INTEGRATED OPTICAL WAVEGUIDE DEVICE
Integrated Optical Device
The present invention relates to an integrated optical device.
It is often necessary to split a waveguide into two outputs, ie to create a Y- branch. This is typically done literally by forming the waveguide into a Y- configuration, as shown schematically in Figure 1. An incoming waveguide 10 divides along its central axis at 12 into two departing waveguides 14, 16. These initially run closely to each other with only a small divergence, to allow the optical mode to settle into the two branches. After a short distance, they then diverge to wherever intended.
However, there are difficulties in this structure. First, the split between the departing waveguides 14, 16 is at the centre of the optical mode. At this location, it presents the maximum disturbance. Second, it is sometimes necessary to subject waveguide structures to thermal oxidation in order to reduce surface roughness. This effectively shrinks the structure by loss of a thin layer from the surface. The effect of this is shown schematically in Figure 1 in dotted lines. Some exaggeration of the effect has been included to allow clear illustration . It will be seen that in the crucial concave formation at which the waveguides separate, the effect of thermal oxidation is to round off the formerly sharp division. This creates a flat reflective surface at the centre of the optical mode, thus increasing scattering from the device and hence its insertion losses. That concave division is also difficult to form with a sharp point ab initio, and thus will be slightly blunt even before oxidation.
M. H. Hu et al report a Y-branch in "A low loss and compact waveguide Y- branch using refractive-index tapering", IEEE Photonics Technology Letters, Vol. 9, No. 2, February 1997. An input waveguide reduces in height adiabatically, eventually disappearing, between a pair of asymmetrically tapered output waveguides which are contiguous on either side. The optical mode is thus forced downwards in that region, and couples into the output waveguides. This arrangement can address the problems of the above. However, it is difficult to fabricate a taper having a smooth and accurate reduction in height.
United States Patent No. 5,818,989 discloses a Y-branching waveguide structure embedded within a clad layer above a substrate of silicon or the like. The branch structure comprises a main waveguide and two branching waveguides that are optically connected to the main waveguide. The main waveguide includes a tapered contracting waveguide portion that gradually contracts along its sides while extending longitudinally, and the branching waveguides have expanding waveguide portions that taper in the direction opposite that of the main waveguide. The minimum width of each of the waveguides disclosed in US 5,818,989 is lμm.
The present invention seeks to provide an improved waveguide branch structure and an improved process for forming the branch structure.
According to a first aspect, the present invention provides an integrated optical device comprising a first waveguide which branches into at least two second waveguides, the branch comprising a structure in which at least one of the second waveguides starts with a taper that is located adjacent the first waveguide, the waveguides having been subjected to a surface etch process after their initial formation.
According to a second aspect, the invention provides a process of forming an integrated optical device, comprising initially forming a waveguide branch structure in which a first waveguide branches into at least two second waveguides, the branch structure being such that at least one of the second waveguides starts with a taper that
is located adjacent the first waveguide, the process further comprising subjecting the waveguides to a surface etch process after their initial formation.
The invention has the advantage that by subjecting the waveguides to a surface etch process after their initial formation, the definition of the waveguide branch structure can be improved. For example, the lithographic process typically used to produce integrated waveguides can often result in what is commonly termed "blurring" of the features due to the finite resolution possible from the lithographic process (i.e. the fact that the resolution cannot be perfect). A subsequent surface etch step after the initial formation of the waveguides can improve the effective resolution of the process, for example. This improvement in the definition of the waveguide branch structure is only possible because the branch structure is not a conventional one having a literal Y configuration as shown in Figure 1 and as explained above. Although the waveguide branch structure disclosed in US 5,818,989 is also not such a conventional structure, it still suffers from the problem of lithographic blur. The present invention ingeniously solves this problem by exploiting the fact that the definition (or resolution) of a tapered branch structure can be improved by a subsequent surface etch process.
In preferred embodiments of the invention, the surface etch process after the initial formation of the waveguides comprises oxidation of the waveguides. The oxidation of the waveguides is preferably carried out by heating the waveguides in an atmosphere comprising oxygen. The heating may, for example, be carried out in a so- called "wet" process in which the atmosphere comprises steam, and the material of the waveguides chemically reacts with the steam (i.e. gaseous or vaporous water). Alternatively, for example, the heating may be carried out in a so-called "dry" process in which a dry atmosphere comprising gaseous oxygen is used. Preferably the heating of the waveguides is carried out to a maximum temperature of at least 700°C, more preferably at least 800°C, even more preferably at least 900°C. Preferably the maximum temperature is no greater than 1200°C, more preferably no greater than 1100°C, for example approximately 1000°C. The heating is preferably carried out in
an oven or furnace under an atmosphere of oxygen, or an atmosphere including oxygen. The heating is preferably carried out for between 0.2 and 10 hours.
Preferably the waveguides of the invention are silicon waveguides.
Accordingly, a third aspect of the invention provides an integrated optical device comprising a first silicon waveguide which branches into at least two second silicon waveguides, the branch comprising a structure in which at least one of the second waveguides starts with a taper that is located adjacent the first waveguide.
Preferably the waveguides of the invention are rib waveguides. A rib waveguide is generally one in which an elongate rib portion of the material of the waveguide extends above slab regions of the waveguide material on opposite lateral sides of the rib portion. By the terms "taper" and "tapering" in the present specification are generally meant a lateral taper or lateral tapering, e.g. in which the width of the rib portion of a rib waveguide decreases (such that the slab regions on opposite lateral sides of the rib portion effectively become closer together).
Accordingly, a fourth aspect of the invention provides an integrated optical device comprising a first rib waveguide which branches into at least two second rib waveguides, the branch comprising a structure in which at least one of the second waveguides starts with a taper that is located adjacent the first waveguide.
It is preferred that the branch comprises a structure in which the first waveguide ends with a taper and the second waveguides each start with a taper that is located adjacent the end taper of the first waveguide.
Thus, the reducing width of the first waveguide again forces the optical mode out and into the second waveguides. These can be located so as initially to accept the evanescent portion of the mode. There need be no disturbance to the central axis of the mode until the tip of the taper of the first waveguide, at which point the optical mode will exist substantially in the second waveguides. Thermal oxidation (as an
example of a surface etch process) will still affect the waveguides, but no concave structures are needed and hence the structure can undergo oxidation without the above difficulties. Indeed, as mentioned above, oxidation will sharpen the tips of the (three) tapers and will generally improve the definition of the branch structure, and may be done deliberately. The waveguides can be of the normal height at all points and hence, for example, the fabrication difficulties of Hu et al are avoided.
It is preferred that the tapers of each waveguide are symmetric. However, depending on the optical properties required, one or more may be asymmetric. Likewise, it is preferred that the start of the second waveguides are substantially aligned with the start of the tapering off of the first waveguide, and also that the end of the first waveguide is substantially aligned with the end of the taper of the second waveguides. Again, however, depending on the optical properties required, this may not be the case.
This structure of a waveguide branch can be employed in a variety of applications, such as a Mach-Zehnder interferometer, a switch, or any application where the waveguide must divide. It can also be used to equivalent effect where waveguides must combine.
It is also possible to use the structure near the edge of a chip or a region thereof to assist in coupling to (for example) the active region of an arrayed waveguide grating. In such situations, the waveguide is split into two short waveguides which are parallel and closely adjacent; these end at the edge of the chip or region. The light leaving the pair will act as a conjoined mode if the waveguides are close enough, the conjoined mode being wider. In some applications, such as band flattening in arrayed waveguide gratings, this is desirable.
Embodiments of the present invention will now be described, by way of example, with reference to the accompanying drawings, in which;
Figure 1, already described, shows a typical Y-branch;
Figure 2 shows a plan view of a Y-branch according to the present invention;
Figures 3, 4 and 5 are sections on ffl-m, IN-IV and N-N, respectively, on Figure 2;
Figures 6a, 6b and 6c show computed transverse mode profiles for the Y- branch of Figure 2 at locations A, B and C (Figure 2) along the structure;
Figure 7 shows, schematically, a Mach-Zehnder interferometer employing the present invention;
Figure 8 shows, schematically, a 1x2 switch employing the present invention;
Figure 9 shows, schematically, a combiner employing the present invention;
Figure 10 shows, schematically, a 4 way divider employing the present invention;
Figure 11 shows a band flattener employing the present invention; and
Figure 12 shows a possible layout for the structure of Figure 11 in an arrayed waveguide grating.
Figure 1 is described above and will not therefore be described here.
Figure 2 shows a Y-branch according to the invention. An incoming waveguide 100 conveys an optical mode which is to be split so as to propagate in two waveguides. In the following, it is assumed that the mode is to be split equally and accordingly the described embodiments will be symmetric. Where this assumption does not hold, appropriate adjustment of the relative dimensions of the waveguides will be required.
At the point where the Y-branch is to take place, the incoming waveguide 100 is formed in a taper 102. Thus, the side walls 106 of the incoming waveguide 100 approach so as to reduce the lateral width of the waveguide 100, eventually to a minimum width at point 108.
Adjacent the taper 102, first and second outgoing waveguides 110, 120 start. The starting point of each waveguide is again formed in a taper, 112 and 122 respectively. The tip 114 of the first outgoing waveguide 110 is approximately adjacent the start of the taper 102 of the incoming waveguide 100, and likewise the tip 124 of the second outgoing waveguide 120. The end of the tapers 112, 122 are approximately adjacent the tip 108 of the taper 102 of the incoming waveguide 100.
The tapers 112, 122 of the outgoing waveguides 110, 120 are located on either side of the taper 102 of the incoming waveguide, closely adjacent. A suitable spacing is between 0.25 and 2μm, which, although achievable using known photolithographic techniques, is produced in accordance with the invention by photolithographic techniques followed by an oxidation step which increases the spacing preferably by about 0.3μm. Thus, at the start of the outgoing waveguides 110, 120, the evanescent part of the optical mode will exert an influence. As the incoming waveguide narrows, the optical mode will gradually be forced out of the incoming waveguide into the underlying substrate. The proximity of the outgoing waveguides then allows the mode to become associated with them, continuing to propagate under their guidance after the incoming waveguide 100 ceases.
Figures 3 to 5 show cross-sections of the structure. Initially, as in Figure 3, only the incoming waveguide 100 is present. Figure 4 shows the cross-section shortly after the incoming waveguide 100 begins to taper; the outgoing waveguides 110, 120 are present but narrow. Figure 5 shows the cross-section after the Y-branch, the incoming waveguide having ceased and the two outgoing waveguides 110, 120 running parallel.
Before the surface etch process, the basic waveguide branch structure as shown in figures 2 to 5 is already present. However, the surface etch process improves the definition or resolution of the branch structure. Preferably the surface etch process comprises oxidation of the waveguides, which are preferably formed from silicon. The oxidation of the waveguides preferably creates a growth of oxide on the surfaces of the waveguides (whether or not a surface oxide layer is already present). The amount of oxide layer growth due to the oxidation step (post waveguide formation) is preferably at least 0.1 μm, more preferably 0.2μm, even more preferably at least 0.3μm, for example approximately 0.35μm. The growth of oxide is achieved by "consumption" of the waveguide material (e.g. silicon) - i.e. by the oxygen chemically reacting with the waveguide material. Consequently, the amount by which the oxide layer grows normally will be substantially the same as the amount by which a surface layer of each waveguide is diminished. Therefore, the rib portion of each waveguide preferably decreases in cross-section (by the oxidation process) by a surface layer of the waveguide disappearing, the surface layer having a thickness of preferably at least 0.1 μm, more preferably at least 0.2μm, even more preferably at least 0.3μm, for example approximately 0.35μm.
Figures 6a, 6b and 6b show computed transverse mode profiles at points A, B and C on Figure 2 respectively. In Figure 6a, an optical mode of the usual type for an Si rib waveguide is shown, existing partly in the rib but mainly in the slab region below and on either side. As the incoming waveguide 100 narrows, the optical mode shifts downward slightly deeper into the slab region, from which it also begins to extend upward into the adjacent outgoing waveguides 110, 120. As the incoming waveguide continues to narrow and the outgoing waveguides 110, 120 become more substantial (Figure 6c), the mode begins to divide - becoming bimodal - and occupies the outgoing waveguides more fully. It can be seen in Figure 6c that the remaining optical power in the incoming waveguide 100 is small. Accordingly, losses caused in the known structure of Figure 1 by the disturbance to the central axis of the mode should in this arrangement be minimal.
The above description has been made in relation to a division of a propagating mode into two separate modes. However, the process is capable of operating in reverse using the same structure, allowing the combination of two propagating modes into a single waveguide. Each will be forced out of the tapering waveguides and will spread into the same adjacent developing waveguide.
Figure 7 shows a Mach-Zehnder interferometer (MZI) constructed using the Y-branch described above. The dimensions of the Y-branch compared to the dimensions of the MZI are small, so in Figure 7 (as with Figures 8-10) the Y branch has been illustrated schematically. An incoming waveguide 130 splits at a Y-branch 132 into two parallel waveguides 134, 136. One waveguide is subjected to a perturbation by device 138 which is able to delay the signal slightly. Suitable devices include pn junctions, where the injected carriers affect the refractive index and locally change the light velocity. Both waveguides then recombine at a further Y-junction 140 and continue as an outgoing waveguide 142. The optical modes will interfere due to the delay to which one was subjected; this means that wavelengths where this delay results in a λ/2 shift will be eliminated. As the number of injected carriers (and hence the refractive index change) can be varied by varying the current, this means that the wavelengths to be accepted or rejected can be selected.
The MZI structure is known, but through the present invention it can be made less lossy. This in turn permits more processing to be carried out on the light signal.
Figure 8 shows a 1x2 switch - a digital optical switch or DOS switch. An incoming waveguide 150 divides at a Y-branch 152 into two outgoing waveguides 154, 156. Close to the branch, one waveguide is perturbed by a device 158 similar to the device 138 of Figure 7, changing the index of that waveguide. As both of the waveguide branches are still quite close at this point, the branches interact. If the structure is suitably designed the light will tend to gradually migrate to the branch with the higher index as the branches separate. When the branches are sufficiently far apart most of light will be localised is the branch with the higher index.
Figure 9 shows a combiner, as referred to above and employed in the MZI of Figure 7. A pair of incoming waveguides 160, 162 meet at a Y-junction 164 and transfer their associated optical mode to an outgoing waveguide 166 in the manner described above.
Figure 10 shows a 4-way divider. An incoming waveguide 170 branches at a Y-junction 172 into first and second intermediate waveguides 174, 176. The first intermediate waveguide 174 then branches again at a Y-junction 178 into first and second outgoing waveguides 180, 182. The second intermediate waveguide 176 also branches at a Y-junction 184 into third and fourth outgoing waveguides 186, 188. In this way an incoming waveguide can be branched into 2" outgoing waveguides. It is also possible to arrange any number of outgoing waveguides, not necessarily 2", but the intensity in the outgoing waveguides will not then be identical.
Figure 11 shows a band flattener. A waveguide 200 is approaching an edge 202, which may be the edge of the substrate or the edge of an arbitrary region within a substrate. Near the edge of the substrate, for example, the waveguide may need to couple into an optical fibre. In an arrayed waveguide grating (AWG), a waveguide may need to end at the edge of a region within the substrate, so as to allow the optical mode to disperse and be coupled into a plurality of waveguides. Before the edge, the waveguide 200 divides in a Y-branch structure 204, as those described previously, into two stub waveguides 206, 208. These extend parallel and close alongside each other for the short remaining distance to the edge 202 where they end.
At the edge, the proximity of the two stub waveguides 206, 208 together with their short length means that the optical modes in each will still be associated. Thus, when they end, a single conjoined mode 210 will be released into the substrate. This mode will have different properties to the mode which would be released from a single waveguide; in particular it will give rise to greater band flattening which may be an advantage in many AWG designs.
Rib waveguides such as those illustrated above are typically formed (in practice) by etching trenches either side of the intended track of the waveguides so as to leave an upstanding rib. Figure 12 shows how the design of Figure 11 could be achieved for an AWG bearing this in mind. The waveguide 200 is defined by etched areas 212, 214 on either side. The edge 202 is defined by the end of the areas 212, 214, and accordingly lies at the end of the waveguide 200. Near the edge, the areas 212, 214 narrow to allow for the stub waveguides 206, 208. A pair of narrow trenches 216, 218 extend from the areas 212, 214 to cut off the waveguide 200, each extending at an angle to end the waveguide 200 with the necessary taper. They then join as trench 220, ending about at the edge 202, and separating the stub waveguides 206, 208.
In this way, the required structure is achieved. The stub waveguides cease to be defined after the edge 202 and the optical mode that they carry is released into the substrate.
It will of course be appreciated that many variations may be made to the above described embodiments without departing from the scope of the present invention. Some such variations are described above; others will be apparent to the skilled person. For example, no other layers are shown in the devices, but these could be provided for unrelated reasons or to provide a level surface. Any low index material such as SiO2 could fill the gaps between the ribs without loss of optical performance. More than two branches could be formed, by providing further adjacent tapered waveguides.