Carbon-coated optical fibers are used here for reducing the luminescence background created by the primary-coating and thus increase the sensitivity of fiber-based spectroscopy systems. The 2-3 orders of magnitude signal-to-noise ratio improvement with standard telecom fibers is sufficient to allow for their use as Raman probes in the identification of organic solvents.
© 2012 OSA
Carbon-coated optical fibers have been developed primarily for use in harsh environments [1–3]. These fibers, with a layer of carbon a few nanometers thick and chemically bonded to the glass surface find sensing applications for instance in oil wells, aerospace and undersea. The main purpose of the carbon film is to prevent water molecules and hydrogen from diffusing into the fiber. Moisture accelerates crack growth on the glass surface, while hydrogen diffusing into the fiber core induces optical loss [4,5].
An apparently dettached field-of-study is that of minimally-invasive spectroscopy using optical fibers. Because typical cross-sections are small (~0.13 mm diameter) and the quality of the optical waveguide excellent, silica fibers are used for light delivery and collection in advanced microscopy, as probes for Raman analysis and in single-cell spectroscopy . In the latter applications, small-core sizes are advantageous, to provide micrometer spatial resolution. In the case of optical analysis using a single fiber for excitation and collection or multiple collection fibers [7,8], one common and serious limitation encountered in fiber-based spectroscopy is the background luminescence of the coating and of the glass fiber itself [9–12]. This luminescence reduces significantly the signal-to-noise ratio of the measurement and ultimately can prevent analysis using optical fibers. Studies with multimode fibers show that luminescence increases with the fiber numerical aperture (NA)  and the coating does not present a serious limitation. For small cores, on the other hand, the polymer buffer is a source of noise and since it is necessary to increase the mechanical strength of the fiber, it is important to minimize the luminescence from the coating. The use of black epoxy resin adhesive and black teflon tubing on a section of an imaging fiber has also been reported to reduce (absorb) the background radiation . Work has also been done on reducing fiber background by choosing the glass type that gives minimal luminescence , using various types of optical interference filters to minimize the fiber background [15,16] and by measuring the anti-Stokes Raman spectra at high-temperature, which eliminates fiber background mainly occurring in the Stokes Raman region .
The effect of a carbon-coating on the luminescence of fibers fabricated industrially and in kilometer-lengths is studied in the present paper. The goal of this study is to try to minimize the disturbance to the measured spectrum of the background imposed by the fiber, and in particular, its coating.
Four types of fiber are used in this study, all with dimensions similar to standard telecom fiber (STF), with 8-μm germanium-doped core and a 125-μm outer diameter. These are an STF (Corning SMF28), a similar germanium-doped-core fiber but with carbon-coating, a depressed-cladding pure-silica-core fiber and a depressed-cladding pure-silica-core fiber with a carbon-coating. All fibers are single-mode at 1.5 µm wavelength but not at 0.491 µm and have NA 0.12-0.14.
Carbon is deposited on the cladding glass surface by chemical vapor deposition from a hydrocarbon gas during fiber drawing. The carbon layer is ~20 nm thick and gives the fiber a dark color, as seen in the inset of Fig. 1 . The bare fiber with its carbon layer is strengthened with an acrylate primary-coating. The broadband absorbance A of the film is estimated by illuminating a standard fiber and a carbon-coated fiber from the side with a HeNe laser and reading the transmitted power Ptransm The transmission through a single carbon-layer at 633 nm wavelength is 75%. Defining A = ln Pi/Ptransm, the typical absorbance of the carbon-coated fibers used here for visible wavelengths is A = 0.29.
The experimental setup is schematically illustrated in Fig. 1. The ~3 mW excitation light source is a continuous-wave solid-state laser (Cobolt Calypso 491) generating 491 nm wavelength by sum-frequency mixing 1.064 nm and 914 nm . The laser light is reflected by a dichroic mirror and coupled into a ~2-m long piece of fiber using a 10x collimating lens. The typical coupling efficiency is 60-70%. The luminescence is transmitted by the dichroic mirror that rejects returning pump light, passes a yellow filter (7 mm OG515) to further remove the remaining pump light, is focused by a 10x collimating lens into a 62.5-µm core multimode fiber and guided into a ~1 nm (0.6 nm) resolution cooled Ocean Optics spectrum analyser (QE65000), operating in the range 530-700 nm. The spectral distance between the laser wavelength and the shortest wavelength accessed by the spectrometer limits the measurements to Stokes shifts greater than 1600 cm−1.
Note that the input lens has NA chosen to increase coupling of the collimated laser light into the core of the fibers. Thus, aside for a small chromatic dispersion, the lens collimates the returning luminescence from the core (not the cladding) towards spectral analysis. Except when using the test fiber to guide light to and from a sample for Raman spectroscopy (below), the remote fiber end is crushed to eliminate reflections. Repeated crushing resulted in identical spectra, so it is concluded that the end-surface does not contribute to the spectra acquired. Data acquisition times are 1 second for relatively strong signals and 10 seconds for weaker signals.
The luminescence of the acrylate primary-coating can be measured free from the contribution from the glass by illuminating it from the side with the focused blue laser beam. In this initial experiment, light does not propagate along the fiber. The result of this preliminary study is illustrated in Fig. 2(a) . The acrylate gives rise to a large number of counts in a smooth spectrum with barely discernible peaks at 555 nm, 580 nm and 630 nm. This large background affects spectral measurements of weak light signals. For comparison, the spectrum is measured from the side under the same experimental conditions for all four fiber types studied, but without the primary-coating. These are shown in Fig. 2(b). In contrast to Fig. 2(a), bare fibers illuminated from the side (pure-silica core or germanium-doped core) give a very flat and negligible background, i.e., the 125-µm thick silica sample produces non-measurable luminescence. Similarly, bare fibers with a thin layer of carbon give a very flat and low-count spectrum, i.e., the carbon-film is non-emmisive in the visible.
Knowing the spectral shape originating from the acrylate coating, the luminescence is studied of the four types of fiber, according to the setup described previously, where the laser light is focused into the core. The comparison is made between an SMF28 fiber with acrylate coating, a pure-silica fiber with acrylate coating, and similar fibers with the 20-nm thick carbon-coated layer between the cladding and the coating. The results are illustrated in Fig. 3 . The spectra produced by SMF28 and pure-silica fiber are seen to be dominated by the strong luminescence from the acrylate. The pure-silica fiber produces a signal about twice as strong as the SMF28, possibly because this particular fiber has a double layer of acrylate, besides being of different origin and age. More importantly, the fibers coated with a carbon-layer create a background two orders of magnitude weaker than those without the carbon-layer.
Examination on an expanded scale of the spectra from the carbon-coated fibers, as seen in Fig. 4(a) , shows three main peaks. Both the narrow peak at 534 nm (1640 cm−1) and the slightly broader peak at 550 nm (2184 cm−1) are present in all spectra of silica fibers taken in this study (sometimes masked by the luminescence from the coating, as in Fig. 3). These are probably associated to Raman scattering in silica, and as showed in measurements with another laser source they are orders of magnitude weaker than the main peak associated to the SiO vibration at 440 cm−1 . The third and main luminescence peak has wavelength ~650 nm, it is broad (100 nm FWHM) and symmetric. This peak most likely originates from the glass core of the fibers and is associated to non-bridging oxygen defficiency centers in germanium-doped SiO2 or in plain SiO2 . In order to confirm that these luminescence peaks come from the glass, the primary-coating of the fibers is removed by immersing the fiber in hot sulfuric acid and rinsing it, and repeating the procedure three times. Figure 4(b) shows the signals from the four types of fiber studied, without the primary-coating. All signals are weak and of similar shape, indicating the presence of the same defect centers in pure-silica fiber (100% silica core) and STF (97% silica core). It is seen that the luminescence of the germanium-doped fiber is considerably (~10 times) more intense than that of the pure-silica fiber. In the absence of the acrylate, the STF with carbon-coating and without carbon-coating show a luminescence signal quite similar to each other. Likewise, without the primary-coating the pure-silica fiber with carbon-film and the pure-silica fiber without the carbon-layer also show similar spectra in shape and amplitude. This indicates that the carbon-film nearly eliminates the contribution of the coating to the fiber background, reducing the number of counts to well below 1%. The fiber luminescence becomes almost completely limited to that from the core.
It is interesting to compare how the luminescence is affected by bending fibers to various radii, since in real-world applications small footprint may require sharp bending. Only fibers protected with the primary-coating can be bent sharply. Although the power transmitted by the fiber is approximately constant for all diameters used, through sharp bending more of the pump light in the cladding reaches the polymer coating and conversely, more luminescence from the coating reaches the detector . The initial 20-cm fiber section is fixated and the remaining fiber length is wound around a cylinder of diameter 10 cm, 6.5 cm, 3 cm or 1.5 cm. Figure 5(a) shows a monotonic increase in luminescence signal for decreased winding radius. Figure 5(b) shows that polymer coating does not contribute to the luminescence measured in fibers with a carbon-layer. All visible signal here comes from the core and the luminescence is constant as long as the guided power is constant, regardless of bend radius.
The noise reduction in carbon-coated fibers is illustrated in an attempt to use SMF28 as a Raman probe. The fiber end is cleaved and a reference trace is taken with the fiber tip placed inside an empty glass vessel. Raman scattering traces for acetone, methanol and ethanol are obtained by inserting the fiber tip into these solvents, contained in identical glass vessels. As shown in Fig. 6(a) the intensity recorded over 10-second integration time in all measurements is large (above 50 000 counts) and the traces are similar to each other. Subtracting the signals from the reference one obtains the traces displayed in Fig. 6(b), plotted on an expanded scale for wavelength and number of counts. The three traces are shifted vertically for easier reading. The Raman peak between 570 and 580 nm associated to a major C-H vibrational mode is barely distinguishable. Note that the noise, proportional to the square root of the number of counts (~230), is comparable to the amplitude of the peaks measured for all three solvents.
The use of a fiber Raman probe incorporating a thin carbon-coating but otherwise similar to SMF28 is advantageous. Figure 7(a) illustrates the raw data obtained with this fiber under similar conditions as in Fig. 6(a). The huge luminescence attributed to the coating is all but eliminated, and the maximum number of counts in all four spectra (three solvents and the reference) does not exceed 1000. The traces are displaced vertically and now the Raman signal at ~570-580 nm assigned to the three solvents is clearly seen. The quasi-periodic spectral peaks seen in all spectra for wavelengths longer than 620 nm is attributed to modal interference in the core. By subtracting the reference signal, as seen in Fig. 7(b), one obtains three spectra that clearly have one, two and three spectral peaks, associated with acetone, methanol and ethanol, respectively. Here, the noise is ~10 times smaller than the signal, again consistent with the square root of the number of counts (~15 counts). The large improvement in signal-to-noise ratio is attributed to the elimination of the large background caused by the acrylate coating.
The results presented in this paper illustrate that carbon-coated fibers developed for the oil and gas sector to render fibers hermetic to diffusion of hydrogen can be advantageously used for fiber-based spectroscopy. The noise level associated with the luminescence of the glass is a couple of orders of magnitude smaller than that caused by the luminescence of the primary-coating. Provided the fiber retains it primary-coating for increased mechanical resistance and ease of manipulation, the ~20-nm thick carbon layer is greatly beneficial in reducing the background noise. The results presented here concentrate on the use of single-mode fiber (at near infrared wavelengths). Although fiber probes (e.g., for Raman spectroscopy) generally consist of multimode fibers with core size equal to or exceeding 50 µm to increase light collection, the study of single cells and bacteria using minimally-invasive optical fibers makes it interesting to migrate to smaller core configurations. It was found in studies not reported here that the background introduced by the luminescence of polyimide is also equally reduced by the thin carbon layer.
The authors gratefully acknowledge Dr. Mårten Stjernström† for valuable help in initial studies of the technique reported here and Professors Fredrik Laurell and Gunnar Björk (Dept. of Applied Physics, Royal Institute of Technology, Stockholm) for support. The authors are also indebted to Acreo’s Fiberlab for the carbon-coated fibers and to the Swedish Research Council (VR) for funding through its Linnæus Center of Excellence ADOPT.
References and links
1. E. A. Lindholm, J. Li, A. S. Hokansson, and J. Abramczyk, “Low speed carbon deposition process for hermetic optical fibers,” in Proceedings of International Wire and Cable Symposium (IWCS), 1999.
2. J. Li, E. A. Lindholm, J. Horska, and J. Abramczyk, Advances in design and development of optical fibers for harsh environments” (Specialty Photonics, 1999). http://specialtyphotonics.com/about/white_papers/Harsh%20Environs.pdf
3. A. Méndez and T. F. Morse, Specialty Optical Fibers Handbook (Academic Press, 2006), Chap. 14.
4. D. A. Pinnow, G. D. Robertson Jr, and J. A. Wysocki, “Reduction in static fatigue of silica fibers by hermetic jacketing,” Appl. Phys. Lett. 34(1), 17–19 (1979). [CrossRef]
5. A. Tomita and P. J. Lemaire, “Hydrogen-induced loss increases in germanium-doped single-mode fibers,” Electron. Lett. 20, 512–514 (1985).
6. A. Wax, M. G. Giacomelli, T. E. Matthews, M. T. Rinehart, F. E. Robles, and Y. Zhu, “Optical spectroscopy of biological cells,” Adv. Opt. Photon. 4(3), 322–378 (2012). [CrossRef]
8. T. F. Cooney, H. T. Skinner, and S. M. Angel, “Comparative study of some fiber-optic remote Raman probe designs. Part I: model for liquids and transparent solids,” Appl. Spectrosc. 50(7), 836–848 (1996). [CrossRef]
9. M. Gallagher and U. Österberg, “Spectroscopy of defects in germanium-doped silica glass,” J. Appl. Phys. 74(4), 2771–2778 (1993). [CrossRef]
10. S. Kannan, M. E. Fineman, and G. H. Sigel Jr., “Defect-related luminescent patterns in bulk silica and optical fibers stimulated by sub-band gap (248 nm) excimer laser radiation,” J. Lumin. 60-61, 433–436 (1994). [CrossRef]
11. T. F. Cooney, H. T. Skinner, and S. M. Angel, “Comparative study of some fiber-optic remote Raman probe designs. Part II: tests of single-fiber, lensed, and flat- and bevel-tip multi-fiber probes,” Appl. Spectrosc. 50(7), 849–860 (1996). [CrossRef]
12. E. Schartner, H. Ebendorff-Heidepriem, and T. Monro, “Sensitive fluorescence detection with microstructured optical fibers,” Proc. SPIE 8028, 802805 (2011). [CrossRef]
14. C. J. de Lima, S. Sathaiah, L. Silveira Jr, R. A. Zângaro, and M. T. T. Pacheco, “Development of catheters with low fiber background signals for Raman spectroscopic diagnosis applications,” Artif. Organs 24(3), 231–234 (2000). [CrossRef] [PubMed]
15. M. L. Myrick and S. M. Angel, “Elimination of background in fiber-optic Raman measurements,” Appl. Spectrosc. 44(4), 565–570 (1990). [CrossRef]
16. R. B. Thompson, M. Levine, and L. Kondracki, “Component selection for fiber-optic fluorometry,” Appl. Spectrosc. 44(1), 117–122 (1990). [CrossRef]
17. S. Dai, J. E. Coffield, G. Mamantov, and J. P. Young, “Reduction of fused-silica-fiber Raman backgrounds in high-temperature fiber-optic Raman spectroscopy via the measurement of anti-Stokes Raman spectra,” Appl. Spectrosc. 48(6), 766–768 (1994). [CrossRef]
18. J. Nordborg and H. Karlsson, “Multiline DPSS lasers – a true Ar-ion alternative” (EuroPhotonics, June/July, 2004). http://www.cobolt.se/Filer/dokument/europhotonics_dual-calypso_july04_cobolt.pdf
19. R. H. Stolen, C. Lee, and R. K. Jain, “Development of the stimulated Raman spectrum in single-mode silica fibers,” J. Opt. Soc. Am. B 1(4), 652–657 (1984). [CrossRef]
20. G. H. Sigel Jr and M. J. Marrone, “Photoluminescence in as-drawn and irradiated silica optical fibers: an assessment of the role of non-bridging oxygen defect centers,” J. Non-Cryst. Solids 45(2), 235–247 (1981). [CrossRef]
21. M. L. Myrick, S. M. Angel, and R. Desiderio, “Comparison of some fiber optic configurations for measurement of luminescence and Raman scattering,” Appl. Opt. 29(9), 1333–1344 (1990). [CrossRef] [PubMed]