Sep 1, 2006 - fiber can be modeled as a cylindrical lens with a radius of 62.5 μm and cladding index ..... The input fibers on the left side are Thorlabs 630 nm.
PHOTOPOLYMER WAVEGUIDE TO FIBER COUPLING VIA 3D DIRECT-WRITE LITHOGRAPHY by Charles Davis Anderson B.S., University of Portland, 2002
A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirement for the degree of Master of Science Department of Electrical and Computer Engineering 2006
UMI Number: 1439421
UMI Microform 1439421 Copyright 2007 by ProQuest Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, MI 48106-1346
This thesis entitled: Photopolymer Waveguide to Fiber Coupling via 3D Direct-Write Lithography written by Charles Davis Anderson has been approved for the Department of Electrical and Computer Engineering
(Robert R. McLeod)
(Allen Mickelson) Date The final copy of this thesis has been examined by the signatories, and we find that both the content and the form meet acceptable presentation standards of scholarly work in the above mentioned discipline.
iii
Anderson, Charles Davis (M.S., Electrical Engineering) Photopolymer Waveguide to Fiber Coupling via 3D Direct-Write Lithography
Thesis directed by Professor Robert R. McLeod
This thesis proposes and demonstrates a novel method of pigtailing optical circuits by accurately aligning a waveguide to a fiber previously embedded in photopolymer. An automated, iterative method locates two embedded optical fiber tips by employing both a reflection and transmission microscope to obtain 3D positions of the fiber tips. This eliminates the need for precise fiber positioning, and low loss pigtailing can be achieved with arbitrarily positioned fibers. The photopolymer pigtailing method is demonstrated by interconnecting two arbitrarily positioned fibers across 30 mm of photopolymer. A 3D direct-write platform enables control over the position, direction, and length of the polymer waveguide interconnect, and aligns the polymer waveguide to within 3 μm of the fiber tip. The path of the waveguide includes 4 s-bends to accurately couple light from the input fiber to the relatively offset output fiber. By employing this method of pigtailing, precise alignment of the fiber is not necessary, and alignment of the interconnecting waveguide is accomplished through an automated system. The photopolymer solution additionally increases reliability of the pigtail by encasing all elements in a mechanically robust package.
iv Acknowledgements I would like to thank Dr. Robert R. McLeod for his patience and diligence in helping me complete this thesis. My parents, brother, and girlfriend Julie Page must be recognized for their willingness to support and encourage me during my research and writing over the past two years. I would also like to thank InPhase Technologies for working with me and for providing their material in our unique packages. Finally, I would like to acknowledge NSF and CCIS for funding this work.
v Table of Contents Figures.......................................................................................................................... vi 1. Introduction.............................................................................................................. 1 2. 3D Diffusion-Mediated Photopolymers................................................................... 6 3. 3D Direct Write Lithography Platform.................................................................... 9 4. Accurate 3D location of encapsulated subcomponents ......................................... 12 4.1 Microscopes ..................................................................................................... 12 4.2 Detection of encapsulated fibers...................................................................... 14 4.2.1 Step 1: Locate polymer extent in Z........................................................... 16 4.2.2 Step 2: Locate fiber cladding edges in Y .................................................. 17 4.2.3 Step 3: Locate fiber depth in Z ................................................................. 18 4.2.4 Step 4: Locate fiber core edges in Y......................................................... 19 5. Experimental Demonstration ................................................................................. 22 5.1 Experimental Sample ....................................................................................... 24 6. Conclusion ............................................................................................................. 32 References................................................................................................................... 35
vi Figures Figure 1: Phase microscope image of an SMF28 optical fiber encapsulated in InPhase photopolymer. ..........................................................................................8 Figure 2: 3D Lithography Platform diagram: The green beam is used to write small localized index changes in photopolymer on high precision 3D stages. A red, non-destructive beam is used for part detection in the material. This combination provides both transmission (Scanning transmission microscopy [20]) and reflection (confocal) feedback signals for accurate element location.................................................................................................................10 Figure 3: The minimum distance from the fiber face a writing spot can achieve before being perturbed by the embedded fiber. ...................................................15 Figure 4: Initial scan in depth to determine the boundaries of the photopolymer. ......16 Figure 5: Vertical centroid position voltage feedback during a 2mm scan of polymer in the vertical direction with an embedded fiber, the arrow indicating the fiber core. ......................................................................................18 Figure 6: Second reflection positioning scan with peak (a) representing the front polymer/glass boundary depth, peak (b) representing the top edge of the fiber, and peak (c) representing both the back polymer/glass boundary and the bottom fiber edge pressed up against the back polymer/glass boundary.......19 Figure 7: Second transverse scan of the fiber between the transverse edges of the cladding revealing the centroid deflection near 2350 μm caused by the core of the fiber............................................................................................................20 Figure 8: Confocal voltage data showing the core of the fiber at the peak, confirming the core location results of the transmission microscope..................21 Figure 9: The fully mounted experimental sample after writing and curing. Two 630 nm single mode fibers are embedded in polymer on the left side of the part and two multi-mode fibers are embedded on the right side..........................26 Figure 10: Differential Index Contrast (DIC) microscope picture of a waveguide written near the single mode 4.3 μm core, aligned by hand using the algorithm presented in this paper, with a misalignment of approximately 6 μm and an output loss of approximately 28 dB. ..................................................28 Figure 11: Differential index contrast microscope image of a polymer waveguide coupled to a single mode fiber with the input laser on. The waveguide was aligned using the computer controlled location algorithm demonstrated in Chapter 4..............................................................................................................29
vii Figure 12: Differential scanning transmission microscope image of a 4.3 μm core single mode fiber with the waveguide written from it.........................................30 Figure 13: Polymer waveguide coupling to the single mode input fiber, incurring bend radius losses, and coupling light into the output multi-mode guide. ..........31
1 1. Introduction
Optical communications has driven the need for complex optical circuitry to process the massive amounts of data traveling through our global networks. As the complexity of the required circuitry grows, optical integrated circuits have become an increasingly attractive alternative to traditional free space optics. In addition to decreasing the size of the system, optical integration increases reliability by fixing all elements in one structure, thereby decreasing the likelihood of alignment errors during both product manufacture and lifetime. Since the integrated circuit must inevitably interface to fiber; cheap, low-loss, stable fiber pigtailing techniques are essential for optical integration. The most common and simplest method to actively align a fiber is to manually adjust precision stages using feedback from an optical power meter. After optimization, the fiber position is fixed with epoxy, solder, or laser welding. On a large scale, this method is labor intensive and prone to long term instability. Instead of fixing the fiber using epoxy, solder, or laser welding, miniature (9 mm) silicon mounts can be fabricated to hold specific elements and be similarly actively aligned to 0.1 μm accuracy [1]. Each fiber must be positioned correctly on the part, and each part must be aligned in six dimensions to assure low coupling losses. However, the silicon mounts must be fixed in place using epoxy, solder or laser welding, making this arrangement similarly unstable and labor intensive. Though smaller than traditional adjustment, this arrangement requires that many components be actively aligned in a manner similar to the macroscopic adjustment method.
2 Silicon v-grooves provide an alternative to active alignment of a set of fibers by providing a lithographically etched slot in which fibers may be placed. Many grooves can be placed on a single substrate, commonly spaced 250 μm apart, thereby reducing the need to align individual fibers [2]. The width of the groove etched into the crystallographic planes of a silicon wafer can be varied to provide the desired height of the fiber with respect to the substrate. Fibers are passively aligned by placing them into the grooves and securing with epoxy. Once secured, all the fibers are actively aligned to the optical circuit in one step, instead of each individually. To compensate for wicking, differential thermal expansion, and surface tension effects of the epoxy, 3 μm silicon nitride clips have been developed to increase alignment reliability [3]. To further reduce alignment uncertainties, silicon v-grooves can be integrated onto a silicon optical bench (SiOB), combining fiber alignment structures and integrated waveguides onto a single chip. Separate etching steps are required to create Si or polymer ridge waveguides and the alignment grooves, so to maintain photolithographic precision, a protective layer of Si3N4 must be deposited over the chip in which “windows” can be opened using reactive ion etching to enable the wet etching process in that area [4]. Fibers can then be passively placed into V-grooves that are photolithographically aligned to waveguides on a silicon integrated optical chip. The application limitations here lie only in the limited versatility silicon offers as optical material. Tolerances of the v-groove alignment procedure are given by the angular alignment of the etch mask to the crystallographic planes; a 0.5˚
3 misalignment produces a 1 μm groove width error [4]. A transverse alignment error of 1 μm for a standard telecom fiber with a 10 μm mode field diameter at wavelength of 1.55 μm results in about 0.25 dB loss. An entirely different approach to fiber pigtailing uses a selective wet etch process to produce male and female “connectors,” allowing a fiber to be easily and accurately coupled [5]. The female end is made by placing a planar, buried, channel waveguide into a hydrofluoric acid (HF) bath to remove the core, while the male end is produced by submersing an optical fiber into the bath to remove the cladding. This plug and socket arrangement allows for passive alignment of the two cores and can potentially provide low coupling losses. However, tolerances and losses have not yet been reported. Another possible solution to fiber pigtailing utilizes diffusion-mediated photopolymers, which are materials that increase index when exposed to sufficient light energy. Thin (120 μm) sheets of photopolymer waveguide arrays have been made by laminating a mask to the material and illuminating the mask. Slots for fiber insertion measuring 120 μm wide and 600 μm long are created by excimer-laser ablation, aligned to the polymer waveguides. Additional layers of polymer are placed above and below the waveguides, and an angle cleaved fiber is placed in the slot. Liquid polymer is then added as an epoxy to fix the fiber in place [6, 7]. Within the tolerances of the multiple mask alignment steps, passive fiber coupling procedures similar to the v-groove with clips can be used. Coupling losses are reported to be typically near 0.5 dB, indicating a lateral misalignment error of 1.75 μm.
4 Thus far, all alignment procedures discussed have focused on accurately positioning a fiber to a previously fabricated waveguide. Self-written polymers demonstrate a fundamentally different approach, where the fiber is embedded into photopolymer and a waveguide is written using the output radiation of the fiber. This yields a waveguide that is intrinsically coupled to the fiber without the need of complicated alignment procedures. Index change of the polymer is strongest at the regions of intense illumination, creating a self-trapping effect similar to spatial optical solitons. The direction of the waveguide can be controlled by introducing an asymmetry across the transverse intensity profile of the focused writing beam [8]; however, no demonstration of this type of waveguide directional control from the output of a single mode fiber is reported in the literature. Waveguide directional control can be achieved in polymer by introducing a second fiber aligned anti-parallel, but slightly offset, to the first. It has been shown that two fibers aligned on a common longitudinal axis, separated by 1 cm, can be coupled by counter-propagating self-writing beams in photopolymer with losses less than 1 dB [9, 10]. With this counter-propagating method, asymmetric transverse intensity profiles are not needed for directional control. The development of higher index at the points of intensity overlap will guide light from one emitting fiber to the other. This has been demonstrated with two fibers separated by 1 mm and transversely offset by 40 μm, resulting in a fiber-to-fiber coupling with 2 dB loss [11]. The self-written waveguide method has a number of attractive features, including intrinsic alignment, robust encapsulation, and adhesion of the polymer
5 material to silica. The combination of these attributes results in a well-aligned waveguide that is directly bonded to the fiber face and cladding. Deformation due to thermal or mechanical stress should not introduce displacement errors, as both the fiber and the waveguide move simultaneously. Encapsulation also provides a protective layer from dust and other contaminates that may cause increases in loss over time. These advantages obtain reliable performance in a hostile environment, potentially offering robust performance in harsh or uncontrolled environments. Reference [12] has demonstrated a multi-mode wavelength division multiplexing communication module suitable for installation in a car, utilizing self-written waveguides to couple embedded, wavelength-selective filters to a multi-mode fiber. This demonstrates that photopolymer is not limited to pigtailing fibers, but can be used as a robust medium for integrated optics. Waveguide placement after fiber positioning, as demonstrated by self-written photopolymers, suggests an alternative approach to fiber pigtailing. Instead of trying to perfectly place a fiber directly in front of a previously fabricated waveguide, the fiber is passively embedded with loose tolerances, and the waveguide is positioned based on the fiber location. Unfortunately, the self-written waveguide process requires very precise control over illumination intensity to produce an invariant single mode guide. In addition, this process does not have the ability to develop waveguides with either an arbitrary path, or an arbitrary length. This paper introduces a method to gain length and direction control by sacrificing the intrinsic alignment inherent in the self-writing technique in favor of a 3D maskless lithography platform. This platform controls the direction of the
6 waveguide in 3D, while still maintaining the ability to locate and write a waveguide from the fixed tip of the fiber. This method offers the potential to connect any 3D arranged components passively and accurately with low loss. This paper first discusses the polymer system used in Chapter 2, then discusses the direct write lithographic optical system that incorporates two discrete detection microscopes in Chapter 3. The data collected from the microscopes is integrated into a fiber location algorithm, presented in Chapter 4. Results of the location algorithm and the complete photopolymer pigtailing device are shown in Chapter 5.
2. 3D Diffusion-Mediated Photopolymers
The photopolymer properties are central to the proposed pigtailing method. There are two distinct requirements; the polymer must be a high-quality encapsulation medium and simultaneously permit index change by optical exposure. Although a number of volume diffusion-mediated polymer systems could potentially meet these goals [13, 14, 15], most commercially available systems attempt to do so in a single polymer formulation, inevitably compromising on both. A more effective method to optimize these functions is to separate them into two cohabitating polymer systems, such as that used in InPhase Tapestry HDS3000. The material uses two, independently curable polymer systems, enabling separate optimization of physical properties and optical sensitivity.
7 The encapsulation polymer has the properties of good adhesion and low curing shrinkage, providing a stable structure for the photosensitive polymer to reside. It is designed to have high optical quality, be highly transparent, and have a curing procedure that is independent of the photosensitive polymer. Curing of this polymer is thermally initiated at room temperature producing a uniform matrix. Nearly all polymers shrink upon polymerization as low molecular-weight monomer joins to form a high-molecular weight polymer chain. Fortunately, prior to curing, the material is liquid and can easily flow to fill spaces and holes created by polymerization driven shrinkage. However, as the monomers continue to bind to form very large polymer chains, the viscosity of the material increases to the point where it no longer flows to compensate for shrinkage. After this point, shrinkage can cause stress-induced birefringence or delamination from glass surfaces. Once cured, it maintains a fixed volume, limiting further shrinkage during curing of the writing stage polymer. When embedding optical fibers into the Tapestry polymer, the low shrinkage produces a uniform optical environment surrounding the fiber, without signs of stress or delamination, or non-uniform scattering that would indicate such features. An embedded SMF28 optical fiber magnified in a phase microscope is shown in Figure 1. Since the physical form of the structural polymer is like rubber, it can tolerate mechanical stress [16]. Heat can cause polymer expansion, but since the fiber is adhered to the polymer, it is expected to move and adjust with its expanding surroundings, maintaining stable alignment to any index features written in the photopolymer.
8
Figure 1: Phase microscope image of an SMF28 optical fiber encapsulated in InPhase photopolymer.
Index structures in the photopolymer are created by illuminating the photosensitive initiator. The single-photon initiator excites monomers to bond with other nearby monomers, creating polymers. This polymerization depletes the local monomer density, inducing diffusion to equalize the local monomer density. The high molecular weight polymer molecules do not diffuse, and thus remain localized to the point of optical writing. The combination of the stationary polymer molecules and the redistribution of the monomer produce a local increase in density. This causes a local index increase, as material index is proportional to its density. The index increase is related to the power incident on the polymer system, which can be adjusted to produce a desired index contrast. An advantage of this system is the
9 ability to produce index structures using low power CW lasers without the need for wet chemical or thermal processing steps. Once all writing has occurred, the sample is “flood cured,” where the entire sample is placed under uniform light. This forces uniform polymerization of the remaining monomer without creating a chemical gradient, and therefore no further index changes occur. This curing process renders the photopolymer insensitive to further illumination, as all monomer has been converted to polymer. It also bleaches out the initiator species, making the material fully transparent. In summary, InPhase Tapestry HDS3000 is a material that responds with an increase of index to 3D patterned optical exposure, while being an effective encapsulant of optical fibers. In the next Chapter, I describe the 3D direct-write lithography platform used to create this optical exposure pattern that can write index features near embedded elements such as fibers.
3. 3D Direct Write Lithography Platform
Control of the polymerization of the photosensitive monomer can be achieved through many writing techniques. Self-writing in photopolymer has been shown to achieve intrinsic alignment, but does not provide an effective means for control of the length or direction of the written waveguide. Focusing a beam into the polymer and writing index features along the path of the beam (axial writing) can produce circular waveguides along the path of the laser, but the length of the waveguide is limited by the working distance of the writing objective. Increases in the working distance leads
10 to a lower numerical aperture (NA), enlarging the minimum spot size and decreasing resolution. We employ a perpendicular writing approach where index features are written in the photopolymer perpendicular to the optical axis, similar to those used in femtosecond waveguide writing in glass [17, 18], shown in Figure 2. While this lithographic technique requires additional beam shaping optics to create a circular waveguide in the material, it does allow writing waveguides with arbitrary length and direction with index features near micron scale. In addition to writing index features, this perpendicular system incorporates two separate detection techniques that can provide a complete set of coordinates of the tip of an embedded fiber, described in the next Chapter [19].
Dichroic Beamsplitter
Frequency doubled Nd:YAG @ 532 nm
Polymer mounted on high precision 3D stages
Position Sensitive Detector
λ/4 waveplate
Photodetector
Polarizing Beamsplitter Pinhole
Y Z
HeNe @ 633 nm Figure 2: 3D Lithography Platform diagram: The green beam is used to write small localized index changes in photopolymer on high precision 3D stages. A red, non-destructive beam is used for part detection in the material. This combination provides both transmission (Scanning transmission microscopy [22]) and reflection (confocal) feedback signals for accurate element location.
11
The InPhase Tapestry HDS3000 photopolymer used is designed for exposure at 532 nm, while it is significantly less sensitive in the longer red wavelengths, such as 633 nm. We therefore employ a two-color system, where the red laser is used for nondestructive detection and location purposes (reading), and the green laser is used to polymerize the material (writing). To achieve a tolerance similar to silicon vgrooves, the relative positioning tolerance between the writing and reading spots must be submicron. Aligning and collimating the beams to be collinear through the same objective lens maintains positional tolerances more reliably than using separate optical trains calibrated to submicron accuracy, as shown in Figure 2. Since both the reading and writing beams share the same objective lens, the objective and NA must be carefully chosen to maintain a diffraction-limited spot over the depth of the sample for both colors. The focusing objective used is a Geltech 350340 aspheric lens because it is designed specifically to focus 685 nm light through 1.2 mm of K3 glass (index 1.518) at a numerical aperature (NA) of 0.62. Unfortunately, spherical aberration is variable with depth in the material and the wavelength of the light, so to maintain a spot reasonably close to diffraction limited throughout 1 mm of polymer encased in 1 mm of glass, the beam is stopped down to NA=0.32. The 1/e2 radius of the waist throughout the photopolymer is approximately 1 μm, varying by a maximum of 0.2 μm. Active aberration correction, such as an adjustable lens system, or more creative techniques [20] have not been employed, but could enable smaller feature size over extended material depth.
12 Low loss, efficient coupling to an optical fiber requires a circular waveguide. Even with a diffraction limited spot, the waveguide produced by this lithographic system is elliptically shaped, due to the depth of focus of the objective lens at NA=0.32 being larger than its transverse spot size. A slit may be placed in the beam before the lens, oriented parallel to the transverse direction of writing such that the NA of the objective is significantly less in the perpendicular direction than that in the parallel direction [21]. This method may be employed to obtain a circular waveguide in photopolymer by significantly increasing the waist size. However, the primary focus of this paper is to demonstrate an accurate technique to locate an embedded fiber core and align a waveguide to that position, and thus will not discuss beam shaping. With the photopolymer mounted on high precision 3D stages, this direct-write lithographic platform can write a waveguide with precise location, direction, and length to micron tolerances. The next section will discuss the two detection microscopes that provide the positional information necessary to align the waveguide to an embedded fiber.
4. Accurate 3D location of encapsulated subcomponents
4.1 Microscopes
To detect the optical fiber encapsulated in a volume of polymer, two detection techniques are required to derive the 3D coordinates of the tip of that fiber. A
13 transmission microscope is used to acquire the transverse coordinates, and a confocal reflection microscope is used to acquire the longitudinal coordinates. Each will be discussed in turn. Differential transmission microscopy detects transverse phase gradients at the focus of an objective by tracking the centroid shift of the beam after a Fourier transform lens. The index structure at the focus can be calculated by de-convolving the optical transfer function of the optical system from the output data collected. Combined with a 3D raster scan system, the low spatial frequency phase profile of an entire volume can be reconstructed [22, 23]. Alternatively, the data collected can be directly analyzed and correlated to the response of embedded elements, such as a fiber. This detection scheme is well adapted to the 3D direct-write lithography system, where the ability for accurate location of an embedded index structure specifically at the focus of an objective is crucial to the alignment of a polymer waveguide to a fiber. Since the detection is sensitive to transverse spatial frequencies that deflect the beam, resolution is strongest when detecting the transverse extent of the fiber, while the depth of the fiber in the polymer is acquired using the reflection microscope. Confocal reflection microscopy complements differential transmission microscopy, as it is most sensitive to high spatial frequency index changes antiparallel to light propagation that efficiently reflect the beam. Fresnel reflections at boundaries are retroreflected back through the system to a beam splitter, where a lens focuses the reflections through a pinhole. Only reflections that occur at the focus can
14 be correctly retroreflected, and thus are the only reflections detected. This leads to very precise depth-sectioning, with a resolution of 1.4λn/NA2 = 14.8 μm in the system demonstrated here [24]. This scanning technique allows for non-destructive axial data collection using a focused beam, ideally matched to our system and needs of detecting buried index structures in photopolymer. The combination of these two microscopes obtains both transverse and axial information, providing the necessary data to accurately locate fibers and align waveguides for optimal coupling.
4.2 Detection of encapsulated fibers
A beam incident on a fiber will be refracted by the curved shape of the fiber due to the index difference between the glass and polymer, and must be discussed before the detection algorithm is presented. The curved top surface of an SMF-28 fiber can be modeled as a cylindrical lens with a radius of 62.5 μm and cladding index of 1.531. When buried in polymer of index 1.481, the power of the modeled lens is Φ≈800 [1/m]. The lensing causes two effects: the focused green writing spot is perturbed by the refraction causing aberration, and the focused red detection spot is shifted axially. Both effects will be discussed in turn. Figure 3 demonstrates the limit at which no aberration occurs and an unperturbed writing spot can be achieved, at the distance r·tan(NA/n) from the tip of the fiber. However, the range over which the beam is refracted, 12.67 μm, is less than the Rayleigh range of the 4.3 μm core fiber used, z0 = 23 μm for 633 nm light. Any
15 small perturbation of the waveguide profile shorter than the Rayeleigh range of the mode should have minimal impact on coupling efficiency.
NA/n=0.2
Fiber Cladding r=62.5 Fiber Core r tan( NA / n ) polymer
Figure 3: The minimum distance from the fiber face a writing spot can achieve before being perturbed by the embedded fiber.
The lensing effect of the fiber also produces an astigmatic axial shift of the reading and writing foci. Using a paraxial ray tracing model, a cylindrical lens with power 800 [1/m] would shift the focus of a NA/n = 0.2 incident beam by 3.1 μm. This is comparable to the Rayleigh range of the writing beam and is thus a potentially serious aberration. However, an artifact of the material system used in this work leads to index features whose depth is much greater the Rayleigh range. Thus, we will ignore this focal shift since it has a negligible effect on the coupling efficiency in this work. Since the lensing effect of the fiber is negligible, the fiber core location with respect to the stage coordinate system can be found directly from microscope data, without the need for additional adjustments. However, fiber location with micron
16 accuracy tolerances requires an iterative approach with the two microscopes totaling four data collection sequences.
4.2.1 Step 1: Locate polymer extent in Z
First, the boundaries of the polymer are found using the confocal detection system. In these experiments, the polymer is encased between two glass plates, and thus the confocal system detects the reflections from the polymer/glass boundary, as shown in Figure 4. The part is translated in Z and the confocally-filtered intensity like that shown in Figure 4 is recorded. The two maxima correspond to the Z stage coordinates of the glass/polymer interfaces. These interfaces are the boundaries of the polymer, and are limits to the depth of the embedded fiber.
Confocal Detector Voltage (V)
First Z Scan 0.03 0.025 0.02 0.015 0.01 0.005 0 1100
1200
1300
1400
1500
1600
Absolute Z Stage Position (μm)
Figure 4: Initial scan in depth to determine the boundaries of the photopolymer.
1700
17 4.2.2 Step 2: Locate fiber cladding edges in Y
Next, the transverse location of the fiber is found using the transmission microscope by observing the centroid on the PSD shift when a fiber is translated past the detection focus. The sample moves with a constant velocity in the vertical dimension (Y), perpendicular to both the axis of the red signal beam and the axis of the fiber. The magnitude of the velocity only affects data resolution, and with the data collection system used it was determined that a velocity of 0.1 mm/s obtained sufficient resolution for micron accuracy. Maximum signal to noise for this data collection step has been experimentally determined to be when the focus of the detection beam is approximately 100 μm in front of the first polymer/glass boundary. This ensures that the fiber being detected is between 100 μm and 600 μm beyond the objective focus, producing a deflection of the centroid of the beam of approximately 8.4 mrad. The signal returned from the scanning transmission microscope shows the extent of the fiber cladding boundaries, and an approximate location of the core can be determined, as shown in Figure 5. The transverse center of the fiber may be estimated by calculating the average value of the graph and recording the intersection of the data between the fiber cladding peaks and the average value. This assumes all perturbations that cause a centroid shift during a scan are antisymmetric about the unperturbed beam, and therefore the average value is an accurate measure of the centroid of the unperturbed beam. This assumption is validated by the observation that the average value of the entire data set, as shown by the blue line, matches the recorded value beyond the
18 edges of the fiber, shown in this example to be between the values of 2265 μm and 2465 μm in stage coordinates, despite the existence of uncontrolled elements such as an air bubble. The transverse center of the core is recorded by the control program for use in the next step. Second Y Scan
PSD Y Centroid Voltage(V)
0.25
Fiber Cladding
0.2 0.15
Fiber Core
0.1
Air Bubble
0.05 0 -0.05 -0.1 2265
Fiber Cladding 2465
2665
2865
3065
3265
Absolute Y Stage Position (μm)
Figure 5: Vertical centroid position voltage feedback during a 2mm scan of polymer in the vertical direction with an embedded fiber, the arrow indicating the fiber core.
4.2.3 Step 3: Locate fiber depth in Z
Now that the transverse coordinates of the center of the fiber are known, the next step is to find the axial coordinates of the fiber. This is done by moving the part so that the focus is approximately centered on the core transversely, and is approximately 30 μm inside the polymer with respect to the front glass/polymer boundary, at position 1600 μm in the example shown in Figure 6. This data collection step translates the stages in the negative Z direction, and the first two peaks encountered represent the front and back surfaces of the fiber, in this example at
19 positions 1350 μm and 1250 μm in stage coordinates. To obtain any reflection signal from the fiber during the confocal scan, the fiber must be centered transversely on the focus to within r·sin(NA/n) = 12.4 μm, where r is the radius of the fiber, and NA is the numerical aperture of the focusing beam. Outside of this range, the reflections from the curved surface of the fiber are directed away from the objective lens and are not retroreflected back through the confocal system to the detector. Since the core is centered within the cladding, the calculated average of the two boundaries is stored in the program as the core depth position in stage coordinates.
Confocal Detector Voltage (V)
Second Z Scan 0.03
c
0.025 0.02
a
b
0.015 0.01 0.005 0 1100
1200
1300
1400
1500
1600
1700
Absolute Z Stage Position (μm)
Figure 6: Second reflection positioning scan with peak (a) representing the front polymer/glass boundary depth, peak (b) representing the top edge of the fiber, and peak (c) representing both the back polymer/glass boundary and the bottom fiber edge pressed up against the back polymer/glass boundary.
4.2.4 Step 4: Locate fiber core edges in Y
With the second confocal depth scan, the depth position of the fiber core is calculated and used in the fourth and final step to determine the transverse location of
20 the fiber core. The part is moved so that the fiber is at a depth approximately 10 μm past the focus of the red beam, and transversely between the edges of the fiber cladding as detected earlier in the first transverse data set. A scan across the width of the fiber cladding, again in the vertical (Y) dimension, yields a trace that identifies the core of the fiber, as shown in Figure 7. This trace shows two peaks that are very similar to the cladding peaks in the first scan in the Y dimension. However, since the extent of the scan did not include the cladding edges, these peaks are the core edges, with their average position being the center of the core. Second Y Scan
PSD Y Centroid Voltage (V)
4 3 2 1 0 -1 -2 -3 2280
2300
2320
2340
2360
2380
2400
2420
Absolute Y Stage Position (μm)
Figure 7: Second transverse scan of the fiber between the transverse edges of the cladding revealing the centroid deflection near 2350 μm caused by the core of the fiber.
With this algorithm, it is possible to position the part so that the focus of the read beam is within a micron of the core of a fiber in both transverse and axial directions. To confirm that the focus is at the core of the fiber, the confocal microscope can be used. When aligned on the core of a fiber, a confocal scan should be able to measure a reflected signal from the index boundary between the core and
21 cladding of the fiber. Similarly to the confocal cladding detection of the third step, reflections from the core will only occur within a small transverse range: r·sin(NA/n) = 0.43 μm. However, for this confocal reflection comparison data, the part is translated in the vertical (Y) dimension instead of the axial (Z) dimension. The confocal data collected during the second Y scan shows the fiber core is in this case at 2344 μm in stage coordinates, as shown in Figure 8, which confirms the core position shown in Figure 7.
Confocal Detector Voltage (V)
Comparison Y Scan 0.016 0.014 0.012 0.01 0.008 0.006 0.004 0.002 0 2280
2300
2320
2340
2360
2380
2400
2420
Absolute Y Stage Position (um)
Figure 8: Confocal voltage data showing the core of the fiber at the peak, confirming the core location results of the transmission microscope.
Automation of this four step algorithm yields an accurate, reliable system for locating embedded fibers in a photopolymer encapsulant to within 1 micron transverse tolerance. A perpendicularly written photopolymer waveguide can be coupled to the core of the fiber and drawn in 3D across an arbitrarily long path to an arbitrarily placed optical element, such as another fiber. Using this fiber detection system enables a fiber pigtailing device to be created out of a photopolymer matrix
22 with low fiber placement tolerances, yet potentially high coupling efficiency, demonstrated in the next Chapter.
5. Experimental Demonstration
This Chapter demonstrates the implementation of a photopolymer based fiber pigtailing system. Two fibers are embedded in the photopolymer discussed in Chapter 2, and then the photopolymer sample is mounted on the direct write lithography platform discussed in Chapter 3, where the fiber locations are determined. A photopolymer waveguide is coupled to the core of the embedded input fiber and routed in 3D to the second fiber embedded in the sample, representing an arbitrary optical element in the photopolymer. The fabrication of the proposed photopolymer pigtail requires a direct write optical system that is well aligned and calibrated such that the green writing spot writes where the red detection spot reads. Determining the diffraction limited performance can be achieved by comparing the focused spot in the focal plane of the objective lens to an ideal Gaussian spot. This can be done using a standard knife scan test. The process of finding the focal plane of the objective lens requires incremental knife scans in the axial (Z) dimension, creating a beam profile. This situation deviates from the standard method in that the knife must be embedded into the polymer to obtain an accurate measurement of the actual writing and reading spots produced by the green and red beams within the polymer. The minimum spot
23 radius of the profile is the focal point. The data collected at this point can be compared to the diffraction-limited intensity profile. In addition to characterizing the diffraction limited performance of the system, the position of the focus for both the red and green beams as a function of material depth can be used to characterize the longitudinal chromatic dispersion of the entire lithographic system, including the objective, glass cover, and polymer. This can be achieved by embedding a knife at various depths in polymer and using the incremental knife scan to find the difference in depth between the red and green foci. Changes in this difference as a function of depth reveal the dispersion of the polymer. To implement this test, I embedded a single razor blade at an angle in depth, so that one test sample can be used to characterize beam profiles at many material depths. The results of the incremental knife scans at 7 depths between 100 μm and 280 μm of polymer showed that the optical system was aligned to within 10% of an ideal diffraction limited focus, and chromatic dispersion was consistently 30 μm, independent of depth. The consistent 30 μm focal shift agrees with ZEMAX® ray tracing simulations based on a non-dispersing polymer and thus dispersion of the polymer is negligible. In addition, the beam profiles confirm that the red and green beams are coincidently aligned and near diffraction limited throughout the depth of the polymer.
24 5.1 Experimental Sample
As explained earlier, the foundation of the fiber to waveguide coupling procedure is to loosely position the fiber in the polymer, then to precisely and automatically align the polymer waveguide to the fiber. This section describes how the fiber is positioned in the polymer which is in turn cast into the external packaging including the glass windows for lithography and microscopy access. In the standard butt-coupling method of fiber pigtailing, the 125 μm fiber cladding is held in a precision silicon V-groove directly at the fiber to waveguide interface, while a strain-relief connection to the plastic buffer is located at the package edge. In this experiment, only the strain relief is required. Its function during the packaging procedure is to position the stripped and cleaved fiber within the polymer during the curing of the matrix. As the matrix cures, it bonds to the cladding and permanently locates the fiber. An inexpensive package that can be adapted for this purpose is made for mounting loose 35 mm slides. The product from Wess Mounts is shown in Figure 9. It consists of a 5 cm x 5 cm black plastic frame with a hinge (top) and clasp (bottom). This frame holds two nominally 700 μm thick windows of glass approximately 460 μm apart. While the manufacturing tolerances of this part are not high, the lithography method does not require them. To hold the fiber at the package edge, notches are carved into the frame using a razor blade. The 900 μm buffer is brought up to the plastic package and cyanoacrylate epoxy (super-glue) is used to fasten the buffer to the package. Ideally,
25 the stripped and cleaved fiber within the package is suspended between the glass panes so that the liquid polymer will encapsulate it on all sides. Occasionally, when the slide mount is closed, the fiber is pressed against the back glass pane. While this leaves the polymer-encapsulated fiber laying on the back glass surface, it was shown earlier that the fiber location algorithm can still locate the fiber core in this extreme case. Once the fiber is secured to the package frame, the liquid precursor of the InPhase TapestryTM polymer is placed on one glass slide and the package is closed to fully encapsulate the fiber in polymer. The viscous liquid spreads to fill the entire region between the slides and flows around the fiber ends which extend several mm into the window region. Since the fiber must be completely encapsulated with no air bubbles or gaps, the liquid encapsulant must be de-gased to minimize waveguide inconsistencies. The sample is placed in a dry nitrogen environment (to reduce O2 indiffusion) while the matrix thermally cures. Multiple fibers are epoxied to the external frame and suspended in the polymer, as shown in Figure 9. The input fibers on the left side are Thorlabs 630 nm single mode patch fibers with a 4.3 μm core diameter, while the output fibers are multi-mode fibers with a 62.5 μm core. Since beam shaping techniques are not used, the direct write polymer waveguide is highly elliptical and multi-mode in the axial direction. Thus, a single mode input fiber is coupled to the elliptical polymer waveguide, which is then coupled to a multi-mode output fiber. With this design, the output coupling loss is reduced due to the elliptically shaped polymer waveguide. Thus, power measurements are representative of the coupling loss at the single mode
26 fiber and waveguide loss. This demonstrates the fundamental location and waveguide placement technique presented here. With these fibers embedded and located somewhere in the polymer, the sample is then placed on the lithographic stage to be scanned and exposed.
Figure 9: The fully mounted experimental sample after writing and curing. Two 630 nm single mode fibers are embedded in polymer on the left side of the part and two multi-mode fibers are embedded on the right side.
Once the polymer sample is mounted on the high precision 3D stages, the automated location algorithm is run as discussed in Chapter 4 on both the input and output fibers. Three dimensional coordinates of both fibers are stored in the computer, so that a waveguide can be written in the photopolymer that accurately couples to the embedded fibers. The part is translated in the Z dimension by 30 μm to
27 compensate for longitudinal chromatic aberration. At this point, all alignment procedures are complete, and waveguide writing can begin. Waveguide writing occurs when the shutter controlling the green beam shown in Figure 2 is opened and the stages translate the part in 3D to create a continuous index feature from the first set of coordinates representing the fiber tip to the second. When writing waveguides in the photopolymer, the beam is placed behind the edge of the fiber so that the overexposure that occurs during the acceleration of the stages does not affect the polymer waveguide. In the samples presented here, the stages accelerate at 10 mm/s2 to 1 mm/s along the axis of the fiber, which is approximately the X direction of the stages. The remaining dimensions (Y and Z) accelerate at a much slower rate (typically 30 μm/s2) to 1 mm/s to create smooth 3D S-bends reducing bend losses in the polymer waveguide. After writing, the sample is flood cured to permanently set the written waveguides and bleach the photopolymer. Figure 10 is a differential index contrast (DIC) microscope image showing a single mode fiber input with a polymer waveguide near the fiber core. This waveguide was aligned manually using visual feedback in transmission, with a misalignment on the order of 6 μm. Incident power of the writing beam at the objective was 400 μW. Total losses in the fiber can be caused by coupling losses and waveguide losses. The coupling losses occur at both the input and output fibers, though they are significantly greater at the input fiber due to its relatively small core diameter. Lateral misalignment of 6 μm at a 4.3 μm core would result in a minimum loss of approximately 33.8 dB for Gaussian modes, and more for the elliptically shaped polymer waveguide mode. Polymer waveguide losses have not been
28 quantified in the InPhase HDS3000 material used, at this time scatter losses and absorption losses are not known. Waveguide bend losses in this sample were not significant due to the fact that the S-bend offsets were small, approximately 830 μm in the Y dimension and 200 μm in the Z dimension, so stage acceleration in these dimensions did not play a large role. The total loss including single mode to waveguide coupling, polymer waveguide losses, and waveguide to multi-mode coupling was approximately 28 dB, indicating loss is dominated by lateral misalignment, and a waveguide alignment of better than the estimated 6 μm.
Figure 10: Differential Index Contrast (DIC) microscope picture of a waveguide written near the single mode 4.3 μm core, aligned by hand using the algorithm presented in this paper, with a misalignment of approximately 6 μm and an output loss of approximately 28 dB.
Misalignment was significantly decreased by automating the alignment procedure, as shown in Figure 11 and Figure 12. In these examples, the transverse
29 misalignment of the waveguide was approximately 3 μm at the input fiber, producing a minimum coupling loss of approximately 8.4 dB. Figure 11 is a DIC image with the brightfield sensitivity significantly increased, showing both the phase contrast of the fiber and written waveguide, but also the output light being coupled from the fiber to the polymer waveguide. This waveguide was written using 125 μW of power at the objective lens. Figure 12 is a differential scanning transmission microscope image of a waveguide written at 250 μW, showing the phase contrast of both the fiber and waveguide. The dark spot in the region of the fiber is a result of index writing in the photopolymer during stage acceleration. Since this large index feature only surrounds the fiber, the guided light is not affected.
Figure 11: Differential index contrast microscope image of a polymer waveguide coupled to a single mode fiber with the input laser on. The waveguide was aligned using the computer controlled location algorithm demonstrated in Chapter 4.
30
Figure 12: Differential scanning transmission microscope image of a 4.3 μm core single mode fiber with the waveguide written from it.
A macroscopic view of the coupling shown in Figure 11 is shown in Figure 13. Translation in the Y and Z dimensions are transversely offset to reduce bend radius losses. The apparent bend losses in the polymer waveguide are due to acceleration values in both Y and Z being 0.5 mm/s2, corresponding to a bend radius of approximately 8 mm. These acceleration values have since been adjusted to 30 μm/s2, corresponding to a radius of approximately 250 mm. Despite the relatively sharp corners, the total loss for this sample was 20 dB, including coupling and waveguide losses.
31
Y translation ends Z translation begins
Z translation ends Y translation begins
Figure 13: Polymer waveguide coupling to the single mode input fiber, incurring bend radius losses, and coupling light into the output multi-mode guide.
Figures 10, 11, and 12 show uniform photopolymer properties around the edges of the fiber, indicating low shrinkage resulting in no detectable birefringent stress or delamination. This enables reliable writing up to the limit of the lithographic process, which from Figure 3 is r·tan(NA/n)=12.7 μm. Since the reliable writing distance from the fiber is less than the Rayleigh range of the fiber, z0=π/λ·w02 = 23 μm, relatively small coupling losses result. Even in the case where the NA of the writing beam is significantly increased to NA=0.6, the lithographic limit to approaching a fiber becomes 26.4 μm, still resulting in low coupling losses. Also apparent in those figures is the ability to produce a waveguide that is accurately aligned to a loosely positioned embedded fiber. We estimate the accuracy
32 of the positioning in these initial samples to be within 3 μm of the core. This could be improved by increasing the NA of the objective for higher resolution in both the transmission and reflection microscope. In addition, further development of the computer program incorporating servo locking could interpret the microscope signals with improved accuracy and interpolate between points leading to better than micron accuracy. Circularizing the written waveguide structures through optical and chemical techniques are currently being investigated to improve the structure of the waveguide produced by the lithographic technique presented here. This is an integral part of producing a viable pigtailing procedure. However, the results presented here demonstrate the potential of a polymer pigtailing device that is compatible with many existing integrated optics platforms, as well as providing a possible platform for other integrated optical components.
6. Conclusion
This paper demonstrated an accurate pigtailing method that aligns a waveguide to a loosely positioned fiber embedded in photopolymer. A simple casting process of pouring photopolymer yields completely encapsulated fibers in a high optical quality material with good adhesion and low stress surrounding the embedded optical fiber. Two precise scanning microscopes were employed to accurately locate the embedded fiber tip in three dimensions. A light induced polymer waveguide was
33 written, aligned to the tip of the fiber without the need for precise positioning of the fiber. Results show high optical quality surrounding the fiber enabling waveguides to be written up to an embedded fiber. To within our ability to detect, the guides were unperturbed, even when the cone of writing light was partially intercepted by the fiber end. Positioning in 3D was achieved using an automated location and writing program, demonstrating an estimated positioning error of 3 μm. In this experiment, light was coupled from a single-mode input fiber, across 3 cm and 3 dimensions, into a positionally offset multi-mode output fiber. A key advantage of direct-write lithography platform is its flexibility, since a mask does not need to be fabricated. This “maskless” lithography is therefore compatible with low production volumes where mask fabrication is inefficient in terms of cost and time. This fits most optical circuit markets today. In addition, the polymer structure can embed a variety of optical components, opening avenues to more complex optical circuitry in the future. An adapted version of the location procedure demonstrated here is expected to be extensible to most optical elements anticipated to be included in a hybrid integrated optical circuit. Telecommunications optical circuits require passive elements such as beam splitters, Faraday rotators, thin-film filters, anisotropic or photorefractive crystals, while also requiring active elements such as laser diodes. Each of these elements can be embedded in the photopolymer and interconnected with polymer waveguides. Further, the photopolymer creates index changes due to incident light, and can therefore support the creation of phase holograms enabling
34 holographic optical elements (HOEs) to be part of the polymer optic toolkit. Since photopolymer can integrate a diverse set of optical elements, and each of these elements can be interconnected without the need of complex alignment procedures, it is an ideal material for integrated optics. Photopolymer provides a means to decrease optical circuit size and cost while simultaneously increasing reliability.
35 References 1. Whitney, Peter S., “Hybrid integration of photonic subsystems,” IEEE Electronic Components and Technology Conference, 578-582 (2002). 2. Boyd, J. T. and S. Sriram, “Optical coupling from fibers to channel waveguides formed on silicon,” Applied Optics 17, (6), 895-898 (1978). 3. Bostock, R. M., J. D. Collierz, R.-J. E. Jansenz, R. Jonesz, D. F. Moorey and J. E. Townsendz, “Silicon Nitride Microclips for the kinematic location of optic fibres in silicon V-shaped grooves,” J. Micromech. Microeng. 8, 343-360 (1998). 4. Hauffe, R., U. Siebel, K. Petermann, R. Moosburger, J.-R. Kropp, and F. Arndt, “Methods for Passive Fiber Chip Coupling of Integrated Optical Devices,” IEEE Transactions on Advanced Packaging 24, (4), 450-455 (2001). 5. Gibson, B., S. Huntington, and J. Love, "Self-aligning method of fiber-towaveguide pigtailing," Opt. Lett. 30, 2858-2860 (2005). 6. Sullivan, C. T., B. L. Booth, and A. Husain, “Polymeric waveguides,” Circuits Dev. Mag. 8, 27-31 (1992). 7. Booth, B. L., “Low Loss Channel Waveguides in Polymers,” Journal of Lightwave Technology 7, (10), 1445-1453 (1989). 8. Kewitsch and A. Yariv, “Self-focusing and self-trapping of optical beams upon photopolymerization,” Opt. Lett. 21, (20), 24-26 (1996). 9. Dorkenoo, K., O. Crégut, L. Mager, F. Gillot, “Quasi-solitonic behavior of selfwritten waveguides created by photopolymerization,” Opt. Lett. 27, (20), 17821784 (2002). 10. Yoshimura, T., J. Roman, Y. Takahashi, W.-C. V. Wang, M. Inao, T. Ishitsuka, K. Tsukamoto, S. Aoki, K. Motoyoshi, W. Sotoyama, “Self-Organizing Lightwave Network (SOLNET) and Its Application to Film Optical Circuit Substrates,” IEEE Trans. on Comp. and Pack. Tech 24, (3), 500-509 (2001). 11. Hirose, N., O. Ibaragi, “Optical Solder Effects of Self-Written Waveguides in Optical Circuit Devices Coupling,” Electronic Components and Technology Conference, 268-275 (2002). 12. Yonemura, M., A. Kawasaki, S. Kato, M. Kagami, “Polymer waveguide module for visible wavelength division multiplexing plastic optical fiber communication,” Opt. Lett. 30, (17), 2206-2208 (2005). 13. M. Schnoes, B. Ihas, A. Hill, L. Dhar, D. Michaels, S. Setthachayanon, G. Schomberger, and W. L. Wilson, “Holographic data storage media for practical
36
systems,” in Proc. SPIE, Practical Holography XVII and Holographic Materials IX, 5005, 2003, pp. 29-37. 14. Richard T. Ingwall and David A. Waldman, Holographic Data Storage, H.J. Coufal, D. Psaltis, G.T. Sincerbox, (Eds.), Springer-Verlag, New York, 2000, Springer Series in Optical Sciences VOL. 76, Chp.: Photopolymer Systems, pp. 171-197. 15. Booth, B.L., Polymers for Integrated Optical Waveguides. “Polymers for Electronic and Photonic Applications,” AT&T Bell Laboratories, 549-599 (1993). 16. M. Schnoes, B. Ihas, A. Hill, L. Dhar, D. Michaels, S. Setthachayanon, G. Schomberger and W. L.Wilson, “Holographic Data Storage Media for Practical Systems,” Proc. SPIE 5005, 29-37 (2003). 17. Davis, K. M., K. Miura, N. Sugimoto, and K. Hirao, "Writing waveguides in glass with a femtosecond laser," Opt. Lett. 21, 1729–1731 (1996). 18. Nolte, S., M. Will, J. Burghoff, A. Tuennermann, “Femtosecond waveguide writing: a new avenue to three-dimensional integrated optics,” Appl. Phys. A 77, 109-111 (2003). 19. A. C. Sullivan, M. W. Grabowski, and R. R. McLeod, "3D direct-write lithography into photopolymer," Appl. Opt., doc. ID 71409 (posted 1 September 2006, in press). 20. Neil, M., R. Juskaitis, M. Booth, T. Wilson, T. Tanaka, S. Kawata, “Active aberration correction for the writing of three-dimensional optical memory devices,” Appl. Opt. 41, (7), 1374-1379 (2002). 21. Ams, M., G. D. Marshall, D.J. Spence, and M. J. Withford, “Slit beam shaping method for femtosecond laser direct-write fabrication of symmetric waveguides in bulk glasses,” Optics Express 13, (15), 5676-5681 (2005). 22. Ayres, M. R., R. R. McLeod, "Scanning transmission microscopy using a PSD," Appl. Opt., doc. ID 69812 (posted 10 July 2006, in press). 23. Kawata, Y., R. Juskaitis, T. Tanaka, T. Wilson, S. Kawata, “Differential phasecontrast microscope with a split detector for the readout system of a multilayered optical memory,” Appl. Opt., 35, 2466-2470 (1996). 24. Pawley, J. Fundamental Limits in Confocal Microscopy. In Handbook of Biological Confocal Microscopy. Pawley, JB (Ed) New York, Plenum Press, 1937, 55-67.
