Tunable spectral slicing filter utilizing sparse grating in Ti:LiNbO3 Renato C. Rabelo*a, Ohannes Eknoyanb and Henry F. Taylorb a Instituto de Estudos Avançados (IEAv), Rod. Tamoios km 5,5, S. J. Campos, SP Brazil 12228-001; b Deptartment of Electrical and Computer Engineering, Texas A&M University, College Station, TX, USA 77843-3128 ABSTRACT A new type of tunable guided-wave spectral slicing filter at the 1530nm wavelength regime is reported. The design allows the selection of equally spaced frequency channels and simultaneously produces nulls that are equally spaced between the selected channels. This makes it attractive for minimizing crosstalk in dense wavelength division multiplexing (DWDM) applications. The spectral selection of the filter is based on co-directional polarization coupling between transverse electric (TE) and transverse magnetic (TM) orthogonal modes in a waveguide by means of a static strain induced index grating. An etalon-like response results from the sparse arrangement of the grating sections as N individual coupling regions in tandem with equal spacing between their centers, yielding N-1 equally spaced nulls between adjacent selected frequencies. Adjustments of the resulting filtering function may be obtained by proper choice of coupling regions’ lengths and spacing. Devices were fabricated using single mode channel waveguides formed by Ti diffusion on x-cut y-propagating LiNbO3 substrates. Static strain from a periodically delineated surface film was used for making N = 6 polarization coupling regions. Electrode patterns centered about the optical waveguide and defined by liftoff were used to tune the filter electrooptically. Experimental results are in good agreement with design theory. Keywords: Integrated optics, Lithium niobate, Tunable filters, Sparse gratings.
1. INTRODUCTION A DWDM optical network demands devices capable of filtering individual channels respecting strictly defined pass bands and stop bands. Among the optical tunable filters, those with smallest response time to a tuning excitation are the ones attracting more interest because they would enable a future dynamically varying DWDM network, in which packet switching and wavelength routing will be routine operations. The electro-optically tunable filter places as one of the fastest response time devices and there is still ongoing research on devices of that category as well as on the competing technologies aimed at satisfying DWDM requirements. Research activities at Texas A&M University have demonstrated electro-optic tunable filters employing Ti-diffused waveguides in lithium niobate substrate (Ti:LiNbO3) [1]. These devices perform optical filtering by using co-directional TE-TM mode conversion. The coupling between these two modes is accomplished by using a phasematched straininduced overlay grating of a silica (SiO2), deposited at high temperature and patterned at room temperature. The difference in SiO2 and LiNbO3 thermal expansion coefficients builds up a strain field that generates a periodic refractive index grating in the waveguide through the strain-optic effect. By properly choosing the SiO2 film deposition parameters as well as waveguide fabrication parameters and grating period, the polarization conversion is made efficient. When uniform gratings are used, the polarization conversion efficiency has a well known spectral response and even though the resulting filtering function is periodic, it is not with respect to the desired parameter (frequency). The filtering function periodicity in the frequency domain would translate into presenting equally spaced nulls and maxima through the spectrum of interest, making it easier to define channel regions in the spectrum and avoid crosstalk between those. In order to circumvent the uniform grating characteristic of unequally spaced nulls and maxima, a new topology of sparse strain gratings was proposed [2]. It is capable of placing the nulls of the spectrum at well known, equally spaced positions, defining channel spectral regions in the stop band and pass band of the filter. This paper reports the first realization of spectral slicing filters that uses sparse strain gratings on Titanium difused Lithium niobate waveguides (Ti:LiNbO3). *
[email protected]; phone 55 12 3947-5460; fax 55 12 3944-1177
Integrated Optics: Devices, Materials, and Technologies XIII, edited by Jean-Emmanuel Broquin, Christoph M. Greiner, Proc. of SPIE Vol. 7218, 72180F · © 2009 SPIE · CCC code: 0277-786X/09/$18 · doi: 10.1117/12.808659
Proc. of SPIE Vol. 7218 72180F-1
2. THE UNIFORM GRATING CASE Uniform grating electro-optically tunable filters have been demonstrated on LiNbO3. In this type of device, a ypropagating channel waveguide is fabricated by titanium diffusion on a x-cut LiNbO3 substrate with parameters controlled in a way it is single mode for both polarizations (TE and TM). On top of the waveguide and along the propagation direction a periodic structure is delineated such it is shown in figure 1.
Ti diffused Channel Waveguide
x Λ SiO2
SiO2
SiO2
SiO2 LiNbO3
y Fig. 1. Schematic diagram (sideview) of a TE-TM polarization converter using a continuous SiO2 strain-inducing grating on a x-cut, y-propagating Ti-diffused LiNbO3 waveguide.
In the absence of the grating, TE and TM modes propagate without exchanging power because of the mode orthogonality property. In the presence of the strain-inducing periodic structure of a convenient spatial period Λ (grating) power exchange (coupling) between the two modes is made efficient. The induced strain arises from the fact that the SiO2 film is deposited on a LiNbO3 substrate, previously heated and kept at an elevated constant temperature during deposition. As the sample is cooled down to room temperature, due to the expansion coefficient mismatch between the two materials a strain field builds up at their interface. The strain propagates into the substrate and waveguide; via the strain-optic effect produces a change in one of the components of the dielectric permittivity tensor that will enable coupling between the two modes with orthogonal polarizations. After cooling down, the silicon dioxide film is patterned by a photolithographic process and then etched at room temperature to define the overlaid structure with spatial period Λ on the waveguide. This results in the periodic variation on the appropriate shear strain component, therefore on the appropriate impermeability tensor component needed to couple the two polarization modes on the waveguide. Polarization conversion under these conditions can be analyzed with the help of co-propagating coupled-mode equations expressed by
dA = − j κ B e jΔy dy dB = − j κ A e − jΔy dy
(1)
where A is the field amplitude for one of the polarization modes and B for the other polarization mode, κ is the coupling 2π is the so-called phase mismatch factor, β TE and β TM are coefficient between the two modes, Δ = β −β ± TE TM Λ the propagation constants for the TE and TM modes, respectively and Λ is the grating spatial period.
Solutions to the set of equations (1) for κ constant along the y direction are known [3]. If it is further assumed that light is only in one polarization state at the grating input (A(0)=1 and B(0)=0) the fraction of optical power coupled to the polarization orthogonal to the input polarization state after propagating through a length L of the grating, often called 2 2 polarization conversion efficiency, will be given by PCE = B ( L ) = sin(κL ) . A typical PCE plot, for a uniform grating on Ti:LiNbO3 waveguide, as a function of frequency detuning with respect to the frequency at which the conversion is maximum is presented in figure 2. For this plot, a coupling coefficient value of κ = 1.496 cm-1 and an interaction length of 1.05 cm were assumed.
Proc. of SPIE Vol. 7218 72180F-2
Polarization Conversion Efficiency
1.0
0.8
0.6
0.4
0.2
0.0 -12
-8
-4
0
4
8
12
Normalized Frequency ((ν−ν0) x 100 GHz)
Fig. 2. Typical polarization conversion efficiency (PCE) plot for a uniform strain-inducing grating on Ti:LiNbO3 waveguide.
The frequency at which the polarization conversion efficiency is maximum the phase mismatch factor vanishes and for this fact is called phase-matched frequency, given by
ν0 =
c Λ nTE − nTM
(2)
where c is the vacuum speed of light and nTE and nTM are the TE and TM modes effective refractive indices, respectively. Once the grating spatial period is defined in the photolithography, the only remaining degree of freedom that allows device frequency tuning is the birefringence ⏐nTE - nTM⏐. This may be accomplished through the temperature dependence of the crystal birefringence or through eletrooptical birefringence tuning. It can be observed from figure 2 that at some frequencies, part of the light will remain on the same polarization state as it entered the device and for a specific frequency all the light will be converted to the orthogonal polarization state (phasematched case). Therefore it is possible to realize a filtering function with this structure, by means of a polarization filtering, for example, with a polarizer placed after the device. It can also be observed in this filtering function (figure 2) that the nulls are not equally spaced in frequency. This fact would make it difficult to define channel bands or avoid cross-talk between channels, say 100 GHz apart and distributed over the same spectral region, after being filtered by this type of device. In an ideal response for filtering equally spaced channels, the passband should contain only one channel within its limits and channels on the stopband would have to fall on nulls of the filtering function to minimize their power on the detected bandwidth (crosstalk). This is the goal of the sparse-grating idea.
3. THE SPARSE-GRATING By sparse grating it is meant a structure with smaller sections of polarization conversion gratings separated by clear sections (no grating) along the desired waveguide like it is pictured in figure 3 for a 6-coupling region sparse grating. The desired periodicity in the rejection band of the filtering function would also require that the overall filtering function be periodic itself, presenting transmission peaks after every certain number of channels (etalon-like). This way, instead of filtering only one wavelength it would slice the communications spectrum, hence the name slicing filter.
Proc. of SPIE Vol. 7218 72180F-3
zˆ
xˆ
L3
L2
L1
Ti diffused Waveguide
L
L
L4
L
L5
L
L6
L
LiNbO3
yˆ
Fig. 3. Schematic diagram of a 6-coupling region sparse-grating TE-TM converter on a Ti:LiNbO3 waveguide.
The analysis of this structure can be done through the matrix description adopted in reference [2]. The main features relating the device structure to the spectral characteristics are described by the number N of coupling regions, the distance L between the centers of two adjacent coupling regions and the lengths Li (where i=1 to N) of each of the coupling regions. For N coupling regions N-1 zeroes in the stopband, placed Δνc apart from each other, will be obtained. This channel spacing will also be linked to the frequency spacing between two adjacent transmission peaks (Free spectral range) given by
Δν FSR =
n gTE
c = N × Δν c − n gTM L
(3)
where ngTE and ngTM are the TE and TM modes group refractive indices, respectively. The determination of the coupling strengths (ζi = κLi) for each region, and as a consequence their length will assume the values shown in Table I of the reference [2] along with the coupling coefficient κ = 1.4996 cm-1, typically obtained on previous experiments with uniform strain-inducing gratings. Figure 4 shows a sparse-grating response spectrum obtained by adopting N=6, Δνc = 100 GHz and ζi = [0.203653; 0.269865; 0.31188; 0.31188; 0.269865; 0.203653] (from ref [2]) It can be noticed from figure 4 that the zeroes of the obtained spectral response are now equally spaced in frequency and placed at the center of channel locations. This integrated-optic topology is very similar in functioning to the bulk polarization interference filter, also know as Šolc filter. On the integrated-optic topology, the polarization coupling is not accomplished in an infinitesimal length like it is in the bulk counterpart. A filter response was calculated by a finite difference numerical method in which the actual lengths of each polarization conversion region was taken into account. The obtained response is plotted in figure 5 and presents a sinc-like envelope as a consequence of the finite lengths of the coupling regions. This is the spectral response that will be used as the theoretical reference for comparison with experimental results.
Proc. of SPIE Vol. 7218 72180F-4
Polarization Conversion Efficiency
1.0
0.8
0.6
0.4
0.2
0.0
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
6
7
Normalized Frequency ((ν-ν0) X 100 GHz)
Fig. 4. Example of a 6-coupling region sparse-grating spectral response for Δνc = 100 GHz.
Polarization Conversion Efficiency
1.0
0.8
0.6
0.4
0.2
0.0 -4
-3
-2
-1
0
1
2
3
4
Normalized Frequency ((ν-v0)/ΔνFSR) Fig. 5. Spectral response of a 6-coupling region sparse-grating where the finite length of the coupling regions was taken into account .
4. EXPERIMENT 4.1 Device fabrication
The tunable sparse-grating TE-TM polarization converters were produced on an x-cut, y-propagating, twelve-millimeterwide by forty three-millimeter-long sample. Before the strain grating was produced, a couple of electrodes patterns had to be placed alongside the waveguide in order to enable electro-optic tuning of the phase-matched wavelength. Electrode patterns were generated using an image reversal photolithographic process, followed by e-beam evaporation to deposit multi-metal layers (400Å Cr/800Å Au/600Å Ti). After deposition, metal electrode patterns were obtained by lifting-off the photoresist layer. The gap between the electrodes is 17 μm.
Proc. of SPIE Vol. 7218 72180F-5
The LiNbO3 sample containing the waveguides and the electrodes was then heated up to 389°C in the e-beam chamber, while under vacuum, by using a pair of halogen lamps as the heat source. The voltage applied to the lamp was adjusted to keep an approximately constant heating rate. The sample was then kept at the final temperature by an on/off controller and a 1.75μm thick SiO2 film was deposited by e-beam evaporation with the substrate at this temperature. After completing the SiO2 deposition, the sample was kept inside the chamber until it reached room temperature, when it was removed and a 6-coupling region sparse grating pattern was delineated over the SiO2 strain film using positive photolithographic process. The grating pattern spatial period Λ (at the coupling regions) was 21μm, resulting in a phasematched wavelength (frequency) around 1545 nm (194.03883 THz). The photoresist was then hardbaked and the SiO2 layer was dry etched almost all the way in a RIE system. Complete dry etching of the strain film was avoided to prevent electrode damage. The remaining SiO2 film was removed by wet etching in buffered oxide etchant (BOE). Figure 6 shows a schematic diagram of the obtained device. The sparse grating mask used in this experiment had the constraint of having a maximum length of 24 mm. Therefore the to increase distance between the centers of adjacent coupling regions was L=4368 μm causing the value of the Δν FSR from the predicted value of 600 GHz (previous section example) to 839.55 GHz and the predicted spacing between nulls in the stopband to increase from 100 GHz (previous section example) to 139.93 GHz. The previously assumed coupling coefficient κ remained the same. Along with the coupling strengths obtained in reference [2] it allowed obtaining the coupling regions lengths. With these values the proof of concept could still be made, and a future design capable of accommodating a longer sparse-grating will satisfy the spectral characteristics presented in the previous section.
Ti diffused Waveguide
zˆ
xˆ
L3
L2
L1
L
L
L4
L
L5
L
L6
L
LiNbO3
yˆ
Electrodes Fig. 6. Schematic diagram of a tunable sparse-grating TE-TM converter on a Titanium diffused channel waveguide on Lithium Niobate (Ti:LiNbO3).
4.2 Device testing
A broadband source is convenient to characterize wavelength selectivity and for this purpose an amplified spontaneous emission (ASE) Erbium doped fiber (EDF) was used. At its output an optical isolator was connected to prevent lasing. The broadband ASE light was butt-coupled to one of the sample’s end facet through a Polarizing (3M PZTM) fiber that could be axially rotated to select either a TE or TM polarization being input to the sample. At the sample output a 20X objective was included to collimate the beam before it went through a bulk polarizer; the bulk polarizer is included to allow selecting the polarization state that would be analyzed in the Optical Spectrum Analyzer (OSA), since the converted and unconverted portions of light are orthogonal to each other; another 20X objective was included to refocus the light and couple to a single mode fiber, which was connected to the OSA input where the wavelength selectivity was measured.
Proc. of SPIE Vol. 7218 72180F-6
The substrate temperature was controlled by placing the sample on a copper plate that was located on top of a ThermoElectric Cooler (TEC). A thermistor was placed in contact with the plate and close to the sample to monitor the substrate temperature. The polarization conversion efficiency (PCE) is calculated as the power fraction on the converted polarization state at the sample output, with respect to the total power at the same point. This must be evaluated for each wavelength in the spectral band of interest, resulting in the spectral response of the PCE. Explicitly
PCE =
Pconv (λ ) Punconv (λ ) + Pconv (λ )
(4)
where Pconv(λ) is the converted light spectrum (orthogonal to the input polarization) and Punconv(λ) is the unconverted light spectrum (same polarization as input). The spectra obtained for each polarization being analyzed may be retrieved from the OSA in a text file format. Equation (4) is then applied to these data to calculate the polarization conversion efficiencies (PCEs).
5. RESULTS Initially the sample was kept at constant temperature (25 oC) and no voltage was applied to the electrodes. For a TM input polarization the spectra shown on figure 7 were measured at the output with the aid of an OSA. Anritsu
(a) Anritsu
(b) Fig. 7. Output spectra from a 6-coupling region sparse-grating TE-TM polarization converter: (a)TM input/TM output (unconverted); (b) TM input/TE output (converted).
Proc. of SPIE Vol. 7218 72180F-7
Similar measurements were carried out for a TE input polarization. With the files retrieved from the OSA for each polarization input and with the help of equation (4) the PCEs could be calculated. Figure 8 shows a comparative plot of these PCEs with the expected theoretical response. It can be seen that they are in very good agreement. o
TE input/TM output T= 25.0 C TM input/TE output Theoretical Response
0
Polarization Conversion Efficiency (dB)
-2 -4 -6 -8 -10 -12 -14 -16 -18 -20 -22 -24 -26 -28 -30 1515
1518
1521
1524
1527
1530
1533
1536
1539
1542
Wavelength (nm)
Fig. 8. Polarization conversion efficiencies (PCEs) for the 6-coupling region sparse-grating TE-TM polarization converter: TE polarization input (red up-triangle) and TM input polarization (blue circle) compared with the theoretical response (solid black).
The measured phase-matched wavelength (center peak) was 1529.16 nm for TM mode input and 1529.1 nm for TE mode input. Achieved polarization conversion efficiencies were 96.1% for TE polarization input and 95.7% for TM polarization input. 5.1 Temperature tuning
The substrate temperature could be changed (heating or cooling) by applying different values of electric current to the thermo-electric cooler on which the sample is placed. The temperature was obtained through the reading of the thermistor resistance. Different values of electrical current were applied to the TEC. It was allowed enough time for the heat exchange to reach a steady state and then the spectra for the polarization modes (converted and unconverted) were measured. 1.0
Polarization Conversion Efficiency
0.9 0.8
o
25 C
o
14 C
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 1515
1520
1525
1530
1535
1540
1545
1550
1555
Wavelength (nm)
Fig. 9. Temperature tuning. Polarization conversion efficiencies for two values of temperature (14 oC and 25 oC) - TE input polarization is shown in red and magenta, while TM input polarization is shown in black and blue.
Proc. of SPIE Vol. 7218 72180F-8
For each temperature (measured by the thermistor resistance) a PCE was calculated the same way as done before. It was observed the shift not only in the center peak wavelength but in the whole filtering function as a result of the temperature change, characterizing temperature tuning of the device. In figure 9 the conversion efficiencies for the lowest and highest temperatures are shown, others were not plotted to avoid excessive cluttering. The obtained temperature tuning rates were
dλ dt = −0.99929 nm/oC for the TE polarization input and
dλ dt = −1.00138 nm/oC for the TM polarization input. Temperature tuning curves for both polarizations are plotted in figure 10. TM to TE conversion TM to TE data linear regression TE to TM conversion TE to TM data linear regression
1542
Center Peak Wavelength (nm)
1540 1538 1536 1534 1532 1530 1528 1526 1524 12
14
16
18
20
22
24
26
28
30
o
Temperature ( C)
Fig. 10. Temperature tuning. Center peak wavelength temperature dependence.
The obtained bandwidth (FWHM) for the center peak was measured 1.044 nm around 1540.19 nm (or 131.9 GHz around 194.645 THz) with a substrate temperature of 14oC. The theoretical response and measured responses are in good agreement (within an experimental error). In figure 8 a little asymmetry in the height of secondary peaks (those in the rejection band) in the lower wavelength side when compared with the ones in the high wavelength side can be noticed. This may be explained by a gradient in the birefringence along the length of the sample (direction of light propagation in the waveguide), that would in turn cause a variation in the phase-matched wavelength of one section of the sparse grating to be slightly different than in the others. The net result is a “sliding” effect in the contribution of that section with respect to the contribution of the other sections in the composition of the whole filtering function, causing the asymmetry. This effect can even be more severe when dealing with longer sparse gratings, where birefringence gradients may be larger under the same fabrication conditions. It was found in the literature [4] reports on similar effects in long uniform grating TE-TM polarization coupling filters, also attributing to birefringence gradients in the device. 5.2 Electro-optic tuning
The application of a voltage to the electrodes deposited alongside the waveguide generates an electric field in the waveguide mainly oriented along the crystalline z direction that changes the waveguide birefringence through the electro-optic effect, thus the phase-matched wavelength. This provides another tuning mechanism for the TE-TM polarization converter and is also done with the highest voltagerefractive index transduction because of the field orientation that takes advantage of the highest electro-optic coefficient in LiNbO3. The spectral characteristics were measured at constant temperature for various applied voltages to the electrodes deposited on the sample. The PCE obtained when applying the two voltage extrema are shown in figure 11 for the TM polarization. TE polarization tuning responses were very similar and are not presented to avoid cluttering of the plot. The center-peak wavelength of the response shifted by 5.76 nm with a variation in the applied DC voltage tuning from -70 V to +70 V, corresponding to a tuning rate of 0.045nm/V for TM input, and 0.039nm for TE input. Polarization conversion
Proc. of SPIE Vol. 7218 72180F-9
efficiency varied somewhat between both situations (from 99.4% at +70V to 91.7% at -70 V). The tuning curve for the center peak wavelength as a function of applied voltage, shown in figure 12, presents a linear behavior for both polarization modes, although the values for the tuning rates are low when compared to previous results in our group. An effect that might account for this difference is the presence of free-charges in the SiO2 strain film, which under the application of external voltage would move and develop an electric field counter acting the applied field, thus reducing the voltage- refractive index transduction. The appearance of free charges in a film expected to be an insulator probably has its roots in the deposition system being used (e-beam evaporation), which is used either for metal or dielectric films. Therefore cross contamination between highly conductive and insulating materials deposited in the same chamber becomes very likely.
Polarization Conversion Efficiency
1.0
0.8
70 V -70 V
0.6
0.4
0.2
0.0 1520
1530
1540
Wavelength (nm)
Fig. 11. Electro-optic tuning. Polarization conversion efficiencies for two values of voltage (-70V and 70V) - TM input polarization is shown. . 1537 1536
Wavelength (nm)
1535 1534
TM input/TE output Linear fit TM input data TE input/TM output Linear fit TE input data
1533 1532 1531 1530 1529 1528 -80 -70 -60 -50 -40 -30 -20 -10
0
10 20 30 40 50 60 70 80
Applied Voltage (V) Fig. 12. Electro-optic tuning. Center peak wavelength voltage dependence.
Proc. of SPIE Vol. 7218 72180F-10
6. CONCLUSION A new type of guided-wave device employing the sparse-grating concept was realized on Titanium diffused Lithium Niobate waveguides. This type of device may find application as filters in DWDM communication systems due to the fact that they are able to generate very narrow bandwidth filters in the passband with adjacent channels in the stopband placed by design on nulls of the filtering function. To realize these filtering functions, the codirectional coupling between orthogonal modes in a waveguide (TE-TM conversion) was explored. The spatially periodic static-strain inducing SiO2 surface film was obtained by depositing a 1.75μm thick SiO2 on an x-cut LiNbO3 substrate, with ypropagation waveguides produced by Titanium diffusion, preheated and kept at 389ºC during the whole deposition. It was let cool down to room temperature and patterned a six-coupling region sparse gratings pattern. The coupling regions spatial period was 21μm. Polarization conversion efficiencies (PCEs) at the phase-matched wavelengths as high as 96% were obtained. Thermal tuning was also used to shift the spectral response, and a thermal tuning rate of about -1.0 nm/°C was measured for both polarizations. Electro-optic tuning was also demonstrated in the straight channel sparse-grating device. A shift for the peak conversion wavelength of 5.76 nm was realized over a tuning voltage ranging from -70V to +70V. The electrode gap was 17μm, and tuning rates of 0.045nm/V for the TM and 0.039nm/V for the TE polarization were obtained.
REFERENCES [1] [2] [3] [4]
P. Tang, O. Eknoyan, and H. F. Taylor, "Rapidly tunable optical add-drop multiplexer (OADM) using a staticstrain-induced grating in LiNbO3," J Lightwave Technol 21(1), 236-245 (2003). Taylor, H. F., "Tunable spectral slicing filters for dense wavelength-division multiplexing," J Lightwave Technol, 21(3), 837-847 (2003). Yariv, A., [Quantum Electronics], John Wiley & Sons, New York,515-538 (1975) F. Cholet, J. P. Goedgebauer and G.Ramankoto, "Limitations imposed by birefringence uniformity on narrowlinewidth filters based on mode coupling," Opt Eng 40(12), 2763-2770 (2001).
Proc. of SPIE Vol. 7218 72180F-11
