INTERFEROMETER FILTERS WITH COMPENSATION STRUCTURE

Abstract

A photonic switch includes a first waveguide including a first region extending between a first coupler section and a second coupler section and a second region extending between the second coupler section and a third coupler section. The photonic switch also includes a second waveguide including a first portion extending between the first coupler section and the second coupler section, the first portion including at least two first compensation sections each having a different waveguide width, and a second portion extending between the second coupler section and the third coupler section, the second portion including at least two second compensation sections each having a different waveguide width. The photonic switch further includes at least one variable phase-shifter disposed in at least one of the first waveguide or the second waveguide.

Claims

1. A photonic switch comprising: a first waveguide including: a first region extending between a first coupler section and a second coupler section; and a second region extending between the second coupler section and a third coupler section; and a second waveguide including: a first portion extending between the first coupler section and the second coupler section, the first portion including at least two first compensation sections each having a different waveguide width; and a second portion extending between the second coupler section and the third coupler section, the second portion including at least two second compensation sections each having a different waveguide width; and at least one variable phase-shifter disposed in at least one of the first waveguide or the second waveguide.

2. The photonic switch of claim 1 wherein: the waveguide width of the two first compensation sections increases sequentially; and the waveguide width of the two second compensation sections increases sequentially.

3. The photonic switch of claim 1 wherein the at least one variable phase-shifter is at least one thermo-optical switch.

4. The photonic switch of claim 1 wherein the at least one variable phase-shifter is characterized by a refractive index that changes as a temperature of at least one of the first waveguide or the second waveguide rises.

5. The photonic switch of claim 1 wherein the at least one variable phase-shifter is disposed in the first waveguide.

6. The photonic switch of claim 5 wherein the at least one variable phase-shifter is a first variable phase-shifter disposed in the first region of the first waveguide.

7. The photonic switch of claim 5 wherein the at least one variable phase-shifter comprises: a first variable phase-shifter disposed in the first region of the first waveguide; and a second variable phase-shifter disposed on the second region of the first waveguide.

8. The photonic switch of claim 1 wherein the at least one variable phase-shifter is disposed in the second waveguide.

9. The photonic switch of claim 1 wherein the first coupler section, the second coupler section, and the third coupler section and the first waveguide and the second waveguide comprise a Mach-Zehnder interferometer (MZI) filter.

10. The photonic switch of claim 1 wherein the first portion includes at least three compensation sections that sequentially increase in width.

11. The photonic switch of claim 1 wherein the second portion includes at least three compensation sections that sequentially increase in width.

12. The photonic switch of claim 1 wherein the at least two first compensation sections and the at least two second compensation sections include respective taper portions.

13. The photonic switch of claim 1 wherein at least one of the first portion or the second portion have a number of waveguide widths that are greater than a predetermined number of perturbations.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0020] FIG. 1A illustrates a simplified plan view of an example Mach-Zehnder interferometer filter including a passive compensation structure, according to embodiments of the disclosure;

[0021] FIG. 1B illustrates a simplified plan view of an example Mach-Zehnder interferometer filter including two passive compensation structures, according to embodiments of the disclosure;

[0022] FIG. 1C illustrates a simplified plan view of an example Mach-Zehnder interferometer filter including a passive compensation structure, according to embodiments of the disclosure;

[0023] FIG. 2 illustrates a single stage of a three-waveguide cascaded third order MZI based filter, according to embodiments of the disclosure;

[0024] FIG. 3 illustrates an incoherently cascaded third-order MZI filter having four stages, according to embodiments of the disclosure;

[0025] FIG. 4 illustrates effective index parameters as a function of waveguide width and height for a silicon-on-insulator waveguide, according to embodiments of the disclosure;

[0026] FIGS. 5A and 5B illustrate a plotted derivative, according to embodiments of the disclosure;

[0027] FIG. 6 illustrates standard deviations for waveguides and couplers, according to embodiments of the disclosure;

[0028] FIG. 7 illustrates the statistical behavior of four-stage cascaded third-order MZI filter without mitigation mechanisms, according to embodiments of the disclosure;

[0029] FIG. 8 illustrates designs to minimize susceptibility to fabrication errors, according to embodiments of the disclosure;

[0030] FIG. 9 illustrates the statistical distribution of cascaded third-order MZI's with asymmetric arm widths, according to embodiments of the disclosure;

[0031] FIG. 10 illustrates the statistical distribution of cascaded third-order MZI's in the absence of coupler variations with respect to fabrication uncertainties, according to embodiments of the disclosure;

[0032] FIG. 11 illustrates the statistical distribution of MZI properties for three waveguide widths, according to embodiments of the disclosure;

[0033] FIG. 12 illustrates the statistical distribution of an MZI filter with four waveguide widths, according to embodiments of the disclosure;

[0034] FIG. 13 illustrates fabrication tolerance achieved using asymmetric widths as well as heights, according to embodiments of the disclosure;

[0035] FIGS. 14A-14D illustrate unconventional cross-sections that are compatible with CMOS-foundry processes, according to embodiments of the disclosure;

[0036] FIG. 15 illustrates the performance of a filter, according to embodiments of the disclosure;

[0037] FIG. 16 illustrates performance of a filter, according to embodiments of the disclosure;

[0038] FIG. 17 illustrates performance of the filter, according to embodiments of the disclosure;

[0039] FIG. 18 illustrates asymmetric widths where width and height variations are independent, according to embodiments of the disclosure;

[0040] FIG. 19 illustrates a filter having asymmetric widths where width and height variations are independent and each stage is correlated, according to embodiments of the disclosure;

[0041] FIG. 20 illustrates an embodiment where width and height variations of every stage are correlated, according to embodiments of the disclosure;

[0042] FIG. 21 illustrates an embodiment where width and height variations are independent but are correlated for all stages, according to embodiments of the disclosure;

[0043] FIG. 22 illustrates an embodiment where width and height variations are independent but are correlated for all stages, according to embodiments of the disclosure;

[0044] FIG. 23 illustrates the performance of a filter, according to embodiments of the disclosure;

[0045] FIG. 24 illustrates yield percentage, according to embodiments of the disclosure;

[0046] FIG. 25 illustrates a simplified plan view of an example Mach-Zehnder interferometer switch including a passive compensation structure, according to embodiments of the disclosure; and

[0047] FIG. 26 illustrates a simplified plan view of an example Mach-Zehnder interferometer switch including a passive compensation structure, according to embodiments of the disclosure.

DETAILED DESCRIPTION

[0048] Some embodiments of the present disclosure relate to a passive compensation structure for a Mach-Zehnder interferometer (MZI) filter that improves the filter's ability to accommodate changes in manufacturing tolerances and/or other perturbations. While the present disclosure can be useful for a wide variety of configurations, some embodiments of the disclosure are particularly useful for cascaded MZI filters that are fabricated using silicon-based structures, as described in more detail below.

[0049] For example, in some embodiments, an MZI filter includes a pair of waveguides that extend between a first and a second coupler section. The first waveguide has a first continuous width along its length. The second waveguide includes a tolerance compensation portion positioned between the first and the second coupler sections. The tolerance compensation portion includes multiple waveguide sections, each having a different width, as explained in more detail below. The compensation portion can reduce a shift in frequency response of the MZI filter that can be caused by various perturbations, including variations in manufacturing widths of the waveguides, manufacturing variations in thicknesses of the waveguides and variations in temperature. In further embodiments the compensation structure can be designed to reduce a shift in frequency response of the MZI filter that can be caused by myriad perturbations while meeting a resonance requirement, as described in more detail below.

[0050] In one example the tolerance compensation portion includes waveguide sections having three different widths, however other embodiments may have a lesser number or a greater number of widths. In this example, the tolerance compensation portion includes a first compensation portion having a second width, a second compensation portion having a third width and a third compensation portion having a fourth width, wherein the fourth width is greater than the third width and the third width is greater than the second width.

[0051] In another example the first waveguide can also have a compensation portion including multiple waveguide sections, each having different waveguide widths. In further examples, the compensation structure can be designed to compensate for a particular number of system perturbations by having a quantity of waveguide widths that is greater than the number of perturbations. In one embodiment the resonance requirement and a number of system perturbations can be accommodated by designing the compensation structure to have at least one more waveguide width than the number of system perturbations. For example in one embodiment a MZI filter can be designed to have insensitivity to width variations and to have a resonance at 1.55 um by having a compensation structure with three different widths, while a compensation structure having two different widths may be used to compensate for width variations only. In further examples, the degree to which the compensation structure can compensate for a particular set of perturbations can be improved by increasing the total number of different waveguide widths, as also described below.

[0052] In some embodiments, lengths and widths of the compensation structure can be determined using one or more compensation equations. More specifically, the first and the second waveguides of the MZI filter simultaneously satisfy:

[00002]${m\lambda }_0=L_1\left(n_1\left({\lambda }_0\right)-{\Sigma }_i{n}_i\left({\lambda }_0\right){\kappa }_i\right)v_{\mathrm{FSR}}=\frac{c}{L_1\left(n_{g1}-{\Sigma }_i{n}_{\mathrm{gi}}{\kappa }_i\right)}\frac{\partial n_1}{\partial X_j}={\Sigma }_i{\kappa }_i\frac{\partial n_i}{\partial X_j}\frac{{\partial }^2{n}_1}{\partial X_j\partial \omega }={\Sigma }_i{\kappa }_i\frac{{\partial }^2{n}_i}{\partial X_j\partial \omega }\frac{{\partial }^2{n}_1}{\partial X_j^2}={\Sigma }_i{\kappa }_i\frac{{\partial }^2{n}_i}{\partial X_j^2}$

wherein:
m=an integral multiple;
λ.sub.0=wavelength of light in first and second arms;
L.sub.1=reference length of first arm;
L.sub.i=length of i.sup.th portion of second arm;
κ.sub.i=L.sub.i/L.sub.1;
ν.sub.FSR=free spectral range;
c=speed of light;
X.sub.1=waveguide width; and
X.sub.2=waveguide thickness.

[0053] In order to better appreciate the features and aspects of the present disclosure, further context for the disclosure is provided in the following section by discussing one particular implementation of an MZI filter that includes a passive compensation structure, according to embodiments of the disclosure. These embodiments are for explanatory purposes only and other embodiments may be employed in other MZI-based filter devices. In some instances, embodiments of the disclosure are particularly well suited for use with quantum computing circuits because of the intractability of using thermo-optic tuning for these applications.

[0054] FIG. 1A illustrates a simplified plan view of an example Mach-Zehnder interferometer filter 100 including a passive compensation structure 102, according to an embodiment of the disclosure. As shown in FIG. 1, MZI filter 100 includes a first waveguide 104 having a first length 106 and extending from a first coupler section 108 to a second coupler section 110. First waveguide 104 has a constant first width 112 along first length 106. A second waveguide 114 includes a compensation portion 116 positioned between first coupler section 108 and second coupler section 110. Compensation portion 116 includes a first compensation section 118 having a second width 120, a second compensation section 122 having a third width 124 and a third compensation section 126 having a fourth width 128. In some embodiments, fourth width 128 is greater than third width 124 and the third width is greater than second width 120. In some embodiments, the width and length of each compensation portion can be determined using one or more compensation equations, as described in more detail below.

[0055] In some embodiments, compensation portion 116 is symmetric along second waveguide 114 and further includes a fourth compensation section 130 having third width 124 and a fifth compensation section 132 having second width 120. In further embodiments, compensation structure 102 may also include a compensation portion positioned within first waveguide 104, as described in more detail below.

[0056] In various embodiments, one or more taper portions can be positioned in-between each compensation section to transition between different waveguide widths. More specifically, in some embodiments, a first taper portion 134 is positioned between first coupler section 108 and first compensation section 118 and transitions to second width 120. A second taper portion 136 can be positioned between first compensation section 118 and second compensation section 122 and transitions from second width 120 to third width 124. A third taper portion 138 can be positioned between second compensation section 122 and third compensation section 126 and transitions from third width 124 to fourth width 128. Similarly, a fourth taper portion 140 can be positioned between third compensation section 126 and fourth compensation section 130 and transitions from fourth width 128 to third width 124. A fifth taper portion 142 can be positioned between fourth compensation section 130 and fifth compensation section 132 and transitions between third width 124 and second width 120. A sixth taper portion 144 can be positioned between fifth compensation section 132 and second coupler section 110 and can transition from second waveguide width 120. In some embodiments, first waveguide 104 can also include one or more taper portions to transition widths between first coupler section 108 to first waveguide 104 and from the first waveguide to second coupler section 110.

[0057] In some embodiments, each compensation section 118, 122, 126, 130, 132 of compensation portion 116 may have a substantially constant width. More specifically, in some embodiments, first compensation section 118 has a constant second width 120, second compensation section 122 has a constant third width 124, third compensation section 126 has a constant fourth width 128, fourth compensation section 130 has a constant third width 124 and fifth compensation section 132 has a constant second width 120.

[0058] In some embodiments, each compensation section can have a particular length, as determined by one or more compensation equations, described in more detail below. First compensation section 118 can have a second length 146, second compensation section 122 can have a third length 148, third compensation section 126 can have a fourth length 150, fourth compensation section 130 can have a fifth length 152 and fifth compensation section 132 can have a sixth length 154.

[0059] In some embodiments, first length 106 of first waveguide 104, length of each compensation section 118, 122, 126, 130 and 132, first width 112 of first waveguide 104 and widths 120, 124, 128, 124, 120 of each respective compensation section 118, 122, 126, 130 and 132 of compensation structure 102 can be determined using one or more compensation equations. More specifically, the first and the second waveguides of MZI filter 100 simultaneously satisfy:

[00003]${m\lambda }_0=L_1\left(n_1\left({\lambda }_0\right)-{\Sigma }_i{n}_i\left({\lambda }_0\right){\kappa }_i\right)v_{\mathrm{FSR}}=\frac{c}{L_1\left(n_{g1}-{\Sigma }_i{n}_{\mathrm{gi}}{\kappa }_i\right)}\frac{\partial n_1}{\partial X_j}={\Sigma }_i{\kappa }_i\frac{\partial n_i}{\partial X_j}\frac{{\partial }^2{n}_1}{\partial X_j\partial \omega }={\Sigma }_i{\kappa }_i\frac{{\partial }^2{n}_i}{\partial X_j\partial \omega }\frac{{\partial }^2{n}_1}{\partial X_j^2}={\Sigma }_i{\kappa }_i\frac{{\partial }^2{n}_i}{\partial X_j^2}$

wherein:
m=an integral multiple;
λ.sub.0=wavelength of light in first and second arms;
L.sub.1=reference length of first arm;
λ.sub.0=central wavelength of light in first and second arms;
L.sub.i=length of i.sup.th portion of second arm;
κ.sub.i=L.sub.i/L.sub.1;
ν.sub.FSR=free spectral range;
c=speed of light;
X.sub.1=waveguide width; and
X.sub.2=waveguide thickness.

[0060] For example, in one embodiment, compensation equations can be used to define a compensation structure for a pump-rejection filter for a quantum computer having the following parameters:

(i) 120 dB of pump rejection at wavelength λ.sub.0=1.55 μm;
(ii) 25 mdB of signal loss; and
(iii) A free-spectral range (FSR) of 2.4 THz.

[0061] In other embodiments other suitable parameters can be defined for an MZI filter, as appreciated by one of skill in the art.

[0062] FIG. 1B illustrates a simplified plan view of an example MZI filter 156 including a passive compensation structure, according to an embodiment of the disclosure. As shown in FIG. 1B, MZI filter 156 is similar to MZI filter 100 illustrated in FIG. 1A. However, in this embodiment, MZI filter 156 includes a compensation portion positioned within each waveguide arm. More specifically, similar to MZI filter 100, MZI filter 156 includes compensation portion 116 positioned within second waveguide 114, however, MZI filter 156 also includes a second compensation portion 158 positioned within first waveguide 160, as described in more detail below. As appreciated by one of skill in the art with the benefit of this disclosure any combination of compensation portions can be employed in an MZI filter and the compensation portions do not need to be the same, or even have similar widths and/or lengths. As described in more detail below, each compensation portion can be uniquely designed according to the compensation equations.

[0063] As shown in FIG. 1B, first waveguide 160 includes second compensation portion 158 that includes a plurality of compensation sections, each having a width and a length as defined by a set of compensation equations, described in more detail herein. Second compensation portion 158 is positioned between first coupler section 108 and second coupler section 110. Second compensation portion 158 includes a sixth compensation section 164 having fifth width 174 and seventh length 176, a seventh compensation section 166 having sixth width 178 and a eighth length 180, and an eighth compensation section 168 having fifth width 174 and seventh length 176. As described above with regard to FIG. 1A, one or more taper portions can be positioned between waveguide sections of different widths to transition from one width to another width.

[0064] FIG. 1C illustrates a simplified model of an MZI filter 172 illustrating geometrical parameters for a set of compensation equations. As shown in FIG. 1C, an MZI filter 172 is shown having two parallel waveguides, each having a particular set of geometric parameters. In general, the phase difference between the two waveguide arms is given by Equation (1).

ϕ(ω)=k.sub.1(ω)L.sub.1−Σ.sub.i=2.sup.n+2k.sub.i(ω)L.sub.i  (Eq. 1)

[0065] In Equation (1), ω is the angular frequency of light, k.sub.i(ω) is the wave number corresponding to the i.sup.th waveguide width at angular frequency ω, while L.sub.i refers to the length of the i.sup.th waveguide. Note that L.sub.i could be negative, in which case it would mean that it is located on the other arm. In one example, L.sub.1, L.sub.2, L.sub.4 are positive while L.sub.3 is negative, then the two arm lengths are L.sub.1+L.sub.3 and L.sub.2+L.sub.4. The simplest case of this class of structures is when each arm has a different but uniform width.

[0066] Several constraints may be satisfied by the filter design. Firstly, the pump with central wavelength λ.sub.0 can be situated at a transmission minimum (since this is a pump-rejection filter). Therefore, the left-hand side (LHS) of Equation (1) corresponds an integral multiple m of 2π at the center wavelength λ.sub.0. Since k.sub.i(λ.sub.0)=2πn.sub.i(λ.sub.0)λ.sub.0.sup.−1, for Equation (2). In writing down the expression for the transmission function, in some embodiments, it is proportional to sin.sup.2(Ø/2). In various embodiments Ø/2=mπ, or ϕ=2mπ.

mλ.sub.0=L.sub.1(n.sub.1(λ.sub.0)−Σ.sub.in.sub.i(λ.sub.0)κ.sub.i)  (Eq. 2)

[0067] In Equation (2), κ.sub.i=L.sub.i/L.sub.1. In addition, in some embodiments, it may be desirable for the filter to possess a predetermined free-spectral range (FSR). The free-spectral range can be obtained by setting ϕ(ω.sub.0+2πν.sub.FSR)−ϕ(ω.sub.0)=±2π. Since the FSR may be smaller than the central angular frequency ω.sub.0, the various k.sub.i can be expanded in a Taylor series about k.sub.i(ω.sub.0), where dk.sub.i/dω=ν.sub.gi.sup.−1=n.sub.gi/c. Here n.sub.gi refers to the group refractive index at the center wavelength λ.sub.0. This yields Equation (3) for ν.sub.FSR.

[00004]$\begin{array}{cc}{v}_{\mathrm{FSR}}=\frac{c}{L_1\left(n_{g1}-{\Sigma }_i{n}_{\mathrm{gi}}{\kappa }_i\right)}& \left(\mathrm{Eq}.3\right)\end{array}$

[0068] To check the validity of Equation (3), a conventional MZI may be considered having arms of differing lengths L.sub.1, L.sub.2 but the same widths. This yields Equation (4) for ν.sub.FSR.

[00005]$\begin{array}{cc}{v}_{\mathrm{FSR}}=\frac{c}{n_{g1}\left(L_1-L_2\right)}=\frac{c}{n_{g1}\Delta L}& \left(\mathrm{Eq}.4\right)\end{array}$

[0069] Next, constraints can be derived that make the system invariant to various sources of perturbation, X.sub.j. This can be achieved by setting

[00006]$\frac{\partial \varphi }{\partial X_j}=0.$

A generic approach can be used in which N+1 waveguide widths are used to mitigate N sources of perturbation. In addition, the resonant wavelength λ.sub.c (defined as the location of the transmission minimum in this case) can be made invariant to perturbations as shown in Equation (5).

[00007]$\begin{array}{cc}\frac{\partial n_1}{\partial X_j}={\Sigma }_i{\kappa }_i\frac{\partial n_i}{\partial X_j}& \left(\mathrm{Eq}.5\right)\end{array}$

[0070] Equation (5) is generally valid for various sources of perturbation. For example, X.sub.1≡w, where w is waveguide width and X.sub.2≡h, where h is waveguide thickness. Additional sources of perturbation can be defined, i.e. X.sub.3≡T, where T is temperature, etc. Each source of variation represents an additional linear equation with unknowns κ.sub.i for a given set of w.sub.i.

[0071] While Equation (5) adjusts the resonant wavelength λ.sub.c (the wavelength at which a transmission minimum is present) to be invariant to perturbation, it does not make the shape of the transmission curve near the minimum invariant. In some embodiments, this condition can be imposed by setting the derivative of ∂.sup.2ϕ/∂ω∂X.sub.j to be constant.

[0072] In some embodiments, an additional condition can be imposed to mitigate N different sources of variation yielding Equation (6).

[00008]$\begin{array}{cc}\frac{{\partial }^2{n}_1}{\partial X_j\partial \omega }={\Sigma }_i{\kappa }_i\frac{{\partial }^2{n}_i}{\partial X_j\partial \omega }& \left(\mathrm{Eq}.6\right)\end{array}$

[0073] Equation (6) also represents a set of linear equations with unknowns κ.sub.i for a given set of w.sub.i. In Equation (6), the order of derivatives is swapped for the sake of convenience since ∂n.sub.i/∂w is readily obtained from the effective-index dispersion of waveguides. Furthermore, Equations (2)-(3) can be reduced to a single equation with unknowns κ.sub.i by dividing Equation (2) by Equation (3) as shown below in Equations (7a) and (7b).

[00009]$\begin{array}{cc}\frac{m{\lambda }_0}{c/\mathrm{FSR}}=\gamma =\frac{\left(n_1\left({\lambda }_0\right)-{\Sigma }_i{n}_i\left({\lambda }_0\right){\kappa }_i\right)}{\left(n_{g1}-{\Sigma }_i{n}_{gi}{\kappa }_i\right)}& \left(\mathrm{Eq}.7a\right)\end{array}$$\begin{array}{cc}\gamma n_{g1}-n_1={\mathrm{\Sigma \kappa }}_i\left(\gamma n_{gi}-n_i\right)& \left(\mathrm{Eq}.7b\right)\end{array}$

[0074] Equations (5), (6) and (7b) represent a set of 2N+1 linear equations in κ.sub.i for 2N sources of perturbation or constraints. If Equation (6) is ignored, then there are N+1 linear equations in κ.sub.i. Thus for a predefined set of N+1 waveguide widths, a solution is yielded by obtaining N+1 values of κ.sub.i. Since the various partial derivatives enumerated above are real, a solution to the above problem is generated. Negative values of κ.sub.i are permitted since they represent that section being present in the ‘other’ arm. Thus, the above problem can therefore be cast into a form MX=B as shown below in Equation (8).

[00010]$\begin{array}{cc}M=\left[\begin{array}{cccc}\gamma n_{g2}-n_2& \gamma n_{g3}-n_3& .Math.& \gamma n_{g\left(N+2\right)}-n_{N+2}\\ \frac{\partial n_2}{\partial X_1}& \frac{\partial n_3}{\partial X_1}& .Math.& \frac{\partial n_{N+2}}{\partial X_1}\\ .Math.& .Math.& .Math.& .Math.\\ \frac{\partial n_2}{\partial X_K}& \frac{\partial n_3}{\partial X_K}& .Math.& \frac{\partial n_{N+2}}{\partial X_K}\\ \frac{{\partial }^2{n}_2}{\partial \omega \partial X_1}& \frac{{\partial }^2{n}_3}{\partial \omega \partial X_1}& .Math.& \frac{{\partial }^2{n}_{N+2}}{\partial \omega \partial X_1}\\ .Math.& .Math.& .Math.& .Math.\\ \frac{{\partial }^2{n}_2}{\partial \omega \partial X_K}& \frac{{\partial }^2{n}_3}{\partial \omega \partial X_K}& .Math.& \frac{{\partial }^2{n}_{N+2}}{\partial \omega \partial X_K}\end{array}\right]& \left(\mathrm{Eq}.8\right)\end{array}$$X=\left[\begin{array}{c}{\kappa }_2\\ {\kappa }_3\\ .Math.\\ {\kappa }_{N+2}\end{array}\right]$$B=\left[\begin{array}{c}\gamma n_{g1}-n_1\\ \frac{\partial n_1}{\partial X_1}\\ .Math.\\ \frac{\partial n_1}{\partial X_K}\\ \frac{{\partial }^2{n}_1}{\partial \omega \partial X_1}\\ .Math.\\ \frac{{\partial }^2{n}_1}{\partial \omega \partial X_K}\end{array}\right]$

[0075] FIG. 2 illustrates a single stage of a three-waveguide Cascaded third order MZI based filter 200 using a solution to Equation (8). Each stage can be incoherently cascaded as shown in FIG. 3 that illustrates an incoherently cascaded third-order MZI filter 300 having four stages 305, 310, 315, 320.

[0076] In some embodiments, it may be considered that the above set of equations do not consider loss or extinction ratio thus it may be possible that the obtained lengths from the above set of constraints violate the parameters of the extinction ratio.

[0077] In some embodiments, the use of more or less than N+1 waveguide widths can be used. In either case, the problem is modified to an optimization problem, i.e. a solution to min(MX−B) is desirable.

[0078] In some embodiments, the transitions in waveguide widths may not considered because the waveguide widths may be marginally different and therefore the transition lengths between these may not be relatively large, approximately 1 micron, in one embodiment. This can be relatively smaller than the length of one of the arms, for example approximately 100 microns, in one embodiment.

[0079] The discussion above disclosed an approach to make the MZI's tolerant to sources of perturbation. The next section discloses a design process including an approach to test the statistical performance of an MZI device.

[0080] The first step is to define the geometry of the device and obtain refractive indices of waveguides as functions of w, h, T . . . and other variables for various angular frequencies co. In some embodiments, this can be accomplished using commercial mode solvers. Upon obtaining this information, it can be stored in the form of look-up tables. To simplify storing the spectral dependencies, the refractive index data can be fit as follows and the coefficients n, ∂n/∂w, ∂.sup.2n/∂ω.sup.2 can be stored yielding Equation (9).

[00011]$\begin{array}{cc}n\left(\omega ,X_1.Math.X_N\right)=n\left({\omega }_0,X_1.Math.X_N\right)+\frac{\partial n\left(X_1,.Math.X_n\right)}{\partial \omega }\left(\omega -{\omega }_0\right)+\frac{{\partial }^{2}n\left(X_1,.Math.X_N\right)}{\partial {\omega }^2}{\left(\omega -{\omega }_0\right)}^2& \left(\mathrm{Eq}.9\right)\end{array}$

[0081] Equation (8) can then be solved to obtain various ratios κ.sub.i. If an exact solution cannot be obtained, variation of the central resonant wavelength Δλ.sub.c can be minimized for given standard deviations in perturbation sources σ.sub.X.sub.j according to Equation (10).

[00012]$\begin{array}{cc}\Delta {\lambda }_c={\lambda }_0\frac{{\Sigma }_j{\sigma }_{X_j}.Math.\frac{\partial n_1}{\partial X_j}\underset{i=2}{\overset{N+2}{-.Math.}}{\kappa }_i\frac{\partial n_i}{\partial X_j}.Math.}{n_{g1}-\underset{i=2}{\overset{N+2}{.Math.}}{\kappa }_i{n}_{\mathrm{gi}}}& \left(\mathrm{Eq}.10\right)\end{array}$

[0082] The value of L.sub.1 can be determined using Equation (3). The second MZI in the third-order MZI will can possess L′.sub.1=2L.sub.1 but the same values of κ.sub.i. Using the obtained values of L.sub.i, the values of t.sub.1, t.sub.2, t.sub.3 may be optimized as well as a number of stages N to meet the specifications of extinction ratio, transmission loss and extinction bandwidth. In some embodiments, extinction bandwidth (BW) may be larger than the central wavelength shift Δλ.sub.c, e.g. BW>>Δλ.sub.c. A Monte-Carlo analysis of the system can be performed by repeating a relatively large number (R.sub.N) of random simulations. The sampling can be conducted with knowledge of correlations in a representative fabrication process. In some embodiments, the process can be repeated until a favorable yield is obtained.

[0083] In some embodiments, numerical methods can be used to develop a MZI filter. The output of a filter can obtained using transfer matrices. A cascaded third-order filter can include directional couplers and the propagation of light in the two arms. A filter can be defined to be third-order when two asymmetric MZI's of differential length ΔL and 2ΔL are cascaded coherently. The transfer matrices for directional couplers and MZI arms are shown in Equation (11).

[00013]$\begin{array}{cc}{M}_c=\left[\begin{array}{cc}t& -\mathrm{jK}\\ -\mathrm{jK}& t\end{array}\right],M_{pm}={\alpha }^{1/2}\left[\begin{array}{cc}{e}^{-j{\varphi }_m}{\alpha }_m& 0\\ 0& 1\end{array}\right]& \left(\mathrm{Eq}.11\right)\end{array}$

[0084] In Equation (11), |t|.sup.2 is the transmission coefficient of the directional coupler. Notably, K=√{square root over (1−|t|.sup.2)} while Ø.sub.r (r=1, 2) corresponds to the differential phase in each of the two asymmetric MZI's that constitute a cascaded third-order filter. α.sub.m=e.sup.−rα′ΔL/2 correspond to the additional losses that accrue due to the differential length in each MZI, while α=e.sup.−rα′L/2 is the common absorption experienced by the nominal length L of the MZI arms. For the general multi-waveguide case, L=min(L.sub.1, Σ.sub.iκ.sub.iL.sub.1) and ΔL=|L.sub.1−Σ.sub.iκ.sub.iL.sub.1|. Note that α′ is the absorption coefficient in units of 1/meter.

[0085] Upon utilizing the above transfer matrices the following expressions for the elements H.sub.mk of the overall transfer matrix of the cascaded third-order filter was obtained. A single third-order filter can be defined by three couplers with corresponding parameters t.sub.1, t.sub.2, t.sub.3 and two phase and absorption terms Ø.sub.r, α.sub.r, where r=1, 2 as shown in Equations (12a), (12b), (12c) and (12d).

H.sub.11(ω)=α[−K.sub.1(ω)(t.sub.2(ω)K.sub.3(ω)+α.sub.2K.sub.2(ω)t.sub.3(ω)e.sup.−jϕ.sup.2.sup.(ω))−α.sub.1t.sub.1(ω)e.sup.−jϕ.sup.1.sup.(ω)(K.sub.2(ω)K.sub.3(ω)−α.sub.2t(ω)t.sub.3(ω)e.sup.−jϕ.sup.2.sup.(ω))]  (Eq. 12a)

H.sub.12(ω)=α[−jt.sub.1(ω)(K.sub.3(ω)t.sub.2(ω)+α.sub.2K.sub.2(ω)t.sub.3(ω)e.sup.−jϕ.sup.2.sup.(ω))+jα.sub.1(ω)K.sub.i(ω)e.sup.−jϕ.sup.1.sup.(ω)(ω)(K.sub.2(ω)K.sub.3(ω)−α.sub.2t(ω)t.sub.3(ω)e.sup.−jϕ.sup.2.sup.(ω))]  (Eq. 12b)

H.sub.21(ω)=α[t.sub.1−jK.sub.1(ω)(t.sub.2(ω)t.sub.3(ω)+α.sub.2K.sub.2(ω)K.sub.3(ω)e.sup.−jϕ.sup.2.sup.(ω))−jα.sub.1t.sub.1(ω)e.sup.−jϕ.sup.1.sup.(ω)(ω)(K.sub.2(ω)t.sub.3(ω)−α.sub.2t.sub.2(ω)K.sub.3(ω)e.sup.−jϕ.sup.2.sup.(ω))]  (Eq. 12c)

H.sub.22(ω)=α[t.sub.1(ω)(t.sub.2(ω)t.sub.3(ω)+α.sub.2K.sub.2(ω)K.sub.3(ω)e.sup.−jϕ.sup.2.sup.(ω))+jα.sub.1k.sub.1(ω)e.sup.−jϕ.sup.1.sup.(ω)(ω)(K.sub.2(ω)t.sub.3(ω)+α.sub.2K.sub.3(ω)t.sub.2(ω)e.sup.−jϕ.sup.2.sup.(ω))]  (Eq. 12d)

[0086] The validity Equations (12a)-(12d) can be shown by verifying that |H.sub.qp(ω)|.sup.2+|H.sub.pp(ω)|.sup.2=1 for q, p=1, 2 under conditions of no loss (i.e. α′=0). This relates to the conservation of energy. The transmission loss and pump-rejection ratios can be calculated in Equations (13a) and (13b), respectively.

[00014]$\begin{array}{cc}{t}_{\mathrm{loss}}=10{\log }_{10}\left(\frac{{\int }_{-\infty }^{\infty }{.Math.E_{\mathrm{out},1}\left(\omega \right).Math.}^2\left(I_s\left(\omega \right)+I_i\left(\omega \right)\right)d\omega }{{\int }_{-\infty }^{\infty }\left[I_s\left(\omega \right)+I_i\left(\omega \right)\right]d\omega }\right)\mathrm{dB}& \left(\mathrm{Eq}.13a\right)\end{array}$$\begin{array}{cc}{t}_{pump}=1000\times 10{\log }_{10}\left(\frac{{\int }_{-\infty }^{\infty }{.Math.E_{\mathrm{out},2}\left(\omega \right).Math.}^2{I}_p\left(\omega \right)d\omega }{{\int }_{-\infty }^{\infty }{I}_p\left(\omega \right)d\omega }\right)m\mathrm{dB}& \left(\mathrm{Eq}.13b\right)\end{array}$

[0087] In these embodiments the waveguides considered are silicon-on-insulator (SOI) strip waveguides, however other embodiments can use different configurations. The material dispersion can be based on the Palik model at room temperature. The dispersion of the effective index can be fit according to Equation (9). In this embodiment the center wavelength λ.sub.0=2πc/ω.sub.0=1.55 μm. The obtained coefficients are plotted in FIG. 4 showing the effective index parameters as a function of waveguide width and height for a silicon-on-insulator waveguide. The obtained effective index is fit to Equation (9).

[0088] A full parameter sweep of the refractive index over angular frequency ω, waveguide width (ω) and thickness (h) is performed. In FIGS. 5A-5B, the derivatives are plotted with respect to w, h of the effective index at the wavelength λ.sub.0=1.55 μm.

[0089] In FIGS. 5A and 5B, the derivative ∂n/∂w is plotted. The value is invariant with thickness but changes dramatically with width. This indicates that waveguide width variations can be mitigated using this approach. On the other hand, while ∂n/∂h does vary with width, it only does so mildly; it varies more with regard to thickness. The magnitude of change is about four times larger than ∂n/∂w. In some embodiments, standard deviations σ.sub.h of the thickness tend be smaller than those of the widths (see Table. 1 showing parameters of simulations), which reduces their impact.

TABLE-US-00001 TABLE 1 Parameters used for simulations Parameter Value Standard deviation of waveguide width (σ.sub.ω) 3 nm.sup.5 Standard deviation of waveguide height (σ.sub.h) 0.5 nm.sup.6 Material index Palik (from Lumerical) Temperature (T) 300K Absorption coefficient (α′) 0.3 dBcm.sup.−1 or 7.5 m.sup.−1 Transmission coefficients (|t.sub.1(ω.sub.0)|.sup.2, |t.sub.2(ω.sub.0)|.sup.2, |t.sub.3(ω.sub.0)|.sup.2) 0.5, 0.75, 0.93 Number of stages 4 Pump, signal and idler distributions Gaussian with 5 GHz e.sup.−2 bandwidth Correlations ω, h for each third-order MZI stage uncorrelated.

[0090] Due to the relative invariance of ∂n/∂h, with respect to w, the strategy of using multiple waveguide widths to mitigate variation in this parameter may not be very efficacious for particular applications. In principle, a solution is possible but the lengths of arms obtained turn out to be in the range of centimeters which can be too large for some applications. Therefore, in some applications that may benefit from small filter sizes, it would be beneficial to reduce the values of σ.sub.h.

[0091] The coupling coefficients of the directional couplers can be determined by obtaining the even and odd modes of the coupled waveguide system. The coupling length can then be determined according to Equation (14).

[00015]$\begin{array}{cc}{t}_m\left(\omega \right)=\sin \left[\frac{\Delta n\left(\omega \right){\lambda }_0}{\Delta n\left({\omega }_0\right)\lambda }{\sin }^{-1}\left(t_m\left({\omega }_0\right)\right)\right]& \left(\mathrm{Eq}.14\right)\end{array}$

[0092] The statistical performance of standard cascaded third-order filters is examined to estimate the yield for such devices. In this approach a Monte-Carlo calculation was employed. Waveguide widths and thicknesses were chosen at random and their effective indices are obtained from the previously generated look-up tables. Similarly, the effective super-mode indices of the couplers are obtained. The coupling coefficients are then calculated using Equation (14) and the parameters from Table 1 are used. The standard deviations for waveguide width σ.sub.w=3 nm and thickness σ.sub.h=0.5 nm are plotted in FIG. 6 showing statistical distributions of effective index n.sub.eff.

[0093] In this embodiment the entire dispersion curve has been shifted. The distribution of effective indices is slightly asymmetric. Therefore, in assuming a 3 nanometer waveguide width standard deviation and 0.5 nanometer standard deviation in thickness, this example evaluates variations more germane to die-to-die or intra-die variations. Therefore, a relevant parameter may be the critical dimension uniformity (CDU).

[0094] The overall performance for a N=4 stage, incoherently cascaded third-order filter can then be obtained. The design described above had the goal of meeting the specifications for a pump rejection filter, that can be, in one example, 120 dB of rejection and 50 mdB of loss. However, from FIG. 7 that shows the statistical behavior of cascaded third-order MZI's without mitigation mechanisms, it can be seen that the mean rejection ratio has shifted to approximately 60 dB and the mean absorption coefficient has shifted to approximately 1800 mdB, however these may have different values in other embodiments.

[0095] In one embodiment, a fabrication tolerant MZI design uses asymmetric widths for each MZI arm. In this particular embodiment it is desired to mitigate variations to both thickness (h) and width (w), so the quantity in Equation (10) is minimized. The results are plotted in FIG. 8 that illustrates designs to minimize susceptibility to fabrication errors. The minimization procedure yields a value of Δλ.sub.c≈700 pm at various values of κ for varying values of w.sub.1 and h=220 nm. Incidentally, the minimization yields ∂n.sub.1/∂w−∂n.sub.2/∂w=0, while being at the mercy of σ.sub.h|∂n.sub.1/∂h−∂n.sub.2/∂h|. Therefore, in some embodiments, σ.sub.h should be reduced.

[0096] As shown in FIG. 9, the statistical distribution of cascaded third-order MZI's with asymmetric widths w.sub.1=500 nm and w.sub.2=540 nm, h=220 nm are illustrated. In some embodiments, this value can be reduced by increasing the height of the waveguides. For instance at h=245.5 nm, Δλ.sub.c≈580 pm. However, it also appears that using thicker waveguides in some embodiments causes the transmission loss to increase due to dispersion. Therefore, over engineering this aspect of the system may not be worthwhile for some embodiments. Using such a configuration, FIG. 9 illustrates the performance for N=4 incoherently cascaded third-order filters. An improvement in performance compared to that depicted in FIG. 7 is evident with the mean rejection ratio shifting to 110 dB and mean loss shifting to 188 mdB.

[0097] Furthermore, if coupler variations with respect to fabrication uncertainties (simply referred to as coupler variations henceforth) are ignored, then the performance is shown in FIG. 10 illustrating the statistical distribution of cascaded third-order MZI's in the absence of coupler variations with respect to fabrication uncertainties. The rejection ratio shifts to 154 dB, while the transmission loss changes to 165 mdB. In some embodiments, this can indicate that coupler variations predominantly produce vertical movements in the spectral response while the index changes produce mainly horizontal shifts. Horizontal shifts affect both rejection ratio and transmission loss, while vertical shifts predominantly affect rejection ratios.

[0098] In some embodiments, while using asymmetric arms can make |∂n.sub.1/∂w−∂n.sub.2/∂w|=0, it may not correlate to a transmission minimum located at λ.sub.0=1.55 μm. In the above embodiments, it is fortuitous that for κ+δκ, the above resonance condition is satisfied. Here, δκ is a relatively small amount of adjustment imparted to κ. Therefore, there may be a residual error of −δκ∂2.sub.2/∂w, which is may be undesirable. However, if two additional waveguide widths are used (i.e. w.sub.2, w.sub.3), then some embodiments may have improved results. This is demonstrated in FIG. 11, where one of the arms contains two widths of 0.5, 0.66 microns. More specifically, FIG. 11 illustrates the statistical distribution of MZI properties for three waveguide widths L.sub.1=22.96 μm, m=58, κ.sub.i=[1, 4.2805, −4.6974] and w.sub.i=[0.5, 0.56, 0.66] microns. In this embodiment, the design can be constrained to satisfy a condition for transmission minimum at λ.sub.0 (Equation (1)), FSR (Equation (3)) and insensitivity to width variations (Equation (5)).

[0099] In FIG. 11, coupler variations are neglected, building on the results from FIG. 10. As can be seen, there is an additional 10 dB improvement in rejection ratio, while an improvement in transmission loss by approximately 70 mdB. While thickness variations may not be mitigated using this approach since ∂n/∂h is not a function of w, the additional constraint of having ∂.sup.2Ø∂w∂ω=0 may be included, which yields the performance in FIG. 12 showing the statistical distribution of an MZI filter with four waveguide widths. L.sub.1=25.51 microns, κ.sub.i=[1, 4.1464, −4.5875, 0.1662] and w.sub.i=[0.5, 0.56, 0.66, 0.76] microns. The shape of the distribution appears to change, although improvements in mean values do not appear to occur.

[0100] In principle, compensation for perturbations in w, h can be simultaneously achieved by choosing arms with different w, h as shown in Equation (15).

[00016]$$\begin{gathered} \begin{array}{cc}\frac{\partial n_1}{\partial w}{|}_{w_1,h_1}=\frac{\partial n_2}{\partial w}{|}_{w_2,h_2} \\ \frac{\partial n_1}{\partial h}{|}_{w_1,h_1}=\frac{\partial n_2}{\partial h}{|}_{w_2,h_2}& \left(\mathrm{Eq}.15\right)\end{array} \end{gathered}$$

[0101] This results in a value of Δλ.sub.c=26 pm. The results are plotted in FIG. 13 illustrating fabrication tolerance achieved using asymmetric widths as well as heights. w.sub.1=500 nm, w.sub.2=535 nm, h.sub.1=220 nanometers and h.sub.2=245 nanometers. The average pump rejection shifts to 174 dB and the average loss is 125 mdB, which is smaller compared to the case when only asymmetric widths without coupler variations (FIG. 10) are considered. Here, too the effect of coupler variations have been ignored. The marginal increase in absorption relative to FIGS. 11 and 12 is that the constraint of fixing λ.sub.c is not satisfied. In some embodiments, waveguide geometries that effectively enable different heights (such as rib waveguides) can be used. Furthermore, some embodiments can use both different heights and multiple widths to further improve performance.

[0102] While obtaining different thicknesses can be challenging in some embodiments, there may be ways to accomplish this by using unconventional cross-sections that are compatible with current CMOS-foundry processes, as shown in FIGS. 14A-14D. In one example embodiment shown in FIG. 14A, a conventional strip waveguide is shown. In FIG. 14B, a cross-section which has an additional silicon nitride or silicon layer on top of the SOI strip waveguide that modifies the effective height is shown. FIG. 14C illustrates a rib waveguide and FIG. 14D illustrates a modified rib waveguide showing two other embodiments that offer height changes.

[0103] The effect of coupler dispersion and insertion loss on the system can now be considered. All the systems are assumed to possess the three-waveguide design from FIG. 11, however other embodiments may have other configurations. In the first embodiment illustrated in FIG. 15, the coupler dispersion is retained while waveguide loss is reduced to 0.1 dB/cm. FIG. 15 illustrates performance of the filter when the loss is reduced to 0.1 dB/cm and with couplers robust to fabrication variations but with varying transmission with respect to frequency.

[0104] With this improvement, the transmission loss has reduced to 93 mdB, while the pump rejection has been minimally altered. On the contrary, when the loss is maintained at 0.3 dB/cm but the couplers are fab-tolerant and also not dispersive, the transmission loss falls below the 50 mdB level as shown in FIG. 16. FIG. 16 illustrates performance of the filter when the loss is 0.3 dB/cm with couplers that are robust to fabrication variations and also with constant transmission with respect to frequency.

[0105] Thus, in order to meet device specifications, in some embodiments, the couplers may be fab-tolerant and broadband. When the loss is also reduced to 0.1 dB/cm with couplers robust to fabrication and also with constant transmission coefficients with respect to frequency, transmission losses reduce to 28 mdB as can be seen in FIG. 17. FIG. 17 illustrates performance of the filter when loss is reduced to 0.1 dB/cm with couplers robust to fabrication variations and also with constant transmission with respect to frequency. In order to reduce transmission loss values below 25 mdB, the number of stages may be reduced to three, although this may also reduce the mean pump rejection ratio. Therefore, in order to further improve the yield, improved process control or reduced values of σ.sub.w, σ.sub.h may be needed.

[0106] As described above, the variations of width and thickness were treated as independent random variables and each stage was assumed to vary independently. In this section the case when the width and height variations are uncorrelated but all stages are well-correlated is evaluated. When the correlation between each stage increases, the spread in performance increases as shown in FIG. 18. FIG. 18 illustrates asymmetric widths where width and height variations are independent and every stage is correlated with a loss of 0.3 dB/cm.

[0107] However, in some embodiments, if the couplers are made insensitive to fabrication, then the performance improves as seen in FIG. 19. FIG. 19 shows an embodiment having asymmetric widths where width and height variations are independent and each stage is correlated. Couplers are considered fabrication tolerant and the loss is 0.3 dB/cm. Using three or four waveguide widths, as was the case in FIGS. 11 and 12, improves performance even more, bringing elements close to specifications in FIG. 20. FIG. 20 illustrates an embodiment where width and height variations of every stage are correlated. In additional to mitigating coupler variations, one embodiment uses three waveguide widths. The waveguide loss assumed in this embodiment is 0.3 dB/cm.

[0108] If develop broadband couplers are developed while maintaining loss at 0.3 dB/cm, the performance improvement is line with trends in the previous embodiments, as shown in FIG. 21. FIG. 21 illustrates an embodiment where width and height variations are independent but are correlated for all stages. The structure uses three waveguide widths and fab-tolerant and broadband couplers and the insertion loss is 0.3 dB/cm. As illustrated in FIG. 22, the loss is reduced to 0.1 dB/cm, which brings the performance to similar levels as shown in FIG. 17. More specifically, even when the correlations are not favorable, the devices have comparable yield. FIG. 22 illustrates an embodiment where width and height variations are independent but are correlated for all stages. The structure uses three waveguide widths, fab-tolerant and broadband couplers as well as a reduced insertion loss of 0.1 dB/cm.

[0109] Serial improvements are summarized that can be achieved for various design improvements shown in Table 3. In some embodiments, broadband, fabrication insensitive couplers enable the system to meet performance specifications. In further embodiments, reducing waveguide losses on-chip may help improve the performance and yield. In addition embodiments having σ.sub.w<3 nanometers and σ.sub.h<0.5 nanometers may be used.

[0110] Table 3 summarizes different embodiments that may have reduced performance and also identifies various strategies that could potentially address the performance. Each point labelled (i)-(iv) in Table 3 is discussed in more detail below.

[0111] (i) In some embodiments, the use of asymmetric arm widths may achieve tuning-free operation of cascaded third-order filters. Use of three or four waveguide widths helps achieve pinning the transmission minimum and also compensates ∂.sup.2n/∂ω∂ω.

[0112] (ii) In some embodiments, the use of multi-waveguide sections can mitigate many sources of variation but due to the invariance of ∂n/∂h to w, this approach may need long device lengths to mitigate thickness variations. In principle, using different waveguide heights can also address thickness variation issues, although this may not be a CMOS-foundry compatible process. Some embodiments may use unconventional waveguide geometries to effectively engineer a height difference.

TABLE-US-00002 TABLE 2 Mean Pump Mean μ.sub.loss + rejection μ − ≥120 Transmission σ.sub.loss ≤25 mdB Design μpump (dB) σ.sub.pump(dB) dB(%) loss (mdB) (mdB) (%) Standard third-order, 4 stage MZI, 60 44 0 1830 3074 0 0.3 dB/cm Asymmetric widths, 0.3 dB/cm 110 92 28 188 280 0 Asymmetric widths and robust 154 130 87 164 237 0 couplers, 0.3 dB/cm Multiple widths, robust couplers, 163 97.4 99.4 110 155.1 0 0.3 dB/cm Standard third-order, 4 stage MZI, 163 98.2 99.43 44 80 2.2 0.3 dB/cm Multiple widths, robust couplers and 163 143.5 97.1 93 144.1 0 0.1 dB/cm loss Multiple widths, robust and broadband 163 141.5 97.4 28 67 67.3 couplers and 0.1 dB/cm loss Multiple widths, robust couplers and 122 104 55.1 22 56 76.2 0.1 dB/cm loss, 3 stages σ.sub.w = 1 nm, σ.sub.h = 0.25 nm and Multiple 188 175 100 10.5 12.5 99.7 widths, robust, broadband couplers, 0.1 dB/cm, 4 stages σ.sub.w = 1 nm, σ.sub.h = 0.25 nm and Multiple 141 131 98.5 7.8 8.8 100 widths, robust, broadband couplers, 0.1 dB/cm, 3 stages Asymmetric widths, correlated stage 111 73.26 36 210 609 0 variations, 0.3 dB/cm Asymmetric widths, robust couplers and 148 102 69 178 329.3 0 correlated stage variations, 0.3 dB/cm Multiple widths, robust couplers and 164 123 78 105 177 0 correlated stage variations, 0.3 dB/cm Multiple widths, robust and broadband 163 120.75 80 44 103 21 couplers and correlated stage variations, 0.3 dB/cm Multiple widths, robust and broadband 163 120 78 28 95 79 couplers and correlated stage variations, 0.1 dB/cm σ.sub.w = 1 nm, σ.sub.h = 0.25 nm and Multiple 187 163 99.2 10.5 13.85 99.4 widths, robust, broadband couplers, 0.1 dB/cm, correlated σ.sub.w = 1 nm, σ.sub.h = 0.25 nm and Multiple 141 124 91 7.8 10.38 99.6 widths, robust, broadband couplers, 0.1 dB/cm, 3 stages

TABLE-US-00003 TABLE 3 Problem Reason Value Strategy (i) Resonance shift Multiple waveguide widths (ii) Sensitivity to height Equal heights 62 mdB Si—SiO2—Si, Si—SiN—Si waveguides (iii) Transmission loss Dispersion Broadband couplers in couplers (iv) Bandwidth of filter Roll-off 2 nm at Reduced waveguide loss −150 dB Additional stages or alternate architectures

[0113] As shown in FIG. 23, in some embodiments, reducing σ.sub.w to 1 nanometer and σ.sub.h to 0.25 nanometer from 3 and 0.5 nanometer respectively enables the specifications to comfortably meet the goals.

[0114] (iii) In some embodiments, the role of coupler dispersion and variations with fabrication may be important. Designing couplers that are more broadband and insensitive to fabrication variations may be needed to make a filter robust to perturbations.

[0115] (iv) In some embodiments, to meet specifications, loss may reach approximately 0.1 dB/cm. This may enable specifications to be exceeded by adding further cascaded third-order MZI filter stages. In further embodiments, using three stages may meet rejection ratio targets while keeping losses below the 25 mdB level.

[0116] (v) In some embodiments, further improvement of fabrication tolerances to σ.sub.w<<3 nm and σ.sub.h<<0.5 nanometer may improve the mean pump rejection to 188 dB and average loss to 10.54 mdB for a four stage cascaded third-order MZI as is seen in FIGS. 23A and 23B.

[0117] FIG. 24 illustrates yield percentage of loss ≤25 mdB and rejection ratio ≥120 db. Variations are correlated, with insertion loss of 0.1 dB/cm, broadband and fab-tolerant couplers and multiple waveguide width arms.

[0118] Although MZI filter 100 (see FIG. 1) is described and illustrated as one particular type of MZI-based photonic device, a person of skill in the art with the benefit of this disclosure will appreciate that compensation structures as described above are suitable for use with myriad other MZI-based photonic devices. For example, in some embodiments the MZI passive compensation structures disclosed herein can be implemented in MZI-based photonic switching devices, as described in more detail below.

[0119] FIGS. 25 and 26 show example MZI-based photonic switches 2500 and 2600, respectively, that include one or more variable phase-shifters and can also include one or more compensation structures. Photonic switches 2500 and 2600 are similar to MZI filter 100 (see FIG. 1), each having two parallel waveguides (2510a, 2510b in FIG. 25, and 2610a and 2610b in FIG. 26), however photonic switches 2500 and 2600 each include one or more phase shifters (2505a, 2505b, 2505c in FIGS. 25 and 2605 in FIG. 26) disposed in one or more waveguides of each photonic switch. Phase-shifters (2505a, 2505b, 2505c in FIGS. 25 and 2605 in FIG. 26) can be implemented a number of ways in integrated photonic circuits and can provide control over the relative phases imparted to the optical field in each waveguide. In some embodiments, variable phase-shifters can be implemented using thermo-optical switches.

[0120] In some embodiments thermo-optical switches can use resistive elements fabricated on a surface of the photonic device. Employing the thermo-optical effect in these devices can provide a change of the refractive index n by raising the temperature of the waveguide by an amount of the order of 10.sup.−5 K. One of skill in the art having had the benefit of this disclosure will understand that any effect that changes the refractive index of a portion of the waveguide can be used to generate a variable, electrically tunable, phase shift. For example, some embodiments can use beam splitters based on any material that supports an electro-optic effect. In some embodiments so-called χ.sup.(2) and χ.sup.(3) materials can be used such as, for example, lithium niobate, BBO, KTP, BTO, and the like and even doped semiconductors such as silicon, germanium, and the like.

[0121] In some embodiments, switches with variable transmissivity and arbitrary phase relationships between output ports can also be achieved by combining directional couplings (e.g., directional couplings 2515a, 2515b in FIGS. 25 and 2615a, 2615b in FIG. 26), and one or more variable phase-shifters (e.g., phase-shifters 2505a, 2505b, 2505c in FIGS. 25 and 2605 in FIG. 26) within each photonic switch. Accordingly, complete (e.g., analog or digital) control over the relative phase and amplitude of the two output ports can be achieved by varying the phases imparted by phase shifters (2505a, 2505b, 2505c in FIGS. 25 and 2605 in FIG. 26). FIG. 26 illustrates a slightly simpler example of a MZI-based photonic switch that allows for variable transmissivity between ports 2620a and 2620b by varying a phase imparted by phase shifter 2605.

[0122] In some embodiments one or more compensation structures can be implemented within MZI-based photonic switches 2500,2600 using compensation equations similar to those described above with regard to MZI filter 100 (see FIG. 1). More specifically, the compensation equations can be used to determine a width and a length of each compensation portion that can be used to reduce a shift in frequency response caused by various perturbations, including variations in manufacturing widths of the waveguides, manufacturing variations in thicknesses of the waveguides and variations in temperature. Similar to the compensation structures described for MZI filter 100 (see FIG. 1), compensation structures can be employed in one or more waveguides (2510a, 2510b in FIG. 25, and 2610a and 2610b in FIG. 26), and each compensation structure can each have a quantity of waveguide widths that is greater than the number of perturbations, however the governing equations may be different for an MZI-based photonic switch embodiment, as described in more detail below.

[0123] The phase relationship in an MZI-based photonic switch embodiment may be described as follows. The first two terms can be the same as MZI filter 100 (see FIG. 1), however a third term corresponding to a sum of various index changes, Δn.sub.j, weighted by various overlap integrals Γ.sub.j, can be added, as described by Equations (16) and (17).

[00017]$\begin{array}{cc}\frac{\left(2m+1\right){\lambda }_0}{2}=n_1\left({\omega }_0\right)L_1-{\Sigma }_i{\kappa }_i{L}_1{n}_i\left({\omega }_0\right)+{\Sigma }_j{\Gamma }_j\left({\omega }_0\right)\Delta n_j\left({\omega }_0\right)L_1& \left(\mathrm{Eq}.16\right)\end{array}$$\begin{array}{cc}{v}_{FSR}=\frac{c}{L_1\left(n_{g1}-{\Sigma }_i{n}_{gi}{\kappa }_i\right)}& \left(\mathrm{Eq}.17\right)\end{array}$

The corresponding compensation equation for the case of an MZI-based photonic switch which requires invariance to width can be described by Equation (18).

[00018]$\begin{array}{cc}\frac{\partial {\varphi }_j}{\partial w}=L_1\left(\frac{\partial n_1}{\partial w}-\frac{{\Sigma }_i{\kappa }_i\partial n_i}{\partial w}+\frac{{\Sigma }_j\partial {\Gamma }_j}{\partial w}\Delta n_j\right)=0& \left(\mathrm{Eq}.18\right)\end{array}$

Equation (18) can be reduced to Equation (19).

[00019]$\begin{array}{cc}\frac{{\Sigma }_i{\kappa }_i\partial n_i}{\partial w}={\Sigma }_j\frac{\partial {\Gamma }_j}{\partial w}\Delta n_j+\frac{\partial n_1}{\partial w}& \left(\mathrm{Eq}.19\right)\end{array}$

In some embodiments, Equations (20) through (22) can be used to account for compensation of higher-order derivatives.

[00020]$\begin{array}{cc}\frac{{\Sigma }_i{\kappa }_i{\partial }^2{n}_i}{\partial w\partial \omega }={\Sigma }_j\frac{\partial }{\partial \omega }\left[\frac{\partial {\Gamma }_j}{\partial w}\Delta n_j\right]+\frac{{\partial }^2{n}_1}{\partial w\partial \omega }& \left(\mathrm{Eq}.20\right)\end{array}$$\begin{array}{cc}\frac{{\Sigma }_i{\kappa }_i{\partial }^2{n}_i}{\partial w^2}={\Sigma }_j\left[\frac{{\partial }^2{\Gamma }_j}{\partial w^2}\Delta n_j\right]+\frac{{\partial }^2{n}_1}{\partial w^2}& \left(\mathrm{Eq}.21\right)\end{array}$$\begin{array}{cc}\frac{\left(n_1\left({\omega }_0\right)-{\Sigma }_i{n}_i\left({\omega }_0\right){\kappa }_i+{\Sigma }_j{\Gamma }_j\left({\omega }_0\right)\Delta n_j\left({\omega }_0\right)\right)}{\left(n_{g1}-{\Sigma }_i{n}_{gi}{\kappa }_i\right)}=\gamma =\frac{\left(m+\frac{1}{2}\right){\lambda }_0}{c/\mathrm{FSR}}& \left(\mathrm{Eq}.22\right)\end{array}$

Generalizing to arbitrary perturbations X.sub.k, the set of compensation equations for MX=B can be described by Equations (23) through (25).

[00021]$\begin{array}{cc}M=\left[\begin{array}{cccc}\gamma n_{g2}-n_2& \gamma n_{g3}-n_3& .Math.& \gamma n_{g\left(N+2\right)}-n_{N+2}\\ \frac{\partial n_2}{\partial X_1}& \frac{\partial n_3}{\partial X_1}& .Math.& \frac{\partial n_{N+2}}{\partial X_1}\\ .Math.& .Math.& .Math.& .Math.\\ \frac{\partial n_2}{\partial X_K}& \frac{\partial n_3}{\partial X_K}& .Math.& \frac{\partial n_{N+2}}{\partial X_K}\\ \frac{\partial n_2}{\partial \omega \partial X_1}& \frac{\partial n_3}{\partial \omega \partial X_1}& .Math.& \frac{\partial n_{N+2}}{\partial \omega \partial X_1}\\ .Math.& .Math.& .Math.& .Math.\\ \frac{\partial n_2}{\partial \omega \partial X_K}& \frac{\partial n_3}{\partial \omega \partial X_K}& .Math.& \frac{\partial n_{N+2}}{\partial \omega \partial X_K}\end{array}\right]& \left(\mathrm{Eq}.23\right)\end{array}$$\begin{array}{cc}X=\left[\begin{array}{c}{\kappa }_2\\ {\kappa }_3\\ .Math.\\ {\kappa }_{N+2}\end{array}\right]& \left(\mathrm{Eq}.24\right)\end{array}$$\begin{array}{cc}B=\left[\begin{array}{c}\gamma n_{g1}-n_1-{\Sigma }_j{\Gamma }_j\left({\omega }_0\right)\Delta n_j\left({\omega }_0\right)\\ \frac{\partial n_1}{\partial X_1}+{\Sigma }_j\frac{\partial {\Gamma }_j}{\partial X_1}\Delta n_j\\ .Math.\\ \frac{\partial n_1}{\partial X_K}+{\Sigma }_j\frac{\partial {\Gamma }_j}{\partial X_k}\Delta n_j\\ .Math.\\ \frac{{\partial }^2{n}_1}{\partial \omega \partial X_K}+\frac{\partial }{\partial \omega }{\Sigma }_j\frac{\partial {\Gamma }_j}{\partial X_k}\Delta n_j\\ .Math.\\ \frac{{\partial }^2{n}_1}{\partial X_K^2}+{\Sigma }_j\frac{{\partial }^2{\Gamma }_j}{\partial X_k^2}\Delta n_j\end{array}\right]& \left(\mathrm{Eq}.25\right)\end{array}$

[0124] Photonic switches 2500 and 2600 illustrated in FIGS. 25 and 26, respectively, and the associated compensation equations are two examples of how compensation structures can be implemented in myriad MZI-based photonic devices. One of skill in the art with the benefit of this disclosure can appreciate that similar compensation structures can be implemented in other MZI-based photonic devices.

[0125] For simplicity, various components, such as the optical pump circuitry, substrates, cladding, and other components of MZI filter 100 (see FIG. 1) are not shown in the figures. In the foregoing specification, embodiments of the disclosure have been described with reference to numerous specific details that can vary from implementation to implementation. The specification and drawings are, accordingly, to be regarded in an illustrative rather than a restrictive sense. The sole and exclusive indicator of the scope of the disclosure, and what is intended by the applicants to be the scope of the disclosure, is the literal and equivalent scope of the set of claims that issue from this application, in the specific form in which such claims issue, including any subsequent correction. The specific details of particular embodiments can be combined in any suitable manner without departing from the spirit and scope of embodiments of the disclosure.

[0126] Additionally, spatially relative terms, such as “bottom or “top” and the like can be used to describe an element and/or feature's relationship to another element(s) and/or feature(s) as, for example, illustrated in the figures. It will be understood that the spatially relative terms are intended to encompass different orientations of the device in use and/or operation in addition to the orientation depicted in the figures. For example, if the device in the figures is turned over, elements described as a “bottom” surface can then be oriented “above” other elements or features. The device can be otherwise oriented (e.g., rotated 90 degrees or at other orientations) and the spatially relative descriptors used herein interpreted accordingly.