Tunable Liquid-Crystal Etalon Filter
SPECIFICATION
Field of the Invention
The invention relates generally to optical filters; in particular, it relates to liquid-crystal optical filters.
Background Art
A need has arisen for low-cost, tunable optical filters. For example, one proposed architecture for the future telephone network based on optical fiber uses wavelength-division multiplexing (WDM) . In WDM , the data of different communication channels (e.g. , multiple voice channels, video channels, high-speed data channels) modulate optical carriers of different wavelength, and all the optical carriers are impressed upon a single optical fiber. The multiplexed optical signals are all distributed to many customer sites, each with its own receiver. Each receiver must be able to pick out one of the multiplexed signals. In a direct detection scheme, an optical filter passes only the selected optical carrier, and an optical detector detects the time-varying (data modulated) intensity of the filtered optical carrier. Preferably, this filter should be tunable so as to easily select different data channels. Channel spacings of as little as 1 nm are being proposed in the infra-red band of 1.3 to
1.5 μ .
Diffraction gratings provide the required resolution but are too expensive and fragile for customer-premise use. It is desired that the tunability be purely electrical and include no moving mechanical parts. Acousto-optic filters have been proposed. They offer superior resolution , tuning range, and ruggedness. However, their cost remain moderately high, and they require significant amounts of expensive RF electrical power.
A liquid-crystal light modulator has been disclosed by Saunders in U .S. Patent 4,779,959 and by Saunders et al. in "Novel optical cell design for liquid crystal devices providing sub-millisecond switching,"
Optical and Quantum Electronics, volume 18 , 1986, pages 426-430. A
modulator blocks or passes the input light. A filter performs a more complicated function by frequency selecting from a broad input spectrum , passing some components while blocking others. Saunders defines a
10 μm optical cavity between two partially reflecting metallic mirrors and fills the cavity with a nematic liquid crystal. The mirrors also act as electrodes for the standard liquid-crystal display configuration in which an applied bias rearranges the liquid-crystal orientation. However, in the etalon configuration of Saunders, the applied bias in changing the effective refractive index of the liquid crystal also changes the resonance condition for the cavity. Both references include a graph showing a bias- dependent optical filtering. Saunders relies upon this effect to intensity modulate a beam of well defined frequency between two intensity values. The liquid-crystal modulator of Saunders could be modified to be used as a tunable filter for a wide bandwidth signal. However, it would operate poorly. For a simple Fabry-Perot etalon, the transmission at a given wavelength λ for radiation incident to the normal of the surface is given by
T (λ) = T- — (1)
(l- p)2+ 4psin2δ
where δ = φ + k0dn. Here T and p are the transmittance and the reflectance and, φ is the phase shift experienced upon reflection, d is the thickness of the uniaxial material, n is the refractive index along the director axis, and I Q is the magnitude of the wave-vector outside the layer. Equation (1) shows that the width of the transmission peak Δλ depends essentially on the reflectivity of the surfaces, while the overall transmission is dictated by the absorption losses.
Saunders uses silver mirrors having reflectivities in the range of 85— 90% . His illustrated transmissions peak at about 50% , an acceptable value for some applications, but the peaks are relatively wide. The widths present little problem when the structure is used as a modulator for a well defined wavelength. However, Saunder's peaks are separated by only a few times the peak widths. Therefore, his structure would be
effective at filtering only a very few channels. For a practically useful device such as those useful in multichannel systems, the reflectivity of the mirrors must be kept above 95% . The peak width of Saunders could be reduced by increasing the thickness of the silver mirrors, thus increasing the mirror reflectivity to above 90% . However, the increased thickness would increase the mirror absorption losses to the point where the peak transmission is unacceptably lowered.
In liquid crystal optical filters as heretofore proposed, the input light directed at the filter must be linearly polarized in a particular direction relative to the orientation of the liquid crystal molecules. Otherwise, the filter cannot be electrically tuned. In some applications, the polarization requirement imposed on the light directed at the filter is easily met. In other applications, however, such as in an optical fiber communication system wherein propagating light is generally elliptically polarized, such a requirement constitutes an undesirable and disadvantageous limitation.
Many applications, particularly in a fiber-optic communication network, require that an optical filter be polarization-independent, that is, that the spectral characteristics of the filter be independent of the polarization of light being filtered.
Summary of the Invention A first aspect of the invention can be summarized as a liquid- crystal etalon optical filter in which the liquid crystal is sandwiched between two dielectric stack mirrors defining the optical cavity. Electrodes associated with each mirror apply a bias to the liquid crystal, thereby changing its dielectric characteristics and thus the resonance conditions of the optical cavity. Preferably, the electrodes are placed behind the mirrors, thus avoiding absorption loss. Variation of the bias applied to the electrodes causes the pass frequency of the filter to be electrically changed.
According to a second aspect of the present invention , the molecular orientation of a nematic liquid crystal material included in a Fabry-Perot etalon is twisted in a prescribed fashion to achieve an electrically tunable polarization-insensitive optical filter. In particular, the
71 17 twist imparted to the molecules of the liquid crystal is established at where n is a positive odd integer.
A third aspect of the invention can be summarized as a liquid- crystal etalon filter or modulator comprising a liquid-crystal filling a Fabry-Perot cavity defined between two end mirrors. One of the alignment layers at the ends of the cavity is a homogeneous aligning agent patterned into two areas aligning the liquid crystal in perpendicular directions within the plane of the alignment layer. The other alignment layer may be similarly patterned and homogeneous layer or preferably may be a uniform and homeotropic alignment layer aligning the liquid crystal perpendicularly to the alignment layer. An input light beam is divided between two areas of the liquid crystal so that both its polarization states are equivalently filtered. Preferably, a birefringent crystal, such as calcite, is affixed to the input side of the filter so as to polarization divide the input beam to the lateral side of the filter with the corresponding polarization.
Brief Descrip tion of the Drawings A complete understanding of the present invention and of the above and other features and advantages thereof will be apparent from a consideration of the detailed description set forth below taken in conjunction with the accompanying drawing, not drawn to scale, in which: FIG . 1 is a cross-section of a first embodiment of the liquid- crystal etalon filter of the invention.
FIG . 2 is a graph illustrating the spectral dependence of transmission of the filter of FIG. 1 at two bias voltages.
FIG . 3 is a graph illustrating the bias dependence of the transmission peaks of the filter of FIG . 1.
FIG . 4 is a cross-section of a second embodiment. FIG . 5 is a plan view of the alignment layer of a third embodiment.
FIG. 6 is a simplified diagrammatic side-view representation of an optical filter made in accordance with the fourth embodiment of the present invention .
FIG . 7 schematically depicts the contours of the longitudinal axes of liquid crystal molecules included in the FIG . 6 arrangement.
FIG . 8 is a cross-sectional view of a fifth embodiment of a liquid-crystal etalon filter of the present invention . FIG . 9 is a plan view of a patterned alignment layer taken along sectional lines 9 — 9 of FIGS. 8 and 11.
FIG . 10 is a illustration of the transmission wavelength as a function of voltage for an example of the invention.
FIG . 11 is a cross-sectional of a sixth embodiment of the invention.
FIG . 12 is a schematic illustration of an optical drop circuit using the polarization diversity filter of the invention .
FIG . 13 is a plan view of an arrayed alignment layer taken along sectional lines 9 — 9 of FIGS. 8 and 11. Detailed Description of the Preferred Embodiments
The optical filter of a first embodiment of the invention is a liquid-crystal etalon filter, in which the end mirrors of the optical cavity are highly reflective, preferably being dielectric interference mirrors.
Example 1 A first example of the first embodiment of the invention is illustrated in cross-section in FIG . 1. Two glass plates 10 and 12 of 1.62 mm thick soda-lime glass are used as substrates. Transparent electrodes 14 and 16 of indium-tin-oxide are deposited on the substrates 10 and 12. Dielectric stack mirrors 18 and 20 are formed on the electrodes 14 and 16. The mirrors 18 and 20 may be either commercially procured from Virgo Optics, Inc. of Port Richey, Florida, in which case they are juxtaposed to the existing structure. Alternatively, they may be deposited on the electrodes 14 and 16 by sputtering. The deposited mirrors 18 and 20 are designed for an infrared filter around 1.5 μm although we have fabricated others that have been centered around 0.3 μm and 1.9 μm. The mirrors we fabricated consisted of four pairs of quarter-wavelength thick layers of different refractive index, specifically Al203 (an insulator) and Si. For the 1.5 μm mirrors, the Al20~ layers were ~ 240 nm thick and the Si layers were — 120 nm thick. The transmission curves of FIG . 2, to
be described later, were fit to theoretical expressions including the reflectance of the mirrors. The so calculated mirror reflectance was ~ 98% , compared to the maximum reflectance of 90% for the modulator of Saunders. Thus, the system requirement of 95% is easily satisfied. Dielectric mirrors of themselves are well known . For example,
Yoo et al. disclose two interference mirrors defining the ends of an optical cavity for a surface-emitting semiconducting laser in U .S. Patent Application, Serial No . 07/510,960, filed April 19, 1990. One of the mirrors consisted of alternating layers of Si and AI2 - , and the other consisted of alternating layers of semiconducting AlAs and GaAs.
Alignment layers 22 and 24 are formed on the dielectric stacks 18 and 20 by the procedure described by Patel et al. in "A reliable method of alignment for smectic liquid crystals", F erroelectrics , volume 59, 1984, pages 137-144. The two assembled structures are then assembled with a precise gap between them according to the following method. Four UV curable epoxy dots were placed over the alignment layer 24 at the corners of one of the structures. The epoxy is previously mixed with 10 μm rod spacers available from EM Chemicals of Hawthorne, New York. The second structure is then placed on the first structure having the epoxy with the alignment directions of the two alignment layers 22 and 24 being parallel. Manual pressure is gently applied to the structures while observing optical interference patterns under monochromatic light. The interference fringes are minimized. This structure is captured by hardening the UV curable epoxy by exposing the structure to UV radiation. The assembled structure is heated to about 100°C and a liquid crystal material 26 is flowed into the gap by capillary action. A nematic liquid crystal, E7, available from EM Chemicals, is used in its isotropic state. The gap is estimated to produce a Fabry-Perot cavity length of about 11 μm between the dielectric stacks 18 and 20. The alignment layers 22 and 24 cause liquid-crystal molecules 28 in the liquid- crystal material 26 to orient with their long axes parallel to the reflectors 18 and 20 and also one set of their short axes to orient parallel to the reflectors 18 and 20 but perpendicular to the long axes. These orientations apply only for no applied bias. In this embodiment, there is
no significant twist of the liquid-crystal molecules from one reflector to the other.
Electrical leads are connected between the electrodes and a voltage generator 29. For our tests, the generator 29 was a computer- controlled programmable voltage source, such as, Wavetek Model 75, which produced a square wave at 1 kHz.
A sheet polarizer 30 can be formed on the outside of either glass substrate 10 or 12 with its polarization direction aligned with the long axes of the molecules 28. However, the filter used in our tests used either unpolarized light or external means for controlling the polarization. Furthermore, the filter can be designed to operate in a band of radiation that avoids the need of a polarizer.
A light-emitting diode producing light at 1.5 μm was used as a light source 32 to test the filter of the embodiment of FIG . 1. In an initial test, no polarizer was used and the transmission spectrum was measured for applied AC bias of 0 V and 4 V, as illustrated in the graphs of FIG . 2. The units of transmission are arbitrary and the baselines have been suppressed.
It is seen that some peaks shift with bias while others remain fairly stationary. By separate experiments, it has been demonstrated that the stationary peaks correspond to light polarized parallel to the short axes of the liquid-crystal molecules 28 and the tunable peaks correspond to light polarized parallel to the long axes. As a result, the filter can also be thought of as a narrow-band polarizer. Because of the high reflectivity of the dielectric stacks 18 and
20, the width Δλ of the transmission peaks is relatively small, -~ 1 — 2 nm, measured as a full width at half maximum . However, the free spectral range (FSR) , which is the wavelength separation between successive transmission peaks, is relatively large, -~- 75 nm. It is determined by the optical thickness n -d of the etalon and by the spectral regions in which it is being operated. This range depends on the order in which the interferometer is being operated, and it is given by
where m is the order. The refractive index of liquid crystals is typically in the range of 1.5 to 1.7. Thus, 1.5 μm light will pass through the etalon having the 11 μm gap when the etalon is used in the 22-nd order. In this wavelength region, the free spectral range for a 11 μm etalon would be about 75 nm. The choice of using an 11 μm thickness was simply due to convenience of fabrication of the actual device. This thickness should ideally be chosen such that the wavelength range of interest is the same as the free spectral range. The tuning range is estimated to be order of 200 nm which corresponds to a change of about 0.2 in the refractive index. The maximum change in the index is equal to the birefringence of the uniaxial liquid crystal.
The tunability of the filter of FIG . 1 is demonstrated by the bias dependence of the peaks' wavelengths, as illustrated in the graph of FIG. 3. The dashed lines represent the transmission of light having a polarization along the short axes of the liquid-crystal molecules 28. There is virtually no tunability of these peaks, and thus this polarization cannot be tunably filtered. The solid lines represent the transmission of light having a polarization along the long axes of the liquid-crystal molecules 28. There is a threshold below which no wavelength shift is observed. This plateau is due to the Freedricks effect and has been observed by Saunders et al. as well. It exists when the directors at both surfaces are parallel to the surfaces. It is possible to eliminate this threshold and control its characteristics by changing the surface tilt of the molecules. For example, no threshold would be exhibited in a hybrid aligned sample in which the liquid-crystal molecules at one of the surfaces lie parallel to the surface and those at the other lie perpendicular. Perpendicular orientation can be achieved by the use of a homeotropic alignment agent, such as, octadecyltriethoxysilane. For such a structure in the low voltage regime, the index will change almost linearly with the applied field.
The tunable operation of the filter can be understood from FIG . 3. If the bias voltage is varied between approximately 1.6 V and 2.6 V, a single transmission peak varies between about 1.528 μm and
1.472 μm without interference from any other peaks, a tunability of 56 nm. To obtain tunability over the range of 1.592 μm to 1.532 μm would require removing the peak at 1.537 μm with the polarizer 40. These tuning ranges can be shifted and widened with optimization of the design.
A calculated transmission spectrum has been fit to the transmission data of FIG . 2. For the zero electric field, the ordinary refractive index is n0 = 1.5 and the extraordinary refractive index is ne = 1.7. On the other hand, for the 4 V applied across 11.32 μm, n0 = 1.5 and nf(E) = 1.536, where nf is the effective refractive index. The power required to operate the filter has been estimated to be in the microwatt range. The switching speed is of the order of milliseconds.
A filter of the present invention has been used in the demonstration of the polarization scrambler disclosed by Maeda et al. in "New polarization-insensitive direction [sic] scheme based on fibre polarisation scrambling," Electronics Letters, volume 27, pp. 10-12, 1991.
The liquid-crystal etalon filter of FIG . 1 is not laterally patterned, but the invention is not so limited. Illustrated in FIG . 4 is a second embodiment of a liquid-crystal etalon 1-dimensional or 2-dimensional filter array. It differs from the filter of FIG . 1 in that at least one of the electrodes is patterned into pixels 40 which extend across the substrate 12. The pixels 40 can be individually contacted at the side of the filter structure. If the associated dielectric stack 20 is deposited on the electrode pixels 40, a planarizing layer 42 must first be deposited to insure the optical flatness of that mirror stack 20.
In a third embodiment, at least one of the alignment layers 22 and 24 is patterned. The aligning procedure, described by Patel et al. , involves depositing a nylon aligning material on the substrate and then rubbing the aligning material in the direction in which the liquid-crystal molecules are to be oriented. Illustrated in FIG . 5 is a plan view of an alignment layer 44 having first portions 46 rubbed in a first direction and second portions 48 rubbed in a different direction , preferably orthogonal
to the first direction. The differential rubbing is easily accomplished for the aligning materials, such as nylon and 1,4 polybutyleneterephathalate. The entire alignment layer 44 is rubbed in a given direction, say the first direction , so that the first portions 46 are given the correct alignment. Then, the entire structure is covered with photoresist and processed using standard lithographic techniques so as to keep the first portions 46 coated with a photoresist material. The exposed second portions 48 are then rubbed in the second direction with the photoresist protecting the already rubbed first portions 46. Removal of the photoresist does not affect the alignment of rubbed first and second portions 46 and 48.
The second alignment layer 22 or 24 can be made to have the same patterned alignments as the patterned alignment layer 44. When the two structures are then assembled, the alignment directions of the opposed portions are made parallel. This requires precise physical alignment of the two alignment layers 22 and 24. Alternatively, the second alignment layer can be made to uniformly induce a perpendicular orientation of the liquid crystals, that is, along the normal of the alignment layer. A homeotropic aligning agent, described previously, will provide this effect. No patterning of the homeotropic aligning agent is required.
The third embodiment of FIG . 5 advantageously provides a polarization independent tunable filter for a single-mode optical fiber. The filter is assumed to be designed to place the untunable passbands (dashed lines of FIG . 3) outside the desired wavelength tuning range. A graded-index lens on the output of the fiber is aligned to the boundary between a paired first and second portions 46 and 48 so that each portion 46 and 48 receives half the light intensity. One portion 46 or 48 blocks a first light polarization but tunably passes a second polarization. Likewise, the other portion 48 or 46 blocks the second light polarization but tunably passes the first polarization. A single direct photo-detector detects the intensity of light passed by both portions. Alternatively, the two frequency-filtered polarizations can be recombined by a graded-index lens onto a second single-mode optical fiber. The intensity passed through both portions 46 and 48 is independent of the polarization of the light. A
3dB loss is incurred by this polarization diversity technique.
The filter of the invention can be used as a short optical pulse generator. A narrow-band light source 32, such as a laser, irradiates the tunable liquid-crystal etalon filter with monochromatic CW light of wavelength λ . The filter of, for example, FIG . 1 is designed to have at least a limited tuning range extending from one side of λ to the other. The bandwidth of the light source 32 should preferably be less than the width of the transmission peak of the filter. To generate the light pulse, a step function DC bias is applied by the voltage generator 29 to the electrodes so as to cover the limited tuning range about λ . Only in the short but finite time required for the finite-width transmission peak to transit λ will the filter transmit the narrow-band light.
In a fourth embodiment of the invention , a liquid-crystal filter can be made insensitive to polarization by imparting a 90° twist (or odd multiples thereof) to the liquid crystal between the two alignment layers.
A specific illustrative optical filter 60 made in accordance with the principles of the fourth embodiment of the present invention is represented in FIG . 6. Optical signals from a source 62 are directed at the left-hand side of the filter 60, as indicated by dash-line arrow 64. By way of example, the source 62 comprises a standard light-emitting diode and an associated optical fiber for applying signals to the filter 60.
The source 62 of FIG . 1 simultaneously supplies multiple input wavelengths in, for example, the range 1.4-to-1.6 μm. Only a selected one of these wavelengths is passed by the filter 60 and delivered in the direction of arrow 66 to a utilization device 68. The device 68 includes, for example, an optical fiber of the type included in a conventional wavelength-division-multiplexed (WDM) communication system .
In accordance with the invention, a signal source 70 shown in FIG . 6 is utilized to apply electrical control signals to the filter 60. In the absence of an applied control signal, the filter 60 will pass to the device 68 an optical signal of a particular wavelength, as determined by the geometry and properties of the constituent parts of the filter 60. Above a specified operating threshold value, control voltages applied to the filter 60 are effective to change its electro-optic properties such that
respectively associated wavelengths supplied by the source 62 are passed by the filter 60 to the device 68. In that way, an electrically tunable optical filter is realized.
The filter 60 of the fourth embodiment shown in FIG . 6 comprises on its left-hand side a glass plate 72 through which input optical signals are transmitted and on its right-hand side another glass plate 74 through which output optical signals are transmitted. Disposed on the inner or facing surfaces of the plates 72 and 74 are optically transparent electrodes 76 and 78, respectively, which comprise, for example, standard layers of indium tin oxide. As indicated in FIG . 6, electrical leads 80 and 81 respectively connect the layers 76 and 78 to the control signal source 70.
The optical filter 60 of FIG . 6 further includes mirrors 82 and
83. Illustratively, the mirrors each include multiple layers of dielectric material, which is a conventional known design. By way of example, each of the mirrors 82 and 83 is designed to have a reflectivity of about 94 to 99.99 percent for the range of wavelengths to be transmitted by the filter 60.
The filter 60 of FIG . 6, including the spaced-apart mirrors 82 and 83, comprises in effect a Fabry-Perot etalon. As is well known, the geometry of such a device can be designed to be resonant at a particular wavelength (and multiples thereof) . In that way, the device can be utilized to provide an output only at specified frequencies.
In accordance with the present invention, the filter of FIG . 6 is electrically tunable by changing the electro-optic properties of a layer 84 of a liquid-crystal material that is contained between the mirrors 82 and
84. Moreover, due to the imposition initially of a particular orientation on the molecules of the liquid-crystal material (described later below) , the operation of the filter 60 over its tuning range is independent of the polarization of input optical signals. In other words, the wavelength selected by the control source 70 for transmission by the filter 60 will be delivered to the utilization device 68 with the same output intensity regardless of the polarization of the input optical signals.
The layer 84 of liquid-crystal material represented in FIG . 6 is, of course, retained in the indicated space by conventional spacer and sealing members (not shown) . Illustratively, the layer 84 comprises a standard nematic liquid-crystal material having elongated rod-like molecules characterized by positive dielectric anisotropy. In one particular illustrative example, the thickness of the layer 84 in the indicated Z direction is only about 10 μm.
Interposed between the liquid crystal layer 80 of FIG . 6 and the mirrors 82 and 83 are so-called alignment layers. A variety of alignment materials suitable for use with liquid crystals are well known in the art. In particular, the mirror 82 includes on its right-hand face a layer 86 of a conventional alignment material, whereas the mirror 83 includes on its left-hand face a layer 88 of a conventional alignment material.
Each of the alignment layers 86 and 88 is effective to impose a particular orientation on molecules in adjacent portions of the liquid- crystal layer 84. Illustratively, each of the alignment layers 86 and 88 is initially rubbed in a particular direction to impose a corresponding orientation on adjacent liquid-crystal molecules. Such rubbing of alignment layers to control the molecular orientation of liquid-crystal materials is well known in the art.
In accordance with the present invention, the alignment layers 86 and 88 of FIG . 6 are designed to impose quiescently a particular twisted structure on the liquid-crystal molecules included in the layer 84. The nature of this twist is schematically depicted in FIG . 7 which shows some of the rod-like liquid-crystal molecules in the layer 84 disposed between the alignment layers 86 and 88.
Illustratively, the liquid-crystal molecules in the layer 84 of
IT
FIG . 7 are oriented to have a 90° or — twist between the alignment
2 layers 86 and 88. In this twisted orientation , molecules at the surface of the left-hand layer 86 are established to have their longitudinal axes parallel to the indicated X axis, whereas molecules at the surface of the right-hand layer 88 are established to have their longitudinal axes parallel to the indicated Y axis. As shown , the longitudinal axes of the molecules
at these two surfaces are displaced 90° with respect to each other. Between these surfaces, the longitudinal axes of the liquid-crystal molecules gradually change in the Z direction from parallelism with the X axis to parallelism with the Y axis. In accordance with the principles of the present invention, the amount of twist imparted to the molecules of a liquid-crystal material n IT included in an optical filter is defined by the term . For the particular illustrative case represented in FIG . 7, n equals 1. More generally, n can be any positive odd integer. Thus, for example, for n= 3, the longitudinal axes of the liquid-crystal molecules would undergo a rotation of 270° in the liquid-crystal layer 84 of FIG . 2 in the direction of the Z axis between the surfaces of the alignment layers 86 and 88.
In the absence of an electrical control voltage applied to the electrodes 76 and 78 of the optical filter 60 shown in FIGS. 6 and 7, both constituent orthogonal polarizations of a particular wavelength (and multiples thereof) will not be resonantly supported by the Fabry-Perot etalon, as determined, for example, by mirror spacing and the electro- optic properties of the liquid-crystal layer 84. Only those particular wavelengths which are resonant will appear at the output of the filter 60. For each polarization condition, however, a different wavelength will be resonantly selected.
Assume now that an electrical control voltage is applied to the depicted filter to establish a Z-direction electric field. Until the magnitude of the electric field in the liquid-crystal layer 84 reaches the well-known Freedricks threshold, the molecular orientation represented in FIG . 7 remains substantially unchanged. Above that threshold (for example, above about 2.0 V for a 10-μ,m-thick liquid-crystal layer 84) , the longitudinal axis of molecules in the central portion of the layer 84 begin to align parallel to the Z axis. Molecules at the surfaces of the alignment layers 86 and 88, however, remain unaffected by the applied field because of strong anchoring forces at these surfaces. The thickness of the unaffected or substantially unaffected surface regions in the liquid-crystal layer 84 is a function of the magnitude of the applied electric field. Thus,
in response to the application thereto of an applied control voltage above the Freedricks threshold, the liquid-crystal layer 84 includes two field- dependent variable-thickness birefringent regions at and near the respective surfaces of the alignment layers 86 and 88. Significantly, the principal optic axis in one such region is disposed at 90° with respect to the principal optic axis in the other region . Accordingly, each of the constituent orthogonal polarization states of an input optical signal will be affected by the same amount irrespective of the input polarization state.
At an operating field strength above that of the Freedricks threshold (for example, above about 2.0 V RMS at IkHz for E7 nematic liquid-crystal material from EM Chemicals, Hawthorne, N .Y. , for a 10- μ/n-thick liquid-crystal layer 84) , the thickness and electro-optic characteristics of the two aforespecified birefringent regions become substantially the same and then track each other as the applied control or tuning voltage is increased further. As a result, for a range of voltages above an operating or high-field value, the herein-described filter structure imparts the same phase change to each of the constituent orthogonal polarizations of an input optical signal. Thus, the wavelength passed by the structure is determined by a particular value of the applied voltage and is independent of the state of input polarization. In one of the birefringent regions, one constituent input polarization disposed approximately parallel to the principal axes of liquid-crystal molecules in the one region is affected while it is unaffected in the other birefringent region. The other or orthogonal polarization component of the input optical signal is substantially unaffected in the first region while it is affected in the other birefringent region. A tunable polarization- insensitive optical filter is thereby realized. In such a filter, the wavelength selected by and corresponding to a particular applied voltage will be transmitted at the same intensity level regardless of its polarization condition.
A specific illustrative embodiment of the present invention includes a 10-μτn-thick twisted nematic liquid-crystal layer 84 of the type represented in FIG . 7. The index of refraction of such a layer can be varied between approximately 1.5 and 1.7 by applying thereto high-field
operating voltages in the range of 2 to 10 V. (For such an embodiment, the Freedricks threshold voltage is about 2 V ) . For each of selected input wavelengths, there is a corresponding operating control voltage that when applied to the filter will permit a particular input wavelength to be transmitted therethrough. Changing the control voltage by a specified amount will cause another input wavelength to be selected for transmission. In the particular specified example, wavelengths spaced apart by 2 nm in a tuning range of 15 nm are respectively selected by changing the control voltage in steps of about 0.5 V. Wavelengths passed by such a filter are characterized by a spectral passband width of about 1 nm.
An optical filter of the type specified above can be switched relatively rapidly from a condition in which it passes one wavelength to a condition in which it passes another wavelength. The switching speed is dictated primarily by how fast the liquid-crystal molecules in the layer 84 can be re-oriented. In turn, this depends on a variety of factors such as the dielectric anisotropy of the liquid-crystal material, the value of the control voltage and the thickness of the liquid-crystal layer 84. Illustratively, switching speeds in the order of milliseconds are feasible in practice. Moreover, the power required to switch such a compact microminiature filter is typically less than a microwatt.
Finally, it is to be understood that the above-described arrangements are only illustrative of the principles of the fourth embodiment of the present invention. In accordance with these principles, numerous modifications and alternatives may be devised by those skilled in the art without departing from the spirit and scope of the invention. For example, in one modification a polarization -in sensitive filter is achieved by combining two layers of liquid-crystal material that are physically separated from each other. In this modification, the two layers are contained within the optical cavity formed by the aforedescribed mirrors and electrode structure and are retained in place by a sandwich structure that includes three spaced-apart glass plates. The optic axes of the two layers are arranged respectively orthogonally to each other so that light of any polarization experiences the same phase shift.
In this modification, it is important that the thicknesses of the two liquid- crystal layers be the same to insure polarization-insensitive operation .
Additionally, it is apparent that the devices described herein can be used as polarization-insensitive spatial light modulators. In operation, two voltages are applied to such a modulator. One voltage is chosen to allow a selected wavelength to propagate through the device. The other operating voltage is designed to cause the device to block the selected wavelength from passing therethrough.
One of the variants of the previously described third embodiment involved perpendicular and parallel alignments of the liquid crystal at the opposed alignment layers so as to produce a polarization- independent tunable liquid-crystal filter. Patel et al. provide a similar disclosure in "Electrically tunable optical filter for infrared wavelength using liquid crystals in a Fabry-Perot etalon," Applied Physics Letters, volume 57, 1990, pp . 1718-1720. This concept is further expanded in a fifth embodiment of a polarization independent tunable liquid-crystal filter 110, as illustrated in cross-section in FIG . 8. The filter 110 is fabricated on two glass substrates 112 and 114, onto which are deposited transparent indium-tin-oxide electrodes 116 and 118. Dielectric stack mirrors 120 and 122 are formed on electrodes 116 and 118, and each consists of multiple pairs of quarter-wavelength thick layers of differing refractive indices to thereby act as interference mirrors for the wavelength of interest.
A homogeneous alignment layer 124 and a homeotropic alignment layer 126 are deposited and buffed on top of the respective stack mirrors 120 and 122. The alignment layers 124 and 126 cause liquid-crystal molecules disposed adjacent to the respective alignment layer to be aligned in a particular direction. In particular, nematic liquid crystals are characterized by orientational order along the average direction of the long axes of the liquid-crystal molecules, called the director n. The alignment layers 124 and 126 establish the director n at the interface with the liquid crystal. The director n then varies smoothly in the liquid crystal between the two alignment layers 124 and 126. If an electrical field is applied across the liquid crystal, the director n becomes
increasingly aligned with the electrical field in the gap between the alignment layers 124 and 126.
According to the fifth embodiment of the invention, the homogeneous alignment layer 124, as illustrated in plan view in FIG . 9, is divided into two portions 130 and 132 divided by an interface 134. Both portions 130 and 132 are formed of a homogeneous aligning agent, e.g. , a nylon or polyester such as 1,4 polybutyleneterephathalate, that causes the director n at that point, and therefore the adjacent liquid-crystal molecules, to be aligned parallel to the surface of the alignment layer 24. However, the two portions 130 and 132 are buffed in perpendicular directions so that one homogeneous portion 130 aligns the liquid-crystal molecules perpendicular to the interface 134 while the other homogeneous portion 132 aligns them parallel to the interface 134. Details of the differential rubbing procedure are discussed in the parent application. On the other hand, the other alignment layer 126 is formed of a homeotropic aligning agent, e.g. , octadecyltriethoxysilane, that causes the director n and the adjacent liquid-crystal molecules to be aligned perpendicularly to the surface of the alignment layer 126.
Following the formation of the critical alignment layers 124 and 126, the two substrates 112 and 114 are formed into an assembly having a gap between the two alignment layers of about 10 μm and a nematic liquid-crystal 136 is filled into the gap, all according to the procedure detailed in the parent application.
When a voltage generator 138 is connected between the electrodes 116 and 118, the applied electric field will cause a change in the effective refractive index in the liquid crystal 136 for light polarized in the direction of the director n, thus affecting the effective optical length of the Fabry-Perot cavity formed between the two mirrors 120 and 122. The direction of the director n transverse to the optical propagation is determined by the alignment direction of the homogeneous portion 130 or 132. The electric field, on the other hand, has no effect on the refractive index for light polarized in the direction perpendicular to the director direction. Thus, the change in effective refractive index depends strongly upon the polarization of the light. Because of the differential buffing
directions in the alignment layer 124, light passing through the portion 130 will be affected only for its electrical polarization component that is perpendicular to the interface 134 while that passing through the portion 132 will be affected only for its electrical polarization component that is parallel to the interface 134. Because of the symmetry, both of the polarization components will be equally affected by the applied voltage as they pass through the respective portions 130 and 132 and the corresponding portions of the liquid crystal 136.
The tunable liquid-crystal etalon filter 110 can be used by optically aligning an input optical fiber 140 with the interface 134 of the homogeneous alignment layer 124. A rod graded-index lens 142, represented functionally in FIG . 8, disperses the beam so that equal amounts of optical energy fall upon the two portions 130 and 132. On the output side, a corresponding lens 144 recombines the filtered light onto an output fiber 146.
Example 2 A dual-polarization liquid-crystal etalon filter 110 was fabricated according to the above procedure. It had a cell gap of 10 μm. Its lateral dimensions were l cm x l cm. Poly 1 ,4 butyleneterephathalate and octadecyltriethoxysilane were used as the aligning agents. The nematic liquid crystal was the previously mentioned E7. The liquid crystal was filled while in its isotropic state into the cell gap of 10 μ using a vacuum filling technique. The alignment layers cannot absolutely determine the alignment direction of the liquid-crystal molecules since there is a degeneracy between the parallel and anti-parallel directions in the homogeneous aligning agent, which would likely result in the formation of multiple domains in each of the homogeneous portions 130 and 132. Multiple domains can be avoided by breaking the symmetry using finite tilt angles at the homogeneous alignment layer 124. However, in the example, a single domain was obtained in each of the homogeneous portions 130 and 132 by filling the gap with liquid crystal 136 while in its isotropic phase and then cooling it while an electric field is applied across it by the electrodes 116 and 118.
The device was characterized at room temperature using an optical spectrum analyzer, a 1.5 μ light emitting diode as a light source, and multi-mode optical fibers. The multi-mode fiber and graded-index rod lens produces a highly collimated with having a diameter of about 400 μm. Use of a single-mode fiber produced a beam diameter of about 250 μ . A programmable and computer-controlled voltage source provided a square wave potential at 1 kHz. The spectral location for the maxima of the transmission peaks are illustrated in the graph of FIG. 10 as a function of the applied voltage. The transmission peaks had widths or pass bands of about 0.9 nm. The transmission spectra show two bands P and Q . The P band does not change with voltage and corresponds to polarization perpendicular to the director n. The Q band, however, does change significantly with voltage. It corresponds to polarization parallel to the director n. The position of the P band is determined by the resonance condition mλ = nsd, (3)
where m is the mode number, λ is the optical wavelength, ns is the refractive index along the short axis of the liquid crystal, and d is the physical thickness of the Fabry-Perot cavity. The applied voltage changes ns by about 12% The figure shows a tuning range of about 100 nm and a free spectral range of about 70 nm for the Q peak. The spectra were remeasured with a polarizer inserted on the input side. Both when the light was polarized perpendicular to the interface 134 and when parallel to the interface 134, the spectra did not differ significantly from those of FIG . 10.
Because the P band is not tunably filtered, the filter needs to be designed with the desired free spectral range avoiding this band.
The filter of FIG. 8 suffers the disadvantage that there is a 3 dB loss associated with the unfiltered polarization or P band. That is, the lens 142 distributes equal amounts of both optical polarizations to the two portions 130 and 132. However, in each portion 130 or 132, only that polarization corresponding to the buffing direction is selected accorded to the wavelength. The other polarization is not affected by the applied
field and is thus necessarily blocked.
A general method of avoiding this 3 dB loss is to recognize that the lens 142 acts as a spatial beam splitter that is polarization insensitive. There are many well known types of polarization beam splitters which separate an input beam into two output beams according to the polarization , for example, a Wollaston prism . Such a polarization beam splitter would receive light from the input optical fiber 140 and deliver the respective polarization components to the respective portions 130 and 132 which filter that polarization. A sixth embodiment of the invention is a particularly advantageous tunable dual-polarization liquid-crystal etalon filter 150 that uses a polarization beam splitter, as illustrated in the cross-sectional view of FIG 11. It is fabricated on substrates 152 and 154 of birefringent material which divides a beam according to its polarization state. On the input side, the birefringent substrate 152 splits the beam from the input fiber into an ordinary beam 156 and an extraordinary beam 158. If the birefringent optical axis is set perpendicularly to the long axis of the interface 134 of the homogeneous alignment layer 124 and preferably at 45° to the input optical axis, the ordinary beam 156 is undeflected while the extraordinary beam 158 is deflected. The light in the ordinary beam 156 is electrically polarized parallel to the interface 134 while that in the extraordinary beam 158 is polarized perpendicularly to it. At the interface between the birefringent substrate 152 and the transparent electrode .116, the two beams 156 and 158 return to their original directions and pass through the two portions 130 and 132 respectively containing the homogeneous aligning agent of the proper alignment. Importantly, the two beams 156 and 158 are well separated at this point so that the diameter and alignment of the input beam are not critical. After passing through the liquid crystal 136, the two beams are recombined by the other birefringent substrate 154 set with its birefringent optical axis set to mirror that of the first birefringent substrate 152. Although a non-polarization beam combiner could be used instead, the two birefringent substrates 152 and 154 are easily matched and provide an integrated assembly. Rather than using birefringent
substrates, birefringent layers can be permanently fixed on glass substrates. The liquid-crystal dual-polarization filter 150 avoids the 3 dB loss because all of the polarization component is delivered to that portion 130 or 132 that can completely filter it. This filter 150 has the further advantage that the P band is eliminated. If a broad-area photodetector is used, the birefringent layer 154 on the output side is not needed.
Example 3 The design of the integrated dual-polarization liquid-crystal filter with polarization beam splitters and combiners was verified by clamping calcite plates of 4 mm thicknesses to the outsides of the soda- lime glass substrates 112 and 114 of the dual-polarization liquid-crystal filter 10 of Example 2. The filter was tested by inputting a beam from a laser and detecting its output. When a manually operated polarization controller rotated the polarization direction of the input light, the output intensity varied by less than 1 dB. On the other hand, when a polarizing sheet was additionally inserted across the input beam, the polarization controller produced 25 dB variations of the output intensity.
Because the dual polarization filter 150 of FIG . 11 is insensitive to polarization, it can be advantageously used in an optical drop circuit. WDM communication systems impress multiple optical signals at wavelengths λj . . .kN on a single optical fiber. The N-fold optical signal will be represented as ∑λ. Various types of filters can be used to select the channel λk from the N channels of ∑λ. However, a simple filter will discard the other N — 1 channels. The optical equivalent of an electronic drop circuit would be advantageous in which the λk channel is removed from the fiber but the other N — 1 channels remain on the fiber. An optical drop circuit 170, illustrated in FIG. 12 and a seventh embodiment of the invention, performs such a function.
An input fiber 172 carries the WDM signal ∑λ, which is divided by a first polarization beam splitter 174 into a first beam 176 and a second beam 178. The illustrated type of polarization beam splitter 174 is a cube of quartz split along a diagonal to form a planar interface. The interface is covered with one or more dielectric layer, and the cube is assembled. Light polarized parallel to the interface is reflected at 90°
while the orthogonal polarization is transmitted. The first beam 176 initially carries the polarization components of the WDM signal that are polarized within the plane of the illustration while the second beam 178 carries the orthogonally polarized components. When the first beam 176 passes through a quarter-wave plate 180, all its frequency components become circularly polarized. The circularly polarized beam 176 is incident on a first liquid-crystal dual-polarization filter 182 having the structure of the filter 150 of FIG . 11 and tuned to the selected wavelength λk. The filter 182 spatially splits the circularly polarized first beam 176 into its two linearly polarized components, passes both linearly polarized components at λ*, and recombines the λk components into a circularly polarized beam . Another quarter-wave plate 184 converts the circularly polarized \k channel to the linear polarization perpendicular to the original polarization, that is, perpendicular to the illustration. The other N — 1 optical channels ∑λ — λ^ which do not pass the filter 182 are reflected with minimal loss of energy. Upon their reverse passage through the first quarter-wave plate 180, they become linearly polarized but perpendicularly to their original polarization . Therefore , the polarization beam splitter 174 reflects them to an unselected output beam 86 received by an output fiber 188.
The quarter-wave plates 180 and 184 can be replaced by magneto-optical devices that rotate the linear polarization of a light beam by 45° and rotate by the same angle when the beam travels in the reverse direction, that is, a 90° rotation for a double passage. Thereby, equal amounts of the two polarizations are presented to the filter 182, and the polarization beam splitter 174 directs the reflected beam away from the input beam . The second magneto-optical device reestablishes the original polarization direction, if this is required.
The structure described so far would operate as an optical drop circuit if the linear polarization of the WDM signal could be controlled. However, to compensate for a lack of such control, similar quarter-wave plates 190 and 192 and dual-polarization filter 194 are needed to perform similar filtering and reflection of the orthogonally polarized components in the second beam 178. The reflected components have their linear
polarization rotated so that they pass through the first polarization beam splitter 174 to the output fiber 188. That is, the beam splitter 174 also acts as a combiner of the two linearly polarized components of the unselected signal ∑λ — kk. Mirrors 196 and 198 bring the two polarization components of the selected channel κk to a second polarization beam splitter 200, which combines them into a selected output beam 202 received by another output fiber 204. Regardless of the polarization state of the input channels ∑λ, the drop circuit 170 performs the same selection into the selected output beam 202 and the same reflection into the unselected output beam 186.
In one array embodiment illustrated in plan view in FIG . 13, the homogeneous alignment layer 124 is divided into a plurality of orthogonally buffed homogeneous portions 130 and 132. The dual- polarization filter 110 of FIG. 8, when combined with the array of FIG. 13, can filter three beams at the interfaces 134 and 160. However, the dual-polarization filter 150 of FIG . 11 can filter only two beams at the equivalent interfaces 134. In another array embodiment, the multiple beams are arrayed along the interface 134 between the two alignment layer portions 130 and 132. One electrode 116 or 118 is patterned to provide separate electrodes for the beams.
Although the above embodiments provide a filter having a narrow enough tunable pass band so as to separate out a portion of a wide spectrum , the invention is equally applicable to a modulator having a relatively wide but tunable pass band. For instance, the mirrors may be metallic, as disclosed by Saunders above, and combined with the electrodes. For purposes of the fifth and sixth embodiments, a modulator will be considered as a special case of a filter.
In the fifth and sixth embodiments, the homogeneous alignment layer was buffed parallel and perpendicular to the interface 134 between the two homogeneous portions 130 and 132. However, the dual-polarization effect can be obtained for orthogonal buffing at different angles with respect to the interface. The embodiments described above used a uniform homogeneous aligning agent in the second alignment layer 120. However, the same effect can be obtained if the
second alignment layer is formed of a homeotropic aligning agent that is patterned similarly to the first alignment layer. Then, the two alignment layers need to be precisely assembled so that their respective interfaces 134 are aligned. The homogeneous aligning agent can be aligned in parallel between the two alignment layers or aligned perpendicularly so as to impart a 90° twist to the liquid crystal across the gap.
The liquid-crystal etalon optical filters of the invention involving interference mirrors provides for a narrow pass band with wide electrical tunability. They are low-powered, economical to fabricate, and rugged.