JPH04219395A - Formation of crystal - Google Patents

Abstract

PURPOSE:To homogeneously produce a singly directed single crystal having a large area in a state free from plurally directed crystals by bringing an original crystal seed into contact with a corner of a depression formed on the surface of a substrate, and having, that is, the mutually vertical surfaces, heating the original crystal seed and subsequently singly solidifying the melted seed, when the single crystalline seed crystal as a starting point is grown on the substrate. CONSTITUTION:A depression 11 having a bottom surface 15 and two side surfaces 13, 14 in a mutually vertical state is formed on a surface of an amorphous insulating substrate 10 consisting of SiO2, etc., and a fine original crystal seed 12 consisting of polycrystalline silicon, etc., is arranged at a corner of the depression in a state that the original crystal seed 12 contacts the three vertical surfaces 13-15. The original crystal seed 12 is heated to form a single crystalline seed crystal, from which as a starting point a single crystal is grown. The method thereby gives high quality single crystals usable for semiconductors, optical elements, etc., on amorphous insulating substrates.

Description

[Detailed description of the invention]

0001

[Industrial Application Field] The present invention is directed to the formation of single crystals used as materials for electronic devices such as semiconductor integrated circuits, optical integrated circuits, magnetic circuits, optical devices, magnetic devices, piezoelectric devices, surface acoustic devices, etc. Regarding how to.

0002

2. Description of the Related Art Conventionally, single crystal thin films used in semiconductor devices, optical devices, etc. have been formed by epitaxial growth on single crystal substrates.

For example, it has been known that Si, Ge, GaAs, etc. can be epitaxially grown on a Si single crystal substrate (silicon wafer) from a liquid phase, a gas phase, or a solid phase. is GaAs
It is known that single crystals such as GaAlAs and GaAlAs can be grown epitaxially.

The semiconductor thin film thus formed can be used in various fields such as semiconductor devices, integrated circuits, and light emitting devices such as semiconductor lasers and LEDs.

[0005]Recently, there has been much research and development into ultra-high-speed transistors using two-dimensional electron gas and superlattice devices using quantum wells, but what has made these possible is
For example, these are high-precision epitaxial techniques such as MBE (molecular beam epitaxy) and MOCVD (metal-organic chemical vapor deposition) using ultra-high vacuum.

In order to perform epitaxial growth on such a single crystal substrate, it is necessary to match the lattice constant and thermal expansion coefficient between the single crystal material of the substrate and the epitaxially grown layer.

For example, it is possible to epitaxially grow a Si single-crystal thin film on sapphire, which is an insulating single-crystal substrate, but crystal lattice defects at the interface due to lattice constant deviations and the epitaxial layer of aluminum, which is a component of sapphire, occur. Diffusion into other materials is a problem when applied to electronic devices and circuits.

[0008] Thus, it can be seen that the conventional method of forming a single crystal thin film by epitaxial growth largely depends on the substrate material. Mathews et al. investigated combinations of substrate materials and epitaxially grown layers (E.
PITAXIAL GROWTH. Academic
Press, New York, 1975 edit
ed by J. W. Mathews).

Furthermore, the size of the substrate is currently as small as about 6 inches for Si wafers, and the increase in the size of GaAs and sapphire substrates has been delayed even further. Furthermore, single crystal substrates are expensive to manufacture, resulting in a high cost per chip.

[0010] As described above, in order to form a single-crystal layer capable of producing a high-quality device by the conventional method, there is a problem in that the type of substrate material is limited to an extremely narrow range.

On the other hand, research and development of three-dimensional integrated circuits in which semiconductor elements are stacked in the normal direction of a substrate to achieve high integration and multifunctionality have been actively conducted in recent years. Furthermore, research and development of large-area semiconductor devices such as solar cells and liquid crystal pixel switching transistors in which elements are arranged in an array on inexpensive glass is becoming more popular year by year.

What these two methods have in common is that they require a technique for forming a semiconductor thin film on an amorphous insulator and forming electronic elements such as transistors thereon. Among these, a technique for forming a high quality single crystal semiconductor on an amorphous insulator is particularly desired.

Generally, when a thin film is deposited on an amorphous insulating substrate such as SiO2, the crystal structure of the deposited film becomes amorphous or polycrystalline due to the lack of long-range order in the substrate material. Here, an amorphous film is one in which short-range order at the level of the nearest neighbor atoms is preserved, but no longer-range order, and a polycrystalline film is one that has a specific crystal orientation. (It is a collection of single crystal grains separated by grain boundaries).

For example, when Si is formed on SiO2 by the CVD method, if the deposition temperature is about 600°C or less, it becomes amorphous Si, and if the temperature is higher than that, it becomes amorphous Si with a grain size of several hundred to several thousand. It becomes crystalline Si. However, polycrystalline Si
The particle size varies greatly depending on the formation conditions.

Furthermore, by melting and solidifying an amorphous or polycrystalline film using an energy beam such as a laser or a rod-shaped heater, a polycrystalline thin film with large grain sizes on the order of microns or millimeters can be obtained (Single-Cry).
stal silicon on non-singl
e-crystallinsulators. Jou
rnal of crystal growth vo
l. 63, No. 3, October, 1983
edited by G. W. Cullen).

When a transistor is formed on the thin film of each crystal structure formed in this way and the electron mobility is measured from its characteristics, it is ~0.1 cm2/V·se for amorphous Si.
c, 1 to 10 for polycrystalline Si with a grain size of several hundred
cm 2 /V·sec, polycrystalline Si with a large grain size obtained by melting and solidification has a mobility comparable to that of single-crystal Si.

From these results, it can be seen that there is a large difference in electrical characteristics between an element formed in a single crystal region within a crystal grain and an element formed across a grain boundary.

In other words, the deposited film on an amorphous layer has an amorphous or polycrystalline structure, and the performance of devices fabricated thereon is significantly inferior to that of devices fabricated on a single crystal layer. . Therefore, its applications are limited to simple switching elements, solar cells, photoelectric conversion elements, etc.

[0019]

[Problems to be Solved by the Invention] There are roughly two methods for depositing a crystal layer on an amorphous substrate.

One is a single crystal (for example, Si) on the substrate.
on which an amorphous insulator (for example, SiO2
), a part of this coating is removed to expose the underlying single crystal surface, and this is used as a seed crystal for lateral epitaxial growth in the gas phase, solid phase, and liquid phase. A method of forming a single crystal region on a layer.

The other method is to grow a crystal thin film directly on a substrate without using a single crystal as the underlying substrate.

As described above, the surface of an amorphous substrate does not have long-range order unlike the surface of a single crystal substrate, and only short-range order is maintained. Therefore, the structure of the as-deposited thin film is, at best, polycrystalline with disordered grain boundary positions. In addition, not only is there no long-range order on the surface of an amorphous substrate, but there is also no anisotropy that determines the crystal orientation (substrate normal direction and in-plane orientation), so it is impossible to control the crystal orientation of the upper layer. Met.

That is, the problems with deposited layers on amorphous substrates boil down to control of grain boundary positions and crystal orientation.

Regarding the control of the grain boundary position, it has been shown that the grain boundary position can be determined by artificially predefining the nucleation position (Japanese Unexamined Patent Publication No. 107016/1983).
Sentaxy (Selective N
(T. Yonehara, Y. Nish)
igaki, H. Mizutani, S. Kond
oh, K. Yamagata, T. Noma an
dT. Ishikawa, Applied Phys.
sics Letters vol. 12, pp. 1
231, 1988). This technology uses S on SiO2.
i3 N4 is localized and becomes a nucleation site,
A single crystal of Si grows and collides with a crystal grown from an adjacent site, thereby forming a grain boundary and determining the position of the grain boundary.

Furthermore, instead of using the "nuclei" spontaneously generated at the nucleation position as described above, a non-single-crystal "primary seed" material is patterned on the substrate in advance. A technique has been disclosed in which the seed crystal is transformed into a single crystal by utilizing the agglomeration phenomenon of the original seed, and this is used as a "seed crystal" (Japanese Patent Application Laid-open No. 1-132117).

In this technique, the plane orientation of the crystal (crystal orientation perpendicular to the substrate) is determined by the interfacial energy (int
Stabilization of surface energy (free surface energy)
) is determined by factors such as stabilization and relaxation of internal stress. However, even with this method, it is very difficult to obtain a completely single orientation, and a plurality of orientations in addition to the main surface orientation often coexist. In particular, the in-plane crystal orientation cannot be determined solely because the nucleation plane or agglomeration plane is amorphous and has no anisotropy.

On the other hand, in 1978, H. I. Smith
showed for the first time that the crystal orientation of KCl deposited on an amorphous substrate could be controlled by artificially imparting anisotropy due to unevenness to the surface of the amorphous substrate using lithography, and G.
raphoepitaxy (H.I.Sm
ith and D. C. Flanders, App
Lied Physics Letters vol.
32, pp. 349, 1978) (H.I.Sm
ith, U.S. Pat. No. 4,333,792).

After that, the grain growth of the Ge thin film (T. Yone
hara, H. I. Smith, C. V. Thom
pson and J. E. Palmer, Appl.
ied Physics Letters vol.
45, pp. 631, 1984), initial growth of Sn (L.S. Darken and D.H. Lown
Dere, Applied Physics Let
ters vol. 40, pp. 954, 198
7), it was also confirmed that the artificial relief pattern on the substrate surface affected its crystal orientation.

However, in graphoepitaxy, KCl and Sn were found to be effective for the orientation of individual crystals separated at the initial stage of deposition, and for continuous layers, crystals were grown by laser annealing after depositing Si ( M. W. Geis, D. A. Fla.
ndersand H. I. Smith, Appli
ed Physics Letters vol. 35
, pp. 71, 1979) and solid-phase growth of Ge (T.
Yonehara, H. I. Smith, C. V.
Thompson and J. E. Palmer,
Applied Physics Letters vo
l. 45, pp. 631, 1984) has been reported.

In the case of Si and Ge, although the orientation is controlled to some extent, the crystal groups are arranged in a mosaic pattern,
Furthermore, grain boundaries exist between the crystals and crystals having slightly different crystal orientations, and the positions of these grain boundaries are disordered, making it impossible to obtain a single crystal uniformly over a large area. This is because, in addition to the fact that the three-dimensional crystal orientations of individual crystals do not completely match, the position of nucleation is not controlled in the surface relief pattern.

The present invention has been made to solve the problems of the prior art described above, and its purpose is to provide a single crystal with a single orientation in which there is no coexistence of multiple orientations other than the principal plane orientation. The object of the present invention is to provide a method that can uniformly and easily form a large area.

[0032]

[Means for Solving the Problems] The present invention provides a method for forming a crystal in which a single crystal seed crystal is placed on a substrate and a single crystal is grown from the seed crystal as a starting point. A concave portion having at least a corner where the three sides touch each other as perpendicular surfaces is formed on the surface of the substrate, an original seed is placed in the corner so as to be in contact with the three sides, and the original seed is heat-treated to form a single recess. This method of forming a crystal is characterized in that a single crystal seed crystal is agglomerated to form a single crystal seed crystal, and a single crystal is grown using the single crystal seed crystal as a starting point.

As mentioned above, the plane orientation of the crystal after aggregation is usually determined by the driving force of minimizing the surface energy of the crystal, the interfacial energy with the substrate, etc. For example, when polycrystalline Si is agglomerated on SiO2, the surface energy of Si is minimum at the (111) plane, so the plane orientation is changed to the (111) plane.
A force is generated that tries to align with 1). On the other hand, Si and SiO
The interfacial energy with 2 is stable at (100), so (
100) is generated.

In addition, since polycrystalline Si itself generally has a (110) orientation, although it depends on its deposition temperature, factors related to its initial state also come into play. As a result, the Si aggregated on SiO2 is relatively strongly oriented in (111), and according to the reflection intensity of X-ray diffraction, (110) exists with an intensity of about 1/2 to 1/3 of (111). In addition, (100), (311), (331),
(422) and other weak peaks are observed.

The reason why the (111) orientation is dominant here is that the agglomerated Si becomes a hemispherical crystal whose surface area is considerably larger than the area of the interface, and that among the factors that determine the orientation, It is thought that stabilization of surface energy is the strongest factor.

[0036] In the conventional method as described above, that is, the method in which original seeds are placed on a flat surface, surface energy, interfacial energy, and many other factors affect the time of agglomeration, so one orientation cannot be completely fixed. It is extremely difficult to control the angle, and even more so, it is almost impossible to control the in-plane orientation.

On the other hand, in the method of the present invention, instead of forming seed crystals by agglomerating original seeds located on a flat surface as in the conventional method, the original seeds are placed in one corner of the recess on the substrate surface. (the corner where the bottom surface and two side surfaces intersect each other as perpendicular planes) and agglomerate, so the three vertical planes of this corner are planes equivalent to the stable orientation structure of the seed crystal formed by agglomeration. It works well as a. In other words, stabilization of the interfacial energy is the main factor that determines the orientation during aggregation, and the in-plane orientation can also be controlled. Furthermore, since the exposed area of the original seeds is smaller than when the original seeds are arranged on a flat surface, it also has the effect of reducing the influence of structural changes due to stabilization of surface energy during aggregation. Mainly due to this kind of action, control of orientation (including in-plane) is reproduced in the system of aggregation of original seeds, and as a result, crystals with a single orientation can be formed easily and well. .

The present invention will be explained in detail below with reference to the drawings.

FIG. 1 is a schematic diagram illustrating a recess (crystal growth region) formed in a substrate and an original seed formed in the recess. FIG. 2 is a partial cross-sectional view of the recess and the original seed illustrated in FIG. 1.

In the example shown in FIGS. 1 and 2,
The original seed 12 has a side surface 13 of the recess (crystal growth region) 11,
14 and the bottom surface 15. In the present invention, the side surfaces 13, 14 and the bottom surface 15 need to be located perpendicular to each other. These three surfaces 13 to 15 function well during aggregation as described above.

The expression that the surfaces 13 to 15 are perpendicular to each other means not only strictly perpendicular but also substantially perpendicular. For example, even if an error within a few degrees occurs due to deviations from the design value due to pattern edge droop or over-etching that occurs in normal photolithography processes, etching processes, etc., it is not perpendicular in the sense of the present invention. , the effects of the present invention can be obtained even in this case.

Even if the planes 13 to 15 are formed with almost completely vertical plane angles, it is impossible for the orientations of many single crystal seed crystals to be in a specific orientation without a deviation of several arcseconds. Therefore, in the present invention, it is sufficient that the perpendicularity is perpendicular to such an extent that, when devices are specifically manufactured in many crystal regions, variations in device characteristics due to variations in orientation do not occur.

[0043] The shape of the original seed 12 is not particularly limited. However, the original seed 12 has three sides 13 to 15 as shown in FIG.
Since the original seed 12 is arranged so as to be in contact with the original seed 12, the contact portions of the original seed 12 are necessarily three vertical planes. The shape of the portion of the original seed 12 that is not in contact with the surfaces 13 to 15 is not particularly limited, and may be any shape that is easy to form.

The size of the original seeds 12 may be small enough to aggregate into a single seed. When the original seed 12 has a rectangular parallelepiped shape as shown in FIG.
is preferably 2 μm or less, preferably 1 μm or less, and optimally about 0.5 μm. If these a and b exceed 2 μm, the original seeds 12 may not be a single body but may be divided into a plurality of pieces and aggregated. Such a plurality of crystals cannot serve as a seed crystal for growing a single crystal. However, whether or not the original seeds 12 are divided during aggregation is largely influenced by the value of the thickness (height) c of the original seeds. If the aspect ratio between a or b and c is large, it will be easy to divide during aggregation.

[0045] Furthermore, it is desirable that a and b be 0.1 μm or more. When these are less than 0.1 μm, the base area becomes extremely small, and orientation control due to the stabilization energy of the interface tends to become difficult. In addition, when using a normal photolithography device, 0.1
Patterning smaller than μm is difficult. And it is not necessarily necessary that a=b.

The thickness (height) c of the original seed 12 is 0.05
The thickness is preferably 0.1 μm or more and 0.4 μm or less, and preferably 0.1 μm or more and 0.4 μm or less. The optimal value of this c is
Determined by the sizes of a and b. In fact, the smaller the value of c, that is, the thinner the original seeds 12, the easier aggregation occurs. However, when c becomes 0.1 μm or less,
As mentioned above, the values of a and b must be made small so as not to break apart during aggregation, and the influence of the side surfaces 13 to 15 for controlling the in-plane orientation becomes small. Furthermore, it becomes difficult to apply a general photo process. Furthermore, when the value of c exceeds 0.5 μm, aggregation tends to be difficult to occur.

[0047] Even if the shape of the original seed 12 is not a rectangular parallelepiped, the suitable size for the parts that come into direct contact with the surfaces 13 to 15 is similar to the above-mentioned values, and if this is used as a guide, good.

Materials for the original seeds 12 include Si, Ge,
Various materials conventionally known as raw seed materials can be used, such as semiconductor elements such as Sn, metals such as Au, Ag, Cu, Pt, and Pd, alloys, compounds, and mixtures.

The base 10 having the recess 11 is made of SiO2
, SixNy, SiONx and other amorphous insulators,
Various types of substrates conventionally known as substrates for crystal formation can be used, and it is useful when the entire substrate or only the surface has a relatively low nucleation density.More specifically, the entire substrate Alternatively, it is particularly useful when using a substrate whose only surface is made of an amorphous insulator.

Next, each step of the method of the present invention will be illustrated using FIG.

First, as shown in FIG. 3(A), the base 3
A recess 31 that will become a crystal growth region is formed at 0. Recess 31
The shape of the surface may be square or rectangular as long as the portion where the original seeds will be placed later is patterned as a desired corner. Although the depth of the recess 31 is arbitrary, it is preferable to set it to the same value as the thickness of the crystal to be grown later. There is no particular limitation on the method for forming the recess 31, but for example, RIE is preferable to so-called wet etching using an etching solution.
(reactive ion etching) etc.
This is preferable because the side surfaces 33 and the like are more perpendicular to the bottom surface.

Next, as shown in FIG. 3(B), original seed material 32' is deposited. As described above, Si, Ge, Sn, etc. can be used as this material, and it does not matter if it is polycrystalline or amorphous. Suitable values for the deposited film thickness and the like are as described above, but the values should not be larger than the depth of the recess 31. This deposition method includes
There are vapor deposition methods, sputtering methods, CVD methods, etc., but since it is undesirable for the original seed material 32 to remain unnecessarily on the side surfaces of the recesses 31 in later steps, vapor deposition methods and sputtering methods with poor step coverage are preferable. is suitable.

Next, as shown in FIG. 3C, the original seeds 32 are left at desired corners in the crystal growth region, and the rest are etched. Etching may be performed by a dry method using RIE or by a wet method using a solution. Suitable values for the size and the like of the original seeds 32 are as described above using FIG. 1.

Next, as shown in FIG. 3(D), the original seeds 32 are heat-treated in a hydrogen atmosphere to cause aggregation.
A single-crystalline seed crystal 36 with uniform plane orientation and in-plane orientation can be formed. The conditions for heat treatment depend on the material of the original seed 32 and its volume, but when the original seed material is a semiconductor element such as Si or Ge, the temperature is usually about 700 to 1100°C. Also, the pressure may be normal pressure, but reduced pressure (several Torr to
200 Torr) is more likely to cause aggregation. Further, doping a large amount of impurities (phosphorus, arsenic, boron, etc.) into the original seeds 32 has the effect of lowering the temperature at which aggregation starts.

Next, as shown in FIG. 3(E), a crystal 37 is selectively grown using the CVD method or the like, centering on the seed crystal 36. For example, if Si crystals are grown using the CVD method, silane-based materials such as SiH4, Si2 H6, S
It is sufficient to use a chlorosilane-based source gas such as iH2 Cl2, SiHCl3, SiCl4, or a fluorosilane-based source gas such as SiH2 F2, SiF4.
Crystal growth can be performed by mixing an etching gas such as H2, HF, etc. in a H2 diluent gas. The temperature in the CVD method may be within a range of approximately 800 to 1200°C, and the pressure may be within a range of approximately several Torr to 200 Torr. The growth is preferably carried out until the recess 31 in the crystal growth region is completely filled. The grown crystal 37 grows with facets unique to single crystals. This crystal 37 grows with facets unique to single crystals. Further, as described above, the growth crystal 37 can be easily determined by the size and shape of the recess 31.

Next, as shown in FIG. 3(F), the recess 3
It is preferable to remove the portion of the crystal 37 that protrudes from the substrate 1 and flatten the base 30 and the crystal 37, because elements and the like can be easily and satisfactorily formed on this flat surface, and circuits can be easily formed.

[0057] This planarization is preferably performed by, for example, a selective polishing method. There are mainly two types of selective polishing methods. One is ``mechanochemical polishing,'' which removes reaction products while causing a chemical reaction with Si using an alkaline solution (Hamaguchi, Endo, Journal of the Japan Society of Applied Physics, Vol. 5).
Volume 6, No. 11, page 1480, etc.). The other method is to utilize the difference in mechanical polishing speed to polish only the material to be polished (corresponding to crystal 37 in Figure 3) with a high mechanical polishing speed with abrasive grains such as colloidal silica. Finish the polishing at a small stopper surface (corresponding to the base 30 in Figure 3).
"Mechanical polishing" (Japanese Unexamined Patent Publication No. 209730/1999). Of these, the latter mechanical polishing is advantageous when the crystal orientation is random, but since the crystals of the present invention have aligned orientations, either method can be used, and from this point of view, the present invention is advantageous. The method is excellent.

[0058]

[Examples] The present invention will be explained in more detail below with reference to Examples. Example 1 First, as shown in FIG. 3(A), a square shape of 20×20 μm 2 was patterned on the surface of a 4-inch fused silica substrate 30. Then, RI this part to a depth of 0.4 μm.
Etched by E to form a recess 31 as a crystal growth region.
was formed.

Next, as shown in FIG. 3(B), LPC
0 polycrystalline Si (original seed material 32') by VD method
．． A thickness of 1 μm was deposited.

Next, as shown in FIG. 3(C), the recess 3
A 1.0 μm x 1.0 μm island-like region was patterned at one corner of No. 1 using a normal photolithography process, and the other portions were patterned using HF:HNO3:
Etching was performed using a solution of CH3COOH=1:40:40, and the remaining island-like regions were used as original seeds 32.

Next, as shown in FIG. 3(D), the substrate 30 on which the original seeds 32 were applied was heated at 1050° C. in a hydrogen atmosphere.
Heat treatment was performed at 100 Torr for 3 minutes. Then the original seed 32
aggregated into a hemispherical shape, and its crystallinity changed from polycrystalline to single crystalline. This aggregated single crystal 36 is then used as a “seed crystal”.
And so.

Next, as shown in FIG. 3(E), starting from the seed crystal 36 formed by aggregation, a Si single crystal (
Growth crystal 37) was selectively grown. The growth conditions were as follows: SiH2 Cl2 as a source gas, HCl as an etching gas, and H2 as a carrier gas.
53 : 1.6 : 100 (l/min), and this was used as a mixed gas at 1030°C and 100T.
orr for 55 minutes. As a result, the Si single crystal (grown crystal 37) reaches the corner diagonal to the corner where the seed crystal 36 is applied.
It also grew beyond the corner and onto a flat surface. The inside of the recess 31 was completely filled with Si single crystal (grown crystal 37).

Next, as shown in FIG. 3(F), the recess 3
The crystal portion that had grown beyond 1 was removed by polishing to obtain a flat Si single crystal 38. This polishing was selectively performed by "mechanical polishing". Under these conditions, colloidal silica having an average particle diameter of 0.01 μm was used as the abrasive grains, and a suspension of this in water was used as the abrasive. On the other hand, an abrasive cloth was placed on the substrate, and while an abrasive was injected little by little, the grown crystal 37 was rubbed against the abrasive cloth and polished. At this time, the pressure is 220g/cm2 and the temperature is 30-40
It was ℃.

When the polished surface of the grown crystal 37 reaches the surface of the fused silica substrate 30, the quartz under the above polishing conditions,
That is, since the polishing speed of SiO2 is significantly lower than that of Si, it is possible to recognize the completion of polishing.

By forming a plurality of crystal growth regions as described above, and performing crystal growth and polishing, an SOI can be formed in which both the in-plane and in-plane orientations are controlled between each crystal, and which is flat over the entire surface of the substrate. A crystal group was obtained.

The obtained crystal is oriented perpendicularly to the substrate (100
), and when measured by X-ray diffraction, no other orientation could be observed. Also, regarding the in-plane direction, EC
P (Electron Channeling Pat
tern), it was found that the crystal growth regions were aligned within ±12° with respect to the (100) equivalent direction along the sides of the concave portions. Example 2 First, as shown in FIG. 3A, a 4-inch Si wafer whose surface was oxidized to 1 μm was used as the base 30, and a rectangular shape of 15 μm in length and 4 μm in width was patterned on the oxidized surface. This portion was then etched by RIE to a depth of 0.3 μm to form a recess 31 as a crystal growth region.

Next, as shown in FIG. 3(B), a Ge film (original seed material 32') was deposited to a thickness of 0.2 μm by EB evaporation.
m deposited.

Next, as shown in FIG. 3(C), the recess 3
A 1.5 μm square area was patterned at one corner of 1, and the other parts were etched by RIE.
The remaining square area was used as the original seed 32.

Next, as shown in FIG. 3(D), the substrate 30 coated with the original Ge seeds 32 was heated for 750 min in a hydrogen atmosphere.
By heat-treating at 80 Torr for 3 minutes, the original seeds 32 were aggregated to obtain single-crystal Ge seed crystals 36.

Next, as shown in FIG. 3E, a Si crystal (grown crystal 37) was selectively grown in the vapor phase using the Ge seed crystal 36 as a starting point to a size of about 20 μm. The growth conditions at this time are ・SiH2 Cl2 /HCl/
H2 =0.53/1.05/100, 80Torr,
The temperature was 900°C for 60 minutes.

Next, as shown in FIG. 3(F), the Si crystal 37 was polished in the same manner as in Example 1 to form a flat SO
I crystal was obtained.

The obtained crystal is perpendicular to the substrate (100)
When measured by X-ray diffraction, no other orientation could be observed. Also, regarding the in-plane direction, ECP
When analyzed, it was found that they were aligned within ±10° with respect to the (100) equivalent direction along the side direction of the recess.

[0073]

Effects of the Invention As explained above, according to the method of the present invention, the three perpendicular faces of the corners in the concave portion of the substrate act well as faces equivalent to the stable orientation structure of the seed crystal formed by agglomeration. Therefore, it is possible to uniformly and easily form a single crystal having a single orientation in which there is no coexistence of multiple orientations other than the principal plane orientation over a large area.

Furthermore, the present invention has the advantage that it can be carried out at a lower temperature than the seed crystal formation method by melting, and that aggregation and nucleus growth can be carried out in a continuous process.

[0075] Since the unidirectional single crystal obtained by the present invention has excellent properties such as electrical properties, it can be favorably used in various devices such as electronic devices, optical devices, magnetic devices, and piezoelectric devices.

[Brief explanation of the drawing]

FIG. 1 is a schematic diagram illustrating a recess (crystal growth region) formed in a substrate and an original seed formed in the recess.

FIG. 2 is a partial cross-sectional view of the recess and the original seed illustrated in FIG. 1;

FIG. 3 is a schematic partial cross-sectional view illustrating each step of the method of the present invention.

[Explanation of symbols]

10 Base 11 Concavity (crystal growth region) 12 Original seeds 13　　　Concave side surface 14　　　Concave side surface 15 Bottom of recess 30 Base 31 Concavity (crystal growth region) 32 Original seeds 32' Original seed material 33　　　Concave side surface 35 Bottom of recess 36 Agglomerated single crystal seed crystal 37 Growing crystal

Claims (5)

[Claims] Claim 1: In a method of forming a crystal in which a single crystal seed crystal is placed on a substrate and a single crystal is grown from the seed crystal as a starting point, an angle where three faces consisting of a bottom face and two side faces touch each other as perpendicular faces. A concave portion having at least a portion is formed on the surface of the substrate, an original seed is placed at the corner so as to be in contact with the three sides, and the original seed is heat-treated to aggregate into a single crystalline seed. A method for forming a crystal, comprising forming a crystal and growing the single crystal using the single crystal seed crystal as a starting point. 2. The method for forming a crystal according to claim 1, wherein the substrate has at least a surface with a low nucleation density. 3. The method for forming a crystal according to claim 2, wherein the substrate has a surface made of at least an amorphous insulator. 4. The method for forming crystals according to claim 1, wherein the temperature of the heat treatment is below the melting point of the original seed. 5. The method for forming a crystal according to claim 1, wherein the heat treatment is performed in a hydrogen atmosphere.