JP2009054771A - Method and device of evaluating defect of semiconductor crystal - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method of rapidly, nondestructively and accurately evaluating a two-dimensional distribution of crystalline structure defects of a semiconductor sample consisting of a wide gap semiconductor. <P>SOLUTION: The method of evaluating a two-dimensional distribution of crystalline structure defects of a semiconductor sample includes processes of: emitting photoluminescence light from the semiconductor sample by irradiating the semiconductor sample with light; and obtaining a two-dimensional distribution of the crystalline structure defects of the semiconductor sample by observing the emitted photoluminescence light. In the method, a semiconductor constituting the semiconductor sample is a wide gap semiconductor, and light for irradiating the semiconductor sample therewith has energy smaller than energy corresponding to a band gap of the wide gap semiconductor. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a semiconductor substrate evaluation method and an evaluation apparatus, and in particular, a high-speed non-destructive two-dimensional distribution of crystal structure defects of a semiconductor substrate such as silicon carbide (SiC) used in a low-loss power device or the like. The present invention relates to a semiconductor substrate evaluation method and an evaluation apparatus that enable highly accurate evaluation.

Due to the growing interest in energy and environmental issues, the development of low-loss power devices has been taken up as a national project.
Forbidden bands called “wide-gap semiconductors” that are considered as semiconductors that exceed the physical limits of conventional semiconductor device materials such as silicon (Si) and gallium arsenide (GaAs) Semiconductors with large width energy are attracting great attention.
Of these “wide-gap semiconductors”, device fabrication using silicon carbide (SiC) wafers, which are now scientifically included in semiconductors, but are sometimes referred to as “semi-insulating”, is also known. It is becoming more active. The semi-insulating characteristics of SiC are obtained by compensating residual impurities that form shallow levels by the deep levels formed by point defects and vanadium impurities in the wafer.
Silicon carbide (SiC), which has excellent physical properties, is the most promising semiconductor material for next-generation low-loss power devices, but the quality of the crystal is inferior to that of current mainstream silicon (Si), resulting in higher quality. (Low defect density) is a problem. In response to the increased device fabrication using silicon carbide (SiC) wafers, in recent years there has been a growing demand for highly accurate evaluation of SiC wafers.

In response to the above demands, conventionally developed evaluation methods are typically 1) light scattering method, 2) etching method, 3) X-ray topography method, 4) transmission electron diffraction method, 5 ) Photoluminescence method is known. Therefore, these will be described individually below.
1) Light scattering method: This method measures light scattering due to crystal structure defects by transmitting light through the wafer. Visible and infrared light is used for SiC wafers. In this method, the light passing through the light is scattered by the defect and the crystal defect is detected. Although this method has a feature of non-destructive inspection, even if the presence of a defect can be confirmed by this method, it is difficult to identify the type of defect.
2) Etching method: This method is a method of visualizing the crystal structure defect portion by immersing the SiC wafer in molten potassium hydroxide (KOH). This method is most commonly used as a means for examining the two-dimensional density distribution of crystal structure defects such as micropipes, dislocations, stacking faults, and inclusions. However, since this method is a destructive test, a device cannot be made using the evaluated wafer itself, and an expensive wafer must always be sacrificed when performing the evaluation. In addition, this method requires the use of powerful KOH heated to 500 ° C, which is also a problem in terms of safety. Further, this method does not know the density distribution of point defects.
3) X-ray topography method: This method uses X-ray diffraction. According to this method, as in the etching method, the two-dimensional density distribution of crystal structure defects such as dislocations, stacking faults, and inclusions can be found, but the density distribution of point defects is not known. This method has the feature of non-destructiveness, but the shape such as the warp of the sample greatly affects the evaluation. Furthermore, high-precision evaluation using this method requires several hours of measurement or the use of a high-intensity X-ray source from a large special facility such as synchrotron radiation.
4) Transmission electron diffraction method: This method has very high resolution with respect to crystal structure defects, but it is practically impossible to perform mapping of the entire semiconductor wafer using this method. This method is a destructive test.

5) Photoluminescence method: This method utilizes the following principle.
FIG. 1 illustrates the mechanism itself in which the photoluminescence process occurs, schematically showing the energy band structure of SiC, and in the forbidden band 3 between the valence band 1 and the conduction band 2. , A deep level 4 and a shallow level 5 are illustrated.
When such a material is irradiated with light having a photon energy larger than the width of the forbidden band 3, as indicated by a thick arrow 10 extending from the valence band 1 to the conduction band 2, excessive electrons are generated in the interband absorption process. -Some of the electrons that have been generated and thus struck into the conduction band 2 are in the shallow level 5 in the trapping process indicated by the arrow 6 and some in the same way by the arrow 8 It is trapped at deep level 4 in the capture process shown.
As indicated by arrows 7 and 9, an event that generates light in the process of recombining with holes in the valence band 1 is generally called photoluminescence, and the light generated by this photoluminescence is spectroscopically analyzed. When analyzed, the type of level that caused the emission of the photoluminescence light can be specified, and the concentration of each specified level can be known by the intensity analysis.
However, in actual photoluminescence, in addition to light emission based on the presence of shallow levels in the photon energy region slightly smaller than the forbidden band energy in the process indicated by the arrow 7 in FIG. 1, exciton light emission, interband transition Although a light emission phenomenon generally referred to as light emission near the band edge, such as light emission, is recognized, in general, the light emission process based on the presence of the shallow level 5 is generally used to represent light emission near the band edge.
Compared with the conventional methods 1) to 4) above, the photoluminescence method can evaluate a minute region on the wafer surface by narrowing the irradiation light, and can also be applied to the active region for device fabrication in the depth direction. It has the excellent feature of non-destructive and non-contact method in that it can perform a temporary evaluation in the corresponding extremely shallow region.
However, the measurement time per wafer by the conventional photoluminescence method is about 30 minutes or more in the case of a typical wafer having a diameter of 50 mm, and it is practically impossible to quickly evaluate many wafers. The fact is that it has not been possible to obtain statistically reliable data.

JP 2006-147848 A JP 2006-349482 A R.E. Stahlbush, K.X. Liu, Q. Zhang and J.J. Sumakeris: “Whole-Wafer Mapping of Dislocations in 4H-SiC Epitaxy” Mater. Sci. Forum, Vol. 556-557, pp. 295-298 (2007) Masahiro Nagano, Noriho Kamada, Shuichi Tsuchida: "PL mapping observation of dislocations and stacking faults in 4H-SiC using UV-LED, 2D-CCD", SiC and related wide gap semiconductor workshop 15th lecture, P -18, p. 48 (2006) M. Tajima, E. Higashi, T. Hayashi, H. Kinoshita and H. Shiomi: "Nondestructive characterization of dislocations and micropipes in high-resistivity 6H SiC wafers by deep-level photoluminescence mapping," Appl. Phys. Lett., Vol 86, No. 5, pp. 061914-1-061914-3 (2005)

The crystallinity of the current SiC is very inferior compared to Si and GaAs. For this reason, in SiC wafers, there are huge cavities called sub-micron diameters called “micropipes”, “dislocations” that are deviations in the regular atomic arrangement of crystals, and disorder in the laminated structure of crystal lattices. There are many crystal structure defects such as “stacking defects”. Since these crystal structure defects may adversely affect the characteristics, yield, and reliability of the device, the reduction of the defects becomes the greatest problem. In addition, the distribution of point defects and impurities that cause the deep levels described above depends on the distribution of crystal structure defects. Furthermore, if there is a crystal structure defect in the device active region, the device is very likely to malfunction.
In the current technology, crystal structure defects exist in the wafer at a fairly high density, and the distribution is extremely non-uniform. In fact, the characteristics of the electronic devices built on the SiC wafer actually vary considerably. As well as lowering the yield of the device, it has become a major obstacle to achieving high reliability and high performance.
For this reason, in this type of field, it is desired to evaluate crystal structure defects in SiC wafer crystals, particularly to detect the density distribution of crystal structure defects along the wafer surface quickly and easily with high spatial resolution. The demand has been made strong. In particular, with regard to the speed of defect evaluation, for the purpose of checking the total number of wafers cut out from an ingot, it is required to measure within one minute per wafer.
Accordingly, an object of the present invention is to provide a method and apparatus for evaluating a two-dimensional distribution of crystal structure defects of a semiconductor sample composed of a wide gap semiconductor such as a SiC wafer at high speed and nondestructively with high accuracy. It is what.

The inventor, when obtaining photoluminescence from a semiconductor sample composed of a wide gap semiconductor such as a SiC wafer, the inventor uses energy smaller than the energy corresponding to the band gap of the wide gap semiconductor as excitation light. The present invention has been reached based on the knowledge that it is possible to use the light that it has.
In the above prior art, attention has been paid to the intensity of the photoluminescence light indicated by the arrow 9 in FIG.
On the other hand, the present inventor has an excitation band 15 in the deep level 14 as shown in FIG. 2, which is smaller than the energy corresponding to the band gap 13 as shown by the arrow 16 in FIG. The present invention has been reached based on the knowledge that electrons in the valence band can be directly beaten to the deep level 14 by light having energy.
The fact that excitation with light having energy smaller than the band gap is possible means that it is possible to use visible light as an excitation light source instead of ultraviolet light used in the conventional method. . Using visible light instead of ultraviolet light as the excitation light source
(1) The optical system can be used with ordinary optical components in the visible light range, simplifying the device,
(2) Ultraviolet rays cause fluorescence (also photoluminescence) from dust on the wafer surface, optical components such as filters and lenses, and cause noise, but this is not the case with visible light ,
(3) Visible light is safer to handle than invisible UV light,
And there are advantages.
By irradiating the entire surface of the wafer with a uniform intensity using a visible light emitting diode array, and taking a deep level photoluminescence image emitted from the wafer with a high-sensitivity CCD camera, an intensity distribution on the entire surface of the wafer can be obtained. it can. In the present invention, since the laser is not scanned to measure each point of the wafer as in the conventional method, the measurement time is greatly shortened.

That is, the present invention includes a step of irradiating a semiconductor sample with light to emit photoluminescence light from the semiconductor sample, and observing the emitted photoluminescence light to observe two-dimensional crystal structure defects of the semiconductor sample. A semiconductor sample crystal comprising a step of obtaining a distribution, wherein the semiconductor constituting the semiconductor sample is a wide gap semiconductor, and the light applied to the semiconductor sample has an energy smaller than the energy corresponding to the band gap of the wide gap semiconductor A method for evaluating a two-dimensional distribution of structural defects is provided.
In the method of the present invention, it is preferable that the light applied to the semiconductor sample is visible light.
In the method of the present invention, the semiconductor may be silicon carbide, and the crystal structure defect may include any of a micropipe defect, a dislocation, a stacking fault, and an inclusion.
Furthermore, the present invention provides a light source for irradiating a semiconductor sample with light to emit photoluminescence light from the semiconductor sample, an observation means for observing the emitted photoluminescence light, and the observed photoluminescence A defect distribution evaluation means for obtaining a two-dimensional distribution of crystal structure defects of a semiconductor sample from light, and a light source that emits light having energy smaller than energy corresponding to a band gap of a wide gap semiconductor. An apparatus for evaluating a two-dimensional distribution of crystal structure defects of a semiconductor sample composed of a gap semiconductor is provided.
In the apparatus of the present invention, the light source preferably emits visible light.
In the method of the present invention using light having energy smaller than the band gap as excitation light, a photoluminescence intensity distribution can be obtained by normal visible light excitation. Since it is visible light excitation, the optical system is simple, and there are few light emission components (noise) from other than the sample, and a clear image can be obtained.

In the present invention, the light having a wavelength smaller than the energy corresponding to the band gap of the wide gap semiconductor is used with respect to the light having a wavelength, such as the shallow level 18 and the deep level shown in FIG. Pay attention to the level 14.
In general, the deep level 14 is in an energy region from a position deeper than the shallow level 18 in FIG. 2 (downward in FIG. 2) to the vicinity of the center of the band gap 13 (sometimes referred to as “midgap level”). Exists. For example, in the case of a semiconductor sample in which SiC is used as a wide gap semiconductor, in order to excite such a deep level as shown by an arrow 16, the band gap energy of SiC is about 3.2 eV. Considering that the shallowest level known for SiC is about 0.1 eV from the bottom of the conduction band, from the mid-gap level to about 0.1 eV deeper than the shallowest level, i.e. about 1.6 Light having a photon energy of about −3.0 eV (wavelength of about 400 to 800 nm) may be used as excitation light. The shallow level 18 is considered to be about 0.1 eV deeper than the lower end of the conduction band regardless of the type of the wide gap semiconductor. Therefore, in the present invention, the shallow level 18 is smaller than the energy corresponding to the band gap of the wide gap semiconductor. As light having energy, it is considered that light having energy smaller by about 0.2 eV than the band gap can be used effectively. Thus, according to the present invention, light in the substantially visible light region can be preferably used (the wavelength region in the visible light region is 400 to 700 nm).
Deep level optical transitions include transitions between electron shell levels created by deep levels called intrashell transitions. In this case, electrons and holes are not directly exchanged with the conduction band and valence band as shown in FIG. 2, but the energy between the electron shell levels corresponding to the photoexcitation process 16 is about 1 eV or more. . Therefore, the photon energy of light that can be used for excitation is about 1.0-3.0 eV (wavelength 400-1200 nm) in addition to the above-mentioned range 1.6-3.0 eV. In other words, when targeting a deep level that exhibits light emission due to such intrashell transition, there is a possibility that the long wavelength side of the excitation light can be handled by extending it to the near infrared region of about 1200 nm.
The photon energy of light emission from the deep level (arrow 17) is accompanied by phonon emission, and is therefore smaller than the minimum energy required for excitation (energy from the valence band to the lowest end of the excitation band 15). Become.
For example, when evaluating a SiC-free semi-insulating wafer of the same type as the wafer targeted by the conventional method, the peak is a deep emission band with a photon energy of 1.3 eV (wavelength 950 nm). . Depending on the type of wafer, it may be possible to target other light emission bands. Examples of such emission bands are the 2.5 eV (wavelength 500 nm) emission band caused by deep boron acceptors or titanium, and the 1.8 eV (wavelength 690 nm) emission band that appears well, although the cause is unknown. There is a 1.35 eV (wavelength 920 nm) emission band caused by a defect called UD3. In addition, the wavelength of the excitation light is selected according to the excitation band of each emission center, and each emission component can be selected and detected by adjusting the wavelength of the transmission band of the bandpass filter (bandpass filter). It is considered possible.

According to the present invention, the defect distribution on the entire surface of the SiC wafer can be known in as little as 100 seconds or less. This is useful for optimizing crystal growth conditions by comparing the occurrence of defects with crystal growth conditions. Since it is a non-destructive / non-contact measurement, it is possible to inspect all wafers before shipping. As a result, since a defective part (defective part) in the wafer can be known in advance, it can be reflected in the device process, and the added value of the wafer can be increased. In addition, the defect distribution is measured for each process of the device, the occurrence state of the defect is known, and the process is improved.
The idea of the present invention can be applied not only to SiC but also to other wide gap semiconductors such as gallium nitride (GaN) and zinc oxide (ZnO). In GaN, a deep-level emission band called a yellow band with a wavelength near 580 nm is known. On the other hand, light below the band gap, for example, green-blue (wavelength 480-500 nm) is used as excitation light. May be used to obtain PL imaging.

FIG. 3 is a schematic configuration diagram of an apparatus configuration example used in the evaluation method according to the present invention.
11 is a SiC wafer sample to be evaluated. The shape of the sample 11 may be an arbitrary shape such as a circle or a rectangle. Although it is desirable that the surface of the sample 11 is flat, the unevenness within the in-focus range of the camera used for observing the PL light is an allowable range.
As the excitation light source 14 for irradiating the sample semiconductor substrate 11 with light, a visible light (wavelength 400 nm to 700 nm) light source having an energy smaller than the forbidden band width of SiC, for example, a light emitting diode array can be used. . As the light source 14, a laser or a laser diode can be used. It is desirable to attach a filter 15 such as an infrared cut filter to the light source 14.
As the PL light detector 16 for observing the photoluminescence light emitted from the sample semiconductor substrate 11, for example, an electronic cooling CCD camera can be used. In front of the objective lens 17 of the detector 16, the excitation light of the light source 14 is cut (for example, the wavelength of 900 nm or less is cut), but the PL light from the sample semiconductor substrate 11 is transmitted. It is desirable to install. For example, when detecting a 1.3 eV (wavelength 950 nm) emission band, a bandpass filter having a transmission wavelength band of 915 to 975 nm is used.
Furthermore, the PL light of the sample semiconductor substrate 11 can normally pass through the sample semiconductor substrate 11 itself. Therefore, it is possible to adopt a mode in which light is emitted from the light source 14 from the back surface side of the sample semiconductor substrate 11 and a PL image is captured by the detector 16 such as a camera installed on the front surface side of the sample semiconductor substrate 11. .
In addition, the sample semiconductor substrate 11 to be evaluated is immersed in an etching solution (not shown) filled in a container such as a plastic petri dish to perform evaluation, whereby the surface recombination of the carrier of the substrate by the etching solution is performed. It seems possible to measure the PL distribution of the substrate more accurately by using the suppression effect.

In the apparatus configuration of FIG. 3, PL light from the sample semiconductor substrate 11 was observed using a light-emitting diode array as the excitation light source 14 and an electronic cooling CCD camera as the detector 16.
As the sample semiconductor substrate 11, an SiC wafer having an additive-free semi-insulating property, a polytype (crystal polymorph due to a difference in repetition period) having a 6H structure and a diameter of 50 mm was used.
As the light emitting diode array, 14 Luxeon V Stars (wavelength: 500 nm) manufactured by Philips Lumileds Lighting Company were used. An infrared cut filter 15 (cutting after a wavelength of 600 nm) was installed on the front surface of the light emitting diode array 14.
As the above-described electronically cooled CCD camera 16, an infrared sensitizing type camera having a number of pixels of 1024 × 1024 and an operating temperature of −70 ° C. was used. The lens of the electronically cooled CCD camera 16 was an industrial lens having a focal length of 25 mm and F1.4. A band-pass filter 18 that transmits a wavelength range of 915 to 975 nm is installed on the front surface of the lens.
With the above configuration, it was confirmed that a deep level emission band having a peak at a photon energy of 1.3 eV (wavelength 950 nm) appeared from the sample 11. This light emission is generally observed in commercially available SiC-free semi-insulating wafers, and is caused by a vacancy that is one type of point defect. In a sample in which other deep level emission appears, the emission band to be measured is extracted by adjusting the wavelengths of the excitation light source 14 and the band pass filter 18 of the apparatus of FIG. 3 according to the difference in the deep level. .
FIG. 4 shows the result of taking the PL image of the above 1.3 eV emission band with an electronic cooling CCD camera. Although the exposure time was only 100 seconds, the two-dimensional distribution of crystal structure defects in the SiC wafer could be observed clearly.

[Comparative example]
For comparison, FIG. 5 shows the result of measuring the entire wafer PL mapping by ultraviolet laser scanning for the same sample by the method described in Patent Document 1 (Japanese Patent Laid-Open No. 2006-147848). The measurement time in this case was about 30 minutes.
In each of the example and the comparative example, exactly the same PL pattern was obtained. From comparison with the results observed by etching, the micropipe appears as a large dark spot and the dislocation appears as a dark linear pattern. It was clearly confirmed.
As described above, according to the present invention, it is understood that the same information can be obtained with a simple device configuration in about 1/20 of the time of the conventional method. The conventional method mentioned here is remarkably superior to the generally used etching method and X-ray topography method, but the present invention can be said to be an epoch-making method that surpasses that.
Of course, from the above method principle and apparatus configuration according to the present invention, it is obvious that the present invention can be effectively used for various semiconductors other than SiC.

It is explanatory drawing which shows the principle of the conventional method. It is explanatory drawing which shows the principle of this invention. It is a schematic block diagram of the apparatus structural example used for the evaluation method by this invention. It is a PL image obtained by the method of the present invention. It is a PL image obtained by the method of the comparative example.

Explanation of symbols

21 Sample semiconductor substrate 22 Excitation light source 23 Infrared cut filter 24 Detector 25 Band pass filter

Claims (5)

A method for evaluating a two-dimensional distribution of crystal structure defects in a semiconductor sample,
Irradiating the semiconductor sample with light and emitting photoluminescence light from the semiconductor sample; and
Observing the emitted photoluminescence light to obtain a two-dimensional distribution of crystal structure defects of the semiconductor sample;
Including
The semiconductor constituting the semiconductor sample is a wide gap semiconductor,
The light applied to the semiconductor sample has an energy smaller than the energy corresponding to the band gap of the wide gap semiconductor,
The evaluation method as described above.   The evaluation method according to claim 1, wherein the light applied to the semiconductor sample is visible light.   The evaluation method according to claim 1, wherein the semiconductor is silicon carbide, and the crystal structure defect includes any one of a micropipe defect, a dislocation, a stacking fault, and an inclusion. An apparatus for evaluating a two-dimensional distribution of crystal structure defects of a semiconductor sample composed of a wide gap semiconductor,
A light source for irradiating the semiconductor sample with light and causing the semiconductor sample to emit photoluminescence light;
An observation means for observing the emitted photoluminescence light, and
A defect distribution evaluation means for obtaining a two-dimensional distribution of crystal structure defects of the semiconductor sample from the observed photoluminescence light;
With
The light source emits light having energy smaller than energy corresponding to a band gap of the wide gap semiconductor;
The evaluation apparatus characterized by the above.   The evaluation apparatus according to claim 4, wherein the light source emits visible light.

Images