Getting the measure of ultrashort pulses

23 Aug 2004

An autocorrelator is often needed for measuring the duration of very short light pulses. Peter Staudt discusses the different types on the market and explains how they work.

From Opto & Laser Europe September 2004

Generating ultrashort laser pulses that last a few femtoseconds is a highly active area of research that is finding applications ranging from micromachining to eye surgery.

For optimum results it is often important to know a laser's wavelength, average power and pulse duration. While it is relatively easy to check the first two parameters with a spectrometer and a power meter respectively, measuring the latter is more complicated.

The problem is that conventional electronic photodetectors that convert a light signal into an electrical current respond too slowly to measure very short light pulses. In fact, the most common device used to determine the duration of sub-picosecond pulses is an optical instrument called an autocorrelator - usually an interferometer in combination with a nonlinear optical material.

Other sophisticated pulse analysis equipment based on so-called FROG and SPIDER techniques can also measure the phase profile, amplitude and spectrum of a pulse. However, these merit their own article and will not be discussed in detail here.

First principles So how does a scanning autocorrelator work? The basic principle is a Michelson interferometer consisting of two arms (typically about 50 mm long), one of which has its optical path length continually changed (scanned) to create a variable optical delay.

An incoming laser pulse arriving at the autocorrelator is split into two replica pulses by a beam splitter. The two pulses then travel along the different arms of the interferometer before they are recombined by the beam splitter and focused to a common point in a nonlinear crystal. The crystal generates an upconverted light signal (often but not always a second harmonic) that is detected and displayed. The trace of the intensity of this mixing signal versus the amount of the optical delay is known as an autocorrelation function (ACF). By making assumptions about the shape of the pulse, its duration can be determined by multiplying the width of the ACF by a form factor.

Different types of autocorrelator introduce the optical time delay in different ways. The three most common techniques use a dispersive material such as special glass, a set of rotating mirrors or a stepper motor/scanner.

The dispersive technique relies on slowing the propagation speed of the pulse by passing it through pieces of rotating glass. The length of the delay relates to the orientation, and thus thickness, of the glass.

Another possibility is to have the two pulses travel different distances. This can be accomplished by using rotating mirrors that control the path length of reflected pulses.

A final option is a variable-length arm. One of the arms is linearly extended by either a fast scanning system with limited scan range, or a stepper motor with slow movement but a wider scan range.

The choice of autocorrelator depends on your needs and budget. The versions with a glass block or rotating mirror operate only in a fixed scan range and thus have limited resolution (>5 fs). The dispersive approach can also induce a large loss, while the device that has rotating mirrors can be more complicated to handle. The third option, the linear scanner, achieves the highest resolution thanks to its variable scan range.

Making a choice The most appropriate autocorrelator can be determined by the following criteria:

Pulse duration range. Pulses shorter than 100 fs are strongly affected by dispersion, so in this case the autocorrelator that uses a dispersive material is not suitable. When analysing such short pulses it is important that each optical element inside the autocorrelator (such as the crystal delay line and the beam splitter) has as little dispersion as possible. This can be achieved by using reflective instead of transmissive optics.

On the other hand, pulses longer than 5 ps place different requirements on the measuring system. In this case the dispersion effects are negligible, but scan ranges (time delays) need to be larger. A useful rule of thumb is that the maximum scan range equals four times the maximum pulse length.

All types of scanning autocorrelator have an upper limit for their scan range, owing to the size constraints of the interferometer cavity. With rotating mirrors a delay of up to 300 ps can be achieved, while a spring-loaded scanner enables the measurement of pulses up to around 150 ps long. Stepper motor translators can measure up to several nanoseconds, but the size of such a device is considerable as a scan of 1 ns needs a physical scan range of about 15 cm.

Power and repetition rate. These two parameters are interrelated and define the sensitivity requirements of the autocorrelator.

For example, consider a train of identical 100 fs pulses with an average power of 100 mW. At a repetition rate of 1 kHz (1000 pulses per second) each pulse reaches a peak power of 100 GW. However, at a higher repetition rate of 100 MHz the peak power is reduced to 1 MW.

To take this into account the sensitivity of the autocorrelator is defined by the product of the peak power multiplied by the necessary average power. Highly sensitive autocorrelators reach a P_PEAKxP_AVERAGE of 10^-7 W². This degree of sensitivity is required, for example, by a fibre laser with an average power of 0.1 mW at a repetition rate of 10 GHz and a pulse duration of 10 ps.

To provide such sensitivity, autocorrelators use highly reflective optics, efficient crystals, low-noise photomultiplier detectors and special noise-reducing data acquisition.

For high-energy pulses at a low repetition rate, such as amplifier pulses, the priorities are entirely different. In this case, photomultipliers are not suitable and photodiodes are used as detectors instead. Depending on the scan frequency of the autocorrelator system it may be necessary to measure laser systems with a slow repetition rate in triggered mode so that only the pulse is measured and not the pause between two pulses.

Wavelength. Another important consideration is the wavelength of the pulses. To work well, the autocorrelator has to feature low-loss, low-dispersion optics suitable for the respective wavelength.

It is vital that the nonlinear crystal can be phase matched at the wavelength in question so that it can generate an upconverted light signal. The detector must also be sensitive to the wavelength of the upconverted signal but insensitive to the fundamental wavelength, which should be filtered out.

The most flexible autocorrelators feature a wide bandwidth and allow switching between wavelength ranges by exchangeable sets of crystals and detectors. Standard wavelength options at APE make it possible to span 420 nm in the visible to 1600 nm in the infrared.

Signal generation. All autocorrelators require a nonlinear crystal, typically BBO, to convert the incoming pulses into an autocorrelator signal. Often the crystal performs second harmonic generation (SHG), which is also known as upconversion or frequency doubling. In this case it is necessary to phase match the crystal to the wavelength range by tilting it, and this can be technically challenging and awkward.

An alternative design that is easier to operate but that sacrifices sensitivity is an autocorrelator that relies on two-photon absorption in a photodiode. This method is more cost efficient than SHG designs and does not require wavelength tuning, but it does have several drawbacks. First, the wavelength range is limited. Second, a background-free autocorrelation function is not available because the laser pulses cannot be filtered.

Budget and flexibility. The main difference between low-cost autocorrelators and more expensive versions is functionality and flexibility. If you only need to measure pulses in the 50 fs to 5 ps range from a Ti:Sapphire laser then a low-cost autocorrelator is probably the best choice and can cost less than 75000. These easy-to-use devices generate an autocorrelation trace that is viewed by connecting the instrument to an oscilloscope. The user measures the width of the trace to determine the duration of the pulse.

The next generation up are stand-alone instruments that use an electronic controller to process and display the results on a small on-board screen. The advantage of these devices is that they can often be configured for different wavelength ranges and can measure a wider range of pulse durations.

Top-of-the-range autocorrelators from 715 000 are the most flexible and enable the user to switch easily between different wavelength ranges and pulse lengths - for example, from <50 fs to >40 ps. These devices often have an integrated position-measuring system and come with changeable optics sets to allow measurements spanning from the ultraviolet to the infrared. Instruments are also available with an integrated spectrometer; these provide simultaneous spectral characterization of the pulse.

For users who wish to know more about their pulses, such as the phase profile, wavelength spectrum or temporal amplitude, other specialized analysis equipment is available. These instruments often rely on pulse analysis techniques known as FROG and SPIDER. Although these devices can provide much more information they tend to have a very narrow wavelength range.

• Peter Staudt is product manager for autocorrelators at APE GmbH, a Berlin-based firm that specializes in short-pulse measurement equipment. www.ape-berlin.de