JPH0465362B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (A) Technical Field The present invention relates to a demultiplexing element that can separate two lights of different wavelengths.

Optical communication uses light as a signal. In order to increase communication capacity, two or more optical signals with different wavelengths are transmitted through the same optical fiber. This is optical multiplex communication. Optical signals of different wavelengths do not interfere with each other within the optical fiber. Transmitted as an independent optical signal.

In this case, an element for separating two lights of different wavelengths from one optical fiber or medium is required at the output end. This is the demultiplexing element.

(b) Prior art The demultiplexing elements used in optical multiplexing communications include (a) those using angular dispersion elements such as diffraction gratings and plasmas, and (b) wavelength selection devices such as dielectric multilayer interference filters. These include those using reflective/transmissive films.

An example of something belonging to category (a) is shown in Figure 8. One input optical fiber 30 and a plurality of output optical fibers 31 are provided in parallel. The end face is flush with the other end and is fixed to one end of the focusing rod lens 32. A diffraction grating 33 is provided at the other end of the focusing rod lens. The light incident on the focusing rod lens 32 from the incident side optical fiber 30 hits the diffraction grating 33.

Light is diffracted, and the diffraction angle differs depending on the wavelength of the light. The light is diffracted into the output side optical fiber 31 in the direction of the diffraction angle corresponding to the wavelength λ of the light.

As the diffraction grating, a blazed diffraction grating with high diffraction efficiency that utilizes selective etching of a silicon single crystal is often used.

Since this demultiplexing element uses a diffraction grating, it has the advantage of being able to demultiplex optical signals of many wavelengths at once. For example, 20 channels of optical signals can be demultiplexed or multiplexed.

However, the positioning of the optical fiber is difficult and manufacturing is difficult. Another disadvantage is that the insertion loss is large.

FIG. 9 shows a demultiplexing element using a wavelength selective reflection/transmission film. Four focusing rod lenses 35 are used. Interference filter films 36 and 37 are provided between the two rod lenses. This is an interference filter made of dielectric multilayer films. It transmits light of certain wavelengths and reflects light of other wavelengths.

Two sets of rod lenses 35 sandwiching interference filter films 36 and 37 are bonded together at positions slightly shifted in a direction perpendicular to the axis.

Four optical fibers A, B, C, and D are fixed to the front and rear end surfaces of the rod lens.

Three types of light with wavelengths b, c, and d are transmitted through optical fiber A.
Suppose we enter from The interference filter 36 has a wavelength b
It transmits only the light of , and reflects the light of c and d. c, d
The light hits the second interference filter film 37. This filter film 37 reflects light c and transmits light d. Therefore, wavelengths b,
Lights c and d are emitted.

This demultiplexing element can only demultiplex about two or three optical signals. However, the insertion loss is small,
It is said that

(c) Problems with conventional technology All of the demultiplexing elements directly utilize the difference in wavelength of light to demultiplex light of different wavelengths. Therefore, the following problems arose.

(1) Extremely high precision is required to manufacture the device, making it expensive.

In the examples of FIGS. 8 and 9, the position of the output side optical fiber must be determined very strictly. This is because the position of the light emitted from the end face of the lens is determined by the wavelength. If it deviates from the normal position by the diameter of the optical fiber's core, no light will be emitted.

(2) The dependence on the angle of incidence is large and adjustment is difficult.
The optical path inside the lens is uniquely determined. If the incident angle deviates from the normal angle, the position of the output point at the output end will deviate.

(3) Since it demultiplexes spatially propagating light, it is basically not suitable for coupling with fiber.

(d) Purpose It is an object of the present invention to provide a demultiplexing element that can be mass-produced and at a low cost.

A second object of the present invention is to provide a demultiplexing element whose adjustment accuracy is gentle during use.

A third object of the present invention is to provide a demultiplexing element that can be easily coupled to an optical fiber.

(E) Waveguided optical beam splitter The waveguided optical beam splitter is a waveguided optical beam splitter by Enohara, Izutsu, and Sueda, “Waveguided optical beam splitter” by the Radiation Science Research Group.
It was first proposed in 1982RS-82-11.

FIG. 3 shows a plan view of the waveguide optical beam splitter.

This is an element with an X-shaped branch, which is a combination of a symmetrical Y-branch and an asymmetrical Y-branch.

It has a shape in which four single mode optical waveguides are coupled at one point.

Each of the optical waveguides that make up the branch is simply written as
It's called a branch. The left half branch and the branch are single mode optical waveguides with the same width. In other words, , is symmetrical in the vertical direction. At the joint Q,
forms an angle θ ₁ .

The branches W and N on the right half of the joint are asymmetrical. form an angle θ ₂ . Branch W is thicker than N. Although the branching angles θ ₁ and θ ₂ are drawn large for ease of understanding, they are actually sufficiently small angles.

Let the direction of light travel be the z-axis. In other words, branching,
The direction of the bisector of is the z-axis.

The x-axis is taken perpendicularly upward to the plane of the paper. Although the light travels along the z-axis direction, a small y-axis direction component is also included.

The optical waveguides, , W, and N are media with the same refractive index. The surrounding cladding is a lower index medium.

To simplify the problem, it is assumed that the x-axis direction is infinitely expanded and the refractive index does not change in the x-axis direction.

In other words, it is assumed that , W, and N are infinitely wide plate-shaped waveguides. Then, the differential coefficient with respect to the x direction becomes 0, and the Maxwell equation is simplified.

Furthermore, since the branching angles θ ₁ and θ ₂ are sufficiently small, each of the branches W, N, and W can be regarded as a five-layer optical waveguide when viewed in the y direction.

The five-layer optical waveguide here refers to the five layers of cladding, core, cladding, core, and cladding. If the cladding in the middle is thick, there is no interaction between the cores.

However, if the middle cladding is thin, there will be core-to-core interaction.

These optical waveguides are single mode. In other words, the width is narrow enough to allow only a single mode, the lowest mode of light, to pass through. With respect to light energy, it is a nodeless mode.

When the middle cladding is sufficiently thick, each core is considered to be an independent medium, so it is almost equivalent to a three-layer structure (cladding, core, cladding) slab waveguide. The lowest order mode can be expressed by the Hermitian function H ₀ (y).

This is true even if the core widths are different. However, the width of the core is required to be smaller than a certain value due to the single mode propagation conditions.

If the intermediate cladding is thin, core-to-core interaction will occur. This field is considered to be a five-layered slab railway. The wave function of light propagating through the two cores is expressed as H ₀ . However, regarding the phase,
There are cases where the two cores have the same phase and cases where they have opposite phases.

When the phases are the same, this is called an even mode. If the lowest order iodine wave function of the branch and branch is , then the even mode can be written as +.

When the phases are opposite, that is, when the phases differ by 180°, this is called an odd mode. Odd mode can be simply written as -.

If the structure is simply a five-layer structure with a thin intermediate cladding, the phase difference φ of the wave function is 0.
Not limited to 180°. Can take continuous random values.

However, since the branches merge at the junction Q, stricter conditions are imposed on the phase difference.

Since the width of the waveguide of the coupling portion Q is the sum of the widths of and, it passes the lowest mode and the second mode. In terms of Hermitian functions, H ₀ and H ₁
may be present at the joint.

FIG. 4 schematically shows changes in the wave function in even mode and odd mode.

It is assumed that the lowest order mode is excited in both branches. These lights approach each other little by little due to a slight divergence angle θ ₁ and eventually coalesce at the joint.

At the junction, a zero-order mode H ₀ and a first-order mode H ₁ exist. Nothing other than H ₀ and H ₁ exists.

Since θ ₁ is small, changes in the wave function occur continuously. In other words, it changes adiabatically. In order to become H ₀ mode at the junction, the wave function must be symmetrical at the branch. In other words, a wave function (with a phase difference of 0) as shown by + must exist at .

In FIG. 4, a mode that is symmetrical with respect to , is called an even mode because it is connected to the H ₀ mode at the coupling part.

On the other hand, the primary mode H ₁ at the junction
A wave function that is smoothly connected is represented by - with respect to a branch. this is
Since it is connected to H ₁ , it is called an odd mode.

Thus, there exists a single mode function for the branch, whose phase difference is random in the absence of a coupling part. However, since there is a coupling part and only two modes exist there, the phase difference is 0.
It is 180°.

Now, the phase constant β is an eigenvalue that describes the mode. This has a physical meaning as a wave number related to light and the propagation direction.

The phase constant is constant for one mode. The lower the mode, the larger it becomes. If this is not a real number, it is a radiation mode and does not propagate, so
Not considered.

The phase constant is a function of the core, cladding refractive index, and waveguide width.

The sum + and difference - of the wave functions mentioned earlier are simplified terms. When the intermediate cladding is narrow, one unified wave function must be considered as a five-layer waveguide.

Since the branches are symmetrical, even and odd modes have the same phase constant at points sufficiently far from the coupling. This is because it is the same as when the branch exists alone.

FIG. 5 shows changes in the phase constant in the optical waveguide of FIG. 4. The horizontal axis is the light propagation direction z. The middle part is the connecting part, the left half is the symmetric branch side, and the right half is the asymmetric branch side W, N side.

It is assumed that the sum of the widths of the branches is conserved. +
It is expressed by the formula =W+N. Here, is the width of the branch, not the wave function.

The phase constant is the same for the core, cladding, and refractive index.
The wider the waveguide, the greater the number of modes. If the width of the branch W is the widest, the phase constant T in the branch W is the largest. Then branch,
The common phase constant U in the branch N is smaller than the phase constant R in the branch N. Thus z=
The value of the phase constant β of −∞ and +∞ is determined. z=±
At ∞, no matter how small θ ₁ and θ ₂ are, as long as they are finite angles, there is no correlation between branches or between W and N. Therefore, at z=-∞, it becomes U, and +
At ∞ it becomes T and R.

Let us consider the connecting part Q.

The higher the mode order, the smaller the phase constant, and the smaller the waveguide width, the smaller the phase constant.

This is because the phase constant of the 0th-order mode in the coupled portion is wider and larger than the phase constant of any branch. This is expressed as V.

The phase constant of the first-order mode is much smaller than the phase constant V of the zero-order mode. Let this be S.

The even mode wave function can be symbolically expressed by +, but as the gap between the branches becomes narrower,
The phase constant approaches V.

This is because the wave function + continues to the zero-order mode function H ₀ in the coupling part Q.

The phase constant of the odd mode wave function - approaches S as the gap becomes narrower.

In this way, the degeneracy of the even mode and the odd mode disappears. Regarding the phase constant, even mode is UV,
The odd mode will draw a US curve.

Such a change in phase constant is reversible.

On the contrary, from the joint Q to the direction of the branch,
It is also possible to propagate light in H ₀ and H ₁ modes.
In this case, the even mode light becomes the zero-order mode in the branch with the same phase as +. Odd mode light has opposite phase 0 as -
Moving into the next mode. The phase constant approaches and degenerates as z→−∞.

It is not common to degenerate. It degenerates because the widths of the branches are the same.

Since the phase constant is degenerate, the energy of the 0th-order mode light starting from 0 is distributed in half to the branches. The energy of the first mode light starting from Q is also distributed in half to the branches.

The equidistribution law is established in this way, and this is because the phase constants are degenerate in the two branches. The phase constant degenerates because the waveguide widths are equal.

Now, the asymmetric branch side on the right side of the joint Q will be explained. It is assumed that the width of the branch W is wider than the width of the branch N.

All that is required is a width inequality W>N. It doesn't matter how many times larger one or the other has to be. The conditions are not strict. This is the important point.

If the light is in the same mode, the wider the width, the larger the phase constant. Since the branches W and N are single mode waveguides, only the zero-order mode light propagates.

Even if it is 0th mode light, W is wider, so
The phase constant of W is larger and becomes T. T is greater than U.

Since N is narrower, the phase constant R of N is smaller.

Since the width of the line is in the order of W→→N, there is a relationship of T>U>R (1) regarding the phase constant.

The phase constant V of the coupling portion is larger than T because it is the zero-order mode in the widest line.
The phase constant S is smaller than R. This is because the sum of the line widths is constant.

Assume that even mode light travels from point Q to branches W and N. In FIG. 4, the even mode wave function is shown on the right and the odd mode wave function is shown on the left at each point.

The wave function on the right at point Q is the even mode wave function. This is the H ₀ mode assuming that the bond is infinitely long. There is one peak,
There are no nodes (zeros).

The even mode phase constant is V. This is the branch W,
When entering N, it is an even function, so the phase difference is 0, which is 0.
Moving into the next mode. However, it is not distributed equally. The even mode energy gradually leaves the branch N and moves to the branch W. Z →
At ∞, all the energy is concentrated in the branch W.

When such a change is seen in FIG. 5, it corresponds to a curve from V to T. No change from V to R occurs.

The even mode wave function has a gradual change in slope. Therefore, the continuity condition of the electric field and magnetic field at the core and cladding boundaries in the narrower branch N cannot be satisfied, and they are localized toward the wider branch W.

If θ ₂ is small, such a change occurs adiabatically, so the conversion loss from V to T becomes almost zero.

Consider that odd mode light propagates from the coupling part Q to the right. This is the wave function shown on the left side of Q (Figure 4). There are two peaks, positive and negative, and one node (zero point). This corresponds to the first-order mode H ₁ when the coupling portion Q has this width and is infinitely long. The phase constant is S.

When the light enters the branches W and N, it is converted into zero-order mode light with a 180° phase difference. The branch N is shown as positive light, and the branch W is shown as negative zero-order mode light. Of course, the opposite is also possible.

Since it was primary mode light, in Q, the core,
The gradient at the interclad boundary is large. This slope is close to the slope at the core-clad boundary in the narrower branch N.

Therefore, the odd mode light fits into the narrow branch N,
Energy is concentrated on branch N.

Since the wider branch W is wide, imposing the boundary condition between the core and the cladding reduces the slope of the wave function, and the wave function of the odd mode with a steep slope is originally difficult to enter the branch W. The negative zero-order mode wave function gradually disappears.

After all, in the limit of z→+∞, all the odd mode light is localized in branch N. In terms of the phase constant, a change from S to R occurs.

The above discussion all assumes that θ ₂ is small and that the adiabatic approximation holds.

Actually, an independent wave function can be defined as Z→
Only in the limit of ±∞. Even at the joint,
Even in a region close to this, it does not mean that an independent wave function exists.

At the coupling part, an approximation is made to a slab line having a three-layer structure with a uniform width, and it is assumed that there are a zero-order mode H ₀ and a first-order mode H ₁ corresponding to this.

In other areas, a five-layer slab line with an infinite series of parallel claddings is approximated. This is an approximation that there is a cladding with a constant width separating each core. In this case, the phase constants of odd and even modes can be calculated, and the wave function can also be calculated.

The phase constant in Figure 5 is obtained by determining the phase constant of each mode as a function of the intermediate cladding width, and then calculating the relationship between the distance in the z-axis direction and the cladding width d (d = 2ztanθ ₂ /2).
It is just a rewriting of the variable to z using .

Therefore, intermediate phase constants, including V and S, are well-defined quantities.

However, if θ ₁ and θ ₂ are sufficiently small, the assumption of parallel slab lines holds true, and the phase constant changes adiabatically.

Such an adiabatic approximation assumption is valid as long as θ ₁ and θ ₂ are small.

Changes in asymmetric branching are also reversible. Since the branch N is narrow, the light propagating from the branch N to the coupling part Q has a large gradient at the boundary between the core and the cladding. Therefore, when a branch W approaches, it demands a negative wave function into the core of the branch W. phase difference
The zero-order mode of 180° grows in the branch W.
When the light reaches the coupling part, it becomes first-order mode, that is, odd-mode light.

If this is all there is to it, it is no different from a simple beam splitter.

The light entering from the branch excites both the zero-order mode and the first-order mode light at the coupling part. Therefore, almost the same amount of these optical energies enters the branch W and the branch N.

The same goes for the light entering from the branch, which enters both branches W and N.

Light entering from either branch is split into two and emitted.

On the contrary, the light entering from the branch W changes to the zero-order mode of the coupling part. The zero-order mode light is divided into two equal parts and emitted into branches. It doesn't change the fact that it's divided into two.

The light entering from the branch N changes to the first-order mode at the coupling part. The first-order mode light is divided into two equal parts, enters the branch, and is emitted from there.

In this way, when light enters from one branch, only two split lights are emitted from the opposite branch, so there is no difference from a simple beam splitter.

However, to the two branches on the symmetric branch side,
At the same time, if light with a phase difference of 0 or 180° can be input, only either the 0th-order mode or the 1st-order mode can exist in the coupling portion Q.

Two incident lights are required, and they must be simultaneous, and the phase difference must be 0 or 180°.

Figure 6 shows light of the same phase entering the branches simultaneously. The wave function can be indicated by +. Since it becomes the zero-order mode of the coupling part and thereafter becomes an even mode, it concentrates on the wider branch W.

FIG. 7 shows light of opposite phases being simultaneously incident on the branches. The wave function can be indicated by -. It becomes the first mode of the coupling part. Since it becomes an odd mode, it concentrates on the narrower branch N.

Comparing FIG. 6 and FIG. 7, it can be seen that this is a selective demultiplexing element.

When we talk about selectivity, we mean phase selectivity, not wavelength selectivity. In Fig. 6, the branch W is in the same phase, and the 7th
In the figure, light is separated into branches N with opposite phases.

(f) Structure of the invention In the above explanation, it has been assumed that the wavelength of light is constant.

The conditions imposed on the branch are: (1) Branch, , W, and N are single mode waveguides. There are no modes higher than the primary mode.

(2) Two modes, a zero-order mode and a first-order mode, can exist in the coupling part Q.

It is. Therefore, light with a wavelength that satisfies these two conditions has the same effect.

The inventor of the present invention has realized that it is possible to create a splitting element using such a waveguide type optical beam splitter.

For light Σ of a certain wavelength λ ₁ , the phase input to the branch is made to be in phase, and for light of a certain wavelength λ ₂ to Π, the phase input to the branch is made to be in opposite phase. Make it.

In this way, the light Σ is split into the branch W, and the light Π is split into the branch N.

So, the question is, how can we create two lights of the same phase? Since the phase difference must not change, a single light source must be used.

Since the wavelength must be constant, a laser becomes the light source. Then, the laser beam is divided into two parts,
Later, they will pass through the branches and merge at the connection point. It is required that the optical path difference between the two paths be constant and that the optical path difference be an integral multiple of the wavelength.

There is also the question of how to create two pieces of light with opposite phases. Since the phase difference and the wavelength must be constant, the same applies to using a laser as a light source, dividing it into two, and then passing the branch. Although they are combined at the coupling part, there is an optical path difference between the two paths, and the optical path difference must be (integer + 1/2) times the wavelength.

This means that a total of four passes are required. However, passing through four paths is difficult in itself,
This is undesirable because changes in the phase difference are likely to occur.

Therefore, make the path common and use only two paths.

Let these paths be K and L. K and L output light in the same phase with respect to the light Σ having the wavelength λ ₁ and in opposite phase to the light Σ having the wavelength λ ₂ . The optical path difference between K and L is an integral multiple of the wavelength for light Σ, and
For this purpose, the wavelength may be set to be a half-integer multiple of the wavelength.

This will be explained using drawings.

FIG. 1 is a plan view showing the configuration of a demultiplexing element of the present invention.

Since this is an assumption of a slab line, it is assumed that the refractive index does not change in the x direction.

The branching element includes a branch, a coupling part Q, a branch W,
It consists of a group of optical waveguides that partially include N.

An input path 10, branch paths 11, 12, for making the same light incident and splitting the same light before the branch,
Branch paths 13, 14, etc. are formed. The input path 10 and the branch paths 11 and 13 constitute a Y branch. The intersection angle θ ₃ is a sufficiently small angle. Y branch 1
Parallel branch paths 12 and 14 continue from the ends of lines 1 and 13. Following the parallel branch paths 12 and 14, there are the aforementioned branches and branches.

The branches intersect at the junction Q with a small intersection angle θ ₂ .

A wide branch W and a narrow branch N are formed so as to form a narrow angle θ ₂ from the joint Q.

Output paths 15 and 16 are continuous to the branch W and the branch N.

The branch points of Y branches 10, 11, and 13 are assumed to be P.
There are two paths from P to Q. One is a path consisting of branch paths 11 and 12 and branches. This is called the optical path K.

The other path is composed of branch paths 13 and 14. This is called the optical path L.

Let n be the refractive index of the core, and let Δl be the difference in length between the optical path K and the optical path L. The difference in optical path is nΔl. When light with wavelength λ passes through optical path K and optical path L from point P to point Q, the phase difference Δφ is expressed by Δφ=1π/λnΔl (2).

For light Σ with wavelength λ ₁ , the phase difference is an integral multiple of 2π. For light Π of wavelength λ ₂ , the phase difference is a half-integer multiple of 2π. That is, nΔl=λ ₁ m (3) nΔl=λ ₂ (s+1/2) (4) where m and s are integers. Here, it is assumed that the refractive indices n for λ ₁ and λ ₂ are the same, but if the difference between λ ₁ and λ ₂ is not very large, they are almost the same. Also, it is not difficult to distinguish between the refractive indices n ₁ and n ₂ .

If the conditions (3) and (4) are satisfied, the light Σ becomes light that is in phase with the branch immediately before the coupling part Q. This excites even modes at the junction. Therefore, among the asymmetric branches, the light advances to the branch W and is emitted from the output end 15.

The light Π becomes light of opposite phase at the branch immediately before the coupling part. This excites odd modes at the junction. Therefore, it progresses to the asymmetric branch N and exits from the output end 16.

Assuming that the locations of λ ₁ and λ ₂ are first determined, the value that can be freely determined in (3) and (4) is the difference Δl between the refractive index n and the optical path length. This must be given precisely with an accuracy on the order of a wavelength.

Δl can be controlled more easily.

A phase delay element having an electro-optic effect is provided in one or both of the optical paths K and L. Figure 2 shows an example of this.

Assume that part or all of the optical path is made of a crystal that has an electro-optic effect. An electrode of length l ₁ is pasted here, and a voltage V is applied between the electrodes. Electrode 1
7, 18, 19, and 20 are illustrated. The following is
This is an explanation of the case where there is one phase delay element.
It's the same thing even if there are two.

Thereby, the refractive index of this portion can be changed. The optical path length increment is the refractive index increment Δn
can be expressed by l ₁ Δn.

Instead of equations (3) and (4), nΔl+l ₁ Δn=λ ₁ m (5) nΔl+l ₁ Δn=λ ₂ (s+1/2) (6).

Since Δn is proportional to the applied voltage V, it is a controllable variable. Equations (5) and (6) are easier to satisfy.

In equations (3) and (4), wavelengths λ ₁ and λ ₂ are first determined. The value of this ratio λ ₁ /λ ₂ is determined with high accuracy. This is made to be the ratio of an integer m to a half-integer (s+1/2), or the ratio of an even number 2m to an odd number (2n+1). In this way, m and s are determined simultaneously. The core refractive index is
It may be determined by the amount of impurity doping and may be variable. However, in many cases, n is considered to be a constant, and in the end, an appropriate optical path difference Δl is determined for m and s.

It is clear from (3) and (4) that the error in nΔl must be less than 1/4 of the wavelength.

The amount that can be controlled by l ₁ Δn is expressed as the product of the maximum value of Δn and l ₁ . Even if the maximum value of Δn is small,
The longer l ₁ , the greater the controllable amount. Let us denote the range of l ₁ Δn by δ(l ₁ Δn). This can be called the maximum value.

If it is about δ(l ₁ Δn)λ ₁ λ ₂ /2(λ ₁ −λ ₂ ) (7), Δn is given so that it satisfies equations (5) and (6) even if there is an error in Δl. be able to.

There are things to be careful of.

3 to 7 and FIG. 1 are all based on the assumption that the core and cladding are continuous in the same cross section in the x direction of the slab line.

However, in FIG. 2, the line does not continue infinitely in the x direction, but instead becomes a thin thin film dielectric line in the x direction.

In reality, the assumption of no change in the x direction is unrealistic. Instead, it is a shallow dielectric branch in the x direction. The slab line assumption is to simplify calculations.

2 on the yz plane without actually expanding in the x direction
Create a dimensional dielectric branch line.

In this way, although the phase constant becomes a little smaller, it is possible to excite the even mode and the odd mode in the branches W and N in the same manner as described above.

The wave function in a Y-branch formed on a plane is often found on the assumption that the same medium exists along the x-axis (perpendicular to the plane of the paper).

According to FIG. 1, the same medium may be considered to be continuous perpendicular to the plane of the paper, or it may be considered to be a thin medium that spreads only on the yz plane.

In FIG. 2, it is necessary to provide electrodes, and the element is naturally thin and planar.

(G) Effect Light Σ with wavelength λ ₁ and light Π with wavelength λ ₂ enter the incident path 10.
Suppose that the incident occurs from Both lights are split in half at the branch point P. The power is divided into two equal parts.

Since the phase difference between the optical paths K and L is an integral multiple of 2π, the light Σ becomes the lowest order mode light with the same phase and the same phase in the branching.

This becomes zero-order mode light at the coupling part. Proceeding to the asymmetric branch W ₁ N, it becomes even mode light and becomes a thick branch W
progress to.

Since the phase difference between the optical paths K and L is a half-integer multiple of 2π, the light Π forms a combination of opposite phases in the branching. This becomes first-order mode light at the coupling part. Proceeding to the asymmetric branches W and N, the light becomes odd mode light and proceeds to the thin branch N.

In the end, the light Σ is branched into the branch W, and the light Π is branched into the branch n.

(h) Effects (1) Since there is no severe requirement for accuracy, mass production is easy and the price can be low.

Since an angular dispersion element or the like is not used to separate lights of different wavelengths, there are no strict conditions regarding the geometrical arrangement.

(2) Adjustment accuracy during use is gradual.

(3) Easy to connect with optical fiber. This is because a dielectric line consisting of a core and a cladding is used.

(4) Can be used as a demultiplexing element in optical wavelength multiplexing communications.

(k) Numerical example As a phase delay element with electro-optic effect,
Assume that LiNbO ₃ crystal is used.

When the electrode spacing is d and the voltage is V, the refractive index change Δn is given by Δn=1/2n ³ γV/d (8). γ is the electro-optic coefficient.

Since n=2.26 γ=32.2×10 ^-12 Δn=1.86×10 ^-10 V/d (9) The length l ₁ of the electrode and the distance d are respectively l ₁ = 5×10 ^-3 m d= Assuming 10×10 ^-6 m, l ₁ Δn=9.3×10 ^-8 V (10). λ ₁ = 1.3m (Σ light), λ ₂ = 1.55μm (Π light)
Assume that This is a wavelength often used in optical communications.

For Σ and Π, Σl ₁ Δn/λ ₁ = V/13.98 (11) Πl ₁ Δn/λ ₂ = V/16.67 (12) When Δl = 0 in equations (5) and (6) Assuming V=42 (volts), we obtain l ₁ Δn/λ ₁ =3.00 (13) l ₁ Δn/λ ₂ =2.51 (14). This corresponds to m=3 and s=2.

[Brief explanation of drawings]

FIG. 1 is a plan view showing the basic form of the demultiplexing element of the present invention. FIG. 2 is a plan view of the demultiplexing element of the present invention using an electro-optical element. FIG. 3 is a plan view of a waveguide beam splitter proposed by Izutsu et al. in the 1982 Radiation Science Study Group, RS-82-11 "Waveguide Beam Splitter". FIG. 4 is a diagram showing changes in the wave function transmitted to even mode and odd mode branches, , W, and N in a waveguide beam splitter. FIG. 5 is a graph illustrating the change in phase constant over the entire waveguide optical beam splitter. FIG. 6 is a diagram showing that light can be separated into branches W by guiding lights of the same phase to the branches. Figure 7 shows the branching of light with opposite phase.
A diagram showing that light can be separated into branches N by guiding the light to N. FIG. 8 is a side view of a known demultiplexing element using a diffraction grating. FIG. 9 is a side view of a known demultiplexing element using an interference filter. ...One of the symmetric branches, ...One of the symmetric branches, W...The wider branch of the asymmetric branches, N...
Narrow branch of the asymmetric branch, Q... joint, P...
...branch point, Σ...light with wavelength λ ₁ , Π...light with wavelength λ ₂ , 17-20... electrode.

Claims (1)

[Scope of Claims] 1. An input path in which light Σ with wavelength λ ₁ and light Π with wavelength λ ₂ are incident, and are symmetrically branched at a branch point P at the end of the input path, and at a coupling part Q. A single mode optical path K and a single mode optical path L having different optical path lengths are symmetrically coupled to each other,
It branches from the coupling part Q where the 0th-order mode and the 1st-order mode can exist, and becomes an asymmetrical branch path consisting of a wide single mode branch W and a narrow single mode branch N, and the The optical path difference between optical paths K and L is the wavelength λ ₁
A wavelength splitting element characterized in that the optical path difference between the optical paths K and L with respect to the light Π is a half-integer multiple of the wavelength λ ₂ . 2. The splitting element according to claim 1, wherein the optical path difference between the optical paths K and L is caused by a difference Δl in the actual length of the waveguides of the branched optical paths K and L. 3. Demultiplexing according to claim 1, wherein the optical path difference between the optical paths K and L is caused by a difference Δn in the refractive index of the waveguides of the branched optical paths K and L at a certain length _l1 . element. 4 All or part of the optical paths K and L are made of crystals having an electro-optic effect, and the difference in refractive index Δn is changed by applying a voltage. branching element.