JPH04332845A - Preparation of sample for observation by transmission electron microscope - Google Patents

Abstract

PURPOSE:To shorten a time spent for a section forming process in a process of preparation of a sample for TEM observation of a section of a semiconductor and the like and to improve the working efficiency of preparation of the sample. CONSTITUTION:Before cut pieces of a semiconductor sample are stuck together in lamination, FIB is applied onto the surface of the sample to provide appropriate marking thereon by etching. Thereby positions of a section in the course of grinding and an object of observation can be made distinct in a section forming process for which an enormous working time is spent usually in a process of preparing a section TEM sample, and a working efficiency is improved sharply. At the same time, excessive grinding is prevented and thereby failure is eliminated.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for preparing a sample for cross-sectional observation in the analysis of internal defects in semiconductor devices by TEM observation.

0002

BACKGROUND OF THE INVENTION Here, TEM observation of a cross section of a semiconductor will be taken as an example and will be described below. FIG. 2 shows a process block diagram of a conventional method for preparing a sample for semiconductor cross-sectional TEM observation. In analyzing internal defects in semiconductor devices using TEM observation,
In order to obtain good results, it is necessary to obtain clear TEM images, and for this purpose, the thickness of the observation area of the sample must be made as thin as possible, and the number of electrons passing through the sample must be increased. An ideal sample for cross-sectional observation of a semiconductor device is shown in FIG. For example, the penetration power of an electron probe at an accelerating voltage of 400 kV is about 2 to 3 μm, so a clear T
In order to obtain an EM image photograph, the thickness of the observation area shown in FIG. 6 needs to be 1 μm or less. Figure 2 shows how to prepare a cross-sectional sample.
As shown in , first, in the "sample cutting" process, the device sample and five silicon wafers are each cut into 5 mm x 5
Cut out into mm size. In the next "bonding", as shown in FIG. 4(1), three silicon wafers 20 and two silicon wafers 30 are bonded to the front and back sides of the cut out device sample 10 using epoxy resin or the like. Then, in "cutting out the sample thin section", as shown in Figure 4 (2), from the completed rectangular parallelepiped, 0.5~
A thin section with a thickness of 1.0 mm is cut out so that the observation target area is included inside the section. Next, in order to expose the cross section to be observed, the cut thin section is roughly polished and buffed from the side closest to the desired cross section to make that surface a mirror surface, and a metallurgical microscope is used to examine the center of the cross section using a TEM. Check whether the cross-sectional structure of the object to be observed exists. If the cross-sectional structure is not confirmed at the center of the surface, repeat rough polishing and mirror polishing again and then observe with a metallurgical microscope until the cross-sectional structure of the object to be observed with TEM appears at the center of the surface. . If the object to be observed is lost due to excessive cutting, if there are multiple samples, all you have to do is cut out the sample and start over, but if it is the only sample, the entire analysis will stop.

After the desired TEM observation cross section is completed as a mirror surface, a disk with a diameter of 3 mm centered on the object to be observed is cut out from the flaky sample using an ultrasonic disk cutter.
Roughly polish the sample from the opposite side to the observed cross section to reduce the thickness of the sample to 10
The thickness should be approximately 0 μm. Further, the thickness of the center of the sample is increased to about 20 μm by ball polishing, mirror polishing is performed, and the shape shown in FIG. 7 is finished by ion milling.

0004

[Problems to be Solved by the Invention] Here, a cross-sectional TEM observation of a semiconductor having an ultra-fine structure will be taken as an example and will be described below. In cross-sectional TEM observation of a semiconductor, it is necessary that the observation target location be present in the center of the observation surface of the sample and that the portion be sufficiently thin. However, it is extremely difficult to ensure that a specific location in the structure of a semiconductor device, which is determined by parameters in μm, is located in this limited location. When preparing a sample according to the above-mentioned conventional example, a thin section of the sample is cut out from the rectangular parallelepiped shown in Figure 4 (1) so that the observation target area is included inside it, and then polished so that the desired cross section appears on the sample surface. do. In this cross-section forming process, in order to accurately represent the desired cross-section on the sample surface, the depth of the desired cross-section from the surface of the thin slice must be precisely known in the μm order. In the formation method, the distance between the surface being polished and the desired cross-section is completely unclear at that point, so it is easy to over-polish, and the sample is abandoned at this point without having to start over again. must not. With great care,
It is possible to obtain a satisfactory cross-section by repeating the process of performing very slight rough polishing, mirror polishing, and checking with a metallurgical microscope, but it is costly due to lack of precision. The amount of time spent on this process is enormous, and work efficiency is extremely reduced. In addition, if you lose the object to be observed due to excessive cutting, if there are multiple samples, you can start over by cutting out the sample, but if it is the only sample, the entire analysis will stop, and the research will be forced to stop. , which will have a significant impact on development. For this reason, it is not easy to analyze identified internal defects in semiconductor devices or the like by TEM observation.

[0005] The sample preparation method of the present invention greatly reduces the working time by accurately knowing the position of the cross section to be observed from the mark etched by the FIB in this cross-section forming step, and prevents the specimen from disappearing due to over-shaving. By preventing this, the purpose is to improve the work efficiency of internal failure defect analysis of semiconductor devices by TEM observation, and related research and development.

[0006]

[Means for Solving the Problems] In the sample preparation method of the present invention, in order to improve work efficiency such as internal failure defect analysis of semiconductor devices by TEM observation, the sample preparation method A mark is made using a focused ion beam (FIB) near the location to be observed, and the sample is polished from the marked side.

[0007]

[Example] Next, TEG (Test El
Embodiments of the present invention will be described with reference to the drawings, taking as an example a cross-sectional observation of AI of element Group). FIG. 1 shows a process block diagram of the process of preparing a sample for TEM observation according to the present invention. First, as in the conventional method, 5mm
Cut out one sample piece of ×5 mm and five silicon wafers. Before pasting them together, a marking is applied around the observation target part of the sample by FIB etching, an example of which is shown in FIG.

[0008] A W-shaped marking made of two right-angled isosceles triangles with a height of 20 μm is placed 20 μm away from the target cross section in the rough polishing direction (perpendicular to the target cross section) using an FIB. Attach as in 3. In Figure 3,
Straight lines a, b are parallel to the exchange of the target cross section and the sample surface. If there is only one marking, there is a risk that if you accidentally polish from the unmarked side, you will end up shaving off the cross section you are looking for. If you cut out the sample thin section as shown in Figure 4 (2), you will not be able to tell which side is marked.
By marking both sides of the cross-section with the desired cross-section in the center, you can sand from either side. In other words, in FIG. 3, accurate polishing can be performed from either the right or left side of the figure.

Now, taking as an example the case where polishing starts from the right side of FIG. 3, how the FIB marking 40 indicates the distance to the target cross section will be explained with reference to FIGS. 3 and 5. go. When the cross section when polishing has progressed to straight line c in FIG. 3 is observed with a metallurgical microscope, it appears as shown in FIG. 5(1). Since the height of the two right-angled isosceles triangles that make up the W-shaped marking is 20 μm, the straight line c in Figure 3
The distance between the two apex angles of the W-shaped marking on the top is 40μm
Therefore, the distance to the target cross section is also 40 μm. That is, if after rough polishing and mirror polishing, the cross section is observed with a metallurgical microscope and it looks like the one shown in FIG. 5(1), it can be accurately determined that the distance to the target cross section is 40 μm. Also, make sure that the cross section being polished is 20 to 40μ from the target cross section.
When it is at a distance of m, it is observed as shown in Figure 5 (2),
If the cross section being polished is parallel to the target cross section, the two distances indicated by the double arrows in the figure will be equal. If the distances x indicated by the double-headed arrows in the figure are not equal, it can be seen that the cross section being polished is not parallel to the target cross section. Here, if the distance indicated by the double-headed arrow is xμm, then the distance to the target cross section is:
Distance to target cross section = 20+20・(1−
x/40) (μm). If it looks like Figure 5 (3), then it is only necessary to remove another 20 μm. The polishing rate during rough polishing of silicon substrate is approximately 2μm/se
Because it is relatively slow at c, even if you manually measure the time while polishing, you can control the depth of polishing quite accurately.

In FIG. 3, markings are made only in the range of 20 to 40 μm from the target cross section, but accuracy can be further increased by marking in a plurality of ranges. An example is shown in FIG. In addition to the markings in Figure 3, 60~
Mark the W shape in Figure 3 with another right isosceles triangle (3 V shapes) in the 80 μm range, and
By marking the 20 μm range with 2 plus (4 V-shaped; 2 W-shaped) and the 200-300 μm range with a right-angled isosceles triangle with a base of 100 μm, the current Polishing can be progressed while gradually grasping the position of the polished cross section, further improving accuracy.

[0011]

[Effects of the Invention] As explained above, if the present invention is applied to the sample preparation process for cross-sectional TEM observation of semiconductors, etc., the time spent in the cross-section forming process can be significantly shortened, and at the same time, mistakes such as over-cutting can be avoided. This will greatly improve work efficiency.

[Brief explanation of the drawing]

FIG. 1 is a process diagram of the present invention.

FIG. 2 is a process diagram of the prior art.

FIG. 3 is a diagram showing an example of marking on a sample surface in the present invention.

[Figure 4] "Lamination" common in Figures 1 and 2
, A diagram showing the process of "cutting out a sample thin section."

FIG. 5 is a diagram for making it easier to understand the explanation of the embodiment of the present invention.

FIG. 6 is a diagram showing another example of marking in the present invention.

FIG. 7 is a diagram showing an ideal completed sample.

[Explanation of symbols]

10 Sample 20, 30 Silicon wafer 40 Marking by FIB

Claims (1)

[Claims] Claim 1: Before the cross-section forming step in the sample preparation process for transmission electron microscopy (TEM) observation, a mark is applied with a focused ion beam near the location in the sample to be observed, and the sample is inspected from the marked side. A method for preparing a sample for cross-sectional observation using a transmission electron microscope, the method comprising polishing.