Report

Microstructured Optical Fibers as High-Pressure Microfluidic Reactors

See allHide authors and affiliations

Science  17 Mar 2006:
Vol. 311, Issue 5767, pp. 1583-1586
DOI: 10.1126/science.1124281

Abstract

Deposition of semiconductors and metals from chemical precursors onto planar substrates is a well-developed science and technology for microelectronics. Optical fibers are an established platform for both communications technology and fundamental research in photonics. Here, we describe a hybrid technology that integrates key aspects of both engineering disciplines, demonstrating the fabrication of tubes, solid nanowires, coaxial heterojunctions, and longitudinally patterned structures composed of metals, single-crystal semiconductors, and polycrystalline elemental or compound semiconductors within microstructured silica optical fibers. Because the optical fibers are constructed and the functional materials are chemically deposited in distinct and independent steps, the full design flexibilities of both platforms can now be exploited simultaneously for fiber-integrated optoelectronic materials and devices.

Optical fibers provide ideal hosts for the manipulation of photons, especially when formed into microstructured optical fibers (MOFs) that enable precise control of photon dispersion (1). Crystalline semiconductors such as silicon provide ideal hosts for the manipulation of electrons, especially when formed into heterostructures that enable precise control of electron transport. It has thus far not been possible to integrate the crystalline semiconductors that form the basis for modern optoelectronics into MOFs, allowing for interaction of such materials with wave-guided electromagnetic radiation over much longer length scales than can be realized in typical planar device geometries (2). Fabrication of such structures would be a major step forward toward all-fiber optoelectronics. The preferred method for depositing semiconductors and metals, including nearly all of the technologically important semiconductors, is chemical vapor deposition (CVD) (3). However, CVD onto the walls of the long, extremely narrow pores in a MOF presents two challenges: Small deviation from perfect conformal deposition anywhere along the length of the pore would immediately arrest deposition, and mass transport of the reactants into and by-products out of such a confined space is prohibitively slow using traditional techniques. We report the fabrication of high-quality polycrystalline and single-crystal semiconductors within the voids of MOFs by high-pressure microfluidic chemical deposition. High-pressure flow, which can be sustained because of the very high mechanical strength of optical fibers (4), overcomes mass-transport constraints and also enables a strikingly uniform, dense, and conformal annular deposition onto the capillary walls, even for pores that reduce to less than 10 nm in diameter.

MOFs are typically fabricated by stacking and fusing arrays of capillaries into preforms centimeters in diameter that are drawn at high temperature (1, 5). This approach allows the complex preform structure to be scalably replicated in fibers that are four orders of magnitude longer and two orders of magnitude smaller in diameter. By appropriately designing the stacked preform, the capillary holes within a single MOF can have a wide range of shapes and sizes in precisely engineered periodic or aperiodic spatial configurations.

We treat the empty pores in a MOF as micro/nanoscale reaction chambers into which we can directly deposit a wide range of technologically important semiconductors and metals with exceptional control, because we can now exploit the decades-long knowledge base developed for CVD onto planar substrates. Very high pressures (10 to 100 MPa) facilitate rapid mass transport through the fiber pores. In a typical experiment (6), a germanium precursor GeH4 flows through the heated MOF at a partial pressure of 2 MPa [much higher than in traditional CVD (3)], along with an inert carrier gas. High pressures and toxic precursors such as GeH4 are safe and practical because the pressure reservoir and the fiber pores have a very small volume. A smooth layer of amorphous germanium begins to deposit on the pore walls as the temperature is ramped up past 300°C. Crystalline grains then nucleate and grow as the temperature exceeds the crystallization point of ∼375°C. As growth proceeds, a remarkably uniform tube forms (see Fig. 1, A to D); as it fills with germanium, a 1.0-μm diameter pore can be narrowed by a factor of 100 down to 25 nm or smaller, tapering open gradually over a deposition length of 70 cm. Silica capillaries drawn to a 100-nm diameter were also successfully filled with germanium to form nanotubes over macroscopic lengths of up to 30 cm, with an inner diameter of less than 10 nm (Fig. 1D).

Fig. 1.

Germanium integrated into MOFs. (A) Scanning electron microscopy (SEM) cross section of a germanium tube within a silica capillary. Scale bar, 500 nm. (B) Germanium deposited within a 1.0-μm pore, with a central hole ∼25 nm in diameter. Scale bar, 200 nm. (C) A germanium wire etched out of a MOF. Scale bar, 2 μm. (D) Field-emission SEM cross section of a germanium nanotube within a MOF. Scale bar, 100 nm. (E) Unfilled honeycomb MOF template. Scale bar, 5 μm. (Inset) Schematic indicating surfaces growing inward at a uniform rate, in a direction parallel to the local normal (green arrows). When the deposition reaches the thickness indicated by the green inner hexagon, the rounded corners have disappeared and the cross section is a perfect hexagon. (F) Honeycomb template after germanium deposition. Scale bar, 5 μm. (G) Cross-sectional TEM of a germanium-filled honeycomb MOF, showing thread-like dislocations within the grains. Scale bar, 500 nm. (H) Selected-area diffraction pattern of the region circled in (G). o is 111; p is 220. Zone axis is <110>.

Deposition into more highly structured pores reveals additional information about this conformal filling. When a large-air-fraction fiber with a honeycomb pattern of holes is filled (Fig. 1E), a spatially ordered array of hexagonal germanium tubes is formed (Fig. 1F). The interior vertices of the hexagonal holes, originally defined by the rounded silica template (Fig. 1, E and G), actually sharpen (Fig. 1F) as growth proceeds. If we assume uniform inward motion of the surface along the local normal during deposition, then the rounded corners of a polygon will become sharp as the thickness of the deposited layer exceeds the radius of curvature of the corner (Fig. 1E, inset). The resulting faceted tubes resemble lithographically patterned micro and nanostructures but are formed in a simple single-stage deposition.

This uniform annular growth down to very small inner diameters, with sharp geometric features, is particularly striking when one considers that the germanium is polycrystalline, with a grain size much larger than the dimensions of these features. Transmission electron microscopy (TEM) and selected-area diffraction patterns (Fig. 1, G and H) reveal that upon heating up to 500°C, the originally amorphous germanium converts into crystalline grains more than 500 nm across (6). These grains are much larger than the final inner diameter of the hexagonal tubes, yet the inner walls remain smooth. micro-Raman spectra of wires inside the silica matrix show just a single mode at 300.8 cm–1, characteristic of pure crystalline germanium.

To demonstrate that we have integrated continuous, electrically active germanium into optical fibers over macroscopic lengths, we have fabricated an 11-mm-long, 5-μm-diameter field-effect transistor (FET) inside a single silica capillary of 94-μm outer diameter (Fig. 2). Transconductance measurements (6) reveal the carriers to be n-type; by interpolating the linear slope of dIdrain/dVgate for a given Vsource (7), the carrier mobility is calculated to be 1.05 cm2/Vs at room temperature. The samples were not intentionally doped, although it should be possible to extrinsically dope during deposition. We can modulate the channel conductance over several orders of magnitude to obtain pinch-off, yielding an estimate of the carrier concentration of 2 × 1015 cm–3 (7). We anticipate that the mobility will approach values more typical of bulk n-type polycrystalline Ge [100 cm2/Vs (8)] upon further optimization of the deposition conditions (e.g., precursor and carrier gas purity and thermal treatment for grain growth).

Fig. 2.

A fiber-integrated germanium FET. The current-voltage characteristics of the FET show the effect of gate bias (as labeled for each curve), with complete pinch-off at –100V. (Upper inset) Etched end of the germanium-filled capillary. (Lower inset) Schematic circuit diagram of the FET.

Electrical characterization of individual wires and tubes within MOF templates is straightforward in comparison with transport measurements on “loose” vapor-liquid-solid (VLS)–grown wires (9, 10), because it does not require lithographic patterning or micromanipulation. An additional advantage is the ease of forming low-resistance ohmic contacts, because up to several millimeters of the protective silica cladding can be etched away (Fig. 2, upper inset) to reveal long sections of bare nano/microwires for metallization by standard techniques. The resulting contacts are thus well decoupled from the transfer characteristics of the device. In contrast, the formation of high-quality ohmic contacts to VLS nanowire devices remains challenging (11). In particular, if sufficiently high-quality contacts are not formed, modulation of the contact resistance by the source-drain bias can give rise to nongated transconduction in such systems. Furthermore, lithographically processed VLS nanowire devices are often annealed to reduce the contact resistance. This annealing can lead to chemical diffusion along the wire (given the proximity of the contacts) and thus make it difficult to control the doping type and level (11). Our approach precludes any contact-induced changes in device behavior because of the large distance between the source, drain, and gate. Such decoupling of extraneous contact effects is critical for obtaining a fundamental understanding of device behavior.

Deposition of silicon from a SiH4 precursor into a honeycomb fiber at 700°C also forms long, smooth hexagonal tubes with sharp vertices (see Fig. 3, A and B). The peak in micro-Raman spectra collected on these silicon wires is downshifted by 2 cm–1 from that expected for bulk silicon (521 cm–1). This downshift is consistent with the differential thermal expansion of silica and silicon between 700°C and room temperature. First-principles total energy calculations of bulk silicon under tensile stress produce a comparable Raman downshift at an isotropic dilation of 0.15%. A roughly isotropic thermal dilation is expected, even though the fiber container is elongated. Taking the silicon bulk modulus of 98.9 GPa, the fiber-integrated silicon wire experiences a substantial negative pressure, –0.15 GPa. A uniaxial or biaxial tensile stress is inconsistent with the maximum possible differential thermal expansion, because the positive Poisson's ratio of silicon ensures that the pressure dependence of the Raman mode is much weaker under such conditions (12). If the wires are etched out of the silica matrix, this stress is relieved and the mode returns to 521 cm–1. In contrast, the germanium tubes show no stress-induced Raman shifts, even though the thermal-expansion coefficient for germanium is 2.3 times as high as that of silicon. The germanium-silica interface does not transfer as large tensile and extensional shear forces as does the silicon-silica interface.

Fig. 3.

Fiber-integrated silicon structures. (A) Hexagonal silicon tubes in a honeycomb MOF template. Scale bar, 1 μm. (B) Long silicon tubes etched out of the template. Scale bar, 100 μm. (Inset) One of these tubes at higher magnification, showing the flat sides of the hexagon. Scale bar, 2 μm. (C) Schematic of lightguiding experiment. The end of a filled fiber is etched to completely expose 2 mm of the central silicon core (blue) and taper a portion of the silica cladding (red). (D) A 633-nm light is guided in the tapered silica cladding but not in the exposed silicon. (E) A 1.55-μm light propagating through the same etched fiber. The light is scattered at the end of the silica cladding but continues to propagate the full length of the silicon core.

A major advantage of the fiber-integrated geometry is the strong coupling of propagating photons to electronic degrees of freedom in the deposited material along the long interaction length of the fiber. Therefore, the waveguiding properties of semiconductor MOF metamaterials are key to their exploitation in the next generation of photonic devices. Approximately 2 mm of the 125-μm diameter silica cladding of a 5-cm-long sample was chemically etched away at one end to allow its optical characteristics to be decoupled from those of the 2-μm diameter silicon core. A 633-nm laser light launched into the other cleaved end-face of the composite fiber cannot propagate beyond the cladding region and into the core (Fig. 3, C and D) because silicon absorbs strongly at visible wavelengths. However, for near-infrared light at energies below the silicon bandgap (1.07 eV), the core is transparent and 1.55-μm radiation can be observed propagating through it (Fig. 3E). By measuring the output power waveguided through the isolated silicon core (6), we place an upper bound of 7 dB cm–1 on the loss within it, comparing favorably to losses of 7 to 9 dB cm–1 reported thus far for optimized planar polycrystalline silicon waveguides (13, 14). The optical losses in semiconductor devices are a strong function of the surface roughness (15), which can limit the performance of lithographically defined rectilinear waveguides because they have an exposed upper surface. However, as material deposits inside a MOF, the area of the exposed growth surface decreases linearly with the inner diameter and becomes zero in the limit of a fully filled pore. In addition, the outer surface of the fiber-integrated semiconductor waveguide is defined by the extremely smooth silica surface of the MOF pore [0.1-nm root mean square roughness (16)]. Therefore, boundary scattering from both inner and outer surfaces is minimized.

Fiber-integrated compound semiconductors, semiconductor heterostructures, and metal-semiconductor heterostructures are also possible. For example, we have integrated annular layers of mid-infrared transparent GeS2 (17), coaxial SiGe heterojunctions, and annular gold/silicon Schottky junctions into MOFs by sequential deposition (Fig. 4). The large discontinuities in refractive index and electronic bandgap in, for example, SiGe heterostructures are a prerequisite for devices such as high-frequency electro-optic modulators that span the mid- to far-infrared (14, 18). Many of the known CVD chemistries for other compound semiconductors (3) and metals (19) can be adapted to the high-pressure microfluidic chemical deposition technique. Perfect crystallinity is not required for many high-quality devices; suitably engineered polycrystalline materials systems can have optical properties nearly equivalent to those of single crystals (20).

Fig. 4.

Heterostructures and single crystals within MOFs. (A) Compound semiconductor GeS2 tube. Scale bar, 500 nm. (B) An 80-nm-thick smooth gold annulus deposited within a silicon tube. Contrast between silica and silicon is low. Scale bar, 2 μm. (C) Silicon cladding on a germanium core. Scale bar, 6 μm. (D) Optical micrograph of an array of gold particles written with a focused 514.5-nm laser beam within a 1.6-μm capillary. (E) SEM micrograph of an 80-μm-long section of a single-crystal silicon wire protruding out of an MOF after etching away the silica cladding. Scale bar, 50 μm. (F) Electron diffraction pattern collected on a cross-sectional slice of the single-crystal wire. o is 131; p is 220.

The high transparency of the MOF templates can be exploited to pattern thermally sensitive or photosensitive precursors along the fiber axis. When combined with the ability to individually address capillaries within spatially ordered arrays (either serially or in parallel), this capability enables patterning in all three spatial dimensions. For example, using a laser oriented perpendicular to the fiber axis and focused through the cladding, we have directly written gold (Fig. 4D) and silicon particles (6) as small as 1.0-μm long at precise locations along the axis of a 1.6-μm diameter capillary.

The superior electron transport and photonic properties of single-crystal semiconductors are desirable for many high-performance devices such as lasers (10) and FETs. To this end, we have modified our high-pressure microfluidic process to enable the template-directed growth of single-crystal semiconductor wires in a manner analogous to VLS growth techniques (9) for single-crystal nanowires, which can be assembled into devices (10, 21, 22). We used laser-written gold plugs inside a 1.6-μm silica capillary to decompose high-pressure silane at 370°C for growth of single-crystal silicon wires. The resulting wires, also 1.6 μm in diameter, grow in the <112> direction (Fig. 4, E and F).

Hierarchical bottom-up organization of nanomaterials into device configurations remains a central challenge in nanotechnology. The CVD deposition and VLS growth of materials within the ordered arrays of capillary holes in MOFs demonstrated here provide an elegant and powerful method to spatially organize functional materials at dimensions down to the nanoscale and allow for cooperative photonic and electronic processes between them. For example, the selective filling [by masking techniques (23)] of adjacent capillary holes in photonic-bandgap fiber waveguides with semiconductor/metallic electrodes (24) can enable electro-optic modulation (14) of the semiconductor refractive index and, therefore, bandgap tuning. In addition, polycrystalline or single-crystal semiconductors, deposited within MOFs, can enable nonlinear frequency conversion in the near- to mid-infrared (25) and serve as direct bandgap gain media for fiber lasers that operate over a range of wavelengths not previously possible. Such devices could be robust, inexpensive, and seamlessly integrated into the existing fiber-based infrastructure. More generally, the ability to simultaneously engineer radial, longitudinal, and compositional complexity within optical fibers, whose own microstructure can be engineered independently, heralds an opportunity to fabricate sophisticated three-dimensional optoelectronic device structures within the fiber geometry.

Supporting Online Material

www.sciencemag.org/cgi/content/full/311/5767/1583/DC1

Materials and Methods

References

References and Notes

View Abstract

Navigate This Article