Advanced techniques measure ultrashort pulses

Driven by potential applications that range from optical communications to materials characterization, the development of ultrashort-pulse sources remains one of the most active areas of laser development.¹ In addition to laboratory devices with ever-decreasing pulse widths, modelocked-semiconductor and diode-pumped solid-state lasers are expected to provide sources that will be used in a number of real-world applications. While receiving significantly less publicity than ultrafast-laser performance records, high-speed detectors and ultrafast pulse-characterization techniques are essential to continued research progress and the ultimate commercial deployment of systems that are based on modelocked lasers.

Techniques and hardware for short-pulse detection and characterization can be divided into two classes based on their response time. For pulsewidths longer than 15 ps, solid-state detectors offer a combination of small size and ruggedness that make them potential candidates for optical communications and other OEM applications.

On shorter time scales, characterization techniques are currently confined to the laboratory environment. Included in this class are streak cameras that can resolve pulses longer than 1 ps and autocorrelation techniques that can be used to analyze pulses as short as a few femtoseconds. In addition to simple intensity autocorrelation, advanced techniques that completely describe the time evolution of the electric field in a short pulse have been developed. With names that are typically used to describe the maturation process of jumping amphibians, FROG, POLLIWOG, and TADPOLE provide the detailed information needed for continued development of ultrashort pulsed lasers.

This article describes techniques for the characterization of modelocked pulses. While there is some overlap between applications for streak cameras, solid-state detectors, and autocorrelation techniques (both traditional and amphibious), there is usually a ‘best choice’ for a particular application.

Detectors combine speed and convenience

Although solid-state detectors are slower than streak cameras and autocorrelators, their small size and ruggedness make them useful for many applications. At the present time, commercial products with bandwidths of 60 GHz and risetimes of less than 10 ps are available from several commercial vendors.

While design details vary among manufacturers, fast response times are achieved by optimizing the detector design with respect to a number of important variables. Principal among these are the decay time for the electrons and holes generated by the input light pulse and the capacitance of the detector structure and associated circuitry. In general, it is desirable to minimize both the carrier lifetime and the capacitance.

In one design, a photoconductive detector is produced by fabricating an interdigitated electrode pattern on low-temperature-grown gallium arsenide (GaAs; see Fig. 1).² Light shining through the electrodes generates carriers in the region between them, thereby in creasing the conductivity. An electric field applied to the interdigitated electrode structure (adjacent electrodes have opposite polarity) is connected to coplanar strip transmission lines of opposite polarity coated on the same substrate. By using a material in which the carriers have a short lifetime and high mobility in combination with a electrode design that minimizes transit-time effects and capacitance, a bandwidth of 375 GHz was achieved in a laboratory prototype. Devices of similar design are currently manufactured by Picometrix (Ann Arbor, MI) and marketed by Newport Corp. (Fountain Valley, CA).

FIGURE 1. In a high-speed photoconductive detector, light passing between the interdigitated electrodes produces carriers in the low-temperature GaAs active region. These carriers produce a current in the strip transmission lines proportional to the incident intensity. The short carrier lifetime, combined with the low capacitance of the electrode structure and transmission lines, has resulted in bandwidths as high as 365 GHz from this structure.

Based on photoelectron emission in an image-converter tube, streak cameras produce an output image in which time is represented as a linear displacement. This is accomplished by rapidly scanning the electron beam inside an image-converter tube. The input optical system of a typical streak camera images light entering the camera into a narrow slit on the photocathode of the streak tube (see Fig. 2). While tightly focusing the input beam along one axis, the input optics have no effect on the intensity distribution in the orthogonal direction. As a result, the camera can accurately reproduce the spatial distribution of light in the exit plane of the spectrograph or along one axis of a spatially inhomogeneous source.

FIGURE 2. Input light striking the photocathode of a streak camera tube produces electrons that are accelerated by the grid electrode. A high-voltage pulse applied to the deflecting electrodes sweeps the electron image of the input beam across the aperture of a microchannel plate on the output end of the tube. A visible output image is produced by a phosphor screen and digitally recorded using a photodiode array.

Light hitting the photocathode produces a narrow beam of electrons that are accelerated by a grid electrode toward the output end of the tube. Deflection electrodes located after this grid are used to rapidly scan the electron beam across the input face of a microchannel plate near the output end of the tube. This element amplifies the electron current while preserving its spatial distribution. A visible output image is generated by a phosphor screen at the output end of the microchannel plate. In modern instruments, the two-dimensional output image is digitally recorded using a detector array for subsequent analysis.

By applying a short, high-voltage pulse to the deflecting electrodes, the slit of electrons produced by the photocathode can be scanned across the output screen in a few hundred picoseconds. Because the linear deflection of the tube is proportional to time, the image on the output face of the tube is a record of the temporal evolution of pulse intensity. Perpendicular to the scan direction, the output image faithfully records the variation of incident intensity on the photocathode.

In a typical commercial device, a time resolution of 2 ps can be achieved when operating in a single-scan mode. Ultimately, the time resolution is limited by space-charge effects within the tube that make the electron-transit time dependent on input intensity. In those cases in which the sample emission is periodic, repetitive scanning is used to increase the camera sensitivity.

In a typical application, a streak camera can be used to record the time evolution of the spectrum from a pulsed sample. By using a streak camera in place of a photographic plate or optical multichannel analyzer in the output plane of a spectrograph, it is possible to observe variations in the source spectrum during the camera scan period.

Measuring amplitude

Carrier lifetime and transit-time effects restrict detection techniques that require the conversion of light to electron current to a minimum resolution between 0.5 and 1.0 ps. For shorter pulses, detailed measurements must be accomplished using fast nonlinear optical techniques before the photon signal is converted into current.

In an intensity autocorrelator, the time-dependent amplitude of a short pulse is obtained by splitting the beam into two components and introducing a variable time delay t between them. Summing the two components in a fast nonlinear crystal produces an output that is proportional to the autocorrelation. The autocorrelation integral for an input pulse with intensity distribution I(t) is given by

A(t) = -I(t)I(t + t)dt (1)

Assuming that the temporal dependence of the input is Gaussian, this expression becomes

A(t) µ I 2 to exp(-t2/2to2) (2)

where to is the width of the input pulse.

In the autocorrelator of Fig. 3, a polarization beamsplitter is used to divide the input pulse into two equal components. One component is imaged into the nonlinear crystal along a fixed path, while the other component travels through a distance that can be adjusted by means of a movable mirror. If the difference in length between the two beam paths is equal to Dx, the time delay t is equal to Dx/c. Thus, by measuring the second-harmonic output power as function of mirror position (or path distance), one can theoretically plot the Gaussian curve described by eq. 2 and determine the temporal width of the input pulse.

FIGURE 3. Intensity autocorrelation can be performed using a simple setup in which an input pulse is split into two components by a dichroic beamsplitter. The reflected component strikes a fixed mirror and passes through the beamsplitter to a nonlinear second-harmonic crystal while the transmitted component is reflected by a second mirror mounted on a translation stage. A time delay between the two components is produced adjusting the position of the second mirror. The second- harmonic power measured by the detector is proportional to the autocorrelation integral.

Autocorrelators similar to the one in Fig. 3 are commonly found in modelocked-laser laboratories. For most wavelength regions of interest, commercial de vices are available from several manufacturers.

Amphibious techniques

While intensity autocorrelation can provide information on the pulsewidth of ‘well-behaved’ pulses, it provides no information on fine structure within the pulse envelope or temporal frequency variations. It has been shown, however, that a complete pulse description can be obtained by recording the spectrum of the autocorrelator output.

Known as frequency-resolved optical gating, or FROG, this technique was originally developed by Rick Trebino and coworkers at Sandia National Laboratories (Livermore, CA).^3,4 Using the output of a FROG measurement, referred to as a spectrogram, it is possible to mathematically construct a complete description of the pulse evolution (at least in those cases where the spatial mode is well-behaved).

Modifications to the FROG technique based on spectral interferometry are useful in those cases in which the input signal is too weak to generate an autocorrelation signal. In these cases, a spectrogram is generated by spectrally analyzing the time correlation between the unknown input and a second, more-powerful pulse with a known spectrum and pulse shape. This technique, which is technically a combination of spectral interferometry and FROG, is known as TADPOLE, or temporal analysis by dispersing a pair of light E-fields. POLLIWOG, or polarization-labeled interference vs. wavelength of only a glint, is the application of TADPOLE to both polarization components of a pulse, thereby providing information on the evolution of the polarization vector with time.

Real-world applications

While ultrafast optics continues to be one of the more interesting and active areas of laser science, the complexity and expense of currently available sources represent significant impediments to their near-term use outside of the laboratory environment. This situation will change, however, as modelocked sources are developed for the OEM marketplace. Given that significant efforts are currently underway to develop ultrafast-semiconductor, diode-pumped solid-state, and fiber lasers for OEM applications, there is a significant probability that robust products will be introduced within the next few years.

Assuming that ultrafast sources live up to their considerable potential, the characterization techniques described in this article will most probably be used for source development and troubleshooting. For example, fast detectors, used in combination with a sampling oscilloscope can be used for the initial characterization of a modelocked source. Intensity autocorrelators can then be used to accurately determine pulsewidth. Both classes of device are simple enough in design to be incorporated into field-portable test equipment for communications and other systems applications.

But the expense and complexity of streak cameras and amphibious autocorrelation techniques will probably restrict their use to the laboratory setting. Streak cameras are uniquely capable of quickly recording the time evolution of a one-dimensional intensity variation and will continue to be used for picosecond chemisty and other similar applications. Because of their sophistication, the use of advanced autocorrelation techniques such as FROG and TADPOLE will probably be restricted to those laboratories interested in obtaining shorter pulses from modelocked lasers and/or looking at very fast nonlinear optical processes.

Overall, the use of ultrafast detection techniques can be expected to grow as modelocked sources continue to be developed. While the time scale and magnitude of this growth are difficult to estimate, those working on the development of ultrafast sources are predicting that there will be significant future markets for their technology.⁵

ACKNOWLEDGMENTS

The author wishes to thank Eric van Stryland and Peter Delfyett of the Center for Research and Education in Optics and Lasers at the University of Central Florida for their informed technical input.

REFERENCES

1. W. Kaiser, Ed. Ultrashort Laser Pulses and Applications, Springer-Verlag, New York (1988).

2. Y. Chen et al., Appl. Phys. Lett. 59, 1984 (1991).

3. R. Trebino and D. J. Kane, JOSA A 11, 2429 (1993).

4. G. Taft et al., Opt. Lett., 20, 743 (1995).

5. Wayne Knox, Laser Focus World 32(6), 135 (June 1996).