically measured in only a portion of a silicon wafer (1% to 10% of an area), defects are often missed when the defect density is low, and the misinterpretation that a GOI defect is absent can occur, even when defects are present. Often, an entire surface is measured to prevent defects from being missed; however, in a common procedure, for example, an area per point is enlarged because the measurement is simple. In this case, however, the problem arises that the exact defect distribution becomes unclear. Thus, with regard to the oxide breakdown voltages in this example, a method was used by laying 14,700 samples, each measuring 2 × 2 mm. 2 The identified pattern was used, and finally the entire surface was measured.

[0042] Fig. Figure 2 shows a measurement result of the oxide breakdown voltages.

[0043] Out of Fig. 2 shows that the result is that, although the three silicon wafers have the same defect distribution, the defect distribution differs depending on each oxide film thickness.

[0044] The error distributions estimated from this result are as follows. 1) Many defects with a size of about 25 nm are present at the edge and none are present in the center. 2) Defects with a size of approximately 10 nm are present across the entire wafer and their density increases towards the edge. 3) Defects with a size of approximately 5 nm have a low overall density.

[0045] As previously described, the detection limit of a conventional measuring instrument (the error density measuring instrument) is approximately 10 nm; however, in the present invention, if an oxide film with an oxide film thickness corresponding to an error size to be detected is formed and the GOI measurement is carried out, an error distribution can be determined even if an error size of approximately 5 nm cannot be detected by the measuring instrument.

[0046] A radial distribution and cavity sizes in each sample were confirmed by using the highly sensitive LST and a Fig. The result shown in Figure 3 was determined.

[0047] Out of Fig. Figure 3 shows that the edge and the center differ in the presence / absence of defects with a size of 20 nm or more. This means the GOI defect at the edge of the wafer has an oxide film thickness of 25 nm. Fig. 2 is to be considered as being caused by defects whose size is 20 nm or greater. Furthermore, it can be assumed that a defect-free product would be located in the center of the GOI of the wafer with an oxide film thickness of 25 nm in Fig. 2 is achieved, but a defective product is found in the center of the wafer with an oxide film thickness of 10 nm in Fig. 2 is achieved due to the influence of errors whose size is 20 nm or less.

[0048] The LST was then used to actually measure defects of 20 mm or more and defects of 10 nm or more in the same wafers as previously described. The radial distribution for each defect size is shown in Fig. 4 shown.

[0049] Out of Fig. Figure 4 shows that defects with a size of 20 nm or more are less prevalent in the center but abundant at the edges. This is consistent with the GOI result when the oxide film thickness is 25 nm. Fig. 2. Furthermore, the defects, whose size is 10 nm or more, are uniformly generated on one plane, and this agrees with the GOI result when the oxide film thickness is 10 nm. Fig. 2.

[0050] The radial distribution determined by the present invention for each error quantity corresponds to the radial distribution of the errors actually measured by the LST as previously described, and it can be assumed that the actual radial distribution of each error quantity is reflected in the error distribution determined by the present invention.

[0051] Furthermore, according to the present invention, if the oxide film thickness is reduced so that it is less than 10 nm, and the GOI is measured, a radial distribution of defects with a size of 10 nm or less, which cannot be detected by the LST, can be determined.

[0052] The present invention is not limited to the preceding embodiment. The preceding embodiment represents an illustrative example, and examples having essentially the same structure and exhibiting the same functions and effects as in the technical concept described in the claims of the present invention are included within the technical scope of the present invention.

Claims

[1] Method for evaluating crystal defects, by which a distribution of crystal defects present in a silicon wafer is evaluated, comprising: Formation of an oxide film with a thickness equal to the size of a crystal defect to be evaluated on the silicon wafer, Measuring GOI characteristics of the silicon wafer, and Determining the distribution of crystal defects with the crystal defect size to be evaluated in the silicon wafer from a measurement result of the GOI features under the assumption that the crystal defects, whose size corresponds to the thickness of the oxide film, are present in an area where the GOI features are impaired. [2] Method for evaluating crystal defects according to claim 1, wherein a plurality of the crystal defect sizes to be evaluated is specified and the distribution of the crystal defects according to each crystal defect size in the silicon wafer is determined. [3] Method for evaluating crystal defects according to claim 1 or 2, wherein the crystal defect size to be evaluated is 10 nm or less.