Integration of Electrooptic Polymers in Optoelectronic Devices

Akzo Nobel Electronic Products Inc., 250 C Twin Dolphin Drive,. Redwood City, CA ... of EO polymer materials to highly integrated waveguide devices. O...
0 downloads 0 Views 740KB Size
Chapter 30

Integration of Electrooptic Polymers in Optoelectronic Devices

Downloaded by STANFORD UNIV GREEN LIBR on April 22, 2013 | http://pubs.acs.org Publication Date: August 11, 1995 | doi: 10.1021/bk-1995-0601.ch030

R. Lytel, A. J. Ticknor, and G. F. Lipscomb Akzo Nobel Electronic Products Inc., 250 C Twin Dolphin Drive, Redwood City, CA 94065

Organic electro-optic (EO) polymer materials offer some new opportunities in integrated optics. This paper reviews the application of EO polymer materials to highly integrated waveguide devices.

Organic electro-optic (EO) polymer materials offer potentially new opportunities in integrated optics for high-performance interconnections. The electronic (1) EO effect in organic materials yields moderate EO coefficients, low dispersion, and low dielectric constants (2). EO polymer materials have been modulated flat to 40 GHz and exhibit few fundamental limits for ultrafast modulation and switching. Polymeric integrated optic materials also offer fabricationflexibility.The materials are spin-coatable into high quality, multilayer films, and can be patterned, metallized, and poled. Channel waveguides and integrated optic circuits can be defined by the poling process itself (3), by photochemistry of the EO polymer (4,5), or by a variety of well understood micro-machining techniques. Multi-layer integrated optic waveguide structures can be fabricated in much the same manner as Si-substrate, multilayer multichip modules. To date, EO polymer materials have been used to fabricate high-speed Mach-Zehnder modulators (6), directional couplers (7), Fabry-Perot etalons (8), and even multitap devices (9). The potential impact of the application of EO polymer materials to highly integrated optical waveguide devices is particularly intriguing with respect to parallel processing systems. High-performance interconnections within optical multichip modules, backplanes, and optical connectors within massively parallel computers can greatly enhance the performance potential as such systems evolve to higher bisection bandwidths, wider data busses, and clock rates above 100 MHz. The incorporation of EO polymer materials into multilayered interconnection substrates built on wafers is of critical importance in the development of dense interconnection networks for modules, backplanes, and connectors. In these applications, EO polymer waveguide switches permit electronic devices to communicate via a waveguide network by driving ultralow-capacitance loads with logic levels signals. To meet standard processing and packaging methodologies of standard electronic components, new EO polymer materials with extremely high thermal stability are required. Initial development proved promising(70,) and today,

0097-6156/95/0601-0414$12.00/0 © 1995 American Chemical Society In Polymers for Second-Order Nonlinear Optics; Lindsay, G., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1995.

30.

LYTEL ET AL.

Electrooptic Polymers in Optoelectronic Devices 415

development continues throughout the world. These materials do not yet meet the EO performance of their lower temperature brethren but are expected to achieve comparable performance in the future. Meanwhile, there are numerous applications of EO polymers which do not require thermal stability at 350° C but instead demand stability and reliability in device performance at temperatures up to 80° C. Such applications include digital data communication links. Very good performance should be achievable with current materials, and even higher performance can be expectedfrommaterials that should be available in the near future.

Downloaded by STANFORD UNIV GREEN LIBR on April 22, 2013 | http://pubs.acs.org Publication Date: August 11, 1995 | doi: 10.1021/bk-1995-0601.ch030

Integrated optical transmitter arrays Optical interconnections have traditionally been implemented as ultra-highperformance serial links. Serialization of all the communications in a massively parallel architecture would be foreboding to most system designers. A primary advantage of EO polymer integrated optics is the ability to take a reasonable parallel approach to optical interconnections. EO polymer waveguide switches and taps occupy only 0.02 mm area and may be tiled to yield large scale integration. To an IC output pin, the switches load the output drivers primarily as only a small capacitive load (100 fF switch, plus via and tape, bump, or wire-bond capacitance). A single switch may be driven by 1-5 volts and be used to switch optical power for routing and distribution. Reflectors can turn light within a plane, providing high packing density and tiling of cellular elements. Finally, subtle modifications in waveguide dimensions, shapes, and polymer refractive indices can produce simple integrated lenses, which will be extremely useful as alignment features for fibers, lasers, and receivers. Active optical transmitter arrays can be fabricated on planar surfaces such as silicon wafers using these methods. They are active because the energy required for the data stream is provided primarily by an external laser and the transducers appear as high-impedance loads to the electrical signals. Arrays could then be integrated into modules as could receiver arrays of similar configuration to provide optical I/O that could be coupled between modules using fiber-ribbon arrays with self-aligning terminals. There are also good approaches being pursued in various projects to use directly-modulated laser arrays for similar applications. These have many of the advantages of parallel optical interconnection and can even be made to look to the signal source like high-impedance transducers through the use of active buffers and separate power supplies. The primary penalty of this approach is however power consumption and particularly the resultant on-module heat generation. As the data rates increase and the energy-per-bit stays steady (to maintain signal integrity), the laser drivers must switch greater and greater energy. This means they must switch more current more rapidly into the impedance of the laser circuit. The on-module power will increase approximately with a linear contribution from the frequency and a quadratic contribution from the peak power. To increase the power in the modulator array, one needs simply to increase the incoming laser power. Since the modulated current in the transmitter array starts out small, its linear increase with frequency to the power consumption has much less impact and the 'higher power' component only contributes a linear increase to power consumption instead of the quadratic increase in the equivalent direct-drive term. Furthermore, the continuous optical energy supplying the modulator array is supplied by fiber and the powerdissipating element (the laser) can be located remotely to the module to isolate its contribution to heating from the active circuitry. The figures in Table 1 show the power required for an eight channel, one-gigahertz/channel optical transmitter using direct modulation vs external modulation for three different power levels. The contributions are separated in terms of modulated (RF) current and low-frequency (DC) current so considerations of switching noise can also be made. External 2

In Polymers for Second-Order Nonlinear Optics; Lindsay, G., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1995.

416

POLYMERS FOR SECOND-ORDER NONLINEAR OPTICS

modulators are modeled as capacitative loads, while the directly modulated laser figures are an average of commercially available specs. Table 1. Power Required for 1 GHz-by-8 Optical Transmitter Arrays

Downloaded by STANFORD UNIV GREEN LIBR on April 22, 2013 | http://pubs.acs.org Publication Date: August 11, 1995 | doi: 10.1021/bk-1995-0601.ch030

METHOD

- 3

dBm

0 dBm

3

dBm

DIRECT MODULATION • RF power, mW • DC power, mW

20 128

80 128

320 128

EXTERNAL MODULATION • RF power, mW • DC power, mW

11.2 53

11.2 105

11.2 210

Reducing modulator size The length of device required for an EO modulator to fully switch or modulate an optical signal from logic-level electrical signals is of order 1 cm in any material system so far demonstrated. The width required is only a few tens of microns. This not only leads to significant space required for an EO transmitter array, but also means the electrodes necessary to drive the modulators must be transmission lines at frequencies around 0.3 GHz or above. There is incentive to decrease the active device length not only to occupy less area, but also to allow the drive electrodes to be designed as simple RF stubs for significantly higher impedance. One approach to this is to increase the EO sensitivity of the materials, and this has long been one of the great promises of polymer EO materials and is being pursued by many groups around the world. Another approach is to make active complementary taps (11). In this approach, a large amount of optical energy is fed into the modulator, and the modulator redirects a small fraction of that energy into one or the other of the two outputs. Since only a small portion of the light is actually being modulated, the active device length can be much shorter, of order 1 mm in current materials, for a given switching voltage. This lets the electrical load of the modulators be modeled as stub elements into the multi-GHz range, rather than requiring transmission lines as in the longer modulators, thereby reducing the power dissipated in both the modulator and the output driving it. The signal level is indicated by which of the two outputs that most of the output signal is directed towards. Since a relatively constant fraction of energy is removedfromthe supply beam, it can be used as the supply for subsequent complementary modulators with little additional noise. Since the signal is represented by the relative levels in two complementary channels, there can be greater tolerance to the additional noise that is introduced. Eventually, the supply beam will become too noisy and/or too weak for further tapping and must be dumped. This means that such a scheme makes much less efficient use of optical energy than other schemes and is only practical in systems where large amounts of single-mode light can be supplied economically, i.e. where wall-plug power is plentiful and power dissipation is the primary limitation on improved system performance. EO polymer waveguide taps may be cascaded and tiled to provide an active optical tap array. This device permits efficient conversion of many electrical signals to optical data streams through the use of a single, CW laser. Tap nodes are small and compact and make very efficient use of the light. An immediate impact to the

In Polymers for Second-Order Nonlinear Optics; Lindsay, G., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1995.

30.

L Y T E L ET AL.

Electrooptic Polymers in Optoelectronic Devices

417

Downloaded by STANFORD UNIV GREEN LIBR on April 22, 2013 | http://pubs.acs.org Publication Date: August 11, 1995 | doi: 10.1021/bk-1995-0601.ch030

system designer is that the IC output drivers need only charge a 100 fF capacitor (plus via and TAB), for any interconnection length and at any frequency up to the link limit. This significantly reduces on-chip line driver power dissipation, and thereby reduces switching noise in the supply current and, consequently, the number of power and ground pins supplying a high-performance IC such as a CMOS microprocessor. An active-tap network can be used with existing chip designs since it requires no more amplification or power than is already present on the chip, but allows the chip designer to scale back output drivers, freeing up more chip real estate and power for logic functions, and reducing delay and noise. Active taps eliminate the problem of how to mount, align, and control hundreds of semiconductor lasers, and utilize instead a few fiber pigtails to CW sources that may be located elsewhere in the system. Conclusion Our analysis indicates that, for data rates of 50 MHz and above, EO polymer optical interconnects would require significantly less power than electrical interconnects for either CMOS or ECL devices. Other performance advantages to the system designer include simple fanout, impedance-matched lines at all frequencies, low noise, low power, low propagation delay, high density, and simple layout. To verify these claims, good stable sources of EO polymers and an acceptable low-loss process are required. We anticipate that such devices will be built and demonstrated in the coming three years. What about other applications of EO polymers? We feel compelled to comment on the potential for polymers to gain the commonplace utilization of optical semiconductors in optoelectronics. Planar polymer waveguide technologies have the ultimate potential to gain widespread use in many electronic and fiber-optic system applications. Passive components could find use as splitters, couplers, multiplexers, and parallel array connectors in trunk, local loop, wide-area, and local-area networks. Applications of EO modulators beyond external modulation of lasers include fast network configuration switches, optical network units in Fiber-to-the-Home (FTTH), modulator arrays for data networks, filters, couplers, multiplexers, digital-analog and analog-digital converters, and pulse-shapers. The market potential for planar polymer waveguides is very large due to low wafer processing costs and potential to achieve low-cost single-mode fiber-attach and packaging. This means polymers may compete well with other technologies in conventional optoelectronic applications. Polymer technologies offer some features that are also available with some (MQW semiconductor) but not other (crystalline) technologies. With polymers, high levels of integration have been demonstrated by using multiple levels of waveguides (12) as well as in-plane and out-of-plane mirrors (13). The potential for low-cost manufacturing, packaging, and assembly arises from the demonstrated ability to perform hybrid integration of single-mode components using lithographically-defined registration techniques. Advanced products include processor multichip modules with high-bandwidth interfaces between CPU and second-level cache, optical mesh routers for massively parallel computers, and 8-12 bit, high-speed A-D's. Cost, reliability, performance, and availability are the main drivers for obtaining and sustaining long-term interest in polymers by systems users. Polymer reliability is seen by customers as a major issue, particularly for EO poled polymers. Reliability needs to be proved with extensive test data of the packaged components, following the well-known standards for telecom and electronic components, in general. It is important to note that laser diodes have achieved success in the market, despite their propensity for drift, low-yields, limited lifetime, and failure. The

In Polymers for Second-Order Nonlinear Optics; Lindsay, G., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1995.

Downloaded by STANFORD UNIV GREEN LIBR on April 22, 2013 | http://pubs.acs.org Publication Date: August 11, 1995 | doi: 10.1021/bk-1995-0601.ch030

418

POLYMERS FOR SECOND-ORDER NONLINEAR OPTICS

market has accepted "correction" methods for laser diode performance, such as thermo-electric coolers, drift compensation circuitry, and elaborate packaging because the total cost of a laser transmitter has been reduced to acceptable levels in many cases. Similar techniques could be applied to polymer devices but will increase cost and reduce reliability. On the other hand, absolute stability of the EO coefficient and waveguide properties to thermal excursions and illumination is essential and not correctable with electronics. This may be an important issue for EO polymers. What about competing technologies? For passive technologies, glass is the main competitor. L1NDO3 and GaAs waveguides, and direct laser modulation provide competition for EO polymers. Underlying all of this is the inertia of electronic systems designers to change their solutions from wires to fiber-based systems: Whenever possible, electronic solutions will be thoroughly examined and selected, if economically feasible and practical. However, high-end communication in all markets is moving toward utilization of the bandwidth offered by optical fiber, and thus the growth of markets for all optoelectronic devices is inevitable. With further development, electro-optic polymers have the potential to faroutdistance inorganic materials in figures-of-merit, and, in fact, already do in some key properties, such as length-bandwidth products. Polymers are not likely to ever exhibit insertion loss as low as glass for passive devices. However, intrinsic performance of polymers, measured against other materials, is not sufficient for judging the potential of the technology. Overall production costs, balanced against performance and reliability, will determine the utilization of polymer waveguide technologies. References 1. S.J. Lalama and A.F. Garito, "Origin of the Nonlinear Second-order Optical Susceptibilities of Organic Systems", Phys. Rev. A 20, 1179 (1979). 2. K.D. Singer and A.F. Garito, "Measurements of Molecular Second-order Optical Susceptibilities Using D C Induced Second Harmonic Generation", J. Chem. Phys. 25, 3572 (1981). 3. J.I. Thackara, G.F. Lipscomb, M . A . Stiller, A.J. Ticknor and R. Lytel, "Poled Electro-optic Waveguide Formation in Thin-film Organic Media", Appl. Phys. Lett. 52, 1031 (1988). 4. G. R. Mohlmann, W.H. Horsthuis, C.P. van der Vorst, "Recent Developments in Optically Nonlinear Polymers and Related Electro-Optic Devices," Proc. SPIE 1177, 67 (1989). 5. M . B . J . Diemeer, F . M . M . Suyten, E.S. Trammel, A . McDonach, M . J . Copeland, L.J. Jenneskens and W.H.G. Horsthuis, Electronics Letters 26 (6) 379 (1990). 6. D.G. Girton, S. Kwiatkowski, G.F. Lipscomb, and R. Lytel, "20 GHz Electro­ optic Polymer Mach-Zehnder Modulator", Appl. Phys. Lett. 58, 1730 (1991). 7. R. Lytel, G.F. Lipscomb, M. Stiller, J.I. Thackara, and A.J. Ticknor, "Organic Integrated Optical Devices", in Nonlinear Optical Effects in Polymers. J. Messier, F. Kajzar, P. Prasad, and D. Ulrich, eds., N A T O ASI Series Vol. 162 (1989), p. 227. 8. C.A. Eldering, A . Knoesen, and S.T. Kowel, "Characterization of Polymeric Electro-optic Films Using Metal Mirror/Electrode Fabry-Perot Etalons", Proc. SPIE 1337, 348 (1990). 9. T.E. Van Eck, A.J. Ticknor, R. Lytel, and G.F. Lipscomb, "A Complementary Optical Tap Fabricated in an Electro-optic Polymer Waveguide", Appl. Phys. Lett. 58, 1558 (1991). 10. J.F. Valley, J.W. Wu, S. Ermer, M. Stiller, E.S. Binkley, J.T. Kenney, G.F. Lipscomb, and R. Lytel, "Thermoplasticity and Parallel-plate Poling of Electro-­

In Polymers for Second-Order Nonlinear Optics; Lindsay, G., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1995.

30.

LYTEL ET AL.

Electrooptic Polymers in Optoelectronic Devices

419

optic Polyimide Host Thin Films", Appl. Phys. Lett. 60, 160 (1992); G.R. Mohlmann ed., Proc. SPIE 2025 (1993) 11. R. Lytel, G.F. Lipscomb, E.S. Binkley, J.T. Kenney, and A.J. Ticknor, "Electro-optic Polymers for Optical Interconnects", Proc. SPIE 1215, 252 (1990). 12. T.A. Tumolillo, Jr. and P.R. Ashley, "Multilevel Registered Polymeric Mach­ -Zehnder Intensity Modulator Array", Appl. Phys. Lett. 62, 3068 (1993). 13. B.L. Booth, "Optical Interconnection Polymers", in Polymers for Lightwave and Integrated Optics". L.A. Hornak ed. (Marcel Dekker, New York), 1992, pp. 231-266. Downloaded by STANFORD UNIV GREEN LIBR on April 22, 2013 | http://pubs.acs.org Publication Date: August 11, 1995 | doi: 10.1021/bk-1995-0601.ch030

RECEIVED April 24, 1995

In Polymers for Second-Order Nonlinear Optics; Lindsay, G., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1995.