Distributed Resonant Cavity Light Beam Modulator

The invention relates to optical devices for the modulation of light beams by the control of light transmission, and in particular to an optical device which blocks light transmission by resonance reflection when the light frequency is substantially equal to the device resonance frequency and which otherwise freely transmits incident light when the device resonance frequency and the light frequency differ.

The advantageous use of light, the visible electromagnetic radiation, as the signal medium in communications and in computer systems is well recognized. Of particular importance is coherent light such as that produced by lasers because of propagation characteristics and bandwidth. Modulation of a laser beam is a rapid and efficient means of information transmission.

Various devices for modulation of a light beam are currently in use. The include semiconductor devices utilizing quantum wells fabricated by molecular beam pitaxy in 3-5 compounds and other switching devices. 3-5 compound, quantum well devices have excellent switching characteristics, but their manufacture involves complicated and expensive technologies. There exists a need for a light modulator which is simple in construction and inexpensive which can readily be fabricated in the form of a 2-dimensional array, and, at the same time, which is capable of switching light on and off at gigahertz rate, a rate much larger than that for most conventional devices.

A review of the prior art reveals one device which involves a grating, namely, a surface emitting distributed feedback laser which uses a grating to support the necessary feedback for laser operation. See G. E. Evans, N. W. Carlson, J. M. Hammer, R. A. Bartolini, Laser Focus World, November 1988, p.97; and W. Streifer, D. R. Seifres, R. D. Burnham, IEEE J. Quantum Electronics, Vol. QE12, No. 7, July 1976. In contrast, the present invention provides for the modulation and control of a beam produced by a laser.

The instant invention is an optical device for the modulation of light transmission through the device by coupling or decoupling the radiation with a resonant mode of the device. For an incident wavelength or frequency equal to the respective resonant wavelength or frequency of the invention, the incident radiation is dominated by resonant reflection and transmission through the apparatus is precluded. For all other values of incident wavelength or frequency, the device is substantially transparent.

In accordance with the present invention, the resonant wavelength strongly depends upon the geometry of the comprising components, the composition of the components, the electrical environment as imposed by external sources of electromotive force (EMF) and the incident radiation. Control by externally applied EMF or radiation is exceptionally fast.

The simplest embodiment of the present invention comprises a first transparent cover layer to which the incident beam is directed perpendicular to the layer; a second layer which is a grating structure of alternate bars of different index of refraction disposed under the cover layer and a third transparent resonant guide layer encapsulating the grating and providing the guide channel for Bragg diffraction in a direction perpendicular to both the grating bars and the incident radiation beam.

The support for the layers comes from the substrate on which they are grown or deposited. Some of the embodiments have the grating bars or teeth exposed to air. Some of the embodiments have electrical contacts through a transparent cover electrode, so a generic cover layer is reasonable. Including a cover layer also enables covering the device with multilayer antireflection coatings, which would improve device performance.

The grating bars have different indices of refraction. There are no opaque regions. This produces more efficient diffraction. Typically, the grating grooves are etched into the guide layer, although a separate grating layer with bars made from a different material than the guide layer can be used. Some embodiments can use the grating layer as the guiding layer as well.

First order coupling of the incident radiation to the grating produces radiation propagating predominantly in the guide layer perpendicular to both the incident beam and the grating bars. Bragg reflection by second order coupling to the grating produces oppositely directed waves which superimpose on the diffracted first order waves to produce a standing wave resonance in the guide layer. Since the resonance in the guide layer depends on the grating geometry and on the optical properties of the guide layer, but not on the lateral dimensions of the guide layer perpendicular to the incident beam, the resonance is distributed along the guide layer.

The guide layer has a higher index of refraction than the adjacent layers for the devices which are based on dielectric waveguides, or is a metal, with a complex index of refraction for surface plasmon based devices. Electromagnetic wave (guided mode) is guided by this part of the structure, but is not entirely contained within it. The wave especially must penetrate into the grating layer for the coupling processes to occur. Bragg diffraction is caused by the periodicity in the grating layer.

The radiation energy of the distributed resonance in the guide layer is diffracted back through the grating and forms a wave propagating in the opposite direction to the incident wave. At resonance, energy from the incident wave is first converted to distributed resonance energy, and second converted to energy in a wave propagating back along the path of the incident wave. Transmission of the radiation through the device is blocked by resonance reflection. As taught by this invention, the incident beam is modulated by controlling the conditions which establish or extinguish a distributed resonance in the guide layer.

Only a portion of the incident radiation is coupled into the cavity. The rest passes through. Because of the resonance effect this energy is stored and accumulates to a level that is larger than the incident beam power. The cavity then emits radiation in the same magnitude but, out of phase with the portion that passes through, which cancels it. The cavity also emits energy in the opposite direction which is the reflected wave.

The energy in the guide distributed resonance determines the reflectivity of the incident wave and thereby determines the transmission of the incident wave through the device. Modulation of the incident radiation beam is determined by the presence or absence of resonance in the distributed resonant cavity.

The requirement for first order coupling by the grating is that the grating period be typically less than a wavelength of the incident radiation. This prevents the loss of resonant energy to higher diffracted orders radiated by the grating.

Using a period less than the wavelength ensures that there are no propagating diffracted orders that radiate at angles other than the normal angle out of the structure. This prevents unwanted energy loss out of the system.

The thickness is dictated by the fact that only one guided mode should be supported by the structure. This prevents another route for unwanted energy loss by coupling to the additional modes (by defects for example). Typically, for one mode it works out that the thickness is on the order of a wave length. So that only one guided mode is supported and resonant energy is not lost by coupling to additional guide modes. Typically the thickness of the guiding film or layer for the distributed resonance is on the order of one wavelength.

By defining a Cartesian coordinate system x, y and z for reference, a beam polarized radiation with electric field along the y-axis (TE polarization) propagates along the x-axis and is normally incident on one side of a first transparent cover layer in the y-z plane. The second element is a grating disposed on the transparent support layer on the side opposite the incoming light, said grating having bars and grooves along the y-axis. The third element is a transparent guide layer imbedding the grating bars and in contact with the first layer in the grating grooves. First order diffracted radiation propagates primarily in the guide layer along the z-axis. Coupling of the normally incident radiation of wavelength .lambda. to the guided mode is characterized by propagation constant .beta.=mK, where m is equal to 1 for first order and K is equal to the 2.pi./.LAMBDA. where .LAMBDA. is the grating period. The radiation propagating with constant .beta. along the z-axis is Bragg reflected by the grating. The value for .beta. for which a standing wave or resonance exists is given by

For second order, m=2. The electromagnetic wave propagating along the z-axis is described by

Using this field representation together with a fourier series representation of the index of refraction in the grating region

yields an infinite set of coupled wave equations for the fundamental and each of the harmonics of the field

where

The quantities .beta., m and K have been previously defined. The electric field is E and n.sub.j is the index of the refraction of the grating layer. This representation of the field in terms in spatial harmonics is known as the Floquet theorem. See A. Yariv, M. Nakamura, IEEE J. Quantum Electronics, Vol. QE-13, No. 4, April 1977, p. 233. When the set of coupled differential equations is solved with the appropriate boundary conditions and, where the result is singular, the field can exist without input. Transverse resonance exists and energy is guided by the structure. See S. T. Peng, T. Tamir, H. L. Bertoni, IEEE Transactions of Microwave Theory and Techniques, Vol. MTT 23, No. 1, Jan. 1975, p. 123. Peng has described transverse resonance for the case of such periodic structures.

If the guided modes with light energy trapped by transverse resonance now interact with the grating, which satisfies the Bragg reflection condition for the mode, then the mode is reflected back on itself or put in different terms, the mode transfers its energy to its counter propagating pair. Since the mode has the same phase velocity forward and backward, a standing wave is created. The structure is now in a state of both lateral resonance along the y-axis of the guide as well as transverse resonance along the x-axis. The lateral resonance condition differs from the transverse resonance condition in that it is a distributed resonance. The standing wave pattern can exist over an area of infinite extent.

The device structure taught by this invention satisfies two condition. First, a portion of incident electromagnetic energy is coupled into the guided mode setting up transverse resonance in the structure. Second, Bragg reflection of the mode is established creating lateral resonance so as to form a resonant cavity distributed in space, a cavity which communicates with the outside world through the coupling properties of the periodic structure.

The distributed resonant cavity is a device that when not at a state of resonance is essentially transparent to the incident wave, since there is no coupling- When the cavity interacts resonantly with the incident wave, the energy confinement raises the reradiated power coupled out of the cavity to a level and with the appropriate phase so that it can interfere with and block the portion of the incident wave that does not couple to the cavity and would normally be transmitted.

A simple ray model can explain the phase shift between the transmitted portion of the incident wave and the reradiated wave at resonance. As shown in FIG. 11, the incident wave, 1, strikes a point on the grating and a portion of the beam is diffracted while the rest, 4, is transmitted normally. Following a ray 2 which is emitted from the grating at point 3 as part of the diffracted wave that couples to the guided mode and travels across the guide layer to surface 5 where it is totally internally reflected and then propagates back across the guide layer to the grating 6. At the grating a portion of the wave 8 is diffracted opposite to the incident beam and a portion 9 is diffracted in the direction of the normally transmitted beam. The phase shift of this diffracted beam 9 relative to the normally transmitted beam 4 is given by the path length traversed by the wave plus the phase shifts associated with the total internal reflection at 5 and the diffractions at 3 and 6. For a well guided mode this phase shift is in the vicinity of 180 degrees. The above is repeated as shown by the incident wave at point 10 and, when resonance is occurring, transmission is blocked, see the action at 9, 11 and 12 at the right side of FIG. 11.

A comparison will now be made of reflectance between transverse resonant case and distributed resonant cavity with both transverse resonance and lateral resonance due to Bragg reflection. This resonant effect is also exhibited by a structure where a guide mode is excited so that the structure is in a state of transverse resonance but where there is no Bragg reflection of the mode. As a result there is no lateral confinement and the mode freely propagates in the z direction. This structure also builds up energy at resonance and blocks transmission in the manner described above. See S. S. Wang, R. Magnusson, J. S. Bagby, M. G. Moharam, J. Opt. Soc. Am. a/Vol. 7, No. 8 Aug 1990, p1470.

The advantage of including lateral resonance by Bragg reflection is the improvement of the energy Confinement in the z direction. This gives the distributed resonant cavity superior bandwidth and angular beamwidth characteristics and allows the device to be fabricated over a smaller lateral area, which are critical considerations for practical applications. FIGS. 12a and 12b show the transmission characteristics of a gaussian beam as predicted by a plane wave analysis, where a coupled mode theory model is used to analyze the transmission of the plane wave components of the beam. See R. F. Kazarinov and C. H. Henry, IEEE Journal of Quantum Electronics, Vol. QE-21, p. 144, Feb. 1985. FIGS. 12a and 12b are a comparison of transmission through a resonant grating structure for a distributed resonant cavity versus transverse resonance only. In FIGS. 12a and 12b, the Gaussian beam waist size equals 1 mm and the coupling parameters are h.sub.2 =200 cm.sup.-1 and h.sub.1 =6 cm.sup.-1. Coupling of the incident wave to the guided wave and Bragg reflection coupling of the backward and forward propagating guided waves are included. Identical coupling parameters are used for both cases. In case 1 a distributed resonance cavity structure with a normally incident Gaussian beam is considered, FIG. 12a, i.e., normal incidence on distributed resonance cavity. This is compared with case 2, where the incident gaussian beam couples to the forward propagating guided wave at an off normal angle such that the structure is in a state of transverse resonance only and Bragg reflection is not considered, FIG. 12b, i.e., incidence at 45 degrees with transverse resonance only. The figures show the better mode energy confinement properties and improved blockage of transmission for the distributed resonant cavity which is in both transverse and lateral resonance versus the case of transverse resonance alone where the guide wave freely propagates and is clearly offset from the incident gaussian beam.

Other and further objects and advantages of the present invention will become apparent from the following detailed description of preferred embodiments when taken in conjunction with the appended drawings.

FIG. 1 is a section through a device according to the invention showing the transparent support layer, the grating and the transparent resonance guide layer.

FIGS. 2a and 2b are schematic illustrations of the coupling of normally incident radiation. FIG. 2a shows first order coupling to the guide layer. FIG. 2b shows second order Bragg reflection of radiation produced by first order coupling.

FIGS. 3a and 3b show prior art resonance cavity structures for comparison with the present invention which is schematically shown in FIG. 3c.

FIG. 4 shows a plot of the reflectance curve against wavelength.

FIG. 5 is another embodiment of the invention showing in schematic cross-section a depletion mode heterojunction diode modulator.

FIGS. 6a and 6b are another embodiment of the invention showing in schematic cross-section a bistable switch using nonlinear medium in a distributed resonant cavity in FIG. 6a. A plot of the reflectance as the function of incident intensity is shown in FIG. 6b.

FIG. 7 is another embodiment of the invention showing in schematic cross-section a dynamic grating switch.

FIG. 8 is another embodiment of the invention showing in schematic cross-section a MOSFET modulator with a surface plasmon distributed resonant cavity.

FIG. 9 is another embodiment of the invention showing in schematic cross-section a Schottky diode modulator with a surface plasmon distributed resonance cavity.

FIG. 10 illustrates the computed reflectance peak at resonance for one embodiment.

FIG. 11 shows schematically the origin of phase shift between incident beam and scattered light from mode.

FIGS. 12a and 12b shows graphically a comparison of transmission through resonant grating structure for distributed resonant cavity versus transverse resonance only. FIG. 12a is for the case of normal incidence on distributed resonant cavity and FIG. 12b is for the case of incidence at 45 degrees with transverse resonance only.

FIG. 13 shows a coherent optical correlator using incoherent to coherent image conversion with resonant reflectance SLM.

FIGS. 14a and 14b show a SLM implemented with an array of integrated resonant reflective modulators in silicon.

FIG. 15 shows a 2D optical neural network utilizing vector matrix multiplication with SLM.

FIGS. 16a and 16b show a silicon wafer stack implementation of 2D optical neural network utilizing the novel resonant reflective modulators.

FIGS. 17a, 17b, 17c and 17d show a silicon wafer and chip layout including communication modules integrated at chip input and output pins using resonant reflective modulators according to the invention.

FIG. 18a shows the lateral dimensions of the grating region versus incident beam.

FIG. 18b shows in cross section a grating termination with metal reflectors.

FIG. 18c shows in cross section a termination with first order Bragg reflectors.

FIGS. 19a and 19b are diagrams showing modified single grating layer structures.

FIG. 20 is a schematic showing a modified grating means consisting of two grating layers.

FIG. 21 is a schematic showing a novel termination device with 2 grating layers.

FIGS. 22a and 22b show graphs which portray characteristics of the termination device of FIG. 21.

Referring now to the drawings and in particular to FIG. 1, there is shown a distributed resonance cavity device 10 for modulation of light beam 11 normally incident on a first semiconductor transparent cover layer 12, implemented in the preferred embodiment in silicon. Silicon is the material of choice and is deemed the best mode for carrying out the invention. Devices made from silicon in accordance with the present invention are the most preferred embodiments. For reference, rectangular coordinate system 18 is provided, and the incident light beam 11 emanates from a laser source (not shown) and is coherent having a preselected wavelength and is propagating along the x-axis. The first transparent cover layer 12 has two surfaces, a first surface 22 receiving the incident light beam and transmitting it to the second surface 23 on which is disposed a grating 13 with grating bars 14 and grooves 15 extending side by side and parallel to the y-axis. The bars and grooves have different indices of refraction and can have different widths and geometries other than rectangular in cross section but the grating must be symmetrical with respect to the x-axis. Embedding the grating and in contact with the first transparent support layer 12 is transparent resonance guide layer 16 which is externally bounded by surface 24. The first layer 12 and the third guiding layer 16 have preselected indices of refraction, n and n.sub.3, respectively. A modulated beam exits from surface 24 as indicated by reference numeral 17. The layer 16 can be supported on a substrate (not shown) that is transparent to the modulated beam 17. The silicon device is provided with a control 19 which is coupled to surfaces 24 and 22 and functions to change the relative indices of refraction of the device to drive it between a resonant guided mode when transmission is blocked and a non-resonant mode when transmission is permitted. Control 19 may be as simple as a bistable voltage source, i.e., two voltage levels with switching between the two voltage levels.

The silicon device can be made by known techniques for developing the various layers on a silicon substrate and using known doping techniques for obtaining the different indices of refraction. Specific examples of the present invention, as will be described in detail hereinafter, will serve to illustrate applications of the invention and also specifics regarding materials and layer thicknesses and the manner by which variation or control of the indices of refraction is accomplished.

Now referring to FIG. 2a in conjunction with FIG. 1, there is shown the action in the distributed resonant cavity. Shown is coupling of normally incident radiation to the guided mode. As shown a normally incident radiation 11 of wavelength .lambda. is transmitted through first layer 12, is diffracted by the grating 13, and produces radiation 20 propagating along the z-axis with propagation constant .beta.=mK for m=1 for first order coupling. The separation of the grating bars is the grating period .LAMBDA., and the grating characteristic number K is 2.pi./.LAMBDA.. Now referring to FIG. 2b in conjunction with FIG. 2a and FIG. 1, the present invention teaches the Bragg reflection of the diffracted radiation 20 in the reverse direction 21 along the z-axis forming a standing wave in the guide layer 16, Bragg reflection of guided mode. The resonance radiation in the guide layer 16 feeds back through the grating in phase relation to the incident radiation such that the incident radiation is reflected when the resonance is excited. Transmission of the light through the device is precluded at resonance. Thus, .beta. .sub.forward -.beta..sub.reverse =2.beta.=mK for m =2, 2nd order. In accordance with the teachings of this invention, an output beam exists in the absence of a resonance and is extinguished when the resonance is excited in the guide. The distributed resonance wavelength of the device depends on the bar spacing .lambda. of the grating and on the optical properties of the layers and the layer thickness.

For a given embodiment of the invention, light of variable wavelength will be transmitted through this device except for values of .lambda. substantially equal to resonance value. For light of a fixed wavelength .lambda., modulation of transmission through the device is achieved by control of the distributed resonance wavelength of the device. In the preferred embodiment, the grating period is less than the wavelength of incident beam 11 and the thickness of layer 16 is less than a wavelength of incident beam 11. However, when using a semiconductor material, such as silicon or CdS, the grating period is preferably less than half the wavelength of the incident beam 11. When using a non-linear medium, such as glass, then the grating period is preferably less than the wavelength of beam 11 but more than half the wavelength of beam 11.

FIGS. 3a and 3b show two well know prior art resonance cavity structures. In the Fabry-Perot device 31 of FIG. 3, a small portion of the incident light beam 30 passes through half silvered mirror 32 and is reflected by half silvered mirror 33. Under resonance conditions, waves are reflected back and forth between the two mirrors with the proper phase relationship to set up a standing wave. The energy build up in the standing wave pattern is sufficient to transmit through mirror 33 a beam with the intensity of the incident beam. The Fabry-Perot device is a bandpass filter at resonance.

In the microwave cavity 35 of FIG. 3b, incident radiation 30 enters the cavity 35 through aperture 36. Away from resonance the impedance mismatch at the aperture 36 causes the beam to be reflected. At resonance conditions, standing waves 34 are produced such that energy builds up in the cavity and radiation can be coupled out through aperture 37. This device again serves as a bandpass filter.

In contradistinction to the prior art, the present invention in FIG. 3c shows the grating 13 and the resonance guide 16. There are no large impedance mismatches as in the Fabry-Perot and Microwave cavities due to the mirrors and apertures. Therefore, the incident beam is freely transmitted through the structure. At resonance, energy is coupled out of the standing wave pattern 20/21 with the proper phase to interfere and cancel the normally transmitted beam. The device of the instant invention may therefore be characterized as a bandstop filter.

Referring to FIG. 4, there is shown the general reflectance characteristics of devices embodying the present invention i.e. a distributed resonant cavity. As shown, a maximum 42 in the reflectance curve 41 is achieved at resonance 43 with wavelength .lambda..sub.1.

Referring to FIG. 5, another embodiment or specific application of the present invention is shown and is in the form of depletion mode heterojunction diode modulator 50. Modulator 50 consists of transparent conducting electrode 51 which overlies a grating 56. A heavily doped p type region 60 acts as a low index substrate. A p+ contact layer 61 is appropriately developed by known techniques on one face of substrate 60. To its opposite face is grown by known methods, a lightly doped n type region, layer 59, of high index alloy. An n+ contact layer or region is appropriately developed on the layer 59 and then etched by conventional means to develop bars 58 and grooves 57 of the grating 56. Electrode 51 fills in the grooves 57. A voltage source and signal source control 54 serving as the means for controlling modulator 50 is connected on its + side by lead 53 to a contact 52 on the surface of electrode 51 and on its - side by lead 55 to a contact 62 on the surface of p+ contact 61. The resonance condition is controlled by imposing a control signal via controller 54 (which includes switching means) to effect changing of the index of refraction in the guide layer 59 (external source 54 of electromotive force, EMF, connected to contact terminals 52 and 62). The electric field produced by the EMF, in particular through the lightly doped, high index semiconductor guide layer 59, within the depletion region of the reverse biased diode depletes the carrier concentration and raises the index of refraction. In other words, the depletion region of the reversed biased diode is shifted within the lightly doped high index alloy region 59. Changing the index of refraction changes the resonant wavelength. Rapid tuning of the device resonance wavelength provides prompt modulation of the incident beam 11 which is readily transmitted through the device for resonance wavelengths other than the wavelength of the incident radiation and emerges as modulated beam 63. By selection of only two control voltages, one producing a distributed resonance cavity in layer 59 to block incident beam 11 and one producing transmission and by providing rapid switching between the two voltages by any simple known switching or gating technique, the modulator 50 can be made responsive to a digital input signal to transduce same into a corresponding digitally modulated beam 63. Controller 54 preferably provides this function using a control signal having a DC component with an AC component superimposed thereon.

A specific example of the depletion mode heterojunction diode modulator 50 is shown in FIG. 5 and consists of the following specific materials. Layer 51, the transparent conducting electrode, can be fabricated from indium tin oxide, In.sub.2 O.sub.3 :Sn, the thickness depending on the desired resistance of the leads (in the range of 1 .mu.m.).

Layer 56, the grating, can be an n+ contact region fabricated from an alloy of composition 10% Germanium, 90% Silicon (Sl.sub.0.9 Ge.sub.0.1) doped with a donor material such as phosphorus to a level of 10.sup.19 per cm.sup.3. Thickness is on the order of 500-1000 Angstroms.

Layer 59, lightly doped n type region high index alloy, is fabricated from the same alloy as layer 56 (Sl.sub.0.9 Ge.sub.0.1) with a donor doping level of 10.sup.17 per cm.sup.3. Thickness is about 0.6 .mu.m. Both layers 56 and 59 are grown epitaxially on silicon substrate 60 by CVD (chemical vapor deposition).

Layer 60, heavily doped p type region, is a silicon wafer of crystallographic orientation (111) which is doped with an acceptor material such as boron with an acceptor doping level greater than 10.sup.18 per cm.sup.3. Thickness is that of a wafer, approx. 150-200 .mu.m.

Layer 61, p+ contact region, is the backside of silicon wafer 60 with an approx. 1 .mu.m thick region of acceptor doping level of 10.sup.19 per cm.sup.3.

Controller 54 operates with a control operating voltage such that the diode is reverse biased between 0-10 volts.

FIGS. 6a and 6b show another embodiment of the present invention in the form of a bistable switch 70. In FIG. 6a, bistable switch 70 consists of substrate 76 on which is developed a non-linear medium which serves as a guide region or layer 75. The surface of region 75 is etched by known means to develop bars 74 and wells or grooves 73 of a grating structure 72. Although not necessary, a transparent cover layer 71 may be developed by known means on grating 72. The guide layer 75 is a non-linear medium (sometimes referred to as chi three) disposed on substrate 76 such that the index of refraction and/or absorbance depends on light intensity. The resonance wavelength of the device is shifted optically by varying the intensity of the incident light beam 11 or using an ancillary beam for this function. The means for varying the intensity of incident light beam 11 are well known and are not shown for sake of simplicity. The resonance traps light in the distributed resonance cavity, builds up large intensities, and such intensities change the refractive index of the material and thus substantially shift the resonance wavelength frequency.

Alternatively, for a saturable absorber, the absorbance is reduced by increasing the intensity in the cavity. The reduction in absorbance allows the intensity in the cavity to build up to stronger levels at resonance. A positive feedback results as illustrated in FIG. 6b where an increase of the intensity builds up the fields in the structure which causes the material properties to change in the direction that brings the system into resonance. FIG. 6b shows graphically reflectance as a function of incident intensity for saturable absorbance. The positive feedback leads to a sharp transition from resonance to non-resonance states as a function of intensity. The resulting hystersis 78-79 creates a bistable element that can be used for optical memory and switching.

As a specific example, the layer 75 is fabricated from cadmium sulfide, CdS, and the substrate 76 is glass. The grating 72 is created by etching the CdS layer 75 to form grooves 73 which are filled only with air. No cover layer 71 is used.

Referring to FIG. 7, another embodiment and specific example of the present invention is shown as a dynamic grating switch 80 consisting of a substrate 82 upon which is developed by known means a non-linear medium 81 serving as the distributed resonant cavity or guide region and contains at its upper surface a grating. The transparent cover layer may be omitted. In this embodiment the grating is written in a thin film layer 85 by two interfering laser beams 83 and 84 which create the bars 86 and grooves 87. The film layer 85 is photorefractive so that it records the intensity pattern of the writing beams as an index gradient. Modulation is achieved by changing the angle of the writing beams 83 and 84 thereby changing the grating period. The advantage of this embodiment is the remote control by optical means while the disadvantage is the relatively slow speed of most photo-reactive materials and the need to change the angles of the writing beams.

In a specific example of the dynamic grating switch shown in FIG. 7, the substrate is a crystal plate of lithium niobate (LiNbO.sub.3) cut with the appropriate crystallographic orientation to support the desired type of guided wave (X or Y cut for TE modes and Z cut for TM modes). Layer 81 is a region at the surface of the crystal plate into which titanium has been diffused. Titanium indiffused lithium niobate is normally prepared by depositing a layer of titanium metal on the surface of the lithium niobate and then heating the sample in the appropriate gaseous environment so that the titanium diffuses into the crystal forming a layer rich in titanium. This region has a sufficient concentration of titanium and is thick enough to support a guided wave. As discussed in the literature impurities such as iron or copper may be desirable with the titanium to support the nonlinear photorefractive effect. The titanium indiffused guide layer supports a single guided mode and exhibits a photorefractive (nonlinear) effect. In addition, the strips 86 and 87 representing the regions of different index of refraction that are segments of the grating written in the photorefractive material extend all the way through the guide region 81 but have been shown with partial extension for sake of clarity.

FIG. 8 shows an embodiment of the invention based on surface plasmons as the guided wave. Surface plasmons are established in the distributed resonance cavity of a MOSFET modulator 90 which consists of a p type silicon substrate 98 having conventionally developed on one face a p+ contact layer 99. The other face of substrate 98 is developed by conventional techniques into a grating 13 with bars and grooves in conjunction with a SiO.sub.2 gate insulator 97. Transparent electrode 91 overlies layer 97. A thin inversion layer 98a exists at the boundary between the gate insulator 97 and silicon substrate 98, of the metal oxide semiconductor field effect transistor (MOSFET), and supports the surface plasmon as the guided mode or wave. Such surface plasmons in very thin layers are known to be of long range. Propagation of the plasmon depends upon the carrier concentration in the inversion layer 98a. The carrier concentration is controlled by an electric field provided by controller 94 which controls an EMF source which is connected via leads 93 and 95 to contact terminals 92 and 96 disposed on opposite external surfaces of the device. The incident input beam 11 (TM polarized) is transmitted as a modulated beam 17 when the device resonance wavelength is unequal to the wavelength of the beam and the input beam 11 is reflected when the device is tuned to the same wavelength as that of the input beam.

In a specific example, the MOSFET modulator 90 of FIG. 8 consists of layer 91, the transparent conducting electrode, fabricated from indium tin oxide (In.sub.2 O.sub.3 :Sn). The thickness depends on the desired resistance of the leads but is in the range of 1 .mu.m.

Layer 97, the silicon dioxide gate insulator, is on the order of 1000 Angstroms thick. The material is silicon dioxide, SiO.sub.2, grown by thermal oxidation of the silicon wafer substrate 98.

Layer 98, the p type silicon wafer substrate, is of crystallographic orientation (100) with an acceptor doping concentration of 10.sup.16 per cm.sup.3. Thickness is that of a typical wafer as described with reference to FIG. 5.

Layer 99, the p+ contact region, is on the backside of silicon wafer 98 with an approximate 1 .mu.m thick region of acceptor doping level of 10.sup.19 per cm.sup.3. The modulator 90 is controlled via controller 94 with an operating voltage so that positive voltage is applied to transparent electrode 91 between 0-5 volts.

The inversion layer supporting the surface plasmons is on the order of an Angstrom in thickness and is situated at the interface between the gate insulator and the silicon substrate as already noted.

FIG. 9 shows another embodiment using surface plasmons in a Schottky diode modulator 100 which consists of an n type silicon substrate 102 having an n+ contact 103 conventionally developed on one face and in conjunction with a semi-transparent thin metal layer 101, a grating 13 including bars and grooves 14-15 on its other face. The grooves are filled with substrate 102. In the Schottky diode modulator 100, the carrier concentration change in the depletion layer between the thin metal layer 101 and the semiconductor 102 is used to modulate the index of refraction. The metal layer 101 supports surface plasmon at the metal-semiconductor interface and provides the metal portion of the Schottky junction. The metal layer 101 must be kept extremely thin so that it does not absorb too much of the incident light. This device as described can function as a very robust tunable filter/modulator. Controller 106 provides the control signal via leads 105, 107 to contacts 104, 108, respectively.

A specific example of the Schottky diode modulator 100 shown in FIG. 9 is as follows. A cover layer of sputtered glass or semiconductor layer 97a covers the thin metal layer 101 to provide protection and modify the propagation characteristics of the plasmon. Layer 101, the semi-transparent thin metal layer, is composed of silver and is approximately 100-150 Angstroms thick. Layer 102, the n type silicon wafer substrate, has a donor doping concentration of 10.sup.17 per cm.sup.3. Thickness is at least several microns. Layer 103, the n+ contact region, is on the backside of silicon wafer 102 with an approx. 1 .mu.m thick region of donor doping level of 10.sup.19 per cm.sup.3. The controller 106 is operated to provide a control operating voltage so that the diode is reverse biased between 0 to -5 volts on silver layer 101.

As noted previously, the thicknesses of the layers of these various devices depends upon the wavelength of the light that is to be modulated. The thicknesses given in the specific examples are for use with radiation having a wavelength of about 1.3 .mu.m., near infrared radiation. The grating teeth height and duty cycle depend on the bandwidth and angular beamwidth characteristics of the desired modulator. Also, in silicon, donor generally refers to phosphorus and acceptor to boron, as these are the materials of choice.

Referring to FIG. 10, the reflectance peak at resonance is illustrated by a plot of computed reflectance values 111 for a specific resonant structure 110 detailed in the lower right corner. The maximum in the reflectance occurs at resonance for which the resonant wavelength is .lambda..

Preferred applications of the device embodiments are shown in FIGS. 13-17 and are shown for spatial light modulation (SLM). In a spatial light modulator the devices are arranged in a two dimensional array (rows and columns) as part of the fabrication process. Each of the devices is addressable optically or electronically (as known) to control the resonant reflection. The lateral resonance due to Bragg reflection is used to provide energy confinement, laterally minimizing the individual element size. Such a spatial light modulator (SLM) using this invention has numerous applications including, incoherent to coherent light conversion for optical correlators, optical computing and optical addressing and interconnects in general.

The main advantage of this invention over existing SLM technology is the narrow bandwidth of the resonant reflectance, which can be made to match the bandwidth of a single mode laser. A small change in optical properties within the modulator is thereby sufficient to shift the reflectance resonance relative to the laser frequency to achieve sizable modulation. As a result, material whose optical properties can be modulated marginally can now be utilized to fabricate modulators. The best example (and most lucrative) is silicon, as shown in the modulator embodiments of FIGS. 5, 8 and 9. These modulators can be integrated into 2 dimensional arrays readily, using technology from the integrated circuit industry. Together with the modulators, associated logic and drive circuits as well as photodetectors, which of themselves are known in the art, can be integrated or connected to give the SLM a broad range of operating characteristics.

The application of an SLM, based on this invention, with a collimated laser beam for incoherent to coherent light conversion is shown in FIG. 13. A laser source 202 via lens 202a and laser collimator 203 feeds coherent light to dielectric mirror 204 and then to a resonant spatial light modulator array 205 and optical correlator 207 wherefrom it is received by a CCD camera 208. The SLM 205 is addressed with an incoherent scene, 201. The incoherent light intensity controls the transmission of the laser light 206 through the elements of the array 205, creating a picture which is presented to the optical correlator 207 and CCD camera 208.

A SLM for this purpose, which is based on this invention is shown in FIG. 14. This SLM is implemented in silicon and takes advantage of integrated circuit fabrication technology to integrate a modulator 215, detector 213 and drive circuitry 214 for each element or pixel 212 in the array 205. The detector 213 can be a photodiode with a suitable amplifier serving as the drive circuit 214. The modulator 215 can be an embodiment such as shown in FIGS. 5, 8 or 9. This device receives an image of incoherent light (arrow L) in the visible, detects it by photodiodes 213, and converts it via amplifiers 214 and modulators 215 to a coherent image in the infrared where silicon is transparent. The transmission of a laser through the modulators 215 is controlled by amplifiers 214, producing a control signal at two stable voltages that allow transmission or produce reflectance in response to the detected outputs of photodiodes 213.

SLMs are widely used for optical computing. FIG. 15 illustrates the application of an SLM in optical computing as part of a 2D optical neural network. The SLM is used to perform vector matrix multiplication as part of an electronically programmable interconnect matrix. An SLM based on this invention, such as the one pictured in FIG. 14 could be used for this purpose, where the modulator is electronically controlled rather than controlled just by a photodiode. The network consists of a known linear LED array 301 whose output passes through cylindrical lens 302 and is received by SLM 303 whose output in turn is gathered by cylindrical lens 304 and passed to linear detector array 305. Block 306 is a conventional drive arranged for the SLM 303 according to the invention and block 307 is a known electronic threshold for driving array 301 from the output of 305.

This 2D optical neural network could potentially be realized as a compact set of stacked silicon wafers, where the cylindrical optics and LED array have been replaced by another SLM 303 and a 2D array of detectors 305 where the SLM presents a column in place of each LED element and the 2D detector array processes a row corresponding to each element in the linear detector array. This arrangement is shown in FIGS. 16a and 16d.

As shown in FIG. 16a and 16b a silicon wafer stack is comprised of SLM 403 (replacing the LED array of FIG. 15), SLM 404 and 2D (two dimensional) detector array 405 separated by spacers 406. Each wafer 403, 404 and 405 has the configuration as shown in FIG. 16b with each modulator or detector element on each wafer identified by reference letter e. Laser 401 via lens 401a and collimator 402 drives the wafer stack with block 407 providing the electronic drive for the SLM 404 and block 409 establishing appropriate thresholds for the SLM 403 from the output of 405.

Integration in silicon, of the resonant reflective modulator together with a detector and appropriate drive circuitry, as in FIG. 14, can also be utilized to place communication modules on chips and on wafers for implementing optical interconnects between such chips on wafers. An example of such an optical interconnect scheme is shown in FIG. 17. Given a suitable architecture, a large number of conventional processors with such communication modules could be interconnected for parallel processing applications. Today, this is a very active field in computing for which the resonant reflective modulator may provide a potential solution to the interconnect problem. As shown, a wafer 501 contains multiple chip areas 502 arranged in rows and columns. Each chip area 502 contains pad input 507 and pad output 506 with connections 514 to a chip 503 for electronic processing. Optical input module 508 is mounted on pad 507 and optical output module 511 is mounted on pad 506. Optical input module 508 contains a photodetector diode 509 which receives and detects light as indicated by arrow L. Diode 509 is driven by a known drive circuit or amplifier 510 whose output is fed to chip 503 for processing. The optical output module 511 contains a resonant reflectance modulator 513 as previously described, fed with a laser, not shown, and controlled by a known drive circuit or amplifier 512 receiving its instructions or control signals from chip 503. The control signals from amplifier 512 determine transmission or reflectance of the laser through modulator 513 thereby transducing the control signals to a digitally modulated output laser beam.

The device descriptions previously given show a grating region which extends laterally over the whole surface of the device. In order for a device to work optimally in accordance with the present invention and properly, the incident beam must be of a lateral dimension that is considerably smaller than the lateral extent of the grating region. More precisely, the distance from the edges of the incident beam to the edges of the grating region must be greater than the Bragg length of the grating as shown in FIG. 18a. The Bragg length of the grating is defined as the reciprocal of the coupling coefficient between the counterpropagating modes as calculated from coupled mode theory. See A. Yariv, IEEE Journal of Quantum Electronics, Vol. QE-9,919, 1973.

The termination devices shown in FIGS. 18b and 18c allow the grating to be terminated at any desired lateral dimension; the device so terminated will modulate completely the incident beam falling on the grating region between the terminations. The terminations are reflectors with appropriate phasematching lengths such that the guided wave, A, that is reflected at the reflector returns to the grating region with the same magnitude and phase as the counterpropagating guided wave B. When A=B at the ends of the grating, the grating will have the properties of the infinite grating even though it has been terminated with a finite size.

The reflector in FIG. 18b is a mirror made by etching through the guiding layer 600 and into the substrate 602, ion polishing the surface to insure smoothness and then evaporating a coating 604 of silver or gold as a metal reflector.

The pit or well 606 created by etching through layer 600 into layer 602 has a vertical or normal side 608 adjacent the grating 612 and an inclined side 610 remote from the grating 612. Although side 610 is shown as inclined, it is not necessarily so as this depends upon the etching technique used. The coating 604 overlaps the top surface 614 of layer 600 as indicated at 616 and 618. The result of the above is etched pits with polished walls. The terminations are spaced a phase matching length 620 on either side of grating 612.

The reflector in FIG. 18c is a Bragg reflector which is a grating 630 that reflects the guided mode in first order as opposed to the second order as is the case for the main grating region of the distributed resonant cavity. Bragg reflection in the first order ensures that there is no radiated wave from the structure and there is complete reflection. The grating period is smaller and the grating can be made deeper than the main grating region, where the grating depth must be controlled to meet bandwidth requirements for the device. The deep grating and smaller period of the reflector result in a smaller Bragg length. Since the reflector must also be longer than the Bragg length the fact that the Bragg length of the reflector can be minimized helps reduce the device size.

The semiconductor devices (MOSFET structures, diodes of FIGS. 5, 8, 9) depend on precise control of the index of refraction and preferably the device should be housed in an environment where the temperature is strictly controlled in order to obtain the best performance. However, changing the device structure to make it less temperature sensitive is also contemplated by the present invention. For example, the use of a cover layer whose temperature dependence is opposite to that of the semiconductor material of the guide layer and the substrate would make the propagation constant of the guided wave temperature insensitive. The temperature effect may also be compensated for electronically by adjusting the bias voltage on the device automatically responsive to a temperature sensor to compensate for temperature induced index changes. A system combining all of the above techniques would produce the optimum result.

Described in the foregoing is the basic concept of the distributed resonant cavity and its embodiment as a tunable filter, which can be used as a light modulator for a narrow spectral bandwidth laser light source. The key points covered are as follows:

1. A layer means which supports one resonant (guided) mode.

2. A grating means which a) establishes first order coupling of the guided mode to the normally incident beam and b) produces a standing wave in the layer means by Bragg reflection of the guided mode by second order coupling by the grating means.

3. The action of modulation is performed by changing the index of refraction of the layer means thereby shifting the resonant wavelength to be equal to the radiation wavelength.

Under these conditions the structure has standing waves both transverse to the layer means (due to the guided mode) and laterally because of the Bragg reflection and can be considered to be a distributed resonant cavity. Energy is built up in the cavity and then coupled out by the grating means in the first order so as to interfere with and block the portion of the incident beam that would otherwise be transmitted when in the non-resonant state as illustrated by FIG. 11.

There will now be described below changes to the grating means to optimize the angular beamwidth characteristics by controlling the grating's fourier components in the single layer grating, or by optimizing the coupling coefficients where the grating means consists of two independent grating layers. A termination device suitable for a grating means consisting of two independent grating layers is also described below.

As discussed above the second order coupling and resulting lateral standing wave can provide a larger angular bandwidth for a given spectral bandwidth and minimize the lateral area of the modulator. This is illustrated using a coupled mode model based on that of Kazarinov et al. in FIG. 12. Not specifically stated is the relationship between angular bandwidth and spectral beamwidth to the coupling constants, which is as follows for the case of strong Bragg reflection (large .sup.h2) ##EQU1## .DELTA..THETA. is the angular beamwidth and .DELTA.k is the spectral bandwidth in terms of the wavevector. All other terms are as defined by Kazarinov et al., and are as follows:

h.sub.1 is the coupling coefficient arising from two first-order diffractions;

h.sub.2 is the coupling coefficient arising from a second-order diffraction;

.xi. is the Fourier coefficient of the grating position x within the grating layer.

k and c are constants.

One can see that to maximize the angular beamwidth for a given bandwidth the coupling constant.sup.h2 must be maximized for a fixed .sup.h2. The coupling constants are directly related to the corresponding fourier components, .xi., of the grating region as given by Kazarinov and discussed above. The second order fourier component must therefor be maximized relative to the first order component while adjusting the grating region thickness to maintain .sup.h1.

For a grating comprised of rectangular bars and grooves as described in the embodiments above, this can be accomplished in one of two ways:

When there is one bar and groove per grating period, .LAMBDA., the relative width of the bar, w, to the period (duty cycle) can be adjusted to give the desired ratio of the fourier components using the following relationship: ##EQU2##

When there are two bars per period, symmetrically placed about the center of the period as shown in FIGS. 19a and 19b, the fourier components are given as follows: ##EQU3##

By the addition of the adjustable parameter, the bar spacing, 2a, the ratio of the fourier components can be further improved over the case with a single bar. Placing the bars so that they are centered a half period apart effectively cuts the grating period in half and removes the first order component. Working in the vicinity of this condition provides large ratios of the second to first order component.

As described above, the grating means consists of a single grating layer which accomplishes both functions of coupling to the layer means. The embodiments describe a way to optimize the grating layer to maximize the angular beamwidth for a desired spectral bandwidth by adjusting the fourier components of the grating to the best value. Another approach is to split the grating means into two grating layers on either side of the layer means as pictured in FIG. 20. Each grating layer of the grating means separately performs one of the functions originally performed by the single grating layer of the grating means. This approach is suitable only for embodiments where a relatively thick layer means is grown so that the perturbations from the bottom grating layer do not reach the top grating layer. It is therefor unsuitable for embodiments based on surface plasmons where the layer means consists of a very thin layer.

The bottom grating layer contains a grating of alternating bars and grooves, which performs only first order coupling of the incident radiation to the layer means. This is accomplished by setting the duty cycle of the grating to 0.5 thereby setting the second fourier component of this grating to zero. This grating layer thereby controls only the coupling constant .sup.h1.

The top grating layer of grating means couples the layer means causing only Bragg reflection of the guided mode, without having any effect on the coupling of the incident radiation to the layer means. This is accomplished by setting the grating period one half the value in the bottom grating layer. The grating performs Bragg reflection in the first order. The fourier component of this top grating layer for first order coupling to the radiation field is now zero. This grating layer thereby controls only the coupling constant .sub.h2.

The grating means still performs the two functions described previously, namely 1) coupling of the normally incident radiation field to the layer means (mode) and 2) coupling to the layer means to cause Bragg reflection of the mode A distributed resonant cavity is again formed, exhibiting the same resonant reflection properties, except that now the coupling constants can be separately adjusted by adjusting the thicknesses and duty cycles of the grating layers The coupling constants can thereby be adjusted separately to maintain a given spectral bandwidth, while maximizing the angular beamwidth.

The layer means must be sufficiently thick to preclude interaction between the two grating layers comprising the grating means. Specifically, the layer means thickness must be greater than the decay length of the evanescent fields of the higher order modes in each grating layer The decay length, .delta., is given by the following formula

n is the index of refraction of the layer means, m is the order of the higher order grating evanescent mode, .beta. is the guided mode propagation constant and K is the grating vector.

Termination devices are discussed above and are illustrated in FIGS. 18b and 18c. The termination devices on either end of the grating region maintain a lateral standing wave in the grating region with the forward and backward traveling mode in the layer means having equal amplitude and phase. This allows the lateral area of the device to be reduced. The termination devices consist of two components, a reflector and a corresponding phase matching length. The reflecting elements described were a mirror and a first order Bragg reflector. The period of the first order Bragg reflector must be set so that the wavelength of the incident radiation sits at the center of the forbidden zone, as discussed below. Fabrication of these termination devices is most easily performed when the termination device and the grating means are made with a single photolithographic step, where both devices are contained on the same mask, as is the case for the termination device utilizing the first order Bragg reflectors. This eliminates the need for submicron alignments with multimask photolithography. When a two grating layer grating means is used as described above the structure may be terminated as illustrated in FIG. 21.

The bottom grating layer is the region of smaller lateral extent, designed to perform first order coupling to the incident beams as described above. The top grating layer consists of three zones (1) a zone which performs first order Bragg reflection, situated above the bottom grating region (2) a second zone of phase matching length at both ends of the first zone and (3) another Bragg reflector zone.

The first zone may extend beyond the lower grating region at either end. In this region, denoted by (b), the upper grating has little effect since the radiation coupled by the bottom grating is at a wavelength which sits at the edge of the forbidden zone for the upper grating alone where there is negligible Bragg reflection when there is no first order coupling to the radiation field. This is shown in FIG. 22. Since there is no Bragg reflection in this region it is unimportant, but provides a buffer area, which insures that the grating region (1) completely overlaps the bottom grating region.

The design of the Bragg reflector zone (3) requires a period such that the grating functions as a first order Bragg reflector for the mode tat the wavelength of the incident radiation. Note that the period is different from that of region (1) so as to position the incident radiation wavelength at the center of the forbidden zone of this region as illustrated in FIG. 22.

Summarizing, the grating means is modified by 1) optimizing the fourier components in a single grating layer, or 2) optimizing two grating layers, where each layer determines one of the two coupling coefficients independently. This change to the grating means is essentially intended for embodiments with dielectric guide layer means as previously described. These modifications allow the angular beamwidth of the device to be optimized for a given desired spectral bandwidth, and as noted, can be substituted in any of the devices, apparatus or modulators previously described.

Although the invention has been shown and described with reference to preferred embodiments, nevertheless, changes and modifications will be obvious to those skilled in the art which do not depart from the spirit or teachings hereof. Such are deemed to fall within the scope and purview of the appended claims hereto.