Lithography-originally a method of creating art by means of a greasy crayon and a stone-has in vastly more-sophisticated form become the standard technique for the fabrication of micron-scale devices. Now researchers at Lucent Technologies-Bell Labs (Murray Hill, NJ) are taking the rubber stamp through the same quantum leap.
Rubber stamp prints microscopic features
Lithography-originally a method of creating art by means of a greasy crayon and a stone-has in vastly more-sophisticated form become the standard technique for the fabrication of micron-scale devices. Now researchers at Lucent Technologies-Bell Labs (Murray Hill, NJ) are taking the rubber stamp through the same quantum leap.
By pouring ordinary liquid silicone over a mold made of lithographically patterned silicon, then baking and peeling the solidified rubber away, the researchers have created reusable stamps with features down to 20 nm in size. The stamps can be used in the fabrication of many kinds of optoelectronic devices-including flat-panel displays, plastic lasers, light-emitting diodes (LEDs), and tunable fiber Bragg gratings-as well as in the making of complete electronic circuits made up of thin-film organic transistors and other components.
"This all got started from a science study of molecules that bond to surfaces in self-assembled monolayers," says John Rogers, physical chemist at Bell Labs. Such a monolayer, when inked onto a rubber stamp and pressed against a thin layer of gold, covalently bonds to the gold and releases from the stamp, he explains. The gold is then etched-leaving a pattern-and the remaining ink is exposed to ultraviolet light and removed. The result is a gold gridwork upon which organic semiconductors or other materials can be deposited.
The technology has potential for the inexpensive manufacture of large-area flexible plastic displays, says Rogers. The Lucent group has used a rubber stamp to produce thin-film transistors integrated with a polymer semiconductor LED, forming a "smart pixel," and is working on the refinements required to fabricate a display 1 sq m in area. One advantage of silicone in this application, says Rogers, is that it can be wrapped around a cylinder to create a rolling stamp. The researchers have already made such a roller, which, at a length of a few inches, Rogers calls "small scale." Because it consists of a silicone layer only 100 µm thick wrapped around a rigid glass cylinder, the distortion created by stamping pressure is at most 1 µm/cm2.
Although the geometry of a thin-film transistor requires a feature size of 0.2 µm at certain locations, the alignment tolerance between process layers is approximately 20 µm-a combination ideal for stamping, says Rogers. He notes that the polymer semiconductor is deposited at room temperature from solution, enabling low-cost manufacture and preventing warpage of the plastic substrate.
Although lateral diffusion of the monolayer ink prevents features of less than 0.2 µm in size from being printed, an inkless silicone stamp has been used to directly pattern features of 20 nm in both plastic and sol-gel glass, useful for devices such as distributed Bragg reflectors and distributed-feedback lasers. The researchers have also created photopumped photonic-crystal lasers that use dye-doped solid plastic as a gain medium and contain photonic crystals consisting of embossed posts in triangular, square, or hexagonal geometries.
"Not only can a rubber stamp be used to stamp a cylindrical pattern on a flat surface," says Rogers, "but it can also do the reverse." He points to optical fiber as a cylindrical shape traditionally difficult to imprint with patterns. Starting with a flat stamp and rolling an optical fiber over it, the Bell researchers have imprinted a microcoil thin-film heater on the fiber that, when heated slightly, changes the fiber`s refractive index-an effect useful for programmable add/drop filters, adjustable gain flatteners, and other fiberoptic devices (see photo). By changing the angle of the fiber relative to the stamp, the researchers can imprint bands, coils, or longitudinal stripes on the fiber.
Although Rogers emphasizes that microcontact printing is strictly at the research stage, he notes that optical fibers containing stamped thin-film heaters used for tunable dispersion compensation have already successfully undergone both linear and nonlinear systems tests at transmission rates of 10, 20, and 40 Gbit/s.
Solar cells give boost to record balloon flight
Hanging in the open air beneath the gondola, a series of solar panels provided the power that helped make the first round-the-world balloon flight a success (see small photo). Twenty photovoltaic cells, built by Solarex (Frederick, MD), provided up to 60 W of electricity each, with a solar conversion efficiency of about 14%. They charged five lead-acid batteries that powered all sorts of electronic equipment aboard the flight, from navigation instruments that linked the crew to the Global Positioning System to laptop computers for e-mailing home base.
The Orbiter 3, sponsored by watchmaker Breitling (Grenchen, Switzerland), carried Bertrand Piccard of Switzerland and Brian Jones of Britain around the world in 20 days to a landing in Egypt. This was the third attempt sponsored by that company and the latest of several attempts to circumnavigate the globe in a balloon. The craft, which used both helium and hot air for lift, was 200 ft high and started out weighing about 10 tons, seven of which were propane for heating the air that helped lift it.
The solar cells had to provide more energy for the craft without adding much more weight. Standard solar cells-comprising a glass substrate, a plastic layer, the cell itself, and another layer of plastic, all placed in a frame-weigh about 15 lb each. The cells that Solarex made specially for the flight-polycrystalline silicon modules sandwiched between two sheets of plastic-weighed in at about 2 lb each.
The cells were arranged in a square configuration, four stacks of five cells each. In this shape, at least one side was always in full sunlight during the day, although those cells not receiving direct light could still convert scattered rays to power. The balloon`s flight path was high enough that overcast skies were not a problem.
For the flight to be successful, Piccard and Jones had to move as fast as they could to complete the trip before they ran out of fuel. This meant spending as much time as possible in the jet stream and minimizing the up-and-down movement that comes from heating and cooling the gas within the balloon. The photovoltaic cells played a role here, as they were used to cool the balloon during the day, maintaining a steadier altitude and cutting down on the amount of helium that had to be released to keep the craft from flying too high.
The Orbiter 3 is to be donated to the Smithsonian Institution`s National Air and Space Museum (Washington, DC). Because the landing damaged the cells that actually flew on the first round-the-world flight, Solarex has been asked to provide new ones, company spokesman Sarah Howell said. Though the cells were custom-made for Breitling, she said, "if anyone else needs them we`ll be happy to oblige."
Tunable devices target complex high-speed nets
As high-speed optical networks evolve in speed beyond the 40-Gbit/s mark and as dense wavelength-division multiplexing (DWDM) technology boosts complexity, the demands on network management devices are expected to increase dramatically-ultimately re quiring a new set of devices to accommodate fluctuations in the network.
As an example, Benjamin Eggleton, a researcher at Lucent Technologies (Holmdel, NJ), has described a WDM system in which signals must be amplified and channels must be added and dropped without affecting other channels. As one channel is dropped or added, however, amplifier performance over the other channels is likely to change slightly, leading to changes in power levels. These changes can then cause variations in the acquired nonlinear phase shift, which can alter the optimum dispersion-compensation configuration for a lightwave network. Consequently, an important thrust of new technologies presented at the Optical Fiber Communication (OFC) conference in San Diego last February was on tunable devices for dispersion compensation and wavelength equalization.
Several devices under development at Lucent and reported during the postdeadline session at OFC included an electrically tunable, dispersion-compensating fiber Bragg grating, an interferometer-based dynamic wavelength equalizer, and a channelized WDM spectral equalizer using lightwave micromachines and autonomous power regulation. Another postdeadline paper, presented by researchers from the Advanced Institute of Science and Technology and from Donam Systems Inc. (Taejon, Korea), described a dynamic erbium-doped fiber amplifier (EDFA) that used a feedback loop to provide automatic gain flattening.
New network management technology presented at OFC was not limited to research results in technical sessions, however. On the exhibit floor, representatives of General Photonics Corp. (Chino, CA) described a commercially available multichannel spectrum equalizer designed to independently control the signal level of up to eight wavelength channels either manually or electrically with a response time on the order of 10 µs.
During the OFC postdeadline session, Eggleton explained that the tunability of the fiber Bragg grating developed at Lucent was based on a thin layer of gold deposited with a tapered thickness from about 500 to 3500 Å over the grating length using electron beam evaporation (see Fig. 1 on p. 20). The variation in thickness resulted in an electrical resistance gradient. When electrical current was applied, the resistive gradient produced a linear temperature gradient and thus a linear chirp along the fiber Bragg grating.
"The magnitude of the grating linear chirp rate, and thus the dispersion, can be continuously varied during device operation by changing the current applied to the film," Eggleton said. "For the operating regime of the device the power consumption was less than 0.5 W." The researchers demonstrated the tunable dispersion-compensating FBG device in a nonlinear 20 Gbit/s single-channel system, in which they were able to maintain optimum performance over a 17-dB variation in launch power with less than 1 dB of power penalty (see Fig. 2).
Semiconductor laser cavities are self-formed
Commercially available compound semiconductor lasers are made of single-crystalline films fabricated by heteroepitaxial growth on lattice-matched substrates. The fabrication technique is both expensive and very restrictive in the type of substrates that can be used.
Researchers at the Materials Research Center of Northwestern University (Evanston, IL) have been exploring lower-cost fabrication of semiconductor lasers by nonepitaxial growth of semiconductor materials on various substrates, including silicon dioxide (SiO2).1 The researchers claim to be the first to observe discrete cavity modes of random lasers and capture images of two-dimensional random laser cavities. This demonstration opens up the possibility of using disordered semiconductor micro structures as alternative sources of coherent light emission.
The lasers developed at the center have cavities that are self-formed due to strong optical scattering in the polycrystalline films. Although scattering is usually considered detrimental to laser action because it induces loss, when strong enough it can help lasing by forming closed-loop paths for light and introducing feedback.
For example, ultraviolet (UV) laser action was demonstrated in 300-350-nm-thick zinc oxide (ZnO) polycrystalline films grown on amorphous fused-silica substrates. The films were deposited by laser ablation with a pulsed 248-nm-output krypton fluoride (KrF) excimer laser. Optical absorption measurement of the films then indicated that the band edge of ZnO was about 0.05 eV lower than that grown epitaxially on a sapphire substrate.
In the absence of any fabricated mirrors, the researchers observed lasing in the films by optically pumping samples with a frequency-tripled modelocked Nd:YAG laser emitting 15-ps pulses at 355 nm with a 10-Hz repetition rate. They found marked differences in the lasing characteristics compared to those of a conventional laser. The foremost was that laser emission from the ZnO films was observable in all directions, with the laser emission spectrum varying with the observation angle. Tests also showed that the pump intensity required to reach the lasing threshold depended on the excitation area.
Transmission-electron-microscopy (TEM) analysis illustrated the structural difference between film grown with this method and film epitaxially grown on sapphire substrates. High-resolution TEM images also indicated that the former structure is populated by defects such as threading dislocation and stacking fault, with dislocation density approaching 1012 cm-2. One benefit of the randomly oriented polycrystalline grain structure of the ZnO films, though, is strong optical scattering.
The experiments characterized the scattering mean-free path in the ZnO films using coherent backscattering. To avoid absorption, the frequency-doubled output (410 nm) of a modelocked Ti:sapphire laser at a 76-MHz repetition rate and 200-fs pulsewidth was used as the probe light. From the angular width of the backscattering cone, the scientists estimated that the scattering mean-free path is on the order of the emission wavelength of ZnO, and because of this short path, the emitted light may return to the scatterer from which it was scattered before, thus forming a closed-loop path (see figure).
The ZnO polycrystalline films produced at the Materials Research Center exhibit large optical gain, with a gain coefficient exceeding 100 cm-1 at a fluence of 2 J/cm2. Increase the pump power, and the gain exceeds the loss first in the low-loss cavities, leading to laser oscillation with lasing frequencies determined by the cavity resonances. As gain increases, the lasing action spreads to the other cavities.
- H. Cao et al., Appl. Phys. Lett. 73(25), 3656 (Dec. 21, 1998).
Keck sharpens its view
Located on Mauna Kea, HI, the twin Keck telescopes, each with a segmented 10-m-diameter primary mirror, are the largest optical telescopes in the world. As with many telescopes, instabilities caused by atmospheric turbulence and minute shifts in mechanical structure place limits on the angular resolution of Keck I and Keck II. But significant technological advances installed on the telescopes this year have pushed both atmospheric and structural instability limits back.
Keck I contains two foci: one at f/15 used predominantly for visible imaging and another at f/25 used for imaging in the near- to mid-infrared. Based on information gathered by a Shack-Hartmann-based phasing camera, the 36 segments of the primary mirror are positionally adjusted to compensate for phase errors introduced by slow structural shifts. However, although the control system holds the phase of the segments stable for weeks at a time, the phasing camera is permanently mounted at the f/15 focus, whereas the measurements most sensitive to phase take place at the f/25 focus.
Faced with this, as well as the desire to have a rapid independent check of the phasing camera, a group headed by Gary Chanan of the University of California, Irvine (Irvine, CA), developed a technique called phase-discontinuity sensing (PDS). This technique senses "piston" phase error (resulting from unwanted translation of the mirror along its normal), in which the error signal is simply the difference between inside-of-focus and outside-of-focus images of a star.
For an 0.8-m defocus, the image appears uniform with no phase difference between the segments and locally nonuniform with piston errors (see Fig. 1 on p. 26). The difference between images of opposite defocus produces a signal with an intensity proportional to phase error within a range of + or -ll/8. By moving the mirror segments in question by 0.7 of their calculated errors and then iterating the process, the researchers have created a simple and practical correction system that homes in on the desired segment positions without overshoot.
In routine operation at Keck since January, PDS allows astron omers to sense and reconfigure the phase of the primary mirror in less than a minute. The technique is limited by atmospheric turbulence to wavelengths longer than 3 µm and by diffraction effects to wavelengths less than 10 µm. "On the other hand," says Chanan, "for segmented optics in space, such as the Next Generation Space Tele scope, [the short-wavelength limit] would not apply, assuming sufficient resolution on the detector."
Keck II has received its own improvement-an adaptive-optics (AO) system that saw first light in February and has produced a stunning tenfold increase in the telescope`s angular resolution (see Fig. 2). Using a Shack-Hartmann wavefront sensor with natural guide stars, the system applies wavefront correction using a fast tip-tilt mirror and a 349-actuator deformable mirror. Peter Wizinowich, optics engineering manager at Keck and developer of the telescope`s AO system, describes how, on the very night the AO optics were trucked up Mauna Kea and installed, Keck II produced a resolved image of a binary star angularly separated by only 0.068 arcsec. Keck II can now resolve to as good as 0.034 arcsec, producing images with Strehl ratios as high as 50% in the K` spectral band.
An artificial-guide-star addition developed by Lawrence Livermore National Laboratory (Livermore, CA) is slated to be installed next year, which will allow the use of AO in regions of the sky where natural guide stars of sufficient brightness are not available. At some point, says Chanan, PDS will be combined with AO, enabling rapid tune-up for highly resolved infrared imaging.
Researchers improve disk lasers
Diode-laser-pumped solid-state disk lasers are promising as high-power sources because of their ability to remove heat efficiently from the gain medium, thereby eliminating many of the thermal-lensing problems inherent in other solid-state laser designs. Invented in 1992 at the University of Stuttgart and the DLR (German Center for Avionics and Space; both Stuttgart, Germany), the thin-disk laser design has since been licensed to 11 companies. Two of these firms have launched disk-laser products-Nanolase (Grenoble, France; see Laser Focus World, Dec. 1998, p. 15) and Jenoptik (Jena, Germany). Meanwhile, the research team at Stuttgart has been working to further improve its original laser design.
The principle of the disk laser is straightforward-the active medium is shaped as a thin disk, with its plane surface mounted onto a heat sink. This surface serves also as one of the laser mirrors so heat is removed efficiently along the optical axis of the laser. Because optical pumping is longitudinal, no transverse thermal gradient de velops in the gain medium, so thermal-lensing effects are negligible. Thermal lensing and thermally induced birefringence in the laser crystal during pumping are difficult design problems associated with conventional high-power solid-state rod lasers. In fact, other laser configurations, such as the slab laser, have evolved as a result of attempts to reduce thermal lensing.
There are complications with the disk-laser geo m etry, however. Pump radiation is not absorbed fully within the thin disk during a single pass because the dopant concentration cannot be made sufficiently high. For the pump radiation to be absorbed completely, it must pass through the disk several times. Hence, to fully exploit the pros pects of scaling the disk-laser output to high powers, the laser design re quires a careful balance of parameters to allow multiple passes of the pump light through the disk.
Ytterbium-doped YAG (Yb:YAG), which emits at 1030 nm, is currently the preferred disk material because of its high absorption of the 940-nm pump light. Other materials have been tested, however, including Nd:YAG, Nd:YVO4, and Tm:YAG. For a given pump power density, Yb:YAG disks can be made thinner than Nd:YAG. Together with its high quantum efficiency, Yb:YAG exhibits excellent optical-to-optical power-conversion efficiency, especially at low temperature. At -70°C, this conversion efficiency was measured at about 65%-it is, however, strongly dependent on temperature and drops to about 45% at room temperature. The reason for this is that the lower laser level becomes about 4.5% populated, which makes the system operate almost as a three-level scheme, so it requires a high pump-power density; this, in turn, increases the operating temperature.
To overcome these obstacles, the researchers investigated various parameter combinations. They calculated the optical efficiency as a function of the crystal thickness for different dopant concentrations and showed that optical efficiency increases with higher dopant levels (see Fig. 1 on p. 30). Another calculation showed, however, that the effect of higher dopant levels is less important than the number of passes of the pump radiation through the disk (see Fig. 2 on p. 32). As a result, the researchers designed a new pump geometry to take advantage of a single paraboloid mirror (see Fig. 3 on p. 32). This allows up to 16 passes (up from eight) so that the pump radiation is almost completely absorbed despite the low thickness of the disk (0.22 mm).
With a resonator length of 50 cm, the researchers achieved a total optical efficiency of 50% (slope efficiency 61%) at 15°C, with M2 remaining constant at less than 1.05 through the whole pumping range up to 63 W (see Fig. 4 on p. 32). With the earlier design, the same efficiency could be obtained only at -43°C. So the new design increased optical efficiency at room temperature from 46% to 58%. Such results are likely to encourage further development of disk-laser systems.
Separately, researchers at the Stuttgart Institut für Strahlwerkzeuge and LAS GmbH (Berlin) reported in a postdeadline session at the 63rd Meeting of the German Physical Society (Heidelberg; 15 March 1999) that they obtained a 7-W CW output at 515 nm from an air-cooled frequency-doubled single-frequency disk laser. These results represent an optical-to-optical efficiency of 75%.
- S. Erhard et al., "Novel Pump Design of Yb:YAG Thin Disk Laser for Operation at Room Temperature with Improved Efficiency," Proc. OSA Trends in Optics and Photonics, Advanced Solid State Lasers, Washington, DC (1999).
Under pressure, CO2 gives harmonic surprise
Choong-Shik Yoo wasn`t looking for a frequency-doubling crystal when he and his colleagues squeezed carbon dioxide (CO2) to a high pressure and heated it with a neodymium-doped yttrium-lithium-fluoride (Nd:YLF) laser. But as the process went on, they noticed their sample emitting a green light.
"We were kind of wondering, `What is this green?`" says Yoo, a physicist at Lawrence Livermore National Lab oratory (Livermore, CA). "At first we thought it was chemical luminescence."
The team members looked at the light with a spectrograph and were surprised to discover it was almost entirely at a wavelength of 527 nm. In other words, it was the second harmonic of the 1054-nm output of the Nd:YLF laser they were using to heat their sample.
Yoo`s team found that its sample, a few micrograms of CO2 in an extended solid phase (CO2-V), was converting 0.1% of the 10-W input. If it worked that well at this low power, Yoo says, think of how it could work if the input were a pulsed laser with a peak power of kilowatts or megawatts. "You can easily imagine this would be a very-high-efficiency conversion."
Yoo and his colleagues were working with Department of Energy funding on the National Ignition Facility of the Stockpile Stewardship and Management Program, trying to increase their understanding of how CO2, the most common byproduct of high-energy explosions, behaves at high pressures and temperatures, so they could improve their models of detonations. This sort of information would also be useful in understanding the chemistry in the interiors of gas giants like Uranus and Neptune. But the frequency-doubling property they discovered holds promise for producing wavelengths that are difficult to achieve by other methods.
Putting the squeeze on
Researchers placed a small amount of CO2 between two diamond tips, which they pressed together to squeeze the gas to a pressure above 40 gigapascals, equivalent to 400,000 atmospheres. At the same time, they used the Nd:YLF laser to heat the CO2 to 1800 K. The result was a sample a few micrograms in size of a CO2 polymer with hardness comparable to cubic boron nitride, the hardest known substance after diamond, which is also a form of carbon. Yoo thinks the polymer will be as good a conductor of heat as diamond, making it able to withstand high temperatures and high laser power. The CO2-V also resembles silicon dioxide, a popular second-harmonic converter.
Because the CO2 absorbs 1064-nm light poorly, the group heated it indirectly, by scattering micron-sized ruby chips in the sample, for instance, and heating those with the laser. In other attempts, they used a platinum foil or a rhenium gasket for heating (see figure on p. 36). To be sure the frequency doubling was not due to some contamination from the heating materials, the group performed 20 experiments and got the same results, regardless of the heating material.
The next step is to create bigger samples of the CO2-V and determine its properties. That, in turn, will allow researchers to start working on ways to create a metastable form, which retains its properties in more ordinary conditions, just as artificial diamond can now be made. The material Yoo`s group created maintains its structure at ambient temperature, but the pressure must be kept above 1 gigapascal.
"Diamond is also one you needed high pressure and high temperature to synthesize, but today people make it at ambient conditions, and in the future I don`t see any reason why we can`t do that with this," Yoo says.
Five technologies vie for DWDM applications
SAN DIEGO, CA-Intense interest in dense wavelength-division multiplexing (DWDM) is creating a hot market for optical subsystems. Five distinct types of DWDM modules were evident at the Optical Fiber Communication conference (OFC) on February 23-25. Each technology has its advocates, but no clear consensus has emerged so far, and companies such as Corning In corporated (Corning, NY) and JDS Fitel (Nepean, Ontario, Canada) are pursuing multiple approaches.
The fundamental problems are a set of difficult and sometimes conflicting goals. Optical systems must separate wavelength channels 200, 100, or even 50 GHz apart, with output wavelengths remaining stable and crosstalk low. System makers would like to add channels by adding new modules, so customers could "pay as they grow," but customers also want low cost per channel and compact modules with low insertion loss. In the end, they wind up balancing tough trade-offs.
One well-established approach is to combine a diffraction grating with a GRIN lens or micro-optics. Dispersive gratings are widely used in optical spectrum analyzers and can be used for DWDM, says Tom Mikes, president of American Holographic (Fitchburg, MA), which supplies the test market. "Doing high resolution in small packages is tricky," he adds, but he thinks it should be possible to produce a 32-channel unit for several thousand dollars. JDS Fitel uses diffraction gratings to provide 100-GHz resolution in high-channel-count systems where thermal stability is essential.
Corning OCA (Marlborough, MA), E-Tek Dynamics (San Jose), and JDS Fitel are among companies offering assemblies of narrow-line interference filters for both 100- and 200-GHz channel spacings. Many assemblies cascade filters, picking off one wavelength at a time, a simple architecture amenable to step-by-step expansion and low channel counts. Going to 50-GHz spacing has proved a tough problem with thin films, but some developers say the market is not yet demanding such tight spacing.
Fiber Bragg gratings can provide 50-GHz resolution and have the advantage of being assembled entirely from fiberoptic components. Fiber grating technology is developing fast, but production remains an issue. One performance issue is the need for a circulator to route both input light and the wavelength selectively reflected by the grating. Fiber gratings have side lobes and rounded transmission peaks, but thin-film filters can square off the peak and block the side lobes.
Coming on strong at OFC was the use of coupled fibers as cascaded Mach-Zehnder interferometers. Different wavelengths emerge from the two output fibers, with the wavelengths interleaved at a spacing that depends on the design (see figure on p. 40). Successive stages with broader interleaving split "odd" and "even" wavelengths until all are separated. Both ITF Optical Tech nologies (Ville St.-Laurent, Quebec, Canada) and Wavesplitter Technologies (formerly AFO, Fremont, CA) focused exclusively on this approach.
Several companies said they are making their own planar waveguide arrays, with silica waveguides on silicon substrates. Monolithic arrays are attractive for high channel counts because a single device can generate 32 or more channels, with resolution to 50 GHz. The technology promises economies of scale for high channel counts, but does not lend itself to adding a few channels at a time, and fiber-to-waveguide coupling losses remain an issue. Fiber-to-waveguide coupling losses remain an issue.
Rydberg atoms excite this IR camera
One drawback of conventional long-wavelength cameras is that the detectors have a broad spectral response. According to Marcel Drabbels at FOM, the Institute for Atomic and Molecular Physics (Amsterdam, the Netherlands), the detectors are, therefore, sensitive to broadband thermionic excitation. A second drawback of existing long-wavelength cameras is a slow temporal response.
Now Drabbels and colleague L. D. Noordam have demonstrated an IR imaging camera at 1 and 55 µm that is based on a Rydberg atom far-IR detector.1 A Rydberg atom has a high principal quantum number and is thus close to being ionized; as a result, even the small energy of an infrared photon is enough to photoionize the atom. Camera sensitivity is determined by the number of Rydberg atoms and the photo-ionization cross section of the atoms (see figure on p. 44). Spatial resolution is 300 µm at a wavelength of 1.06 µm. Effective exposure time could possibly vary from 1 ns to 1 µs.
In the IR camera, which the researchers believe has potential as both an ultrafast detector and a wavelength-selective detector, the photocathode consists of atoms excited to a Rydberg state by a sheet of ultraviolet (UV) radiation from a pulsed dye laser or Q-switched Nd:YAG laser. Infrared radiation directed through a mask onto the film of Rydberg atoms ionizes the atoms. The resulting ions are accelerated toward a position-sensitive detector consisting of a pair of multichannel plates with a phosphor screen and a charge-coupled device (CCD) camera interfaced with a PC. The result is basically an IR contact print of the mask using Rydberg atoms as the photosensitive film. Changing the UV excitation wavelength alters the principal quantum, number of the Rydberg atom, thus turning the wavelength response of the camera.
Exposure time can be controlled at the film and the position-sensitive detector. In the current design, the camera becomes sensitive to IR radiation only after the atoms have been excited to a Rydberg state by a UV laser pulse. Using a subnanosecond-pulsed laser for this step, the film can be activated within a nanosecond. Switching off the electric field across the excitation region at the desired time deactivates the Rydberg film. An alternative approach to control exposure time is to apply pulsed voltage to the microchannel plates of the detector. Based on these data, the researchers estimate that the system`s gate times can be as small as 1 ns.
Because the one-photon ionization of atoms and molecules is, in general, not strongly wavelength dependent, the camera is, in principle, sensitive to all IR wavelengths, provided the IR photon energy is larger than the binding energy of the Rydberg electron. The wavelength window is continuously tunable across a range of 1 to 100 µm. By choosing a fairly high Rydberg state, the long-wavelength cutoff of the camera can be as long as 100 µm.
- Marcel Drabbels and L. D. Noordam, Appl. Phys. Let. 74, 1797 (March 29, 1999).
EU micromachining project ends successfully
A three-year program funded under the European Union (EU) BRITE EURAM scheme to look at laser-based submicron machining has come to a successful completion. The COMPALA (cost-effective series and mass-production of high-precision microparts and optical structures in molds by laser submicron-machining) project finished in early 1999 with an impressive list of innovations, including laser-cutting widths of 10 µm in 0.2-mm metal sheet, nontapering laser-drilled holes of 10 µm diameter in 1-mm metal sheet, and the design of workstations based on high-power copper-vapor lasers (CVLs) to achieve these results (see Fig. 1). Project partners are Philips CFT and TNO Industry, (both Eindhoven, The Netherlands), Robert Bosch GmbH (Stuttgart, Germany), Oxford Lasers Ltd. (Abingdon, England), and Oxford University (Oxford, England).
The COMPALA project, coordinated through Philips CFT, aimed to develop new tools and processes for the fabrication of very small metal parts. In short runs this was accomplished by direct machining of the parts. For mass production, the CVL-based system was used to mill molds from which galvanic replicas can be made using existing technology (MIGA). The laser-based processes developed during the project include three-axis drilling, five-axis milling based on both Q-switched Nd:YAG and CVL technologies, and microstructuring of metallic masters with 140 microparts on one substrate ready for replication several thousand times. Work-handling systems were also developed, as was a prototype industrial CVL.
Copper-vapor-laser development concentrated on producing high powers with low-divergence beams in laser systems that have high reliability and low maintenance. Improvements from this project led to a CVL with a maintenance interval of 2000 hours and an availability greater than 95%. A new master-oscillator/power-amplifier configuration halved the footprint of the existing device and facilitated easy access for maintenance. The unit produced 40 W, with a pulse-repetition frequency of up to 22 kHz, beam diameter of 25 mm, and divergence of less than 100 µrad. Active power stabilization allows the output power to be locked to a user-set value of better than plus or minus 0.5% over 24 hours.
The CVL was integrated into a purpose-designed workstation, which allowed several different micromachining processes to be investigated. These included hole drilling, where holes as small as 1.5 µm were obtained, and contamination-free cutting, which is clean-room compatible. Milling processes included surface texturing of molds, which was extended to create periodic submicron structures in molds and the production of MIGA molds.
Following the successful completion of the research phase, it is anticipated that a further two years of on-site testing at end-user facilities will lead to the commercial introduction of complete systems in 2001. However, technological spinoffs will be introduced earlier. All of the partners intend to move forward with the new technology. At least one of the potential application areas for CVL-based micromachining-the dril ling of fuel-injection nozzles-seems set to become commercially viable within the next few years (see Fig. 2).
Laser aims to analyze rocks on Mars
Collecting rock and soil samples for laboratory analysis is not particularly difficult-unless, of course, the samples you`re interested in are tens of millions of miles away on Mars. All the information NASA has gathered so far on Martian geology has come either through remote sensing by orbiting spacecraft or through direct contact, as when Pathfinder rolled up and stuck an alpha-proton x-ray spectrometer (APXS) against the rock named Yogi.
Now scientists at Los Alamos National Laboratory (LANL; Los Alamos, NM) are working on an instrument for characterizing samples at an intermediate range-within about 60 ft of a Mars lander. In a three-year, $1.1 million project, LANL researchers are developing a prototype of a laser-induced breakdown spectroscopy (LIBS) instrument to fly to the red planet early in the next century.
The technology for LIBS is not new. Project leader David Cremers has been improving the technique at LANL for the last 18 years. In 1996 he and his colleagues developed a backpack-sized portable instrument for analyzing contaminants in soil. And in 1997, Coyote Mining and Environmental Instruments (Los Alamos, NM) was formed with a license from LANL to market LIBS systems for mining. But this is the first attempt to build such an instrument for remote use that is small and lightweight enough to be part of a NASA payload.
The instrument works by focusing laser pulses on a spot in order to vaporize a small amount of material. The resulting plasma gives off light that the spectrograph collects and analyzes, telling geologists the makeup of the rock or soil (see image on p. 55).
"The analyses are very quick," Cremers said. "We can do an analysis in several minutes." That compares to the several hours it took the APXS to get readings.
Tests so far have been performed with a Q-switched Nd:YAG laser emitting at 1064 nm in pulses of 10 ns at energies of less than 100 mJ/pulse. A beam expander focuses the pulses on the sample, and a spectrograph and an optical array detector process the resulting spectra. Using a target chamber that simulates the low-pressure carbon dioxide atmosphere of Mars, researchers have been able to take measurements at 19 m. That kind of stand-off distance could allow a Mars rover to sample rocks it can see but cannot approach because of rough terrain, for instance, or high on a cliff face.
"Another advantage is that all the techniques up to now have been basically passive," Cremers said. Most studies have looked only at the surfaces of Mars rocks, surfaces that might not be truly indicative of the rock`s composition because the outer layer is weathered or covered with dust. A series of 20 to 50 laser shots can bore through the weathered layer, typically 30 to 100 µm thick, to analyze the rock underneath.
The final product may not use a Nd:YAG laser if another type meets the requirements for power and size. The information that can be gleaned from the plasma does not depend on the kind of laser used to create it.
Researchers at LANL are also working on a similar instrument that might be linked with LIBS. Laser-induced-plasma ion-mass spectrometry works on similar principles, except that individual atoms vaporized from the sample by the laser are taken up and analyzed. This technique works only in the near-vacuum found on the Moon and asteroids. The atmosphere of Mars, though thin by terrestrial standards, is too dense. A Soviet mission in the late 1980s planned to use laser-induced-plasma ion-mass spectrometry on the Martian moon Phobos, but both spacecraft in that mission were lost before their work could begin.
The researchers will eventually test their instrument on a NASA rover in the Mojave desert. If NASA approves a LIBS device, it may ride on a mission as early as 2003.
Multiple wavelengths may reduce cost of LANs
High speed and low cost are the watchwords as demand for bandwidth in local-area networks (LANs) continues to grow toward 10 Gbit/s. But currently available 10-Gbit/s solutions-designed for long-haul telecommunications-are prohibitively expensive for local applications. So coarse wavelength-division multiplexing (CWDM)-achieving 10 Gbit/s through four parallel 2.5-Gbit/s channels-has emerged as a possible low-cost alternative for both enterprise backbones and some short-haul implementations. And multiple-wavelength vertical-cavity laser and detector arrays are being explored as potential enabling technologies.
Connie Chang-Hasnain has taken a leave from the University of California (UC)-Berkeley to develop components and subsystems for WDM as president and CEO of Bandwidth Unlimited (Hayward, CA). During a presentation at the Optical Fiber Communication (OFC) conference in San Diego, CA, last February, Chang-Hasnain described device technologies originally developed at UC-Berkeley that show promise for important roles two or three years hence in multigigabit LAN systems.
The multiple-wavelength, vertical-cavity surface-emitting-laser (MW-VCSEL) design is based on two distributed Bragg reflectors (DBRs) sandwiched around a quantum-well active region containing a Fabry-Perot cavity, she said. The thickness of the cavity is determined by parameters in the growth process and, in turn, determines the wavelength emission spectra (see Fig. 1 on p. 56). VCSELs represent just one of the laser-array alternatives being considered for CWDM applications, she said, but the VCSELs offer a range of advantages including single Fabry-Perot mode operation because of the short cavity; a wide range of possible wavelengths; high coupling efficiency into optical fibers; ease of integration into one- and two-dimensional arrays; capability for wafer-scale fabrication and testing; inexpensive packaging; and small geometry.
To ensure process reliability in device fabrication, the Berkeley researchers used an intelligent growth process based on laser reflectometry. A similar process also was used to develop multiple-wavelength detector arrays (see Fig. 2).
The primary advantage of CWDM is that high-resolution wavelength control is not necessary, according to Mark Schwager, product manager at Bandwidth Unlimited. "So it`s a cost function," he said. "With dense WDM you need very precise wavelength control, which means you need to have temperature control of the lasers and you need to spend a lot of time in managing those wavelengths to make sure they don`t step on one another and cause crosstalk and so on. So with coarse WDM you lose the need for all that control.
"The challenge remaining with VCSELs is getting them at the right wavelengths," Schwager said. "For the enterprise area where you`re only going 500 m, making the 850-nm VCSEL is fairly straightforward. What hasn`t been proven yet is moving on to longer-distance, single-mode implementation."
Laser-generated x-rays watch atoms move
Studying the interaction of atoms in materials is not easy, especially when you consider that they move 1 trillion times faster than the blink of an eye. Now, however, researchers at the University of California-San Diego (UCSD; San Diego, CA) report they can capture this movement using laser-based x-ray detection equipment. Their method directs ~30-fs pulses of 800-nm light generated by a Ti:sapphire laser into a gallium arsenide (GaAs) crystal.
According to Kent Wilson, director of the testing laboratory involved, the 200-mJ pulses, each lasting about 30 fs, heat a thin layer on the surface so fast that the atoms do not have time to move. Acting much like a microscopic sonar gun, the subsequent expansion of this hot, high-pressure layer sends a very fast sound pulse traveling deep into the crystal.1
At the same time, the laser also directs 30-fs, 50-mJ pulses onto a moving copper wire in a vacuum to produce x-ray pulses of very short duration. Once delivered to the crystal surface at a set time after atom movement begins, the x-rays diffract from the moving atoms, penetrating roughly 2 µm into the bulk along the surface normal, to produce what Wilson calls a movie of the moving sound pulse (see figure).
In essence, the researchers were able to follow a 100-ps acoustic pulse as it propagated through the x-ray penetration depth in the GaAs crystal. So far as the researchers know, this is the first time anyone has been able to collect this type of data on bulk dynamics in such an absorbing material.
Although the equipment will work for a variety of materials and natural processes, testing used the semiconductor material because it is available in large samples of very high crystalline quality. In addition, its physical param-eters are well documented, as are research data related to the material`s ultrafast electronic properties.
"What makes this technology promising is that in a relatively short time you may be able to watch in detail how molecules assemble and disassemble during chemical reactions," said researcher Andrea Cavalleri. "We should be able to time-resolve the motions of single atoms and take pictures of the motions in these elementary chemical reactions."
Wilson believes that a variety of applications could benefit from real-time exploration of such atom interactions. Physicists would have a tool to explore the behavior of solid-state materials, and chemists could study how atoms move in biologically important materials.
- C. Rose-Petruck et al., Nature, 310 (March 25, 1999).
Incorporating news from O plus E magazine, Tokyo
Single-photon light source spurs quantum investigations
STANFORD, CA-Researchers at the Yamamoto Quantum Fluctuation Project have successfully developed a semiconductor device that can emit single photons at a constant frequency. The team fabricated the device at the Stanford University Center for Integrated Systems and the Edward L. Ginzton Laboratory. It consists of a gallium aluminum arsenide chip covered by a regular array of posts, each potentially a source of photons (see Fig. 1). Single photons are produced in quantum wells near the base of each post.1
Quantum computers, quantum communications devices, and high-precision photodetectors all need single-photon light sources. Conventional single-photon light sources operate by decreasing laser light intensity until a maximum of one photon is emitted per time step. However, the number of photons contained in laser light is not constant because it is subject to quantum fluctuations. If the light intensity is decreased to a maximum of one photon per time step, then in most cases no photons are emitted. This approach is extremely inefficient.
In contrast, this new device can reliably emit a single photon (or even two or three photons) at every time step, even at frequencies of 10 MHz. In other words, quantum fluctuations are controlled, and a squeezed photon-number light source can be manufactured.
Each post has a minuscule tunneling p-n junction that is subject to various voltages. The key point is that there are three quantum wells. The shape of each well is determined by the voltage applied to the junction (see Fig. 2).
The electron in the quantum well on the n side of the junction enters the middle well because of the quantum-mechanical-tunneling effect. The applied voltage is set so that the energy level of the middle quantum well is perfectly matched to the electron energy. Because of this, the electron tunnels with a very high probability-the resonance tunneling effect. The shape of the potential changes once the electron enters the middle well, and the resonance conditions are no longer met. Therefore, new electrons cannot enter the middle well.
Next, the applied voltage is changed so the Coulombic potential of the electron, the structure of the semiconductor, and the energy levels of the quantum wells (determined by the applied voltage) work together to match the energy of the positive hole. The positive hole then goes from the p-side well to the central well due to resonance tunneling.
The electron and the positive hole reunite in the middle well and produce a photon, which is emitted from the top of the post. The timing of the tunneling of the electron and the positive hole can be completely controlled by the adjusting the applied voltage. Therefore, single photons can be reliably emitted at precise frequencies. Because this mechanism is similar to revolving doors, project members have named this device the single-photon turnstile element.
The turnstile can produce a stream of 1 to 10 million photons per second, while the researchers say an improved version in the works has the capability of increasing the transmission rate 200-fold. A next-generation turnstile design will replace the quantum well in the post with a quantum dot. To detect the regulated beams of the turnstile, a visible-light photon counter is in development with the capability of detecting individual photons nine times out of ten without noise.2,3
The work was sponsored by Japan`s Exploratory Research for Advanced Technology (ERATO) program.
- J. Kim et al., Nature 397, 500 (Feb. 11, 1999).
- J. Kim et al., Appl. Phys. Lett. 74(7), 902 (Feb. 15, 1999).
- S. Takeuchi et al., Appl. Phys. Lett. 74(8), 1063 (Feb. 22, 1999).
Tunable filters improve in the blue
SENDAI-At the engineering department of Tokoku University a research group led by Tatsuo Uchida has developed a tunable color filter that exploits the double refractive properties of nematic liquid crystals. The transmission wavelength can be continuously changed by adjusting the applied voltage. This device can extract specific wavelengths from two-dimensional image information and can be used in imaging spectrometers.
The structure of this narrowband color filter is fundamentally the same as that of a birefringent, or Lyot, filter, consisting of three liquid-crystal cells sandwiched be tween four polarizers (see figure). The liquid-crystal cells are electrically compensated bend (ECB) cells in which double-refractive properties change when placed in an electric field. Simply put, an applied voltage can change the difference in refractive index between the light polarized parallel to this direction and the light polarized perpendicular to this direction. This difference in refractive index creates a phase difference between the two polarizations of light known as retardation.
For example, in the first cell sandwiched between two polarizers, the polarization axis is on a 45° angle relative to the optical axis. The transmission intensity is high only when the wavelength is such that the phases of both polarized beams coincide at the second polarizer.
However, this effect alone does not make a narrowband bandpass filter-which is where the second cell comes into play. This cell has twice the thickness of the first cell and twice the retardation effect. Therefore, the transmission conditions are met even when the wavelength is half that of the first cell. When both filters are used, the refractive index is represented by the product of the two curves so that the bandpass region narrows. The Lyot filter repeats this pattern. The ratio of thicknesses between the first filter and the third filter is 1.5, not the conventional 4. The first polarizer is positioned perpendicular to the other polarizers. When the same voltage is applied to all the filters, the retardation effect is proportional to the thickness of the filters.
The research group has previously developed a prototype filter with proper retardation properties by using three cells of the same thickness with voltage controlled individually across each cell. This new construction is, however, far simpler and compensating for the temperature dependence of liquid crystal mobility is much easier.
The properties of the device as a color filter have been tested. For an applied voltage of 1.4-1.9 V, the central transmission wavelength changes continuously from 480 to 680 nm. Even in the worst case-the blue region-peak transmission was approximately 40%. (Because polarizers are being used, transmission can never be greater than 50%.) Tunable color filters using liquid crystals have been marketed in the past; however, one of the main disadvantages was low transmission in the blue region.
The group also has tested the device using optically compensated bend (OCB) cells instead of ECB cells. The results were similar and, because OCBs have very fast reaction times and low-incident-angle dependence, they have great potential in high-speed applications.
Interferometric beam increases resolution
Continuing improvements in optical lithography have made possible the volume manufacture of semiconductor wafers with feature sizes as small as 0.18 µm. Due to demands for ever-increasing computing power, the pressure is on to develop techniques that extend the usefulness of optical lithography to even smaller linewidths.
Limits in optical materials and lens design require that the newest techniques include "tricks" to increase the resolution to beyond that of the standard lithographic system that uses chrome-on-glass masks and on-axis illumination. Some of these tricks include phase-shift masks, off-axis illumination of various sorts, and optical proximity correction. Now augmenting this lineup is a technique called imaging interferometric lithography (IIL).
Developed by Steven Brueck and his group at the University of New Mexico (Albuquerque, NM), IIL increases the spatial frequency cutoff of a standard lens by angularly shifting the zero-order component of the light passing through the photomask far enough from the axis that it is blocked by the lens pupil stop, after which the zero-order component is reintroduced under the lens. By doing so, high-frequency information that is normally lost passes through the lens pupil and yet combines properly with the zeroth order. For standard orthogonal semiconductor chip geometries, three exposures are required for IIL: an x-offset exposure, a y-offset exposure, and a standard on-axis exposure to provide the low-frequency information.
In an initial experiment done with simple optics, Brueck has confirmed the predicted boost in resolution given by IIL. Starting with two achromatic doublets and a square pupil stop, he created an afocal optical system with 2X reduction and a numerical aperture (NA) of 0.04 (see Fig. 1). Although far below the 0.7 NA of a leading lithographic lens, the experiment was designed to demonstrate a relative resolution improvement that would apply as well to more-sophisticated lenses. Using collimated light from a single-frequency, 364-nm argon-ion laser for illumination, the system imaged a 6 x 6-mm area of a standard chrome-on-glass photomask containing both dense and isolated lines. The result was a more-than-doubled resolution when used in the IIL mode (see Fig. 2 on p. 71).
Brueck envisions that, in use, four beams with plane wavefronts will be directed in underneath a lithographic lens at a high angle with respect to the lens axis to provide the zeroth order. Four beams rather than two are required, he explains, due to the requirement that through-focus image position be maintained. The ultimate resolution of such a system depends on the features, he notes, ranging from l/4 for simple gratings to l/3 for complex geometries. This translates into 65 nm for complex features imaged at 193 nm, potentially pushing resolution into "next-generation" territory with the use of proven fused-silica optics.
"It`s only one of a number of approaches that the industry is looking at [to increase resolution]," Brueck cautions. He notes that depth of focus for IIL is as low as l/2-the greatest obstacle for the technique. But, he adds, "the installed base for optical lithography is enormous. Optics will [be used] for as long as it will work. Let`s see what optics can do."
Lidar sensor sees forest and the trees
Biomass, the amount of living matter, such as vegetation, in a given area, is important information for scientists studying global warming. Growing forests absorb carbon from the atmosphere and can slow warming, while deforestation returns carbon to the air. The latest version of a remote-sensing instrument can now give researchers information on biomass with unprecedented accuracy and coverage.
The Laser Vegetation Imaging Sensor (LVIS), designed and operated by scientists at the NASA Goddard Space Flight Center (Greenbelt, MD), is an airborne laser altimeter that maps not only the top canopy of trees, but also the terrain below and the vegetation in between. J. Bryan Blair, the Goddard engineer who developed LVIS (pronounced "Elvis"), has worked on several such instruments and says the technique is not a new one-laser altimetry goes back to the early 1960s. But LVIS records the pulse shape (waveform) for each laser shot and has a wider swath and greater precision than previous devices.
An important innovation is that LVIS uses a single detector and digitizer to record both the outgoing and returning laser pulses, eliminating a major source of range ambiguity. The ground return signal is located by searching the recorded signal buffer with real-time software. The transmitter, a solid-state, diode-pumped Nd:YAG oscillator-only laser (Cutting Edge Optronics; St. Louis, MO), emits >5-mJ, 10-ns pulses at 1064 nm and at repetition rates up to 500 Hz. Galvanometers drive mirrors that control the output beam and the receiver field of view. The telescope has an 8-in.-diameter aperture, and the detector is a silicon avalanche photodiode.
The digitizer begins recording slightly before the output pulse and continues for 120 µs (18 km in range), well after the latest possible return from the surface. Whereas Blair`s previous altimeter, the Scanning Lidar Imager of Canopies by Echo Recovery (SLICER), recorded waveforms that were accurate to only 11 cm, the range and waveform information from LVIS is accurate to better than 1 cm. "Digitizer-only ranging and waveform recording is actually a really huge improvement," Blair said.
The new instrument is the only altimeter to record the outgoing pulse shape, which allows researchers to compensate for its effects on the range data caused by changes in the shape of the outgoing pulse. Whereas SLICER scanned a 50-m swath, LVIS scans swaths of 1 km. The instrument uses 80 beams in its standard mode, and the swath pattern can be programmed to vary depending on a researcher`s needs.
Ralph Dubayah, a geographer from the University of Maryland (College Park, MD), worked with Goddard scientists last year to map 500 sq km of dense forest around La Selva Biological Research Station in Costa Rica in footprints 25 m in diameter. By analyzing the data collected by LVIS, Dubayah, Blair, and colleagues were able to estimate the density of leaves and branches in addition to knowing the shape of the canopy and the terrain (see figure).
Meanwhile, Goddard and Maryland scientists compared their data with a previous high-resolution study that mapped part of the same area with a small-footprint, range-only laser altimeter. The two sets of data matched within 1 m horizontally and 10 cm vertically. The next step will be the Vegetation Canopy Lidar (VCL), a satellite-based system to be launched by NASA in mid-2000. The VCL won`t be able to do the detailed surface mapping that LVIS does-the speed of the satellite will prevent that. Instead, it will sample forests to measure biomass on a global scale.
Lasers map material profiles
No machined part is completely smooth. To varying degrees, its surfaces will have microscopic peaks, valleys, and curves that are either inherent in the material or the result of some manufacturing process. Thus, the designer, manufacturer, or even the end user of that component must ask, "How smooth is smooth enough?"
The right surface texture can mean the difference between a part that functions successfully or not at all. For example, without enough deep valleys, an automotive engine component may seize up because its surface does not retain enough of the oil flowing through it to lubricate efficiently.
The problem is, the smoother the part must be, the higher the associated costs usually are to produce required surfaces. The tools to measure surface profile or roughness also vary in both sophistication-technology can be noncontact or contact-and cost. One possible alternative is a laser surface profiler developed at Sandia National Laboratories (Albuquerque, NM) that, while relatively low in cost compared to other surface-profiling technology, still provides a fairly sophisticated qualitative image of surface roughness (see figure).
In contrast with systems that measure surface height to evaluate roughness, the laser surface profiling technique directly maps the slope of a sample surface. According to scientists An-Shyang Chu and M. A. Butler, engineers mount the sample to be measured on a translation stage and then focus a beam from a helium-neon laser on the surface with an angle of incidence of about 45°. The beam reflects, then impinges on a position-sensitive detector at an angle of incidence equal to the angle of reflection.
This reflected beam is scanned across the sample surface, and changes in the surface-normal direction of an angle d deflect the reflected beam through an angle of about double that. It is this change in direction of the reflected beam that the position-sensitive detector measures. The data are stored in computer memory and then converted to an 8-bit gray-scale representation. The end results are images that qualitatively show variations in surface slope.
The system filters the spatial noise of the laser beam profile by passing the beam through a 10X objective and then a 25-µm pinhole. The beam is then recollimated to ~2.5 cm in diameter by a lens with a 15-cm focal length so that it matches the f/number of the final focusing-lens set. The scientists use a combination of an achromat and an aplanatic meniscus to focus the laser beam on the sample. Combined with the expanded incident beam, this lens combination produces a diffraction-limited spot size of about 5 µm. A microscope with a long working distance is used to visually optimize the spot size on the sample.
The laser profiling system reportedly can measure changes in surface slope from 0.01°-4° with a spatial resolution on the order of the laser spot size, or roughly 5 µm. From these data, the scientists note that the vertical resolution of the surface topology is estimated to be less than 1 nm. There are some practical factors limiting measurement performance, though. One is that the sample surface must be either a reflective or semireflective type. For more details, contact Chu at email@example.com.
Tomography finds flaws in composite parts
Optical fiber makes microsoldering device
As electronic components get ever smaller, the tools for manipulating them must shrink as well. Researchers at the University of South Florida (USF; Tampa, FL) have now created a microscopic version of a soldering iron, consisting of a laser and an optical fiber.
The researchers used laser-heated pedestal growth techniques, applying a carbon dioxide laser to melt the tip of a seed rod and grow a crystal of undoped yttrium aluminum garnet (YAG). Then, with the YAG fiber as a seed, they grew a tip on the end doped with 5% neodymium (Nd; see photo). Normally, a Nd:YAG laser contains 1% neodymium to generate the laser light efficiently.
"In our case, we want the opposite. We want as much [light] converted to heat as possible," said Nicholas Djeu, a professor of physics at USF who described the microsoldering device in a recent issue of Applied Optics.
The tip Djeu and fellow researcher Jonathan H. Herringer made was able to heat up to nearly 1800°C. "Of course, for the soldering application, you don`t have to get it very, very hot," Djeu said, just a few hundred degrees to melt whatever soldering agent is chosen. "For what we did, we only put in a few hundred milliwatts of power."
The pair tested the device by attaching a piece of gold wire, 25 µm in diameter, to a 35-µm-wide gold trace on a Kapton substrate with indium as the soldering agent. They made a piece of fiber about 6 cm long that tapered to a tip 20 µm in diameter. They put the polished input end of the fiber into a standard fiberoptic coupler, which they connected to a piece of fused-silica fiber coupled to the output of an argon-ion laser. The test was performed under a microscope in a nitrogen atmosphere that prevented the indium from oxidizing.
With the tip, the researchers picked up a small chip of indium and placed it on the gold trace, then shot 0.26 W of laser power into the fiber to melt the indium into a drop. They put the wire next to the drop, then heated it with 0.32 W of laser power. The result was a microscopic solder joint.
The team used an argon-ion laser because it was handy and because the argon line at 415 nm just happens to be absorbed by the tip they made, Djeu said. In practice, he said, researchers performing microsoldering would probably use a diode laser.
Although soldering generally requires a tip 20 or 30 µm in diameter, Djeu and Herringer were able to fabricate tips as small as 30 nm in diameter by etching the tips with a heated acid mixture. The potential for high heat and small tips means the technology could also be useful for nanofabrication. "We haven`t really tried anything other than the microsoldering," Djeu said.
But other researchers working on similar principles have experimented with microscopic etching. For instance, some have tried putting dents in a surface with a tapered glass fiber for recording purposes. In principle, Djeu said, his technique should allow researchers to modify surfaces on a very small scale.
Intracavity bench tests KTP crystals
The nonlinear optical crystal potassium titanyl phosphate (KTP) is frequently used to convert the 1064-nm fundamental wavelength of a Nd:YAG laser to its 532-nm second harmonic. Its excellent nonlinear properties make KTP particularly suitable for generating green light during continuous-wave (CW) operation up to the watt output-power level.
To increase its frequency-doubling efficiency, a KTP crystal is generally placed inside the laser cavity. In this case, however, the harmonic output is highly sensitive to the presence of intracavity components that induce losses. This property can be used by KTP-crystal manufacturers to evaluate crystal performance and homogeneity. Test results from an intracavity-doubling setup can help the manufacturers improve the quality and performance of future crystal production.
F. Balembois, P. Georges, and A. Brun from the Institut d`Optique Theorique et Applique (Orsay, France) have collaborated with H. Albrecht and D. Lupinski from Cristal Laser Compagnie (Chaligny, France) to develop a laser test bench for testing KTP crystals that are used for intracavity second-harmonic generation. Based on a diode-pumped laser source, this bench allows the researchers to identify and catalogue different KTP defects as well as compare their setup to a standard extracavity test bench.
The experimental bench is built around a diode-pumped Nd:YVO4 laser (see Fig. 1 on p. 84). The diode laser delivers 4 W of 808-nm light from a
1 x 500-µm output facet. The emitted beam is collimated and focused by two objective lenses, and a 3X magnification anamorphic prism pair reshapes the pump beam in the direction parallel to the diode junction. This optical arrangement allows the researchers to obtain a focused 170-µm-diameter spot inside the Nd:YVOz,4. To allow testing of KTP homogeneity, the test crystal is mounted on two translation stages that move the crystal in a direction perpendicular to the optical cavity axis.
Tests show a 10% root-mean-square instability in the green output power from an uncoated KTP crystal and a 1.5% instability from a coated crystal. According to the researchers, this high instability in the output from the uncoated crystal is due to the high refractive index of the KDP, which produces an 8% reflection from each uncoated face and leads the crystal to act as an intracavity Fabry-Perot etalon-thus reducing the number of longitudinal modes.
It is easiest and most economical for KTP manufacturers to measure conversion efficiency on uncoated crystals. To improve output-signal stability during testing of uncoated crystals, a piezoelectric transducer behind the fourth mirror is fed by a square-wave electrical signal and oscillates at 1 kHz. The green intensity fluctuations are thereby reduced to 2% rms, allowing accurate testing of uncoated KTP.
The intracavity bench can distinguish different types of defects in KTP crystals, including two that may reduce green output power. One defect is localized in the bulk material (flaws) or at the surface (dust or streaks) and induces losses inside the laser cavity leading to reduced output at both 1064 and
532 nm. The second type of defect-index inhomogeneity at the fundamental or harmonic wavelength-leads to changes in phase-matching conditions for efficient second-harmonic generation. Even when the fundamental intracavity power is not affected by the second type of defect, the green output power can be reduced strongly. An easy method of identifying these two defects consists of simultaneously recording the infrared (IR) and green power while scanning across the mirror coated onto the Nd:YVO4 crystal (see Fig. 2).
The intracavity bench can be compared with an extracavity bench. Because extracavity testing requires more energy, the extracavity bench contains a Q-switched Nd:YVO4 laser that operates at a 30-KHz repetition rate and delivers 150-ns pulses with 500-mW average power. The output beam is focused in the KTP, and a prism allows separation of the green and fundamental beams.
Scanning the same crystal in the two configurations produces different results (see Fig. 3 on p. 85). Such experiments demonstrate the very high sensitivity of intracavity operation to losses and to index mismatch as compared to extracavity operation. Other tests made on a "good" crystal-one that has passed successfully the frequency-doubling test-have confirmed the superiority of intracavity tests.
Glass cutting advances toward SID `99
Visitors to the Society for Information Display meeting in San Jose, CA, this month can expect major improvements in laser glass-cutting technology over the technology and systems that were discussed and displayed last year.
Recent advances in two different approaches to laser glass cutting were presented in technical presentations at Display Works `99 last February in San Jose by representatives of Schott Glas (Mainz, Germany) and Accudyne (Palm Bay, FL)-two of the three exhibiting laser glass-cutting companies at SID `98 (Anaheim, CA). The third laser glass cutter at SID `98, Precision Technology Center (PTG; Lake Mary, FL), did not exhibit or present new technology at Display Works `99 but is planning to attend SID `99, according to a company spokesperson.
Accudyne has shifted from the single-laser-beam design presented at SID `98, to a dual beam system (a scribe beam followed by a break beam) to increase cutting speed. The company has produced two beta units for the United States Display Consortium (USDC), according to Brian Hoekstra, who presented the Accudyne results at Display Works `99.
The Accudyne system passes light from a 200-W CO2 laser through an optical path of four zinc selenide elements that shape the beam to increase cutting speed, prevent scorching of the glass substrate, and facilitate full separation after cutting (see Fig. 1 on p. 88). During the cutting process, the two heating beams are followed by a cooling stream of helium gas (that can also be mixed with water) to quench the substrate and facilitate the glass separation. Glass-cutting speeds achieved to date with the system, which is class-10 cleanroom compatible, depend on type of glass, glass thickness, coolant mixture, and breaking method (see table on
p. 90). Edge quality obtained with laser and mechanical methods has a high degree of smoothness (see Fig. 2).
Unlike Accudyne, Schott has met commercial glass-cutting speeds using a single laser beam and a different
beam shape, according to Christoph Hermanns. He said Schott uses a 10.6-µm CO2 laser to heat the glass along the desired scribe line, followed immediately by a jet stream of cooling air. The Schott team has taken this design from the prototype and test stage described at SID `98 to actual products, of which four have been built and sold since September 1998, said Hermanns. A primary current application is cutting thin glass panels for plasma displays. Other promising application areas include automotive glass, vacuum tubes, and glass tubes for medical and chemical applications. The output power of the lasers in the Schott systems varies from 100 to 600 W depending on glass thickness, according to Hermanns.
As with the Accudyne system, the Schott systems can provide either laser scribing prior to mechanical breaking or full laser breaking. Cutting speeds achieved so far by the Schott system for production-quality systems range from 1 to
10 m/min. for full breaking and from 10 to 20 m/min. for laser scribing and mechanical breaking.
Typical cut velocities for laser glass cutting
|0.7 mm soda lime||NA||1000 mm/s|
|1.1 mm soda lime||490 mm/s||800 mm/s|
|1.1 mm D263||450 mm/s||650 mm/s|
|1.1 mm||OA-2 270 mm/s||500 mm/s|
|1.1 mm 1737||120 mm/s||300 mm/s|
|1.1 mm soda lime||200 mm/s|
|1.1 mm D263||170 mm/s|
|1.1 mm OA-2||110 mm/s|
|1.1 mm 1737||70 mm/s|
Fiber Bragg gratings manage soliton dispersion
Researchers at Pirelli Cavi e Sistemi SpA (Milan, Italy), working with MCI Worldcom (Jackson, MS), recently evaluated a four-channel bidirectional 10-Gbit/s (OC-192) dispersion-managed soliton system over a 450-km field-installed link between Chicago, IL, and St. Louis, MO.1 The equipment required to upgrade the system from optoelectronically regenerated 10-Gbit/s non-return-to-zero (NRZ) coding to optically amplified 10-Gbit/s return-to-zero coding included a unique dispersion compensator in the form of long, chirped fiber Bragg gratings.
The field trial reflects the increasingly tough competition in the industry and new requirements from service providers. As a result, manufacturers are moving to 10-Gbit/s channels, and 16-channel 10-Gbit/s wavelength-division-multiplexing (WDM) systems will soon emerge. At this high bit rate and with 40-Gbit/s channels following soon, soliton transmission may offer a real advantage over traditional NRZ coding.
In solitons, nonlinear distortion and fiber chromatic dispersion perfectly balance each other out if the correct pulse power duration is chosen. Computer simulations and laboratory experiments have shown that soliton-like transmission leads to robust error-free system operation over a wide input power range, even in the presence of fiber links with high polarization-mode dispersion (PMD; see Laser Focus World, May 1999, p. 145).
Soliton systems attracted strong interest when the concept of dispersion management was first introduced in 1994. In a dispersion-managed soliton system, fiber chromatic dispersion is partly compensated for in approximately 95% of every fiber span; nonlinear effects compensate for the remaining dispersion. Soliton systems appear to be quite tolerant to undercompensation but are impaired by overcompensation.
Current technology offers two solutions for dispersion compensation: dispersion-compensating fiber (DCF) and chirped fiber Bragg gratings (CFBGs). Long CFBGs are the most promising devices because they combine a competitive price and reduced size with negligible nonlinearities, low losses, and low PMD. Chirped fiber Bragg gratings cannot provide 40-nm wavelength coverage on a single device-as do DCFs-but this fact is not a disadvantage because recent dense-WDM systems subdivide the total number of channels into sub-bands that are treated separately. A single grating could be designed to account for the dispersion compensation of a single sub-band, including third-order dispersion.
In CFBGs, the useful bandwidth, BW, and dispersion, D, are strictly related to the grating length, L, through the expression: BW x D = 2Ln/c, where n is the fiber refractive index and c is the speed of light in a vacuum.2 Chirped FBGs are currently manufactured with a chirped phase mask. Because of the limited mask length (100 mm) CFBGs are generally considered narrow-bandwidth devices.
"Thanks to technology advances, however, the grating length has been increased far beyond the mask size limit, and high-quality 1-m-long gratings have been realized by Pirelli via a proprietary manufacturing procedure," says Fausto Vaninetti, staff scientist at Pirelli. Thus, bandwidth in excess of 6 nm with dispersion of 1300 ps/nm can be achieved in a single device.
In the bidirectional 10-Gbit/s soliton system trial connecting Chicago and St. Louis, the installed fiber had a total differential group delay (DGD) of 9.2 ps; dispersion compensation was achieved by using CFBG modules as well as DCF units (see figure). The test showed the potential competitive advantage of CFBGs over DCF modules because the CFBGs allowed for a higher tolerance to signal power level and a reduced PMD effect. Bit-error-rate plots showed the excellent sensitivity achieved with each channel, confirming the robustness of the system.
- N. Robinson et al., Tech. Dig. OFC `98, PD19.
- M. Cole et al., "Design and application of long continuously chirped fiber gratings," IEE Optical Fiber Gratings, London, 16/1 (1997).
Optical probe measures MBE temperature
Molecular beam epitaxy (MBE) is used for making quantum-well lasers and other III-V semiconductor devices. Unlike ion implantation-typically used for silicon-based devices-MBE offers finer control for growing thin films. One of the problems with MBE, however, has been the lack of repeatability due partly to difficulty measuring the temperature of the substrate. Typically the temperature is measured by a thermocouple. To achieve uniform growth, the thermocouple cannot touch the substrate. But distance from the substrate reduces the accuracy of the measurement enough to cause difficulty in reproducing processes from chamber to chamber, or even from day to day. Also, accurate temperature models are essential for developing theoretical models of MBE growth of these semiconductors.
Researchers at the University of Arkansas (UA; Fayetteville, AK), together with CI Systems (Migdal Haemek, Israel) and Riber Inc. (Rueil Cedex, France), report building a much more precise temperature sensor for MBE machines based on optical transmission through the substrate.1 The researchers use a fiber bundle and quartz lightpipe to direct light through the back of the rotating substrate in a vacuum chamber used for molecular-beam epitaxy. Transmitted light is collected through a viewport in front of the substrate and directed through a lens into a fiber bundle. Chopped light differentiates the signal from stray light and other noise.
Because the spectrum depends on the bandgap of the material, and because the bandgap changes with temperature, the transmitted spectrum gives the substrate temperature. For gallium arsenide, this technique measures temperatures from 0°C to 700°C to within plus or minus 2°C, and the measurement can be updated about once a second. In comparison, thermocouple readings were found to be off by as much as 15°C.
Using spectroscopy to determine the temperature is not a new idea. Other groups have used the idea in a number of ways. Pyrometers measure temperature by detecting the blackbody radiation from the material-but stray light reduces the accuracy of this method. Another scheme uses the substrate heater filaments as a broadband light source, measuring the light transmitted-but problems remain; the source cannot be modulated, stray light is still a problem, and it does not work if the heater is turned off.
Methods based on a light source outside the high-vacuum chamber include band edge reflection spectroscopy, in which white light is directed at the substrate through the pyrometer viewport and back-reflected light carries information about the substrate temperature. This method works well for bare substrates, but suffers interference effects with thin films.
The best accuracy is achieved by directing white light behind the substrate using a lightpipe, then gathering the light through the pyrometer viewport. Another group reported doing exactly that, but only for a fixed substrate.2 In many MBE machines, however, the substrate is mounted on a manipulator that rotates it from the position in which it is introduced into the chamber to the growth position, around a perpendicular axis. The UA group designed a system that could follow the rotation using a quartz rod bent in a right angle and a fiber bundle designed for high-vacuum compatibility.
This set up allowed the researchers to compare the thermocouple response to the system response-they found that when the wafer is cooling, the thermocouple cooled even faster. Future research may focus on the temperature gradient across the wafer-the current arrangement can measure temperature only at the center of the wafer. To detect the temperature at several spots, multiple lightguides would be needed.
The maker of the MBE system, Riber, plans to offer the thermometry device as an option with its customized machines. An NSF-funded partnership between UA and Lucent Technologies (Murray Hill, NJ) will foster exchanges of researchers to use the new facility.
- J. Vacuum Science and Technology (Jan.-Feb. 1999).
- J. A. Roth and others at Hughes Research Laboratory, Malibu, CA.