JPS6338269A - End face type photodiode - Google Patents

Abstract

PURPOSE:To be able to input an input light from the end face of an element and to easily hybrid integrate it thereby to obtain a high sensitivity in an end face type photodiode by forming an optical guide structure made of an N-type semiconductor buffer layer, an N-type semiconductor optical absorption core layer and a P-type diffused layer. CONSTITUTION:The structure of an end face type photodiode is formed by sequentially laminating an N-type InP buffer layer 33 (3mum thick) having 10<16>/cm<3> of carrier concentration, an N-type InGaAsP optical absorption core layer 34 (2mum thick) having 10<14>/cm<3> of carrier concentration and a P-type diffused region 35 (Zn-diffused InP of 3mum thick) having 10<18>/cm<3> of carrier concentration on an N<+> type InP substrate 32 having 10<18>/cm<3> of carrier concentration. The layer 34 has larger refraction than the layer 33 disposed therearound, the region 35 and an N-type InP upper layer 42. Thus, a planar optical guide 43 is formed. N-type InP window layers (5mum of thick in (z) direction) having 10<15>/cm<3> of carrier concentration are provided to protect the ed face (boundaries cd and ef) at both ends of the guide 43.

Description

DETAILED DESCRIPTION OF THE INVENTION "Field of Industrial Application" The present invention relates to an edge-type photodiode for a hybrid optical integrated circuit.

``Conventional technology'' In recent years, optical components used in the fields of optical communication and optical information processing have been developed in bulk, based on conventional mirrors, lenses, and prisms, from the viewpoint of higher functionality, mass production, and miniaturization. It has evolved from shaped optical components to optical integrated circuits based on optical waveguides. Optical integrated circuits include monolithic optical integrated circuits in which a semiconductor film is laminated on a ■-■ group semiconductor substrate I- to form light-receiving and emitting elements and optical waveguides, and LiN 110.
There are hybrid optical integrated circuits in which various individual optical components such as light receiving and emitting elements are incorporated into a J optical waveguide or a 6F optical waveguide. Hybrid optical integrated circuits (J) have the characteristic of being highly functional and easy to obtain because they can be combined with optical components and leading wave circuits made from materials that are optimal for receiving, emitting, and leading waves.

PTN photodiodes and avalanche photodiodes have been proposed as receiving elements for use in this J, UNA hybrid optical integrated circuit. The driving voltage of the avalanche photodiode is as high as several tens of square meters, whereas the drive voltage of the PIN photodiode-F is as low as several square meters, making it suitable for hybrid optical integrated circuits.

FIG. 4 shows an InGaAs P-P IN photocoup (ode structure) as an example of a conventional planar photodiode used as a light-receiving element for a long wavelength band of 1 μm or more. Middle 1 is n+-1nP substrate, 2!
J: n1nP buffer layer, 3 is n-r n GaAs
I) Light absorption layer, 4 is n-1nP window layer, 5 is p-diffusion region, 6 is 5iaN4 insulating film that also serves as iJ surface protection, 7 is
- electrode pole, 8 is a lower electrode, 9 is a drive power source, 10 is a load resistor, 11 is a coupling capacitor, 12 is an output terminal, and 13 is input light. When the input light 13 is introduced into the PrN photodiode which is reverse biased by the driving power source 9, electrons and poles are generated in the n-InGaAsP light absorption layer 3 and become a photocurrent, which causes the load resistance 10 and the coupling. Output 'J: Ai (an electrical signal is output from 12) via the capacitor 11.

Above this -) planar l n Ga As P-1'
A conventional method for hybrid integration of an IN phytodiode into a leading wave circuit is shown in FIGS. 5(a) and 5(h).

In the figure, 14 is a planar type 1) I N phytodiode, 15
is a substrate, a 4° optical waveguide constituting a 16iJ optical waveguide circuit, 1
7 is a core, 18 is a crat, 19 is a reflective mirror block, 20 is a reflective surface, and 21 is a guided light.

Reflection mirror block 19 ('', Δ1, etc. is deposited on the 45° inclined surface of the glass block to form a reflection mirror. Reflection surface 2
has 0. Push one reflective mirror beam onto the end of the optical waveguide 1G and fix it, and then place a flat PIN photodiode I4 on a part of it with its light-receiving surface facing down. Hybrid integration of arranged configurations is being carried out. The guided light 21 is reflected by the reflecting surface 20 and enters the planar type 1) I N −/t todium auto 14 to become an electrical signal. 1. However, in this method, it is necessary to use the reflective mirror block [19] and the optical waveguide 16, the reflective mirror block 19, and the flat PIN photodiode 14 must be precisely aligned and fixed. Because of the necessity, the process is complicated and mass production is difficult, and the complicated structure makes it difficult to miniaturize.

As a way to solve these drawbacks, the edge-type PIN
A method of hybrid integration of photodiodes by direct bonding at the end of an optical waveguide is considered. Figure 6 (a
) is a perspective view for explaining the end face direct bonding, and FIG. 2(b) is a sectional view.

In the figure, 22 is an end face type PIN photodiode-1ζ, 23
is a low melting point metal thin film for fixing the photodiode (for example, A
u-Zn vapor deposited film). The end face type PIN photodiode 22 is a light receiving element that inputs light from the end face. After placing this end face type PIN photodiode 22 on the substrate 15 and aligning it with the optical waveguide 16, the end face type PIN photodiode 22 is heated and the low melting point metal film 23 is melted.
The IN diode 22 is fixed. Such a hybrid integration method has the advantage of being highly mass-producible and capable of miniaturization.

By the way, a dedicated film guide suitable for hybrid integration by such direct end-face bonding has not yet been developed. 4. In a simple two-way ping-pong optical communication system, AI Ga in the 0°8 μm wavelength band
There have been only reports of examples in which the As laser diode 1 is used as an edge-type photodiode. [person, A
lping and R, Tc11.100 Mbit
/s Sem1duplex 0ptical Fib
er TransmissionJ, Lightwav
e Tcch, Experimentimcnt Ijsing
GaAs/GaAlAs 1. aser Trans
ceivers LT-2, 663 (1984), ) Figure 7 shows the Al
GaAs C3P (Chanr+aled 5ubst
raLeplanar) shows the structure of a laser diode. In the figure, 24 is a fin-GaAs substrate, 25 is n-Δl
GaΔS buffer Y layer, 26 is AtGaAs active layer,
27 iJ: p-Δ1GaAs crat layer, 28 is n
-GaAs gap layer, 29 is p-diffused region, 30 (J
The upper 7u pole, 31, is a soil part electrode.

Since such a laser diode 1 is used by passing current in the forward direction, the resistance value of the element is low and it is weak against reverse bias, so it is used in a zero bias state. When light is incident on the GaAs active layer 26, a light receiving sensitivity of 0.15 A/W and a response speed of 100 Mb/s are reported.

I-Problems to be Solved and)] However, when such a laser diode is used as a photodiode, the light-absorbing AlGa
Since the As active layer 26 has a very thin thickness of 0.37 zm or more, it cannot sufficiently absorb input light by 4.0 mm, and has the drawback of poor light receiving sensitivity.

In addition, the AI GaAs active layer 26 is normally of P type, and the boundary with the n Δ1 GaAs buffer layer 25 (the seventh
A pn junction is formed at the boundary indicated by a and b in the figure) and acts as a capacitance. Such a p-n junction is directly under the p-diffusion region 29 as shown in the figure (instead of from the left end a to the right end 1 of the laser diode) over a long distance ll! , ]ζ
3,1 ill, which increases the capacitance due to the formation of
, as mentioned above, the laser diode cannot be reverse biased, and the thick depletion that can be obtained with a normal 4-layer auto There is no formation of a layer (a layer with extremely low density of electrons and poles), and l'l I
+The capacitance of the junction increases. -1-Large capacitance due to the reason in Section 21 reduces the response speed I (1(l
[1] The problem of limiting to about Ml+ / s1
We have described the case where a ΔlGaΔS laser diode is used as a photodiode.
Exactly the same problem will occur if a GaAs P laser is used for ITI.

This invention was made with these circumstances in mind.

1-1 of this invention is to be able to directly receive the emitted light from the end of the waveguide, to sufficiently absorb the emitted light in order to increase the light receiving sensitivity and to increase the response speed. The purpose of the present invention is to present an end-face type autotomate with a new structure that allows the application of reverse high voltage with low junction capacitance.

1-One stage for solving the problem (The edge-type photodiode of the first invention consists of an n-type buffer layer made of a semiconductor ((() material, an n-type light absorption=1a layer, and a p-type has a planar optical waveguide consisting of a diffusion region of It is characterized by being configured as follows.

The edge type photodiode of the second invention is a semiconductor +A
an n-type buffer layer consisting of f=4, a rectangular n-type light absorption core layer, a p-type diffusion region, a part of the buffer layer, the light absorption core layer, 3L and the diffusion region. Fill both sides of the channel optical waveguide consisting of an n-type buried layer (
1°, and a co-1' conductor window layer is provided at both ends of the channel optical waveguide (J, r), the end of the ir diffusion region is located inside the +'+if semiconductor window layer and the buried layer. It is characterized by being configured as follows.

In terms of structure, the edge type photodiode according to the present invention has a light absorption region having a core and a 4-waveguide structure so that light can be input from the edge, and a 1. The main features are that a rounded window layer is provided at the end and that the pn junction area is reduced, resulting in high light receiving sensitivity and high response speed. Conventional planar type
The light absorbing region and the window layer are simply laminated at λ 1 (−7,, l), which is suitable for directing light perpendicular to the element surface manually by 4 ゛. In this invention, the structure differs in that the light absorption region described in ``1'' has a structure in which the optical waveguide and the window layer are oriented in the lateral direction of the element. pn so that 1 is applied
The difference is that the junction region has a −1° loss from the optimal structure.Also, when the laser diode is used as an edge-type photodiode, the lateral propagation of the waveguide in the light absorption region is changed to a structure suitable for light reception. The reverse bias voltage is
Using a high-resistance semiconductor resistor so that i> is applied, the angle becomes 1"°4. 1 [Function 1 End type according to the present invention) t l-guiot (J,
, pn junction, is affected by the L capacitance 11t (j4-, and absorbs light with high sensitivity.

[Example] Hereinafter, an example of the present invention will be described based on FIGS. 1 to 3.

1 and 2 are rounded diagrams illustrating an embodiment of the first invention, in which figure (a) is a perspective view and figure (b) is a figure (
Cross-sectional view taken along line A-A' in a), same figure (c)
is a sectional view taken along line BB' in FIG.

In the figure, 32 is an n''-1nP substrate, 33 is an n-InP buffer layer, 3/I is an n-In Ga As P light absorption core layer, 35 is a P-diffusion region, 36 is an n-InP window layer, and 37 is an n-InP window layer. Si3N4 insulating film, 38 is an upper electrode, 39 is a lower electrode,
40 is a non-reflection coating film, 4X is input light, 42 is n
-TnP upper layer, 43 is a planar optical waveguide. Here, as the optical waveguide 16 shown in FIG.
Now, the case of hybrid integration of the end-face type fort diode for long wavelength band (wavelength 1 μm or more) according to the present invention shown in FIG. 1 will be explained.

The structure of the end face type foot diode 1 of this example is shown in Figure 1 (
In b), carrier concentration 1i 10”/cll
l” of n”1nPJil; Above the plate 32, an n-1nP buffer layer 3 of Geetria concentration I
3 (,'(ttm7), Gearia concentration 10 days/cm'
n-1nGaAsP light-absorbing core layer 34(2)im thick)
The p-diffusion region 35 (,'l Itm F; (Zn-diffused InP) having a 1-Year concentration IQ18/cm' is laminated in this order in a sawtooth structure.

Here, the absorption edge of the light absorption core layer is 1.55 μm, and n
I n o, 5eGa n, ++ΔS Q, 9.
+) o, fl11 light absorption core layer 34 was used. n-Tn
The GaAsP light absorption layer 34 is formed by surrounding n-1n
P buffer layer 33, p-diffusion region 35 and n-TnP
l = larger refraction than the sublayer 42 (see figure (c)); these form a planar optical waveguide 43; At both ends of this planar optical waveguide 43, a carrier concentration of 10' 5 is applied to protect the end faces (interfaces c, d and ef in the figure).
/ cm 3 n-In P window layer (thickness in X direction 5
μm) 36 is provided. Furthermore, in order to protect and electrically insulate the upper surface of the element, an Si3N+ insulating film 37 is formed on the n-TnP window layer 36 and on the end of the p-diffusion region 35.
An upper electrode 38 of Au--Zn is provided on the - side of the p- diffusion region 35, including a part of this 5isN4 insulating film 37. Further, a lower electrode 39 of Au-8n is provided on the bottom surface of the n-In P substrate 32. A non-reflective coating film 40 is deposited on the side where input light 41 is input.

If the refractive index of InP is Ns and the refractive index of air is Na, then
(Na −Ns)” has a refractive index value of 5 and a film thickness λ/[4
It is known in optics that the reflectance can be minimized when a film of (Na Ns )"') is added. Here, λ is the wavelength in vacuum. In device manufacturing, For use, Si with a refractive index of 1.789
An OX film was deposited to a thickness of 0.182 μm.

The structure of the p-diffusion region 35 is determined in consideration of reverse bias voltage resistance and reduction of capacitance of the pn junction. That is,
The p-diffusion region 35 penetrates into the n-InP window layer 36, thereby obtaining reverse bias voltage resistance. Since deterioration of crystallinity is unavoidable when forming the end of the planar optical waveguide 43 (interface between cd and er in the figure) by an etching process, the interface of the p-diffusion region 35 coincides with the interface of this end. In this case, the reverse bias voltage resistance of the pn junction formed at this interface deteriorates, making it impossible to apply a sufficient operating voltage. Therefore, n-TnP with good crystallinity
The p-diffusion region 35 is formed so that a pn junction is formed in the window layer 36, so that no reverse bias voltage is applied to the end of the planar optical waveguide 43. Specifically, the p- diffusion region was made to penetrate approximately 2 μm. In the lateral direction (X direction) of the element, in order to reduce the capacitance caused by the pn junction, as shown in Figure (c), n-1
++C; n- layered on the aAsP light absorption core layer 34-A
[Instead of diffusing Zn over the entire surface of the nP upper layer 42,
Keep the width to the minimum necessary. Specifically, the width was set to 20 μm. In addition, in the depth direction (X direction), n-1nG
The interface of the p diffusion region 35 is set to n-1nG so that a depletion layer is formed in the aΔsP light absorption core layer 34 by a pn junction.
The two surfaces of the aAsP light absorption core layer 34 are aligned with each other.

Next, the action will be explained.

The input light 41 passes through the non-reflection coating film 40 and n-l n
After passing through the P window layer 36 and being optically coupled to the planar optical waveguide 43, the light propagates through this waveguide while being gradually absorbed. Light absorption is performed only by the n-1nGaAsP light absorbing core layer 34.

The n-1nP buffer layer 33 and the p-diffusion region 35 are transparent. Electron-hole pairs generated by the light absorbed by the n-InGaAsP light absorption layer 34 become photocurrents, which are detected as electrical signals. The optical energy of input light 4I is PO
Then, the optical energy Pa absorbed in the optical waveguide is - α1, r'a = Po (1-a).

Here, α is the absorption coefficient in the planar optical waveguide 43,
(2) 7 is the length of the planar optical waveguide 43 (length ce in the figure). The absorption coefficient α is calculated from the field distribution of propagating light in the planar optical waveguide 43 and the absorption coefficient (approximately 1000/cm) of the n-rnGaAsP material itself, and is approximately 200/c.
It is m. By applying a sufficient reverse bias voltage, the n-In
The light-receiving sensitivity S (A/W) when the depletion layer is expanded over the entire thickness direction of the GaAsP light absorption layer 34 is given by -α1°S=q/hν(1-e). here,
q is the electron charge, 11ν('' photon energy.

Figure 2 shows the calculated value (solid line) and experimental value (white circle) of the light receiving sensitivity in the ideal state (light is completely absorbed and all becomes photocurrent, vertical shape e) when the optical waveguide length is changed. .

dotted line). As the input light, the light emitted from the quartz-based θν path described above was used. The reason why the sensitivity of the experimental value is lower is that some of the electron-hole pairs generated in the n-InGaAsP light absorption layer 34 recombine and do not contribute to the photocurrent. ,Responsible. In a normal Fort diode, when the light receiving sensitivity in the ideal state is less than 80%,
Being fucked. As can be seen from the figure, as the Brehner optical waveguide 43 is lengthened, the light-receiving sensitivity improves, but at the same time, the capacitance due to the p-n junction formed at the interface of the p-diffusion region 35 increases, so the response speed decreases. Become slow. Therefore, in actual device fabrication, the optical waveguide length is determined based on the balance between sensitivity and response speed. Specifically, sensitivity 0
．． 7 A/W, response speed IGb/s (+) In order to obtain n junction capacitance less than PF (capacitance added at the time of packaging: 0.5PF), the optical waveguide length was set to 90 μm to obtain the performance as designed. .

Next, the manufacturing method will be described.

On the (+00) plane n±Inp substrate 32, an n-In1) buffer layer, an n-InG
a Shrimp membrane is laminated in this order on the AsP light absorption core layer and the n-InPn part. Each of these layers has a carrier concentration of I
The shrimp are grown in a non-doped manner so that the O157cm is as low as 157 cm'. Next, in order to form the n-InP window layer 36, the above laminated shrimp film (potassium dichromate + acetic acid + water) is
n-1 along the <110> direction using an etching solution.
A groove corresponding to the nP window layer 36 is created. Next, this groove is filled with n-1nP by liquid phase epitaxy. Next, a Si:+Ila film is deposited on the negative side of the element by the CVD method, and an insulating film 37 of Si*N as shown in FIG. A pattern is formed using a buffered hydrofluoric acid etching solution. This 5lsN4
The insulating film 37 is ultimately used as an electrical insulating film to prevent the voltage from being applied to the end of the element, but at the same time it is used as a mask for forming the p- diffusion region 35. S! The 3N+ insulating film 37 is formed so as to cover the end of the planar optical waveguide 43 in consideration of the diffusion length of Zn. After forming the Si 3N 4 insulating film 37, the device is subjected to 7.
N and P persons were vacuum sealed in a glass ampoule, and Zn was diffused in an electric furnace. During this step, the p-diffusion layer formed at the bottom of the n-1nP substrate 32 by Zn-diffusion is removed by polishing.

Next, an 111-part electrode 38 is formed by depositing Δu-Zo 4 on the top of the element. Core 17 of optical waveguide 16 in FIG.
Due to the positioning U' between the n-1nGaΔsP light absorption core layer 34, the plating method may be used as needed.
Further, plate the nl-electrode with Au to a desired thickness. Next, Au-8n is deposited on the element 1z portion to form a lower electrode 39 (4). Next, n-1nP
To make the window layer 36 have the desired thickness (approximately 577 m in two directions): Cut into cleavages while searing. In the lateral direction (Y direction) of the element, the width of the hybrid integration hole (usually 50 to 100 μm) is also cut by cleavage. next,
A non-reflection coating film 40 of SiOx is deposited on the input light side.

The refractive index of SiOx is adjusted by the oxygen partial pressure within the Herger.

Can you manufacture an edge-type photodiode using the following process?

As a method for hybrid integration of this end face type photodiode at the end of the optical waveguide 16, the end face direct honding method described as the conventional technique with reference to FIG. 6 is used.

Although the edge type photodiode of the present invention has been described in 1-1 above, InGaAst) can be used as the material of the light absorption core layer. Also, the .j° method for each layer should of course be changed depending on the waveguide (14 prefectures) used in the leading wave circuit.

FIG. 3 is a diagram illustrating an embodiment of the second invention, and FIG. 3 (a) is a +J perspective view, and FIG.
) is a cross-sectional view taken along the line 8-Δ' of FIG.

The symbols in the figure are the same as those in FIG. 1 of the embodiment of the first invention described above, or 44 for the n l n Ga As l' bi-absorption core layer, 45 for the n-T n P buried layer, and 45 for the n-T n P buried layer. The optical waveguide 46 is different. The real l111i example is
In this case, the light emitted from the silica optical waveguide I6 (in the direction of the seedlings, or suitable for narrow cases such as 571 m width) Since it spreads greatly in the direction -, the first
As shown in the embodiment of the invention (FIG. 1(C)), a planar optical waveguide 43 without a lateral confinement effect has an effect of less than The energy decreases, resulting in a decrease in light-receiving sensitivity (Q). Therefore,
As shown in FIG. 3(c), by forming an n-InGaΔsP light absorption layer 44, a total absorption layer 44, and a six-channel optical waveguide 46 with a refractive index distribution in the lateral direction (y direction), the input light can be The light receiving sensitivity can be improved by confining energy directly below the p-diffusion region 35.

Manufacturing method (4) is almost the same as the embodiment of the first invention described above, but differs in the following points.

■ To form the n -1n P window layer 36, a laminated shrimp membrane (
n-1n P layer/n-In Ga As P light absorption layer/n-1n P flat layer/n-1nP substrate)
When forming the groove 114, a groove corresponding to the buried layer 45 in FIG. 3(c) is also formed at the same time. Next, by liquid phase epitaxial method, the n-1nP window layer 36 and the n-TnP
The buried layer 45 is simultaneously buried with n-1nP.

(2) The interface of the p- diffusion region 35 is n as shown in FIG. 3(c).
A Si3N4 insulating film 37 is formed within the -1n P buried layer 45 to be used as a mask for Zn diffusion. The reason is that crystallinity easily deteriorates (gh and i in the figure).
J) This is to prevent formation of Ni-11 heavy alloy.

In this way, this embodiment has the feature that high sensitivity can be obtained even when the width of the optical waveguide constituting the leading wave circuit is not large enough, although the lateral movement is slightly complicated and manufacturing is troublesome. .

Hereinafter, in the embodiments of the first and second inventions (J, end-face type using InP optical semiconductor for long wavelength band), the present invention is not limited to this.
Using a GaAs optical semiconductor, i.e., n-Ga
As substrate, n-A I GaΔS buffer layer, n-
For Zn diffusion into the GaAs optical absorption core layer, nΔlGaAs epitaxial film, a 1; filter diffusion layer, n-Al GaΔS window layer, n-AI GaAs buried layer, etc. are used to diffuse Zn into the short wavelength band (0 , 87zm) is also available. The manufacturing method is not limited to the liquid phase epitaxial method, but may also be used such as the organic metal vapor deposition method (MOCVD) or the molecular heel epitaxy method. In addition, it is obvious that the edge-type photodiode of the present invention can be applied to \Ibrid integration into silica-based optical waveguides (not J but other glass) waveguides, I-iNbOl optical waveguides, and semiconductor guideways. It is. .

[Effects of the Invention] As explained below, the edge photodiode according to the present invention has an optical waveguide structure consisting of an n-type semiconductor buffer layer, an n-type semiconductor buffer layer, an absorption core layer, and a p-type diffusion layer.
Therefore, human-powered light can be input from the element end face, and as a result,
It has the advantage of facilitating hybrid integration, and of being highly sensitive since the light is successively absorbed as it propagates through the optical waveguide.

In addition, semiconductor window layers are provided at both ends of the optical waveguide.
- By locating the pn junction formed by the diffusion region in the semiconductor window layer with good crystallinity, a large reverse bias voltage can be applied, resulting in high sensitivity and high response speed. There are advantages.

In addition, since the p-diffusion region can be narrowed to a width corresponding to the width of the input light, no pn is formed at the interface of the p-diffusion region.
There are 111 points that the capacitance due to bonding can be reduced by one or more and a high response speed can be obtained.4.

[Brief explanation of drawings]

FIG. 1 is a diagram showing the structure of an edge-type photodiode showing an embodiment of the first invention, and FIG. Figures 4 to 3 are diagrams showing the structure of an end-face type diode. Figure 4 shows a conventional planar foot diode. Figure 5 is a diagram showing a method for hybrid integration of planar photodiodes 1. (Figure 1 is for direct honing.)
Figure 7 is a diagram showing the rehybrid integration 4-filter method.
FIG. 1 is a diagram showing the structure of a laser diode 1 used as a laser diode. 3 2-n ” - T n P 3+'; board, 3 3
...n-1 n P buffer γ layer, 34...
・n-T n C: aΔsP light absorption=1γ layer, 35・
- p-diffusion region, 3 [i -n - 1 n P window layer, 37 ・Si
JN+ insulating film, 38... Upper electrode, 39 1st electrode, 40... Non-reflective coating film, 41... Human power light, 42... n-1nP upper layer,
43...Planar optical waveguide, 44...r+-1nGaΔsP light absorption core layer,
45...n-InP buried layer, 46...channel optical waveguide. Figure 6(b) Figure 7

Claims (1)

[Claims] 1. A planar optical waveguide comprising an n-type buffer layer made of a semiconductor material, an n-type light absorption core layer, a p-type diffusion region, and an n-type upper layer; 1. An edge type photodiode, characterized in that a semiconductor window layer is provided at both ends of a planar optical waveguide, and an end of the diffusion region is located inside the semiconductor window layer. 2. An n-type buffer layer made of a semiconductor material, a rectangular n-type light absorption core layer, a p-type diffusion region, a part of the buffer layer, the light absorption core layer, and the diffusion region. It has a channel optical waveguide consisting of an n-type buried layer filling both side surfaces, a semiconductor window layer is provided at both ends of the channel optical waveguide, and an end of the diffusion region is located inside the semiconductor window layer and inside the buried layer. An edge-type photodiode characterized in that it is configured to be located at.