Optimum parameters for CO/sub 2/ laser-assisted poling of optical fibers

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 6, JUNE 2002

965

Optimum Parameters for CO2 Laser-Assisted Poling of Optical Fibers Paul Blazkiewicz, Associate Member, IEEE, Member, OSA, Wei Xu, Member, IEEE, and Simon Fleming

Abstract—CO2 laser-assisted poling is a new technique that potentially allows rapid poling and spatially selective poling. The effect of scan-rate, multiple scans, and power fluctuations on CO2 laser-assisted poling were investigated. For a beam irradiance of 52 W cm 2 the minimum necessary dwell time was 0.55 s; multiple scans have no effect unless the poling conditions are changed. CO2 laser-assisted poling was found to be sensitive to perturbations such as laser power fluctuations, which were detrimental. Under optimized conditions the maximum electrooptic coefficient achieved was 0.4 pm/V. Fiber devices up to 90 cm in length have been poled. Index Terms—Electrooptic devices, optical components, optical fiber devices, optical fiber materials, optical fibers.

I. INTRODUCTION

S

ILICA glass plays a key role in photonic systems because of its excellent optical properties, such as low loss, low fabrication cost and high photorefractive damage threshold. Unfortunately, silica, being centro-symetric, has no intrinsic linear electrooptic coefficient or second-order nonlinearity, which is a limitation for active devices [1]. However, it is possible to artificially induce a second-order nonlinearity or linear electrooptic coefficient in silica by poling. Poling can be achieved by applying an electric field across a dielectric material, whilst using an excitation source to activate the poling mechanism. In structurally anisotropic media this leads to the reorientation and alignment of dipoles and domains; this type of poling is referred to as dipole orientation. This type of poling process occurs in ferroelectric and pyroelectric materials. A second mechanism for poling is due to formation of space-charge layers in the dielectric. These spacecharge layers produce a frozen-in field that acts on the surrounding dielectric. This frozen-in field caused by the trapped (third-order nonlinearity) to prospace charge, acts on the duce a nonzero (second-order nonlinearity). Recent work suggests that the space-charge model is highly applicable to thermally poled silica [1]–[5].

Manuscript received December 12, 2000; revised October 16, 2001. This work was supported by the New Energy and Industrial Technology Development Organization of Japan and by the Australian Research Council. The work of P. Blazkiewicz was supported by an Australian Postgraduate Award (scholarship). P. Blazkiewicz and W. Xu were with the Australian Photonics Cooperative Research Centre, Optical Fiber Technology Centre, University of Sydney, New South Wales 2006, Australia. They are now with Innovative Specialty Optical Fibers and Components (iSOFC), Salem, NH 03079 USA. S. Fleming is with the Australian Photonics Cooperative Research Centre, Optical Fiber Technology Centre, University of Sydney, New South Wales 2006, Australia. Publisher Item Identifier S 0733-8724(02)05392-6.

Fig. 1.

Diagram of typical twin-hole fiber cross section (not to scale).

Thermal poling has become one of the most popular and reliable methods of poling silica glass and it provides values of 1 pm/V [5 ]. A disadvantage of thermal poling in optical fibers is that it is slow, on the order of many minutes to reach saturation. Also thermal poling has been typically limited to uniform poling due to uniform heating obtained from conventional heating methods. This makes the fabrication of periodically poled structures in glass difficult [6]. A recent development in the field of poling of silicate glass has been CO laser-assisted poling (CLAP) [6]. CLAP is a novel method, which has been demonstrated to be a very rapid and localized method of poling silicate optical fibers. In this paper, the CO laser beam scan-rate was varied to find the minimum dwell time required. The effect of multiple scans and laser power instability was examined. Finally, poling of long length twin-hole fiber devices using the CLAP method was demonstrated. II. EXPERIMENTAL SETUP A. Fiber Details The twin hole fiber used was fabricated using the modified-chemical-vapor-deposition (MCVD) technique. It had a MCVD deposited germanosilicate core codoped with phosphorus, and a cladding formed from fused-silica tube (General Electric Optin-100). Fig. 1 illustrates the geometry and orientation of the fiber with respect to the CO laser beam. The hole diameter was 108 m, the hole-to-hole spacing was 16 m, and

0733-8724/02$17.00 © 2002 IEEE

966

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 6, JUNE 2002

Fig. 2. Diagram of CO laser-assisted poling experiment.

the closest hole to core spacing was 1 m. Aluminum wires were inserted via side entry and had a diameter of 50 m [1]. In our fibers, the core is intentionally displaced toward one of the holes. If a positive voltage is applied to the hole closer to the core, we define this as positive poling. If the polarity is reversed, then we define this as negative poling. B. Mach–Zehnder Interferometer The CO laser-assisted poling experiment was set up as shown in Fig. 2. A Mach–Zehnder interferometer was used for measurement of the evolution of the E-O coefficient in the optical fiber. A translation stage was used to scan the 3-mm CO laser beam along 7 cm of the optical fiber while a poling voltage was applied. This Mach–Zehnder setup allows us to measure the average electrooptic coefficient of a fiber device of set length. When using this Mach–Zehnder for in situ measurements, we obtain a linear growth of the average electrooptic coefficient as the laser beam scans along the length of the device if the E-O is uniform along the scan. If the device is scanned without a poling voltage, no growth in the electrooptic coefficient is observed. III. RESULTS AND DISCUSSION A. Scan Rate and Multiple Scans Using a constant beam irradiance of (52 2) W cm , we varied the laser beam scanning velocity between 0.80 and 17.3 mm s . This varied the dwell time of the 3-mm-wide beam from 3.75 to 0.173 s. Device lengths used were 7.0 cm long and the poling voltage was kept constant at 3.0 kV. We used 3 kV to avoid electrical breakdown through the glass, which commonly occurred at 3.5 kV [6]. Fig. 3 shows a plot of residual electrooptic coefficient after poling versus the dwell time used while scanning the CO laser beam. Fig. 3 is plotted using a logarithmic scale on both the horizontal and vertical axis. Fitting straight lines to the saturated region and the nonsaturated region gives a rollover point for the optimum dwell time to be approximately 0.55 s. A dwell time less than this gives a residual electrooptic coefficient below

Fig. 3.

Plot of electrooptic coefficient versus laser-beam dwell time.

the maximum possible; any value larger than 0.55 s produces a maximum residual electrooptic coefficient. The maximum saturated values obtained for this experiment were approximately 0.15 pm/V. This value was obtained after averaging the results of a number of fiber devices poled under identical conditions. For standard thermal poling, one can consider three poling time regions [4]. The first is underpoling, where poling has been stopped too early before optimum overlap between the space charge region and the core has occurred. This generally occurs for fibers poled for less than 20 min. Then there is the poling time period where saturation has occurred; this is when there is optimum overlap between the space charge field and the core. This occurs for poling times in the range 30–60 min. Finally, the last poling time interval is overpoling; in this case, the space charge region has been pushed too far into the glass and no longer has optimum overlap with the fiber core. With the CLAP method of poling, we have only observed under poling and optimum poling for the range of dwell times under consideration. It has been previously demonstrated with resistive thin film heaters that it takes in excess of 0.4 s to uniformly heat optical fiber cross sections [7]. The reason for underpoling with short dwell times may be due to the actual time

BLAZKIEWICZ et al.: CO LASER-ASSISTED POLING OF OPTICAL FIBERS

Fig. 4.

967

Evolution of optimized positive CO laser-assisted poling.

it takes for the heated section of the fiber to reach the optimum poling temperature. If the dwell time is too short, then the optimum temperature cannot be reached and then the residual electrooptic coefficient is less than the maximum or saturation value that can be achieved. In addition, it has been observed that the residual electrooptic coefficient after the first scan cannot be increased if the same poling conditions are repeated with multiple scans. This is true even for fibers CLAP poled under unoptimized conditions and which have not reached the saturation value after the first scan. This is in contrast to the behavior in thermal poling, where poling can be interrupted and the fiber device underpoled, and it is possible to continue poling and saturate the value of the residual electrooptic coefficient.

Fig. 5. Plot of electrooptic coefficient versus applied external voltage for device poled under optimized conditions.

B. Frozen-In Voltage and Laser Instability From previous work on thermally poled devices, it has been observed that the maximum residual electrooptic coefficient is achieved when the total frozen-in voltage due to the space charge is equal in magnitude to the poling voltage. This occurs only when the poling conditions have been optimized. Otherwise, the frozen-in voltage measured is always smaller than the magnitude of the applied poling voltage. Fig. 4 shows the typical optimized E-O evolution achieved during exposure of the fiber with a positive applied voltage. First, the E-O coefficient jumps up when the DC poling voltage is applied. When the beam is unblocked and the scan begins, the E-O coefficient grows rapidly. When the scan ends and the beam is blocked, the E-O coefficient stops growing. Finally, upon turning the poling voltage off, a residual E-O coefficient remains, in this case 0.4 pm/V. The irradiance used for the device in Fig. 4 was (54 2) W cm in the center of the optimum irradiance window and using a dwell time of 2.2 s [6]. The total electrooptic coefficient is a combination of the residual electrooptic coefficient and the DC-induced electrooptic coefficient, which is observed instantly when applying an electric field across the silica glass. Measuring the total electrooptic coefficient for a number of applied voltages and then plotting the total electrooptic coefficient versus the applied voltage, gives a straight line whose slope is proportional to

Fig. 6. Plot of electrooptic coefficient verses applied external voltage for device poled under nonoptimized conditions.

the DC of the silica glass. The horizontal intercept of this line gives the frozen-in voltage, which we can compare to the poling voltage used. The vertical intercept gives the residual electrooptic coefficient at zero applied volts after poling. Fig. 5 is a plot of electrooptic coefficient versus applied voltage before and after CLAP poling of the device used in Fig. 4. Here, we can see that a 3 kV voltage is needed to fully cancel the effect of the frozen-in voltage. This indicates that the frozen-in voltage is 3 kV and equal to the poling voltage. From the vertical intercept, the residual electrooptic coefficient with no externally applied voltage is 0.4 pm/V. In practice, one typically obtains smaller values of electrooptic coefficient due to laser power fluctuations and a slight misalignment of the laser beam. This causes the irradiance to vary spatially along the fiber, causing the E-O to vary spatially along the poled region. This reduces the average frozen-in voltage and average E-O of the poled fiber device. Fig. 6 is a plot of E-O verses applied voltage before and after CLAP poling, where the irradiance fluctuated across the scan. Here, the poling voltage was 3 kV and we needed to apply only 2.3 kV to cancel the effect of the frozen-in voltage.

968

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 6, JUNE 2002

C. CO Poling of Long Length Devices It has been proposed that CLAP could be used to pole long lengths of fiber [6]. Using the minimum saturation time of 0.55 s, the fastest poling rate would be 0.33 m/s, which is one to two orders of magnitude slower than a typical fiber-drawing rate. This would suggest that CO laser poling would need to be done after fiber pulling, as the pulling rates would have an insufficient dwell time for maximum poling in addition to other problems that may arise. We have been able to CLAP pole device lengths up to 90 cm, which have been limited by our optical bench width and translation stage length. When measuring the average residual electrooptic coefficient of such long devices, it is difficult to obtain values larger than 0.1 pm/V even with optimized conditions. The reason for the lower electrooptic coefficient is believed to be due to additional variations of beam irradiance while scanning along the fiber device. To circumvent this problem, it would be necessary to keep the laser beam stationary and translate the fiber, or to keep the fiber stationary and physically translate the laser while keeping the distance between the laser and the fiber constant. A third method would be to expand the laser beam to decrease its divergence and then refocus closer to the correct power density at the point of exposure. IV. CONCLUSION In this paper, we have investigated the effects of scan rate, multiple scans, irradiance, power fluctuations, and the usefulness of CLAP when poling long device lengths. It was found that for a 3-mm-wide Gaussian beam, the minimum necessary dwell time needed was approximately 0.55 s. We believe this dwell time corresponds to the time it takes to uniformly heat the cross section of the fiber to the correct poling temperature. Also, multiple scans had no effect on the final residual electrooptic coefficient unless one of the poling parameters was changed for any subsequent scans. The maximum electrooptic coefficient that we have obtained for the CLAP technique has been 0.4 pm/V. When testing such devices, we have found that the frozen-in voltage is equal in magnitude to the poling voltage indicating that the poling conditions have been optimized. The residual electrooptic coefficient obtained using the CLAP technique is very sensitive to perturbations such as laser power fluctuation. When such perturbations occur during poling, then the final average electrooptic coefficient of the device is reduced due to a reduction in the average frozen-in voltage.

We have poled up to 90 cm lengths of fiber using the CLAP technique obtaining average electrooptic coefficients of approximately 0.1 pm/V. Lower average electrooptic coefficients were obtained due to variations in the laser beam irradiance along the fiber scan. These variations translated into a spatially varying electrooptic coefficient along the poled fiber device. Any regions not poled using an optimum irradiance have a lower electrooptic coefficient which decreases the average electrooptic coefficient. To obtain optimum results when using the CLAP technique, it is imperative that a stable laser source is used, its irradiance is kept within the optimum range over the entire scan length, and the beam dwell time is kept above 0.55 s. REFERENCES [1] D. Wong, W. Xu, and S. Fleming, “Charge dynamics and distributions in thermally poled silica fiber,” in SPIE Conf. Optical Devices for Fiber Communication, vol. SPIE 3847, Boston, MA, Sept. 1999, pp. 88–93. [2] T. G. Alley and S. R. G. Brueck, “Visualization of the nonlinear optical space-charge region of bulk thermally poled fused silica glass,” Opt. Lett., vol. 23, no. 15, pp. 1170–72, 1998. [3] P. G. Kazansky, A. R. Smith, P. St. J. Russell, G. M. Yang, and G. M. Sessler, “Thermally poled silica glass: Laser induced pressure pulse probe of charge distribution,” Appl. Phys. Lett., vol. 68, no. 2, pp. 269–71, 1996. [4] D. Wong, W. Xu, S. Fleming, M. Janos, and L. Kai-Ming, “Frozen-in electrical field in thermally poled fibers,” Opt. Fiber Technol., vol. 5, no. 2, pp. 235–41, 1999. [5] V. Pruneri, F. Samoggia, G. Bonfrate, P. G. Kazansky, and G. M. Yang, “Thermal poling of silica in air and under vacuum: The influence of charge transport on second harmonic generation,” Appl. Phys. Lett., vol. 74, no. 17, pp. 2423–25, 1999. [6] P. Blazkiewicz, W. Xu, D. Wong, J. Canning, M. Åslund, and G. Town, “Carbon-dioxide laser assisted poling of silicate-based optical fibers,” Opt. Lett., vol. 25, no. 4, pp. 200–202, 2000. [7] J. A. Rogers, P. Kuo, A. Ahuja, B. J. Eggleton, and R. J. Jackman, “Characterization of heat flow in optical fiber devices that use integrated thin-film heaters,” Appl. Opt., vol. 39, no. 28, pp. 5109–5116, 2000. [8] W. Xu, M. Janos, D. Wong, and S. Fleming, “Thermal poling of boroncodoped germanosilicate fiber,” IEICE Trans. Commun., vol. E82-B, pp. 1283–1286, Aug. 1999.

Paul Blazkiewicz (S’98–A’01), photograph and biography not available at the time of publication.

Wei Xu (M’99), photograph and biography not available at the time of publication.

Simon Fleming, photograph and biography not available at the time of publication.