JP5189661B2 - Inspection method of semiconductor layer - Google Patents

Description

  The present invention relates to a method for inspecting a semiconductor layer formed on a substrate.

A mixed crystal represented by the composition formula Al _x In _y Ga _1-xy N (0 ≦ x ≦ 1, 0 ≦ y ≦ 1, x + y <1) such as gallium nitride GaN, aluminum nitride AlN, and indium nitride InN. Nitride semiconductors, which are generic names, are generally mechanically robust and chemically stable, have high thermal conductivity, and excellent heat dissipation. Semiconductor devices using such nitride semiconductors, for example, a high electron mobility transistor (HEMT) combining an AlGaN layer / GaN layer or a laser diode (LD: laser diode) combining an InGaN layer / GaN layer Is expected as a high-power element.

  On the other hand, the melting point of a nitride semiconductor is very high. For example, the melting point of AlN is 3273K (Kelvin), the melting point of GaN is 2000K or higher, and the melting point of InN is 1373K (reference: Shiro Sakai, Group III nitride semiconductor, Isao Akasaki, Chapter 1, Bafukan, 1999) . Therefore, it is relatively difficult to grow a nitride semiconductor layer having high crystallinity. In fact, it is known that cracks of the order of nm occur on the surface of the nitride semiconductor layer due to subtle differences in growth conditions. Such a crack causes an increase in gate leakage current and deterioration of pulse response characteristics in the HEMT device. Therefore, quantitative evaluation of the crack density is extremely important in the manufacture of nitride semiconductor devices.

  Conventionally, quantitative evaluation of surface conditions such as crack density has been mainly performed using an atomic force microscope (AFM). The AFM measures the displacement of the cantilever by measuring the reflected light from the cantilever when the cantilever is displaced according to the interatomic force between the probe fixed to the tip of the cantilever and the atoms on the sample surface. Then, the cantilever or the sample is scanned, and the cantilever or the sample is moved in the vertical direction so that the displacement of the probe becomes constant. By converting the control signal at that time into an image, the surface state (state of unevenness) of the sample can be measured in atomic order.

  Such an AFM has an advantage that the surface state can be directly evaluated. However, the throughput at the time of data acquisition is low, the AFM apparatus itself is extremely expensive, and is not suitable for introduction to a mass production line.

  In addition, scanning tunneling microscopes (STMs) and Kelvin force microscopes (KFMs) are known as methods that can directly evaluate the surface state, both of which have the same problems as AFMs. There is.

  Therefore, there is a demand for a technique capable of sensitively measuring the surface state of a semiconductor layer in a short time with a relatively simple configuration.

  Cracks generated in the semiconductor layer greatly affect the surface crystallinity. Therefore, it is possible to indirectly evaluate the surface state of the semiconductor layer by measuring a parameter relatively sensitive to crystallinity among physical parameters of the semiconductor layer.

  Among the parameters sensitive to crystallinity, a parameter generally used is the half width of the X-ray diffraction pattern. This half width has the property of increasing as the crystallinity of the semiconductor layer decreases, and is relatively easy to measure, so it is well suited for crystallinity evaluation at the stage of bulk crystal before forming various semiconductor layers. It's being used.

  However, since the X-ray diffraction pattern reflects not only the crystal surface but also the internal state of the crystal, the change in half-value width depends greatly on the internal state change of the crystal and is very sensitive to the change in the crystal surface state. is not.

Japanese Patent Laid-Open No. 2-307046 JP-A-7-92236 JP 2003-224171 A

  As described above, the full width at half maximum of the X-ray diffraction pattern is a physical quantity reflecting not only the crystal surface but also the state change inside the crystal, and is not suitable for evaluating only the crystal surface state. In addition, when a plurality of semiconductor layers having different compositions are stacked on a substrate, the X-ray diffraction pattern reflects the crystal state of all the semiconductor layers and the crystal state of the substrate. It is difficult to distinguish only information.

  An object of the present invention is to provide a semiconductor layer inspection method capable of accurately measuring the surface state of a semiconductor layer in a short time.

The method for inspecting a semiconductor layer according to the present invention includes irradiating light to a semiconductor layer formed on a substrate;
Applying a predetermined frequency modulation to the semiconductor layer such that the physical properties of the semiconductor layer change;
Detecting reflected light from the semiconductor layer;
Extracting a modulation frequency component from the detected reflected light signal;
Measuring the reflection spectrum peculiar to excitons generated in the semiconductor layer by changing the wavelength of the irradiation light ;
And fitting in consideration of a broadening factor that quantitatively expresses the broadening of the reflection spectrum unique to the exciton .

  According to the present invention, a reflection spectrum peculiar to excitons generated in a semiconductor layer by light irradiation is measured. When there are few crystal defects such as cracks, the exciton spectrum obtained becomes steep, while when there are many crystal defects, the exciton spectrum becomes broad. Therefore, the surface state of the semiconductor layer can be evaluated by analyzing the broadening factor of the exciton spectrum.

  Therefore, since the present invention uses an optical spectrum analysis method, compared with the conventional AFM observation method and X-ray diffraction pattern method, the throughput at the time of data acquisition is high, the inspection apparatus itself has a relatively simple configuration, and mass production. Introduction to the line is easy.

Embodiment 1 FIG.
FIG. 1 is a flowchart showing an example of a surface inspection method according to the present invention. Here, as a typical test sample, as shown in FIG. 2, a nitride semiconductor having a HEMT epitaxial structure in which a buffer layer 2, an undoped GaN layer 3, and an undoped AlGaN layer 4 are sequentially formed on a substrate 1. An example using an element will be described.

  First, in step a1 of FIG. 1, the probe light is irradiated to the surface of the undoped AlGaN layer 4 located at the uppermost layer of the sample. Next, in step a2, an optical reflection spectrum peculiar to excitons generated in the undoped AlGaN layer 4 by light irradiation is measured.

  The reflection spectrum measurement method is: a) a method using monochromatic light as the probe light, and measuring the intensity change of the reflected light from the sample while changing the wavelength of the probe light, and b) a light having a continuous spectrum as the probe light. In addition, a method of measuring the reflection spectrum by splitting the reflected light from the sample can be employed.

The calibration of reflectance is performed by the following procedure. First, the reflected light intensity I _ref (λ) for the wavelength λ is measured using a reference sample having a known reflectance R _ref (λ), for example, a metal-coated mirror. Next, the reflected light intensity I _sample (λ) for the wavelength λ is measured for the inspection sample. Then, according to the definition of reflectivity = (intensity of light reflected on the sample surface) / (intensity of light incident on the sample surface), (intensity of light incident on the sample surface) is I _ref (λ) / R _ref It is given by (λ). Therefore, the reflectance R _sample (λ) of the inspection sample is expressed by the following equation (1).

Here, R _ref (λ) is known, and I _ref (λ) and I _sample (λ) are obtained by actual measurement. Therefore, the reflectance R _sample (λ) of the inspection sample is calculated by calculation using a computer or the like. Can be obtained.

Next, in step a3 of FIG. 1, the broadening factor of the reflectance spectrum R _sample (λ) of the inspection sample is analyzed.

  Here, an exciton is a kind of electron-hole pair generated in an insulating material or a semiconductor material. In particular, it is known that a nitride semiconductor is stable at room temperature. An exciton is a kind of elementary excitation in a solid, and a characteristic structure is observed in an optical spectrum in the vicinity of photon energy corresponding to the energy. If there is inhomogeneity on the sample surface, the spectrum unique to excitons will be broad and this spectral width can be quantified using a broadening factor.

If it is assumed that the non-uniformity of the sample follows a Gaussian distribution, the refractive index n _AlGaN (λ) of the undoped AlGaN layer 4 can be defined phenomenologically by the following equation (2).

This equation (2) calculates the refractive index function in an arbitrary surface state by superposing the refractive index function inherent to the material (here, equivalent to n _AlGaN (λ)) and the distribution function (here, Gaussian distribution). This is a model. In this calculation model, the standard deviation σ of the Gaussian distribution is a parameter corresponding to the broadening factor.

  Next, if the refractive index n (λ) of the sample is given, the reflectance spectrum R (λ) can be easily obtained using various reflection spectrum calculation models such as a transfer matrix method. As a simple example, when the undoped GaN layer 3 is sufficiently thick and only the interference effect in the undoped AlGaN layer 4 needs to be considered, the reflectance spectrum in the vicinity of the exciton energy in the undoped AlGaN layer 4 is expressed by the following formula ( 3).

Here, n _Air (= 1) and n _GaN are the refractive indexes of air and GaN, respectively. The physical quantity δ is a phase difference between two lights that cause interference, that is, light reflected by the uppermost surface of the undoped AlGaN layer 4 and light reflected by the AlGaN / GaN interface, and is expressed by the following equation (4). The

  Therefore, the broadening factor can be extracted by analyzing the reflection spectrum using the above equation (2) and the equations (3) and (4) known so far.

  Next, in step a4 in FIG. 1, the surface state of the undoped AlGaN layer 4 is evaluated based on the extracted broadening factor. When the calculation model of Expression (2) is used, the wider the exciton spectrum, the larger the standard deviation σ as a broadening factor, and it can be determined that the surface state of the undoped AlGaN layer 4 is more uneven. On the other hand, as the exciton spectrum becomes steeper, the standard deviation σ as a broadening factor becomes smaller, and it can be determined that the surface state of the undoped AlGaN layer 4 is more uniform.

  FIG. 3 is a graph illustrating an example of a reflectance spectrum of a nitride semiconductor. The vertical axis represents reflectance (%), and the horizontal axis represents wavelength (nm). A solid line indicates a sample without a surface crack, and a broken line indicates a sample with a surface crack.

  The center of the exciton spectrum is about 320 nm, and the broken-line spectrum appears wider with reference to the solid-line spectrum. The broadening factor of each spectrum can be numerically calculated by a computer or the like using the above-described equations (2) to (4). Therefore, it is possible to quantitatively evaluate the surface state of the sample using the obtained broadening factor.

  In the above description, the nitride semiconductor HEMT element is exemplified as the inspection sample. However, the present invention also applies to transistor elements, light emitting elements, light receiving elements and the like using other insulating materials and semiconductor materials, for example, Si, Ge, GaAs, and the like. The invention is applicable. In this case, the wavelength of light to be irradiated may be appropriately selected according to the exciton spectrum specific to the material.

In the present invention, the test sample is expressed by the composition formula Al _x In _y Ga _1-xy N (0 ≦ x ≦ 1, 0 ≦ y ≦ 1, x + y <1), such as GaN, AlN, and InN. It is preferable to use a nitride semiconductor which is a generic name for mixed crystals. This is because a nitride semiconductor has a large energy band gap and excitons exist stably at room temperature, so that the exciton spectrum becomes clear and the surface state can be evaluated with high accuracy.

  In this embodiment, the method of measuring the exciton spectrum of the sample with the reflection optical system has been described. However, when the sample is sufficiently thin compared to the penetration length of the probe light and the substrate is transparent to the probe light. It is also possible to measure the exciton spectrum using a transmission optical system.

Embodiment 2. FIG.
FIG. 4 is a block diagram showing an example of the surface inspection apparatus according to the present invention. The surface inspection apparatus includes a light source 11, a spectroscope 12, a sample stage 21, a detector 32, a computer 50, and the like.

  The light source 11 generates light having a continuous spectrum including a wavelength necessary for spectrum measurement. The spectroscope 12 divides the continuous spectrum from the light source 11 and outputs monochromatic probe light. The wavelength of the probe light can be continuously changed according to a control signal from the computer 50. The probe light from the spectroscope 12 is condensed to a desired spot diameter on the sample S by the condenser lens 13, and the measurement area of the sample S is defined.

  The sample stage 21 holds the sample S, and a three-dimensional position (XYZ direction) or a three-dimensional angle (pitch, yaw, roll) of the sample S can be adjusted by the adjustment mechanism 22. The adjustment mechanism 22 may be a manually operated type or a type operated in response to a control signal from the computer 50.

  The reflected light from the sample S is collected on the detector 32 by the condenser lens 31. The detector 32 has a function of converting the light intensity into an electrical signal. The current signal from the detector 32 is converted into a voltage signal by the subsequent current amplifier 33 and digitized by the voltmeter 34 in the subsequent stage. Converted to a signal.

  The computer 50 takes in the detection signal from the voltmeter 34, stores it in a memory or the like, and displays data on the display 51 as necessary.

  Such a surface inspection apparatus operates according to the flowchart shown in FIG. First, the sample S, for example, the nitride semiconductor element shown in FIG. Next, the computer 50 transmits a control signal (for example, a scan start wavelength, a scan end wavelength, a scan speed, etc.) to the spectroscope 12, and starts spectrum measurement.

  In the sample S, phenomena such as light absorption, reflection, and interference occur according to the wavelength of the probe light. In the present invention, attention is focused on a change in the optical reflection spectrum peculiar to excitons generated by light irradiation.

  The computer 50 captures the detection signal output during spectrum measurement, stores it in a memory or the like, and converts the detection signal into a reflectance spectrum using the above-described equations (3) and (4) after the measurement is completed. It displays on the display 51 as needed. Furthermore, the broadening factor of the exciton spectrum is extracted by applying the formula (2) and analyzing the reflectance spectrum of the sample S.

  The computer 50 stores in advance data relating to a standard sample whose relationship between the surface state (for example, crack density) and the broadening factor of the exciton spectrum is known. Therefore, the surface state of the sample S can be quantitatively evaluated by comparing the broadening factor of the sample S measured this time with the broadening factor of the standard sample as a reference.

  Subsequently, when changing the measurement area of the sample S, the adjustment mechanism 22 is operated to move the sample S, and the irradiation position of the probe light is changed. Then, spectrum measurement and analysis are performed according to the same procedure as described above. By repeating such in-plane scanning and spectrum measurement of the sample S, it becomes possible to evaluate the defect distribution on the entire surface of the sample S.

Embodiment 3 FIG.
FIG. 5 is a block diagram showing another example of the surface inspection apparatus according to the present invention. The surface inspection apparatus includes a light source 11, a sample stage 21, a spectroscope 35, a multichannel detector 36, a computer 50, and the like.

  The light source 11 generates a continuous spectrum probe light including a wavelength necessary for spectrum measurement. The probe light is condensed to a desired spot diameter on the sample S by the condensing lens 13, and the measurement area of the sample S is defined.

  The sample stage 21 holds the sample S, and a three-dimensional position (XYZ direction) or a three-dimensional angle (pitch, yaw, roll) of the sample S can be adjusted by the adjustment mechanism 22. The adjustment mechanism 22 may be a manually operated type or a type operated in response to a control signal from the computer 50.

  The reflected light from the sample S is collected in the spectroscope 35 by the condenser lens 31 and further input to the multichannel detector 36. The spectroscope 35 includes a diffraction grating, a prism, and the like, and has a function of spatially decomposing continuous spectrum light according to wavelength. The multi-channel detector 36 is configured by a linear array or the like in which a large number of light receiving surfaces are arranged in a straight line, and detects the intensity distribution of light spatially resolved by the spectroscope 35. Therefore, when the multi-channel detector 36 is used, wavelength scanning is not required, so that the speed of spectrum measurement can be increased.

  The controller 37 processes the output signal from the multichannel detector 36 and outputs it as a detection signal to the computer 50.

  The computer 50 takes in the detection signal from the controller 37, stores it in a memory or the like, and displays data on the display 51 as necessary.

  Such a surface inspection apparatus operates according to the flowchart shown in FIG. First, the sample S, for example, the nitride semiconductor element shown in FIG. Next, when the probe light from the light source 11 is irradiated toward the sample S, the multichannel detector 36 outputs a reflection spectrum from the sample S.

  The computer 50 takes in the detection signal from the multi-channel detector 36 and stores it in a memory or the like. After the measurement is completed, the computer 50 converts the detection signal into a reflectance spectrum using the above equations (3) and (4). It displays on the display 51 as needed. Furthermore, the broadening factor of the exciton spectrum is extracted by applying the formula (2) and analyzing the reflectance spectrum of the sample S.

  The computer 50 stores in advance data relating to a standard sample whose relationship between the surface state (for example, crack density) and the broadening factor of the exciton spectrum is known. Therefore, the surface state of the sample S can be quantitatively evaluated by comparing the broadening factor of the sample S measured this time with the broadening factor of the standard sample as a reference.

  Subsequently, when changing the measurement area of the sample S, the adjustment mechanism 22 is operated to move the sample S, and the irradiation position of the probe light is changed. Then, spectrum measurement and analysis are performed according to the same procedure as described above. By repeating such in-plane scanning and spectrum measurement of the sample S, it becomes possible to evaluate the defect distribution on the entire surface of the sample S.

Embodiment 4 FIG.
FIG. 6 is a configuration diagram illustrating an example of an optical path correction mechanism for reflected light. This optical path correction mechanism can be applied to the surface inspection apparatus shown in FIGS.

  The optical path detection device includes a split optical element 38, an optical position detector 39, a differential amplifier 40, and the like.

  The split optical element 38 is composed of a beam splitter or the like, and has a function of taking out part of the reflected light from the sample S. The optical position detector 39 is constituted by a PSD (position sensitive device) or a split type detector, and detects the position of the light extracted by the split optical element 38. Although an example using a two-divided detector in which two light receiving surfaces A and B are arranged is shown here, a four-divided detector in which four light receiving surfaces are arranged in the XY direction may be used.

  The differential amplifier 40 is composed of an operational amplifier or the like, and differentially amplifies the two signals SA and SB output from the optical position detector 39 and outputs a differential signal of SA-SB.

  By using such an optical path detection device, it is possible to detect the optical path deviation of the reflected light with high accuracy while hardly affecting the reflected light from the sample S. Moreover, it is preferable to set the optical path length from the sample S to the optical position detector 39 via the split optical element 38 as long as possible, and the sensitivity to the optical path shift of the reflected light can be increased.

  The optical path correction mechanism adjusts the three-dimensional position (XYZ direction) or the three-dimensional angle (pitch, yaw, roll) position of the sample stage 21 based on the optical path detection device described above and the difference signal corresponding to the optical path deviation. The adjustment mechanism 22 is configured.

  Next, the feedback operation of the optical path correction mechanism will be described. When a wafer on which a semiconductor layer is formed is used as the inspection sample S, it is necessary to consider the flatness of the wafer. In particular, when the wafer is warped, the optical axis of the reflected light may fluctuate when measuring the in-plane distribution data. For example, as shown by the one-dot chain line in FIG. 6, the original optical path of the reflected light is shifted to the optical path shown by the solid line in FIG.

  In particular, in the case of multichannel photometry shown in FIG. 5, if the optical axis of the reflected light fluctuates, the incident angle to the spectroscope 35 changes and the light intensity distribution in the light receiving array of the multichannel detector 36 shifts. A wavelength shift of the optical spectrum occurs.

  As a countermeasure, during the in-plane scanning of the sample S, the optical path detector detects the optical path deviation of the reflected light from the sample S, while the adjustment mechanism 22 eliminates the optical path deviation or By adjusting the angle, measurement errors due to wafer warpage can be reduced or eliminated.

  FIG. 7 is a flowchart showing the operation of the optical path correction mechanism. First, in step b1, a part of the reflected light from the sample S is divided by the dividing optical element 38. Next, in step b2, the optical position detector 39 receives the divided light and detects the position of the light. Next, in step b3, it is determined whether or not the differential signal SA-SB from the differential amplifier 40 is zero. If the difference signal is not zero, the process proceeds to step b4, the adjustment mechanism 22 adjusts the position or angle of the sample stage 21 in the direction in which the difference signal becomes zero, and then returns to step b2 to again detect the optical position and the sample. Repeat the stand adjustment. When the difference signal converges to zero by such a feedback operation, the process proceeds from step b3 to step b5 and spectrum measurement of the sample S is started.

  Subsequently, when changing the measurement area of the sample S, the adjustment mechanism 22 is operated to move the sample S in the plane, and the irradiation position of the probe light is changed. Then, after performing the optical path correcting operation shown in FIG. 7 to eliminate the optical path deviation of the reflected light, spectrum measurement and analysis are performed according to the same procedure as described above. By sequentially repeating the in-plane scanning of the sample S, the optical path correcting operation, and the spectrum measurement, it is possible to evaluate the defect distribution on the entire surface of the sample S.

  FIGS. 8A and 8B are photographs respectively showing images obtained by observing the AlGaN / GaN HEMT epitaxial structure shown in FIG. 2 with an AFM. The gray scale indicating the shading of the image corresponds to the height of the unevenness on the surface. The time required to acquire these AFM images was 30 minutes each.

  The sample A shown in FIG. 8A has a flat surface, no cracks, and can be used as a standard sample. Sample B shown in FIG. 8B has irregularities on the surface, and a large number of cracks exist.

  FIG. 9 is a graph showing optical reflection spectra of samples A and B. The vertical axis represents reflectance (%), and the horizontal axis represents photon energy (eV). The solid line shows Sample A without surface cracks, and the broken line shows Sample B with surface cracks. These spectra were measured using the surface inspection apparatus shown in FIG. The time required for measurement of each spectrum was 3 minutes corresponding to 1/10 of AFM observation.

  The center of the exciton spectrum is about 3.87 eV (≈320 nm), and the spectrum of the sample B where the crack exists (solid line) shows stronger broadening than the spectrum of the sample A without crack (broken line). You can see that it shows.

  10A and 10B are graphs showing the results of fitting the optical reflection spectra of the samples A and B shown in FIG. A solid line shows the experimental value of the reflection spectrum shown in FIG. 9, and a white circle is a calculated value fitted by a computer or the like using the above-described equations (2) to (4).

  Both the fitting curve of sample A shown in FIG. 10A and the fitting curve of sample B shown in FIG. 10B agree very well with the experimental values in the vicinity of the exciton spectrum center. As a result of calculating the broadening factor for these fitting curves, Sample A was 25 meV and Sample B was 45 meV. Therefore, it can be seen that the broadening factor increases as the surface crack density increases.

  In this way, when the surface state of the sample is inspected, by calculating the broadening factor of the exciton spectrum, quantitative evaluation can be performed in a shorter time and with higher accuracy than in the conventional AFM observation method.

Embodiment 5 FIG.
In the above, the quantitative evaluation method of the semiconductor surface based on the reflection spectroscopy was described. This is a method of quantifying the surface state based on the broadening factor of the reflection spectrum. In other words, the surface state is quantified using the sharpness of the spectrum shape as an index. The sharpness of the spectral shape corresponds to the intensity of the differential signal. Therefore, if the differential signal of the reflection spectrum, that is, the modulated reflection spectrum is measured, it is considered that the surface state can be quantified using the intensity of the obtained spectrum as an index. In general, the differential signal is highly sensitive to changes. Therefore, it is possible to extract information on the surface state that cannot be obtained with a normal reflection spectrum. In the following, a quantitative evaluation method for a semiconductor surface, which is based on modulated reflection spectroscopy, will be described.

  There are various types of modulated reflection spectroscopy. Among these, PR (Photo reflectance) spectroscopy (light modulation reflection spectroscopy) is the most simple technique that is non-destructive and non-contact. A feature of PR spectroscopy is that the sample surface is irradiated with excitation light in addition to probe light for detecting reflectance. The carriers generated by the excitation light slightly change the internal electric field of the sample due to the shielding effect. In general, the optical constant of the sample also changes as the internal electric field changes. PR spectroscopy is a spectroscopic technique that measures a minute change in optical constant as a modulation reflectance. The detected PR signal can be roughly classified into two types: FK (Franz-Keldysh) vibration and third-order differential shape signal under the condition that there is only an internal electric field that does not change the band structure.

  Here, when e is an elementary charge, h is a Planck constant, F is an internal electric field strength, and μ is an electron-hole converted mass, an electro-optic constant represented by the following formula (5) and a lifetime broadening factor Γ When two indices are used, the FK vibration and the third-order differential shape signal are signals observed in the cases of the equations (7) and (8), respectively.

  A schematic diagram of a spectroscopic device for measuring the PR signal is shown in FIGS. The PR spectroscopic system takes the form of an apparatus in which an excitation light source, a modulator, and a lock-in amplifier for measuring a modulated reflected signal are added to the reflection spectrum measuring apparatus shown in FIG.

  The probe light optical system includes a white light source 111, a condenser lens 112, a spectroscope 113, a condenser lens 114, and the like. The white light source 111 is composed of, for example, a lamp, and generates continuous spectrum light including a wavelength necessary for spectrum measurement. The spectroscope 113 divides the continuous spectrum from the white light source 111 and outputs monochromatic probe light. The wavelength of the probe light can be continuously changed according to a control signal from the computer 140. The probe light from the spectroscope 112 is condensed to a desired spot diameter on the sample S by the condenser lens 114, and the measurement area of the sample S is defined.

  The excitation light optical system includes an excitation light source 121, an excitation light stabilizer 122, an excitation light filter 123, a modulator 124, a condenser lens 125, and the like. The excitation light source 121 is composed of a laser light source, for example, and generates excitation light having a wavelength shorter than the band gap wavelength of the sample S so that free carrier absorption occurs in the sample S. The excitation light stabilizer 122 stabilizes the excitation light power from the excitation light source 121. The excitation light filter 123 is a band pass filter that passes only the excitation light wavelength and cuts noise light. The modulator 124 modulates the excitation light intensity based on a predetermined reference frequency signal. The light that has passed through the modulator 124 is condensed to a desired spot diameter on the sample S by the condenser lens 125 so as to substantially coincide with the irradiation area of the probe light.

  The detection optical system includes a condenser lens 131, an optical filter 132, a signal detector 133, and the like. The reflected light from the sample S is collected by the signal detector 133 by the condenser lens 131. The intermediate optical filter 132 is a long pass filter that shields the excitation light and allows only the probe light to pass, and is removable as necessary.

  As shown in FIG. 12, the signal processing system includes a current / voltage converter 134, a bandpass filter circuit 135, a DC voltmeter 136, a lock-in amplifier 137, a computer 140, a display 141, and the like. The signal detector 133 has a function of converting the light intensity into an electric signal, and the current signal from the detector 32 is converted into a voltage signal by the subsequent current / voltage converter 134. The band pass filter circuit 135 extracts a DC component corresponding to the reflectance R and an AC component corresponding to the modulation reflectance ΔR from the detection signal. The DC voltmeter 136 measures the voltage of the DC component of the detection signal, converts it to a digital signal, and outputs it to the computer 140. The lock-in amplifier 137 uses the reference frequency from the modulator 124 to extract a frequency component that matches the reference frequency from the AC component of the detection signal, converts it to a digital signal, and outputs the digital signal to the computer 140. The computer 140 takes in the reflectance signal R from the DC voltmeter 136 and the modulated reflectance signal ΔR from the lock-in amplifier 137, stores it in a memory or the like, and displays data on the display 141 as necessary.

  The PR spectrum is measured according to the flowchart shown in FIG. Hereinafter, the PR spectroscopic measurement procedure will be described in detail according to the flowchart shown in FIG.

  The probe light for detecting the reflectance of the sample S is obtained by introducing light from the white light source 111 into the spectroscope 113 and making it monochromatic. The monochromatic probe light is applied to the sample S through the condenser lens 114 (step c1). On the other hand, the excitation light is converted into excitation light whose intensity periodically changes over time by the modulator 124, and then irradiated to the sample S (step c2).

  The lock-in amplifier 137 that detects the PR signal requires phase adjustment. In order to perform phase adjustment, the optical filter 132 in front of the detector is removed from the optical axis so that the excitation light scattered on the sample S enters the signal detector 133. The phase of the lock-in amplifier 137 is adjusted so that the signal detected in this state is synchronized with the reference frequency signal sent from the modulator 124 (step c3). After the phase adjustment, the scattered excitation light is blocked by inserting the optical filter 132 on the optical axis so that only the probe light enters the signal detector 133 (step c4).

  After the above settings, move to data measurement. First, the spectroscope 113 is swept, and the probe light reflected from the sample S is converted into an electric signal by the signal detector 133 (step c5). The obtained signal is divided into a direct current component corresponding to the reflectance R and an alternating current component corresponding to the modulation reflectance ΔR through the band pass filter circuit 135 (step c6), and measured by the direct current voltmeter 136 and the lock-in amplifier 137, respectively. (Step c7). The computer 140 calculates ΔR / R using each measurement value, and obtains a PR spectrum by plotting it as a function of photon energy or wavelength (step c8).

  FIG. 14 shows a flowchart for obtaining the defect density from the PR spectrum thus obtained. First, after measuring the PR spectrum as described above (step d1), the amplitude of the third-order differential shape signal or FK vibration caused by the layer of interest is read from the obtained spectrum (step d2). The FK vibration generally indicates a shape composed of a plurality of vibrations. At the time of analysis, the accuracy of the value obtained from the analysis can be ensured by selecting the vibration with the best signal-to-noise ratio, usually the vibration with the largest amplitude. The vibration with the maximum amplitude is vibration near the band gap energy of the layer of interest. Therefore, it is optimal to read the amplitude from this vibration. Hereinafter, the amplitude of the FK vibration and the third-order differential shape signal will be simply referred to as PR signal intensity unless otherwise specified.

  Finally, the procedure proceeds to a procedure for obtaining the defect density from the obtained PR signal intensity (step d3). In the spectrum obtained as described above, the amplitude of the FK vibration caused by the target layer changes depending on the state (crystallinity, etc.) of the target layer. Therefore, PR measurement is first performed on a standard sample whose defect density is known in advance, and a conversion curve (FIG. 15) is prepared with the horizontal axis representing the defect density and the vertical axis representing the PR signal intensity. Based on this conversion curve, the defect density can be obtained by comparing the PR signal intensity of the sample to be evaluated with a calibration curve.

Embodiment 6 FIG.
The PR spectroscopy described above is the most widely used among the modulated reflection spectroscopy. In practice, however, there are cases where PR spectroscopy cannot be applied. For example, there is no light source that can excite the sample. In addition, when the sample exhibits strong light emission characteristics by excitation light irradiation, the original PR spectrum cannot be obtained because the light emission component works as a disturbance component. There is CER (Contactless electroreflectance) spectroscopy (non-contact electric field modulation reflection spectroscopy) as a modulation reflection spectroscopy applicable in such a case. A schematic diagram of the CER spectrometer is shown in FIG. Note that the signal processing system is not shown because the configuration shown in FIG. 12 can be used as it is.

  The probe light optical system includes a white light source 211, a condenser lens 212, a spectroscope 213, a condenser lens 214, and the like. The white light source 211 is composed of, for example, a lamp, and generates continuous spectrum light including a wavelength necessary for spectrum measurement. The spectroscope 213 splits the continuous spectrum from the white light source 211 and outputs monochromatic probe light. The wavelength of the probe light can be continuously changed in accordance with a control signal from the computer 140 shown in FIG. The probe light from the spectroscope 112 is condensed to a desired spot diameter on the sample S by the condenser lens 214, and the measurement area of the sample S is defined.

  The electric field application circuit includes an AC power source 221 that generates an AC voltage having a predetermined reference frequency, a pair of electrodes 222 and 223 arranged so as to sandwich the sample S, and the like. The electrode 222 arranged on the light incident side of the sample S is generally called a transparent electrode, is formed of an electrode material that can transmit probe light, and is electrically connected to one terminal of the AC power source 221. The other terminal of the AC power source 221 is grounded. The electrode 223 arranged on the back surface of the sample S is also grounded through the sample table.

  The detection optical system includes a condenser lens 231 and a signal detector 232. The reflected light from the sample S is collected by the signal detector 232 by the condenser lens 231. The detection signal from the signal detector 232 is sent to the signal processing system shown in FIG.

  When an AC electric field is applied between the two electrodes 222 and 223 sandwiching the sample S at the time of CER spectroscopy measurement, electric field modulation is caused inside the sample S. The optical reflectance modulation caused by this electric field modulation is detected via the probe light. Since the obtained modulated reflectance is equivalent to the PR signal, the CER spectrum analysis / defect density calculation flow is the same as the flowchart of FIG. 14 and is shown in FIG.

  First, regarding the CER spectroscopic measurement procedure, the probe light for detecting the reflectance of the sample S is obtained by introducing the light from the white light source 211 into the spectroscope 213 and making it monochromatic. The monochromatic probe light is applied to the sample S through the condenser lens 214. On the other hand, an AC electric field is applied to the sample S via the pair of electrodes 222 and 223, and the reflected light from the sample S is modulated. In this state, the phase of the lock-in amplifier 137 is adjusted so that the detection signal from the signal detector 232 is synchronized with the reference frequency signal transmitted from the AC power source 221.

  After the above settings, move to data measurement. First, the spectroscope 213 is swept, and the probe light reflected from the sample S is converted into an electric signal by the signal detector 232. The obtained signal is divided into a direct current component corresponding to the reflectance R and an alternating current component corresponding to the modulation reflectance ΔR through the band-pass filter circuit 135 and measured by the direct current voltmeter 136 and the lock-in amplifier 137, respectively. The computer 140 calculates ΔR / R using each measured value, and obtains a CER spectrum by plotting it as a function of photon energy or wavelength.

  Next, after measuring the CER spectrum as described above (step e1), the amplitude of the third-order differential shape signal or FK vibration caused by the layer of interest is read from the obtained spectrum (step e2). The FK vibration generally indicates a shape composed of a plurality of vibrations. At the time of analysis, the accuracy of the value obtained from the analysis can be ensured by selecting the vibration with the best signal-to-noise ratio, usually the vibration with the largest amplitude. The vibration with the maximum amplitude is vibration near the band gap energy of the layer of interest. Therefore, it is optimal to read the amplitude from this vibration. Hereinafter, the amplitude of the FK vibration and the third-order differential shape signal will be simply referred to as CER signal strength unless otherwise specified.

  Finally, the procedure proceeds to a procedure for obtaining the defect density from the obtained CER signal intensity (step e3). In the spectrum obtained as described above, the amplitude of the FK vibration caused by the target layer changes depending on the state (crystallinity, etc.) of the target layer. Therefore, CER measurement is first performed on a standard sample whose defect density is known in advance, and a conversion curve is prepared in which the horizontal axis indicates the defect density and the vertical axis indicates the CER signal intensity. Based on this conversion curve, the defect density can be determined by comparing the CER signal intensity of the sample to be evaluated with the calibration curve.

Embodiment 7 FIG.
In the CER spectroscopy, an electrode transparent to probe light used for measurement is essential. However, there are cases where a transparent electrode cannot be obtained in the ultraviolet region. ITO (In ₂ O ₃ : Sn), which is one of the most popular In ₂ O ₃ based transparent conductive films as practical transparent conductive films for electronic devices, has a base crystal band gap energy of 3.75 eV at room temperature. And has transparency only in the visible light region. Similarly, in the case of a ZnO-based and SnO ₂ -based transparent conductive film, the band gap energy of each parent crystal is 3.44 eV and 3.70 eV. Therefore, CER spectroscopy cannot be applied to a multilayer structure composed of a material having a larger band gap energy than the transparent electrode.

  Even if the sample is not applicable to CER spectroscopy and PR spectroscopy, if they are composed of piezoelectric material, stress / electric field modulation reflectance spectroscopy called PZR (piezoreflectance) spectroscopy (piezo modulation reflectance spectroscopy) By applying one of the above, the modulated reflection spectrum can be measured.

  A schematic diagram of the PZR spectrometer is shown in FIG. Note that the signal processing system is not shown because the configuration shown in FIG. 12 can be used as it is.

  The probe light optical system includes a white light source 311, a condensing lens 312, a spectroscope 313, a condensing lens 314, and the like. The white light source 311 is composed of, for example, a lamp and generates continuous spectrum light including wavelengths necessary for spectrum measurement. The spectroscope 313 splits the continuous spectrum from the white light source 311 and outputs monochromatic probe light. The wavelength of the probe light can be continuously changed in accordance with a control signal from the computer 140 shown in FIG. The probe light from the spectroscope 312 is condensed to a desired spot diameter on the sample S by the condenser lens 314, and the measurement area of the sample S is defined.

  The stress application circuit includes an AC power source 321 that generates an AC voltage having a predetermined reference frequency, a piezoelectric element 322 that is driven by the AC voltage, and the like.

  The detection optical system includes a condenser lens 331, a signal detector 332, and the like. The reflected light from the sample S is collected on the signal detector 332 by the condenser lens 331. The detection signal from the signal detector 332 is sent to the signal processing system shown in FIG.

  In PZR spectroscopy, stress is generated by the piezoelectric element 322 attached to the back surface of the sample S, and periodic internal strain is induced in the sample S. In the case of a piezoelectric sample, a piezoelectric field is generated in response to this periodic stress, and internal electric field modulation occurs. The optical reflectance modulation caused by the electric field modulation is detected via the probe light as in the case of the modulated reflection spectroscopy. As in the case of CER spectroscopy, the obtained modulated reflectance is equivalent to the PR signal, and therefore, the PZR spectrum analysis / defect density calculation flow is the same as the flowchart of FIG. 14 and is shown in FIG.

  First, regarding the PZR spectroscopic measurement procedure, the probe light for detecting the reflectance of the sample S is obtained by introducing light from the white light source 311 into the spectroscope 313 and making it monochromatic. The monochromatic probe light is applied to the sample S through the condenser lens 314. On the other hand, periodic stress is applied to the sample S by the piezoelectric element 322, and the reflected light from the sample S is modulated. In this state, the phase of the lock-in amplifier 137 is adjusted so that the detection signal from the signal detector 332 is synchronized with the reference frequency signal sent from the AC power supply 321.

  After the above settings, move to data measurement. First, the spectroscope 313 is swept, and the probe light reflected from the sample S is converted into an electric signal by the signal detector 332. The obtained signal is divided into a direct current component corresponding to the reflectance R and an alternating current component corresponding to the modulation reflectance ΔR through the band-pass filter circuit 135 and measured by the direct current voltmeter 136 and the lock-in amplifier 137, respectively. The computer 140 calculates ΔR / R using each measurement value, and obtains a PZR spectrum by plotting it as a function of photon energy or wavelength.

  Next, after measuring the PZR spectrum as described above (step f1), the amplitude of the third-order differential shape signal or FK vibration caused by the layer of interest is read from the obtained spectrum (step f2). The FK vibration generally indicates a shape composed of a plurality of vibrations. At the time of analysis, the accuracy of the value obtained from the analysis can be ensured by selecting the vibration with the best signal-to-noise ratio, usually the vibration with the largest amplitude. The vibration with the maximum amplitude is vibration near the band gap energy of the layer of interest. Therefore, it is optimal to read the amplitude from this vibration. Hereinafter, the FK vibration and the amplitude of the third-order differential shape signal will be simply referred to as PZR signal intensity unless otherwise specified.

  Finally, the procedure proceeds to a procedure for obtaining the defect density from the obtained PZR signal intensity (step f3). In the spectrum obtained as described above, the amplitude of the FK vibration caused by the target layer changes depending on the state (crystallinity, etc.) of the target layer. Therefore, PZR measurement is first performed on a standard sample whose defect density is known in advance, and a conversion curve is prepared in which the horizontal axis indicates the defect density and the vertical axis indicates the PZR signal intensity. Based on this conversion curve, the defect density can be obtained by comparing the PZR signal intensity and the calibration curve of the sample to be evaluated.

Next, the applicability of the modulated reflection spectroscopy described above to surface morphology analysis will be discussed. Here, an example of defect density estimation using PR spectroscopy will be shown. 20A to 20C show samples A, B, and C each having an Al _0.2 Ga _0.8 N / GaN heterostructure grown on three types of sapphire substrates that were evaluated. A surface AFM image is shown. The thicknesses of the AlGaN layer and the GaN layer are the same for all samples.

  The AFM image of sample A shown in FIG. 20A shows that many cracks are present on the AlGaN surface. Such cracks are hardly observed in the sample B shown in FIG. 20B and the C shown in FIG. However, a large number of pits are detected in the AFM image of the sample B. Generally, this pit is accompanied by threading dislocations, and it is considered that impurities gathered around the threading dislocations are triggered when the pits are formed.

  FIG. 21 shows PR spectra of samples A, B, and C measured at room temperature. The vertical axis represents ΔR / R obtained by dividing the modulation reflectance ΔR by the reflectance R, and the horizontal axis represents photon energy. The characteristic spectral structure appearing at the photon energy of 3.4 eV is attributed to the signal originating from the GaN layer because its position is almost equal to the band gap energy of GaN. Therefore, the vibration structure that continues from the photon energy of 3.8 eV is attributed to the FK vibration of the AlGaN layer.

  The amplitude of the FK vibration caused by the AlGaN layer decreases in the order of samples C, B, and A, which corresponds to the deterioration of the surface morphology shown in FIG. Therefore, it can be concluded that amplitude analysis of FK vibration can be applied as a means of quantifying surface morphology.

  In the surface morphology analysis by reflection spectroscopy described above, the significant difference with respect to the presence or absence of pits on the surface of the AlGaN layer was small, but surface morphology analysis using PR spectroscopy was more sensitive than normal reflection spectrum measurement. It can be seen that this is a simple evaluation method. If a PR curve measurement is performed on a plurality of standard samples whose defect density is known in advance, and a conversion curve obtained by plotting the PR signal intensity against the defect density is prepared, the PR signal of the sample to be evaluated Obviously, the defect density can be estimated from the strength.

  Finally, it will be explained that the optical reflection and PR spectrum measurements described herein are more sensitive to surface morphology compared to previously used crystal characterization techniques.

FIG. 22 shows ω-2θ X-ray diffraction spectra of samples A, B, and C near the Bragg angle of GaN (0004) reflection. The vertical axis represents the X-ray diffraction intensity (logarithmic display, arbitrary unit), and the horizontal axis represents the diffraction angle 2θ (arcsec). The position of the broken line corresponds to the Bragg angle of Al _0.2 Ga _0.8 N (0004) reflection obtained by calculation. In the calculation of the Bragg angle of Al _0.2 Ga _0.8 N (0004) reflection, the Al _0.2 Ga _0.8 N layer grows in a pseudomorphic manner (pseudomorphic) with respect to the GaN layer. Assumed.

From the comparison between the position of the broken line and the X-ray diffraction pattern, the peak structure near 3100 (arcsec) is attributed to Al _0.2 Ga _0.8 N (0004) reflection. The peak position of Al _0.2 Ga _0.8 N (0004) reflection varies slightly depending on the sample. This indicates that the AlGaN layer composition ratio of each sample is slightly shifted from each other. Except for this difference in peak position, the Al _0.2 Ga _0.8 N (0004) reflection X-ray diffraction pattern differs from the reflection and PR spectrum in that its shape is almost independent of the AlGaN surface morphology. From this, it can be concluded that the reflection spectrum measurement and the PR spectrum measurement are more suitable for the evaluation of the surface morphology.

It is a flowchart which shows an example of the surface inspection method which concerns on this invention. It is sectional drawing which shows a typical test | inspection sample. It is a graph which shows an example of the reflectance spectrum of a nitride semiconductor. It is a block diagram which shows an example of the surface inspection apparatus which concerns on this invention. It is a block diagram which shows the other example of the surface inspection apparatus which concerns on this invention. It is a block diagram which shows an example of the optical path correction mechanism of reflected light. It is a flowchart which shows operation | movement of an optical path correction mechanism. FIGS. 8A and 8B are photographs respectively showing images obtained by observing the AlGaN / GaN HEMT epitaxial structure shown in FIG. 2 with an AFM. It is a graph which shows the optical reflection spectrum of samples A and B. 10A and 10B are graphs showing the results of fitting the optical reflection spectra of the samples A and B shown in FIG. It is a block diagram which shows an example of PR spectroscopy apparatus. It is a block diagram which shows an example of PR spectroscopy apparatus. It is a flowchart which shows an example of a PR spectroscopy measurement procedure. It is a flowchart which shows an example of the procedure which calculates | requires defect density from PR spectrum. It is a conversion curve which shows the relationship between a defect density and PR signal strength. It is a block diagram which shows an example of a CER spectrometer. It is a flowchart which shows an example of the procedure which calculates | requires defect density from a CER spectrum. It is a block diagram which shows an example of a PZR spectroscopy apparatus. It is a flowchart which shows an example of the procedure which calculates | requires defect density from a PZR spectrum. It is the surface AFM image which observed the samples A, B, and C which have the AlGaN / GaN heterostructure each grown on three types of sapphire substrates. It is PR spectrum of sample A, B, C shown in Drawing 20 (a)-Drawing 20 (c). 21 is an X-ray diffraction spectrum of ω-2θ of samples A, B, and C shown in FIGS. 20 (a) to 20 (c).

11 light source, 12 spectrometer, 13 condenser lens, 21 sample stage,
22 adjustment mechanism, 31 condenser lens, 32 detector, 33 current amplifier,
34 Voltmeter, 35 Spectrometer, 36 Multi-channel detector,
37 controller, 38 splitting optical element, 39 optical position detector,
40 differential amplifiers, 50 computers, 51 displays,
111, 211, 311 white light source, 112, 114, 125, 131, 212, 214, 231, 312, 314, 331 condenser lens,
113, 213, 313 Spectrometer, 121 excitation light source, 122 excitation light stabilizer,
123 excitation light filter, 124 modulator, 132 optical filter,
133, 232, 332 signal detector, 134 current / voltage converter,
135 Band pass filter circuit, 136 DC voltmeter,
137 lock-in amplifier, 140 computer, 141 display,
221, 321 AC power source, 222, 223 electrode, 322 piezoelectric element.

Claims (4)

Irradiating the semiconductor layer formed on the substrate with light;
Applying a predetermined frequency modulation to the semiconductor layer such that the physical properties of the semiconductor layer change;
Detecting reflected light from the semiconductor layer;
Extracting a modulation frequency component from the detected reflected light signal;
Measuring the reflection spectrum peculiar to excitons generated in the semiconductor layer by changing the wavelength of the irradiation light ;
And a step of fitting in consideration of a broadening factor that quantitatively expresses the spread of the reflection spectrum peculiar to the exciton, and a method for inspecting a semiconductor layer.   2. The semiconductor layer inspection method according to claim 1, wherein in the modulation applying step, the semiconductor layer is irradiated with excitation light modulated at a predetermined frequency.   2. The semiconductor layer inspection method according to claim 1, wherein, in the modulation applying step, an electric field modulated at a predetermined frequency is applied to the semiconductor layer.   2. The semiconductor layer inspection method according to claim 1, wherein in the modulation applying step, a stress modulated at a predetermined frequency is applied to the semiconductor layer.

Images