Up-conversion Time Microscope

We have demonstrated a temporal imaging system with a novel time lens which magnified (slowed down) 100 Gb/s optical data by a factor of twelve, to a rate of 8.55 Gb/s. The function of a time lens is to impart a quadratic phase modulation or linear frequency sweep to the waveform under study. Our approach to achieving time lens action is to up-convert the waveform under study using a linearly swept pump, thus imparting a linear frequency sweep to the waveform. This technique allows for much greater frequency sweep rates and hence shorter focal times than can be obtained with electro-optic modulators. Additionally, the increased bandwidth that can be obtained optically instead of electro-optically should result in higher resolution in a temporal imaging system.

The direct measurement of ultrashort light pulses and waveforms is limited by current technology to several picoseconds resolution. New techniques for improving resolution generally concentrate on faster photodetectors and sampling electronics which are, of course, very important and form the foundation for all high speed optical measurements. There is, however, the possibility of modifying the optical waveform before it is detected by effectively slowing it down to a speed that is within the capabilities of state-of-the-art instrumentation. This is the approach that we have been developing and it is called ``temporal imaging'' in direct analogy with spatial imaging.

The objective in both spatial and temporal imaging is to expand or compress a field or waveform and preserve its overall profile, be it a function of space or of time. The technology of the former has been around for several centuries and is familiar to most everyone. The technique of temporal imaging is based on a one-to-one analogy between the processes that occur in the spatial case and those that occur in the temporal case. For example, the phenomenon of diffraction has a dual in the equation that describes the dispersion of light pulses in dielectric media. In addition, there is a one-to-one correspondence between the action of a conventional lens and a quadratic phase modulation applied to a time waveform. When these simple elements are combined in proper measure, a system is formed which can magnify or demagnify optical time waveforms. This is the essence of temporal imaging.

The up-conversion time microscope.

Figure 1. Schematic diagram of the up-conversion time microscope system. The input waveform I(t) was a 4 bit 100 Gb/s word. Input and output dispersion was obtained by two sets of diffraction grating pairs. The up-conversion time lens was achieved by sum frequency generation of the input waveform and a linearly chirped pump E(t). The measured temporal image of the 100 Gb/s 1101 input test pattern was a 1011 waveform at 8.5 Gb/s.

With this brief description of the operation of an up-conversion temporal imaging system in mind, we now proceed to describe the origins of the signals used in this system.

Generation of the input signals.

In order to test the system, a high speed test waveform was needed. Since digital optical pattern generation is not easily accomplished at 100 Gb/s, and certainly not commercially available, we resorted to the simple expedient of using the reflections from air-glass interfaces to appropriately space picosecond pulses in the form of a binary test pattern. The test pattern generator is shown in Figure 2. The source is a modelocked Nd:YAG laser which produces 71 ps pulses. These pulses are coupled into 70 m of single mode polarization maintaining fiber and emerge chirped and spectrally broadened due to self-phase modulation with a bandwidth of 620 GHz. The chirping of the pulses in the fiber serves two purposes. First, we needed to create reasonably short pulses from the initial pulsewidth of 71 ps. Calculations indicated that 70 m of fiber would produce enough bandwidth to compress the pulses to the 1-2 ps range. Second, some fraction of the uncompressed linearly chirped pulses could serve the dual purpose of the pump for the up-conversion time lens. Thus, upon exiting the fiber, the chirped pulses pass through a 1 mm thick uncoated fused silica etalon and then through an uncoated fused silica wedge. The reflections off the front and back surfaces of the etalon combine collinearly with the reflection off the front surface of the wedge at a very shallow angle with respect to the input beam. This reflected beam is now composed of three overlapping chirped pulses. The first two are separated by a measured round-trip transit time of 9.2 ps in the etalon. The third pulse comes from the reflection off the front surface of the wedge which is positioned behind the etalon so that the pulse arrives 30 psafter the first pulse. The beam is finally passed through a pair of diffraction gratings adjusted to optimally compress the chirped pulses to an autocorrelation width of 2.1 ps. The pulse sequence forms the digital word 1101.

Figure 2. Schematic diagram showing the arrangement for generation of the 100 Gb/s 1101 test pattern (Signal) and the chirped pump for the up-conversion time lens (Pump).

The configuration of the compression gratings in Figure 2 that produced the optimumly compressed pulses was used to calculate the chirp rate of the pump signal. Then from this known strength of the time lens, the input and output dispersions in Figure 1 were determined which would provide a desired magnification of M=-10 and satisfy the temporal imaging condition.

Measured results.

The temporal image was measured with a high-speed photodiode (20 ps impulse response) and 34 GHz sampling oscilloscope. Because of low conversion efficiency to the second harmonic in the time lens and 20% throughput efficiency of the output dispersive network, electronic amplification of the detected signal was necessary. Two amplifiers with a total gain of 35 dB from 2-20 GHz were used. Figure 3a shows the measured image. It is a 1011 pattern with 351 ps between the first and last pulse. The waveform is a time-reversed and expanded replica of the input demonstrating a magnification M=-11.7. The ringing following the bit pattern and the increased width of the pulses is due to the limited bandwidth of the amplifiers.

To verify the magnification as well as the overall validity of the time microscope, we modified slightly the input word and observed the change in the image. The etalon was moved 0.60 mm further away from the wedge, moving the first two pulses of the test pattern 4 ps earlier in time. This resulted in an additional 48 ps delay of the last two pulses of the temporal image (the change from Figure 3a to Figure 3b), indicating a magnification of M=-12. The arrival time of first pulse (originating from the wedge) and the separation between the pulses from the etalon surfaces were unchanged, as expected.

Figure 3. Measured temporal images. (a) The original temporal image is a 1011 pattern with 351 ps from first to last pulse, a magnification of M=-11.7. (b) When the two pulses from the etalon were advanced 4 ps the new temporal image showed a corresponding 48 ps additional delay.

To our knowledge this is the first demonstration of an up-conversion temporal imaging system. The resolution of this system was around 5 ps referred to the input, limited by the bandwidth of the amplifiers. With higher power pumps signals or better throughput efficiency the amplifiers would not be necessary. There are also other techniques besides self-phase modulation for obtaining highly chirp, wide bandwidth pump signals which can be used to make stronger time lenses and systems with better resolution. Temporal imaging systems with 100 fs resolution may be obtainable with this up-conversion time lens approach.

bennett@leorg.ucdavis.edu