Patent Application
20200379315
None.
The present invention relates generally to optical phased arrays. More specifically, it relates to photonic crystal optical phased arrays.
Optical beam steering techniques are important in projection systems, 3D printing, various sensing devices, and light detection and ranging (LiDAR). Applications of LiDAR include autonomous vehicles, aerial mapping, atmospheric measurements, and security systems. Currently, most optical beam steering devices use rotating lasers or mirrors to perform scanning. In addition to being bulky and expensive, these are also sensitive to vibrations and alignment error. The desire for low-cost, compact beam optical steering systems has motivated research of integrated devices based on microelectro-mechanical mirrors, acoustic-optic effects, or phase-change materials to accomplish steering. Yet, these devices typically lack in speed, scan-range or beam divergence.
Another approach to optical beam steering is to use an optical phased array (OPA). An OPA uses an array of antennas, each emitting light with a phase corresponding to that of the wavefront of the desired angled beam. By actively changing the phase relations between the antennas, the wavefront can be altered, allowing the beam to be steered. A major challenge in the design of an OPA is that the antennas should be spaced very close to each other, i.e., preferably less than 1 μm for optical wavelengths, because larger spacing results in side lobes in which energy is lost and which are typically detrimental for the intended application. This creates a challenge of fabricating on a very small scale the antennas, the optical waveguide that feeds the antennas, and the phase modulators that control the phases of light emitted from the antennas.
To allow a higher density of antenna elements in the array, a conventional 2D optical phased array does not use a separate phase modulator for each individual antenna, but instead uses a phase shifter for each of line of antennas. These phase shifters, however, only allow the array to be actively steered along one axis (in the transverse direction). Steering along the second axis (in the longitudinal direction) is performed by controlling the wavelength input to each line of antennas. By combining these techniques, a dense optical phased array can steer a beam in two directions.
Steering with wavelength, however, currently has limitations when using strip waveguides on the conventional silicon-on-insulator platform. To obtain a 14° range of steering angles would require a 100 nm bandwidth, which is beyond the capabilities of compact, low-cost integrated lasers. Moreover, a 14° degree range is still small considering that a LiDAR system for autonomous vehicles typically wants to address a 360×30 degree field of vision.
In summary, current OPAs fail to meet the specifications needed to be applicable in low-cost systems. The steering range of these OPAs is often too small or requires a wavelength tuning range beyond what is achievable with a compact integrated laser.
In one aspect, the present technology provides an optical phased array based on dispersion engineered waveguides. By using dispersion engineered waveguides, the phased array can be more sensitive to the wavelength, that is, by changing the wavelength, the beam emitted from the optical phased array can be steered over a larger angle compared to phased arrays based on typical strip waveguides.
Realizing such a phased array involves overcoming challenges in designing dispersion engineered waveguides with near linear dispersion relation, designing a mode converter that couples light from the input waveguide mode to the dispersion engineered waveguide mode, designing antennas integrated in the waveguide to radiate out light from the waveguide, all while ensuring that the antenna and waveguide are spaced close to each other (e.g. 2 μm). These are all non-trivial optical design problems.
The dispersion engineered phased array can scan a large angle with a smaller wavelength range compared to phased arrays based on strip waveguides. The larger angle range is an improvement compared to existing optical phased arrays. Moreover, the smaller wavelength range lowers the requirements on the laser.
The dispersion engineered phased array can be used as a beam steering device for compact LiDAR or ranging systems, line-of-sign communication systems and projection systems which rely on a scanning beam. As the beam deflects light with a different wavelength under different angles it can also be used as a compact spectrometer.
In one embodiment a 2D phased array comprises a splitter that divides the input light from a wavelength tunable laser to the modulators; modulators that tune the phase of the individual waveguides for transverse steering; and dispersion engineered phased array that can sweep the longitudinal direction by changing the wavelength.
In one aspect, the invention provides an optical phased array device comprising a photonic crystal having a dispersion engineered slow light waveguide region; a mode coupler region capable of optically coupling an input waveguide to the dispersion engineered slow light waveguide region; and optical antenna regions integrated within the dispersion engineered slow light waveguide region; wherein the dispersion engineered slow light waveguide region has a substantially linear dispersion relation within a predetermined operational bandwidth of the optical phased array device; wherein the optical antenna regions are formed by an alteration of a periodic structure of the photonic crystal; wherein the optical antenna regions are capable of radiating light out from the dispersion engineered slow light waveguide region. The optical phased array device may also include multiple such photonic crystals stacked next to one another. The multiple photonic crystals may be regions of a single monolithic photonic crystal. The optical phased array may also include an optical splitter having multiple outputs and multiple phase modulators connected to the multiple outputs and to the multiple photonic crystals.
The dispersion engineered slow light waveguide region preferably has an average group index substantially larger than 7 within a predetermined operational bandwidth. For example, the average group index may be substantially equal to 25. The dispersion engineered slow light waveguide region preferably has a dispersion relation that deviates less than 10% or 20% from linear within a predetermined operational bandwidth of the optical phased array device.
The optical phased array device may also include an input waveguide region. The optical phased array device may also include a layer above the photonic crystal waveguide capable of facilitating radiation of light out from only one side of the dispersion engineered slow light waveguide region. The optical antenna regions in some implementations have different radiation scattering strengths.
In one embodiment of the invention, an optical phased array is provided based on a slow light waveguide capable of steering a large angle range with a limited bandwidth. The dispersion relation of the slow light waveguide is optimized to be near-linear with an average group index of 25 in the intended operating bandwidth. Additionally, a mode converter and antennas efficiently couple light in to and out of the slow light waveguide. The resulting OPA is capable of steering the optical beam 19° by scanning from 1543 nm to 1566 nm.
The term “slow light waveguide” is a well-known term in the art that refers to waveguides in which light propagates at a very low group velocity due to its interaction with the medium in which the propagation takes place. Simple slow light waveguides suffer from strong higher-order dispersion and limited bandwidth, limiting their use dramatically. In contrast, embodiments of the present invention provide a dispersion engineered optical phased array including a slow light waveguide, the slow light couplers, and antennas integrated into the waveguide. The performance of the OPA is significantly superior to prior devices.
To fully appreciate the present invention and the design challenges it overcomes, the following description introduces principles of optical phased arrays and inverse design techniques.
Optical Phased Arrays
An OPA radiates a beam in a given direction when the phase relation between its antennas matches the phase profile of that beam. This is illustrated in the case of the 1D phased array shown in
For a 2D OPA, as shown in
The phase relation between the antennas along the longitudinal direction of the waveguide can be understood by the grating equation:
k
airx
=β+m·k
ant. (1)
where kairx is the projection of the wavefront wavevector on the x-axis, i.e. the waveguide direction, β is the wavenumber of the waveguide mode, m is an integer, and kant=2π/dant is the wavenumber of the periodically placed antenna with period dant. As the wavelength changes, so do β and kair.
The dispersion relation of the waveguide determines how sensitive the OPA's radiation angle is to a change in wavelength. If the dispersion curve has a low slope, then the waveguide's wavenumber will change significantly and, as a result, so will the radiation angle. This is the case for a slow light waveguide, which is characterized by a low group velocity, vg=∂ω/∂k or a large group index ng=c/vg. An example of a dispersion relation for a slow light photonic crystal waveguide is shown in
Designing a full 2D OPA which can steer longitudinally by changing the wavelength may be performed by optimizing several photonic components. First of all, we require slow light waveguide that operates in a specific bandwidth, e.g., the tuning range of the laser used for the OPA setup, and provides a substantial change in the wavenumber within that bandwidth. Additionally, it is desirable to have a linear or near-linear dispersion relation in this bandwidth. (Here, “near linear” is defined as variation from linearity by less than 10% over the operational wavelength range.) Secondly, a mode converter is used to couple a strip waveguide to the photonic crystal (PhC) waveguide. The slow light mode can have little overlap with that of a strip waveguide, and the different dispersion relation can lead to strong impedance mismatch, both of which lower the coupling efficiency between the waveguides. A mode converter is thus used to facilitate this transition. Finally, the device needs to gradually radiate the light from the photonic waveguide via an antenna. Simple partially etched notches, used in OPA's relying on strip waveguides, are not an option here. We thus design an antenna structure in our slow light waveguide.
Before discussing the slow light OPA elements, we shortly review photonic inverse design problems. Inverse design problems for optical devices typically are of the following form:
minimize fobj(p,E1, . . . ,En)
with respect to p,E1, . . . ,En
subject to the constraints
h
EMi(p,Ei)=0,i=1, . . . ,n, and hfab(p)=0. (2)
The objective function, fobj, defines a figure of merit, which needs to be minimized as a function of a parameterization, p, which describes the device structure, and the optical fields, Ei, for n different modes indexed by i. The optical fields are linked to the structure by constraining the field solutions to Maxwell's equations, as represented by hEM. Furthermore, the structure can be subjected to additional constraints, most commonly a fabrication constraint, hfab to ensure the design can be fabricated.
Slow Light Waveguide
The aim of the slow light waveguide design is to obtain periodic structure with a nearly-flat, linear dispersion relation, within a certain bandwidth. Different figures-of-merit can be devised for this. The most direct way is to minimize the absolute difference between the group velocity of the structure and the desired group velocity for in a series of wavevectors, ki. While effective, this method is mocomputationally heavy and as such, not ideal for a 3D structure. Alternatively, for a series of wavevectors we can optimize the difference between the frequencies. In this case the objective function becomes:
f
obj(p)=F({(ωi(p)−ωi+1(p)−Δωi)2|i=1, . . . ,n−1}), (3)
where ωi is the angular frequency associated with ki, and Δωi is the target frequency difference. Considering we want to reach a certain group index, ng, the Δωi will be (ki−1−ki)/ng. F can be the sum of the set, the maximum or the softmax.
The underlying physics constraint, hEM, for this optimization problem describes the dispersion relation, which results from the eigenvalue problem:
h
EMi=∇k×(1/μ)∇kEi−ωiε(p)Ei=0,i=1, . . . ,n, (4)
where μ is the magnetic permeability, which we take to be one here, and ε(p) is the permittivity of the structure, parameterized by p. ∇k× is the curl-operator taking the Bloch boundary condition for wavevector, k, into account.
Photonic Crystal Waveguide Coupler
For the couplers, we optimize the transmission in the waveguide mode. Considering we normalized the input power of the source, the objective then becomes:
f
obj(E1, . . . ,En)=Σi=1, . . . ,n1−(Ci·Ei)2, (5)
where Ci represents a vector which will integrate the power transmitted in waveguide mode for ωi.
Here the fields, Ei, result from the sources, Ji, corresponding either the input waveguide mode or a vertically incident Gaussian beam, in case of waveguide coupler or grating coupler, respectively. As such, the EM constraint becomes:
h
EM=∇×(1/μ)∇×Ei−ωi2ε(p)Ei+jωJi=0 (6)
Slow Light Waveguide
The slow light waveguide is optimized based on the optimization problem in eq. 2 with eq. 3 as objective functions. We rely on simple sum function for F.
The optimization is done using eight modes to form the objective function in eq. 3. The trajectory of the angular frequencies during the optimization can be seen in
Slow Light Waveguide Coupler
To route light on-chip to the slow light OPA, one would most commonly use a simple strip waveguide. The modal field profile of the PhC waveguide, which can be seen in
Slow Light Waveguide Antenna
To use the slow light waveguide as an OPA, we integrate antennas in the waveguides, as depicted in
Light is emitted in the slow light mode from the right side of the design area 400. On the left, we evaluate the transmitted light in the slow light mode by relying on the mode overlap objective function, eq. 5. In addition, we also evaluate light emitted in the far field. For this, we project the local fields in a slab above the design region onto a sphere. An extra objective function, fffobj, is therefore added to the previous overlap objective function. The form of this objective is as follows:
f
ffobj(E1, . . . ,En)=Σi=1, . . . ,nTff−∫Ω(½)|Fff(Ei)|2, (7)
nowhere Tff is the target transmitted light in the far field and Fff is the near-to-far-field transformation. Ω is the region of interest in the far field, i.e., the window on the sphere over with we integrate the projected fields. The window used in the optimization is the shaded region 410 shown in
The optimization starts with an initial parametrization that matches the PhC waveguide structure with some small random alterations. After optimization, we obtain for the antenna region 420 the structure shown in
Slow Light Optical Phased Array
The mode coupler region 500, waveguide region 502, and antenna regions 504 through 506 are combined into a 1D OPA line, as shown in
A 2D phased array can be achieved by stacking the waveguides of
Alternative Design:
Combined Antenna and Dispersion Optimization
The design of the antenna can also be combined with the dispersion design. Designing the antenna as a stand-alone defect in the dispersion engineered waveguide has the disadvantage of the antennas interacting when closely spaced. If we want to have very close-packed antennas we can also take up the radiation behavior in a dispersion optimization.
We can do this by optimizing a mode that lies in the light cone 700, as illustrated in
Using the complex frequencies, we can now optimize the combined dispersion/antenna structure by the following problem.
Minimize with respect to p
Σk=1, . . . ,n(Re{ωtk}−Re{ωk})2+Σk=1, . . . ,n(Im{ωtk}−Im{ωk})2
subject to the constraint
∇k×(1/μ)∇k×Ek−ωk2ε(p)Ek=0,k=k0, . . . ,kn. (8)
The first term in the objective function above accounts for the mode's dispersion by trying to match the real part of the frequency with a target frequency, i.e., a target dispersion relation. The second term in the objective function accounts for the loss. We try to match the imaginary part of the frequency to a target value, which corresponds to the desired radiation loss we want in our OPA.
Practically we can start from a slow light waveguide design as the one obtained above. We now reoptimize this design by altering two or more periods of this design. By changing the periodicity of the waveguide, the bands are pushed in the light cone and couple to free space modes. Starting from the larger period design we now reoptimize using the optimization problem in eq. 8.
An example of an optimization process initialization can be seen
Fabrication
The resulting OPA designs can be fabricated using semiconductor fabrication processes. The designs can be patterned in resist using a lithographic process, such as ebeam or optical lithography (for example ArF immersion, or extreme UV lithography), and subsequently etched in the material systems for which the OPA was designed. These fabrication methods also allow for multi-layer designs. One part of the design, e.g., the slow light waveguide, could be made in one layer, while after deposition processes, a different part, e.g., the antennas, could be fabricated in a different layer. Silicon-on-insulator is the most evident material to work in, yet the proposed method does not rely specifically on this material system. The design process could also be used for other materials that are compatible with semiconductor fabrication processes such as, for example, lithium niobate, III/V-materials, silicon carbide or silicon nitride.
The OPAs of the present invention provide a large longitudinal steering range controlled by a change in wavelength. The OPA design is based on slow light waveguides in which we integrate antennas. All the design challenges in making this OPA were addressed using inverse design methods. The slow light waveguide was optimized to have a near-linear dispersion over a broad wavenumber range. To couple light to the slow light waveguide, we design an efficient mode converter that facilitates the coupling between a strip waveguide and the slow light waveguide in the operation bandwidth. To radiate light from the slow light waveguide, we designed compact antenna structures into the slow light waveguide. Combined, these components form an OPA which obtains performance not possible previously. In one implementation, the OPA can steer the optical beam 19° by scanning from 1543 nm to 1566 nm. Other embodiments may have even larger scanning range by pushing this OPA design further using the inverse design methods presented here. Relying on these design methods, the performance of compact integrated OPAs can now achieve and outperform current optical beam steering systems. This enables low-cost optical scanning functionality in a wide variety of application.
This application claims priority from U.S. Provisional Patent Application 62/853,526 filed May 28, 2019, which is incorporated herein by reference.
| Number | Date | Country | |
|---|---|---|---|
| 62853526 | May 2019 | US |
