Open Access
Add to BibSonomy
Abstract
Asymmetric optical multimode splitters based on optical waveguides were fabricated in glass substrates using a field-assisted ion exchange process. Accompanying simulations, conducted to study the light propagation, revealed the possibility to realize asymmetric splitters based on waveguides with different width. In the exchange process, broad mask openings in the blocking layer are compared to those consisting of closely spaced parallel lines with various widths. The waveguide profiles of the resulting splitters were recorded and the optical losses and splitting ratios were determined for a wavelength of 850 nm. Additionally data transmission tests were conducted and showed the suitability of the splitters for a bandwidth of 28 GBit/s.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Integrated optics constitutes an increasingly important field of research. Its possible applications range from neuromorphic processors [1] and quantum computing [2] to biosensors [3] and optical data transmission [4,5]. One established technology platform for integrated optics, which is especially well suited for data transmission purposes, is the ion exchange in glass. A variety of passive and active optical components has been demonstrated using this technique [6]. This involves singlemode and multimode components. The former are mainly used in the telecom sector over long distances and have also made inroads into other data transmission applications. Multimode, on the other hand, plays an important role in the short-range datacom sector, for example in local area networks. Due to the lower alignment requirements, the installation is more economical.
However, the realization of some multimode components, including asymmetric splitters, have only been possible with significant drawbacks, including high optical losses [7]. These asymmetric splitters are interesting for the monitoring of information transmission in datacom applications, where only a small fraction of the total radiant power in the waveguide is required. Additionally they might be used in cascading tap couplers with other than $2^n$ output ports.
2. State of the art
Detailed descriptions and analysis of the ion exchange in glass can be found in various publications [8,9]. Due to two different mechanisms, the replacement of ions in the glass with a different species leads to a local change of the refractive index, which enables the creation of light guiding structures (waveguides). The first effect is the different ionic radius, which causes a local deformation of the glass matrix. A large ion replacing a smaller one forces the structure around it to expand, which usually lowers the refractive index. A smaller replacement ion causes a collapse of the structure around it, leading to a higher index. This ion size effect also leads to stresses in the glass structure, which usually induce birefringence [10]. The second effect, which is usually dominant, is the difference in electronic polarizability. According to the Lorenz-Lorentz equation [11] a higher polarizabilty results in a higher refractive index. This effect can be described with the following equation [12]:
where $\Delta n$ denotes the refractive index change and $n$ the absolute refractive index. With the particle concentration of the invading ion species $C_{B}$ and the electric permittivity of vacuum $\epsilon _0$. Additionally $\alpha _B$ and $\alpha _A$ represent the polarizabilty of the invading and the native species respectively.In most glasses the native ion species is Na$^+$, while there are many different ions available as invading species. Ag$^+$ is widely used, because its high polarizability allows for a large refractive index increase. Since both ions are similar in size, we neglect the birefringence effect. The introduction of Ag$^+$ is usually facilitated by diffusion at elevated temperatures, which allows Ag$^+$ from a AgNO$_3$ melt to enter the glass substrate and Na$^+$ inside the glass to leave it. This thermal ion exchange can be supported by an external electric field, which forces the positively charged ions into the glass. The external field accelerates the diffusion process and reduces the time necessary for the exchange process. Additionally the field leads to steeper Ag$^+$ concentration and thus refractive index gradients.
Disregarding the ion size effect, Eq. (1) can be used to estimate the maximum increase in refractive index for our case: $\Delta n_{max} \approx 0.08$. This value is corroborated by the experiment (see Supplement 1, Fig. S1(a)). For the creation of channel waveguides it is necessary to localize the exchange process to certain areas of the substrate surface, where the waveguides should be located. This can be achieved by applying a blocking layer to the surface, which is then structured using photolithography, to allow localized diffusion. We recently presented a two-dimensional model to describe this whole process [13].
After this first ion exchange step, waveguides are present directly at the surface of the glass substrate. However, for practical purposes it is necessary to transfer them deeper into the glass, to insulate them from surface effects, which would cause high optical losses. Additional process steps are necessary to achieve this. First the blocking layer is removed. Afterwards a second ion exchange is conducted over the whole substrate surface. This exchange uses Na$^+$ ions from a NaNO$_3$ melt, which are introduced into the glass. An electric field support is used to push the Ag$^+$ ions, which form the waveguide, from the surface deeper into the glass. Due to additional isotropic diffusion processes the concentration gradient of the waveguide flattens out, forming a gradient-index (GRIN) profile. Correspondingly the maximum refractive index change decreases by a factor of about 2.5 in our case (see Supplement 1, Fig. S1(b)).
3. Asymmetric multimode splitters
3.1 Design of mask structuring
For the creation of unidirectional asymmetric multimode waveguide splitters we based our design on the following basic premise. If the two output arms of the splitter have different widths, the radiant power will spread unevenly between those two arms, with the wider waveguide carrying a larger share of the aggregate power.
The arms of the splitter follow two mirror symmetric paths (see Fig. 1). These paths start in the same point on the input side and diverge until they reach their final distance (pitch), which was chosen to be 250 $\mathrm {\mu }$m to allow for the attachment of optical fibers. To connect the input and output ports, the paths follow a continuously differentiable function to minimize losses. This function is defined piecewise in four intervals with a constant input segment, followed by a circular arc with a 100 mm radius and a length of 1 mm along the optical axis (z-direction in Fig. 1) as the splitting section. The circular arc was chosen to ensure a swift separation of the two splitter arms. The next section is a fifth degree polynomial with a length of 12 mm to avoid abrupt changes in curvature and connect it to the constant output segment.
Fig. 1. Simulated refractive index distribution in the splitter structure with the different sections denoted on the right.
To realize a different diameter of the output arms, a single mask opening of 2 $\mathrm {\mu }$m width was used to create the narrow arm while the mask structure of the wide arm has been varied to gauge the impact of mask patterning. In Fig. 2, the two mask structures, we used can be seen in detail. One of them has a broad mask opening without patterning (a), the other is patterned (b). The exact shape of the patterning was guided by two principles. First the minimum reliable feature size of the photolithographic process we used, was 2 $\mathrm {\mu }$m. Additionally electric field distortions described by Mrozek [14] may lead to undesired concentration maxima at the edges of the waveguide. To counter this phenomenon, we increased the width of the mask openings in the middle. For the same reason, two additional 3 $\mathrm {\mu }$m wide openings running parallel to the two splitter paths, each at a distance of 250 $\mathrm {\mu }$m from the nearest mask opening were added to the patterned openings.
Fig. 2. Dimensions of the utilized mask structures in $\mathrm {\mu }$m. In the case of a broad mask opening (a) the lithographic mask (dark green) is interrupted by one single wide opening. In the patterned case (b), additional masked areas subdivide the mask opening into several parallel lines.
To examine the assumption of realizing asymmetric multimode splitters based on waveguides with different widths, a whole splitter structure was simulated, with the mask design shown in Fig. 3. Based on the diffusion model described in Supplement 1 a stack of about 150 two-dimensional Ag$^+$ concentration maps were calculated with COMSOL Multiphysics [15]. Using the assumption of a linear relationship between this concentration and the refractive index as described by Findakly [12], refractive index maps can be computed, as shown in Fig. 1. It can be seen, that the narrow arm has a higher maximum refractive index. This might lead to more light being guided in this arm, than expected by the waveguide width alone and thus limit the asymmetry of the splitting [16].
Fig. 3. Mask structure with patterned openings for the splitter fabrication. The openings for the narrow arm are shown in magenta, while those for the wide patterned arm are colored in cyan.
To examine the light propagation throughout the splitter, a simulation using the scalar finite-difference beam propagation method (BPM) [17] was conducted. We implemented this method in the Python language to simulate light distribution in the integrated optical chip. The result can be seen in Fig. 4, where the power has been integrated along the axis perpendicular to the substrate surface. The simulation was conducted at a wavelength of 850 nm and yields a splitting ratio of about 75:25. This supports the assumption, that it is possible to manufacture asymmetric multimode splitters based on waveguides of different width.
Fig. 4. Light propagation through an asymmetric splitter simulated with BPM. Note the logarithmic color scale.
3.2 Experiments
For the experiments we used glass wafers with a thickness of 1.5 mm and a diameter of six inches. The glass that was used has a chemical composition, which was specially tailored for ion exchange experiments. It is similar to BGG31 [18].
A titanium layer of 150 nm was deposited on the wafer surface and subsequently structured via a photolithographic process. Afterwards the first ion exchange step was performed. For this purpose the wafer was placed between two baths of molten AgNO$_3$ salt. To realize the electric field support, a DC voltage was applied to the two baths by placing a wire in each salt bath and connecting both to a voltage source.
Following this first ion exchange the titanium mask was completely etched off the wafer surface and a second exchange was performed with a NaNO$_3$ melt. To reduce the risk of an electrical breakdown, the voltage was lowered in three stages. The parameters for both exchanges are listed in Table 1.
Table 1. Exchange parameters for both ion exchange steps
After the ion exchange process was completed, the wafer was diced into several chips. Their edge faces were polished to allow optical access to the input and output of the waveguide structures located on the chip.
3.3 Waveguide profiles
In Fig. 5 the comparison between waveguides derived from a broad (a) and a patterned mask opening (b) can be seen. The images show the Ag$^+$ concentration map of the wide arm of each splitter, which were recorded with backscatter SEM [19]. A significant difference in shape is noticeable, the waveguide created from a broad mask opening is more convex and extends to a greater depth, while the patterned mask opening yields an oblate cross-section.
Fig. 5. Ag$^+$ concentration in the cross section of a waveguide created with a broad (a) and a patterned mask opening (b). The images were obtained using backscatter SEM.
3.4 Splitter characterization
The optical inputs and outputs were coupled with graded-index multimode fibers (Ø50 $\mathrm {\mu }$m Core / Ø125 $\mathrm {\mu }$m Cladding, OM4, 0.200 NA). A detailed description of the coupling process can be found in Supplement 1. This enabled the measurement of parameters relevant to data transmission. Thus, the excess loss and the splitting ratio of the entire fiber coupled devices were determined using LED and VCSEL light sources with 850 nm, the corresponding results are shown in Table 2.
Table 2. Optical properties of fiber coupled asymmetric multimode splitters
From these measurements it is clear, that light transmission from a VCSEL source suffers consistently lower excess losses than from the LED source. This can be attributed to the shape of the modal field excited with the different light sources. While the VCSEL excites mainly modes of low order, which are transmitted near the center of the waveguide respectively the coupling fibres, the LED additionally excites modes of higher order, which propagate in the outer zones [20]. Since the shape and the diameter of the ion exchanged waveguides in the glass cannot be perfectly matched to the attached glass fibers, the higher modes incur a significantly higher loss at the coupling site. A direct comparison of the structures shows that the patterned one has lower attenuation than the broad one. This may indicate a closer fiber match, in particular for the junction from the wide arm to the attached fiber. This might also explain the shift in the splitting ratio. Additional differences in the waveguide shape, especially in the splitting segment, cannot be ruled out. The difference in splitting ratio between the experimental values and the simulation in subsection 3.1 is probably also due to these factors. Another factor, which was not addressed in the simulation, but might play a role in the diffusion process is a partial blocking of ion diffusion in the second exchange step, which has been described in [21].
To test whether the patterned optical splitter meets the high demands of modern data links in terms of bandwidth, the bit error rate (BER) was measured using a VIAVI ONT-800 Optical Network Tester. An intensity modulated non-return-to-zero signal (2$^{7}$-1 pseudorandom binary sequence) was transmitted at 28 GBit/s using an 850 nm VCSEL. The optical input power was lowered in various steps and the proportion of incorrectly transmitted bits at both outputs was counted. The results are shown in Fig. 6. To give a reference, a 1 meter graded-index multimode fiber (Ø50 $\mathrm {\mu }$m Core / Ø125 $\mathrm {\mu }$m Cladding, OM4, 0.200 NA) was measured as well with this setup. Compared to the fiber, the narrow arm of the tested splitter shows a degradation of the signal integrity as expected. It has a BER, which is about three times higher for an optical input power of 8 dBm. However, even at lower optical input power, the BER is significantly lower than the forward error correction threshold of $5 \cdot 10^{-5}$ required for short distance optical communication [22].
Fig. 6. Signal integrity for a standard multimode fiber and both output arms of the asymmetric multimode splitter for different optical input powers at 28 GBit/s. In addition to the measured values, the corresponding exponential fits are shown.
4. Conclusion
The fabrication of asymmetric optical multimode splitters via field-assisted Na$^+$/Ag$^+$ exchange in glass has been demonstrated. Simulations were conducted based on the beam propagation method, which suggest that asymmetric power splitting can be achieved by varying the optical waveguide width. The exchange parameters for the corresponding experiments were chosen to minimize adverse effects of space charge zones forming underneath the mask. Splitters based on broad and patterned mask openings were fabricated and compared. A significant difference in the waveguide profiles was measured. This demonstrates, that mask patterning offers an additional degree of freedom in the design of ion exchanged optical waveguide structures in glass. The two splitter types were coupled to optical fibers and tested. Both exhibit asymmetric power splitting, however it can be expected to achieve even more uneven splitting ratios by using a finer mask patterning. Excess loss was lower in the splitter created from a patterned opening. This can be explained by closer fiber match on the output side. Additional research is necessary to reduce losses further, for example by adding tapered waveguides to the input and output ports of the splitters to achieve a better fiber match [23]. Furthermore, BER measurements showed that transmission rates of 28 GBit/s are possible with minimal losses.
Funding
Bundesministerium für Bildung und Forschung (03FH008PX5).
Acknowledgments
Major gratitude is given to Julian Schwietering and Norbert Arndt-Staufenbiel from Fraunhofer Institute for Reliability and Microintegration IZM in Berlin. They supported us with the RNF measurements and were very open to a scientific exchange of ideas.
Furthermore, we greatly appreciate the efforts of Mirko Abt and ADVA Optical Networking SE, who provided the measurement equipment and expertise for the data transmission tests.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this study are not publicly available at this time but may be obtained from the authors upon a reasonable request.
Supplemental document
See Supplement 1 for supporting content.
References
1. Y. Shen, N. C. Harris, S. Skirlo, M. Prabhu, T. Baehr-Jones, M. Hochberg, X. Sun, S. Zhao, H. Larochelle, D. Englund, and M. Soljacic, “Deep learning with coherent nanophotonic circuits,” Nat. Photonics 11(7), 441–446 (2017). [CrossRef]
2. S. Slussarenko and G. J. Pryde, “Photonic quantum information processing: A concise review,” Appl. Phys. Rev. 6(4), 041303 (2019). [CrossRef]
3. C. Chen and J. Wang, “Optical biosensors: an exhaustive and comprehensive review,” The Analyst 145(5), 1605–1628 (2020). [CrossRef]
4. S. Aleksic, “The future of optical interconnects for data centers: A review of technology trends,” in Proceedings of the 14th International Conference on Telecommunications, O. Dobrijevic and D. Matija, eds. (2017), pp. 41–46.
5. H. Schröder, J. Schwietering, G. Böttger, and V. Zamora, “Hybrid photonic system integration using thin glass platform technology,” J. Opt. Microsystems 1(3), 033501 (2021). [CrossRef]
6. J.-E. Broquin and S. Honkanen, “Integrated photonics on glass: A review of the ion-exchange technology achievements,” Appl. Sci. 11(10), 4472 (2021). [CrossRef]
7. P. Karioja, S. Tammela, A. Tervonen, and S. Honkanen, “Passive fiber optic bus using bidirectional integrated optic bus access couplers,” Opt. Eng. 34(9), 2551–2559 (1995). [CrossRef]
8. A. Tervonen, B. R. West, and S. Honkanen, “Ion-exchanged glass waveguide technology: A review,” Opt. Eng. 50(7), 071107 (2011). [CrossRef]
9. J.-E. Broquin, “Ion-exchanged integrated devices,” in Integrated Optics Devices V, vol. 4277 of Proceedings of SPIEG. Righini and S. Honkanen, eds. (2001), pp. 105–117.
10. A. Brandenburg, “Stress in ion-exchanged glass waveguides,” J. Lightwave Technol. 4(10), 1580–1593 (1986). [CrossRef]
11. H. A. Lorentz, “Ueber die Beziehung zwischen der Fortpflanzungsgeschwindigkeit des Lichtes und der Körperdichte,” Ann. Phys. 245(4), 641–665 (1880). [CrossRef]
12. T. Findakly, “Glass waveguides by ion exchange: A review,” Opt. Eng. 24(2), 242244 (1985). [CrossRef]
13. D. Schaeffer, D. Klenkert, J. Stauch, F. Brand, W. Foss, J. Schwietering, P. Wachholz, and M. Kufner, “Simulation of field-assisted ion exchange in glass regarding the space-charge density and pseudo-mixed-alkali effect,” Appl. Opt. 60(22), 6632–6638 (2021). [CrossRef]
14. P. Mrozek, E. Mrozek, and T. Lukaszewicz, “Side diffusion modeling by the explicit consideration of a space-charge buildup under the mask during strip waveguide formation in the Ag+-Na+ field-assisted ion-exchange process,” Appl. Opt. 45(4), 619–625 (2006). [CrossRef]
15. A. B. Comsol, “Comsol multiphysics v6.0®,” www.comsol.com (2022).
16. Y. Shani, C. H. Henry, R. C. Kistler, R. F. Kazarinov, and K. J. Orlowsky, “Integrated optic adiabatic polarization splitter on silicon,” Appl. Phys. Lett. 56(2), 120–121 (1990). [CrossRef]
17. G. Lifante Pedrola, Beam Propagation Method for Design of optical Waveguide Devices (John Wiley & Sons, Inc, 2015), Chap. 2.
18. N. Fabricius, H. Oeste, H.-J. Guttmann, H. Quast, and L. Ross, “BGG 31: A new Glass for Multimode Waveguide Fabrication,” European Fibre Optic Communication & Local Area Networks Exposition6, 59–62 (1988).
19. K. Heinrich, “Interrelationships of sample composition, backscatter coefficient, and target current measurement,” Adv. X-Ray analysis 7, 325–339 (1963). [CrossRef]
20. X. Chen, S. Bickham, J. Abbot, J. Coleman, and M.-J. Li, “Multimode Fibers for Data Centers,” in Handbook of Optical Fibers, G.-D. Peng, ed. (Springer, 2019), pp. 41–97.
21. B. Pantchev, P. Danesh, and Z. Nikolov, “Field–assisted ion exchange in glass: The effect of masking films,” Appl. Phys. Lett. 62(11), 1212–1214 (1993). [CrossRef]
22. K. Li, X. Chen, J. E. Hurley, J. S. Stone, and M.-J. Li, “High data rate few-mode transmission over graded-index single-mode fiber using 850 nm single-mode VCSEL,” Opt. Express 27(15), 21395–21404 (2019). [CrossRef]
23. J. C. Campbell, “Tapered waveguides for guided wave optics,” Appl. Opt. 18(6), 900–902 (1979). [CrossRef]
