More and more Americans are cruising the information highway. They are accessing the Internet and sending electronic mail throughout the world. They are looking forward to the day when they can tune in to hundreds of cable television channels. Eventually, they will use their telephones and television sets to call up their favorite movies to watch immediately or store for later viewing. They will be able to receive airline schedules, news, and catalog shopping information when they want them.

To speed communication and retrieval of information as the number of users increases, the evolving information highway must be constructed to expand its high-speed data traffic. A possible solution to this capacity problem is a promising new material developed by researchers at Oak Ridge National Laboratory. The material is being studied for possible use in optical switches for networks of glass, or optical, fibers that are as fine as human hair. Preliminary results indicate that the ORNL material could pave the way for much greater use of the emerging information highway.

Using a technique called molecular beam epitaxy (MBE), Rodney McKee and Fred Walker, scientists in ORNL's Metals and Ceramics Division (M&C) Division, have developed an optically clear ceramic film for transmitting waves of light. This "waveguide" material consists of barium titanate, deposited one atom-layer at a time on a crystal of magnesium oxide. They have also shown that the waveguide material can be deposited on silicon. Their goal is to develop an optical switch consisting of silicon on which is deposited magnesium oxide, the barium titanate waveguide, and appropriately configured electrodes.

The light-transmitting ability of the barium titanate films has been verified in tests at ORNL, Wright-Patterson Research and Development Center in Ohio, Hughes Aircraft in California, and AT&T in New Jersey. In these tests a laser is used to transmit a 1-millimeter-wide beam of light through the waveguide medium, but ultimately a 1-micrometer-wide channel, or optical path, will be used.

Even with today's technology, an optical fiber uses less than 1% of its theoretical capacity, or bandwidth, for carrying information. In other words, the information highway has room for considerably more high-speed data traffic if the appropriate technology is developed.

"Switches using this type of waveguide could increase the amount of information carried by fiber-optic cables by 100 times," McKee says. "Since the advent of lasers in the early sixties, scientists have dreamed of replacing electrical circuits with all-optical circuitry or integrated electro-optical devices in silicon-based technology. With these materials, our dreams may soon become reality."

Over the next few decades, copper wire for telecommunications will be replaced with optical fibers made of glass. Glass is much cheaper and more common than copper. Even more important, glass fiber can carry 250,000 times as much information in the form of light from lasers as does a standard copper telephone wire conducting electrical signals. The reason: glass conducts light signals of much higher frequency--number of waves per second--than copper can for electrical signals. The higher the frequency, the more information can be packed into light signals.

The ORNL achievement will contribute to recent developments that have enabled the transmission of increased amounts of information by light. They include miniature lasers on silicon chips (laser diodes), improved optical fibers that carry light longer distances, and optical switches, which convert electrical signals to light signals.

Such technology must economically produce key components of the information highway. It should use a single fabrication process to incorporate laser diodes, microcircuits, electrodes, and optical switches on a silicon chip, or substrate (the physical structure that provides electrical and mechanical support for microelectronic components). Silicon is the choice support structure for these components because of its purity, crystal perfection, and associated electronic and physical properties.

Optical Switches

Although a number of material systems are being studied, the waveguide material chiefly in commercial use today for optical switches is lithium niobate. However, this material is damaged after repeated passage of laser light, making the laser signal too weak for the switch to work. Another drawback is that no way has been found to integrate this material with other components on silicon. Currently, lithium niobate cannot be deposited as a thin film on silicon as can barium titanate. ORNL's ability to grow barium titanate as a single-crystal thin film one layer of atoms at a time can produce a well-ordered and nearly defect-free material with electro-optical properties potentially 35 times better than those of lithium niobate. In other words, a barium titanate switch could be 35 times smaller or use 35 times less electricity than a lithium niobate switch.

Each piece of digital information traveling on the information highway is called a binary digit, or bit. A bit is either a 1 or 0; a combination of 8 bits, which can define a letter or number, is called a byte. Bits can be transmitted into a glass or copper-wire cable by a switch. If a switch is on, it lets through a high-intensity light or electrical signal that represents a 1; if the switch is off, it passes a low-intensity signal that represents a 0. Each combination of high-intensity and low-intensity signals represents a word, number, or other data.

Today digital pulses, or bits, of information are sent down optical fiber by gating--turning on and off--laser diodes, tiny sources of light of different frequencies (representing different telephone conversations, for example). More advanced technology uses lithium niobate to gate light from a continuous laser source. Although lithium niobate switches are faster than switched lasers, the data transmission rate is still limited to a frequency level of 1 to 10 gigahertz. With barium titanate (which has a larger "bandwidth"), the data transmission rate can reach 1000 gigahertz, permitting about 100 times more information to be transmitted.

In optical guided wave devices such as optical switches, electric fields are pulsed to switch the light signals on and off in the waveguides, which have a much faster response than the laser diodes. The reverse is also true: Light signals to waveguides on silicon chips are rapidly converted to electrical signals.

In an optical switch using barium titanate, light is split into two paths. An alternating electric field is applied to the electrodes of one of the beam paths. When the field is on, it changes the phase of the light signal in one path relative to the other path; in other words, the field causes the waveguide material to shift the positions of the crests and troughs of the light waves passing through. When the two beams of light recombine, they mutually interfere, reducing the intensity as the light exits. When the field is off, the split light beams recombine as a signal with the same intensity they had entering the switch. The resulting combinations of increases and decreases in light intensity serve as rapid "on-off" switches, greatly increasing the amount of information transmitted.

McKee and Walker are developing a prototype optical switch using barium titanate as a planar waveguide with aluminum electrodes attached to do the on-and-off switching. They are also working with AT&T on an optical switch in which narrow channels for light about the size of the wavelength of visible light are etched into the ORNL waveguide material. It is these channel waveguides that allow practical optical circuits to be integrated into the material. This new class of waveguide structures and their potential for integration into silicon-based technology, if exploited in a timely manner, could advance communications technologies, benefiting consumers, industry, and the federal government.

Formation of Thin Films

Since 1989 McKee and Walker have been using the demanding technology of MBE to form these light-transmitting ceramic thin films on an optical isolation layer of magnesium oxide on silicon. Although their research and approaches to this materials development problem have been at a very fundamental level, they have developed practical technology while making scientific advances. As a result of their work, they have received one patent, applied for three more, and published a series of journal articles in Applied Physics Letters and Physical Review Letters.

The ORNL researchers were the first to grow high-performance thin-film ferroelectric oxides with low losses of light. They use MBE to build each film layer by layer, creating a desired ordered structure to enable the transmission of light. To produce the optical waveguide on silicon, the researchers first grow a transition oxide a few atom layers thick on a single-crystal silicon substrate. This transition oxide is a critical component that provides a structural match to magnesium oxide (MgO). Magnesium oxide was chosen because of its optical properties, crystal structure, and stability. As an isolation layer, it separates the waveguide from the substrate, preventing the silicon from absorbing and dissipating light passing through. Also, its atoms can be lined up with those of silicon, the unique transition oxide, and barium titanate, permitting properly oriented growth of these materials.

MgO is an alkaline earth oxide like barium oxide, calcium oxide, and strontium oxide. The ORNL researchers thought that atoms of any of these materials could be lined up with those of silicon and barium titanate, the waveguide material. The researchers used barium oxide (BaO) first as an isolation layer, but they found that it reacted with water vapor during the deposition process, forming barium hydroxide. In their experiments they found that MgO worked much better as an isolation layer.

Barium titanate is both a perovskite (a mineral with a roughly cubic crystal structure like that of silicon and MgO) and a ferroelectric (a material in which the positive and negative ions move apart, setting up electric fields within the crystal). The interaction between the internal electric field and the applied external field controls the material's ability to influence the information content in a guided light wave.

The ORNL group formed a transparent film of alternate one-atom-layer-thick deposits of titanium oxide (TiO2) and BaO. The researchers aligned the atoms of TiO2 with respect to the atoms in the MgO layer. Next they added a BaO layer in proper alignment with the TiO2. They repeated the process many times, fabricating a barium titanate film thick enough to transmit a thin laser light beam. The light was efficiently conducted through the waveguide with practically no loss of intensity. In addition, the researchers showed that the waveguide film and isolation layer could be deposited on a silicon substrate. They are now working on growing these films on larger-diameter silicon to reduce the cost of each optical switch and make the manufacturing process more economical.

In making a BaTiO3 film, why is a TiO2 layer rather than a BaO layer deposited first on the MgO layer? The reason, McKee says, is that the arrangement of the atoms makes for a good electrostatic fit, or proper "ionic coordination," between the layers. The positively charged titanium ions in the waveguide layer interfacing with the isolation layer are attracted electrostatically to the negatively charged oxygen ions in the MgO layer, and the positively charged magnesium ions are attracted electrostatically to the negative oxygen ions in the TiO2 layer. Depositing a BaO layer on the MgO layer first does not work; when placed near each other, the atoms in both layers (barium atoms on magnesium atoms, and oxygen atoms on oxygen atoms) will repel each other because of like charges. Deposition of the TiO2 layer, which is formed by vaporizing a titanium sample in low-pressure oxygen, is followed by deposition of a BaO layer, then a TiO2 layer, then a BaO layer, and so on to form BaTiO3.

"This technique is called planar stacking," McKee says, "because layers are stacked to form a compound of both a desired composition and structure. We hope to improve the properties of existing compounds and synthesize new materials this way."

Layered Structure

An obvious example of a waveguide is a mirror placed upon a mirror so that only the silver backing is visible. If light is passed between the two mirrors, it will be reflected between them as it moves from one end to the other. The most familiar waveguide is the optical fiber. Light is guided along an optical fiber by undergoing total internal reflection at the interface between the air and the glass fiber.

Glass fibers and other transparent materials can form waveguides if their index of refraction is significantly higher than that of the surrounding medium. In such a case, light trying to escape the waveguide into the medium will be bent back, or internally reflected, into the waveguide. The best waveguides minimize optical losses.

Because of the importance of the refractive index in keeping the light inside the waveguide, the ORNL researchers in collaboration with David Zelmon of Wright-Patterson Research and Development Center in Ohio developed a four-layer waveguide structure. The first three layers are silicon, MgO, and BaTiO3. The fourth layer, air, acts as an isolation layer like MgO.

"To make a ferroelectric waveguide," says McKee, "we could not simply deposit barium titanate on silicon to make a three-layer waveguide (with air as the top layer) because silicon's index of refraction is higher than that of barium titanate. Because silicon will absorb and dissipate light, we had to put down an optical isolation layer between the substrate and guiding layer. We chose magnesium oxide because its index of refraction is lower than that of barium titanate and because its atoms can be lined up with those of silicon and barium titanate, making a good epitaxial film. Our tests show that magnesium oxide is an effecive isolation layer."

The index of refraction for BaTiO3 is 2.4, significantly higher than the refractive indices for the two isolation layers. For MgO it is 1.8, and for air it is 1. The differences keep the light traveling in the BaTiO3 layer.

Molecular Beam Epitaxy

Using MBE techniques to make the thin films, the researchers heat a magnesium oxide crystal grown by Lynn Boatner of ORNL's Solid State Division. The crystal's MgO molecules vaporize in a chamber containing a silicon substrate in ultrahigh vacuum. Because the chamber is kept at extremely low air pressure, the vaporized MgO molecules reach the silicon without encountering many air molecules. Titanium and barium are vaporized and reacted with oxygen to make alternating layers of titanium oxide and barium oxide, which are aligned to form a barium titanate waveguide film.

"To produce a thin film with good optical quality," McKee explains, "we maintain stringent synthesis and processing requirements at the near-interface regions. We have to regulate the arrival rate of the oxygen, the temperature of the metal being vaporized, and the temperature of the substrate. Source-shuttering MBE techniques give us control at the atomistic level of materials development."

To make defect-free, crystalline films in which all the atoms are epitaxially aligned, the ORNL researchers monitor film formation using reflection high-energy electron diffraction (RHEED). RHEED directs a high-energy electron beam at the sample at a very low angle (1 to 2 degrees) to obtain detailed surface information. If the monitoring technique indicates poor atomic alignment and crystalline imperfections, then the researchers change temperatures, pressures, and other parameters to improve film quality.

"When RHEED electrons strike the atoms of the film," says McKee, "they are deflected to a phosphor screen, and the points of light there show the arrangement of the atoms in a crystalline pattern and indicate whether the film is growing across the substrate surface in an orderly way. Distortions in the electron pattern can indicate defects in the crystalline structure, telling us that we must fine-tune the film-forming process."

Evaluating the Waveguide

A variety of ORNL resources have been called upon to analyze the material and optical properties of the thin-film waveguide. Kathi Alexander, a materials researcher in the M&C Division, has been studying the waveguide's crystalline lattice structure using the Hitachi field-emission gun transmission electron microscope at ORNL's High Temperature Materials Laboratory. She hopes to determine the structure of the magnesium oxide-barium titanate interface that provides optical clarity. Scanning electron microscope images produced in the M&C, Solid State, and Health Sciences Research divisions have shown that the thin-film waveguide's optical properties are superior to the bulk materials now used.

"Our group at ORNL," says McKee, "has achieved the first demonstration of optical clarity in thin-film barium titanate. Characterization of our waveguides produced by growing nearly perfect thin films of barium titanate on magnesium oxide has shown that their ability to channel light with little loss in intensity is as good or better than that of bulk titanium-drifted lithium niobate waveguides. This materials synthesis achievement points the way towards development of a whole new class of waveguides--thin-film perovskites on alkaline earth oxides.

"This ORNL development is important for several reasons. First, the bandwidth achievable for these thin-film waveguides is about 100 times that of a bulk ferroelectric like lithium niobate. Second, the electro-optic coefficient is 35 times larger than that of lithium niobate and 300 times larger than that of gallium arsenide. Such a high-performance characteristic would enable fabrication of smaller, faster, and more efficient devices. Third, we have shown that, unlike lithium niobate, our barium titanate-magnesium oxide composite waveguide can be deposited on silicon and that, fortunately, the problems we anticipated in integrating our waveguide material with silicon to make practical devices can be avoided."

The Silicon Solution

When efforts were initiated to try integrating these waveguide structures with silicon, McKee, Walker, and others were concerned about the structural change that occurs when the BaTiO3-MgO films cool after being deposited on silicon. The question was this: although the waveguide material is chemically compatible with silicon, is it structurally compatible after deposition as a film?

In the fall of 1994, McKee and Walker demonstrated that this new class of waveguides could be integrated with silicon. McKee reported this success at the Materials Research Society annual meeting in November 1994. He also reported their finding that the structural change that caused concern actually favors the fabrication of a silicon-based waveguide!

When deposited on silicon, the BaTiO3 film is in its cubic phase. As the BaTiO3 film and silicon substrate cool, the film tries to shrink but stays in place on the substrate. As a result, the microstructure of BaTiO3 is stretched--from a cubic to a tetragonal configuration. In other words, if the structural change could be greatly magnified, you would see stacks and rows of cubic boxes turned into stacks and rows of upright milk cartons.

To relieve this tension, the molecules shift 90 degrees C or "lie down," thereby aligning their long axis with the surface of the plane. Viewed another way, the "milk cartons" are laid down on their sides. "This transformation is critical to waveguide development," McKee says. "The new alignment makes it possible to obtain barium titanate's optimum electro-optic coefficient. It allows us to easily orient the applied electric field with the crystal's internal ferroelectric field and optical polarization of the traveling light wave signal--that is, the orientation of the light wave vibrations. Thus, this favorable structural change will enable us to integrate of our waveguide material with silicon and to control the waveguide structure so that practical silicon-based optical guided wave devices can be built."

McKee explains the technical basis for his concern about integrating the waveguide material with silicon. "The electro-optic properties of barium titanate are directionally dependent," he says. "So we were concerned about the spontaneous polarization-the collective displacement of charged ions--and the domain structure--regions in which positive and negative ions separate, setting up electric fields--in the crystalline film of barium titanate.

"We knew that barium titanate's spontaneous polarization and domain structure could couple to thermal strain in the silicon substrate, making it difficult to achieve the best possible electro-optic response. But Nature has been kind to us. We found that a unique and potentially advantageous domain structure can be obtained for our barium titanate--magnesium oxide waveguides if we grow them on silicon."

For their silicon demonstration, McKee and Walker received support from ORNL's Laboratory Director's R&D Fund. They are using this funding to develop methods of depositing waveguide material on silicon at a lower temperature--a range of 300 to 500 degrees C instead of 500 to 800 degrees C. A lower temperature is required to make silicon chips that integrate microcircuits, laser diodes, and optical switches using a single economical fabrication process. To accomplish this goal, McKee says, adjustments must be made in the oxygen pressure of the film-growth chamber and the arrival rate and time for oxygen in contact with the vaporized metals of magnesium, barium, and titanium.

McKee and Walker will continue work with collaborators to achieve the correct thicknesses of the light-guiding and isolation layers for the four-layered waveguide structure, to develop better optical switches, and to make possible the economical manufacture of silicon chips that integrate optical switches with laser diodes and microcircuits. These developments could help the United States win the global race to find ways to significantly speed up communications and computations, greatly improving the information superhighway.

Where to?

[ Next article | Search | Mail | Contents]