JPH05224043A - Parallel optical waveguide - Google Patents

Abstract

PURPOSE:To provide an optical parallel processing method which reduces optical crosstalk and light loss in an optical coupling part constituting an optical computer and a device which performs parallel information processing. CONSTITUTION:Projection light from an LD array 8 is collimated by a microlens array 7 into collimated light, and position adjustment and fixation are performs so that light beams from the respective LD array elements are made incident on the respective waveguides of a parallel optical waveguide 5. The input light 6 is reflected successively at right angles to the waveguide substrate surface by slanting grooves in the respective waveguides and made incident on a surface type optical amplifying element array 4. After optical amplifying and modulating operation by this surface type optical amplifying element array 4, the light is made incident on the parallel optical waveguides 3 and converged by the waveguides, and then made incident on a PD array 2. Those parallel optical waveguides are used to compactly constitute the device which performs product sum arithmetic and parallel optical conversion using the optical amplification and modulation.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optical interconnect, and more particularly to a parallel optical waveguide as a basic component for optically parallel processing information between electronic devices (LSI etc.). The present invention is also used as a basic component of optical neurocomputing having an optical arithmetic processing unit and a learning function in optical computing.

[0002]

2. Description of the Related Art In the conventional supercomputer technology, the maximum operation speed Max is about 1 Gop / s, and even a super parallel machine represented by a connection machine has a parallel degree of about 10 ⁴ . Optical computers and neurocomputers have been proposed as future extensions of this computing speed and parallelism. The most basic of these is the optical implementation of Hopfield by N. H. Farhat. Model, Applied Optics Volume 24
Pages 10 to 1469 to 1475 (Optical I
mplementationof Hopfield Model, Appl.Opt., Vol.2
4, No. 10, pp. 1469-1475).

This is the input vector Χ (n) and the storage matrix W.
This is for performing the product-sum operation with (n, m) by light. The input vector is expressed by the light intensity of the light emitting element, is expanded in one direction by a cylindrical lens, and the storage matrix (for example, transmittance) is modulated. The structure is such that the output vector Y (m) is obtained by converging in the orthogonal direction. Using this basic structure, the output vector is multiplied by a threshold function and then fed back to the input vector side to perform optical computing.

Further, as an example of applying the Hopfield model to an optical associative memory, details are reported in Technical Research Report OQE87-174 (1988), pages 39 to 45 of the Institute of Electronics, Information and Communication Engineers.

A memory correlation matrix is realized by multiplexing and driving different frequencies or phases for the positive and negative values of the stored information, or by alternately time-sharing the positive and negative values. As an example of the patent application publication No. There is a proposal by No. 2-45815.

All of these are examples in which the input vector, the storage vector, and the output vector are realized by spatial optical coupling.

[0007]

[Problems to be Solved by the Invention]

1. In the above-mentioned conventional technique using optical spatial coupling such as the Hopfield type mutual coupling type model, the device increases in volume as the degree of parallelism increases, and stable operation of the element is achieved when each optical component is mounted. No consideration is given to the installation method of the cooling mechanism for this purpose. An object of the present invention is to provide a mounting method in which the volume of the apparatus does not increase so much even if the parallelism is increased and the cooling mechanism can be easily mounted.

2. Further, the conventional method of branching light in parallel using an optical system has a problem that optical coupling loss is large and light crosstalk is likely to occur. An object of the present invention is to provide a method of optically coupling light from an optical element without loss and branching in parallel in a matrix.

3. It is another object of the present invention to provide a compact parallel optical waveguide substrate having an optical / electronic functional device and having high functionality.

[0010]

[Means for Solving the Problems]

1. In order to achieve the above-mentioned object, the present invention uses a substrate transparent to signals, provides a rectangular waveguide in the input light direction by cutting or the like, and forms an oblique groove in a direction perpendicular to the waveguide. .. The diagonal groove becomes deeper from the input light side,
The input light can be reflected in parallel in the vertical direction of the substrate.

2. Further, in the present invention, the waveguide and the grooved portion are metalized to improve the efficiency of confining light and the reflectance, thereby increasing the efficiency of parallel branched light.

3. Further, an electronic circuit electrode pattern is formed on the back surface of this parallel waveguide substrate so that the optical / electronic functional device can be mounted, and is integrated with the optical / electronic functional device.

[0013]

[Action]

1. By the parallel optical waveguide substrate that branches in the direction perpendicular to the input signal direction, the multi-parallel input signals can be branched into matrix signals without requiring a large space space. Further, since the parallel optical waveguide substrate and the planar functional device can be surface-mounted on the substrate in a laminated form, a cooling mechanism can be easily attached.

2. It is apparent that the efficiency of confining light in the waveguide and the reflectance can be increased by attaching metallization or the like to the surface of the parallel optical waveguide substrate to increase the reflectance. At this time, since there is almost no crosstalk between the rectangular waveguides, there is almost no crosstalk between parallel signal lights from the array element, which is important when performing the Hopfield type sum-of-products calculation.

3. In addition, the light on the back side of the parallel optical waveguide substrate
Since the electrode pattern for the electronic functional device can be provided, it can be integrated with the optical / electronic functional device, and a parallel optical waveguide substrate having a high function can be provided.

[0016]

1 shows an example applied to a parallel waveguide type optical arithmetic unit as a first embodiment of the present invention.

The parallel optical waveguides 3 and 5 are formed by using borosilicate glass, which is a transparent material for signals, and forming rectangular portions by groove processing parallel to the signal light transmission direction to form optical waveguides. .. This waveguide has a rectangular shape with a height of 200 μm and a width of 150 μm. The pitch between the rectangular waveguides is 250 μm. Further, oblique groove processing of 45 degrees is performed in the direction perpendicular to the rectangular waveguide. This depth gradually increases by 50 μm from the input light side to form four grooves. A gold vapor deposition film is formed on the surface of the rectangular waveguide and the diagonal groove processing part in an amount of about 0.
The thickness is 5 μm, and the efficiency of optical confinement in the waveguide is improved.

The parallel optical waveguides 3 and 5 sandwich the surface amplification element array 4 with the waveguide forming side as a back surface, and the transmission directions of the rectangular waveguides are orthogonal to each other, and Au / Sn fusion bonding is performed by flip chip bonding. is doing. The parallel optical waveguide 3 is further bonded to the alumina substrate 1 by Pb / Sn. On this alumina substrate 1, P
The D array 2, the LD array 8 and the heat sink 9 are bonded with a high melting point Pb / Sn solder as shown in FIG.

The input light 6 from the LD array 8 becomes collimated light by the microlens array 7, and the position of the light of each LD array element is adjusted so that the light of each LD array element corresponds to each waveguide of the parallel optical waveguide 5. After that, the microlens array 7 is L
It is fixed on the D array 8. The input light 6 is reflected by the oblique groove in each waveguide in a direction perpendicular to the waveguide substrate surface. The reflected light is incident on the surface-type optical amplification element array 4, is subjected to the optical amplification and modulation action, and then is incident on the parallel optical waveguide substrate 3, is condensed for each waveguide, and becomes the output light 8 and is incident on the PD array 2. It

The product-sum operation is performed by using the light intensity of each element from the LD array 8 as an input vector, the surface amplification element array 4 as an operation matrix, and the light receiving sensitivity of each element of the PD array 2 as an output vector. Was completed. Similarly, by applying the amplification / attenuation action of each element of the surface amplification element array 4, optical communication between the input Xi and the output Yj, that is, optical exchange can be performed.

Next, as a second embodiment, an example in which a parallel optical waveguide is applied to information processing between electronic devices such as LSI will be described.

FIG. 2 shows a parallel waveguide type optoelectronic information processing apparatus. The parallel optical waveguide substrate 15 is the same as that shown in FIG. An electrode pattern 16 of Cr / Pt / Au is formed on the surface of the parallel optical waveguide substrate 15 opposite to the waveguide portion, and the microprocessor LSI 14 is flip-chip bonded to the electrode pattern by Au / Sn. Four of the LSIs 14 are mounted in parallel on the parallel optical waveguide 15, and one PD array 13 corresponds to one of these LSIs, and flip-chip bonding is performed in parallel by Au / Sn. The signal light 19 emitted from the LD array 18 travels in parallel for each waveguide similarly to that shown in FIG.
Incident in parallel. This light is converted into an electric signal and transmitted to an adjacent LSI as an input signal. The output signal electrically processed by the LSI 14 is taken out through the electrode pattern 16 of the parallel waveguide 15.

The light of the LD array 18 is aligned with the microlens array 17 so as to become collimated light, and then fixed on the LD array 18 on the substrate. Similarly, the parallel optical waveguide 1 in which the LSI 14 and the PD array 13 are bonded
5 is bonded to the substrate 11 via the spacer 12 with Pb / Sn solder.

As a result, the parallel waveguide type optoelectronic information processing apparatus is compact and can be cooled without taking a spatial space, and parallel signal processing to the processor LSI has become possible.

[0025]

By using the parallel waveguide substrate of the present invention, the device for performing the sum-of-products calculation and the optical computing can be made more compact as compared with the conventional space-coupling system, and the optical loss with excellent crosstalk characteristics can be achieved. It is possible to realize a device with less power consumption. Further, by using the parallel optical waveguide substrate, signals to a large number of adjacent LSIs can be operated in parallel. In this method, the optical / electronic device is integrated with the parallel optical waveguide substrate and further mounted on the substrate,
A cooling mechanism can be attached and a highly reliable device can be realized.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a parallel waveguide type optical arithmetic unit showing an embodiment of the present invention.

FIG. 2 is a configuration diagram of a parallel waveguide type optoelectronic information processing apparatus showing an embodiment of the present invention.

[Explanation of symbols]

1 ... Alumina substrate, 2 ... PD array, 3, 5 ... Parallel optical waveguide substrate, 4 ... Planar optical amplification element array, 6 ... Input signal light, 7 ... Microlens array, 8 ... LD array, 9 ...
Heat sink, 10 ... Output light, 11 ... Substrate, 12 ... Spacer, 13 ... PD array, 14 ... Microprocessor L
SI, 15 ... Parallel optical waveguide substrate, 16 ... Electrode pattern,
17 ... Microlens array, 18 ... LD array, 19
… Signal light.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Yoshiko Matsuoka 1-280 Higashi Koikeku, Kokubunji, Tokyo Inside Central Research Laboratory, Hitachi, Ltd. (72) Inventor Aki 1-280 Higashi Koikeku, Kokubunji, Tokyo Hitachi Ltd. Central Research Laboratory (72) Inventor Yuichi Ono 1-280, Higashi Koikekubo, Kokubunji City, Tokyo Inside Hitachi Central Research Laboratory

Claims (4)

[Claims] 1. A parallel optical waveguide comprising a substrate transparent to an input signal light, a rectangular waveguide provided on the substrate, and an oblique groove formed in a direction perpendicular to the waveguide. Waveguide. 2. The oblique groove is gradually deepened from the input light side,
The parallel optical waveguide according to claim 1, wherein input light can be reflected in a direction perpendicular to the parallel optical waveguide substrate. 3. The parallel optical waveguide according to claim 1, wherein the waveguide portion and the oblique groove processed portion are metallized to improve light confinement efficiency and reflectance. 4. The parallel optical waveguide according to claim 1, wherein an electronic circuit electrode pattern capable of mounting an optoelectronic functional device is formed on the back surface of the substrate.