Primary Photophysical Processes in Chromoproteins - Fluorescent Proteins: From Understanding to Design - page 42

imaging. These aspects are covered elsewhere in this volume, particularly in the
chapter by Jung. However, ultrafast fluorescence techniques (i.e., for decay pro-
cesses faster than 100 ps) have proven especially useful in understanding the
unusually rapid radiationless decay of the GFP chromophore (Sect. 3 ) and the
kinetics of the unique ESPT reaction in avGFP (Sect. 4 ). Thus, it seems pertinent
to outline the principles of the up-conversion method.
The starting point is a Kerr lens mode-locked titanium sapphire laser. These
compact commercially available devices provide a train of ultrafast pulses (typi-
cally at a repetition rate of 70-100 MHz) with excellent stability [ 33 ]. The appara-
tus in our laboratory uses a Coherent Micra 10 source, which provides (following
pulse compression, see below) pulses as short as 18 fs. The central wavelength is in
the region 760-860 nm, the output power about 1 W and the bandwidth is typically
100 nm. Such a broad bandwidth is essential to support very short pulses, a
consequence of the time-energy uncertainty relation.
The measurement of such short pulses is non-trivial, but well-established meth-
ods are available [ 34 ]. The most common is background-free autocorrelation. In
this method, the output of the laser is split into two identical beams at a beam
splitter. One beam is routed through a fixed delay, while the second is routed
through a delay of variable path length. Typically a mirror is mounted on a delay
stage which has a positional accuracy of better than 1 m allowing control of the

delay time between the two pulses with an accuracy of better than 6.7 fs. The two
beams are made parallel and focussed to overlap spatially in a thin non-linear
crystal, which generates second harmonic radiation (i.e., for an 800 nm centred
input beam the output is at 400 nm, generated from two 800 nm photons). The
400 nm output from the crystal appears in three beams, the second harmonic of the
two individual input beams plus a signal generated from both beam (each con-
tributing one photon). These three signals emerge spatially separated due to the
different phase matching conditions [ 34 , 35 ]. Clearly, the two beam signal can only
be generated when both pulses are overlapped in the crystal in time as well as in
space. Thus, a measurement of the second harmonic intensity while scanning one
pulse in time with the other is fixed yields a convolution of the temporal profile of
one pulse with the other, the pulse autocorrelation. The autocorrelation serves as the
instrument response function for deconvolution of the ultrafast fluorescence up-
conversion traces (see below). However, the actual width of the laser pulse can be
recovered from the autocorrelation if a pulse shape is assumed. Typically, a sech 2
profile is assumed for mode-locked titanium sapphire lasers, in which case the
pulsewidth is 0.65 times the autocorrelation width.
Although the intensity autocorrelation is a simple and effective means of char-
acterising laser pulses, additional information may be used to better characterise the
pulse. At the simplest level the laser spectrum can also be measured and the time
bandwidth product (pulsewidth times spectral width in frequency units,

m

)
determined; for a sech 2 pulse this should take the value 0.315. Greatly superior
methods of ultrafast pulse analysis are now available, which simultaneously char-
acterise time structure, phase, spatial profile and amplitude of the laser pulse. These
methods are described in more detail elsewhere [ 34 ].

Dn . Dt