A twin-hole fiber was provided with Au-Sn alloy electrodes and thermally poled at 255 °C with 4.3 kV applied during 155 minutes. An electric field 6×107 V/m was recorded. The poled fiber was cleaved and etched, revealing that the depletion region overlapped the entire core, was wedge shaped and pointed towards the cathode. The recorded profile closely followed the spatial distribution of the poling field.
©2004 Optical Society of America
Since the first observation of second harmonic generation in optical fibers there has been much interest in the process of poling fibers to create a residual linear electrooptic coefficient. Most reports on poling of optical fibers [1, 2] involve thermal poling , known to be the most repeatable and reliable method to induce a relatively large second-order nonlinearity in glass. However, in comparison with ferroelectric crystals and poled polymers, the nonlinear coefficient is still very low (~0.5 pm/V)  and the induced nonlinear layer thin, which makes it difficult to guarantee complete overlap between the recorded electric field and the core of the fiber. Another difficulty of poled fibers is that the fabrication is manual and time consuming, involving for example polishing of d-fibers  or insertion of thin wires into the fiber holes . The technique of using alloy electrodes  makes it possible to produce many devices in parallel. Typically, ten fiber devices are filled with metal simultaneously in less than 1 minute. The method can be further developed for industrial manufacturing. Using alloy electrodes enables the production of electrodes more than 1 m long, so that even a weak nonlinearity is potentially sufficient to create a useful device. Furthermore, the electrodes fill up the entire hole and are in contact with the glass in the whole active region, which increases the reproducibility and efficiency. Fibers with long alloy electrodes have been used with the Kerr-effect [5, 6], but so far alloy electrodes have not been used with poling, which requires temperatures comparable to the melting point of the alloy. To the best of our knowledge the present paper demonstrates for the first time that it is possible to thermally pole a fiber with alloy electrodes. The non-linear layer created has comparable width as in thermally poled bulk samples and overlaps the entire core.
The fiber used here had outer diameter 125 µm and was fabricated by drilling a MCVD preform. A high temperature acrylate coating was applied to make the fiber heat resistant during electrode insertion and poling, and to keep it transparent after the thermal treatments. The fiber was designed with the anode hole placed closer to the core to ensure good overlap between the core and the poled region (Fig. 1). In order to minimize the loss caused by the electrodes the core was highly doped with Ge, to give a refractive index step Δn=0.025 and N.A.=0.27 which confines the optical mode mainly to the core. The diameter of the core is 3.6 µm and the fiber is single-mode at 1550 nm. The edge-to-edge distance between the core and the nearest hole is 4.6 µm and farthest hole 10.1 µm. The two holes are 29.5 µm in diameter and have a separation d=18.3 µm.
The alloy used to create the electrodes consisted of 80 % wt Au and 20 % wt Sn, with a melting point of 280 °C and resistivity 1.6×10-5 Ω cm. This alloy was chosen because it can be inserted at 300 °C in molten form and poling can take place with the metal in its solid phase at a temperature 250–270 °C, close to the optimum for poling . Previously, other alloys tested by us (BiSn and AgSn) with melting points (137 and 220 °C) below the poling temperature failed to produce a second-order nonlinearity. This can be attributed to the injection of cations from the molten electrode during poling, which prevented the formation of a depletion layer. The AuSn alloy was placed in a small pressure cell heated to 300 °C and pushed into the holes of the heated fiber by applying a pressure of 6 bars . The length of the electrodes in the present work was 20 cm, a compromise between maximizing the phase shift and keeping the loss caused by the electrodes reasonably small. Electric contact with the electrodes was established by side-polishing the fiber  and then attaching 20 µm thick gold coated tungsten wires to the alloy using conductive epoxy.
A hotplate was used to heat the contacted fiber to 255 °C. The fiber was poled by applying +4.3 kV to the anode electrode and grounding the cathode. After 155 minutes the heater was switched off and the fiber allowed to cool down to room temperature with the voltage still applied . All fibers were poled with the high-temperature resistant primary coating in place and unharmed.
In order to characterize the effect of poling, the fiber component was used as the active arm of a ~1m long, unstabilized 2×2 Mach-Zehnder interferometer (MZI) [1, 4, 9]. Since the main envisaged application is in communications, for example in protection switching, the characterization was carried out at 1550 nm, where the fibers used were single mode. For all experiments a tunable laser amplified by an EDFA was used as a light source. The MZI consisted of two 3-dB couplers with a standard single-mode fiber as the passive arm. Characterization of the active fiber was done both before and after poling by applying a +3 kV linear ramp to the anode electrode and detecting the optical intensity in one of the two output ports, described by (assuming total extinction):
where Io is the modulation amplitude, L is the length of the active region, λ is the operating wavelength of the interferometer (1.55 µm), no is the unperturbed refractive index, χ (3) is the third order nonlinear coefficient and Eext and Erec the externally applied- and the recorded field, respectively. Δϕo is the residual phase change due to the difference in arm lengths and thermal/mechanical instabilities. Before poling, Erec =0 and Δϕ∝ while after poling Δϕ;∝(Eext +Erec )2. An externally applied voltage is assumed to give an electric field inside the fiber, Eext =Vext/d, where d is the distance between the holes. Likewise, it is assumed that Erec =Vrec/d.
In Fig. 2 it is seen that for the unpoled fiber Δϕ≈1.5 π for 3 kV applied, while for the same applied voltage, Δϕ≈2.5 π for the poled fiber. For this 20-cm long fiber, Vπ≈1.37 kV at 1550 nm. The inset of Fig. 2 shows the quadratic relation between the phase and Vext derived to give the best fit of equation 1 to the two experimental curves shown in Fig. 2. The parabolas give a value to χ (3) and Erec, which determine the opening and the position of the minimum of the curves, respectively. We obtain a good fit for a value of χ (3) in the range (2.1 ±0.3)×10-22 m2/V2 before and after poling, in relatively good agreement with the value expected for Ge-doped silica.
We do not attribute too much significance to the slightly lower value obtained after poling here (i.e., χ (3) apparently decreases), since we measured in a few other devices an apparent marginal increase, which could be attributed to a change in the electric field distribution during poling. Furthermore, we found that a fit to such a limited phase excursion is very imprecise when determining χ (3) and Erec. Fig. 3 illustrates the small difference in fitting curves obtained when Vrec is varied widely from 1200 V to 1950 V, leading to a χ (3) value that varies from 1.86 to 2.36×10-22 m2/V2. The conclusion one draws is that such a fit is too inaccurate for the determination of Vrec, and an alternative measurement is desired.
To independently measure the field recorded after poling, the fiber was kept in the same MZI as described above, and probed with a 400V AC peak-to-peak voltage at 1 kHz. The peak-to-peak intensity modulation resulting from the AC voltage was recorded for different applied DC voltages. By slightly adjusting the wavelength of the laser, the interferometer was placed in quadrature before each measurement. Minimum modulation of the optical signal is achieved when the external DC bias cancels the recorded field. In this case, the external voltage was measured to be -1.1 kV, as seen in Fig. 4, and thus the recorded voltage V rec=+1.1 kV.
The spatial distribution of the effective nonlinearity recorded was studied in order to attempt to optimize the poling procedure. To this end, the acrylate coating was removed and the fiber cleaved several times in the ~20 cm long poled region. In order to reveal the recorded electric field the cleaved ends of the fiber pieces were etched in 40 % hydrofluoric acid (HF) for 45 seconds [3, 10]. The fiber pieces were then rinsed in deionized water and dried at 150 °C in air, after which the samples were inspected with an optical phase contrast microscope. The cleaves were not always perfect because of the presence of the two solid metal electrodes that filled the holes.
Etching revealed that in the present experiment the depletion region is wedge-shaped and points towards the cathode (Fig. 5, left), in contrast to the round shape previously reported . The depletion layer extends furthest between the electrodes - 13 µm. To the best of our knowledge this is by far the widest depletion region recorded in a fiber. It is obvious from Fig. 5 that here the poling time was excessive, and that the depletion region is unnecessarily wide. The recorded field in the core could be increased with shorter poling times. At the side of the anode opposing the core the depletion layer is 10.2 µm thick, indicating that the electric field during poling must have been almost as strong as between the two electrodes, which is consistent with the result of . The migration speed of the alkali ions that gives rise to the depletion region has been reported to depend on the electric field . The field distribution in the fiber during poling was simulated (Fig. 5, right), and good agreement with the observed etched pattern was observed when the outer surface potential was assumed to be equal to the cathode potential (electrically grounded).
The wide wedged-shaped depletion region profile can be attributed to the field distribution during poling , which is in turn associated with the fiber design. The fact that the alloy electrodes are inserted into the 125 µm fiber as a liquid makes it possible to design relatively small holes, where wire insertion over long lengths would be difficult.
Since it is possible that the field distribution between the holes becomes inhomogeneous after poling because of the higher resistivity in the poled region due to charge depletion, it is not possible to directly determine the recorded electric field. However, the field can be estimated by neglecting the influence of the different resistivities after poling and assuming that the field distribution between the two electrodes (d=18 µm) obeys the condition ∫E dl=0, verified experimentally in . As shown in ref , the electric field is created both inside and outside the depletion region in thin poled bulk samples. In the present case, the measured voltage V rec=1.1 kV gives E rec=V rec/d=0.61×108 V/m in the depletion region. In the 5 µm thick region between the depletion region and the cathode, from the condition ∫E dl=0, one infers E=-(13/5) E rec=-1.59×108 V/m. This implies that the charge displaced by poling in the fiber is sufficient to create a field equal to the sum 2.2×108 V/m, comparable to the values reported for bulk samples. The recorded electric field couples with χ (3) to give an effective electrooptic second-order nonlinearity expressed as:
With E rec=0.61×108 V/m as calculated above, the second-order nonlinearity experienced by the core is 0.02 pm/V, approximately a factor of 10 lower than previously reported . This is believed to be caused by uneven temperature in the hotplate used to heat the fiber, causing different parts of the fiber to reach the optimum recorded field at different points in time. Work is underway to improve the oven to overcome this problem.
In conclusion, a twin-hole fiber provided with alloy electrodes was thermally poled. 1.1
kV was needed to cancel the recorded field, which was estimated to be 6×107 V/m across the core. The electrooptic response of the fiber component was tested in a fiber Mach-Zehnder interferometer at 1550 nm, and Vπ~1.37 kV after poling. The poled fiber was cleaved and etched, revealing that the depletion region overlapped the entire core, was wedge shaped and pointed towards the cathode, with a profile similar to the electric field distribution during poling. The successful use of alloy electrodes opens the doors to manufacturing of long and homogeneous poled fiber devices. Work is under way to reduce the switching voltage by optimizing the poling time and increasing the length of the devices to ~1m.
Financial support by the European Commission IST project GLAMOROUS (2000-28366) is gratefully acknowledged. We are indebted to M. Fokine, O. Tarasenko and L-E Nilsson (Acreo) for early experimental help and to M. Kristensen, J. Fage-Pedersen (COM, Denmark) and B. Lesche (UFJF, Brazil) for fruitful discussions. We would like to thank in particular Y. Quiquempois (USTL, France) for valuable comments and input to the work.
References and Links
1. X.-C. Long, R. A. Myers, and S. R. J. Brueck, “A Poled Electrooptic Fiber,” IEEE Phot. Tech. Lett. 8, 227 (1996) [CrossRef]
4. T. Fujiwara, D. Wong, and S. Flemming, “Large Electrooptic Modulation in a Thermally-Poled Germanosilicate Fiber,” Phot. Tech. Lett. 7, 1177 (1995) [CrossRef]
5. M. Fokine, L. E. Nilsson, A. Claesson, D. Berlemont, L. Kjellberg, L. Krummenacher, and W. Margulis, “Integrated fiber Mach Zehnder interferometer for electro-optic switching,” Opt. Lett. 27, 1643 (2002) [CrossRef]
6. L. Li, R. D. Birch, and D. N. Payne, “An all fibre electro-optic Kerr modulator,” in IEE Colloquium on Àdvanced Fibre Waveguide Devices79, p 10/1–4 (1986)
7. R. A. Myers, S. R. J. Brueck, and R. P. Tumminelli, “Stable second-order nonlinearity in SiO2-based waveguides on Si using temperature/electric-field poling,” in Doped Fiber Devices and Systems, Proc. SPIE2289, 158 (1994)
8. D. Wong, W. Xu, S. Flemming, M. Janos, and K.-M. Lo, “Frozen-in Electrical Field in Thermally Poled Fibers,” Opt. Fib. Tech. 5, 235 (1999) [CrossRef]
9. P. G. Kazansky, P. St. J. Russel, and C. N. Pannell, “Optical fibre elecrets: observation of electro-acousto-optic transduction,” Electron. Lett. 30, 1436 (1994) [CrossRef]
10. P. Blazkiewicz, W. Xu, D. Wong, and S. Flemming, “Mechanism for the thermal poling in twin-hole silicate fibers,” J. Opt. Soc. Am. B 19, 870 (2004) [CrossRef]
11. D. Faccio, V. Pruneri, and P. G. Kazansky, “Dynamics of the second-order nonlinearity in thermally poled silica glass,” Appl. Phys. Lett. 79, 2687 (2001) [CrossRef]
12. W. Xu, J. Arentoft, D. Wong, and S. Flemming, “Evidence of Space-charge Effects in Thermal Poling,” Phot. Tech. Lett. 11, 1265 (1999) [CrossRef]
13. Y. Quiquempois, A. Kudlinski, and G. Martinelli, “Zero potential condition in thermally poled silica samples: Evidence of a negative electric field outside the depletion layer,” Submitted for publication in J. Opt. Soc. Am. B (2004)