JP2003249186A - Observation method and observation equipment by scanning transmission electron microscope - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To realize a computing means which can optimize an amount of focal gap in STEM and which can erase a virtual image and receive a feedback of a condition by substantially shortening the period of time to obtain a computed image and by comparing an image of an electron microscope to the computed image right after observation in an observation method and an observation equipment by a scanning transmission electron microscope. <P>SOLUTION: The observation is performed by a scanning transmission electron microscope observation equipment equipped with a field emission electron gun, a convergence means to converge an electron ray emitted from the field emission electron gun onto a sample, a deflecting means to scan a convergence electron ray irradiating the sample, a projecting means to project a sample image, and a measuring means for a sample thickness. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a scanning transmission type of an atomic structure image or an element composition distribution image suitable for identifying a crystal structure of a minute region or a sample constituent element in a bulk material, a thin film material, an interface structure and the like. The present invention relates to an observation method and an observation device using an electron microscope.

[0002]

2. Description of the Related Art Needless to say, not only semiconductor devices but also magnetic heads and magnetic media used in magnetic recording devices have been miniaturized year by year, and the characteristics at the material interface, for example, have come to greatly affect the characteristics. Structural control at the atomic level is required. To realize this, it is necessary to identify the crystal structure and composition at the atomic level, but currently there is no means for that.

A high resolution transmission electron microscope, which has been widely used in the past,
ision electronon microsco
pe: HR-TEM) is used for observation at the atomic level when developing semiconductor devices and materials, and is useful for obtaining many findings.

However, it is impossible to discriminate only by observing the atomic species of each atom by HR-TEM.

Regarding the electron microscope image taken,
By using a transmission electron microscope image calculation method based on the dynamic diffraction theory based on sample thickness, crystal structure, atomic physical property parameters, etc., a calculation image with different imaging conditions is constructed to identify the sample constituent atomic species and crystal structure. Although it has been done, various images appear in the calculated image due to minute parameter changes, and since the calculation is performed under the artificially set conditions,
It is difficult to judge the exact structure.

Further, in the experiment, when an image is formed using elastically scattered electrons, the interference fringes of the electron wave reflecting the crystal structure are observed, so that an accurate atomic image cannot be obtained.
Moreover, the atomic species cannot be identified from the contrast intensity of the atoms.

Further, regarding the simulation,
Usually, the multi-slice method is applied, but in this method, the experimental parameters and the simulation calculation parameters do not match, and accurate evaluation cannot be performed.

On the other hand, a scanning transmission electron microscope (sc
anning transmission elect
dark field STEM image observation, which is one of the observation methods based on ron microscopy (STEM), includes observation of images using elastically scattered electrons and thermal diffuse scattering (th
armal diffuse scattering:
Observation of an image obtained by capturing electrons, which are inelastically scattered at a high angle (~ 50 [mrad]) by TDS, using a ring type detector (high angle a).
Both of the nulular dark fields (HAADF) can be performed.

When TDS is used, an image is formed by electrons directly scattered by atoms, and the scattered electrons have accurate positional information of atomic species and atoms. It has been used as a means of grasping the position.

The means is also called the Z-contrast method because the scattering intensity depends on the atomic number Z. An atom having a larger atomic number Z has a stronger scattering intensity on the higher angle side and a smaller amount has a smaller scattering intensity. The strength is weak.

However, although the dark field STEM observation method has been researched for a long time, there is no unified view on the interpretation of the experimental results and the simulation results, and the experimental results are reproducible. Was low and no quantitative interpretation was possible.

In addition, in the dark field STEM observation method, it has not been possible to identify the exact atomic species and atomic positions by defining the experimental conditions and performing the simulation calculation using the experimental parameters. It was

In the dark field STEM observation method, it has been considered so far that the observed image directly reflects the atomic position and atomic species. It has become clear that contrast appears as a virtual image.

In order to solve the above problems, it is necessary to have a calculating means for optimizing the amount of defocus in the STEM and eliminating the virtual image.

Also in the simulation method,
By relying on the conventional multi-slice method, the calculation time is long, and it is impossible to compare the calculated image immediately after observation and feed back the condition immediately, but this must also be eliminated.

[0016]

In the present invention, the STEM is used.
We have realized a calculation method that enables optimization of the amount of defocus in the image and elimination of virtual images, and the time for obtaining the calculated image is greatly shortened, and the electron microscope image is compared with the calculated image immediately after observation. So that the conditions can be fed back.

[0017]

In the dark-field STEM observation method, the electron beam incident conditions, the sample thickness, and the scattered electron uptake angle into the annular detector greatly depend on the observation result. It is necessary to accurately measure the parameters and grasp the experimental conditions, and also in the simulation calculation, it is necessary to carry out using the same parameters as the experimental results.

In the present invention, all the parameters necessary for the simulation are actually measured by an experiment, and the experimental result and the simulation calculation result are compared to obtain a more accurate elemental composition at the atomic level. Makes it possible.

Usually, in the simulation calculation, the intensity distribution is calculated by applying the multi-slice method. However, when the dark field STEM image simulation is calculated by the multi-slice method, the sample thickness is calculated for each point of the image pixel. It is necessary to calculate the wave field in the vertical direction, and it is necessary to perform Fourier transform many times during that time, which requires a huge amount of calculation time. It is almost impossible to feed back to and obtain an element distribution image.

In the present invention, by applying the Bethe method, it is possible to omit the Fourier transform for each image pixel, and it is possible to significantly reduce the calculation time. By adopting the means, it has become possible to obtain the element distribution image by comparing the experimental result and the simulation image for it in real time.

The experimental image and the simulation image are formed by the contributions of elastically scattered electrons due to Bragg diffraction and inelastically scattered electrons due to TDS, and the intensities are the sum of the two, and in the experiment, Since it is impossible to obtain a pure Z contrast image based on TDS, it is necessary to consider the effects of both in the simulation.

Further, from the incident electron beam intensity distribution depending on the lens condition, a virtual image appears in the captured image, or the intensity is added to the wrong position of the atom, and the problem that an accurate composition distribution cannot be obtained is pointed out. Has been done.

For this, the electron beam intensity distribution can be predicted from the experimental conditions, and by deconvoluting the dark field STEM image, the virtual image and the erroneous intensity distribution can be removed to obtain an accurate observation result.

FIG. 1 is a diagram for explaining the principle of dark field STEM observation. In the figure, 1 is a TEM sample, 2 is an annular detector, and 3 is an EELS (electron electron).
rgy loss spectroscopy detector, α is a half-angle of convergence [m
rad] and t are the thickness [nm] of the sample 1, β ₁ and β ₂ are scattered electron uptake angles [mrad], and β is the electron uptake angle [mrad] in the EELS detector 3.

The dark-field STEM observation method is a method in which a finely converged electron beam is incident on a sample and scanned to obtain information for each atomic point. A line is made incident on the sample 1, and the electrons scattered at a high angle in the sample 1 are taken into the annular detector 2. The scattered electron uptake angle can be changed by a scattering angle control lens (see FIG. 2), and is set inside β ₁ and outside β ₂ of the annular detector. Since the scattered electron uptake angle differs depending on the element, it is necessary to set the scattered electron uptake angle according to the material according to the purpose with the scattering angle control lens. In the dark-field STEM observation method, the scattering intensity due to TDS increases as the thickness of the sample increases, so it is necessary to accurately measure the sample thickness.

FIG. 2 is an explanatory view of a main part of a standard scanning transmission electron microscope. In the figure, 11 is a field emission electron gun, 12 is an electron beam converging lens, 13 is a scanning deflection coil, and 13 is a scanning deflection coil.
Reference numeral 14 is an objective lens, 15 is a TEM sample, 16 is a scattering angle control lens, 17 is an annular detector, and 18 is an EELS detector. Although not shown, the scanning transmission electron microscope is provided with projection means for imaging the intensity and position (x, y, z) of the electrons detected by the annular detector 17.

Since the field emission electron gun 11 has a high coherence and generates an electron beam of high brightness, it has a high current density and a small beam size, that is, an electron beam of 1 [nm] or less can be obtained. . The acceleration voltage in such a scanning transmission electron microscope is 100 to 300 [kV].

The electron beam generated by the electron gun 11 is focused into a conical shape by the converging lens 12 and the objective lens 14 and irradiates the TEM sample 15. At this time, the resolution is determined by the focusing method of the electron lens, that is, the focusing lens 12 and the objective lens 14.

The function of scanning the electron beam is provided by the converging lens 12
And the scanning deflection coil 13 installed between the objective lenses 14.

The electron beam applied to the TEM sample 5 is scattered at various angles depending on the atomic number, and the angle at which the scattered electrons are taken into the annular detector 17 is controlled by the scattering angle control lens 16. To do.

The thickness of the TEM sample 5 in the electron beam transmission direction is measured by the EELS detector 18.

FIG. 3 is a flow chart for explaining an observation method using a scanning transmission electron microscope according to the present invention.
Hereinafter, the flow from (1) to (12) is as follows.

(1) Spherical aberration coefficient measurement Electromagnetic lens (for example, converging lens 1) that constitutes an electron microscope
2 and the objective lens 14, etc. (see FIG. 2), the spherical aberration coefficient (C _S ) is different for each manufacturer depending on the manufacturer, or
Different due to different lens performance. In dark-field STEM observation, C _S largely depends on the beam diameter and shape of the focused electron beam, and has a great influence on the image resolution.
Therefore, in each device and lens condition, C _S
Needs to be measured. The measurement method is derived from the phase transfer function and Ronchigram, and the measured C _S is used for the simulation.

(2) Measurement of electron beam incident angle The electron beam incident angle also largely depends on the beam diameter and shape of the convergent electron beam, and has a great influence on the image resolution. Therefore, the electron beam incident half-angle (α) under the beam conditions used for observation
Needs to be measured. α is calculated by obtaining the divergence angle of the beam from the electron beam diffraction line and dividing it by half the angle, and the obtained α is used for the simulation.

(3) Electron beam incident condition setting simulation C _S , α, acceleration voltage E ₀ , defocus amount Δf is used to calculate an electron beam incident condition setting simulation, and the optimum beam condition is determined based on the simulation. To do.

(4) Setting of electron beam incident conditions Match the experimental conditions to the optimum simulation conditions.

(5) Setting of Detector Intake Angle The detector take-in angle is set in order to determine the integration range in the scattering intensity calculation. First, the crystal structure, composition, and electron beam incident direction of the material are input to calculate the scattering cross section.

(6) Scattering intensity calculation Range of detector uptake angle (β ₁ −
β ₂ ) is integrated and the strength is calculated. This calculation is performed by the Bethe method based on the dynamic electron beam scattering theory.

(7) TEM sample thickness measurement The TEM sample thickness (t) also has a great influence on the dark-field STEM image intensity, so it must be measured. The thickness is EELS or CBED (convergent beam elect).
ron diffraction) method is applied and measured. Select the part with the optimum sample thickness as the experimental condition.
The measured t is used for the simulation.

(8) Image capture condition setting The image capture condition takes a long time to capture if the magnification and the pixel size of the digital image are large. Therefore, the sample drift amount is measured and the optimum condition is set. The magnification and the number of pixels used in the experiment are applied to the simulation.

(9) Dark-field STEM observation Dark-field STEM observation is performed by setting the conditions described above.

(10) Dark-field STEM simulation image By setting the above parameters, a dark-field STEM simulation image is created.

(11) Comparative dark field STEM observation results and dark field STEM simulation images are compared to verify the elemental composition distribution at the atomic level.

(12) Deconvolution From the verification result of the elemental composition distribution at the atomic level, a virtual image due to the electron beam intensity distribution is removed by deconvolution.

By adopting the above means, in the case of crystals,
By observing a dark-field STEM image, the structure and composition can be evaluated simultaneously from the captured image, and EDX (energy dispersive X), which is popular for performing composition analysis by TEM or STEM.
It is possible to obtain the composition at the atomic level, which is not possible with the -ray spectroscopy method or the EELS method. Furthermore, in order to improve the quantitative accuracy when observing the elemental composition image, EDX or EELS can be used to measure the position image. The uptake time is long, and it takes 30 [minutes] to 1 [hours] or more.
It takes less than 10 seconds to capture one screen, and not only the observation time can be shortened but also the influence of sample drift, which is a big problem in high-magnification observation, can be reduced.

[0046]

BEST MODE FOR CARRYING OUT THE INVENTION FIG. 4 is a diagram showing a result of a simulation performed for determining an optimum electron beam incident condition.

The calculation was carried out by using the following equations (1) and (2) according to the method used by Pennycock.

[0048]

[Equation 1]

Simulation conditions are: TEM acceleration voltage 200 [kV], objective lens spherical aberration coefficient (C _S ), convergence half-angle (α): 6.8, 8.8, 12.
5, 15, 17 [mrad], and the amount of deviation (Δf) of the focus of the objective lens: −40, −50, −60, −70.

As a result, the convergence half angle is 12.5 [mra.
d], the strength is high and the half width is narrow.
Since an electron beam of [nm] was generated, the convergence half angle was set to 12.5 [mrad] even under the experimental conditions.

In the simulation calculation, the contribution of elastically scattered electrons at each detector uptake angle and the contribution of inelastically scattered electrons due to TDS were calculated separately, and in comparison with the final experimental image, both were added. The combined images were compared. still,
The convergence half-angle is preferably 10 to 15 [mrad], and if it is less than 10 [mrad], the image cannot be seen,
Further, if it exceeds 15 [mrad], a virtual image will appear, and this point will be apparent from FIG. 15 described later.

FIG. 5 shows the scattering angle dependence of the scattering cross section of Si.
It is a diagram showing, and in the case of Si, s = 0-10 [nm^-1]
Elastic scattering is dominant at the capture angle of 10
[Nm ^-1], Inelastic scattering due to TDS is dominant
Becomes

S is the atomic scattering angle calculated by theoretical calculation, the unit is [nm ^-1 ], and θ is also the atomic scattering angle, but the unit is [mrad]. Convert from s to θ using equation (3) [mra
d] In units.

The scattering angle s and the scattering angle θ are expressed by the following equation (3).
It's like a relationship.

[0055]

[Equation 2]

As described above, it is necessary to set the optimum annular detector intake angle for each element.

The atomic scattering factors of elastic scattering and inelastic scattering due to TDS were calculated by the following equation (4).

[0058]

[Equation 3]

Here, f is an atomic scattering factor of elastic scattering f ′: an atomic scattering factor of inelastic scattering due to TDS M _S : Dehai-Waller factor.

The atomic scattering factor calculation based on the high angle approximation is
Weickenmeier and H.W. Using the high-angle scatter fitting parameters by Kohl, "Weic
kenmeier and H.M. Kohl, Acta
Cryst. A47 (1991) 590-597 ".

The scattering cross section was calculated using the following equation 5.

[0062]

[Equation 4]

The scattering cross section is the detector taking-in angle β _{1 to} β
Since it integrates over ₂ , the scattering intensity will depend on the capture angle and capture range.

FIG. 6 is a diagram showing the defocus amount (Δf) dependence of the Si (110) high resolution image. The left side shows the experimental result, the upper right side shows the experimental result with noise removed, and the lower right side shows the simulation result. Are shown respectively.

As shown, the defocus amount Δf =
At −40 [nm], the Si dumbbells are not separated, and at Δf = −65 [nm], further,
At Δf = −75 [nm], the Si dumbbells are separated, but it is possible to visually recognize that a contrast appears between the dumbbells where there are no atoms. Therefore, the image resolution is low and a virtual image appears unless the defocus amount Δf is optimized.

FIG. 7 shows the line of the scattering intensity of the Si dumbbell.
It is a diagram showing a profile, and the experimental results and the calculation results are in good agreement.

As described above, the condition for converging the electron beam cannot be uniquely determined, and by continuously capturing the image while changing the focus, the electron is reliably converged on the sample, and the atomic level is obtained. The conditions under which the quantification can be evaluated are determined.

FIG. 8 is an explanatory view showing the change in the scattering intensity at the position of the incident electron beam, and shows the case where the electron probe is placed on the Si atom and the case where it is placed in the absence of the Si atom.
It can be seen that the electron beam scattering excitation process differs depending on the defocus amount Δf.

FIG. 9 shows a sample used for observation: Si (110)
And a sample: an explanatory view showing a structure of GaAs (110), and in Si (110), Si called a dumbbell structure
A pair of atoms can be seen, and the distance between Si atoms is 0.136 [n
m], which is extremely short.

Therefore, the Si dumbbell structure is observed by TEM under normal conditions (for example, acceleration voltage = 200 [k
V], C _S = 1.0 [mm], atomic resolution = 0.23
[Nm]) is impossible to separate, and G
In the case of aAs, since the atomic numbers of Ga (Z = 31) and As (Z = 33) are close to each other, it was similarly difficult to identify them at the atomic level by the Z-contrast method.

FIG. 10 is a diagram showing the calculation results of the scattering cross section of the elastic scattering electron intensity of Si and GaAs used in the experiment and the inelastic scattering intensity by TDS. Both Si and GaAs are elastic scattering on the low angle side. The intensity of electrons is high, and the intensity of inelastically scattered electrons is low. The intensity reverses as the scattering angle becomes higher, and the intensity of the inelastically scattered electron increases. In order to obtain an image with an intensity dependent on the atomic number, it is necessary to predict the scattering cross section for each element from the calculation of the scattering cross section so that elastically scattered electrons do not enter the annular detector, and set the uptake angle depending on the element. Further, according to FIG. 10, Si and G
In comparison with aAs, it can be observed that the scattering intensity increases as the atomic number increases, and the electrons are scattered at higher angles.

FIG. 11 is a diagram showing the measurement result of the sample thickness, and the measurement was performed by applying the EELS method. In order to obtain the absolute value of the sample thickness, it is necessary to measure the scattering angle (β) of the electrons entering the EELS detector. The measurement result is 1
It was 2.5 [mrad]. In addition, in the sample thickness measurement by the EELS method, “RF Egerton, Elec
tron Energy Loss Spectros
copy in the Electron Micro
Scope, Plenum New York, 302
-307 (1981) ".

FIG. 12 is a photomicrograph showing an example of the dark field high resolution STEM image (element distribution image) observation of Si (110), and the take-in angle of the annular detector is 20 to 50.
In [mrad], the Si pair cannot be separated even if the sample thickness is changed, and thus 35-90 [mrad], 95-1
By setting it to 30 [mrad], Si pairs can be separated and a clear dumbbell structure can be observed.

By the way, in the data of FIG. 12, 35 to 50 are obtained when the take-in angle of the annular detector is 20 to 50 [mrad] and 35 to 90 [mrad].
It seems that there is a contradiction in this point because there is an overlap in the range of [mrad], and it is said that the Si pair cannot be separated on the one hand and it can be separated on the other hand. Factors are involved.

That is, the electron uptake angle changes depending on how the annular detector is set, and the angle has a great influence on the integrated intensity ratio of the scattering cross section and the inelastic scattering cross section. According to the above 35-50 [mr
In the range of [ad], Si dumbbells may or may not be observed.

Table 1 below shows the ratio of the integrated scattering cross section to the elastic scattering inelastic scattering cross section ratio with respect to the detector incorporation angle of each atom.

[0077]

[Table 1]

In the present invention, the angle at which electrons are taken in by the annular detector may be set to an angle at which the contribution of inelastically scattered electrons due to thermal diffuse scattering is larger than the contribution of elastically scattered electrons. However, preferably, it is 50 to 200.
It is good to select in the range of [mrad].

The reason for this is that in the case of Si, by setting the low angle side (inner angle) to 50 [mrad], the inelastic scattering cross section due to heat diffused scattering is more than the scattering cross section due to elastic scattering. Grows larger.

When all substances and materials are applied, as shown in FIG.
(Elastic) and (Inelac
At the intersection of (tic), the intensities of the two become the same, and if the angle is larger than that, the intensity due to inelastic scattering becomes larger than the intensity due to elastic scattering.

Therefore, since the total intensity can be obtained by integrating the scattering cross section from the vicinity of the intersection to a high angle, the intensity due to the inelastic scattering becomes dominant in the image as a whole.

The high angle side (outer angle) is 2
When set to 00 [mrad], the scattering cross section due to inelastic scattering of almost all substances becomes close to zero at that value.

Conventionally, only the inner angle is defined, and the outer angle is not defined, and it is taken at infinity. Since this converges as a value, it is theoretically not an error, but in the case of performing simulation calculation. However, unless the range of the capture angle is limited, a huge calculation time is required due to the approximation to infinity.

As described above, the scattering cross section due to inelastic scattering is close to zero, requires a huge amount of calculation time, and contains a substance having a large Z compared to Si,
Considering that the intensity can be obtained up to about 200 [mrad], the electron uptake angle is 50 to
It is appropriate to set it to 200 [mrad].

FIG. 13 is a photomicrograph showing an example of the dark field high resolution STEM image (element distribution image) observation result of GaAs (110). As in the case of Si observation, the annular detector incorporation angle is 20 to 50 [. mrad], the dumbbell structure could not be observed, and 35-90 [mrad], 9
Ga and As can be distinguished by setting it to 5 to 130 [mrad]. In the case of GaAs, Si
In the thin region, that is, the sample thickness is 45 [n
m] can be observed.

FIG. 14 shows Si (11) observed under various conditions.
It is a figure which shows the dark field STEM image (micrograph of an element distribution image) of 0) contrasted with the calculation image of the dark field STEM simulation result calculated by the Bethe method. Dark field STE
The M image intensity I ^HA (R) can be found in the following equation (6) by the wave field O (R) that appears when the electron beam hits the atomic column and the function P (R) that represents the incident electron beam intensity distribution. It is approximately represented by performing the convolution operation.

[0087]

[Equation 5]

Here, P (R) is the same as that in the equation (1), and O (R) is represented by the following equation (7).

[0089]

[Equation 6]

That is, the image intensity is calculated by calculating the electron probe shape and the wave field of electron scattering at each probe position (R). When α = 12.5 [mrad], defocus amount Δf
, An accurate Si atom image appears at Δf = −60 [nm], and a virtual image appears between Si atoms by making it underfocus, and α = 17 [m]
rad], it largely depends on the sub-peak and shape of the incident electron beam, and compared with the case of α = 12.5 [mrad],
Virtual image appears remarkably, and Si is further on the underfocus side.
The shape of the atom seems to extend laterally and does not show an accurate atomic image.

FIG. 15 shows a deconvolution method used for removing a virtual image, a simulation image, and HAA.
It is a figure showing a DF-STEM observation image (micrograph of an element distribution image).

The image intensity of the dark field STEM image is generated by the convolution of the wave field excited in the atomic column depending on the position of the electron beam probe and the incident electron beam. Therefore, the electron beam intensity distribution can be removed by deconvoluting the electron beam intensity distribution at each point of the image, and the virtual image shown in FIG. 6 or 15 can be removed. In the figure, a simulation image before and after convolution of Si (110) and an observation image are shown, but an accurate Si (110) atomic image can be obtained by deconvolution. In this way, a slight deviation in the incident condition can be removed by deconvoluting the electron beam intensity distribution.

FIG. 16 shows Si (1) according to the dark field STEM method.
10) High-resolution image observation result (micrograph of element distribution image:
(Right figure), image simulation result (left figure),
It is a figure showing the intensity profile comparison result.

As is clear from the observation results, a clear dumbbell structure in which Si atom pairs were separated was visually recognized. Also,
In order to verify the experimental results more accurately, a dark field STEM image simulation was carried out in which the electron beam scattering intensity based on TDS using a kinetic theory based on the Bethe method was applied. As a result, the experimental results and the simulations show a good agreement, and the spatial resolution reaches 0.136 [nm]. The experimental and simulation conditions at this time were as follows: electron beam incident half-angle 12.5 [mrad], annular detector take-in angle 35 to 90 [mrad], sample thickness 80 [nm],
Spherical aberration coefficient is 1.0 [mm], defocus amount is -60
[Nm].

FIG. 17 shows GaAs by the dark field STEM method.
It is a figure showing a (110) high resolution image observation result (micrograph of an element distribution image: right figure), an image simulation result (left figure), and an intensity profile comparison result.

It was confirmed that, unlike the case of Si, a difference in strength appeared between the left and right sides of the dumbbell. When the intensity line profiles of the experiment and the simulation were compared, the intensities of Ga and As at each atomic position were substantially the same. At this time, the experimental condition and the simulation condition are that the electron beam incident half angle is 12.5 [mrad], the annular detector take-in angle is 95 to 130 [mrad], and the sample thickness is 40 [n.
m], the spherical aberration coefficient is 1.0 [mm], and the defocus amount is −
It was 60 [nm]. As described above, according to the Z-contrast method, it is possible to identify the concept and the composition at the atomic level for substances having close atomic numbers.

[0097]

INDUSTRIAL APPLICABILITY In the observing method and the observing apparatus using the scanning transmission electron microscope according to the present invention, it is possible to observe the dark field STEM image and simultaneously evaluate the crystal structure and the composition from the captured image. Moreover, the composition can be grasped at the atomic level, and even when the atomic position image is captured when observing the elemental composition image, a sufficiently high quantitative accuracy can be obtained. Since it takes less than 10 seconds to capture one screen, the observation time can be shortened, and the influence of sample drift, which is a problem when observing at high magnification, can be reduced.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of the principle when performing dark field STEM observation.

FIG. 2 is an explanatory view of essential parts showing a standard scanning transmission electron microscope.

FIG. 3 is a flow chart for explaining an observation method using a scanning transmission electron microscope according to the present invention.

FIG. 4 is a diagram showing a result of a simulation performed for determining an optimum electron beam incident condition.

FIG. 5 is a diagram showing the scattering angle dependence of the scattering cross section of Si.

FIG. 6 is the amount of defocus (Δ) of a high-resolution image of Si (110).
f) It is a figure showing a dependency.

FIG. 7 is a diagram showing a line profile of scattering intensity of Si dumbbells.

FIG. 8 is an explanatory diagram showing a change in scattering intensity at an incident electron beam position.

FIG. 9: Samples used for observation: Si (110) and samples:
It is explanatory drawing showing the structure of GaAs (110).

FIG. 10 is a diagram showing the calculation results of the scattering cross section of the elastic scattering electron intensities of Si and GaAs used in the experiment and the inelastic scattering intensities by TDS.

FIG. 11 is a diagram showing a measurement result of sample thickness.

FIG. 12 is a micrograph showing an example of a dark field high resolution STEM image (element distribution image) observation result of Si (110).

FIG. 13: Dark field high resolution STE of GaAs (110)
It is a microscope picture showing an example of an M image (element distribution image) observation result.

FIG. 14 is a diagram showing a dark-field STEM image (micrograph of an element distribution image) of Si (110) observed under various conditions, in comparison with a calculated image of a dark-field STEM simulation result calculated by the Bethe method.

FIG. 15 is an explanatory view of a deconvolution method used for removing a virtual image, a simulation image, and HAADF-ST.
It is a figure showing an EM observation image (micrograph of an element distribution image).

FIG. 16 is a diagram showing a Si (110) high-resolution image observation result (micrograph of an element distribution image), an image simulation result, and an intensity profile comparison result by a dark-field STEM method.

FIG. 17: GaAs (110) by dark field STEM method
It is a figure showing a high resolution image observation result (micrograph of an element distribution image), an image simulation result, and an intensity profile comparison result.

[Explanation of symbols]

1 TEM sample 2 Annular detector 3 EELS detector α Convergence half-angle t which is half the value of electron beam convergence angle Thickness of sample 1 β ₁ and β ₂ Scattered electron uptake angle β EELS detector 3 Acquisition angle 11 Field emission electron gun 12 Electron beam converging lens 13 Scanning deflection coil 14 Objective lens 15 TEM sample 16 Scattering angle control lens 17 Annular detector 18 EELS detector

Continued front page    (72) Inventor Kazuto Watanabe             17-8 Horiguchi, Kanazawa-ku, Yokohama-shi, Kanagawa (72) Inventor Takashi Yamazaki             218 Santanda-cho, Asahi-ku, Yokohama-shi, Kanagawa F-term (reference) 2G001 AA03 BA11 CA03 DA01 DA09                       JA02 KA01 KA11 KA20 LA11                       MA05 NA07                 5C033 HH05 MM01 NN03 SS01 SS06                       SS07

Claims (5)

[Claims] 1. An electron beam accelerating voltage is 100 to 300 [kV].
A field emission electron gun, a focusing means for focusing an electron beam emitted from the field emission electron gun on a sample, a deflecting means for scanning a focused electron beam for irradiating the sample, and a projection for projecting a sample image. A scanning transmission electron microscope observing apparatus comprising: a means and a sample thickness measuring means. 2. An incident electron beam is made into a conical shape by an electron beam converging means composed of a two-stage or three-stage converging lens, and a half-angle of convergence on a sample is preferably in the range of 10 to 15 [mrad], preferably 12 [mrad]. mrad], a convergent electron beam is incident on the sample, and the electron scattered at a high angle of 35 [mrad] or more is detected in order to project the sample image by the electron beam scattering intensity at one point of the electron beam incident on the sample. Imaging Electromagnetic Lens Controllable over a Wide Range at Intake Angles of Electrons in Annular Detectors with Different Diameters, where the Inelastically Scattered Electron Contribution Due to Thermal Diffuse Scattering is Larger Compared to the Elastically Scattered Electron Contribution A system is used to control and capture the electrons scattered at a specific scattering angle, and the mechanism for adjusting the focus conditions is used to continuously change the focus of the detected electrons to establish a precise atomic position. Defocus amount as a condition in which the last intensity is generated -5
In the range of 0 [nm] to -70 [nm], preferably -6
An observation method using a scanning transmission electron microscope, characterized in that it includes a process of taking an image by setting it to be 0 [nm]. 3. A process for measuring a sample thickness by applying electron energy loss spectroscopy (EELS method) or convergent electron diffraction method (CBED method) after detecting electrons by a ring-shaped detector is included. An observation method using a scanning transmission electron microscope according to claim 2. 4. A crystal structure of a material based on a high angle approximation, a physical property parameter of constituent atoms, a convergence half angle used in an experiment, an annular detector uptake angle, a sample thickness, an acceleration voltage, a defocus amount, and an objective lens spherical aberration. The dark field STEM image simulation is performed by the image simulation in which the electron beam transmission path in the sample is estimated by the calculation by the Bethe method using the experimental parameter including the coefficient as the calculation parameter, and the dark field STEM image observation obtained in the experiment is performed. An observation method using a scanning transmission electron microscope according to claim 2, further comprising a process of comparing the result. 5. In the dark field STEM simulation, a calculation method and an experimental evaluation method that incorporate the effects of elastically scattered electrons due to Bragg diffraction and inelastically scattered electrons due to thermal diffuse scattering of nuclei are adopted, and 5. A scanning transmission electron microscope according to claim 2, wherein a virtual image is removed by deconvoluting the effect of the electron beam intensity distribution from the visual field STEM observation image and the simulation result. Method.