2 PROCEEDINGS OF THE IEEE, VOL. 57, NO.

1, JANUARY 1969

Picosecond Laser Pulses

A. J. DEMARIA, MEMBER, IEEE, WILLIAM H. GLENN, JR., MEMBER, IEEE,
MICHAEL J. BRIENZA, AND MICHAELE.MACK

Invited Paper

AbstraeZ--Tbe broad bandwidth a d long storage lifelimes oFNd”: tion of what is believed to be the second generation of short-
glas.adrnbylasershvemadepossiMetLeg~tkmofpicosecoadLser timeduration,. high-peak-power laser pulses, i.e., laser
~h.riagpePLponersmexcessofowgigrrPrttlradrepetitimlrtesm
pulses .having picosecond time duration and peak power in
themierowrvermge.Tbelwefilrs.ppticPtioaareasoftbesepokesiudnde
resePrehiammlhearoptics,tmnimt~ofatotnicandndecdar excess of lo9 W [16].
systens, optiepnr g-td plpsmrs, spectroscopy, rpngia& epticrl ilb There are numerous reasons why researchers have be-
f o n n n t i o a ~ ~ h i g h - s p e e d p b o t ~ ~ y . ’ z h i s p a p e r r e v i e n s s e v e f acome
l interestedin the generationof picosecond light pulses
experimentfl teshdqws for generating, measwing, a d u t i b i q these with gigawatts of peak power. Picosecond light pulses can
be convenientlygenerated in laser media having wide
spectralbandwidths and long fluorescent lifetimes, such
I. INTRODUCTION as ruby or Nd:glass. The availability of optical pulses of

1 N 1961 HELLWARTH proposed an experimental tech- such high power and short time duration has aroused con-
nique for generating large output bursts of radiation siderable interest among military, academic, and industrial
from laser devices [l 1. This .experimental technique, researchers. For example, a pulse of 10- s at a wavelength
now called laser Q-switching,.wasfirst achievedby McClung of 1 micron has a lengthin free space of 0.03 cm and there-
and Hellwarth in 1962 with a Kerr cell [2], by Collins and fore. offers the possibility of measuring long distances to
Kisliuk with a rotating disk [3], and by DeMaria, Gagosz, fractions of millimeters. An event would have to move an
andBarnard with anultrasonic-refractionshutter [4]. appreciablefraction of the velocity of light inorder to
Rotating mirrors and prisms, Pockels cells, and saturable realize the full potentials of such pulses in high-speed
absorbers have also been utilized in Q-switching experi- photography applications. Electrical pulses having a rise
ments [5l-[7]. The availability of these Q-switched optical time of less than lO-’’s, amplitudes of 60-100 V, and repeti-
pulses has made possible the experimental investigation of tion periods as short as 1.5 ns have been obtained through
such phenomena as optically generated plasmas [8], opti- the use of fast photodiode detectors.These electrical pulses.
cal harmonicgeneration [9]- [ IO], stimulatedRaman, were previously unattainable and should find application.
Brillouin, and Rayleigh-wing scattering [ll],’ photon for determining the location andseverity of internal reflec-
echoes [ 121, self-induced optical transparency [13], optical tions in wide-bandwidthtransmission systems, studying
self-trapping [14], andopticalparametricamplification propagation delay, transient response of wide-bandwidth
[15]. Examples of applications of these short-duration, systems, etc.Theapplication of high-energy picosecond
high-peak-power Q-switched pulses include semiactive pulses to controlled thermonuclear plasma, optical radar,
guidance,ranging,illumination, high-speed photography opticalinformationprocessing,spectroscopy,nonlinear
and holography, and material working and removal. opticalproperties of materials,transientresponse of
The minimum pulse widths obtainable with existing Q- quantum systems, and ultrashort acoustic shock research
switching techniques are limited to approximately s appears very promising.
because of the required pulse buildup time. Peak powers of
11. GENERATION OF PICOSECOND LASER PULSES
approximately 5 x lo8 W without any additional stages of
amplification have been obtained with variousstraight- A . Basic Operating Principles
forward, Q-switching, experimentalarrangements. This It iswell known that a feedback loop, as illustrated by
paper will review the generation, measurement, and utiliza- Fig. l(a), encompassing an amplifier, a filter, a delay line,
and anonlinear element that provides less attenuation for a
ManuscriptreceivedJune 17, 1968;revisedOctober28,1968.The high-level signal than for a low-level signal, behaves as a
technical concepts andachievementsdiscussed herehave evolved over regenerativepulsegenerator [17]. When theloopgain
a period of two years, under the sponsorship of the U. S. Air Force Cam-
bridgeLaboratories,Office of AerospaceResearch;ProjectDefender exceeds unity, a pulse recirculates indefinitely around the
(under the joint sponsorship of the Advanced Research Project Agency, loop and each traversal gives rise to an output pulse at the
the Office of Naval Research, and the Department of Defense); and the output terminal. It is evident that such a pulse would soon
U. S. Army Missile Command. This inviredpaper is one of a seriesplanned
on topics of general interest.-The Editor. be degraded unless the effects of noise and distortion can
The authors are with the United Aircraft Research Laboratories, East be counteracted.Thenonlinear element (called an “ex-
Hartford. Conn. 06108 pandor” by Cutler [ 171) has theeffect of 1) emphasizing the
Stimulated thermal Rayleigh scattering hasalso been observed with
Q-switched ruby pulses. See D. H. Rank et a/., “Stimulated thermal peak region of the recirculating pulse while reducing the
Rayleigh scattering,”Phys. Rev. Letr., vol. 19, pp. 828-830, October 1967. lower amplitude regions, 2) discriminatingagainst noise
DEMARIA ET A L . : PICOSECOND LASER PULSES 3

that it have a line width equal to or greater than the laser

line width, and 3) that the dye recovery time be shorter than
the loop-time-delay of the laser.
A simplified explanation of the operation of the regenera-
tivepulse laser oscillator illustrated by Fig. l(b) can be
REFFOR given with the aid of Fig. 2.If an optical carrier frequency
vo along with two sidebands at vokf are superimposed, an
amplitude modulation of the light results at a frequency
f = A f with some peak-to-peak variation M , and a peak
intensity I, [see Figs. 2(b)and 2(c)]. Whenthis beam is passed
through a saturable absorber having the typical character-
(b)
Fig. 1. Equivalence o f ( a ) an electronic and (b) an optical
istics illustrated by Fig. 2(a), the initial sinusoidal amplitude
regenerative pulse generator. fluctuation of the input beam will be found to be distorted,
the peak-to-peak excursions of the fluctuation will be in-
creased, i.e., M 2> M , , and the time duration of the fluctua-
tion will be shorter as a result of the nonlinear transmission
characteristics of the saturable absorbers.With the sharpen-
ingof theamplitudevariation,additionalsidebandsare
added to the spectrum. In an optical cavityof length
L =c/2A f this process is repeated over and over again by
reflecting the light beam back and forth between two
INTENSITY
mirrors placed on both sides of the saturable absorber cell.
The fluctuation will continue to sharpen until a discrete
pulse is circulating in the cavity. The laser media provide
gain to compensate for the residual saturation loss of the
absorber and the mirrors. The circulation rate Af is given by
c/2L. The pulse will eventually acquire a steady-state width
Ar determined by the bandwidth of the laser media. The
repetitive output pulse train emitted from the laser mirrors
will havediscrete spectral components defined by the Fabry-
Perot resonances of the cavity extending +mAf on either
k- A V 4 side of v o for a bandwidth Av, as illustrated by Fig. 2(c).
Fig. 2. Operation of a saturable absorber in the time This result is to be expected since it iswell known, from
and frequency domain.
Fourier's theorem, that any repetitivepulse train can be
represented by a series of discrete sinusoidal functions hav-
ing integrally related frequencies and fixed phase relation-
and reflections, and 3) acting to shorten the pulse until the ships. The frequencies are all multiples of Ah and the
pulse width is limited by the frequency response of the cir- narrower the pulse width Ar, the larger the bandwidth re-
cuit. The output of the regenerative oscillator has a pulse quired to reproducethe repetitive pulse[16],[19],[20],
rate equal to the reciprocal of the loop delay, pulse widths Le., Ar 5 l/Av.
equal to the reciprocal of the overallsystem bandwidth, The system schematically shown in Fig. l(a) lends itself
and a center frequency determined by the filter frequency. to simplified analysis by repeated application of Fourier
Utilizing this technique, Cutler was able to generate micro- transforms, applying the frequency or amplitude character-
wave pulses having a carrier frequency of 4 GHz and pulse istics of each element, and finally equating the character-
widths of 2 ns. istics of the returning signal to the characteristics of the
An ordinary laser possesses all the basic elements of the assumed initial signal. Consider a signal entering the ex-
regenerativepulse generatoroperated inthe microwave pandor given by S,(t)e'$'. The signal leaving the expandor
regionwith the exception of the expandor element. The is S,(T)= K [S,(t)]qd$', where K is a constant indicating an
laser medium serves as the amplifier, the combination of the amplitude change of the signal and the superscript can be
Fabry-Perot resonances and the linewidthof the laser taken to indicate a nonlinear operation, not necessarily a
transition serve as the filter, and the time required for an powerlaw[17]. Notice that the nonlinear operation was
optical pulse to traverse twice the distance between the performed only on the envelope portion of the signal and not
reflectors serves as the loop-time-delay [see Fig. I(b)]. The on thephase portions. If this signal is now passed through a
optical analog of the electronic expandor circuit element is filter having a frequency function F(co), the output from the
a saturable absorber, such as the reversible bleachable dye filter has the form F3(co)=F(m)F2(m),where
solutionscommonly used as laserQ-switches[18]. The
fundamental requirements of the saturableabsorber are (1)
1) that it have an absorption line at the laser wavelength, 2 )
4 PROCEEDINGS OF THE IEEE, JANUARY 1969

is the transformof S2(t)to the frequency domain. The trans- For y=O, the pulse width is a minimum; AT,,,^,, N l/Av. A
formation of the signal F3(w)into thetime domain gives large phase curvature gives rise to a pulse whose instan-
taneous frequency sweeps through a bandwidth larger than
S,(t) = -
2n lrn
F3(w)P'dw.

The amplifier and circuit losses give a net gain G so that

(2)
the
The
reciprocal of the pulse width; or Az=Azmi,(q2+ 1)/q1l2.
phase measured with respect to the pulse time, in
general, is found to change from pulse to pulse in this analy-
sis. In the observation of a series of pulses, the analysis shows
theoutput of the amplifier is S,(t)=GS,(t). The signal that the phasewould appear to move continuously through
S,(t) leaving the delay element is delayed by T, and brings
the pulse at a rate9 / rad/s,
~ where 9 is the phase shift change
us back to the expandor input.We now require that S,(t) be
between pulses. Theanalysisshowsthatthe frequency
equal to S,(t), except possibly for instantaneous phases, so components of the repetitivepulse train aregiven by
that
v, = N I T - e/(27tT), (7)
where N is the number of carrier half-wavelengths in the
loop. Equation (7) shows that the frequency components
. J - [s,(t- T ) ~ q e - ~ ( t - T e-'
) '($1- ')dl are notexactly harmonically related. Thephase is not found
-m to be a continuous extrapolation from previous pulses as is
obtained by modulating a C W signal ; nor is it directly re-
where 9 is the phase shift of the optical wave relative to the
lated to thepulse time as would be obtained by shock excit-
pulse time. Given a filter characteristic F(w), an expandor
ing an oscillatory circuit; noris it random and uncorrelated
nonlinear law q, and time delay T, then (3) specifies the time
as is obtained from pulsed oscillators.
function S,(t)e"'.
The frequency sweep or "chirp" described by Cutler for
In the event a function S,(t) is found to be a replica of
the microwave regenerative pulse generator case has re-
S,(t), a solution to (3) has been found. Gaussian functions
cently been experimentally observed by E. B. Treacy [21],
are well suited for such types of solution for (3) if a power
[22] in Nd3+ :glass lasers. The discovery that picosecond
law for q is assumed. Cutler assumed Gaussian character-
pulses from dye mode-lockedglass :Nd3+ lasers are chirped
istics for the pulse envelope andfilter functions and apower
explainstheorder of magnitudediscrepancy observed
law for q [17]. Under these assumptions, the solution for
between the spectralwidth and thepulse widths'of Nd :glass
(3) gives the pulse rate Af to be equal to the reciprocal of
regenerative pulse generators [16]. Treacy used the disper-
the group time delay l/r =2n/fi=dwjd4 around the loop,
sive characteristics of two cascadedopticalgratings to
and the pulse width AT to be equal to
compress x4 s pulses from a dye mode-locked
4 glass:Nd3+ laser down to 4 x s [23]. These com-
pressed pulses consisted of just over 100 optical cycles. It
is notknownatthistimewhetherthe laser orthe dye
medium is primarily responsible for the chirp character-
istic of the output pulses of dye mode-locked glass:Nd3+
lasers.
A quantitative feeling of the magnitudeof the sharpening
experienced by an opticalpulse in passing through a satura-
where Am is the bandwidth at the 1 neper point of a Gaussian ble absorber can be obtained by the simple calculations
filter, andfi and y are thecoefficients of linear and quadratic given below. Assume a pulse whose time duration is long
terms in the power series expansion of the loop's phase compared to asaturableabsorberrelaxation time. The
+
shift 4 = a + fi(w - wo) y ( o -coo)' + . This power series absorber will act as an intensity dependent absorber. Such
expansion of the phasein the feedback loop canbe obtained a case has been treated by Hercher [24]. Theintensity
by taking a Taylor series expansion of the refractive index transport equation is given by
of the media within the feedback loop,

where the propagation time has been neglected, a. is the

and substituting into the phase relation 4 = 2nLn/c, where small-signal absorption coefficient, and I, is a saturation
L is the length of the media within the feedback loop. The intensity, i.e., the intensity at which the absorption coeffi-
existence of the squared frequency phase term causes the cient of the dye is reduced to one-half of its small-signal
instantaneous frequency to changethroughthe pulse as value. The solution to this equation is .
given by v = v,, +(bt/n), where b/n is the sweep rate. The
linear frequency sweep rate is given by 1' e'{=) = TI' $io)
(X) (0) (9)
b - .(? - 1) Y (6) where T the
is small-signal transmission of the finite length
+1 +
( 2 / ~ ~ y )2 * ~ dye cell, and Z'(0) and Z'(x) are
the
normalized
intensities
DEMARIA ET AL.: PICOSECOND LASER PULSES 5

1.o
at the entrance and exit as a function of distance in the dye
cell, respectively.
A computer experiment can be performed to illustrate 0.8
the operation of Fig. 2 using (9). Fig. 3 illustrates the pulse j;
z
Y
shape normalizedto unity intensity after every tenth pass of
a trial Gaussian-shaped pulse initially having a peak inten-
4 0.6

sity of 0.1 I , through a saturable absorber cell having asmall-

signal transmission of T=0.7. The absorber is most effec-
tive in sharpening when the intensity is in the neighborhood
of I,. In the actual operation of a laser, the absorber con-
centration is muchweakerand the transmission is cor-
respondingly higher, of the order of 0.95.
0
It is also of interest to obtain a quantitative feeling of the t
magnitude of the sharpening when an optical pulse is
Fig. 3. Opticalpulsesharpening as afunction of passesthrougha
propagated through a two-level system in which the dura- saturable absorber having a short relaxation time with respect to the
tion of the optical pulse is lessthan the relaxation time of the pulse duration.The assumedparameters inthe calculation were
absorber. A rate equation treatment of this case has been intensity of the initial Gaussian pulse 0.1 Zs,gain between successive
passes G = 1.4, and dye transmission T = 0.7.
given by several authors [25], [26]. Their solution is applica-
ble to the saturable gain in the laser medium as well as
the saturable absorption in the absorber. When the relaxa-
tion time is assumed infinite, an exact solution of the rate
equations is possible. For this case, it is more appropriate
to deal with the integrated pulse intensity

U ( t )=
s1m
Z(t)dt

rather than the intensity. The solution for the saturable

(10)'
20 30 40 50
PULSE NUMBER
60 00

two-level medium is (a)

1
WOut(t)= In [l + G(ewin(')- l)],
when the normalization W(t)=U(t)2o/hv is used, o is the
cross section per active atom, andG = exp [a&] is the small-
signal gain or attenuationof the medium.
Fig. 4 illustrates the computersimulation of a laser-
medium saturable-absorber combination utilizing (1 1) for
both the active medium and the absorber. Thecross section z
oL of the laser atoms was taken to be much smaller than Fig. 4. Computer simulation of an optical pulse regenerative oscillator.
for the absorber molecules oa, i.e., hv/2o,= 16 J/cm while The assumed parametersin the calculation weremirror reflectivity 0.8,
hv/20a=0.016 J/cm. The loss due to the reflectivity of the startingpulseenergy=lasersaturation energy,laser saturation
laser mirrors was assumed to be 0.2, the initial laser gain energy EL=16 J/cm, dye saturation energy=10-3 EL, and the dye
relaxationtimewasassumedmuch,muchgreater than the pulse
2.0, the absorber transmission0.8, and the starting Gaussian duration.
pulse energy 1.6 x J/cm. Fig. 4(a) illustrates the ampli-
tude increase of the pulse and the gain decrease of the laser model between the twentieth and fortieth passes. Referring
medium as a function of time. The absorber was assumed to Fig. 4(a), it is seen that this is just the region where the
to relax to its initial state between the successive passes of pulse energy reaches and exceeds the absorber saturation
the pulse but the laser gain decreases as a result of all pulses energy.
that have passed through it. The amplitudes of the pulse It can be concluded that much can be learned about the
train for the first 30 passes are small, in the range of 0.1 to operation of the laser saturable-absorber combination by
1.0 hv/20a. When the energyfinallygrows toapproxi- means of such computer simulation. Neither of the two
mately the assumed absorber saturation energy (0.016J/cm), models described here is capable of explaining the forma-
a rapid increase in pulse energy takes place. The growth in tion of the initial pulse. In addition, the absorber concentra-
amplitude is finally limited by the onset of saturation in tion is much weakerin the actual operationof such systems
the laser. The laser gain then begins to decrease. In this than the values assumed in these simulations. As a result,
region, the energy of the pulses reaches a maximum and the sharpening action is much weaker. It is apparent that
then begins to decrease. Fig. 4(b) illustrates the sharpening a more elaborate model of the dye is needed. In addition,
of the pulses as a function of the number of passes.The peak light amplification in saturableabsorbers must also be
intensity of each pass has been normalized to unit energy. taken into account, as will be discussed in later sections of
The most pronounced sharpening is found to occur in this this paper.
6 PROCEEDINGS OF THE IEEE, JANUARY 1969

Laser oscillators consist of a resonantsystem withdimen- passive modulators eliminates the need for such critical
sions that arelarge compared to the oscillating wavelengths. adjustments2 In addition, the saturable absorberserves as a
Consequently, mode density is high and there are Q-switch and therefore makes possible the generation of
m = 2LAl/iL: axial interferometer resonanceswithin the pulses having peak powers in the gigawatt range.
line width A i of a laser transition having a center wave-
length loand reflectors separated by an optical length L.
In the normal operation of a laser, these modes are to a B. Experimental Results
great extent uncoupled and therefore have no fixed phase The relatively narrow spectral line widths of gas lasers
relationship between themany discrete oscillating fre- limit pulse widths to the order of lo-’’ s [27], [31]. The
quencies of the laser. In the above discussion, the required broader line widths available in solid-state lasers such as
coupling was supplied by a passive modulator, i.e., the Nd3+:glass [18], [28], [36], ruby [29], [37], and
saturable absorber. However, the required mode-coupling YAG:Nd3+ [38] have producedconsiderablynarrower
necessary to lock the phase of the axial modes, i.e., mode- pulses having pulse widths in the picosecond range. The
locking, can also be supplied by an active modulator. This Nd3+:glass is of particular interest for the generation of
operation canbest be described as follows. Assume that the ultrashoit pulses because of its broad oscillating line of 100
Fabry-Perot mode vo, nearest the peak of the laser gain A to 200 A at i o = 1.06 p. With a mirror spacing of 1.5 m.
profile, will begin to oscillate first. If an amplitude or phase there exist approximately 6 x lo4 Fabry-Perot interferome-
modulator operating at a frequency Af is inserted into the ter resonances across the200 A line width [or 2 m = 6 x lo4
laser’s feedback interferometer,thecarrier frequency vo spectralcomponents in Fig. 2(c) are possible], anda
will develop sidebandsat &A$ If the modulating frequency pulse width in the10- l 3 s range could theoretically be
u i s chosen to be commensurate with the axial mode fre- obtained.
quency separation A f =c/2L, the coincidence of the upper Simultaneous @switching and mode-locking experi-
(vo +Af) and the lower ( y o -Aj) sidebands with the adjacent ments with saturable dyeshavebeen performed with
axialmode resonances will couple the vo -Aft vo, and Nd3 :glass laser rods having lengths from 12.2 cm up to
+

vo +Af modes with a well-defined amplitude and phase. 100 cm [18]. One major requirement in these experiments
+
As the vo Af and vo - Af oscillations pass through the was that the ends of the rod have Brewster’s angle ends to
modulator, they will also become modulated and their side- eliminate back reflection at theend surfaceof the rods. The
bands will couple the v o f2Af modes to the previous three dye concentration, optical pumpingintensities, and mirror
modes. This process continues until all axial modes falling reflectivities could be adjusted to obtain long pulse trains
within the oscillating line width are coupled. The construc- withlow pulseamplitudes,short pulse trains with cor-
tive and destructiveinterference of these simultaneous respondingly higher pulse amplitudes, or single or multiple
phase-locked modes can be described by the interference of Q-switched mode-locked pulse trains. The dye cell lengths
Fourier-series components in the construction of a repeti- used m our experiments were typically from 1 cm to 0.1
tive pulse train. cm. For the best results, the dye cell was placed at Brew-
Active time-varying-loss [27]-[29] and reactive [30] ster’s angle andthe dyewasplacedin contact with one
modulators havebeen utilized fo-r mode-locking laser of the mirrors. Practically all the dyes useful in the genera-
oscillators. Ultrasonic standing-wave diffraction cells [27], tion of ultrashort laser pulses decompose whenexposed
[28], [31&[34] and KDP Pockelscells [29] havebeen to ultraviolet light so that glass filters are useful for dye
utilized as active-loss modulators, whereasonlyPockels cellwindows in obtaining longdyelife.These comments
cells have been utilized as reactance modulators [30], [35]. hold equally well for both the ruby and the Nd3’ :glass
Hargrove, Fork, and Pollack were the first to couple the experiments.
modes of a laser for the generationof ultrashort light pulses Eastman 9740 or 9860 saturable dyes have been used in
[27]. Theyemployed a visible He-Ne laser and an ultra- our Nd3 :glass experiments, and cryptocyanine, dicarbo-
+

sonic diffraction modulator [32] when locking of the axial cyanine iodine, or dicyanine A in our ruby experiments.
modes had been achieved ; they report the virtual elimina- Eastman 9740 has an upper limit relaxation time of 25 to
tion of both random andsystematic amplitude fluctuations 35 ps [39], and 9860 an upper limit of 6 to 9 ps [40]. These
arisingfromthe drifting of the phases of theindividual dyes shouldtherefore bewell suited for obtaining pulse
modes and a fivefold increase in peak power. Theoretically, repetition rates up to 10” or 1OI2 Hz. Of the commonly
the peak power of the pulses should be n times the average used ruby Q-switchingdyes such as chloro-aluminum
power, where n is the number of coupled modes. phthalocyanine, vanadyl phthalocyanine, andcrypto-
Mode-locking alaser oscillator with an active modulator cyanine, only, cryptocyanine has arelaxation lifetime
requiresthe critical adjustment of mirror spacing and shorter than the recirculation time of a pulse of light be-
modulating frequency as well as compensationforany tweentwo laser Fabry-Perot mirrors separated by a rea-
perturbations in opticallength of the feedback interferome-
ter. Such compensation is of particular importance in the
mode-locking of large solid-state lasers as a consequence Theseadjustmentshaverecently been eliminated by regenerative
electronic feedback techniques. See G. R. Huggett, “Mode-locking of CW
of the variation in optical length of the rods during the lasers by regenerative RF feedback,” Appl. Phys. Left.,vol. 13, pp. 186-187,
optical pumping flash. The use of saturable absorbers as September 1968.
DEMARIA ET AL.: PICOSECOND LASER PULSES 7

sonable length [41]. Thesuitability of cryptocyanine in TABLE I

methanol for the generation of ultrashort laser pulses was PICOSECOND RUBYLASERPULSESWITH Vmous DYESAND SOLVENTS
first reported by Mocker and Collins [37]. Mode-locking
of ruby lasers withcryptocyanine in nitrobenzeneand
ethanol has also been reported [42]-[43]. Asubstantial
improvement in consistency of operation can be obtained
Cryptocyanine (1,l 'diethyl+'-
by using acetone as a solvent instead of methanol. The peak dicarbccyanine
acetone
iodide)
of the absorption line for this dye-solvent combination
coincides with the ruby line, while in the methanol solution DDI (1,7'diethyl-2,2'- methanol
dicarbocyanine iodine)
the absorption peak shifts by 120 A to 7060 A. In addition, water
it is possible that by virtue of the lower solvent viscosity
of acetone, the recovery time of the dye in the acetone solu- dicyanine A
tion may be shortened [MI.
Twoother dyesclosely related tocryptocyanine also
mode-lock the ruby laser effectively [45]. These dyes are
dicyanine A and l,l'-diethyl-2,2'dicarbocyanine iodide
(DDI). Nosolvent was found in which-the absorption peak
SWEEP,SPEED
of either of these dyes coincided with the ruby line. The
best solvent for dicyanineA was dimethyl sulfoxide
(DMSO). The absorption peakfor this dye-solvent com-
bination occurred at 6770 A.
DDI can be used ineither water or methanol. Its absorp-
tion peak is at 7030 A in water and at 7060 8, in methanol.
Thewater solutions are unstable with a half-life of ap-
proximately 1 hour. Representativedata of the pulse widths
and peak powers obtained with the three dyes found to
mode-lock the ruby laser effectively in our work appear in
Table I [45]. The "nominal" power outputs arethose which
were found to optimize the performance of the laser. The
output power levelcan be varied to some extent by changing
the dye concentration. Several of the results listed in the
table are worthy of special note. The first is that with DDI
in either water or methanol, pulses nearly as short as those
presently obtained with the mode-locked neodymium-glass
laser can be obtained. Another is that one can choose the
pulse width desired by choosing the proper dye. In all
cases, a small pulse-widthbandwidthproduct was ob-
served, evincing the fact that a frequency chirp in the output Fig. 5. Oscillograms of the typical output of an optical
pulses does not occuras per the Nd: glass laser. regenerative pulse oscillator.
One problem was noted with the ruby regenerative pulse
oscillator. It was found that the destruction of the dielectric
mirror occurred more often than for the Nd3+ :glass sys- rod of excellent optical quality. The placement of the ruby
tem. A Galilean telescope was included in the ruby laser rod within the cavity was not found to be critical.
cavity to double the beam diameter, thereby reducing the Ultrashort light pulses with a Beckman& Whitley model
optical flux density and preventing the destruction of the 4-40 thin-film Q-switch were also unknowingly observed by
99+ percent reflector. In addition, the Galilean telescope Hercher in his attempts to obtain single-mode operation
permits compensation forthe thermal lensing effects within from a ruby laser [46]. It was not realized at that time that
the ruby rod arising from the optical pumping by the ad- saturableabsorbers with long relaxation timestend to
justment of the separation of the telescope elements. narrow the spectral width while saturable absorbers with
The Galilean telescope consisted of a 50 mm focal length fast relaxation times tend to broaden the spectral width of
plano-concave lens and a100 mm focallength plano-convex lasers.
lens. Both lenseswere anti-reflection-coated. The spacing Fig. 5(a) isan oscillogram of the early portions of a slow-
of the lenses was adjusted for a minimum beam divergence buildup, Q-switched Nd: glass laser pulse train. The sweep
as evidenced by far field patterns. From the final spacing speed is 5 ns/div with the oscillogram covering 3 x IO-* s.
of the lenses, the focal length of the thermal lensing was The oscillogram illustrates the tendency of the saturable
estimated to be approximately 4 m. The beam divergence dye toemphasize the highest amplitude fluctuation oc-
was well below 1 mrad for a 1.20m diam, 5 cm long .Verneuil curring at the initiation of laser oscillation and to shorten
rod of fair quality, or a 1 cm diam, 10 cm long Czochralski this fluctuation pulse width as it successively propagates
8 PROCEEDINGS OF THE IEEE, JANUARY 1969

through the saturable absorption cell. The initial periodic

amplitude fluctuations are caused by the beating of two
axial modes separated by a frequency2Af: A beat frequency
at 2Af would correspond to the existence of two pulses in
the feedback cavity at the:ame time. At the end of 3 x IO-*
s, the secondary peak of the amplitude fluctuation is practi-
cally completely eliminated. The pulse width is reduced to
0.5 x s, and the pulse repetition period equals 2.5 ns or
A f . An oscillogram of an entire pulse train at asweep speed
of 200 ns/div is given in Fig. 5(b). An oscillogram of a pulse
train at a sweep speed of 10 ns/div is given in Fig. 5(c). The
recorded pulse half-widths are approximately 0.5 ns,
limited by the rise time of the photodetector and traveling-
wave oscilloscope combination. The fall time is larger than
the rise time as a result of critical damping of a resonance
at approximately 850 MHz within the I T T model
F4018(S-1) biplanarphotodiodes utilized in the UAC
Research Laboratories Model 1240 Phototransducer.
The use of a sampling oscilloscope greatly facilitates the -4 2ns I-
pulse-width measurementsof CW mode-locked lasers. The (b)
Fig. 6. Oscillogram of the output of an optical regenerative pulse oscilla-
relatively short overall time duration of approximately tor taken with (a) a 0.13 11s and (b) a 0.3 11s rise time traveling-wave
50 to 500ns of simultaneously Q-switched andmode- scope and a 90 ps detector.
locked lasersexcludes the use of sampling. scopes. Un-
fortunately, traveling-wave scopes require large-signal in-
putsand have considerablynarrowerbandwidths than
sampling scopes. For example, the Tektronix Model 519 use of a 16.5 m.long glass rod oscillator and a 76 cm long
traveling-wave scope hasapproximately 10V/cm sensi- glass rod amplifier has resulted in the generation of pulses
tivity and 1 GHz bandwidth, whereas millivolt sensitivity having apeak power between 10 GW and 100 GW and time
andbandwidths inexcessof10 GHz are available with durations between 10 ps and 2 ps, respectively. A discussion
sampling scopes. Nevertheless, an instrument-limited direct of the picosecond time duration measurement technique
measurement of0.15 ns rise time has been obtained [36J will be given in Section 111. The use of nonlinear optical
for a simultaneously Q-switched and mode-locked pulse measurement techniques tobe described (Section 111)
Nd3+: glass laser [see Fig. 6(a)]. The measurement utilized a revealed that the 2 ns and 0.75 ns pulse widths obtainedwith
modified Tektronix Model 519 traveling-wave scope having the 76 cm and 45 cm long rods were in actuality a grouping
a bandwidth of 3 GHz with a sensitivity of 217 V/div, and of picosecond pulseswhose repetition period was much
an ITT F4014 diode. If the rise time of the scope, i.e., 0.13 shorter than the 0.5 ns response time of the detection sys-
ns, is computed out of the measurement, a pulse width less tem. A grouping of three such pulsesis illustrated by
than 90 ps isobtained out of the detector. Theuse of ultra- Fig. 6(a).
short laser pulses with fast detectors can produce electrical Energy gain versus input energy data for a 76cm long
pulses up to 100 V in amplitude, with widths less than 90 ps glass rod amplifier utilized in the amplification of pico-
and repetition times to 2 ns. Such electrical pulses should second pulsesis shown in Fig. 7. The lower curve is the
find applications in the electronics industry to measure the actual energy gain, i.e.,the ratio of measured output to
characteristics of wide-bandwidth systems[47]. For ex- input energies, of the overall amplifier system. The upper,
ample, the ringing of astandard 1 GHz traveling-wave i.e., normalized, curve is the gain of the glass laser medium
scope when excited by a F4014 photodiode is illustrated by without losses. Eachexperimental point represents an
Fig. 6(b). The reason for the grouping of three pulses illus- average ofsix to eight separate measurements. Eastman
trated by Fig. 6(a) will be given later. 9740 saturable absorber was utilized for optical isolation
A 76 cm long by 1.8 cm diam Nd3 :glass rod was simul-
+
between the oscillator and the amplifier. The energy mea-
taneously mode-locked andQ-switched with a 1 cm dye cell surements were taken with ballistic thermopiles and lean
containing 16 cm3 of solvent and 5 cm3 of dye. The Q- heavily toward the conservative sidebecauseenergyloss
switched pulse envelope consisted of only six pulses, each due togas breakdown at the focus point within the thermo-
having a typical pulse width of approximately 2 ns time pile wasnot taken into account.
duration. The energy output was 44 J with 72 kJ input en- Fig. 8(a) is a schematic diagram of a pulse regenerative
ergy for an average energy per pulse of a little over 7 J. laser oscillator having a l o w 3 resonant reflectorwith a
A 47 J pulse train has been obtained with the use of a 45 thickness d , = 1 cm and surface reflectivities 5 percent and
cm long glass rod oscillator in conjunction with a 76 cm 35 percent, respectively. The distance between the 99'
long glass rod amplifier. The time duration of the individual percentand 35 percent reflectivity surfaces was120 cm.
pulses of the train were typically 0.75 ns for this case. The The simpldied operating model of Fig. 8(a) can be given
DEMARIA ET AL.: PICOSECOND LASER PULSES 9

s-ourpuT
102 99.9% 3:57. 75%

5
I
1 +21.6 crn (a)

20 ns/div (b)

1 2
f
0
Fig. 9. Oscillograms of the output of an optical regenerative pulse
oscillator operated with two coupled cavities.

p 10'
i detection system is much slower than 2 d,/c, the pulse width
of the individual grouping of pulses having a fundamental
5 repetition period of 2 d,/c will appear toincrease as a result
of the number of pulses within the grouping increasing with
time. Fig. 8(b) is an oscillogram of the output pulse of the
experiment illustrated in Fig. 8(a) which demonstrates this
effect. The response timeof the detection system was0.5 ns.
2 It isbelieved that t h mechanism is responsible for the
104 2 5 105
AMPLIFIER P U M P E N E R G Y - JOULES multiple pulsing illustrated in Fig. 6(a).
Fig. 7. Energy gain as a function of pumping energy for a 76 cm
If the thickness and Q of the resonant reflector are in-
long Nd3+:glass laser amplifier. creased to the values shown in the schematic of Fig. 9(a),
the individual pulses can be resolved with currently avail-
able detection systems. Fig. 9(b) and (c) presents oscillo-
grams of the resulting pulse trains at sweep speedsof 20 and
-b 120 cm4- 10 ns/div, respectively. For the case shown by Fig. 9(c),
the two coupled cavities were not exact multiples of one
another and phase shifts at the nodal points of the modu-
lating envelopecan be observed. The useof these tech-
niquescan be utilized to generate pulse rates extending
well into the microwave region. If such multiple pulsing is
to be eliminated, these data clearly reveal the necessity of
placing all surfaces within a regenerative laser oscillator
d2 cavity at Brewster's angle. Similar attention must be paid
to backscattering in the dye cell and other components of
the laser cavity, as well as to backscattering back into the
SWEEPSPEED 20 nr/div laser cavity from external objects.
Fig. 8. Oscillograms of the output of an optical regenerative pulse If the dye cell inside a laser is moved a distance dl from
oscillator operated with a resonant reflector. one of the mirrors, double pulsing is obtained [16]. The
time separating the twopulses correspondsto the time
t = 2 d l / c . The pulse-crossover points alternate between a
as follows. Assume that at t =O two pulses exist at x =dl point at a distance dl in front of each reflector inside the
whose pulse width AT<<d J c . After t=O the pulses travel laser cavity. In general, Harrach and Kachen have found
in opposite directions, and the pulse traveling to theleft that the pulse repetition rate of the laser is mc/2L when
bounces back and forth between the two surfaces of the m = L/d, is an integer [48].
resonant reflector for a time determined by the Q of the For some applications longer .intervals between, pulses
resonator. For the illustrated case, the delay time is ap- are desirable. A longer pulse repetition time can be obtained
proximately 50 ps. If the Q is low enough, we can assume by inserting an optical delay line as part of the cavity of a
only one extra pulse is produced by the resonant reflector. regenerative laser oscillator. It has been reported that such
For all practical purposes, we can then assume that two long optical delay lines can be made compact by reflecting
pulses travel to the right and left of the resonant reflector the beam repeatedly between two spherical mirrors without
separated by a time t = 2 dl/c 2: 0.1 ns, with the first pulse interference between adjacent beams [49]. The diffraction
havingamuch larger amplitudethan the second. Each losses of such a delayline are much lower than for an open
time the two pulses pass through the resonant reflector the beam because of the periodic focusing of the mirrors.
process is repeated, givingrise to 2(m+ pulses for m Besides effectively increasing the optical distance between
traversals in the laser cavitv. If the remonse time of the the cavity mirrors, this technique can yield an automatic
10 PROCEEDINGS OF THE IEEE, JANUARY 1969

nonmechanical digital scan of the pulsed laser output [so] RI nr

if the output from one of the delay line mirrors is utilized. - 84 cm{ -
Fig. 10 illustrates &he arrangement usedin one experi-
ment. The radius of curvature was 1 m for mirrors R4 and
R,. Mirrors R3 and R, were used to inject the optical beam
into the optical delay line through a 1.8 cm aperture in
mirror R, in such a way that the repeated reflections traced
out a circle on the surfaces of mirrors R, and R,. For most
cases in our experiments, the light beam slope selected by ~

the angular adjustment of R, was so chosen that the light 73.6 crn
beam made six round trips between R, and R, andthe 85 cm ____I
trajectory traced out a six-spot circular pattern on each (a)

mirror of the delay line. After the twelfth pass in the delay
line, the lightbeamwasreflectedback onto itself and
through the delay line for another 12 traversals. Fig. 10(b)
illustrates a pulse separation of approximately 7 1.5 ns.
C. Single Picosecond Pulse Generation
There are many applications where only a single ultra-
short light pulse is required rather than a train. The first
method for obtaining a single hgh-peak-power laser pulse (b)
Fig. 10. Schematicdiagram of anopticalregenerativepulseoscillator
having a pulse width narrower than those obtainable from incorporating a folded optical delay linefor obtaining a long pulsating
standard Q-switched lasers was proposed by Vuylsteke [51]. period.
This method involved @switching a laser with mirrors of
100 percent reflectivity on bothends of the cavity and, at the
peak of the pulse, rapidly switching the output mirror from MI
MODE - LOCKING
DYE CELL
OUTPUT TO
EXPERIMENT
/
,’
100 percent to 0 reflectivity. In this manner, the optical
energy stored within the cavity would be dumped in the
time required for the round-trip transit time. This type of
laser operation, called the pulse transmission mode, was
reported to have resulted in pulse widths of 4 ns [52]. The
experimentaltechnique of combining the simultaneous
Q-switching and mode-lockingwith the pulse transmission
Fig. 1 1. Experimental schematic for selecting a single pulse
mode is, we believe, the one capable of yielding the highest from a train of ultrashort pulses.
peak power with the shortest pulse width [53]-[57].
A schematic diagram of an early single pulse selection
experiment isgivenin Fig. 11. The reflectivityof both tion, a 112 voltage is required across the polarizing switch
mirrors is high. In addition to the dye cell, a Glan prism [54]. This latter configuration is preferred when the mini-
polarizer anda polarizing switch such as aPockels or mum pulse width is desired [56].
Kerr cell are inserted into the feedback path of the laser. The main problemarea in performing a single ultrashort
The polarizing switchwas initially unenergized, andthe pulse selection experiment is the high-voltage pulser. The
polarizer was adjusted for maximum transmission. A pulse rise time of the high voltage supplied to the polarizing switch
of radiation then “bounces” back and forth between the must be faster than the time required for the light pulse to
two reflectors.The leakage radiation from one of the mirrors make a round-trip traversal between the two mirrors. The
is used to trigger a high-voltage pulser at a predetermined high-voltage pulser utilized in our experimentswas a Marx-
optical pulse amplitude. The high-voltage pulser energizes bank pulse generator common to manyhigh-energy physics
the polarization switch to its 1/4 voltage. After the pulse experiments. Avalanche transistors assembled in a Marx-
has made one traversal through the energized polarizing bank configuration wereused toproducea pulse in the
switch, the polarization of the pulse is rotated 45”. For a 1 kV range, which in turn was used to trigger a spark gap
two-trip traversal of the polarizing switch, the polarization holding off a voltage in the 15 kV range. An open-ended
of the pulse is rotated 90”. The Glan prism prevents the transmission line was then used to double this voltage to 30
propagation of the pulse with this polarization in the sys- kV to drive the Kerr cell polarizing switch.
tem, andthe propagation direction of the pulse is redirected A single subnanosecond pulse has been obtained with the
as illustrated. In effect, the high reflectivity of the cavity is experimental arrangement shown in Fig. 11 [16], [17]. The
suddenly changed to a low reflectivity and the pulse stored recorded full width atthe half-intensity points was measured
within the cavity is suddenly “dumped.” The experiment with a 0.5 11s response time detection system and found to
+
can also be performed with the polarization switch and be 0.63 ns withan energy content of J. The use of the two-
Glan polarizer outside the cavity. For this latter configura- photon absorption measurementtechniques to be described
DEMARIA ET AL.: PICOSECOND LASER PULSES 11

in Section I11 reveals that these pulses usually consisted of a

substructure of picosecond pulses. The amplification of such
single pulses witha 75 cm long amplifier has resulted in 1.8 J
of energy [ S I . Since the damage threshold as a function of
power decreases with decreasing pulse widths, the further
addition of stages of amplification should make possible
the generation of picosecond pulses having peak powers
well in excess of 1013 W. The use ofthese pulses incontrolled
thermonuclear research appears particularly promising.

111. MEASUREMENT TECHNIQUES

FOR PICOSECOND
LASERPULSES
A . Introduction
SQUARELAWDETECTOR
Direct measurement of the duration of relatively long
optical pulses is most often made by displaying, by means Fig. 12. A Michelson interferometer, a typical linear
of an oscilloscope, the output of a suitable photodetector optical instrument.
illuminated by the optical radiation.High-spied photo-
diodes and oscilloscopes have been utilized in this manner
to obtain direct pulse-width measurements.of 1.5 x 10- l o s B. Linear Optical Pulse- Width MeasurementTechniques
for a mode-locked CW YAG laser [38] and a Nd3+: glass 1) Theory :For purposes of illustration, let us consider a
laser [16]. A power spectrum measurementof the oscillating linear optical instrument such as the Michelson interferom-
laser bandwidth made by means of a scanning Fabry-Perot eter diagramed in Fig. 12. An incident pulse having an
interferometer yielded an indirect measurement of amplitude E ( t ) is split into two pulses, each with an ampli-
7.6 x lo-" s for the YAG experiment, and a grating spec- tude E(t)/$. Each of these pulses is made to traverse a
trometer yielded 2 x s for the Nd3+: glasslaserex- separate orthogonal armof the interferometer. After travers-
periment. Since it is not expected .that direct electronic ing their respective paths 2 0 , and 2D,, the pulses are re-
techniques will be capable of measuring time durations combined on a squarelaw detector such as a photographic
down to 10- l 3 s, new measuring techniques had to be plate or photodetector. If Dl # D,, the pulses can be repre-
found for the measurement .of the time duration of pico- sented by
second laser pulses.
It will be shown in part B of this section that any linear
interferometer measures the autocorrelationfunction of the
pulse amplitude. Since the power density spectrum and the
amplitude autocorrelationfunction are a Fourier transform and
pair, knowledge of one uniquely specifies the other. The
power density spectrum is usually measured with a spec-
trometer. Theequivalence arises from the fact thatboth
instruments used to measure these twoparameters are where T =2(D, - D l ) / c and Eo(t) is the slowly varying en-
linear optical systems. The linear interferometer can also velope of the pulse with respect to w. The intensity incident
be used to measure the coherence length AD and the co- on the detector is given by
herence time A t c ; these two, parameters are related to the
Z(t, t) = +[Eo@)+ E,(t - T)/,. (12)
spectral bandwidth Aw and to each other by the following
relationship : The responseof the detector is assumed to be slow compared
to the pulse duration or the delay time T , so the output
AD 2: -7zc
= 2. At c
signal S ( T )of the detector is given by
Aw 2
r m
Since the relation AoAT, 2 271 provides only a lower limit to S(t) = J Z(t, r)dt
a pulse duration for a given Aw? it can be concluded that -m

any measurement taken with a linear optical system can = W(1 + A(z))
provide only information establishing a lower limit to the
time duration of a pulse. where W is the pulse energy and A ( t )is the autocorrelation
It will be shown in part C that measurements taken with function of the pulse amplitude, i.e.,
nonlinear optical instruments can provide information foT
determining the actual time duration of a pulse. Nonlinear
optical instrumentsmeasure the autocorrelation of the
W = 1 X

J - X
Eg(r)dt

pulse intensity. and

 12 PROCEEDINGS O F THE IEEE, JANUARY 1969

A(T) = (15) I
J;mEiwr

When T =0, then S(o)/W = 2 , and when T is large enough so

that no overlap between E,(r) and E,(t) exists, then S(t)/W

rn
= 1. From Fourier analysis, we find that

1
A(?) =- ejw')P(o)12do (16)
&. -03

where IP(o)12is the power density spectrum of the original

laser pulse [58]. Since the power density spectrum and the
amplitude autocorrelationfunction are a Fourier transform
pair, knowledge of one uniquely specifies the other. The
power density spectrum is the quantity measured with a
Fig. 13. For a pulse (a) having a bandwidth Av determined by the width
spectrometerand the amplitudeautocorrelation is the of the pulse Arc= l/Av, a linear optical instrument gives a signal-to-
quantity measured with an interferometer. The two results background ratio (b) of 2 for r=O and 1 for T>> AT<.For a pulse (c)
are essentially equivalent and the equivalence arises from having the characteristic AT> IIAv, the identical result is obtained for
thecontrastratioandpulse-widthmeasurement.Theinstrument
the fact that both instruments are linear optical systems. cannot distinguish between the two cases.
Similar considerations hold for any linear optical instru-
ment.
It is clear from Fig. 12 that aninterference pattern will be _---- MODE-LOCKED
0-SWITCHED
observed in the planeof the detector, if the interfering pulses
NORMAL
have a relative retardation in time equal to T =0, 2L/c,
4L/c, . . . ,i.e., if the difference inlength of the interferometer
arms is D,-Dl =0, L , 2 L ; - . , where L is the separation
between the laser reflectors. The visibility of the interference
pattern changes with the relative delay time of the pulses,
is a maximum at D , -Dl = 0, L, 2L,. . . , and gradually falls
to zero in moving away fromthese points. The difference in
path lengths, AD, over which the interference pattern is
visible is called the coherence length. The coherence length
andthe spectral bandwidth of the pulse are related by
2ADlc = 271/Ao AT^, where AT, is the coherence timeof the
pulse.
The coherence time A q will equal the duration of the
pulse only in the special case where the entire spectral
content of the pulse is due to the short duration of its
envelope. In other words, the relation AUAT 2 271 provides
only alower limit to the pulse duration. Fig. 13schematically
summarizes the results of measurements with a linear opti- 1.os 1.06 1.07 1.08

cal instrument for pulses having a pulse-width (AT) band- A IN MICRONS

width (Av) relationship given by AT= Arc= I/Av and by Fig. 14. Comparison of thespectralCharacteristics of an Nd3+ :glass
laser in various modesof operation.
AT>> I/Av.
2 ) Experiments: Fig. 14 illustrates the spectral charac-
teristics of the output from a12.2 cm long by 0.95 cm diam Q-switched and mode-locked Nd: glass laser with the dye
Brewster-ended Nd3+:glass rod operated normally, Kerr cellplaced at Brewster's angle revealed auniformly dis-
cell @switched,and as a regenerative pulse oscillator. tributed 180 A wide spectrum with long leading and trailing
Eastman 9740 saturable absorberwas utilized as the optical edges. The increase in spectral width results from the
expandor element. The spectra were obtained with a 3.4 m tendency of the saturable absorber expandor element to
Jarrell-Ash spectrometer. The Kerr cellQ-switch type of distribute the energy evenly throughout the spectral line
operation yielded a uniformspectral width of approximately width of the laser medium by the generation of sidebands
50 A. The overexposure of the film by overlappingthe at the resonances of the Fabry-Perot interferometer. To a
spectra of four or more Kerr.cel1 Q-switchedpulses still first approximation, the minimum pulse width obtainable
revealed apowerspectrum of approximately 50 8, with with the harmonic content revealed by the spectral data is
sharply defined ends. The spectrum of the simultaneously 2 x s, with acorresponding peak power of 1O'O W.
DEMARIA ET AL.: PICOSECOND LASER PULSES 13

A close observation of the spectrum of the simultaneously visibility of the fringes of pulses occurring in the middle of
Q-switched andmode-locked laser of Fig. 14 reveals a the pulse trainrecorded by an image-convertor streak
relatively intense line at 1.06 microns, i.e., the peak of the camera, as a function of the length variation of one leg of
Nd3+ gain line. This sharp line is believed to arise from the the interferometer of Fig. 12 [16]. For a length variation
relatively slowbuildup rates of the modes within a passively greater than 8 x cm, the fringes disappeared. The
Q-switched laser. For example, in afast switched laser minimum pulse width calculated from a coherence length
oscillator, i.e., switchedwith a Kerr cell, rotating prism, of 8 x is 5 x
cm s, as expected from the discus-
etc., the pulse buildup takes approximately 10 to 40 loop sion on theory (part B-1) in this section. These data are in
transits, whereas for a passivelyQ-switchedlaser this excellent agreement with the spectral data of Fig. 14.
buildup requires typically several hundred to a thousand In general, it was found that when the modes of the laser
transits [59]. Such a long buildup time for the passively were only partially locked in phase (as observed by a con-
Q-switched case favors the existence of relatively few axial siderable dc level in the oscilloscope traces), a correspond-
modes at line center for a long time before an appreciable ing reduction in spectral content of the pulses was noted
sharpening takes place. When a streak imageconverter 1161. It is important to note that even under normal Q-
camera was used in conjunction with a 3.4 m Jarrell-Ash switched or free-running operations, the spectral content
spectrometer, the sharp line at 1.06 microns did not appear of a Nd3+:glass laser pulse is considerably greater than the
in the time resolved spectral data taken for each individual observed pulse widths would indicate (see Fig. 14).As a
pulse [16]. The spectral content of each pulse was found to result, it is to be expected that amplitude or phasefluctua-
be 100 A to 120 A. A channeled spectrum was obtained in tions corresponding to a timeduration equal to the inverse
any case where additionalreflecting surfaces were placed in u€the bandwidth of the spectrum must exist in the output.
the cavity. For example, with the dye cell placed normal to It is also expected that a finite probability exists in a series of
the laser axis, a grouping of picosecond pulses separated in experimental runs that acertain number of axial modes will
time by the optical thickness of the dye cell was obtained statistically have the properphase as to yield periodic
and it correspondedtothe reciprocal of the frequency amplitude fluctuations riding on the envelope of the Q-
separation of the channeledspectrum. The grouping of switched pulse. In general, the periods of these periodic
pulses had a repetition frequency equal to the axial mode- fluctuations are equal to the round-trip traversal time of the
spacing frequency of the cavity. cavity and can thus be easily observed on an oscilloscope.
A still closer inspection of the spectrum of the simul- It has been recently reported that picosecond pulses nor-
taneously Q-switched and mode-locked laser of Fig. 14 re- mally appear in free-running and Q-switched ruby,
veals an unsymmetrical distribution of the spectral content Nd: glass, and Nd: YAG lasers even when periodic ampli-
about line center. The spectrum shows a tendency to expand tude fluctuations were either not observed on an oscillo-
toward the longwavelength region undermode-locked scope or, if observed, had a width longer than the response
conditions. We believe this effect is caused by the fact that time of the detection system [60]-[62]. The conclusion of
the peak of the laser line profile falls on the slope of the the picosecond pulsating output of free-running or Q-
long wavelength side of the absorption line of the Eastman switched ruby, Nd: glass, or Nd: YAG lasers was based on
9740 saturable absorber [16]. In effect, this provides less the data obtained from two-photonabsorption-fluorescence
absorption, andtherefore higher overall system gain toward experiments. A reviewof the two-photonabsorption-
the longer wavelength region of the oscillator spectrum. fluorescence technique for the measurementof a pulse's time
This effect is not expected to be as pronounced with East- duration in the picosecond range will be given in part C
man 9860 dye because of the closer coincidence of the peak of this section. It willsuffice tostateat this point that
of the laser and dye spectral line. extreme care mustbe taken in interpreting the fluorescence
In order to determine the spectral extent of the mode- track in such pulse-width measuring experiments[63].
locked pulses, a movable diode and slit were placed in the It iswell known that the nonlinear properties of the
exit focal plane of a spectrometer [36]. Data on the tem- amplifying laser medium can give rise to self-locking of the
poral behavior of 1 A and 0.5 A selected regions of the modes within gas and solid-state lasers [31], [64]-[69]. For
spectra were obtainedandcomparedwith the simul- the He-Ne laser case, it has been shown that stable mode-
taneously recorded data of the input pulse train. It was locking operation is most stable under conditions of low
found that the oscilloscope pulse-width measurements of laser excitation [67]. Stable operationcan be obtained
the radiation emitted by the limited spectral region were under high-excitation conditions by using an auxiliary
still instrument-limited. This is to be expected since the discharge tube containing pure Ne as a nonlinear loss ele-
number of modes available in these limited apertures is ment inside the laser cavity [70]. Self-locking of lasers is
still sufficient to produce pulses of the order of 3.6 x lo-" not very reproduciblebecause it is presumably difficult,
and 1.8 x 10-l' s. Since no change in pulse shape or width except in carefully controlled situations, to couple a large
was observed in random samplings of 0.6 percent and 0.3 number of axial modes [31], [66], [69]. Usually self-mode-
percent portions of the total input spectral range, it can be locking is an occasional occurrence and difficult to repro-
concluded that modecoupling is extensive over the entire duce. Whenself-mode-locking repetitively occurs with a
spectral output. solid-state laser Q-switched system, it usually is peculiar
An experiment was performed on the variation of the to that one laser system and not necessarily reproducible
14 PROCEEDINGS OF THE IEEE, JANUARY 1969

with other lasers of the same type. Needless to say, self-

mode-locking of lasers is not well understood at this time
even though considerable effort has been expended on this
topic [31], [60],[64]-[72].
C. Nonlinear Opticul Pulse- Width Measurement Techniques
1) Theory: In part B of this section, itwas noted that
linear optical systems can yield only a lower limit on the
pulse width of optical pulses, whereas measurements per-
formed with nonlinear optical instrumentscan yield the
actual pulse width. The mannerin whicha nonlinear optical
system can perform a true measurement of a pulse can be
explained as follows. Suppose one passes the two output
pulses E , ( t ) and E,@) from the linear interferometer of Fig.
12 through a nonlinear optical crystal. The second harmonic (C) (d)
output from the nonlinear optical crystal will be given by Fig. 15. For a pulse (a) having a bandwidth Av determined by the width
of the pulseAT<= I!Av, a nonlinear optical instrument gives a signal-to-
*"E(t) = [E,(f) + E2(f - 5 ) l 2 (17) backgroundratio (b) of 3 for T =0, unity for T >>AT,, andatrue
pulse-width measurement. For a pulse (c) A T > l/Av. a measurementof
if one neglects the constant representing the second har- the true pulse widthAT and the coherence width can simultaneouslybe
monic generationefficiency of the crystal. The outputsignal obtained.Theshadedareas of (b) and (d) representinterference
fringes.
S ( t ) fromadetectorhavinga slow responsetime with
respect to w and t is given by

S ( t ) = J:m/"E(t)12dt = 2"W(1 + 2G(t)) (18)

where *"W is the second harmonic pulse energy and G(T)

is the autocorrelation function of the pulse intensity, i.e.,
OPTIC AXIS
20W = j:aE4(t)dt (19) Fig.16.Schematicexperimentalarrangement for measuringpicosecond
laser pulses with the second harmonic technique.

and ps/mm at i = 1.06. The two orthogonalpolarized pulses are

then caused to interact in a nonlinear optical crystal, in this
Jy/2(r)E2(r - t)dt case ADP, to produce second harmonic radiation 5300 at A.
G(T)= (20) The orientation of the nonlinear crystal is chosen such that
no second harmonicis produced by either component of the
J;*E'(t)dt pulse acting alone, but only when both are present. The
method thus, measures the "overlap" of the pulse with a
When t = 0, S(T)/~"W = 3, and when T is large enough so that delayed replica of the pulse. When a delayis introduced such
no overlap between E , @ ) and E 2 ( t ) exists, S ( T ) / ~ " W 1.= that no overlap of the two pulses occurs, the second har-
Measurement of the variation of S ( T ) / ~ " W as a function of 7 monic drops tozero.
gives the timeduration over which the energy of the pulse is For ADP, the angle at which the second harmonic is
distributed. Fig. 15 schematically summarizes the results of proportional tothe product of two orthogonalpolarizations
measurements with a nonlinear optical instrument for the was reported by Weber [73] to be 61'12'. This angle pro-
two basic types of pulses AT= AT^= l/Av and A7 >> l/Av. duces phase matching for this combination of vertical and
A nonlinear interferometer can also be constructed in horizontal polarization and produces second harmonic ra-
which S ( Z ) / ~ " , W0=when z is large enough so that no over- diation with horizontal polarization. If we let input signal E i
lap between E , ( t ) and E2(t) exists. Suppose one passes an to the ADP crystal be represented by
optical pulse througha birefringent crystal of length L
with the optical polarization 45' with respect to the optic E,@) = E,(t)e*' + E,(t - T)e*('-') (21)
axis (see Fig.16). The crystal will resolve the single pulseinto
the second harmonicsignal 2"EH(t)is given by
two pulses of equal amplitude and orthogonal polarization.
One polarization component of the pulse propagates as an 2w E,@) = E,(t)E,(t - T)eb(2r-r)
(22)
ordinary ray and the other as an extraordinary ray. For a
crystal of thickness L, the delay T so introduced is given by where E H ( t )and E,(t - t ) are the slow-varying envelopeof
z = t(ne- no) c, where ne and no are the extraordinary and the horizontaland vertical polarized electric fieldswith
ordinary indices of refraction. For calcite, the delay is 0.59 respect to w. The constantrepresenting the second harmonic
DEMARIA ET AL.: PICOSECOND LASER PULSES 15

generating efficiency has also beenneglected. If T>> AT, t

then ’,EH=O, and if T = O , then Z”E, =max. The output
signal from a detector having a slow response time with
respect to 2w is given by

S(T) = [:mEi(t)Ei(t - T)dt.

(23)

Since a separate experimenta1,runis required for each data t A r f l/Av

point and since each experimental run is independent of
previous runs, normalization is required in the dataprocess-
ing. A convenient meansof normalizing is to take apulse of
any polarization andlet it alone produce second harmonic
energy. This reference signal can be represented by (19).
(C) (d)
The experimentally measured quantity is S(T)/*[~W=Z(r)
Fig. 17. For a pulse (a) AT=AT,= l/Av, an orthogonally polarized non-
as a function of T by selecting different lengths of delay linear optical instrument gives an infinite signal-to-background ratio
crystals. When T = O , S ( T ) / ~ , W= 1, and when T is large (b) for r=O, zero for T>>AT,, and a true pulse-width measurement.
enough so thatnooverlap between E H ( t ) and E , ( ~ - T ) For a pulse (c) AT> l/Av, a measurement of the true pulse width AT
and thecoherence widthcan simultaneously be obtained with no back-
exists, S ( T ) / ~ ” W = OThe
. nonlinearoptical system utilized in ground signal. The shaded areas in (b) and (d) represent interference
the measurement of picosecond pulses in [ 4 3 ] , [ 7 4 E [ 7 6was
] fringes.
essentially one of the nonlinear systems described above.
Fig. 17 schematically summarizes the results of measure-
ments with the orthogonally. polarized nonlinear optical
instrument for the two basic types of pulses AT = AT^ = l / A v
and AT>> I/Av.
Recently, Giordmaine et al. [ 7 7 ] , [ 7 8 ]reported the two-
photon absorption-fluorescence technique for the measure-
ment of picosecond laser pulses. This technique is also an
aCAMERA

intensity-correlation system. Its simplicity is its major ad-

vantage over the second harmonic systems described pre-
viously. A schematic diagramof the two-photon absorption-
fluorescence measurementexperimentalarrangement is
illustrated byFig.18. The correlation of one pulse with
succeeding pulses (cross-correlation) illustrated by Fig.
Fig. 18. Experimental arrangement
for
performing
intensity
cross-
18(a), or the correlation of a pulse with itself (autocorrela- correlationsandintensityautocorrelationmeasurements of pico-
tion) illustrated by Fig. 18(b),can be obtained. Both cross- second pulses by the two-photon absorption-fluorescence technique.
correlation and autocorrelation experiments have been per-
formed by the authors and identical results were obtained
in the two measurements.
If we consider the center of the cell to be the origin (z = 0)
I , c/; { E ( t - :) + E:( t+ f) + 2E:(t - :$(t + f)
and denote pulses approaching the cell from the left and
+
right as * [l + C COS 2kz)’I

[ ( f) ( + f)
4 E: t - - E2t - (25)

Et
( 3
-- sin(kz- w t ) and
( + -3
E, t sin(-kz- at),
+ El(t -$Ei(t + f)]cos2ir}-
respectively, the intensity I , in the dyecellwill then be The photographicfilm essentially records the value of I , but
given by it is time averaged through the photographic process and
spatially averagedover several optical wavelengthsdue
to the limited resolution of the film. The value recorded by
(24) the film is
+ E:(t + f).
If we assume that the fluorescent intensity I, is proportional
to the intensity squared, I , a Z ~ , thenthe fluorescent in-
tensity is given by
16 PROCEEDINGS OF THE IEEE,JANUARY 1969

The normalization of (26)gives

which is identical to (18). If E,(t)=E,(t) at z=O is assumed,

a bright vertical line will be recorded with a contrast ratio
of 3 to 1 with respect to the background. This line will have
a width AL=Az,c/n, where AT is the pulse duration, c is the
velocity of light, and n is the refractive index of the two- 40°
0.6 0.7 0.8 0.9 1.0 1.1
photon absorption-fluorescence dye.

90m\
A- MICRONS
It is important to realize that d a noise pulse having a
time duration A T s and a spectral bandwidth of Am rad,
1
- PHASE MATCHING IMPOSSIBLE
in which AT>>2rc/Aw, ispassed throughatwo-photon
absorption-fluorescence cell, a bright line having a length
corresponding to the coherence time A ~ , 1 : 2 7 ~ / Awill w be
recorded. The evaluation of the pulse duration by this 50°
70a I I I I

method requires caution because avery similar fluorescence

structure is obtained from the radiation of an ideally mode- 1.0 1.04 1.08 1.12 1.16 1.20
locked laser and from a free-running laser with the same A - MICRONS
oscillating bandwidth [63]. Theproper interpretationof the Fig. 19. A plot of the second harmonic phasematching angles
data depends strongly on the contrast ratio in the photo- as a functionof wavelength for KDP andLiNbO,.
graphic record. Fora bandwidth-limited short pulse
(AT'v l / A v ) , the maximum ratio of (27) is 3. For a free-
running laser, the maximum ratio of (27) was calculated
by Weber to be 1.5 [63]. It cannot be concludedat this
point that the output of a free-running or Q-switched ruby,
Nd :glass, or Nd :YAG laser normally consists of repetitive
picosecond pulses. On the other hand, there are sufficient
data available in the literature to support the conclusion
that the output of regenerative pulselaser oscillators
utilizing saturable absorbersconsists of pulses in the IO-',
to lo-'' s range.
2 ) Experiments: In the analysis of the nonlinear optical
system for the measurement of picosecond pulses, it has
been assumed that phase matching in the nonlinear crystal
is maintained over the entire bandwidth of the laser pulse.
This sets an upper limit to the thickness of the crystal or a 530A0i
lower limit to the time resolution. ADP and KDP crystals +-loo A
4-
having a thickness of 1 mm provide resolutions of approxi- R= Imm
mately 1 ps. It is important to note thatthe group velocities 8- 41'
of the fundamental and second harmonic frequencies must
be approximately equal in order to have the envelope of ae- 0.20
Fig. 20. The second harmonic spectrum of the picosecond pulses
the second harmonicpulse equal to the square of the funda- generated by an Nd3+:glass laser in 1 mm thick KDP crystals.
mental pulse [79]-[82]. If the group velocities of the two
harmonically related frequencies are not equal, the second
harmonic pulse will have a flat top with a time duration that the spectral width of the second harmonic picosecond
L,/[cg(2w)]-L,/c,(w), where LC is the crystal length, and pulses can be conveniently adjusted by varying the phase
cg(w) and ~ ~ ( 2are 0 )the group velocities of thefundamental matching angle.
andsecondharmonic frequencies, respectively. Fig. 19 Themeasurement of picosecond laser pulses by the
illustrates plots of the phase matching angles as a function second harmonic technique was performed by Armstrong
of wavelength for KDP and LiNbO,. Note that for KDP [ 7 4 ] , Glennand Brienza [75], andMaier, Kaiser, and
the curve is relatively flat around io= 1.06 microns, there- Giordmaine. [76]. Glenn and Brienza also found that the
fore making it possible to phase match easily over several early pulses in the train of pulses emitted by their laser were
hundred angstroms. Fig. 20 illustrates the spectrum of the shorter than the later pulses in the train. Since their laser
second harmonic of a Nd: glass laser's picosecond pulses pulses had a variable pulse width, an additional parameter
generated by a 1 rnm thick KDP crystal. The three spectra was needed to characterize the pulses in their measurement.
were taken around the phase matchingangle condition with This characterization was needed so that similar pulses in
a changeof 0.2" between each of the three shots. It is evident successive firing of the laser could be compared.The
DEMARIA ET AL.: PICOSECOND LASER PULSES 17

SINGLE

z
> "h
4120x10-'2s I--
+ 240x10-'* s +

(b) (d)
PULSE DURATION-5x10 s
-Iy
0 5 10 152 5 2 0 30 Fig. 23. Photographs of picosecond laser pulses taken with the two-
PULSE NUMBER photon absorption-fluorescence measurement technique.
Fig. 21. Thebehavior of theratio of thereferencesecondharmonic
energy tofundamental energysquared (i.e., conversion efficiency)
as a functionof the number of pulses of an Nd3+:glass laser. limiting width as determined from the spectral width is of
the order of 10- l 3 s. Thedye shortens the initial low-
amplitude pulses to the same limiting widths. The pulse
amplitudethen begins to grow rapidly. For sufficiently
high intensity, the dye willbleach completely at the be-
ginning of each passage of the pulse. Since the dye is satu-
rated for most of the pulse width, it will no longer be effec-
tive in shortening the pulses. Thereafter, the width of the
pulse is determined by the physical properties of the laser
cavity. Such factors as dispersion of the glass rod and dye
cell, inhomogeneities of the optical path, transverse modes,
- LONGEST WUES SHORTEST PUUES- diffraction limit, energy storage properties of mirrors, and
0 '
20 30 40 50
saturation effects of the gain profile all contribute to the
(REFERENCE S.H. ENERGY) ;A1l~lTRARy growth of the time duration of the pulses. The cumulative
(FUNDAMENTAL ENERGY)' effects grow as the square root of the number of passes
Fig. 22. The variation of the intensity correlation as a function of con- which could account for a near saturationin the growth of
version efficiency of the picosecond pulses emitted by an Nd3+: glass
laser. the pulse width as was experimentally observed [75].It has
not been definitely determined at the present time whether
all pulse regenerative laser oscillators have variable pulse-
parameter chosen was the conversion efficiency,i.e., the width outputs as a function of time.
ratio of the energy of a reference second harmonic pulse Fig. 23 presents typical data obtained by the two-
W, to the square of the energy of the fundamental pulse photon absorption-fluorescence measurement technique.
W:. The highest value of this parameter occurs atthe start Fig. 23(a) shows the bright fluorescent line due to the over-
of the pulse train and then decreases uniformly with the lap of a pulse with itself or with its immediate predecessor.
increasing number of pulses. This behavior is illustrated by The fact that thebright line isthe overlap point of two pulses
Fig. 21 [75]. The intensity correlationratio S(z)/W,., was confirmed by the observation that a movement of the
plotted as a function of W,/ W: for several different delays, reflecting mirror in (a) or the dye cell in(b), through a given
is illustrated byFig.22. The energy correlation function distance, caused the bright line to move a corresponding
for a pulse of given width may be determined by drawing a distance in the dye cell. Rhodamine 6G was utilized as the
vertical line on the graph and reading the value of the cor- dye medium. In anethanol solution this dye has its primary
relation as a function of the delay time. If the correlation absorption peak at a wavelength that is quite close to the
function is assumedto be Lorentzian, then the shortest 5300 A second harmonic of the Nd :glass laser. In addition,
pulses obtained in this set of data had a full width at half- it has another strong absorption linewhich peaks very
maximum of 8 x Treacy [23] has also noticed a close to the second harmonicof the ruby laser line.The two-
periodic variation in pulse widthfrom pulse to pulse photon absorption and subsequent fluorescence at 5500 A
throughout the pulse train.The periodic variation was of this dye is relatively much larger than for the 1,2,5,6-
noticed by comparing the second harmonic signal generated dibenzanthracene (DBA) dye previously used insuchex-
by the compressed picosecond pulses with the second har- periments [77]. Anadditionaladvantage of Rhodamine
monic signal generated by the uncompressed pulses. 6G is that the experiment can be performed by direct ir-
A tentative explanation for the behavior of the pulses radiation ateither 1.06 microns or 6943 A without requiring
recorded in [75]can be summarized as follows. The action the conversion of these wavelengths into their second
of the saturable dyein the laser cavity tends to shortenthe harmonic before irradiating the dye. The use of Rhodamine
pulse by reason of its nonlinear absorption. Thetheoretical 6G dye in this experimental technique has made possible
18 PROCEEDINGS OF THE IEEE, JANUARY 1969

the measurement of a single ultrashort pulse with only one IV. RECENT EXPERIMENTS
PERFORMED
WITH
firing of the laser. PICOSECOND
PULSES
Under certain conditions multiple bright lines were ob-
tained as shown in Fig. 23(c) and (d). The optical path A . Measurement of Ultrashort Decay Times
length between the bright lines corresponds to one-half of Since early in the 1930’s it has been possible to measure
the actual pulse separation in thelaser output. The time fluorescence lifetimes shorter than 1 ns [85]. However, the
scales in Fig. 23 have taken this factor into account, and fluorometers devised for these measurements are complex
therefore repment the actualpulse separation in time.These and cumbersome. Moreover, inasmuch as the actual decay
multiple pulses are caused by mode selection within the curve is not observed, the measurements are indirect. The
laser bandwidth as previously described in this paper (see high intensity and short duration of mode-locked pulses
Figs. 8 and 9). The presenceof the reflecting mirror of make them ideal for a number of different types of lifetime
Fig. 18(a) can act as a secondary cavity external to the main measurements.
laser cavity, and gives rise to a channeled spectrum in the To observe a fluorescentdecaytime, a short-duration,
frequency domain or multiple pulsesinthetime domain high-intensity source is required to excite the fluorescence
[see Fig. 23(c)]. Experiments were also performed with the and a wide-bandwidth detection system is needed to detect
insertion of optical flats into the laser cavity normal to the it. Regenerativepulselasers, as discussed earlier in this
laser axis. Multiple bright lines were again obtained with paper, are anideal excitation source for such measurements.
the separation of the lines in time equal to twice the optical The use of the fundamental and second harmonic of ruby
thickness of the flat inthe laser [see Fig.23(d)]. Pulse widths and neodymium lasers [86] offerspromise of revolutionizing
as short as 2 ps and as long as 25 ps were recorded by this these measurement techniques [41]. With the continuously
technique. It should be noted, however, that the photograph mode-locked lasers and crossed-field photomultiplier [87]
of Fig. 23 is an average over the entire pulse train and not used in conjunction with sampling oscilloscopes, an overall
the width characteristic of a single pulse inthe train. detection rise time as short as 60 ps can be obtained. With
Weber [63] has noted the danger in drawing the conclu- Q-switchedlasers and traveling-waveoscilloscope,rise
sion that normally free-running or Q-switched wide- times of the order of a few tenths of a nanosecond can be
bandwidth lasers (such as ruby, Nd:glass, or YAG: Nd) obtained. In addition, the use of picosecondlaserpulses
consist of picosecond pulses on the basis of data obtained has also made possible the direct measurement of popula-
solely by the two-photon absorption-fluorescence measure- tion decay time in the picosecond region for the first time
ment technique [60]-[62]. He haspointed out thata unique [39], [40]. Picosecond pulses have also been used for prob-
assignment can be made only if the contrast ratio is known ing molecular orientation dynamics in liquids by the use
with great accuracy, and has shown that the fluorescence of pulse times on the scaleof the molecular orientation
record from N modes of equal-amplitude, equal-frequency time [88].
separation and random phase relationships yields a maxi- The experimental method used to measure such short
mum contrast ratioof 1.5. For the same case, but with fixed decay times is as follows. Absorption of a very intense light
phase relationships between the modes, a maximum con- pulse of ultrashort duration by a dilute sample of dye
trast ratioof 3 is obtained with this experimental technique. molecules will prepare all the molecules in the light path in
This second case is the mode-locked case. Similar results an excited electronic state. The suisequent decayof the
have also been obtained by Klauder et al. [83] and their population of this state is then probed by a “probe” pulse
measurement of the contrastratio of the mode-locked of light whichcan be delayed continuously to arrive before,
pulses from glass: Nd3+ yielded 2 instead of3. This dis- during, or after the intense “preparing” pulse. The trans-
crepancybetween theory and experiment could arise be- mission of the sample for the probe beam is proportional at
cause of the simplified plane wave approximation utilized each instance to the concentration of the ground state dye
in the analysis instead of the statistical approach utilized by molecules. As the sample is prepared in its excited state,
Ducuing and Bloembergen in their study of fluctuation in this transmission willrise abruptly. As the ground state
nonlinear optical processes[84]. repopulates through decay from the excited state, transmis-
The simplicity ofthe two-photon absorption-fluorescence sion will decrease again. Table I1 illustrates the lifetimes
technique is one of its major advantages if a measurement of Eastman 9740 [39] and 9860 [40] saturable dyes directly
of the contrast ratio does not have to be taken. If the pulse- measured by this technique. These two lifetimes are im-
width measurement obtained by this technique is greater portant to the use of these dyes in the generation of pico-
than the inverseof the spectra bandwidth of the pulse, second pulses. For example,theselifetimes indicate that
then the measured width is the actual pulse width and a pulse repetition rates as high as lo-” s are possible with
contrast ratio measurement does not have to be performed. these dyes.
If a periodic pulse train consisting of subnanosecond laser Mack investigated the fluorescencelifetime of several
pulses are displayed on an oscilloscope, then the measure- dyes commonly used for Q-switching ruby lasers and for
ment of the pulse width obtained from the two-photon liquid lasers pumped by ruby second harmonic [41]. The
absorption-fluorescence can be considered to be the true experimental arrangement was particularly simple. The dye
pulse.widthwith a highdegree of certainty without the cell was irradiated by the second harmonic ruby radiation
contrast ratio measurement. for the measurements of Table 111, and directly by the ruby
 DEMARIA ET A L . : PICOSECOND LASER PULSES 19

TABLE I1
PICOSECOND ABSORPTION
DECAY TIME
OF EASTMAN
9740 AND 9860
SATURABLE ABSORBERS
AS MEASURED AND [40]
BY REFERENCES [39]
where r,~is the viscosity, T the temperature, anda the effective
Deduced Direct molecular radius. Values for r of 10- to s have
Measurement
Measurement
Reference
Dye been measured in liquid by various indirect techniques.
(PS) (PSI When an optical pulse duration AT is greater or equal to r,
9740 the anisotropic molecules can rotate in responseto the
[40 9860 6 1 9 pulse, but when AT < r, the molecules cannot respond and
the self-focusing threshold will increase. Shapiro, Giord-
TABLE I11 maine, and Wechtclearly showed in their experiments a
FLUORESCENT
NANOSECOND LIFETIME
OF SOME LIQUID
LASER definite tendencytoward less Raman scattering when r
DYFS AS MEASUREDBY REFERENCE [41]
became longer than AT [88].
FluorescentFluorescent B. Light AmpliJication in Saturable Absorbers
LifetimeDye Peak Solvent
( 4 (ns) When power at a pump frequency and one or two side-
Acridine red ethanol
2.4 5800
bands is incident on a saturablepower absorber, the power
Acridine yellow ethanol
5.2 5050 absorbed by the saturable absorber will vary at the differ-
Sodium fluorescein ethanol6.8 5270 ence frequency. If the magnitude of the absorption is non-
Rhodamine 6G ethanol
5.5 5550 linear with respect to the power absorbed, the resulting
Rhodamine 6G water 5.5 5550
Acridone ethanol 11.5 4370 changes in absorption will in turn alter the amount of power
Anthracene methanol 4Ooo 4.5 transmitted through the saturable absorber.If the recovery
time of the saturable absorber is faster than the beat fre-
TABLE IV quency between the pump and the sidebands, the absorp-
NANOSECOND OF SOME Q-SWITCHING RUBY
FLUORFSCENTLIFETIME tion will vary at the difference frequency and thus pump
LASERDYB AS MEASUREDBY REFERENCE [41]
energy into the sidebands. In principle, this process can
Absorption FluorescentFluorescent producemoreoutput signal at the sidebandsthan was
Solvent Peak Lifetime
Peak present originally. This technique was first utilized to am-
(A) ( 4 (ns) plify millimeter waves [89].
~~~

CAP ethanol 6700 7550 10.1 Mack [90] has found that a saturable absorber driven
CAP methanol 6710 7550 10.3 into saturationby an intense light pulse from a mode-locked
CAP chloronaphthalene 6970 7380 8.0 ruby laser can amplify a weakerlight pulse simultaneously
VP nitrobenzene 6980 - 4.1
VP chloronaphthalene 7010 ~
4.2 incident upon the medium. Energygains as high as 20 times
cc methanol 7060
0.5 7400 have been observed with a path length of only 1 cm in the
CAP = chloro-aluminum phthalocyanine
saturableabsorbing dye solution. Previous theoretical
VP = vanadyl phthalocyanine [91] and experimental[92] investigations of hole burning in
CC = cryptocyanine. saturable absorbers have indicated that the weak wave at-
tenuation ismerely reduced as compared to that for the
radiation for the measurementsshown in Table IV. A strong wave, i.e., the hole in the absorptionspectrum
copper sulfate solution andasharp cutoff interference reaches only to the zero attenuation point. The fact that
filter were placed in front of the photodetector to isolate amplification of the weak wave can be achieved means that
the detector from the scattered ruby and second harmonic the hole in the absorptionspectrum actually penetrates
radiation, respectively. The overall responsetime of the through to the negative absorption region. It has since been
detection systemwas0.45 ns. The decaytimeswere de- found that in addition to the gain from the saturable ab-
termined by fitting an exponential to each of the decaying sorption process described above, gain from a new stimu-
portions in the oscillographs. The values should be accurate lated thermal scattering process was also taking placein
to f20 percent. The fluorescent lifetimes of Rhodamine 6G these experiments [93]. This new stimulated thermal scatter-
is particularly interesting since it is so useful in two-photon ing process has also been observed in linear absorbers [93].
absorption-fluorescence measurement of picosecond pulses An intuitive understanding of the optical amplification
as discussed in Section 111. In addition, ithas been pumped process in saturable absorbers canbe gained with the aid of
by the second harmonic picosecondpulses of Nd lasers to Fig. 24. If two colinear optical waves of different frequency
generate ultrashort pulses at its fluorescent wavelength in and of substantially different intensity are incident on such
the yellow portion of the spectrum [86]. anabsorber,a small amplitudemodulation at the beat
It is now well known that the observed gain in excess of frequency will occur in the dye. If we assume that the in-
the calculated values arising in the stimulated Raman tensity is sufficientto drive the dye into the nonlinear region
effect canbeexplained by self-focusing of the pumping of its transmission transfer function and that the relaxation
laser light which arises primarily from the Kerr effects. The time of the dye is sufficiently fast to follow the modulation
response time of the Kerr effect is given bythe Debye rota- frequency, the nonlinear transmissivity will enhanceand
tion time r for molecules driven by an ac field : distort the modulation as shown in Fig. 24. The distortion
20 PROCEEDINGS OF THE IEEE, JANUARY 1969

t
70% HIGH
LEVEL
=12W LINES/mm
TRANSMISSION

Fig. 25. Schematicdiagram of twoplane blazed gratingswith 1200

10’ lines/mm used tocompresschirpedpicosecondpulsesemitted by a
IIAMSMISSION
lo2 I I I I I I I I -
dye mode-locked Nd3+:glass laser.
lo2 lo3 10‘ I@ 10’10’10’ 10’ 1olo roll

INCIDENT MlENSnY (W/CM2 )

Fig. 24. Simplified explanation of opticalgain in saturableabsorbers
by means of thetransmissioncharacteristics of cryptocyaninein
methanol.

indicates the generation of additional sidebands. The en-

hancement of the modulation indicates that the initial
sideband intensity has grown relative to the intense compo-
nent even though the overall intensity of the output beam is
less than the input beam.
C . Compression of Picosecond Pulses
The fact that ultrashort optical pulses generated by an - 4 4 . 4 ps t-
Nd: glasslaser are typically 4 to 10 x s longandhave Fig. 26. Microdensitometertraces of acompressedanduncompressed
a spectral width of the order of 100 A suggests that part of pulse taken with the two-photon absorption-fluorescence technique.
the observed spectral width is due to some kind of phase
increasing function of wavelength A. It is easy to see that
modulation of the carrier wave. One is led to such a conclu-
an inputpulse with a positive carrier frequency sweep willbe
sion from the fact that for such a bandwidth, one would
compressed by sucha grating pair since the long wave-
expect the pulse length to be an orderof magnitude smaller.
length components arrive at the input first andtakea
Treacyhas recently foundthataquadraticterm in the
longer time to traverse the system. It is also easy to see that
phase 4(t) exists in the picosecond pulses emitted by dye
the time delay between the two extreme frequencies of the
mode-locked Nd3+:glass lasers[21], [23]. Thisphase
frequency swept pulse is proportional to the separation of
variation &t)= wt + I/2gt2 corresponds to a linear sweep
the gratings.
of the carrier frequency d+/dt = o + fit, where the constant
Fig. 26 illustrates microdensitometer traces from photo-
/? is positive. This optical frequency sweep is analogous to
graphs of the uncompressed and compressed pulses taken
the frequency swept wave trains emitted by chirp radar.
with the two-photon absorption-fluorescence technique by
In chirp radar systems, the transmitted pulse is of rela-
Treacy. The wide pulse is the pulse prior t o enterinflbe
tively long durationAT,during which timethe instantaneous
gratings, and the narrow pulse is the pulse leaving the grat-
frequency is swept over some range A$ The return pulse is
ings. The pulse is compressed to approximately 1 ps. Later
passed through adispersive network providing differential
a
data have resulted in pulse widths down to4 x 10- l 3 s. It is
delay AT over the frequency rangeA$ As a result, the energy
not presently known whether the dye or the laser medium
at the beginning of the pulse is delayed so as to reach the
is primarily responsible for generating the chirp character-
end of the network at the same time as the energy at the
istic of Nd3+:glass laser pulses. Treacy has also used the
end of the pulse. The duration of the compressed pulse
narrow compressed pulse to optically sample the uncom-
(AT,)is of the order of AT,? Af or atime-bandwidth
pressed pulse for displaying the shape of its intensity en-
product of unity. The pulse normally emitted by a dye
velope [23].3
mode-locked Nd3+:glass laser has a Arb$? 10 to 20.
Giordmaine et al. [97] proposed using an ultrasonic
Fig. 25 shows the essential elements of the optical dis-
frequency translator[98], [99] for chirping the mode-locked
persion deviceused tocompress the picosecond laser
pulses from a He-Ne or an Nd3+ :YAG laser. The He-Ne
pulses. A pair of plane blazed gratings with 1200 lines per chirped laser would require 50 km of bromobenzene for the
mm arearranged with their face and rulings parallel so ~~

that light of any wavelength,- after diffracting- twice,will Itwasrecentlybrought to our attentionthat picosecondpulsesfrom
exit in a beam to the input beam. since different aNd:glass laserwerealsoutilized to samplethesubnanosecondpulses
emitted by a He-Ne laser by M. A. Duguay and J. W. Hansen, “Optical
components Of a beam are diffracted through sampling of subnanosecondlightpulses,” Appl. Phys. Lett., vol 13, pp.
different angles, the transit time is approximately a linearly 178-180, September 1968.
DEMARIA ET A L . : PICOSECOND LASER PULSES 21

dispersive element to compress the pulses. The 50 ps pulses solutions exhibit laser action whenexcitedwith a short-
from a Nd3+ :YAG laser would be compressed down to duration, high-intensity pump pulse [ 1031-[1061. Pump
0.4 ps after going through33 m of bromobenzene or 25 m of pulses from Q-switched lasers andfrom specially con-
nitrobenzene. structed flash tubes have beenemployed. The outputspectra
from such dye lasers are quite broad, extending insome
D. Generation of Ultrashort Acoustic Pulses
cases over a few hundred angstroms.
The transient surface heating of materials by repetitive, It is wellknown that aperiodic modulation of the gain of a
high-power, picosecondlaser pulses is capable of generating laser medium will lead to mode-locking if the frequency of
acoustic shocks of very short time duration [loo]. Ready the modulation is equal to, or a multiple of, the difference
[loll has calculated that the thermal gradients produced frequency between longitudinal modes of the laser, and if
by the absorption of a typical Q-switched ruby laser pulse the upper state laser lifetime is shorter than, or comparable
could be as high as lo6 deg/cm, with a temperature rate of to, the optical loop transient time of the laser cavity. If the
change as high as 10I6 deg/s. In a comprehensive treatment pumping signal for a dye laser consists of a mode-locked
of the problem of transient surface heating, Whte E1021 train of pulses, the gain of the laser will have a periodic
has shown thatthe conversion efficiency by whichthe sound variation with a period equal to the spacing between the
is produced varies linearly with the incident peak power pumping pulses. If the length of the dye laser cavity is equal
density and is inversely proportional tothe first powerof the to, or a submultiple of, the length of the cavity of the laser
sound frequency. It can therefore be expected that repeti- producing the pump pulses, then the mode-locking condi-
tive, high-peak-power, picosecondlaser pulses can generate tion will be satisfied and the output of the dyelaser will
acoustic shocks having harmonic componentswell into the consist of a series of pulses.
microwave region [ 1001, Initial experiments were performed with Rhodamine 6G
The outputof a pulse regenerative Nd : glass laser having and Rhodamine Bdyes [86]. In ethanol solution these dyes
a repetition rate of 200 MHz, pulse widths between have absorption peaks at 5260 A and 5500 A, respectively.
and lo-" s, and an average energy per picosecond pulse The peaks are sufficiently broad to allow efficient pumping
of approximately 1 mJ was used to irradiate a metal film with the 5300 8, second harmonic of the Nd laser. The ex-
deposited on the end of an LiNb03 bar. The entire pulse perimentalarrangement for laser-pumped,mode-locked
train contained from 100 to 150 individual pulses and lasted dye lasers is as follows. A 2.5cm diam by 5 cm long dye
about 0.4 to 0.6 p s . The thermal stressing caused by the cell was placedin a laser cavity having a 1 m radius of curva-
partial absorption of the laser pulse train by the metal film ture mirrors with a reflectivity in excess of90 percent be-
propagated acoustic shocks into the crystalline bar.The tween 5000 A and 6000 A. A KDP crystal was also placed
shocks traveled the length of the LiNb03 bar and were re- in the dye laser cavity. The pulse regenerative Nd pumping
peatedly reflected from the two ends of the bar, thus form- laser beam was propagated through one of the mirrors and
ing an echo pattern. Each time the shocks werereflected irradiated the KDP crystal, thereby producing the second
from the end of the bar oppositethe thin-film end, a voltage harmonic radiation required for pumping the dye. The
was generated by the piezoelectric properties of the LiNb03 fluorescencelifetime of the Rhodamine 6G is5.5 ns (see
crystal. This voltage was detected by a radioreceiver having Table 11). The Ndlaser pulseperiod was approximately 5 ns.
an IF bandwidthof 8 MHz. The receiverwas tuned to Laser action in the dyewasobservedwith its optical
multiples of the 200 MHz laser pulsing rate and the radio cavity length equal to 1, 4,and 3 times the length of the
pulse displayed on an oscilloscope. At room temperature, pumping laser cavity. The dyelaser output showsvery
several echoes were observed of the fifteenth harmonic at 3 smallfluorescence until the pumping signal reaches
GHz. As evidenced by the sharpness of the tuning of the threshold, at which point the dye laser breaks into oscilla-
radio receiver, the sound was strictly confined to the har- tion and produces a series of pulses at the repetition rate
monic frequencies of the laser pulse repetition frequency. determined by its cavity length. The laser output could be
It is felt that, besides providing a convenient method of varied from 5600 A to 5900 A by changing the composi-
producing discrete sets of sound frequencies well up into tion of the solutions. The wideuseof the stagger-tuning
the microwave region, this techniquedemonstrates the principle often utilized in the RF region can be utilized to
possibility of producingultrashort acoustic pulses with great advantage in obtaining spectral lines of enormous
ultrashort light pulse^.^ widths by the proper mixing of several organic dyes [107].
E. Variable Wauelength Picosecond Pulses F. Optical Rectijication Studies with Picosecond Pulses
The availability of ultrashort optical pulseswith any Optical rectification wasfirstobserved by Bass et al.
desired wavelength would greatly increase their applicabil- in the form of an induced dc voltage across a crystal of
ity to studies of nonlinear transient optical effects, spectros- potassium dihydrogen phosphate during the passage of an
copy, and lifetime measurements. One method of obtain- intense ruby laser pulse through the crystal [108]. Con-
ing an echo pattern. Each time the shocks werereflected siderable dficulty is normallyencountered inthe un-
lasers with the second harmonic ruby of Nd: glass lasers. ambiguous observation of the effect because of the short-
It has been reported that a large number of organic dye duration, low-voltage signals produced which, besides re-
quiring broad-band and thus insensitive electronic equip-
More recent work has resulted in the direction of 10 GHz signals ment, must be distinguished from electrical noise from the
in LiNbO, at room temperatures. laser and spurious pyroelectric and acoustic signals. The
22 PROCEEDINGS OF THE IEEE, JANUARY 1969

periodic pulsating characteristics ofpulse regenerative plasma plume, atoms in very high states of ionization can
lasersallow the useof highsensitivity and selectivityof be obtained. A spherically symmetricexpandingplasma
radio receivers in detecting the effect. The large signal-to- ball can be generated by laser beam irradiation of a single
noise ratio available with such RF detection enables one to solid particle. The plasma generated fromthe particle has a
identifyeasily (and thus eliminate) spurious acoustic sig- set of properties which makes it of particular interest for
nals by their repetitive echo characteristics. In addition, the study of the interaction of an expanding plasma with a
signals from relatively slow effects such as the pyroelectric magnetic field. Significant containment of the laser-
effect can be eliminated by detecting the microwave har- irradiated single-particle plasma with low-intensity mag-
monics of the rectification signals. netic confinement hasbeenachieved todateto indicate
Harmonically rich electrical pulses generated in crystals one use of laser-produced plasmas.
of LiNbO, and KDP have been observed up to approxi- Basov and Krokhin [120] calculated that laser powers in
mately 10 GHz [109]. The experimental arrangementwas as excessof lo9 W wereneeded to heat a laser-generated
follows. The output of an Nd pulse regenerative oscillator lithium deuteride plasmauptoatemperature at which
having a repetition rate of 275 MHz was directed through thermonuclearneutronemission may be observed. Un-
crystals of LiNbO, or KDP either along or perpendicular fortunately, optical damage tooptical components resulting
to the Z-axis of the crystals. A flat-ended coaxial probe from such large optical intensities made it difficult to carry
fixed against the surface normal to the Z-axis was used to out suchexperiments atthat time. It was subsequently
sense the induced fields present at the surface. The RF sig- noticed that the power damage threshold increased with
nals were measured to be approximately 2 mV and were decreased pulse duration. The possibility of reaching
detected by a superheterodyne receiver having a 20 MHz thermonuclear neutron emission temperatures withhigh-
bandwidth. intensity, ultrashort laser pulses presented itself when single
The observed signals were recorded with a signal-to-noise picosecond laser pulses became available [120]-[122]. The
ratio of up to 20 to 25 dB and followed the general features use of an Nd:glass single picosecond laser pulse generator in
of the envelope of the laser pulse train. Signals were re- conjunction with five cascaded 60 cm long amplifiers has
corded at each harmonicfrequencyfrom 275 MHz to resulted in the generation of 20 J of energy in’lO-” s [123].
9.076 GHz and were limited only by the equipment avail- Preliminary results of the use of these high-energy, single
able. The amplitude of the received signals followed the sin ultrashort pulsesin generating thermonuclearneutron
28 dependence as the angle 8 between the plane of polariza- emission from laser-heated lithum deuteride surfaces have
tion of the light and the x-axis of the KDP crystal was been reported [123]. A total number of four coincidence
rotated. neutron detections were reported in an experimental series
Thebroad-bandresponse of the optical rectification of 14shots. This numberwas calculated to be 20 times higher
cffects, coupled with greatly increased sensitivity obtained than the probability of obtaining accidental coincidence of a
with radio receivers and repetitive picosecond laser pulses, background pulse with the ultrashort laser pulse.
suggests the use of the effects for detecting picosecond light
pulses and for generating millimeterwave radiation. A
CONCLUSIONS
unique characteristic of such amicrowave generating system
would be its broad-band passive generating element, i.e., In approximately two years, a 5 order of magnitude jump
the nonlinear crystal, whose operating frequencycanbe has been made in our ability to generate high-peak-power,
easily changed and precisely specified by adjustment of the ultrashort laser
pulses, i.e., from s to s, and
a
laser cavity length. The upper frequency of the device is 3 order of magnitude jump in our ability to generate high
limited only by the harmonic content of the available peak powers, i.e., from lo9 W to 10” W. To keep up with
ultrashort laser pulses. these breakthroughs, researchers havedevised techniques
for measuring the time duration of ultrashort pulses which
G . Optically Generated Plasma Studies represent a 3 order of magnitude improvement over past
Using the focused radiationfromhigh-powered, Q- direct measurement capability, i.e., from lO-’Os to 10- l 3 s.
spoiled lasers, highdensity, high-temperature plasmas have Picosecond laserpulseshaveproved to be valuable as
been produced in gases [8], [110E[114], from solid surfaces scientific tools in controlled thermonuclear research, in the
[115]-[118], and from singlemicron-sizedsolid particles generation of acoustic shockwaves, in nonlinear optical
[119]. The resulting plasmas have been employed in funda- experiments, in the measurement of picosecond relaxation
mental studies of radiation-matter interactions, to excite times, and in the measurement of the response timeof
high-temperature gas reactions, to produce extremely thin molecular systems. It is obvious that researchers have not
vapor-deposited coatings, as spectral sources for micro- yet even broken the surface in finding scientificapplications
analysis, and to study lughly excited ions. The use of lasers for picosecond laser pulses. As of this writing, no break-
permits the generation of plasmas with a wide range of through has been made in the practical application of these
composition, density, ionization, andtemperatureprop- “bullets of light,” but the potential application of these
erties. Laser-generated gas breakdownplasmascan be short laser pulsesin information processing, rangmg,
produced with electron densities greater than l O I 9 cm-, topography mapping, and hgh-speed photography appears
andtemperatures exceeding 100eV. In the solid surface extremely promising.
DEMARIA ET AL.: PICOSECOND LASER PULSES 23

REFERENCES [30] S. E. HarrisandR.Targ.”FM oscillation of the He-Ne laser.”

Appl. Php. Lett.. vol. 5. pp. 202-204. November 1964.
[ I ] R. W. Hellwarth. ”Control of fluorescent pulsations.” in Adrances [31] M. H. Crowell. “Characteristics of mode-coupled lasers.” IEEE J .
in QuantumElectronics. J.R. Singer, Ed. New York: Columbia Quantum Electronics. vol. QE-1. pp. 12-10, April 1965.
University Press. 1961, pp. 334341. [32] A. J. DeMaria. “Ultrasonic-diffraction shutters for optical maser
[2] F. J. McClung and R. W. Hellwarth. “Giant optical pulses from oscillators.” J . Appl. Phys.. vol. 34. pp. 29842988. October 1963.
ruby,” J . Appl. Phys., vol. 33. pp. 828-829. March 1962. [33] L. C. Foster. M. D. Ewy. and C . B. Crumly. “Laser mode locking
[3] R. J. Collins and P. Kisliuk. “Control of population inversion in by anexternaldoppler cell,” Appl.Phys.Lett.. vol. 6. pp. 6 8 ,
pulsed optical masers by feedback modulation.” J. Appl. Phys., vol. January 1965.
33. pp. 2009-201 1. June 1962. [34] A. J. DeMaria and D. A. Stetser. “Laser pulse-shaping and mode-
[4] A. J. DeMaria. R. Gagosz. andG . Barnard. “Ultrasonic-refraction locking with acoustic waves,” Appl. Phys. Lett., vol. 7, pp. 71-73,
shutter for optical maser oscillators.” J . Appl. Phys.. vol. 34. pp. August 1965.
453456. March 1963. [35] G. A. Massey, M. K. Oshman, and R. Targ, “Generationof single-
[ 5 ] P. P. Sorokin. J. J. Luzzi, J. R. Lankard. and G. D. Pettit. “Ruby frequency light using the FM laser,” Appl. Phys. Lett., vol. 6. pp.
laser Q-switching elements using phthalocyanine molecules in solu- 1C-11. January 1965.
tion.” IBM J . Research and Derelop.. vol. 8. pp. 182-194. April [36] D. A. Stetser and A. J. DeMaria, “Optical spectra of ultrashort
1964. optical pulses generated by mode-locked glass:Nd lasers.” Appl.
[6] P. Kafalas. J. I. Masters.and E. M. E. Murray.“Photosensitive PIIJS. Letr.. vol. 9, pp. 118-120, August 1966.
liquid used as a nondestructive passive Q-switch in a ruby laser.” [37] H. W. MockerandR.J. Collins, “Modecompetitionand self-
J . Appl. Phxs.. VOI.35, pp. 1349-2350, August 1964. locking effects in a Q-switched ruby laser.” Appl. Phys. Lett., vol.
[7] L. M. Frantz and J. S. Nodvik. “Theory of pulse propagation in a 7, pp. 27C-273. November 1965.
laser amplifier,” J . Appl.Phxs.. vol. 34. pp.23462349. August [38] M.DiDomenico.Jr.,J.E. Geusic. H.M.Marcos,andR. G.
1963. Smith. “Generation of ultrashort optical pulses by mode-locking
[8] R. G. Meyerand, Jr.. and A. F. Haught, ”Gas breakdown at optical the YAG:Nd laser.” Appl. Phys. Lett.. vol. 8, pp. 18C-183, April
frequencies.” Phys.Rer..Lerr.. vol. 11, pp. 4 0 1 4 2 . November 1966.
1963. [39] J. W. Shelton andJ. A. Armstrong, “Measurementof the relaxation
[9] P. A. Franken and J. F. Ward, “Optical harmonics and nonlinear time of theEastman 9740 bleachable dye,” IEEE J . Quantum
phenomena.” Ret. Mod. Phys.. vol. 35, pp. 2>39. January 1963. Electronics(Correspondence), vol. QE-3. pp. 696497, December
[lo] N. Bloembergen. Nonlinear Optics. New York: W. A. Benjamin, 1967.
Inc.. 1965. [a] R. I. Scarlet. J. F. Figueria, and H. Mahr, “Direct measurement of
[I 1] R. Y. Chiao, E. Garmire, and G. H.Townes. “Raman and phonon picosecond lifetimes,” Lab. of AtomicandSolid-state Physics,
masers.” Quantum Electronics and Coherent Light. Proc. Internat’l Cornell University, Ithaca,N. Y.. Tech. Rept.24, May 1968.
School of‘ Physics, Enrico Fermi.Course X X A I , P. M. Miles and [41] M.E. Mack. “Measurement of nanosecond fluorescence decay
C. H. Towns, Eds. New York: Academic Press, 1964, pp. 326338. time.” J . Appl. Phys., vol. 39, pp. 248S2484, April 1968.
[I21 I. D. Abella. N. A. Kurnit. and S. R. Hartmann. “Photon echoes.” [42] V. I. Malyshev. A. S. Markin,and A.A. Sychev, “Mode self-
P/z.vs. Ref... vol. 141. pp. 391406, January 1966. synchronization in the giant p u l ~of a ruby laser with a broad
[I31 S. L. McCall and E.L. Hahn. “Self-induced transparency by pulsed spectrum,” Sotiet Physics-JETP Lett., vol.6. pp. 3435. 1967.
coherent light,” Phys. Ret. Lett.. vol. 18. pp. 908-911. May 1967. [43] I. K. Krasyuk, P. P. Pashkin. and A. M. Prokhorov, “Ring ruby
[I41 E. Garmire, R. Y. Chiao. and C. H. Townes. “Dynamics and char- laser forultrashort pulses.” SorietPhysics-JETPLett., vol. 7,
acteristics of the self-trapping of intense light beams,” Phys. Rev. pp. 89-91. February 1968.
Lett.. vol. 16. pp. 347-349, February 1966. [44] M.L.Spaethand W. R. Sooy. “Fluorescence and bleaching of
[15] J. A. Giordmaine and R. C. Miller. “Tunable coherent parametric dyes for a passive Q-switch laser,” J . Chem.Phys., vol. 48, pp.
oscillation inLiNbO,atoptical frequencies.” Phys.Rer.Lett., 2315-2319. 1968.
vol. 14, pp. 973-976. June 1965. [45] M.E.Mack,“Mode-lockingtherubylaser,” IEEE J. Qunntum
[I61 A. J. DeMaria, D. A. Stetser, and W. H. Glenn, Jr.. “Ultrashort Electronics (Correspondence), vol. QE-4, pp. 1015-1016, December
light pulses.” Science. vol. 156. pp. 1557-1568. June 1967. 1968.
[I71C.C.Cutler.“The regenerative pulse generator.“ Proc.IRE, [46] M. Hercher, “Single-mode operation of a Q-switched ruby laser,”
vol. 43, pp. 14CL148. February 1955. Appl. Phys. Lett.. vol. 7, pp. 39-41, July 1965.
[18] A. J. DeMaria, D. A. Stetser. and H. Heynau. “Self-mode-locking [47] H. A. Heynau and A. W. Penney. Jr.. “An application of mode-
of laser with saturable absorbers.” Appl. Ph,vs. Lett., vol. 8, pp. locked, laser generated picosecond pulses to the electrical measure-
174176, April 1966. ment art,” I€€€ Internar’l C o w . Rec.. pt. 7. vol. 15. pp. 8 6 8 6 ,
[19] M. DiDomenico. Jr., “Small-signal analysis of internal (couphng- 1967.
type) modulation of lasers,’’ J . Appl. Phys.. vol. 35. pp. 387C-3876, [48] R. Harrach andG. Kachen. “Pulse trains frommode-locked lasers,”
October 1964. J . Appl. Phys., vol. 39, pp. 2482-2483, April 1968.
[20] A. Yariv. “Internalmodulation in multimode laser oscillators.” [49] D. R. Herriott and H. H. Schulte, “Folded optical delay lines,”
J . Appl. Phys..vol. 36. pp. 388-391. February 1965. Appl. Opt., vol. 4. pp. 883-889, August 1965.
[21] E. B. Treacy. “Compression of picosecond light pulses.” Phys. [50] A. J. DeMaria, W. H. Glenn, Jr.. and D. A. Stetser. “An extended
Lett., vol. 28A. pp. 3435, October 1968. pathand digitally scanned optical pulse generator,” Appl.Opt.,
[22] A. J. DeMaria. “Mode-locking opens door to picosecond pulses,” VOI. 7, pp. 1405-1407. July 1968.
Electronics. vol. 41, pp. 112-122. September 16, 1968. [51] A. A. Vuylsteke. “Theory of laser regeneration switching,” J . Appl.
[23] E. B. Treacy.“Measurement of picosecond pulses substructure Phys., vol. 34. pp. 1615-1622. June 1963.
using compression techniques,” Appl. Phys. Lett.. to be published. [52] W. R. Hook. R. H. Dishington, and R. P. Hilberg. “Laser cavity
[24]M.Hercher.“Ananalysis of saturableabsorbers.” Appl. Opf., dumping using time variable reflection.” Appl. Phys., Lett.. vol. 9,
vol. 6. pp. 947-954, May 1967. pp. 125-127. August 1966.
[25] L. M. Frantz and J. S. Nodvik. “Theory of pulse propagation in a [53] A. W. Penney. Jr., and H. A. Heynau, “PTM single-pulse selection
laser amplifier,” J . Appl. Phys.,vol. 34. pp. 23462349. August1963. froma mode-locked Nd3+:glass laser using a bleachable dye,”
[26] R. Bellman, G. Birnbaum, and W. G. Wagner. “Transmission of Appl. Phys. Lett., vol. 9, pp. 257-258. October 1966.
monochromatic radiation in a two-level material.” J. Appl: Phys. [54] M. Michon, J. Ernest, and R. Auffret, “Pulsed transmission mode
vol. 34. pp. 78C-782, Apnl 1963. operation in the case of a mode-locking of the modes of a non-Q-
[27] L. E. Hargrove, R. L. Fork, and M.A. Pollack, “Locking of He-Ne spoiled ruby laser,” Phys. Lett.. vol. 21, pp. 514-515, June 1966.
laser modes induced by synchronous intracavity modulation,”Appl. (551 A. J. DeMaria. R. Gagosz, H. A. Heynau. A. W. Penney, Jr.. and
Phys. Lett.,vol. 5. pp. 4-5. July 1964. G . Wisner. “Generationandampldication of asubnanosecond
[28]A.J.DeMaria.C. M. Ferrar. and G. E. Danielson. Jr.. “Mode laser pulse.” J. Appl. Phyx.. vol. 38, pp. 2693-2695, May 1967.
locking ofa Nd3+doped glass laser,” Appl. Phys. Lett.. vol. 8, [56] N. G.Basov. P. G. Kryukov, V. S. Letokhov. and Yu. V. Senotskii,
pp. 22-24. January 1966. “Oscillation and amplification of ultrashort pulses of coherent
[29] T. Deutsch, “Mode-locking effects in an internally modulated ruby light,” Internat’lQuantumElectronicsConf: (Miami, Fla.. May
laser,” Appl. Phys. Letr..vol. 7, pp. 8C-82, August 1965. 1968), pp. 81 1-812.
24 T H E PROCEEDINGS O F IEEE, JANUARY 1969

[57] R. N. Lewis. E. A. Jung. G. L. Chapman. L. S. Van Loon, and T.A. [86] W. H. Glenn, M. J. Brienza, and A. J. DeMaria, “Mode-locking of
Romanowski,“Some high voltagepulsetechniques in use at an organic dye laser,” Appl. Phys. L e t t . , vol. 12, pp. 54-56, January
Argonne,” IEEETrans. Nuclear Science, vol.NS-13,pp. 84-88, 1968.
April 1966. [87] M. B. Fisher and R. T. McKenzie, paper presented at the Internat’l
[ 5 8 ] W. B. Davenport, Jr., andW. L. Root, An Introduction to the Theory Electron Devices Meeting, Washington, D. C., October 1967.
of Random Signals and Noise. New York: McGraw-Hill, 1958, pp. [88] S. L. Shapiro, J. A. Giordmaine, and K. W. Wecht, “Stimulated
87-99. RamanandBrillouinscatteringwithpicosecondlightpulses,”
[59] W. R. Sooy. “The natural selection ofmodes in a passive Q-switched Phys. Rec. Lett., vol. 19, pp. 109S1095, November 1967.
laser,” Appl. Phys. Lett.,vol. 7, pp. 3637, July 1965. [89] B. Senityky, G. Gould, and S. Culter, “Millimeter wave amplifica-
[ 6 0 ] M. A. Duguay, S. L. Shapiro, and P. M. Rentzepis, “Spontaneous tion by resonance saturation,” Phys. Rec., vol. 130, pp. 1 4 6 C 1 4 6 5 ,
appearance of picosecondpulses in rubyandNd:glass lasers,’’ May 15, 1963.
Phys. Ret.. L e t t . , vol. 19, pp. 10141016, October 1967. [90] M. E. Mack, “Light amplification in saturable absorbers,” Appl.
[61] S. L. Shapiro, M. A. Duguay, and L. B. Kreuzer, “PicosecoRd sub- Phvs. Lett., vol. 12, pp. 329-330, May 1968. Presented at the Inter-
structure of laserspikes,” Appl.Phys.Lett., vol.12, pp. 36-37, nat’l Quantum ElectronicsConf.,Miami,Fla.,May 17,1968,
January 1968. Paper 174-10.
[62]M. Bass and D. Woodward, “Observationofpicosecondpulses [91] S. E. Schwartz and T. Y. Tan, “Wave interaction in saturable ab-
from Nd:YAG lasers,” Appl.Phys.Lett., vol. 12, pp.27S277, sorbers,” Appl. Phys. Lett.,vol. 10, pp. 4 7 , January 1967.
April 1968. [92] B. H.Soffer and B.B. McFarland,“Frequency-lockinganddye
[63] H. P. Weber, “Comments on the pulse width measurement with spectral hole-burning in Q-spoiled lasers,”Appl. Phys. Lett.,vol. 8,
two-photon excitation of fluorescence,” Phys. Lett., vol. 27A, pp. pp. 1 6 1 6 9 , August 1966.
321-322, 1968. [93] M. E. Mack, to be published.
[63a] H. P. Weber and R. Dandliker, late paper presented at the Inter- [94] J. R. Klauder, A. C. Prince, S. Darlington, and W. J. Albersheim,
nat’l Quantum Electronics Conf., Miami, Fla., May 1968. “The theory and design of chirp radars,”Bell Sys. Tech. J., vol. 39,
[ 6 4 ] W. E. Lamb, Jr., “Theoryof an optical maser,” Phys. Rm.,vol. 134, p. 745,1960.
pp. A1429-AI450, June 1964. [95] J. R. Klauder, “The design of radar signals having both high range
[65] C. M. Ferrar, “Q-switching and mode-locking of a CF,I photolysis resolution and high velocity resolution,” Bell Sys. Tech. J., vol. 39,
laser,” Appl. Phys. Lett., vol. 12, pp. 381-383, June 1968. p. 809, 1960.
[ 6 6 ] H. Statz and C. L. Tong, “Phase-locking of modes in lasers,” J. [96] W. B. Mims, “The detection of chirped radar signals by means of
Appl. Phys., vol. 36, pp. 392S3927, December 1965. electron spin echoes,” Proc. IEEE, vol. 51, pp. 1127-1134, August
[67] P. W. Smith, “The self-pulsing laser oscillator,” IEEE J. Quantum 1963.
Electronics, vol. QE-3, pp. 627435, November 1967. [97] J. A. Giordmaine, M. A. Duguay, and J. W. Hansen, “Compres-
[68] T. Uchida andA. Ueki, “Self lockingofgaslasers,” IEEE 1. sion of optical pulses,” IEEE J. Quantum Electronics, vol. QE-4,
Quantum Electronics, vol. QE-3, pp. 17-30, January 1967. pp. 252-255, May 1968.
[69]A. G.Foxand P. W. Smith,“Mode-lockedlaser andthe 180“ [98] C. G. B. Garrett and M. A. Duguay, “Theory of the optical fre-
pulse,” Phys. Ret..L e t t . , vol. 18, pp. 826828, May 1967. quency translator,” Appl. Phys. Lett., vol. 9, p. 374, 1966.
[70] A. G. Fox, S. E. Schwarz, and P. W. Smith,“Use of neon as a [99] M. A. Duguay and J. W. Hansen, “Optical frequency shifting of
nonlinear absorber for mode-locking a He-Ne laser,” Appl. Phys. mode-locked laser beam,”IEEE J. Quantum Electronics, vol. Q E 4 ,
L e t t . , vol. 12, pp. 371-373, June 1968. pp. 47748 1, August 1968.
[71] H. Statz, G.A. DeMars, and C.L. Tong, “Self-locking of modes in [IO01 M. J. Brienza and A.J. DeMaria, “Laser-induced microwave sound
lasers,” J. Appl. Phys., vol. 38, pp. 2212-2222, April 1967. by surface heating,” Appl. Phys. Lett., vol. 11, pp. 44-46, July 1967.
[72] C. L. Tong and H. Statz, “Maximum-emission principle and phase- [ 101] J. F. Ready, “Effects due to absorptionof laser radiation,”J. Appl.
locking in multimode lasers,”J. Appl. Phys.,vol. 38, pp. 2963-2968, Phys., vol. 36, pp. 462468, February 1965.
June 1967. [IO21 R.M.White,“Generation of elastic waves by transientsurface
[73] H. P. Weber. “Method for pulse width measurement of ultrashort heating,” J. Appl. Phys., vol. 34, pp. 3559-3567, December 1963.
lightpulsesgenerated by phase-lockedlasersusingnonlinear [lo31M. Bass and T. F. Deutsch,“Broadbandlightamplificationin
optics,” J. Appl. Phys.,vol. 38, pp. 2231-2234, April 1967. organic dyes,” Appl. Phys. Lett., vol. 11, pp. 89-91, August 1967.
[74] J. A. Armstrong, “Measurement of picosecond laser pulse widths,” [lo41 B. H. Soffer and B. B. McFarland, “Continuously tunable, narrow-
Appl. Phys. Lett.,vol. 10, pp. 1 6 17, January 1967. band organic dye lasers,” Appl. Phys. Lett., vol. 10, pp. 2 6 2 6 8 ,
[75] W. H. Glenn and M. J. Brienza, “Time evolution of picosecond May 1967.
optical pulses,” Appl. Phys. Lett., vol. 10. pp. 221-223, April 1967. [IOS] M.L.Spaeth and D. P. Brotfeld,“Stimulatedemissionfrom
[76] M. Maier, W. Kaiser, and J. A. Giordmaine, “Intense light bursts polymethinedyes,” Appl. Phys. Lett.,vol. 9, pp. 17%181, September
in thestimulatedRamaneffect,” Phys. Rec. Lett., vol.17, pp. 1966.
1275-1277, December 1966. [lo61F. P.Schafer, W. Schmidt, and J. Volze, “Organicdyesolution
[77] J. A.Giordmaine, P. M.Rentzepis, S. L. Shapiro,andK. W. laser,” Appl. Phys. Lett., vol. 9, pp. 306-309, October 1965.
Wecht,“Two-photonexcitation of fluorescene by picosecond [lo71A.J.DeMariaandC.Miller,“Linear,wide-bandwidthoptical
light pulses,” Appl. Phys. Lett., vol. 11, pp. 216218, October 1967. amplifiers by staggered tuning,” J. Opt. Soc. Am., vol. 58, pp. 467-
(781 P. M. Rentzepis and M. A. Duguay, “Picosecond light pulse dis- 472, April 1968.
play using two different optical frequencies,”Appl. Phys. Lett.,vol. [lo81 M. Bass, P. A. Franklin, J. F. Ward, and G. Weinreich, “Optical
11, pp. 218-220, October 1967. rectification,” Phys. Ret.. Lett.,vol. 9, pp. 446448, December 1962.
[79] W. H.Glenn,Jr.,“Parametric amplificationofultrashortlaser [ I 0 9 1 M. J. Brienza, A. J. DeMaria, and W. H. Glenn, “Optical rectifica-
pulses,” Appl.Phys.Lett., vol.11, pp. 333-335, December 1967. tion of mode-locked laser pulses,” Phys. Lett., vol. 26A, pp. 390-
[SO] R. C. Miller, “Second harmonic generation with broadband optical 391, March 1968.
maser,” Phys. L e t t . , vol. 26A, pp. 177-178, January 1968. [110] A. J. Alcock and S. A. Ramsden, “Two wavelength interferometry
[81 ] S. L. Shapiro, “Second harmonic generation in LiNb03 by pico- of a laser-induced spark in air,” Appl. Phys. Lett., vol. 8, pp. 187-
second pulses,’’ Appl. Phys. Lett.,vol. 13, pp. 19-21, July 1968. 188, April 1966.
[82] J. Comly and E. Garmire, “Second harmonic generation from short [l 1 1] R. W. Minck and W. G. Rado, “Investigation of optical frequency
pulses,” Appl. Phys. Lett., vol. 12, pp. 7-9, January 1968. breakdownphenomena,” Proc.Physics of Quantum Electronics
[83] J. R. Klauder, M. A. Duguay, J.A. Giordmaine, and S. L. Shapiro, Conf., P. L. Kelley,B. Lax, and P. E. Tannenwold, Eds. New York:
“Correlation effectsin thedisplay of picosecondpulses by two- McGraw-Hill, 1966, pp. 527-537.
photon techniques,” Appl. Phys. Lett., vol. 13, pp. 174-176, Septem- [112] R. G. Meyerand, Jr., and A. F. Haught, “Optical energy absorption
ber 1968. andhighdensity plasmaproduction,” Phys. Rm. Lett., vol.13,
[84] J. Ducuing and N. Bloembergen, “Statistical fluctuations innon: pp. 7-9, July 1964.
linear optical processes,” Phys. Rec., vol.133, pp. A149SA1502, [113] A. J. Alcock, P. P. Pashinin, and S. A. Ramsden, “Temperature
March 1964. measurements of a laser spark from soft-X-ray emission,” Phys.
[ 8 5 ] P.Pringsheim, Fluorescence and Phosphorescence. New York: Ret.. Lett., vol. 17, pp. 528-530, September 1966.
Interscience, 1949, pp. 10-17. [114] M. M. Litvak and D. F. Edwards. “Spectroscopic studies of laser-
PROCEEDINGS OF
VOL.
THE IEEE. 57. NO. 1. JANUARY 1969 25

producedhydrogenplasma,” IEEE J. QuantumElecrronics, vol. by laserbeamirradiation of single solid particles.” Phys. Fluids,
QE-2, pp. 48-92. September 1966. vol. 9. pp. 2047-2052, 1966.
[I 151 J. F. Ready. “Effects due to absorption of laser radiation,”J . Appl. (1201 N . G. Basov and 0. N. Krokhn. Soriet Physics-JETP Letr.. vol.
Phys.. vol. 36. 1965. 46. pp. 171-172. 1964.
[ I 161 P. Langer. G. Tonon. F. Floux. and A . Ducauze. “Laser induced [I21 ] R. V. Ambartyumian. N. G. Basov. V. S. Zuev, P. G. Kriukov, and
emission of electrons, ions, and X rays from solid targets.” IEEE V. S. Letokhov. Soriet Physics-JETP Letr., vol. 4. pp. 19-21, July
J. Quantum Electronics, vol. QE-2, pp. 499-506. September 1966.
1966. [122] N. G. Basov, V. S. Zuev, P. G. Kriukov, V. S. Letokhov. Yu. V.
[117] Yu. V. Afanasyev, 0. N. Krokhin, and G. V. Skilzkov, “Evapora- Senatsky. and S. V. T.Chekalin. Soriet Physics-JETP Leu.. vol. 54,
tion and heating of a substance due to laser radiation,” IEEE J . pp. 3-5, 1968.
Quantum Electronics, vol. QE-2, pp. 483486, September 1966. [I231 N. G. Basov. P. G. Kriukov, S. D. Zakharov, Yu. V. Senatsky. and
[I 181 B. C. Fawcett, A. H. Gabriel,F. E. Irons,N. J. Peacock, and P. A. H. S. V. Tchekalin. “Experiments on the observationof neutron emis-
Saunders. “Extreme ultraviolet spectra from laser-produced plas- sion at a focus of high-power laser radiation on lithium deuteride
mas.” Proc. Phys. Soc.. vol. 88. pp. 1051-1059. 1966. surfaces.” IEEE J. Quantum Elec:ronics, vol. QE-4, pp. 864867,
[119] A. F. Haught and D.H. Polk, “High-temperature plasmas produced November 1958.

Physical Basis of Noncatastrophic Degradation

in GaAs Injection Lasers

Abstract-A study was madeof the gradual degradation in the output of damage of the facets [l], [2]. It has been shown that the
GaAs injection lasers in the course of operation at 300°K and below. The damage threshold is related to the optical flux density
degradation process was found to be a bulk, rather than a surface, effect rather than tothe current density in the junction [ 2 ] .
during which the near-fieldemissioa patterns decrease in uniformity. Except
in rare cases, thereis no external evidence of mechanical damage to the de-
Here we are concernedwith the gradualdegradation
vices. This has heen coofirmed by scanning electron microscopy observations process which occurs at power levels lower than those re-
of the facetsbefore and aftergradualdegradation.Softeningofthe I-V quired for catastrophic damage. The object is to establish
characteristicsoccurs in additiontoincreases in thethresholdcurrent the nature of the gradual degradation process in GaAs in-
density and decreases in both the stimulated and the spontaneous exterior jection lasers. While degradation in GaAs incoherent elec-
differentialefficiency. No evidencewasfoundforsignificantchangesin
either the junction impurity profile or the optical emission spectra. From a
troluminescent diodes has been investigated in detail [3],
detailed analysis of the change in the threshold currents and the efficiency, [4], laser degradation has received far less attention. In this
it is concludedthattheinternalquantumefficiency is graduallyreduced paper we will investigate the following questions.
during laser operation.This decrease is primarily attributedto the formation
of nmadiative recombination centersin the recombination region. In addi- 1) Is the degradation a surface effect which can be elimi-
tion, the optical loss is increased in some lasers. An evaluation of the factors nated by surface coatings such as oxides?
affecting the degradation rate indicates that isit a superlinear function of the
2 ) Is it connected with changes in the mechanical condi-
current density of operation. Furthermore, the degradation rate is strongly
in5uwced by theinitialjunctionquality-laserswhichinitiallyexhibit tion of the device, asin catastrophic degradation?
highlynonuniformemissionpatternsdegradefasterthanthosewhichare 3) Are fundamental material changes involved (for ex-
relatively uniform. ample, the formation of lattice defects which would
I. INTRODUCTION reduce the radiative efficiency in the energy range of
interest)?

1 T IS KNOWN thatthe optical power output of injection

lasers may permanently degrade under certain operat-
ing conditions. The power degradation given
level may be rapid (a few pulses) in a manner denoted as
at acurrent
4) How do the operating conditions affect the degrada-
tion rate?

It will be shown that the degradation process is due to

“catastrophic” or gradual over many millions of pulses. bulk effects involving fundamental material changes. The
The catastrophic modeof failure is the result of mechanical dominant factor is the gradual reduction in the internal
Manuscript received June 10, 1968; revised August 19,1968, and
radiative quantum efficiency. In addition, evidence exists
October 4, 1968.The research reported in this paper was jointly sponsored for anincrease in optical loss in the junction region in some
by the U. S. Army Electronics Command, Philadelphia offices, Philadel- lasers.
phia Pa., under Contract DAAB05-67-C-2709, and by the RCA Labora- After consideration of experimental details (Section II),
tories, Princeton, N. J.
The authors arewith the RCA Laboratories, David Sarnoff Research we describe in Section 111 the phenomenological changes
Center, Princeton, N. J. 08540 which accompany the degradation process. The factors