1. Introduction

Advanced time-resolved microscopy techniques make often use of fluorescent organic molecules or labeling quantum-dots with lifetimes in the range from few- to tens of nanoseconds. This is one example of application area where a variable repetition rate mode-locked laser source is desired. One needs to reduce the repetition rate of the pump laser to prevent superimposing luminescence from consecutive pulses and at the same time have high enough repetition rate to minimize the data acquisition time. It is therefore valuable to have an adjustable pulse selection mechanism for gating the mode-locked laser pulses.

Light gating in the nanosecond domain has changed little over several decades. The bulk solution is generally employed of using deflection in an acousto-optic modulator (AOM) or polarization rotation in an electrooptic modulator (EOM) placed between crossed polarizers. The bulk approach provides high extinction ratios. However, in-fiber modulation would allow simplifying most systems where light is coupled to an external bulk modulator, and in particular fiber lasers. Typical AOMs lead to large excess loss, are expensive, have large size and if coupled as bulk components, increase the cost of manufacturing by requiring alignment and mechanical parts. If optical fibers could perform active gating, then one could expect low losses, low cost, ease of manufacturing and a significantly improved laser system.

Unfortunately, all-fiber single pulse selection at high repetition rates (hundreds of kilohertz and above) is not an easy task. Acousto-optic fiber devices based on a piezoelectric element placed against the fiber are often used for phase modulation in gyro applications at relatively low speeds (kHz) [1], for Q-switching [2, 3] or mode-locking of fiber lasers [4, 5]. However, the modulation achieved is sinusoidal and not digital as best suited for pulse selection and piezoelectric elements also tend to exhibit hysteresis and long-term drift. Furthermore, the long risetime limits their use for single pulse selection to hundreds of kHz. A high-speed nanosecond fiber switch has recently been reported to gate and select individual mode-locked pulses from a laser [6, 7], based on heating an internal electrode to induce birefringence [8]. However, the repetition rate achieved is also limited to below tens of kHz.

In the present letter, an electrooptic silica fiber switch based on a Sagnac interferometer is reported, capable of selecting individual pulses from a mode-locked laser in the ~1 MHz range. Although the work here is carried out at 1.5 µm for convenience, simple extension to shorter wavelengths (~1 µm) is anticipated.

2. The fiber phase modulator

Single pulse selection at adjustable rates is achieved with electrooptic fibers. While untreated silica fibers only exhibit a weak quadratic Kerr effect (third-order optical nonlinearity) [9, 10] it is possible to induce the linear electrooptical effect by poling [11–14]. Here, for thermally poling the fiber [15–18] we use electrostatic charging [19], where the anode of the high-voltage supply is connected to both internal electrodes of the fiber and the surrounding air supplies the electrons that effectively ground the fiber side-surface. Previous work [19] shows that Ge-doped fused silica preforms contain cations in sufficient concentration to allow recording a strong electric field (>10⁸ V/m) by electrostatically charging the drawn fiber. After recording the linear electrooptic effect in the fiber, the two internal electrodes can be used at room temperature in a conventional way to apply a desired electrical gating signal to induce a significant phase change.

A 78 cm piece of suitable twin-hole fiber is provided with internal electrodes by melting Au_0.8Sn_0.2 at 280°C and pumping it as a liquid into the fiber holes. After cooling to room temperature, the electrodes become solid. The beginning and end of the fiber are free from metal, so that standard splicing allows coupling light into and out of the metal-filled fiber with ~0.5 dB splice loss. The total insertion loss of a 78 cm fiber component is ~4 dB loss at 1.5 µm and ~1.3 dB loss at 1 µm. This fiber with electrodes is then side-polished, electrically contacted, and heated again to 265°C while subjected to high voltage bias in the unconventional electric circuit named above that lacks a true contacted cathode [19]. By measuring at 1.5 µm wavelength more than 10 fibers of length ~78 cm and determining a mean value V_π = 150 V, the average nonlinear coefficient recorded after 195 minutes is found to be χ⁽²⁾_eff = 0.26 pm/V. It can be emphasized that the electrooptic phase modulators thus constructed are robust, reproducible and exhibit no measurable decay in the second order nonlinearity over a period of several (~7) years.

3. Experiments

Electrical characterization

Making use of a longer electrooptical fiber decreases V_π but increases the loss. Longer electrodes also mean longer risetime due to larger capacitance. The importance of the length of the device and the associated capacitance is illustrated in Fig. 1 , where the risetime of a fiber with 78-cm long electrodes is studied. Figure 1(a) shows the 10 V step from a 50 Ω pulse generator (blue trace) that is applied to the electrodes of the fiber and the current flowing to the device (green trace), monitored with a Tektronix CT-2 current probe. The current decays approximately exponentially. One can estimate the average voltage developed across the electrodes as $V(t)=1/C{\displaystyle \int idt},$ (red trace). The estimated voltage reaches 50% of the full value after ~7 ns, implying a RC time constant of 10 ns. Gating a higher voltage and switching off the bias after a short time to utilize the steep slope in the beginning of the response is a way to reduce the risetime of the device and to reach V_π more rapidly, as shown in Fig. 1(b). There, a 13 ns voltage pulse is applied to the modulator. The peak voltage is reached after 16 ns and the pulse has 14 ns full width at half maximum (FWHM).

 

Fig. 1 (a) Measured input voltage (blue) and current (green) when applying a step function to the modulator. The voltage developed over the electrodes is calculated (red) as the normalized integral of the current. (b) Same as previous with the application of a 13 ns pulse.

Download Full Size | PDF

Further manipulation of the shape of the electrical pulse applied to the fiber electrodes allows modifying the rise- and falltime of the voltage developed, but in this work no further effort is placed on modifying the electrical response of the devices from that shown in Fig. 1(b).

Amplitude modulation with Sagnac loop

A fiber Sagnac interferometer [20] is used in order to transform phase- to amplitude modulation, gated with the activation of the electrooptic fiber in the loop. Previous results obtained with an electrooptic fiber in a Sagnac interferometer [21] suggest that the configuration is appropriate for on-off gating. A schematic of the setup is displayed in Fig. 2(a) , where a polarization controller (PC) is used to bias the phase to minimize the transmitted signal measured at photodiode (PD) prior to the application of voltage pulses. The fiber coupler recombines equal optical powers to the output even if the coupling ratio deviates from 3 dB. In addition, all wavelengths experience approximately the same delay, since the clockwise and counter-clockwise propagating beams travel along the same short length loop and dispersion is limited to ~102 fs/nm (assuming a 30 ns loop). Therefore the bandwidth of the switch is large and fs pulses are supported [22]. Most importantly, the interferometer is robust against mechanical and thermal fluctuations. The gating performance of the loop probed by CW laser light is illustrated in Fig. 2(b). The switching efficiency is weakly wavelength dependent from 1500 nm to 1560 nm because of the spectral dependent loss of the fiber with electrodes, the polarization controller and the fiber coupler.

 

Fig. 2 (a) Schematic diagram of a Sagnac fiber interferometer with an electrooptical fiber as phase control element and (b) Transmission of the Sagnac loop in the temporal and spectral domain when the modulator is driven by 105 V pulses without a 50 Ω resistive load.

Download Full Size | PDF

The visible tail in Fig. 2(b) is due to the phase modulator being driven here by a pulse generator in an open-end configuration. The long tail is related to the return to zero of the voltage pulse delivered and can be adjusted with the use of matching resistors, not attempted in this early experiment but implemented below.

In the present study, traveling-wave aspects of the electrooptic component are not considered, assuming that the length is relatively short (propagation time ~3 ns) compared to the Sagnac loop transit time (~30 ns) and the electrical drive pulse duration.

Pulse selection at 1 MHz

The loop can work in a push-pull mode by placing two similar phase modulators symmetrically in the Sagnac interferometer, one of which with reversed polarity, as illustrated in Fig. 3 . Here, the phase-shift gained by the light traveling clockwise in one modulator is added to the phase-shift gained by light traveling in the counter-clockwise direction in the second modulator. In this arrangement, the switching voltage is approximately halved, and the measured V_π at 1.5 µm is ~80 V. This Sagnac loop has a total transit time of ~55 ns with a ~35 ns delay between the modulators.

 

Fig. 3 Setup for selecting every 7th pulse from a pulsed DFB laser. The light transmitted by the circulator is either reflected or transmitted by the Sagnac loop before detection in photodiode

Download Full Size | PDF

This setup is used to perform pulse selection at a rate 1 MHz. The laser source used in this demonstration is an amplified DFB laser emitting 4 ns pulses at 1550 nm. By choosing a laser diode rather than a mode-locked laser increases the flexibility of the system, which can be run as chosen here at a repetition rate of 7 MHz simulating a laser with roundtrip time ~140 ns.

Synchronization is achieved electrically, not by adjusting the loop length. The voltage pulse has 13 ns FWHM and is delivered through a 50 Ω coaxial cable and terminated by a 50 Ω resistance across the electrodes. For the ideal setting of the PC in the loop the voltage pulse gives rise to two identical transmission windows of the interferometer separated by the modulator delay ~35 ns [21]. The refractive index of the two electrooptic fibers is dependent on the average electrical-power because heat develops as the voltage-pulse travels across the finite resistance (~235 Ω) of the fiber-electrodes. The drift is therefore compensated by adjusting the polarization controller in the loop.

By tuning the delay electronically, every 7th pulse of the input train illustrated in Fig. 4(a) is transmitted through the Sagnac interferometer. The use of a circulator between the laser source and Sagnac loop allows comparing the transmitted and reflected signals, as shown in Fig. 4(b). It is seen that the loop provides for good contrast, measured in more detail in the following.

 

Fig. 4 (a) The optical input signal to the Sagnac loop, and (b) the reflected (blue) and transmitted (red) signal. The curves are normalized as the reflected signal experiences an extra loss when passing the circulator. The measured peak voltage developed over the fiber electrodes is 131 V.

Download Full Size | PDF

Transfer function and extinction ratio

The transfer function obtained with the gated interferometer is displayed in Fig. 5(a) , measured by varying input voltage U over the fiber-electrodes and reaching maximum at V_π. The modulation is calculated as P_transm(U)/P_incident, where the power transmitted P_transm(U) and incident P_incident are corrected for splice losses, etc. The modulation at V_π switching is measured to 99.2%. Note that the input voltage is higher than the voltage developed across the electrodes, as shown in Fig. 1.

 

Fig. 5 (a) Transfer function of Sagnac loop as a function of the input voltage. The inset shows how the transmission is measured from the time-resolved reflected signal, illustrated for an input voltage 65 V. (b) The extinction ratio as function of the input voltage applied to the two phase modulators. The inset shows how the extinction ratio is measured in the transmitted signal, for the point designated in the graph with a modest extinction ratio of 16 dB.

Download Full Size | PDF

The corresponding extinction ratio is displayed in Fig. 5(b). It is calculated as 10*log(P_leak (U)/P_transm(U)), where P_leak(U) is the peak intensity of the pulses leaking through the interferometer in the absence of a voltage pulse and P_trans(U) is that of the transmitted pulses. The extinction ratio varies depending on how well the PC is set in the loop. However, it is typically between −25 and −30 dB around the V_π switching voltage.

For convenience, the component here is connectorized rather than spliced, and small sub-reflections are created. These are visible in the inset of Fig. 5(b) next to the leaked pulses.

4. Conclusions and discussion

It is shown that an all-fiber setup with two phase modulators in a Sagnac loop can be utilized to select single pulses at a repetition rate up to 1 MHz from a laser running at 7 MHz. The setup has a modulation depth of 99.2% and up to −30 dB extinction ratio. While this value is insufficient if the pulses are used for direct sample excitation, it could be acceptable if the laser pulses are mixed in a nonlinear process (e.g., second harmonic generation) after pulse selection. It should be noted that the configuration here is non-resonant and works well at arbitrary repetition frequencies below 1 MHz as well.

The repetition rate of the setup is currently limited by the electronics generating the voltage pulses. Behind the problem with the drive electronics lies the difficulty in recording a larger second order nonlinearity in the silica fibers (here, χ⁽²⁾ ~0.26 pm/V), and work is ongoing to increase the nonlinear coefficient achieved. With present values, in order to keep the drive voltage low one needs to make use of relatively long components. Ideally one would want to be able to switch laser pulses at up to 100 MHz with low loss. At 1.5 µm wavelength the insertion loss is already significant (~4 dB due to the extended mode-field outside the core interacting with the electrodes), while the 3 dB cut-off frequency for the 78 cm fiber modulators is only 16 MHz. Operation at shorter wavelengths, for example 1 µm, brings about lower loss due to better mode confinement and lower switching voltage. Nevertheless, major improvements in the technique may require the additional use of traveling-wave propagation, which should greatly improve the time-response of the electrooptical fibers.