IR FIBERS: Silicon core is highly crystalline

Scientists at Clemson University (Clemson, SC) have figured out how to fabricate a silicon (Si)-core, glass-clad optical fiber using a conventional fiber-draw process.

Dec 1st, 2008
Th Ir 01

Scientists at Clemson University (Clemson, SC) have figured out how to fabricate a silicon (Si)-core, glass-clad optical fiber using a conventional fiber-draw process.1 In addition, the resulting fiber has a highly crystalline core–in fact, measurements show that the Si grain size is large enough that the fiber core behaves experimentally as a single crystal. The fiber has potential for mid-IR power delivery, terahertz-radiation waveguiding, and nonlinear-fiber devices.

In collaboration with others at the University of California-Los Angeles, Northrop Grumman Space Technology (Redondo Beach, CA), and Elmira College (Elmira, NY), the Clemson researchers started with a silica cladding preform 50 mm in outer diameter and having a 3.5 mm hole down its center; the preform consisted of three concentric silica tubes (only because a single tube with the required inner and outer diameters was not available). The cylinder was joined to a piece of solid-silica rod to seal the cylinder’s end. A rod of single-crystal Si, 3 mm in diameter, was inserted into the cylinder to complete the preform.

The fiber was drawn at a temperature of 1950°C, a conventional draw temperature for telecommunications fibers. At this temperature the Si melts completely, while the silica cladding becomes viscous. John Ballato, a Clemson professor and director of the university’s Center for Optical Materials Science and Engineering Technologies (COMSET), where the fabrication was carried out, notes that a single-crystalline Si rod was used as the core preform simply because they had that available. “Because the Si ultimately melts, a polycrystalline sample (or a doped sample) could also be used,” he says.

A silica-clad fiber has a highly crystalline Si core (top, seen in cross section). The fiber was fabricated using a draw tower at COMSET (bottom). (Courtesy of Clemson University)
Click here to enlarge image

The researchers drew fibers with diameters ranging from 1 to 2 mm and cores from 60 to 120 µm (see figure), and selected several 5-cm-long bubble-free sections of fiber for their measurements. X-ray diffraction (XRD) analysis of the fiber’s core showed only crystalline Si, and a lack of evidence of twinned crystals. “That was a surprise to us,” says Ballato. “We expected possibly an amorphous core, due to the high quench rate, or polycrystallinity. Again, the x-ray diffraction indicates single crystallinity but that analysis only investigates a given volume of sample; so there could be polycrystalline regions, or the grain size could be large such that it still appears single-crystalline to the x-rays. Either way, it is highly crystalline.”

Oxygen in there somewhere

Energy-dispersive spectroscopy revealed that the Si core has a concentration of up to 17 atom percent (at%) oxygen in the core (the result of diffusion from the cladding). A comparison of Raman spectra of the core and the predrawn Si show only a slight change to the peak after drawing, which may arise from the thermal-expansion mismatch between the core and cladding.

“This (the 17 at%) is more oxygen than silicon should be able to contain without precipitating out as something else,” says Ballato. “The phase diagram says that, at room temperature and that oxygen content, the stable phase is silicon and cristobalite (crystalline SiO2). It should be noted that neither the XRD nor the Raman found evidence of the oxide precipitates. We postulated that this was because both the x-ray and Raman scattering intensities from SiO2 are considerably smaller than from Si so, if they are there as one might presume, they were overwhelmed by the Si signal.”

Optical-transmission measurements made on a polished 5 cm section of fiber at a wavelength of 2.936 µm showed (after taking into account Fresnel reflections from the fiber ends) a loss of 4.3 dB/m. The loss could stem from thermally caused microcracks, from SiO2 precipitates, or both. If entirely from precipitates, a calculation shows that the precipitates would have to be about 6 nm in size (in truth, though, the researchers do not yet know what form the oxygen takes in the core).

Future directions for the researchers include boosting the fiber’s transmission, fabricating smaller-diameter fibers, and experimenting with different cladding materials. “The cladding plays a big role,” notes Ballato. “There would be less diffusion, hence oxygen content, if the cladding glass were better matched in its draw temperature to the melting point of the Si core. Also, if its thermal expansion was matched to that of the Si, then it would be in a lower stress state and strength would improve.” In addition, if the cladding were made of glass that was transparent at longer wavelengths, the fiber’s mode “tails” would attenuate less, leading to the possibility of single-mode fibers.

The high Raman gain coefficient of single-crystalline Si (three to four orders of magnitude higher than for glass fibers) makes the Clemson fiber ideal for Raman and other nonlinear optical devices. As for IR transmission, Ballato sees an opportunity in mid-IR biomedical fibers for power transmission at 3 µm–a wavelength for which better passive fibers currently are needed.

John Wallace


  1. J. Ballato et al., Optics Exp. 16(23) 18675 (Nov. 10, 2008).

More in Research