Patent Application
20070035800
The invention relates to optical modulators in general and particularly to optical modulators that use light to modulate light and that employ materials, such as polymers, having large nonlinear optical characteristics.
The structure of telecommunications systems today is fundamentally different from that of transistor-based electronics. Broadly speaking, computation is not done today commercially in the optical domain; computation and logic is done with transistor-based logic. Fiber optics are often used for transmitting high speed data streams over longer distances, while slower and shorter-distance communications continues to be dominated by electronic signaling which is often done with copper wires or with short range wireless communication systems, such as WiFi. This is partially a result of the high cost of optical devices, and partially a result of the complexity and cost of the electronics required for high-bit-rate applications.
The very fastest commercially available optical detectors and modulators available today are limited by free-carrier diffusion speeds and by the speed of the supporting electronics to the Gigahertz frequency range. The speeds of such devices have been relatively static for several years, and cannot be expected to increase dramatically in the near future. The electronics to generate high-rate bit streams, and the amplifier electronics required in order to recover high speed signals from high-speed detectors are both quite complex and expensive at speeds exceeding approximately 10 Gb/s.
The field of nonlinear optics is extremely rich in results, and has been around for many years. Basically the premise of nearly all measurements in the field is that one introduces a sufficiently high power flux (or “fluence,” a term of art) in an optical material, it is often possible to excite nonlinear behavior, meaning that the properties of the material change with the input optical power. This kind of effect is very often described through the use of, for instance. Chi2 (χ2) and Chi3 (χ3) which are material dependent constants that describe the strength of two of the relevant nonlinear optical activities of a material. Some nonlinearities, which are material dependent, will work at the full optical frequency, while others are slower. Recently, engineered organic materials have begun to be used for nonlinear optics, because they can be designed to have extremely large χ2 and χ3 moments.
It would be desirable to be able to perform computations or analog signal processing purely in the optical domain, without the data stream having to be converted into an electrical signal by a detector. There is a need for systems and methods that can fully exploit the optical properties of materials that exhibit large χ2 and χ3 moments without having to provide excessive amounts of optical power to do so.
In one aspect, the invention relates to an apparatus for modulating light with light. The apparatus comprises a substrate having an insulating surface; a high index contrast waveguide adjacent the insulating surface, the high index contrast waveguide having a first input port for receiving a first input light beam having a first frequency, a second input port for receiving a second input light beam having a second frequency different from the first frequency, a third input port for receiving a third input light beam with a third frequency different from at least one of the first and second frequencies, and an output port for providing an output light beam; and a cladding adjacent the high index contrast waveguide, the cladding comprising a material that exhibits an enhanced nonlinear optical coefficient. The high index contrast waveguide and the cladding are configured so that, when the first input light beam is provided as a first continuous-wave laser beam having a first frequency, and the second input light beam is provided as a second continuous-wave laser beam having a second frequency, the output light beam appearing at the output port includes a modulated signal at the third frequency having a modulation frequency equal to a difference between the first frequency of the first input light beam and the second frequency of the second input light beam.
In one embodiment, one of the first continuous-wave laser beam and the second input light beam is amplitude modulated. In one embodiment, the first and second input light beams are provided at the same input port.
In another aspect, the invention features an apparatus for modulating light with light. The apparatus comprises a substrate having an insulating surface; a high index contrast waveguide adjacent the insulating surface, the high index contrast waveguide having a first input port for receiving a first input light beam having a first frequency, a second input port for receiving a second input light beam having a second frequency different from the first frequency, and an output port for providing an output light beam; and a cladding adjacent the high index contrast waveguide, the cladding comprises a material that exhibits an enhanced nonlinear optical coefficient. The high index contrast waveguide and the cladding are configured so that, when the first input light beam is provided with an amplitude modulation at a predefined frequency, and the second input light beam comprises no amplitude modulation, an output light beam includes an amplitude modulation at the predefined frequency on the second light beam at the second frequency.
In one embodiment, the enhanced nonlinear optical coefficient is an enhanced χ3 coefficient. In one embodiment, the substrate is a silicon wafer. In one embodiment, the insulating surface is a layer comprises silicon and oxygen. In one embodiment, the high index contrast waveguide adjacent the insulating surface is silicon.
In one embodiment, the cladding adjacent the high index contrast waveguide is an optical polymer. In one embodiment, the cladding comprises a material that exhibits an enhanced nonlinear optical coefficient is an electro-optic polymer material. In one embodiment, the high index contrast waveguide is configured as a Mach-Zehnder interferometer having at least two arms. In one embodiment, the first and the second input light beams interact in one arm of the Mach-Zehnder interferometer.
In one embodiment, the apparatus for modulating light with light further comprises an optical cavity that enhances an optical field strength of at least one optical beam. In one embodiment, the optical cavity that enhances an optical field strength of at least one optical beam comprises a ring configuration. In one embodiment, the optical cavity that enhances an optical field strength of at least one optical beam comprises a grating configuration. In one embodiment, the optical cavity that enhances an optical field strength of at least one optical beam comprises a Fabry-Perot configuration.
In one embodiment, one of the first input light beam and the second input light beam comprises a combination of a first pump light beam having a first pump frequency and a second pump light beam having a second pump frequency, the combination of the first pump light beam and the second pump light beam providing a modulation source beam having a selected frequency corresponding to a difference between the first pump frequency and the second pump frequency.
In one embodiment, an optical logic gate comprises at least one apparatus for modulating light with light. In one embodiment, the logic gate is configured as a NAND gate. In one embodiment, the logic gate is configured as a XOR gate. In one embodiment, a latch comprises at least one optical logic gate. In one embodiment, the logic gate is configured as an AND gate. In one embodiment, the logic gate is configured as an OR gate.
In a further aspect, the invention provides an apparatus for modulating light with light. The apparatus comprises a substrate having an insulating surface; and a high index contrast waveguide adjacent the insulating surface, the high index contrast waveguide comprises a waveguide core exhibiting a third order optical nonlinearity, the high index contrast waveguide having a first input port for receiving a first input light beam having a first frequency, a second input port for receiving a second input light beam having a second frequency different from the first frequency, and an output port for providing an output light beam. The high index contrast waveguide configured so that, when the first input light beam is provided with an amplitude modulation at a predefined frequency, and the second input light beam comprises no amplitude modulation, an output light beam includes an amplitude modulation at the predefined frequency on the second light beam at the second frequency.
In yet a further aspect, the invention provides an apparatus for modulating light with light. The apparatus comprises a substrate having an insulating surface; and a high index contrast waveguide adjacent the insulating surface, the high index contrast waveguide comprising a waveguide core exhibiting a third order optical nonlinearity, the high index contrast waveguide having a first input port for receiving a first input light beam having a first frequency, a second input port for receiving a second input light beam having a second frequency different from the first frequency, a third input port for receiving a third input light beam with a third frequency different from at least one of the first and second frequencies, and an output port for providing an output light beam. The high index contrast waveguide is configured so that, when the first input light beam is provided as a first continuous-wave laser beam having a first frequency, and the second input light beam is provided as a second continuous-wave laser beam having a second frequency, the output light beam appearing at the output port includes a modulated signal at the third frequency having a modulation frequency equal to a difference between the first frequency of the first input light beam and the second frequency of the second input light beam.
In still another aspect, the invention features an optical transistor. The optical transistor comprises a substrate having an insulating surface; a high index contrast waveguide adjacent the insulating surface, the high index contrast waveguide having a source input port for receiving a first input light beam having a first frequency, a gate input port for receiving a second input light beam having a modulation frequency, and a drain output port for providing an output light beam; and a cladding adjacent the high index contrast waveguide, the cladding comprises a material that exhibits an enhanced nonlinear optical coefficient. In the optical transistor, an amount of modulation on an output beam that is provided at the drain output port is greater in absolute power swing than an amount of modulation on the second input light beam provided at the gate input port.
In an additional aspect, the invention provides an optical transistor. The optical transistor comprises a substrate having an insulating surface; and a high index contrast waveguide adjacent the insulating surface, the high index contrast waveguide comprising a waveguide core exhibiting a third order optical nonlinearity, the high index contrast waveguide having a source input port for receiving a first input light beam having a first frequency, a gate input port for receiving a second input light beam having a modulation frequency, and a drain output port for providing an output light beam. The amount of modulation on an output beam that is provided at the drain output port is greater in absolute power swing than an amount of modulation on the second input light beam provided at the gate input port.
In yet another aspect, the invention provides a method of optically processing light. The method comprises the steps of providing a structure that comprises: a substrate having an insulating surface, a high index contrast waveguide adjacent the insulating surface, the high index contrast waveguide having a first input port for receiving a first input light beam having a first frequency, a second input port for receiving a second input light beam having a second frequency different from the first frequency, a third input port for receiving a third input light beam having a third frequency different from at least one of the first frequency and the second frequency, and an output port for providing an output light beam; and a cladding adjacent the high index contrast waveguide, the cladding comprising a material that exhibits an enhanced nonlinear optical coefficient; providing a first continuous-wave laser beam having a first frequency; providing a second continuous-wave laser beam having a second frequency different from the first frequency, providing a third continuous-wave laser beam having a third frequency different from at least one of the first frequency and the second frequency, and observing an output light beam at the output port, the output light beam comprises a modulated signal having a modulation frequency equal to a difference between the first frequency of the first input laser beam and the second frequency of the second input laser beam.
In one embodiment, at least two of the first, the second and the third input light beams are provided at the same input port. In one embodiment, at least one of the steps of providing a first input laser beam having a first frequency, providing a second input laser beam having a second frequency different from the first frequency and providing a third continuous-wave laser beam having a third frequency different from at least one of the first frequency and the second frequency involves providing an input laser beam using an input waveguide that communicates with the high index contrast waveguide with a coupler.
In still another aspect, the invention features a method of optically processing light. The method comprises the steps of: providing a structure that comprises: a substrate having an insulating surface, a high index contrast waveguide adjacent the insulating surface, the high index contrast waveguide having a first input port for receiving a first input light beam having a first frequency, a second input port for receiving a second input light beam having a second frequency different from the first frequency, and an output port for providing an output light beam; and a cladding adjacent the high index contrast waveguide, the cladding comprises a material that exhibits an enhanced nonlinear optical coefficient; providing a first input light beam having a first frequency, and having an amplitude modulation at a predefined frequency; providing a second unmodulated input light beam having a second frequency different from the first frequency, and observing an output light beam having an amplitude modulation at the predefined frequency on the second light beam at the second frequency.
In one embodiment, at least one of the steps of providing a first input light beam having a first frequency and providing a second input light beam having a second frequency different from the first frequency involves providing either input light beam using an input waveguide that communicates with the high index contrast waveguide with a coupler.
The foregoing and other objects, aspects, features, and advantages of the invention will become more apparent from the following description and from the claims.
The objects and features of the invention can be better understood with reference to the drawings described below, and the claims. The drawings are not necessarily to scale, emphasis instead generally being placed upon illustrating the principles of the invention. In the drawings, like numerals are used to indicate like parts throughout the various views.
Panel A of
Panel B of
Panel C of
Panel D of
We describe herein an all-optical modulator in a scaleable, integrated silicon system that works at ultrafast speeds, up to 2.6 Terahertz. Such a device is expected to be useful as a practical ultrafast wavelength converter. The intensity modulation of light with light in silicon-polymer integrated waveguide devices is based upon the all-optical Kerr Effect, the same ultrafast effect used in four-wave mixing. It is also expected that the integrated silicon system described will allow the interconnection of the all-optical modulator with electronic systems for interfacing the all-optical modulator with more conventional electronic systems, for example to provide user input and output capabilities.
By integrating optical polymers through evanescent coupling to high-mode-confinement silicon waveguides, the effective nonlinearity of the waveguide is greatly increased for cross-phase modulation. The combination of high mode confinement, multiple integrated optical components, and high nonlinearities produces all-optical ultrafast devices operating at power levels compatible with telecommunication systems. These devices are a first step toward large scale integrated ultrafast optical logic in silicon. They operate at frequencies that are two orders of magnitude faster than present silicon devices.
High index contrast waveguides as described herein are useful to concentrate light in order to enhance nonlinear optical effects in various materials so that such effects can be employed to manipulate light (or more generally electromagnetic radiation) at low power levels, as compared to conventional systems and methods that employ nonlinear optical materials. The manipulation of electromagnetic radiation or light can be useful to provide a variety of components that perform operations on light such as rectification and logic operations in a manner analogous to the same operations which are provided using electronic devices operating on electrical signals. For example, an input a light wave to be processed is impressed onto the component. The light wave has at least one parameter characterizing the light wave, such as one of an intensity, a polarization, a frequency, a wavelength, and a duration (e.g., a pulse length, or in the case of continuous wave light, an effectively infinite duration). After the input light wave is processed (or interacts with the waveguide and the clad nonlinear optical material adjacent to the waveguide), an output signal is observed. In a circumstance where the input signal has been processed, the output signal has at least one parameter that is different from at least one parameter characterizing the input light wave, including possibly an electrical output signal when the input light wave had no electrical signal component (e.g., optical rectification).
As described in greater detail herein, the present invention provides methods and structures that exhibit enhancement of the nonlinear effects in various electro-optical materials that is sufficient to make the nonlinear effects accessible with continuous-wave, low-power lasers. As is described herein the waveguide is coated or clad with another material which provides or exhibits an enhanced nonlinear optical coefficient, such as certain kinds of organic electro-optical materials that can be specifically designed to operate in various regions of the electromagnetic spectrum. It is to be understood that if the high contrast waveguide core material itself exhibits a sufficiently large nonlinear optical coefficient of the correct order, e.g., a χ2 or a χ3 coefficient, the cladding may be omitted and the waveguide core itself can provide the nonlinear optical effects of interest. We have demonstrated that some designs of high index contrast waveguides are designed to concentrate light in the cladding. In some embodiments, the waveguide is a split waveguide. In some embodiments, the split waveguide is coated with a material which provides an enhanced nonlinear optical coefficient. In some embodiments, the two sides of the split waveguide also comprise electrodes that are used for polling a χ2 material introduced into the gap. As described herein, in some embodiments, the dispersion of a waveguide is engineered to enhance the optical power in the mode by slowing the propagation of the light. In some embodiments the waveguides are segmented waveguides. As discussed herein, the waveguide can provide optical field enhancement when the structure is arranged into a resonator, which in various embodiments can be either a ring resonator or a linear resonator. It is believes that appropriate claddings can comprise one or more of glass, semiconductor, quantum dots, saturable absorbers, quantum dots doped into an organic mains, electro-optic materials such as polymers and dendrimers, polymers or other organic materials providing large χ3 coefficients, or other nonlinear optical material to provide large optical nonlinearities through field enhancement in the cladding. In some embodiments, the systems and methods of the invention can be used to provide a tunable infrared source. In some embodiments, by using a low power tunable laser and a high power fixed wavelength laser as the inputs, it is possible to produce a high power coherent tunable source. The tunable source can be a widely tunable coherent source. In addition, using systems and methods of the invention, the use of an incoherent input light source can result in an incoherent tunable source. With the provision of on-chip feedback, the systems and methods of the invention can be used to provide devices that exhibit optical self-oscillation. In some embodiments, the central high index waveguide comprises an amplifying medium, such as a gallium arsenide stripe laser. In some embodiments, where the cladding material exhibits nonlinearities, the laser can be operated as a pulsed source. In some embodiments, systems and methods of the invention can be constructed to provide optical logic functionality, such as optical AND or optical flip-flops. It is believed that systems and method according to the invention can be employed to create optical NAND, OR, NOR and XOR gates, and optical latches, or optical memory. In some embodiments, the systems of the invention can further comprise pump lasers integrated onto the same chip. In some embodiments, the systems of the invention can further comprise off-chip feedback or amplification for frequency conversion or pulse generation. In some embodiments, an additional electrical signal is coupled into the structure to provide active modelocking.
We have developed a set of tools for concentrating light to a high degree by using silicon or other high index contrast waveguides, and we have fabricated devices that demonstrate some of the many applications that can be contemplated when such nonlinear materials are exploited. In particular, by utilizing split waveguides, we are able to greatly enhance the optical fields in the cladding of a tightly confined waveguide, without greatly enhancing the optical losses of the same waveguide. Combining the high field concentrations available from the split waveguides with the high nonlinear activity of nonlinear optical polymers permits the development of nonlinear optical devices operating at much lower optical input power levels than are possible with conventional free space or chip based systems. We have demonstrated four-wave mixing (which is based upon χ3), as well as optical rectification (based on χ2), in such waveguides. Using these waveguides it is possible to decrease the power levels needed to observe significant nonlinearities to the point where, by contrast with conventional nonlinear optics, it can be done with non-pulsed, continuous wave lasers.
Chi2 (χ2) and Chi3 (χ3) based optical effects can be used in particular to build on-chip optical parametric oscillator (“OPO”) systems, where two input wavelengths can be mixed together to produce sum and difference frequencies. These frequencies can be either higher or lower than the input frequencies, and can be made tunable. These effects work for frequencies from the ultraviolet and X-ray regime all the way out into the far infrared and microwave, and in fact can work down to DC in some cases, particularly with optical rectification.
The material of which the high index waveguide is made can be any material having a high index that is reasonably transparent at the wavelengths of interest. This can include but is not limited to silicon, gallium nitride, indium phosphide, indium gallium nitride, gallium phosphide, diamond, sapphire, or the various quaternary III/V and II/VI materials such as aluminum gallium arsenide phosphide. III/V denotes materials having at least one element from column III of the periodic table of elements (or an element that is stable as a positive trivalent ion) and at least one element from column V (or an element that is stable as a negative trivalent ion). Examples of III/V compounds include BN, AIP, GaAs and InP. II/VI denotes materials having at least one element from column II of the periodic table of elements (or an element that is stable as a positive divalent ion) and at least one element from column VI (or an element that is stable as a negative divalent ion). Examples of II/VI compounds include MgO, CdS, ZnSe and HgTe.
We will now present a more detailed description of the systems and methods of the invention, including successively the mechanical structure of exemplary embodiments of high index waveguides, exemplary embodiments of cladding materials having large nonlinear constants χ2 and χ3 and their incorporation into devices having high index waveguides, exemplary results observed on some of the fabricated devices that are described, and some theoretical discussions about the devices and the underlying physics, as that theory is presently understood. Although the theoretical descriptions given herein are believed to be correct, the operation of the devices described and claimed herein does not depend upon the accuracy or validity of the theoretical description. That is, later theoretical developments that may explain the observed results on a basis different from the theory presented herein will not detract from the inventions described herein.
Exemplary High Index Waveguide Structures
Resonators comprising high-Q microrings were fabricated from thin silicon-on-insulator (SOI) layers. Measured Q values of 45 000 were observed in these rings, which were then improved to 57 000 by adding a PMMA cladding. Various waveguide designs were calculated, and the waveguide losses were analyzed.
Microring resonator structures as laser sources and as optical filter elements for dense wavelength division multiplexing systems have been studied in the past. The silicon-on-insulator (SOI) structure described here is particularly advantageous. It has low waveguide loss. One can extrapolate an uncoupled Q value of 94 000 and a waveguide loss of 7.1 dB/cm in the unclad case, and −6.6 dB/cm in the PMMA clad case, from the respective measured Q values of 45 000 and 57 000. Although higher Q values have been obtained for optical microcavities, we believe that our geometry has the highest Q for a resonator based on a single mode silicon waveguide. It is also noteworthy that a large amount of power appears outside the core silicon waveguide, which may be important in some applications. The modes that are described herein have approximately 57% of the power outside the waveguide, as compared to 20% for a single-mode 200-nm-thick silicon waveguide, and 10% for a single-mode 300-nm-thick silicon waveguide.
In the embodiment now under discussion, wafer geometries were selected that minimize the thickness of the SOI waveguiding layer as well as the buried oxide, but still yield low loss waveguides and bends. A number of different waveguide widths were compared by finite difference based mode solving. The geometry used in the exemplary embodiment comprises a 500-nm-wide waveguide formed in a 120-nm-thick silicon layer, atop a 1.4 μm oxide layer, which rests on a silicon handle, such as a silicon wafer as a substrate. Such a configuration supports only a single well-contained optical mode for near infrared wavelengths. The dispersion characteristics are shown in
These modes were determined by using a finite difference based Hermitian eigensolver, described further herein. It is possible to calculate the loss directly from the mode pattern with an analytic method valid in the low-loss limit. The waveguide loss at 1.55 μm calculated in such a fashion is approximately −4.5 dB. This loss figure was in agreement with the extrapolated results of FDTD simulation.
Because a loss of −4 dB/cm is attributed to substrate leakage, the waveguide loss can be improved by the addition of a cladding, which tends to pull the mode upwards. This notion is supported by the measured decrease in waveguide loss upon the addition of a PMMA cladding. It can be shown that the substrate leakage loss attenuation coefficient is nearly proportional to
e−2√{square root over (n2eff−no2k0A)}
if k0 is the free space wave number, neff is the effective index of the mode, no is the effective index of the oxide layer, and A is the thickness of the oxide. In the present case, the e-folding depth of the above-mentioned function turns out to be 180 nm, which explains why the substrate leakage is so high.
SOI material with a top silicon layer of approximately 120 nm and 1.4 μm bottom oxide was obtained in the form of 200 mm wafers, which were manually cleaved, and dehydrated for 5 min at 180° C. The wafers were then cleaned with a spin/rinse process in acetone and isopropanol, and air dried. HSQ electron beam resist from Dow Corning Corporation was spin coated at 1000 rpm and baked for 4 min at 180° C. The coated samples were exposed with a Leica EBPG-5000+ electron beam writer at 100 kV. The devices were exposed at a dose of 4000 μc/cm2, and the samples were developed in MIF-300 TMAH developer and rinsed with water and isopropanol. The patterned SOI devices were subsequently etched by using an Oxford Plasmalab 100 ICP-RIE within 12 mTorr of chlorine, with 800 W of ICP power and 50 W of forward power applied for 33 s. Microfabricated devices such as the one shown in
The theoretical development of the expected behavior of a ring resonator system has been described in the technical literature. In the present case the dispersion of the waveguide compels the addition of a dispersive term to the peak width. We take λ0 to be the free space wavelength of a resonance frequency of the system, n0 to be the index of refraction at this wavelength, (δn/δλ)0, the derivative of n with respect to λ taken at λ0, L to be the optical path length around the ring, a to be the optical amplitude attenuation factor due to loss in a single trip around the ring, and finally t to be the optical amplitude attenuation factor due to traveling past the coupling region. In the limit of a high Q, and thus
(1−α)1 and (1−t)
1,
we have
The waveguide mode was coupled into a ring resonator from an adjacent waveguide. As shown in
From the mode-solving results for the unclad waveguides, we have (δn/δλ)(1.512)=−1.182 μm−1, and n(λ=1.512)=1.688. Using this result and the earlier relations, the waveguide loss can be calculated from the measured Q value. Specifically, an extinction that is at least −22 dB indicates that a critically coupled Q in this geometry is greater than 38 500, which then implies a waveguide loss of less than −7.1 dB/cm. In similar fashion, the PMMA clad waveguide resonator with a Q of 57 000 but only −15.5 dB of extinction allows a worst case waveguide loss of −6.6 dB/cm. This also implies an instrinsic Q of 77 000 for the unclad resonator, and an intrinsic Q of 94 000 for the PMMA clad resonator.
These devices have a slight temperature dependence. Specifically, the resonance peak shifts correspondingly with the change in the refractive index of silicon with temperature, moving over 2 nm as temperature shifts from 18 to 65° C. The Q rises with higher temperatures slightly, from 33 k at 18° C. to 37 k on one device studied. This shift can probably be explained entirely by the dependence of Q on the effective index.
We now describe the design, fabrication and characterization of high Q oval resonators based on slot waveguide geometries in thin silicon on insulator material. Optical quality factors of up to 27,000 were measured in such filters, and we estimate losses of −10 dB/cm in the slotted waveguides on the basis of our resonator measurements. Such waveguides enable the concentration of light to very high optical fields within nano-scale dimensions, and show promise for the confinement of light in low-index material with potential applications for optical modulation, nonlinear optics and optical sensing. As will be appreciated, the precise geometry of a resonator (or other kinds of devices) is frequently a matter of design, and the geometry can be varied based on such considerations as length of waveguide, area of a chip, and required interaction (or required non-interaction), such as coupling (or avoiding coupling) with other waveguide structures that are present in a device or on a chip. In some embodiments, the waveguide can be a closed loop, such as at least one ring or at least one oval shaped endless stripe. As has been explained, optical energy can be provided to such a closed loop, for example with an input waveguide.
One can form high quality factor ring or oval resonators in SOI. In these SOI waveguides, vertical confinement of light is obtained from the index contrast between the silicon core and the low index cladding and the buried silicon dioxide layer, whereas lateral confinement can be obtained by lithographically patterning the silicon. The majority of the light tends to be guided within the silicon core in such waveguide. Although the high refractive index contrast between silicon and its oxide provide excellent optical confinement, guiding within the silicon core can be problematic for some applications. In particular, at very high optical intensities, two-photon absorption in the silicon may lead to high optical losses. Moreover, it is often desirable to maximize the field intensity overlap between the optical waveguide mode and a lower index cladding material when that cladding is optically active and provides electro-optic modulation or chemical sensing.
One solution to these problems involves using a slot waveguide geometry. In a slot waveguide, two silicon stripes are formed by etching an SOI slab, and are separated by a small distance. In one embodiment, the separation is approximately 60 nm. The optical mode in such a structure tends to propagate mainly within the center of the waveguide. In the case of primarily horizontal polarization, the discontinuity condition at the cladding-silicon interface leads to a large concentration of the optical field in the slot or trench between the two stripes. One can predict that the electric field intensity would be approximately 108 √P V/m where P is the input power in watts.
In the 1.4-1.6 μm wavelength regime, the waveguide geometry is single mode, and a well-contained optical mode is supported between the two silicon waveguide slabs. There is some loss that such an optical mode will experience even in the absence of any scattering loss or material absorption due to leakage of light into the silicon substrate. The substrate loss can be estimated semi-analytically via perturbation theory, and ranges from approximately −0.15 dB/cm at 1.49 μm to about −0.6 dB/cm at 1.55 μm for the SOI wafer geometry of the present embodiment.
Oval resonators were fabricated by patterning the slot waveguides into an oval shape. An oval resonator geometry was selected in preference to the more conventional circular shape to enable a longer coupling distance between the oval and the external coupling waveguide or input waveguide. See
Predicting coupling strength and waveguide losses for such devices is not easy. Many different coupling lengths and ring to input waveguide separations were fabricated and tested. It is well known that the most distinct resonance behavior would be observed for critically coupled resonators, in which the coupling strength roughly matches the round trip loss in the ring.
An analytic expression for the quality factor of a ring resonator was presented in equation (1) hereinabove.
Also, the free spectral range can be calculated via:
Here, L is the round trip length in the ring, and n0 and λ0 are the index of refraction, and the wavelength at resonance, respectively. The derivative of the effective index with respect to the wavelength at the resonance peak is given by (δn/δλ)0, and it can be shown that this term is roughly equal to −0.6 μm−1 from the 1.4-1.6 μm spectral range for the slot waveguides studied here.
We have observed a quality factor of 27,000 in a device fabricated with a slot size of 70 nm, a ring to input waveguide edge to edge separation of 650 nm, and a coupling distance of 1.6 μm. The radius of the circular part of the slotted oval was 50 μm. This resonance was observed near 1488 nm, and the resonance peak had an extinction ratio of 4.5 dB.
Since this resonance is nearly critically coupled, the waveguide loss can be estimated using equation (1) as −10 dB/cm. We can also use equation (2) to further validate our theoretical picture of the ring resonator. The observed free spectral range of this resonator was 2.74 nm, while equation (2) predicts 2.9 mm. This discrepancy is most likely due to small differences in the fabricated dimensions as compared to those for which the numerical solutions were obtained.
To further validate the waveguide loss result, several waveguide loss calibration loops were fabricated with varying lengths of the slot waveguide, ranging from 200 to 8200 um in length. A total of five center slot waveguide devices were studied for each of the 50, 60 and 70 nm slot widths. Linear regression analysis on the peak transmission of each series yielded waveguide loss figures of 11.6±3.5 dB/cm for the 50 nm center waveguide, 7.7±2.3 dB/cm for the 60 nm center waveguide, and 8.1±1.1 dB/cm for the 70 nm center waveguide. These figures are in agreement with the loss estimated from the oval resonator. Since the theoretical loss due to substrate leakage is much lower than this, it is clear that a great deal of loss is due to surface roughness and possibly material absorption. It is believed that engineering improvements will decrease this loss further. For sensing and modulation applications as well as use in nonlinear optics, the high optical field concentration that can be supported in the cladding material of the slotted waveguide geometry should be very advantageous when compared to more conventional waveguides.
By comparison,
Other variations on the geometry of waveguides are possible.
The segmented waveguide structure can also be modeled as regards its expected properties, which can then be compared to actual results.
By utilizing the same type of design methodology as was used for the segmented waveguides, one is able to able to construct structures that provide electrical isolation without substantial optical loss.
Exemplary Results for Waveguides with Cladding Materials
Two types of integrated nano-optical silicon waveguide structures were used for this demonstration. The first type of structure was a series of ring resonator structures, which allowed an estimation of the waveguide loss of the nonlinear material. The second type of structures used were long runout devices, which comprised a simple waveguide loop with distances on the order of 0.7 cm. Characterization of loss could be done passively.
For the actual nonlinear testing, a Keopsys EDFA was used to boost two lasers to a high power level, on the order of 30 dBm (1 Watt) or more.
The materials used for the demonstrations were clad on waveguides configured as previously described herein. The chromophore identified as JSC1 is shown by its chemical structure in
1. JSC1/APC: The chromophore JSC1 is doped into amorphous polycarbonate (APC) with the loading of 35 wt %. The solvent we used is cyclohexanone, and concentration of overall solid in this solution is 14 wt %.
2. AJL21/PMMA: The chromophore AJL21 is doped into PMMA with the loading of 40 wt %. The solvent used was 1,1,2-trichloroethane, and solution concentration was 10 wt %.
3. AJL21 monolithic films: The chromophore AJL21 is coated by itself monolithically. The solvent was 1,1,2 trichloroethane, and the concentration was 10 wt %.
4. AJC212 monolithic films: The chromophore AJC212 was coated by itself monolithically. The solvent was cyclopentanone, and concentration was 11 wt %. This film may have wetting problems, as evidenced by periphery shrinkage after baking.
Passive Results
Waveguide loss was measured for each of the four die. Intrinsic waveguide loss with a cladding having an index of 1.46 is about 7 dB/cm. A cladding with n>1.46 would lower this figure slightly. The total loss and the estimated loss due to the polymer are presented separately. This is based on subtracting 7 dB from the polymer, and then multiplying by three, because the polymer causes approximately as third as much loss as it would for the mode if it were in a bulk material, because not all of the optical energy interacts with the polymer.
Losses were found to be as low as less than 1 dB/cm and as high as 69 dB/cm for the various formulations of polymer and chromophore.
Active Results
The intrinsic nonlinear response of our EDFA and optical test system was measured to determine a baseline for measurements on devices.
A Die 1 loop device with 7000 μm of runlength produced about 29 dB of conversion efficiency (that is, sidebands were 29 dB down from peak at end of run).
The noise level on some of these scans is higher than others because some were taken with faster scan settings on the optical signal analyzer.
Semi-Analytic Results
The slowly varying approximation can be used to generate the characteristic equations to predict the conversion efficiency. Let a0(z), a1(z) and a2(z) be the amplitudes of the 3 wavelengths involved in a given four-wave mixing interaction. Let w2=2*w0−w1. Approximately, E is 108 V/m for 1 Watt of power. so if we take E=a0(z)*108 V/m then a0 is power normalized to be 1 watt when |a0|=1. The characteristic equations are:
The quantity f is taken as an unknown fraction which reduces the effect of the nonlinear material due to the fact that some of the optical energy is not in the optical region, but in the waveguide core. It is estimated that f is about 0.1, with an uncertainty of perhaps a factor of 2.
The phasor factor turns out to have an oscillation period on the order of a meter for the waveguides under consideration, and can be ignored. Based on a numerical integration, the χ3 values for the strongest materials—found on dies 1 and 2—were found to be on the order of 10−22 (m/V)2
The devices that were tested were observed in all cases to eventually fail, either when ramping up the power levels or after extended testing. It is believed that the problem is caused by heating damage. Fortunately the damage seems not to extend to the silicon waveguides. This means that devices that fail in this way can be recovered by stripping the polymers, and then being recoated. With additional experience, solutions for the problem of this damage problem may be identified and solved.
It is unfortunate that the waveguide loss in the die 1 material is so high, because it is a material that exhibits extremely high χ3. Nevertheless, reasonable efficiencies were demonstrated with material exhibiting a lower χ3. It would be advantageous to identify a material with a value of χ3 that is larger by a factor of 10 or so. It would also be advantageous to lower the waveguide loss slightly. With these two adjustments, it would be possible to enter the “strong coupling” regime, so that one might observe 100% conversion in lengths <0.5 cm. One likely possibility would be to lower the optical loss of the Die 1 material, JSC1.
In this example, we describe a system and process that provide low power optical detection and modulation in a slotted waveguide geometry filled with nonlinear electro-optic polymers and present examples that demonstrate such methods. The nanoscale confinement of the optical mode, combined with its close proximity to electrical contacts, enables the direct conversion of optical energy to electrical energy, without external bias, via optical rectification, and also enhances electro-optic modulation. We demonstrate this process for power levels in the sub-milliwatt regime, as compared to the kilowatt regime in which optical nonlinear effects are typically observed at short length scales. The results presented show that a new class of detectors based on nonlinear optics can be fabricated and operated.
Waveguide-based integrated optics in silicon provide systems and methods for concentrating and guiding light at the nanoscale. The high index contrast between silicon and common cladding materials enables extremely compact waveguides with very high mode field concentrations, and allows the use of established CMOS fabrication techniques to define photonic integrated circuits. As we have already explained hereinabove, by using slotted waveguides, it is possible to further concentrate a large fraction of the guided mode into a gap within the center of a silicon waveguide. This geometry greatly magnifies the electric field associated with the optical mode, resulting in electric fields of at least (or in excess of) 106 V/m for continuous-wave, sub-milliwatt optical signals. Moreover, since the slotted geometry comprises two silicon strips which can be electrically isolated, a convenient mechanism for electro-optic interaction is provided. Such waveguides can be fabricated with low loss. We have previously described systems that provide losses below −10 dB/cm.
In the present example, we exploit both the high intensity of the optical field and the close proximity of the electrodes for several purposes. First, we demonstrate detection of optical signals via direct conversion to electrical energy by means of nonlinear optical rectification. An exemplary device comprises a ring resonator with an electro-optic polymer based χ2 material deposited as a cladding. Inside the slot, the high optical field intensity creates a standing DC field, which creates a virtual voltage source between the two silicon electrodes, resulting in a measurable current flow, in the absence of any external electrical bias. Though optical rectification has been observed in electro-optic polymers, typically instantaneous optical powers on the order of 1 kW are needed for observable conversion efficiencies, often achieved with pulsed lasers. The exemplary embodiment provides measurable conversion with less than 1 mW of non-pulsed input, obtained from a standard, low power tunable laser operating near 1500 nm.
In one embodiment, systems and methods of the invention provide standard Pockels' effect based modulation, which is similarly enhanced by means of the very small scale of our device. The close proximity of the electrodes, and ready overlap with the optical mode, causes an external voltage to produce a far larger effective electric modulation field, and therefore refractive index shift, than would be obtained through conventional waveguide designs. In one embodiment, the modulation and refractive index shift is provided by tuning the resonance frequencies of a slot waveguide ring resonator.
Device Fabrication
Waveguide Fabrication
The devices described in this example were fabricated in electronic grade silicon-on-insulator (SOI) with a top layer thickness of 110 nm and an oxide thickness of 1.3 microns. The silicon layer is subsequently doped to approximately 1019 Phosphorous atoms/cm3, yielding resistivities after dopant activation of about 0.025 ohm-cm. Electro-optic (“EO”) polymers were then spin-deposited onto the waveguide structures and subsequently poled by using a high field applied across the slot in the waveguide.
Lithography was performed using a Leica EBPG 5000+ electron beam system at 100 kv. Prior to lithography, the samples were manually cleaved, cleaned in acetone and isopropanol, baked for 20 minutes at 180C, coated with 2 percent HSQ resist from Dow Corning Corporation, spun for two minutes at 1000 rpm, and baked for an additional 20 minutes. The samples were exposed at 5 nm step size, at 3500 μC/cm2. The samples were developed in AZ 300 TMAH developer for 3 minutes, and etched on an Oxford Instruments PLC Plasmalab 100 with chlorine at 80 sccm, forward power at 50 W, ICP power at 800 W, 12 mTorr pressure, and 33 seconds of etch time. The samples were then implanted with phosphorous at normal incidence, 30 keV energy, and 1×1014 ions/cm2 density. The sample was annealed under a vacuum at 950C in a Jipilec Jetstar rapid thermal annealer. The samples were dipped in buffered hydrofluoric acid in order to remove the remnants of electron beam resist from the surface.
After initial optical testing, the samples were coated with YLD 124 electro-optic polymer, and in one case with dendrimer-based electro-optic material. The samples were stored under a vacuum at all times when they were not being tested, in order to reduce the chances of any degradation.
Synthesis of YLD 124 coating solution
To a solution of 0.56 g (0.96 mmol) of 1 and 0.36 g of 2 (1.1 mmol) in 1.5 mL of THF was added 6 mL of absolute ethanol. The mixture was stirred for 6 h at room temperature. The precipitate was collected by filtration and washed by ethanol and methanol. The crude product was dissolved in minimum amount of CH2CI2. The resultant solution was added dropwise to 100 mL of methanol. The product (0.76 g) was collected as dark green precipitate. Yield was 90%. 1H NMR (CDCl3): 8.05 (t, J=13.6 Hz, 1H), 7.45-7.58 (m, 5H), 7.38 (d, J=8.9 Hz, 2H) 6.93 (d, J=15.9 Hz, 1H) 6.79 (d, J=15.9 Hz, 1H), 6.70 (d, J=8.9 Hz, 2H), 6.40-6.25 (m, 3H), 3.80 (t, J=5.8 Hz, 4H), 3.59 (t, J=5.8 Hz, 4H), 2.42 (s, 2H), 2.40 (s, 2H), 1.04 (s, 3H), 0.98 (s, 311), 0.90 (s, 18H), 0.04 (5, 12H). MS (ESP): 879.48 (M+H). UV-Vis (THF): 765 nm. m.p. 173° C.
One part of YLD 124 was mixed with three parts of APC (Poly[Bisphenol A carbonate-co-4,4′-(3,3,5-trimethylcyclohexylidene)diphenol carbonate]). The mixture was dissolved in cyclopentanone. The total solid content (YLD 124 and APC) is about 12%. The resultant solution was filtered through a 0.2 pm filter before being used on the device to provide a cladding layer comprising the chromophore YLD 124.
Measurement Results
Optical rectification based detection
Measurements were performed with single-mode polarization maintaining input and output fibers, grating coupled to slotted waveguides with an insertion loss of approximately 8 dB. Optical signal was provided from an Agilent 81680a tunable laser and in some cases an erbium doped fiber amplifier (“EDFA”) from Keopsys Corporation. A continuous optical signal inserted into a poled polymer ring results in a measurable current established between the two pads, which are electrically connected through a pico-Ammeter. In the most sensitive device, a DC current of ˜1.3 nA was observed, indicating an electrical output power of ˜10−9 of the optical input power (5×10−12 W of output for approximately 0.5 mW coupled into the chip). Control devices, in which PMMA or un-poled EO material was substituted, show no photocurrent.
The fact that there is no external bias (or indeed any energy source) other than the optical signal applied to the system of this embodiment demonstrates conclusively that power is being converted from the optical signal. To establish that the conversion mechanism is actually optical rectification, we performed a number of additional measurements. A steady bias was applied to the chip for several minutes, as shown in Table IA. A substantial change in the photoresponse of the device was observed. This change depends on the polarity of the bias voltage, consistent with the expected influence of repoling of the device in-place at room temperature. Specifically, if the external bias was applied opposing the original poling direction, conversion efficiency generally decreased, while an external bias in the direction of the original poling field increased conversion efficiency.
To further understand the photo-conversion mechanism, 5 EO detection devices were poled with both positive and negative polarities, thus reversing the direction of the relative χ2 tensors. For these materials, the direction of χ2 is known to align with the polling E field direction, and we have verified this through Pockels' effect measurements. In all but one case, we observe that the polarity of the generated potential is the same as that used in poling, and the +V terminal during poling acts as the −V terminal in spontaneous current generation, as shown in Table IB. Furthermore, the polarity of the current is consistent with a virtual voltage source induced through optical rectification. It was observed that these devices decay significantly over the course of testing, and that in one case the polarity of the output current was even observed to spontaneously switch after extensive testing. However, the initial behavior of the devices after polling seems largely correlated to the χ2 direction.
Part A of Table I shows the dependence of the steady state observed current after room temperature biasing with various voltage polarities for one device. The device was originally polled with a ˜12 V bias, though at 110 C. With one exception, applying a voltage in the direction of the original polling voltage enhances current conversion efficiencies, while applying a voltage against the direction of the polling voltage reduces the current conversion efficiencies. It should be noted that the power coupled on-chip in these measurements was less than 1 mW due to coupler loss.
Part B of Table I shows the behavior of several different devices immediately after thermal polling or cycling without voltage. Measurements were taken sequentially from top to bottom for a given device. The only anomaly is the third measurement on device 2; this was after significant testing, and the current observed was substantially less than was observed in previous tests on the same device. We suspect that the polymer was degraded by repeated testing in this case.
A number of measurements were performed to attempt to produce negative results, and to exclude the possibility of a mistaken measurement of photocurrent. The power input to the chip was turned on and off by simply moving the fiber array away from the chip mechanically, without changing the circuit electrically, and the expected change in the electrical output signal of our detector was observed. A chip was coated in polymethylmethacrylate and tested, resulting in no observed photocurrents. Also, when some of the devices shown in Table I were tested before any polling had been performed; no current was observed.
We used a lock-in amplifier to establish a quantitative relationship between the laser power in the EQ material and the photo-current, and achieved a noise floor of about 0.2 pA. This resulted in a reasonable dynamic range for the 10-200 pA photocurrent readings.
Panel A of
As another demonstration of the dependence of the output current on the amount of light coupled into the resonator, we also tuned the laser frequency and measured the output current. As can be seen in
The conversion efficiency from our first measurements is thought to be several orders of magnitude below the ultimate limit, and can be explained by the high insertion losses in our system. In the present embodiment, 75% of the input power in the fiber is not coupled onto the chip. Our low-Q resonators only provide a limited path length within which light can interact with the electro-optic material. Furthermore, by design a great deal of the light in the resonator will be dumped to an output port, and not absorbed. It is expected that with further design and higher Q resonators, the efficiency of these devices can be greatly increased. It is, however, important to note that nothing about this effect depends on the presence of rings. The rings provide a convenient and compact device for observing these effects, but one could just as easily observe optical rectification by using other geometries, such as a long linear, polymer coated, split waveguide, with each side connected to an electrical pad.
Pockels' Effect Modulation
At DC, the Pockels effect was measured by applying varying voltages to the device and observing the device transmission as a function of wavelength. For devices having operative modulation, the resonance peaks were shifted, often to a noticeable degree. To counter the systemic drift due to temperature fluctuations, a series of random voltages were applied to a device under test and the wavelength responses noted. The intersection of a resonance peak and a certain extinction, chosen to be at least 10 dB above the noise floor, was followed across multiple scans. A 2d linear regression was performed, resulting in two coefficients, one relating drift to time, and one relating drift to voltage.
At AC, a square wave input voltage was applied across the device. The input wavelength was tuned until the output signal had the maximum extinction. It was determined what power levels were implied by the output voltage, and then the observed power levels were fit to a wavelength sweep of the resonance peak. This readily allowed the tuning range to be calculated. We successfully measured AC tuning up to the low MHz regime. The limitation at these frequencies was noise in our electrical driving signal path, and not, as far as we can tell, any rolloff in the modulation process itself.
By utilizing new dendrimer-based electro-optic materials, we have achieved 0.042±008 nm/V, or 5.2±1 GHZ/V for these rings. This implies an r33 of 79±15 pm/V. This result is better than those obtained for rings of 750 micron radius, which we believe to be the best tuning figure published to date. By contrast, our rings have radii of 40 microns. We credit our improvement over the previous results mainly to the field enhancement properties of our waveguide geometry.
Terahertz All-Optical Modulators
Ultrafast optical effects are available with response times in the femtosecond regime, while the fastest transistors are limited to the picosecond timescale. It would be desirable to use nonlinear effects to perform data processing tasks, and to have the ability to provide multiple optical transistors in devices to perform ultrafast logical or computation operations.
The basic component of such a system is expected to be an optical logic gate, similar in function to a transistor in the electronics world. Such a gate, which may be referred to as an optical transistor, is expected to provide signal gain—it is expected to allow a small change in input power at a gate to generate a large swing in output power at a drain. There has been extensive work over the past two decades on nonlinear optics aimed at producing such an element. However, practical devices have proven to be elusive. For an optical transistor to be a practical device, it is advantageous for it to have some or all of the performance characteristics enumerated below.
1. Integration
A compelling characteristic of transistors is the ease with which many can be integrated onto a single substrate, as is commonly done today in the semiconductor industry. The marginal cost of producing multiple transistors on the same chip is quite low. For scaleable optical logic to be applied commercially, it is advantageous to be able to inexpensively making many devices connected to one another on the same chip in order to create complex logic circuits.
2. Size
There has been extensive work over the past two decades on nonlinear optics on the benchtop. Although it is possible to construct a wide variety of nonlinear devices, such as optical parametric oscillators, these are generally constructed using discrete components that are individually assembled on an optical bench, which may be provided with vibration damping apparatus just to keep the components from becoming misaligned by accident. On the macro scale, it is difficult to imagine an ultrafast logic system based upon free space optics. The most basic component of a digital logic system is the ring oscillator. Such a device consists, in its most basic form, of a single inverter fed back into itself. Such a device will oscillate at a frequency determined by the ultimate speed of the fundamental element, and the latency of signal propagation around the ring.
Using the present technology, a optimized benchtop nonlinear free-space optical system constructed from individual components might have a path length of the order of 10 cm around such a loop. As a result, the clock frequency of the fastest computations that one could expect to perform in such a system is expected to be limited by the speed of light to around 3 GHz. That is not to say that the response time of the individual elements is limited to that speed; but this is the limit to the speed of feedback within the system based on the physical size of the loop. Due to limitations of this type, optical logic elements, if they are to take advantage of the extremely high speeds available in nonlinear optical effects, need to have path lengths on the order of hundreds of microns or less. To achieve a Terahertz optical ring oscillator, one is expected to need a path length on the order of 300 microns.
3. Speed
For an all-optical logic technology to be commercially important, it is desirable for it to be able to compete in terms of speed with the fastest practical transistors, which have Ft values in the hundreds of GHz today. This means not only the response speed of the individual devices, but the rate at which signals can be fed back into the system needs to be on that same order.
4. Optical Power
Although power levels available in fiber lasers have increased dramatically over the past few years, single-mode continuous-wave fiber amplifiers and lasers with power levels above 30 dBm cost tens of thousands of dollars at present. If each optical transistor requires one of these lasers, then making complex logical circuits will be prohibitively expensive. As a result, it is desirable that an optical logic technology allow the use of single-mode semiconductor lasers operating at power levels below 100 mW, since such lasers are comparatively inexpensive.
One potential solution to such problems is to use pulsed semiconductor lasers, where the laser outputs all of its optical power over very short pulse times. Such a laser provides orders of magnitude in enhancement of the peak optical field, without an associated increase in the total output power. Although this is an attractive idea, at least one problem with such an approach presents itself: The achievable bit rate is limited by the repetition rate of the laser system. Although there is work being performed on high repetition rate lasers to try to have them operate in the hundreds of Gigahertz, such systems are not yet commercially available. In view of the limitations of present day lasers, it is desirable for an optical transistor to operate with continuous-wave light at power levels below 100 mW, where it is possible to inexpensively purchase arrays of Indium Phosphide single mode lasers.
An element that meets the criteria described above makes it possible to construct the optical equivalent of optical amplifiers and transistors. It is expected that one or more of such elements will allow the demonstration of digital logic operating in the optical domain. A first step toward this is to produce an all-optical modulator in a scaleable, integrated silicon system that works at ultrafast speeds. Such a device is expected to be useful as a practical ultrafast wavelength converter.
Device Design Elements: Four-Wave Mixing in Silicon
Our ultrafast modulators take advantage of cross-phase modulation based upon third order nonlinear effects. Such effects are dependant on the cube of the electric field within the waveguide.
D=ε0(εyE+χ2E2+χ3E3)
The χ3 value for silicon is quite small. At telecommunications wavelengths it is reported as being approximately 4×10−21 (M/V)2, as compared to values as high as approximately 3×10−20 (M/V)2, which we have extrapolated from our measured data in polymer systems. A χ3 with an ultrafast response is the source of the cross-phase modulation which produces four-wave mixing. The devices are based upon the silicon-polymer hybrid system discussed earlier. The basic waveguide geometry is a fully etched ridge waveguide of 120 nm silicon on insulator. The waveguides provide substantial modal overlap with the cladding polymer material; on the order of ⅓ of the optical power propagates in the cladding for our standard 0.5 micron wide stripe waveguides.
In an all-polymer waveguide system, close to 100 percent of the light will overlap with the polymer waveguide core, because the waveguide can be composed only of polymer in the core. One might expect, because of the low overlap with the active cladding in the silicon devices, that they would have substantially reduced nonlinear optical activity, compared to all-polymer systems. Despite this factor of about ⅓, the silicon waveguides have a substantial advantage over conventional all-polymer waveguides, namely their high mode confinement. Because the mode is confined in an area that is smaller than one square micron, compared to the approximately 50-100 square microns of a typical polymer waveguide, the E field in the cladding is greatly enhanced. In fact, 1 mW propagating in a conventional single mode fiber or polymer waveguide produces an approximate peak E field of 3×104 V/m, whereas the same optical power propagating in our polymer-clad silicon waveguide produces an approximate peak E field of 3×106 V/m in the cladding.
In order to assess the feasibility of creating all-optical modulators, we first fabricated and tested devices for four-wave mixing. These devices comprise a long waveguide, into which two input wavelengths are coupled. A number of different polymers were coated onto these chips in order to assess their performance, in terms of both optical losses and χ3 coefficients. Waveguide runout length was 7 mm, and they were pumped with light at two wavelengths near 1550 provided with a Keopsys EDFA providing 28 dBm of power.
By contrast, when the devices were coated with the nonlinear optical polymers, the efficiency of the conversion increased by as much as two orders of magnitude.
Modulator Devices
The precise behavior of the entire nonlinear system is complex: not only is there a phase shift in the source signal, but the four-wave mixing process also produces sidebands from both the source and the gate laser, which must be considered in analyzing the behavior of the device. In the low-conversion regime, only a small fraction of the source wavelength is converted into other wavelengths via four-wave mixing, and the primary effect of the four-wave mixing process is to phase shift the signal wave by a small amount. It is straightforward to show the effect of the gate signal on the source signal in this regime. Let ws, wg and as, ag be the frequencies and amplitudes of the source and gate lasers, respectively, and the sideband Wa=ws+(ws−wg). Then,
Here, we have assumed that we can neglect second- and higher order derivatives of the waveguide effective index, which is reasonable for our waveguide geometry. Similar expressions exist for the other sideband at wg+(wg−ws), and for the effect on ag. In the low-conversion regime, the effect of the gate signal is to phase shift the source signal. The power level of the source signal is unchanged. Note also that the dispersion of the waveguide does not affect this phase shift.
Thus, the bandwidth of the device is limited solely by the effect of dispersion on the intensity-modulation envelope of the gate signal as the latter travels through the waveguide. The dispersive properties of the waveguide used for this device are well known from both theory and experiment, and it is straightforward to show that, for a 1-cm-long device, minimal pulse distortion in the gate will occur up to intensity modulations of 1 THz (as explained below in the Methods section).
We now describe the basic principle of operation of the modulator devices we have constructed. An unmodulated signal wavelength is coupled into a Mach-Zehnder interferometer, where it is split substantially equally, each beam passing along one of two paths or arms. A modulated optical signal, which we refer to as a gate signal, is also coupled into one of the two interferometer paths. This modulated optical signal causes a change in the index of refraction of the arm to which it is coupled. This results in a phase delay for the half of the signal wavelength that is traveling through that arm. The two signals that pass through the two arms are recombined at coupler and are provided as an output signal at an output of the Mach-Zehnder interferometer. One observes a change in the output power of the device after the final coupler that recombines the two halves of the signal. Having changed the relative phase of one of the arms, the transmitted power is altered by the presence of the gate signal. For purposes of exposition, we can analogize the optical signals to the signals in a transistor, such as an FET. We can relate the input optical signal provided at the input port, to a source signal provided at a source in a transistor. We can relate the gate signal applied to the Mach-Zehnder to the control signal applied to the gate of an FET. We can relate the output signal of the Mach-Zehnder to a signal appearing at a drain terminal of an FET.
It was necessary to design components which could be used to combine different wavelengths on at least one arm of the interferometer, to combine the gate and source wavelengths. In the present embodiment, we elected to use 3 dB directional couplers on both arms, which are highly repeatable and broadband, though they incur a substantial optical loss. In order to make the Mach-Zehnder approximately balanced, these couplers were placed on both arms, even though only one is actually used in the operation of the modulator. In addition, the Mach-Zehnder is designed to be intrinsically unbalanced, with one arm having a length of 1.2 cm, and the other a length of 1.24 cm. In other embodiments, it is possible to use muxes which will couple two different wavelengths into the same waveguide with relatively low loss. In our experience, we found that such devices were less acceptable in the processes we employed.
Device Fabrication
Our devices were fabricated in 120 nm thick silicon-on-insulator layers by using electron beam lithography at 100 kV with a Leica EBPG 5000+ beamwriter. Beam step sizes were set to 5 nm, allowing the use of a 360 micron field.
A direct-written hydrogen silsesquioxane hardmask was used to define the waveguides, and the devices were etched with a chlorine-based recipe in an Oxford PlasmaLab 100 inductively coupled plasma reactive ion etcher. A number of different polymers were coated, in order to determine which chromophore loading would exhibit the best nonlinear behavior. The samples were then briefly dipped in buffered HF, in order to remove the resist from the surface and in order to provide a slight undercut, improving the modal overlap with the cladding.
The devices discussed herein were coated with AJLS102/APC polymer at 30 percent chromophore loading and 14 weight percent concentration, with a solvent of cyclohexanone. The spin rate was 1800 rpm for approximately 1 minute. The samples were then baked in a vacuum oven overnight in order to remove any residual solvent. This third-order nonlinear polymer proved to be robust over several days of testing, and formed clean films over centimeter scale areas. Measurements were performed with a computer-controlled stage system which allowed light to be coupled into the chips using single-mode couplers.
Each device was initially exercised with a swept-wavelength laser and diode detector, in order to determine the optical loss and wavelength response between the gate and drop port, and between the source and drop ports. The modulators were then characterized for their response to modulation in the Terahertz by using two lasers simultaneously as the gate.
The devices were measured using a planar, wafer-scale optical test setup. This setup allows for relatively low-loss single-mode coupling into the devices. Losses are typically on the order of 5 dB, with an uncertainty of around I dB. More importantly, this setup allows for the automated measurement of very large numbers of devices. Initially, the devices were tested passively, with low optical power, in order to characterize their linear response.
Observation of Terahertz Modulation
Generating test signals at Terahertz frequencies electrically is not a straightforward proposition. There are no modulators available that could be used to add such an amplitude modulation to a continuous-wave laser signal. Nevertheless, in order to characterize these devices at speeds in the Terahertz frequency range, we needed a Terahertz test tone.
Two lasers were combined together to form the gate, providing an intensity modulation envelope in time at the difference frequency. This is a purely linear process, associated with the superposition of two beams that are sinusoidal in time and space, where the two beams are at different frequencies, and is quite similar to the well-known Vernier effect. Assuming the lasers have equal power, this gives an envelope function at the difference frequency with the average intensity given by
I(T)=1+cos(Δwt)
This results in an ultrafast change in the index of refraction of the left arm of the interferometer of
For example, two laser signals co-propagating in the same waveguide with wavelengths near 1550 nm whose wavelengths are spaced apart by 0.8 nanometers will provide an intensity modulation in the waveguide at a frequency of about 100 GHz. Given the fact that our input optical couplers are broadband, accepting signals over about 20 nm of bandwidth, this turned out to be a feasible way to generate very high frequency test signals.
The frequency of the gate signal can be controlled by varying the wavelength spacing between the two gate lasers, one of which operates at λ0 and the other of which operates at λ1. The two laser signals are first combined using a 3 dB coupler. The combined signal is fed through a polarization maintaining erbium doped fiber amplifier (EDFA). The source wavelength (designated as λs in
It is believed that the fundamental limitation on such devices is dispersion in the waveguides. As a traveling-wave device, phase matching conditions may be violated at very high speeds. Through careful dispersion engineering of waveguides, it is expected that the observed behavior will be improved. It is of note that there are a number of peaks exhibited which are neither modulation sidebands nor input laser frequencies. These peaks appear to be the result of the source wavelength phase-modulating the gate frequencies, which is to be expected in a four-wave mixing circumstance.
One key feature of the data shown in
E(t)=A cos(w(t−L/c)+Δφ(t))
On the other hand, an unbalanced Mach-Zehnder with phase modulation on a single arm can be written as:
In the time domain, the difference between these two devices would be easily visible. Unfortunately, for our highest speed measurements, equipment to perform this measurement was not available, and we only had frequency domain information. Fortunately, as can be seen from these two equations, and as we have shown in our measurements, a Mach-Zehnder modulator can be distinguished from a pure phase modulator by tuning the signal wavelength and observing the extinction of the central peak, independent of the amplitudes of the side bands.
It is possible to relate the phase shift occurring in the modulator to the extinction ratio between the central peak and the sideband. Taking f as this ratio, the relationship is, to a very good approximation:
Δφ=4√{square root over (f)}
For the modulator shown in
One of the features of the all-optical data is that the optical spectrum analyzer only provides amplitude information, but does not give any phase information. Although we could indirectly see that we had an intensity modulator, from our ability to change the bias point of the unbalanced MZI without changing the magnitude of the modulation sidebands, there was no direct confirmation that what was being generated was pure amplitude modulation. In fact, the peaks generated by our devices are at exactly the same wavelengths as the peaks that would be generated by a pure phase modulator based on the same effects.
In order to confirm our interpretation of the results we obtained from our all-optical modulator, we constructed a second embodiment, making use of more-conventional microwave measurement technology. In general, passive microwave components are characterized by frequency-dependant S-parameters.
These S-parameters specify the phase and amplitude coupling between an input and an output port of a given device at a given frequency. The tool used to measure such S-parameters is called a Vector Network Analyzer (VNA). Such a system works by sending an RF test tone of known amplitude and phase into one port of a device, and measuring the amplitude and phase of the returned signal. By sweeping the frequency of the test tone, it is possible to characterize the spectrum of S-parameters across a very wide bandwidth.
In the second embodiment, of course, our signals are not purely microwave. They are, at least within the device, modulated onto an optical carrier, whose frequency is in excess of 100 THz. Fortunately, there exists an apparatus designed to do radio frequency measurements of optical components, originally developed for testing telecommunication components. The particular apparatus we used is an Agilent 8703B, which is a VNA that has both optical output and input. The optical output is a continuous-wave diode laser which is then fed through an integrated electro-optic modulator. The optical input is a calibrated photodiode with bandwidth in excess of 20 GHz.
In this embodiment, the source wavelength was a single continuous-wave laser boosted in power by an EDFA. For the gate wavelength, we used a second laser, which was modulated by the optical modulator built into the 8703B and then boosted in power by a separate EDFA.
It is important to note that both the source and gate wavelengths are coupled out of the optical modulator at the drain. In order to distinguish between the various laser wavelengths in the all-optical Terahertz measurements described previously, we used a wavelength filter downstream from the modulator device. Output was coupled either through this filter and into the detector port of the 8703B, or directly from the drain to the optical spectrum analyzer.
The conditions of the measurement were controlled so that at the input to the phase modulation region the source power was 10 dBm, and the gate power was 17 dBm. Control measurements were performed with a simple PMMA cladding. When PPMA was used, we observed that there was no modulation detected at the output on the source wavelength. Data was then taken for devices with the nonlinear optical cladding.
The Agilent 8703B is designed to measure the transfer of an intensity modulation from one electrical- or laser-signal port to another such port. Such a system is typically used to measure the performance of modulators and detectors as a function of frequency. It reports the results of the measurement as an optical S parameter, which is defined as follows. Let I (t) be time domain intensity measured by the photodiode. Then, the optical S parameter is defined in dBo as
Here, T is an integration period for the measurement and f is the frequency at which the optical S parameter is being measured. N is a normalization offset selected such that, if the gate laser was simply directed into the photodetector directly (a ‘through’ measurement), the optical S parameter would be 0 dB.
In
In addition, the same measurement was performed, but the output was routed to the optical spectrum analyzer (OSA) instead of the VNA. This allowed the observation of the sidebands on the input signal and those on the output of the device. The modulation of the source wavelength at the output of the device is clearly visible in the sidelobes shown in
Furthermore, the horizontal line superimposed on
The left panel of
These devices are the first low power, planar, continuous-wave, ultrafast intensity modulators implemented in a silicon-compatible materials system to our knowledge. With an improvement of waveguide losses to the 0.1 dB/cm regime, which is expected to be possible with improved lithography and etching, and continued improvement in the χ3 values of the engineered optical polymers, devices similar to these are expected to comprise the elements of ultrafast optical logic. In such logic devices, small gate signals could control large source signals. It is expected that such devices will provide optical transistors operating in the THz. It is believed that these devices will function as both wavelength converters and as logic gates. It is expected that such devices will be used to switch data streams from one carrier wavelength to another. In some embodiments, this will allow for cascaded logic. Furthermore, with a sacrifice in the ultimate speed of the devices, it is possible to utilize resonant enhancement to reduce optical input power levels. Through the process control available in modern silicon fabrication facilities, it is expected that ultrafast integrated photonics in silicon will become a reality.
Theoretical Discussion
Optical Rectification Theory
The general governing equation of nonlinear optics is known to be:
Di=ε0(εyEi+χijk2EjEk+ . . . )
Our EO polymers are designed to exhibit a relatively strong χ2 moment, ranging from 10-100 pm/V. In most χ2 EO polymer systems, the Pockel's effect is used to allow the electric field (due to a DC or RF modulation signal) to modify the index of refraction. In such systems the modulating electric field is typically far in excess of the electric field from the optical signal and the term that produces the material birefringence is the only term of importance in the above equation.
Our waveguides, however, have a very large electric field as most of the radiation is confined to a 0.01 square micron cross section. It can be shown that the electric field is approximately uniform in the transverse direction, with a magnitude of
where P is the optical power in Watts. At large optical fields, the non-Pockels terms involved in the governing nonlinear equation cannot be neglected. For coherent input power, at a given location in the waveguide, the optical field is:
Eoptical(t)=A cos (wt+θ)
The term
will therefore contain not only frequency doubled components, but also a “DC” component. This phenomenon is known as optical rectification. We believe that this DC component provides a likely explanation for the photo-current that we observe. Because we have positioned electrodes (the two sides of the slot waveguide) at precisely the bounds of the induced field, the effect of optical rectification takes a small slice of the optical power and converts it into a virtual voltage source between the two arms. This in turn induces a current that we can measure and is linearly proportional to the input power Eoptical2.
Now let us consider the solution to Maxwell's equation in more detail. Our system can be approximated for this discussion as having two dimensions, with both the optical and DC electric field in the x direction and propagation in the z direction, for instance. Let us imagine that the χ2 is nonzero and small for a tiny region from 0 to w in the x dimension. χ2 is sufficiently small that the electric field due to the optical mode is still uniform. Let us imagine the system has no charge anywhere. The optical electric field can be written as E=Ae(ikz−iwt)+c.c. where c.c. indicates a complex conjugate. Let us further assume that the rectified DC field is of real amplitude C and uniformly directed in the x dimension on (0, w), and 0 elsewhere.
Other than the divergence condition, Maxwell's equations are still satisfied by this system. But at the edge of an interface on the interior, the DC frequency component of Dx, the displacement electric field, is discontinuous. At x0, we have:
Dx−=0
Dx+=ε0(εyC+χ2C2+2χ2|A|2)
We neglect χ2C2 because we expect the amplitude of the rectified field to be far smaller than that of the optical field. Clearly, the boundary condition of zero divergence can only be satisfied if Dx+ is 0. Then,
Thus the direction of the rectified field is reversed compared to the direction of χ2. Note that there is no particular direction associated with the optical field as it is continually oscillating. As we have seen, this rectified DC field would then, if acting as a virtual voltage source, create an effective positive terminal on the positive polling terminal.
Analysis of Data for Optical Rectification
To derive the magnitude of the expected photocurrent, we assume that the χ2 magnitude relating to the Pockels' effect is similar to that for optical rectification. A measurement of χ2 can then be obtained from the direct observation of the electro-optic coefficient by the standard measurements described earlier. The typical measured tuning value of 2 GHz/V yields approximately 50 pm/V.
In the best case, devices with 6 dBm of input power returned approximately 1.4 nA of current. With Qs ranging from 3k to 5k, and assuming approximately 7 dB of insertion loss in the input grating coupler on one of our chips, in the best case as much as 0 dBm might be circulating in a resonator on resonance. This implies a peak electric field due to the optical signal of approximately 3.1×106 V/m. The induced static nonlinear polarization field is then nearly 1000 V/m, which amounts to a voltage drop of 14×10−5 V across a 140 nm gap. If this voltage is assumed to be perfectly maintained, and the load resistance is assumed to be 5 MΩ, then 28 pA would be generated, about a factor of 100 less than is observed in the largest measurement made, but within a factor of 20 of the typical measurement of 352 pA for 6 dBm of input. Significantly, because the generated current is quadratic in E, it is clear that the current will be linearly proportional to the input intensity. This is in accordance with our observations. The best results for optical rectification were obtained with YLD 124/APC polymer, whereas our best Pockels' Effect results were obtained with the dendrimer materials.
Significantly, the sign of the output current matches that which would be predicted by nonlinear optical rectification, as discussed above. Specifically, since positive current emanates from the positive terminal, the rectified E field has a sign reversed from the χ2 and the polling E field. It is well established that the χ2 direction tends to align with the direction of the polling E field. Because of this, the rectified field acting as a voltage source will produce an effective positive terminal at the terminal that had the positive polling voltage.
We do not yet fully understand the current generation mechanism. In particular, it is not clear what provides the mechanism for charge transport across the gap. The APC material in which the nonlinear polymer is hosted is insulating, and though it does exhibit the photoconductivity effect due to visible light, it is unclear whether it can for near-infrared radiation. Photoconductivity due to second harmonic generation may play a role in this effect. It is certainly the case, however, that current flows through this gap; that is the only region in the entire system where an electromotive force exists. Also, photoconductivity alone is not adequate to explain the reversal of the current coming from the detector devices when the poling direction is reversed, nor the conversion of the optical input into directed current in general. The only mechanism to our knowledge that adequately explains this data is optical rectification.
If we assume that it will be possible to achieve a 10-fold improvement in the Q's of the resonators, while still getting more than 10 dB of extinction, then the intensity circulating in such a ring would be about 13 dB up from the intensity of the input wave. By comparison, with a Q of about 1000 and high extinction, the peak circulating intensity is about the same as the intensity in the input waveguide. Therefore, it is reasonable to expect that it will be possible to get at least 10 dB of improvement in the circulating intensity, and thus in the conversion efficiency, by fabricating higher Q rings.
By combining the nano-scale slotted waveguide geometry with electro-optical polymers having high nonlinear constants, we have obtained massive enhancement of the optical field. That has in turn enabled us to exploit nonlinear optical processes that are typically only available in the kW regime in the sub-mW regime. This difference is so considerable that we believe it represents a change in kind for the function of nonlinear optical devices. In addition, it is believed that this hybrid material system provides systems and methods for creating compact devices that exploit other nonlinear phenomena on-chip.
Optical rectification based detectors can have many advantages over currently available technology. In particular, such detectors are expected to function at a higher intrinsic rate than the typical photodiode in use, as the optical rectification process occurs at the optical frequency itself, on the order of 100 THz in WDM systems. The absence of an external bias, and the generation of a voltage rather than a change in current flow, both provide certain advantages in electronic operation. We also believe that a device based on nonlinear optical rectification will not suffer from the limitation of a dark current. This in turn can provide WDM systems that will function with lower optical power, providing numerous benefits. Similarly, our demonstration of enhanced modulation using these waveguide geometries provides useful components for future communications systems.
Nonlinear Phase Modulation Mechanism
To best understand the limitations of the nonlinear phase modulation mechanism, one must consider both the waveguide loss and dispersion of our waveguides. Together, these two parameters establish a limitation on the amount of phase shift that can be obtained, or the rate of change of the phase shift that can occur. Our waveguide loss has been extensively calibrated through other test structures, and is known to be −7 dB/cm. The effective index of these waveguides near 1550 nm is well modeled by
As noted, this expression corresponds to a group velocity dispersion of about −4,000 ps nm−1 km−1, a value typically encountered for such waveguides. Note that, unlike a device dependent on four-wave-mixing conversion, in our device the momentum mismatch noted in equation (2) does not affect the phase shift experienced by the source wavelength. Thus, the only limitation on the bandwidth of the device is the tendency of a pulse to scatter based on dispersion, and the tendency of pulses on different wavelengths to travel at different velocities. It is straightforward, using Fourier analysis, to analyze the time-domain behavior of such pulses in the waveguide. The pulses used in the simulation shown in
The limitations on the nonlinear response can be best characterized by introducing the concept of effective length. The effective nonlinear length of a given waveguide is the equivalent length of a lossless, dispersion free waveguide. For a modulation frequency of ωm, the effective length can be written as
It can be shown that for our waveguides, an asymptotic value for Leff of 0.5 cm is approached for L>0.7 cm. It turns out that ωm must approach 10 THz before any significant change in effective length occurs. This indicates that our device has an operation speed than can exceed the measured 2.6 THz, and approach 10 THz, and that the current limitation on device performance is the waveguide loss.
Derivation of Equation (3)
This is a justification for the expression relating the fractional wavelength conversion to the phase shift.
We begin with the standard equation of nonlinear optics, modified slightly for the dispersive nature of the waveguide. Here χ3 is in (m/V)2
Here, Ex is the collection of all modes:
Ex=Σai exp(ikiz−iωit)+ai*exp(−ikiz+iωit)
Note that Ex is always real valued, as is required for use with the equations of nonlinear optics. Using the standard slowly-varying amplitude approximation, all of the linear terms are assumed to vanish for propagating modes, except for:
Proceeding further depends on the particular nonlinear process under study. In the case of energy conversion by four wave mixing, taking ω2 as the generated frequency, and ω1, ω0 the pump beams, when ω2=2 ωi−ω0, we have:
The final term, however, is simply a measure of the so-called momentum mismatch between the 3 waves. As we have shown elsewhere, for our waveguide at length scales around 1 cm, and for frequencies less than 10 THz, this mismatched value is not significant, and the integral is nearly equal to L.
Under this assumption (|a1|=|a0|), and under the condition that the two laser beams are nearly equal in power, the experimentally measured f fraction is then:
In the case where one beam (ω2) is intensity modulated by another beam (ω1), the process can be written as:
It is seen that, as seen by beam 2, propagation amounts to multiplication by a complex value with modulus 1. The absolute value of the argument of this value is readily seen to be twice f. That is,
Finally it will be noted that the expression used hereinabove was twice this 4√{square root over (f)}. This is because the measured f is artificially suppressed by a factor of ¼ due to passing through the final 3 dB coupler. The produced sideband is reduced 3 dB, while the signal laser is magnified by 3 dB, since we biased the Mach-Zehnder at maximum.
We conclude by stressing advantageous economic aspects of our invention in various embodiments. Because our devices can be fabricated in planar electronics grade silicon-on-insulator, using processes compatible with advanced CMOS processing, it is expected that devices embodying these principles will be less expensive to fabricate.
While the present invention has been particularly shown and described with reference to the structure and methods disclosed herein and as illustrated in the drawings, it is not confined to the details set forth and this invention is intended to cover any modifications and changes as may come within the scope and spirit of the following claims.
This application claims priority to and the benefit of co-pending U.S. provisional patent application Ser. No. 60/708,109, filed Aug. 12, 2005, and priority to and the benefit of co-pending U.S. provisional patent application Ser. No. 60/713,132, filed Aug. 31, 2005, each of which applications is incorporated herein by reference in its entirety.
| Number | Date | Country | |
|---|---|---|---|
| 60708109 | Aug 2005 | US | |
| 60713132 | Aug 2005 | US |
