JPH027920B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention relates to a liquid phase epitaxial (LPE) method for growing multiple epitaxial crystal layers of InGaAs, InGaAsP, InP, etc. on an InP substrate. Long-wavelength (1.0 to 1.6 μm) optical communications are currently attracting attention, and the light source, the InP-InGaAsP double heterojunction laser, is progressing rapidly. On the other hand, the current situation is that there is no satisfactory photodetector in the long wavelength band. The main oscillation wavelength range of GaAs-GaAlAs lasers is
0.8μm to 0.87μm, silicon is used as a photodetector
PD or APD is the most widely used and shows excellent characteristics, but it is difficult to detect light in the wavelength range of 1 μm or more, and from 1.1 μm where optical fiber has good transmission characteristics.
It cannot be used in the 15μm wavelength range. Also,
Although Ge-APDs can be used for wavelengths of 1.1 μm or more, they have large dark currents and excessive noise, making them undesirable as photodetectors for optical communications, so -compound semiconductor APDs are required. Currently, for this 1.1μm to 1.5μm wavelength range,
There are reports of mesa structures having a light absorption region and a multiplication region in a semiconductor region of the same composition made of InGaAs, InGaAsP, GaAsSb, etc., but all of these have low multiplication and large dark current. However, an innovative APD structure has been proposed that can perform high reverse bias operation with low noise. The first is Tokugansho.
Specification of Japanese Patent Application No. 53-87850 (Japanese Unexamined Patent Publication No. 55-13990)
13957), it is based on a heterostructure in which a light-transmitting window is formed on the light-absorbing layer. Application No. 54-39169 (Unexamined Japanese Patent Publication
55-132079). FIG. 1 is a cross-sectional view of the structure shown in Japanese Patent Application No. 54-39169 (Japanese Unexamined Patent Publication No. 55-132079) "Semiconductor Device 8".
This is an example of fabrication using InGaAsP-based materials.
First, an n ⁺ InP layer 1 with a thickness of several μm is formed on the n ⁺ −InP substrate 11Y by liquid phase epitaxial (LPE) method.
2, then a film thickness of 5 μm and an impurity concentration of 2×10 ¹⁶
cm ^-3 n-type In _0.79 Ga _0.21 As _0.47 P _0.53 layer 13 (hereinafter
Abbreviated as InGaAsP. ) Furthermore, impurity concentration 1×
An n-type InP layer 14 of 10 ¹⁶ cm ^-3 is epitaxially grown. Next, a selective diffusion mask 15 of Si ₃ N ₄ , SiO ₂ or the like is applied, and Cd diffusion is performed to obtain a p-type region 16 and a pn junction surface 17. Furthermore, an insulating Si ₃ N ₄ or SiO ₂ film 151 is formed again, and after patterning an electrode extraction window, a p-type electrode 19 is formed, and an n-type electrode 20 is formed on the back surface of the InP substrate 11. The structural features of the APD made in this way are p
- The n-junction 17 is in the InP layer 14, and the InGaAsP layer 13 is only formed when a reverse bias is applied.
The reason is that it is located so far that a depletion layer spreads inside it. It is detailed in the aforementioned Japanese Patent Application No. 54-39169 (Japanese Unexamined Patent Publication No. 55-132079) that by doing so, an APD having excellent breakdown characteristics can be obtained. As a photodetector covering the entire long wavelength range (1.0 to 16 μm), a structure using an InGaAs light absorption layer epitaxially grown on InP is preferable.
That is, the InGaAsP layer 13 shown in FIG. 1 may be replaced with an InGaAs layer. However, InGaAs
When InP is grown on top by LPE, it is difficult to obtain a heterostructure because the InP growth solution causes the InGaAs to melt back. Therefore, a layer structure as shown in the cross section in FIG. 2, that is, an InGaAsP anti-meltback layer 22 is formed on the InGaAs layer 221 provided on the InP layer 12, and then an InP layer 23 is formed. is necessary to obtain the desired layer structure. However, forming the InGaAsP layer 22 for anti-meltback is also extremely difficult. It is extremely difficult to prevent this melt-back phenomenon. i.e. −0.5
In the process of slow cooling at a cooling rate of less than °C/min
InGaAs layer 21 is grown, and then InGaAsP layer 2 is grown.
When InGaAs layer 21 is grown, the InGaAs layer 21 usually disappears due to etchback. Also, the cooling rate is −0.5
If the temperature is set at ℃/min or more, the etch will be mild. That is, FIG. 3 shows an angle-polished cross section of an InGaAsP layer 22 with a forbidden band width of 0.95 eV epitaxially grown on an InGaAs layer 21. An interfacial transition layer 31 due to melt back exists, and a good interface is observed. is not obtained. The thickness Δd of the interface transition layer 31 shown in FIG. 3 is ~
Up to 3000A. The present invention provides a method for growing an InGaAsP layer on an InGaAs layer by LPE without melt back, without an interfacial transition layer, and with an extremely flat interface between the InGaAs layer and the InGaAsP layer. . The gist of the present invention is to heat the film at 620°C to 650°C on InGaAs crystal.
The forbidden band width is less than 1.05 eV at the growth starting temperature of
When epitaxially growing an InGaAsP layer, the growth rate is calculated by dividing the growth thickness by the square root of the growth time.
The method is characterized by liquid phase epitaxial growth at a rate of 0.2 μm/sec ^1/2 or more. The numerical basis shown in the gist of the present invention will be explained below using a specific example. In the structure shown in FIG. 2, it is necessary that the depletion layer of the pn junction 17 created by Cd diffusion exceeds the InGaAsP anti-melt back layer 22. That is, during reverse bias operation as a photodetector, the depletion layer must expand by at least d ₁ +d ₂ . Since the n-type impurity concentration obtained with InP, InGaAsP, and InGaAs is 1 to 2×10 ¹⁶ cm ⁻³ , the width of the depletion layer at the reverse bias breakdown voltage is considered to be 4 to 2 μm. therefore inevitably
The thickness of the InGaAsP layer 22 is ~1 μm or less. Figure 4 shows the growth thickness d of the InGaAsP layer with a forbidden band width of 0.95 eV obtained at a temperature of 640°C and a cooling rate of 0.2°C/min.
In the case of d1μm, the relationship is d∝√, and the growth is under diffusion-controlled growth conditions. Of course, the experimental result in Figure 4 is 0.2/
Since it is not possible to form an InGaAsP layer on an InGaAs layer under conditions of Naturally, if the conditions are d1 μm, diffusion law growth will occur, and it is appropriate to use the growth rate d/√ as a parameter to describe the growth rate of the InGaAsP layer. It is qualitatively considered that melt back does not occur when the rate of growth is faster than the rate of dissolution.
Therefore, in order to change the growth rate d/√, the fifth
Crystal growth was performed on a sliding continuous epitaxial board using the temperature program shown in the figure. InGaAs, InGaAsP for sliding boats,
Each growth solution for the InP layer was 4 mm thick, and the composition of each solution was as shown in the table.

[Table] First, the temperature program will be explained. After preparing each growth solution together with the substrate InP in a boat, the boat is placed in a reaction tube and the temperature is raised to T _H , and then maintained from point A to point B for a certain period of time t _H =2 Hrs. After this, from point B, cooling rate α = 1~5℃/
Perform rapid cooling for min, and reduce the cooling rate to β at point C of 640°C.
= 0.2℃/min and at the same time onto the InP substrate.
The InGaAsP growth solution was brought into contact with the InGaAsP layer for t ₁ = 2 min, and then, at point D, the InGaAsP growth solution was transferred onto the substrate InP on which InGaAs was grown, and the InGaAsP layer was grown for t ₂ = 20 sec.
After growth, move the InP growth solution at point E, grow the InP layer for t ₃ = 40 min to growth point F, remove the InP growth solution from the growth-finished substrate, and immediately cool it down to continue continuous LPE growth. finish. In the above explanation of the temperature program, the temperature T _H changed from 645°C to 655°C. The growth rate d/√ changes by changing the temperature T _H and the cooling rate α. Figure 6 shows the growth rate d/√ and the interfacial transition layer △ shown in Figure 3.
It depicts the relationship between d. The forbidden band width Eg of the InGaAsP layer is shown when it is 0.95 eV (○ mark) and when it is leV (● mark). increase the growth rate d/t
If it is 0.2 μm/sec ^1/2 or more, Δd=200 Å or less can be obtained. This value of △d=200Å or less is the third
The angle polishing method shown in the figure is at the limit of measurement, and there is substantially no interfacial transition layer due to melt back. In this example, the conditions for the growth rate d/√ to be 0.2 μm/sec ^1/2 or more are obtained when the cooling rate α is 3° C./min and the temperature T _H is 650° C. or more. The conditions for growth rate d/√0.2μm/sec ^1/2 are
This is necessary when growing InGaAsP on InGaAs at a growth starting temperature of 620 to 650°C for the InGaAsP layer, but this does not apply at low temperatures below 620°C. However, when the growth temperature is lowered to below 620℃, InP
It is difficult to reduce the n-type impurity concentration of InGaAsP and InGaAs layers. Also, shown in the table
The forbidden band width of the InGaAs layer obtained from the InGaAsP growth solution composition is ~1.0 eV, but the growth rate is d/√
To grow an InGaAsP layer on InGaAs without melt back under the condition of 0.2 μm/sec ^1/2, it is possible to grow an InGaAsP layer with a forbidden band width of 10.5 eV or less.

[Brief explanation of drawings]

Figures 1 and 2 show cross sections of a photodetector having a p-n junction in InP. In Figure 1, an InGaAsP layer 18 is used as the light absorption layer.
Figure 2 shows a structure in which an anti-melt back InGaAsP layer 22 is arranged with the InGaAs layer 21 as an absorption layer, and Figure 3 shows the angle of the layer cross section showing the interface where melt back occurs when InGaAsP is formed on InGaAs. Micrograph showing the polished surface, Figure 4 is InGaAsP
The relationship between growth time (t) and growth thickness (d) when growing a layer is shown. FIG. 5 shows a temperature program in liquid phase epitaxial growth, where the horizontal axis represents time and the vertical axis represents temperature. FIG. 6 shows the relationship between the interfacial transition layer Δd produced by etchback and the growth rate d/√. In the figure, 21 is an InGaAs layer, 22 is an InGaAsP layer, 2
3 represents the InP layer.

Claims (1)

[Claims] 1 When an InGaAsP layer with a forbidden band width of 1.05 eV or less is epitaxially grown on an InGaAs crystal at a growth starting temperature of 620°C to 650°C, the growth rate is 0.2 μm/sec ^1/ as the growth thickness divided by the square root of the growth time. A liquid phase epitaxial growth method characterized by liquid phase epitaxial growth with a value of ² or more.