GRATINGS ON PLANAR LIGHTWAVE CIRCUITS
Field of the Invention
Bragg Gratings (BG) are commonly formed within the core of optical fibers, which are usually fused doped silica materials. This so-called Fiber Bragg Grating (FBG) is actually a section of the whole fiber that has a longitudinal periodic variation in refractive index (n). When a multi-wavelength light source propagates through a FBG, the grating can reflect specific wavelengths defined by the period of the refractive index change. The present invention relates to a new technique for accurate fabrication of distributed gratings in waveguide structures, particularly planar waveguide structures.
Background to the Invention
Fiber Bragg Gratings (FBG) are produced by exposing the doped silica core to a high-energy illumination source, e.g. the UV light pulses generated from an Excimer Laser (KrF [248nm], ArF [193nm]). The absorption causes a permanent change in the n of the fused silica. The ease of this irreversible transformation depends on the power of the light pulse and the photosensitivity of the fused silica, which is largely affected by the dopant type. To invoke the fiber core to 'transform', i.e., change its n upon light irradiation, the fiber core has to undergo special treatment to make it 'photosensitize'. This used to be accomplished by increasing the germania (or germanium) content in the glass ingot or by hydrogenation. However, the small cores of the high-germania makes them not compatible with standard communication fibers, and the process of hydrogenation not only adds complexity, risk and cost but also tends to produce high-loss gratings after irradiation. Recently, the boron-doped core (Fibercore Limited's PS1250/1500) has become more popular since the produced core itself is intrinsically photosensitive. This allows gratings to be written onto the fiber core without the hydrogenation step. Currently, there are two industrial methodologies to produce FBGs:
Interferometric method: As illustrated in Figure 1(a), a single laser beam is split into two beams, which interfere at the fiber and generate a periodic pattern. Interference fringe is very sensitive to optical alignment and mechanical vibrations. The UV source must have a long temporal and spatial coherence. But this method is highly flexible for creating various types of gratings.
Phase-Mask method: As illustrated in Figure 1(b), an UV laser beam passes a diffraction grating, the interference pattern between the -1st and +1st order
diffraction generates the required intensity pattern at the fiber. This method is very stable against misalignment and vibrations and can be easily automated. Therefore, the Phase Mask writing method is perfectly suited for mass production. The UV source should display a reasonable spatial coherence length. Figure 1 (c) illustrates a schematic of a FBG.
The key application of FBGs in modern optical telecommunication is Wavelength Division Multiplexing (WDM). Owing to their flexibility (performance varies with either change in grating periods or refractive index of periods, or both), FBGs are thus employed in many WDM components, for example, (a) to multiplex/demultiplex numerous signal wavelengths,
(b) to add/drop single channels,
(c) for dispersion compensation (chirped gratings),
(d) or for stabilizing pump lasers.
The filtering characteristics of FBG satisfies the demand of narrow bandwidth, and high isolation of > 30 dB simultaneously. The attractiveness of putting BG on optical waveguide is not confined only to fibers, but to planar waveguides as well, e.g., Silica on Silicon (SoS) waveguides.
The idea again is to create periodic variations in the n of the planar waveguide. There are a few ways to achieve this condition. One way is to perform optical lithography and dielectric etching into the waveguide to create vertical cavities, which repeats after certain interval length, along a portion of the waveguide. This is illustrated in Figure 2 which shows the process sequence of etching gratings inside the waveguide. The symbols nw, naιr, Ltotai, and Lp represent the refractive index of waveguide material, refractive index of air, the length of the grating section and the grating period.
The other is to make use of special inorganic polymers, instead of silica, that are capable of altering their refractive index across a certain range via photon excitation process (Redfern Polymer Optics' Inorganic Polymer Glass) to make waveguides. By the similar phase mask technique as described above, gratings can be written directly into the inorganic polymer-waveguides (interference Patterning
(IP)).
In a Bragg grating, the Bragg reflected wavelength (λBragg) is defined as: λβragg = 2.A7e/7-Λeχpeoted (1)
whereby neft = the effective mode refractive index in FBG or waveguide; and Λexpected = the expected grating period
From equation (1), it is obvious that the selection of which wavelength to be reflected is controlled by the value of netr and Λexpeoted. However, owing to variations in process conditions, it is very difficult to get precise neff and Λexpected. The process- attainable nefτ * and Λβχpeoted * are often, if not always, different from our desired nβfr and Λe peoted- Table 1 summarizes some effecting factors from acquiring the precise nΘff and ΛeXpected in silica and inorganic polymer waveguides.
Table 1 : List of some factors responsible for netf and Ltotai variations in etched gratings and interference patterned (IP) gratings on a tunable waveguide (or fiber).
For a tunable waveguide (or fiber) filter, these effecting factors short-change its functionality and performance, but most important of all, undermines its reliability. Figure 3 illustrates the four most important characteristics required from a tunable waveguide (or fiber) filter affected by the effecting factors described in Table 1.
The current process of writing gratings on waveguide or fibers inevitably faces one or more of those issues as described in Figure 3. One culprit for such problems is the netr variation. The inhomogeneity of the UV-illumination source, and the dopant distribution within the fiber are the likely causes for netr variation in FBG. The inhomogeneity of the UV-illumination source, and the different degree of refractive index alterations incurred in the inorganic polymers will be the susceptible causes for nβfτ variation in the polymeric waveguides. There would be no additional leeway on attempting to minimize the variation in neff by repeated UV exposures, i.e., the neff variation still stay. Henceforth, in the case of FBGs, to acquire the correct wavelength to transmit through the UV-irradiated fiber, some fibers after being written with gratings, are pulled longitudinally at one end. This changes the Ltotai, which then indirectly alters the Λeχpeoted (since ΛeχPe0ted is proportional to Ltotai)-
Alternatively, other methods such as inducing an expansion/contraction of the grating length to incur the much-wanted change in ΛeχPected by thermal heating/cooling. In an empirical form, the aforesaid relations can be mathematically defined as follows:
netf* (Euv, fp, Uuv) - neff (Euv, fp, Uuv) + Δn (Euv, fp, Uuv) (4a)
whereby both netr and netf* values are affected by the energy of the UV- source (Euv), the frequency of the UV pulses (fp)and the uniformity of the UV-source (Uuv)- The symbol Δn represents the change in effective refractive index.
Uotai* (Dg) = Lwai (Dg) + ΔL (Dg) (4b)
whereby both Ltotai and Ltolaι* values are affected by the dimensions of the etched or fringe-pattern gratings (Dg). For etched gratings, Dg will be affected by lithography and etching. For fringe-pattern gratings, the grating dimensions on the phase mask will affect Dg. The symbol ΔL represents the variation in total grating length.
Uotai* (Dg, Ta) = iai (Dg, Ta) + ΔL (Dg, Ta) (4c)
whereby both Ltotaι and Ltota* values are affected by the dimensions of the gratings and applied process temperature (Ta).
Uotai* (Dgι Pa) = Ltaw (Dg, Pa) + ΔL (Dg, Pa) (4d)
whereby both Llolaι and Ltota* values are affected by the dimensions of the gratings and applied pressure for longitudinal pulling (Pa).
Under normal process conditions, λβragg* = 2. neff* (Euv, fp, Uuv)-( Ltotai* (Dg) / Ntotal) ≠ λβragg (5a)
However, it is plausible to get equality if we impose an external stress factor like temperature, or pressure or both:
λβragg* = 2. neff* (Euv, fp, Uuv)-( Ltotai* (Dg, Ta) / Ntotal) = λβragg (5b) λβragg* = 2. neff* (Euv, fp, Uuv)-( Uotai* (Dg, Pa) / Ntotal) = λβragg (5c)
λβragg* = 2. neff* (Euv, fp, Uuv)-( Ltotai* (Dg, Ta, Pa) / Ntøtal) = λβragg (5d)
For the case of waveguides, the whole tuning process is even more complicated. Some other forms of applying stress / strain loading to the waveguide would be necessary. This could involve serious bulk micromachining, or surface micromaching procedures be done on the platform whereby the waveguides reside. The heating and cooling of the fibers to effect the change in Λeχpected can also work for the waveguides. The major advantage of having gratings on the fiber or on the waveguide is to allow us to use only the wavelengths that are reflected back by the period of the refractive index change in the fiber or waveguide. In this a way, the availability of grating, like for the case of Bragg gratings, allows us to enjoy the option of wavelength tuning amongst many others, on our waveguides and fibers. Under ideal circumstances, the reflected Bragg wavelength(s) from the
FBGs, or the written waveguides should have a narrow bandwidth and a central peak position that we specifically desired. The reason for this is that we can have this exact and known signal wavelength channel to transmit into another waveguide or fiber after it has been gain-amplified. The aforesaid bandwidth actually refers to the Full Width Half Maximum (FWHM) bandwidth as shown in Figure 4. This requirement would imply very precise machining of the absolute gratings. In other words, any physical variation amongst the produced gratings would be forbidden. In reality this will be impossible. Variations from process conditions, which translate to physical variations in gratings, are inevitable. Note the Bragg reflected wavelength (λBragg) defined in equation (1) is:
λβragg = 2.πeff.Λeχpected
whereby neff = the effective mode refractive index in the waveguide, which is affected by the dimensions (dry etching) of the generated gratings; and Λexpected = the expected grating period, which is affected by the phase mask grating period, and lithography;
A continuum optical source emits lights with many different wavelengths, and each wavelength occupies some bandwidth. The continuity arises when the wavelengths overlap with its proximity neighbors. An optical source placed at the input end of the waveguide would therefore imply many continuous waveforms, spanning across a lambda (λ) range, are entering into the waveguide
simultaneously. These many waveforms will comprise of those that we like to reflect back (Bragg wavelengths), and those that we like to transmit on. If we have gratings etched (or 'written') on the waveguide in the manner as described by equation (1), ' wavelengths satisfying the conditions depicted by equation (1) will be reflected, and those that do not, will be transmitted. As shown in Figure 4, a small percentage of input light (~ 0.01%) is reflected from each grating face. Identical grating (with same period) will reflect light with similar phase that can constructively interfere.
Each grating contributes to the Bragg reflection, and the Bragg wavelengths reflected by the gratings takes a sine function since the gratings are written with a certain period across a section of the waveguide. The design and the dimensions of the gratings determine the amount of Bragg reflection back. For typical gratings, each grating will reflect approximately 0.01% of the Bragg wavelength (see Figure 4). If the gratings are etched deeper into the waveguide, the level of reflection will be higher. When the gratings are made totally identical and written accurately across a waveguide section in a periodic momentum, the reflected lights will be of identical phase, and constructively interfere to produce a Bragg wavelength peak with a narrow bandwidth. The bandwidth mentioned over here is the Full Width Half Maximum (FWHM) bandwidth that can be seen in Figure 4.
In addition, the amount of Bragg back reflection will be directly proportional to the total number of grating periods written across the media section. In other words, if each grating is made to reflect 0.01% of the input light, y number of gratings will reflect (0.01 x y)% of the input light.
Total Bragg reflection (%) = 10"2.y (6)
In addition, owing to the constructive interference of the back reflected Bragg wavelengths, the FWHM bandwidth tends to be smaller with greater number of gratings.
FWHM I, y t (7)
Therefore, if we have 1000 gratings, we can have around 10% reflection of the Bragg wavelength. We can also attain narrower bandwidths with increasing number of periods since these periodic reflected waves interfere constructively. As mentioned previously, the design and the dimensions of the gratings can change the amount of back Bragg reflection. When we etched the gratings deeper to make each of them to reflect 0.02% of the input light, then according to equation (6), if we still
have 1000 gratings, we will have almost 20% Bragg reflection. Even though neff varys with this new grating dimension, the impact is usually mild enough for the FWHM bandwidth to remain constant.
From equation (6), it also indicates that if for those etched gratings that can each reflect 0.02% of the Bragg wavelength to produce the same amount of 10% reflection as in the previous case for 1000 shallow gratings, we will only need 500 of such deeper gratings. In other words, the same level of functionality achievable by a long section of shallow gratings is now attainable by a shorter section of deeper gratings. This attractive feature grants us the liberty of producing the same amount of Bragg reflection from either long or short waveguides. However, it has to be reminded that the smaller number of gratings comes with a broader FWHM bandwidth cost, (see equation (7)).
In reality, the fabricated gratings are usually physically similar and not identical. Henceforth, the reflected lights posses phases only similar to each other, but not identical. Upon superimposition, these reflected lights will generate a Bragg wavelength peak having a broader bandwidth. The physical variations within the fabricated gratings can also account for the shift in its peak position, and the appearance of other Bragg reflected wavelengths that barely fulfill equation (1) conditions. Collectively, these are few of the derived consequences when we have Δneff, and ΔΛeχpeotθd ≠ 0. Δneff, and ΔΛexpeo.ed respectively refer to the variations in effective mode index of waveguide and variations in the grating period that are made non-zero terms by the unavoidable process variations. When we have nonzero Δnθff, and ΔΛeχpeoted terms, we will have non-zero ΔλBragg term. This then translates to a shift in the position of the reflected Bragg wavelength. Henceforth, better processes that produce minimum Δneff, and ΔΛeχpected need to be developed.
Process variations can result in some gratings, being etched deeper than others, into a silicon dioxide planar waveguide. Such dissimilarities between the gratings lead to non-zero Δneff, and ΔΛeχpected terms. Figure 5 presents schematically a planar waveguide with periodic gratings written on it. The gratings are usually etched a depth of few tens of nm. The gratings are not exactly etched to the same depth and width, which give rise to small non-zero Δneff, and expeoted terms. Lets first assume that the gratings exist at the correct frequency, i.e., ΔΛeXpected = 0, and make Δneff our primary subject of interest in our subsequent discussion.
Then from equation (1), we then empirically formulate the following:
Δλβragg OC Δneff (8)
Both the refractive index variations in the waveguide core material, Δncore, and fabricated gratings, Δngratiπg, contribute to Δneff, i.e.,
Δneff α (Δn0ore + Δngrating), (9)
and both the values for both Δncore and Δngrating are in the nominal order of around 10" . If the desired Bragg wavelength has a central peak position at 1550nm, and Δneff happens to be in the order of 10'4, the Bragg wavelength will register a central position between 1549.85 nm and 1550.15nm (@ 1550 ± 0.15nm).
For an operational ITU transmitter working @ 50GHz, the central peak wavelength will be around 1550 ± 0.40nm. A Distributed Feedback Laser (DFB) laser has the ability to tune its wavelength by temperature control - a Δλ shift of 0.1 nm per 1 degree Celsius change (+ 0.1 nm for -1°C, and - 0.1 nm for +1°C). Therefore, if a 1550nm DBR laser is to be employed as an ITU transmitter working @ 50GHz, a ± 3 °C temperature control on the device will just be necessary. Nevertheless, the coupling loss of the laser optical mode into a Single Mode Fiber (SMF) would be great.
If a planar silica waveguide is engaged to ensue better coupling of laser light into a fiber, then we need to ascertain that the Bragg gratings are etched onto the dielectric waveguide in a specific manner, which only reflects the 1550nm wavelength. In most cases, the reflected wavelength position should not be too far off from the desired central position only if the gratings are done correctly. If we are keen on fine-tuning the wavelength position, we will have to apply similar thermal control techniques to the planar waveguide. This step, nevertheless, poses the greatest challenge since many glass or glass-like composite waveguides only show a Δλ shift of 0.01 nm or less per 1 degree Celsius change. As a consequence, the waveguide might need to be heated up or cooled down, so as to raise or lower its temperature by some 40°C. Most commonly, a heater (e.g. heater wire) is placed beneath the planar waveguide to heat it up. As for cooling, forced convention of cold air is adopted. However, these sources for effecting temperature variations to the waveguides could be ineffective. A micro-heater might not be able to provide that huge amount of heat, while the forced convention volume of cold air might be too small to remove dissipated heat efficiently. In this aspect, greater emphasis will be placed on writing and positioning the gratings accurately on the waveguide. Nonetheless, the precision of the grating dimension and position will be limited by
the resolution and capability of the machining tools. In short, either we compromise with the large coupling losses incurred when using a DFB laser, or we have to live with the de-centered peak wavelength offered by the waveguide with physical gratings.
Brief Description of the Drawings
Examples of the present invention will now be described with reference to the accompanying drawings in which:-
Figure 1 shows known methods of forming a Bragg grating within an optical fiber;
Figure 2 illustrates the fabrication of a grating structure in a waveguide by a lithography and etching process;
Figure 3 highlights the problems associated with known techniques for grating fabrication in waveguides; Figure 4 illustrates the monitoring of the optical spectra of the transmitted and reflected signal from a Bragg grating;
Figure 5 illustrates the typical dimensions of a periodic grating structure etched in a planar waveguide; and,
Figure 6 shows the process flow for fabricating gratings in a waveguide according to the present invention.
Detailed Description
The fundamental factors which undermine the wavelength tuning capability of the waveguides are:- (1) the variations in the refractive index of the waveguide core material,
(2) the variations in the periodic gratings' dimensions and etched depth, and,
(3) the precision of the gratings' location.
The present invention seeks to address these problems so as to afford a simple but reliable manufacturing process to write gratings on waveguides. Figure 6 illustrates a process flow for fabricating a distributed grating in or on a waveguide in accordance with the present invention.
As shown, the present invention proposes the application of athermal film materials to build the top portion of the waveguide. Reason being that these materials offers excellent resistance against material aging. Examples of some athermal film materials are inorganic polymers, glass or glass-composites such as BSG, PSG or BPSG. Some of these materials can be spun-on, which improves thickness uniformity, surface smoothness and process cost. The refractive index of
the employed athermal film has to be similar to the bottom waveguide material, and the thickness of the athermal films will also have to be thicker than the etched depth of the intended gratings. The athermal film can be the whole waveguide layer, if possible. In addition to the athermal films, the invention also allows for other thermal materials such as chemically vapor deposited dielectric films.
The athermal materials will be doped with boron, or germania, or boron and germania. The encompassed dopant concentration will depend on what level of photosensitivity is required for the waveguides. The doping step can be incorporated during the concoction stage (for the spun-on materials), or during the deposition step.
The versatile electron beam (E-beam) writing/masking technique is used to create the resist-grating mask. Different grating sections of different periodic variations can be produced onto the waveguides with high accuracy and lower cost, since the phase mask is done away with. The E-beam writing/masking technique offers flexibility and maximum resolution.
Physical gratings (permanent) are plasma etched into the waveguide materials. End-point detection (EPD) technique can be implemented to accurately control the depth of the etched gratings that is required.
The mask used to define the waveguide dimensions is then aligned to the same set of E-beam markers used to define the E-beam gratings. This is to ensure that the gratings reside directly onto the waveguide.
Etching gas chemistries are properly selected to anisotropically etch the waveguide. Several possible combinations are CHFa/CF Ar/O∑, CI2/Ar/O2 and CI2/BCl3/Ar/O2. EPD can again be employed during the waveguide etching. Subsequently, the E-beam resist can be removed. The availability of these etched gratings already offers a wavelength tuning option, if the waveguide is to offer final wavelength-tuning properties.
A light is then guided into the written waveguide, and the Bragg wavelength spectrum is monitored. If the wavelength λBragg* position deviates from the desired central location, the fine wavelength position shifting may be performed via UV- tuning. The energy of the UV-source (generated by a excimer laser) can be systematically altered to induce gradual dopants activation within the waveguide. The nΘff* magnitude will change with the level of dopant activation. Consequently, λβragg* can be tuned until it equates to λBragg by continuously monitoring the Bragg wavelength spectrum throughout the UV-excitation process. A setup is needed to perform this delicate tuning procedure.
With reference to Table 1, we can minimize the impact of the factors which affect the characteristics of the grating in a negative way by employing better techniques and more stable materials. The present invention mitigates the unwanted effects by careful selection of process techniques and materials as shown in Table 2.
Table 2
Thus, important features of the present invention are:-
(i) E-beam lithography is used to write the gratings. The resolution and versatility offered by the E-beam writing technique is higher and better than optical lithography, i.e., tighter grating periods, and no need of chrome or phase masks.
The value of ΔL incurred from patterning the whole grating section via E-beam lithography is smaller than those obtained from conventional lithography. With smaller ΔL, the ΔΛeχpecte will be small as well.
(ii) E-beam lithography allows several types of resist-grating mask with different periodic variations can be patterned onto the same waveguide. In this way, chirped gratings can be made, or waveguides that can offer optical isolation. (iii) End-point detection technique is employed during grating etch to determine accurately the etch duration which gives the best grating depth uniformity, (iv) Athermal film materials are used to build the waveguides. The conventional film materials suffer a thermal shift of 0.1nm/°C, whilst the athermal films only register a thermal shift of 0.01nm/°C. In terms of temperature variations during waveguide operation, the athermal waveguides will provide more superior performance. In addition, the athermal waveguides are more resistant to thermal aging, so as to achieve consistent quality performance.
(v) An additional built-in filter is included beneath the UV-source to homogenize the UV-power (intensity).
(vi) MOST IMPORTANT: Precise wavelength tuning mechanism is achieved via a two-step process. E-beam gratings are engaged for coarse wavelength tuning, and UV-induced boron (or others) dopant activation for fine wavelength tuning. There is no need of external loading to change Ltotai*. In addition, any Λexpected variations we received from E-beam writing, and dry etching, we can compensate by varying Δneff. Hence, we can consistently achieve precise Bragg wavelength position with high precision. The whole concept and sequence of the wavelength tuning process is different from conventional techniques. We do not heat or cool the Waveguides to high or low temperatures so as to incur wavelength shift, and it will also be quite tedious to do so since these waveguides have such low thermal shifts - 0.01nm/°C.
(vii) The whole setup for in-situ monitoring the output wavelength while simultaneously, performing gradual dopant activation via UV-illumination. The UV excitation for waveguide is terminated when the monitored grating wavelength reaches the desired precision.
(viii) The versatility of this general technique to produce other planar lightwave circuits, e.g., spot size converter stack, with much ease, better performance and more functionality.