IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. QE-10,NO.

3, MARCH 1974 375

band wider than 1 THz. Above 100 atm and for inversions R. Ely and T. K. McCubbin, Jr., “The temperature dependence of
the self-broadened half-widthof the P-20 line in the 00” 1-10’0 band
greaterthan 1.05, the P- andR-branches merge into a Of CO,,” Appl. Opt., V O ~ .9, pp. 1230-1231, 1970.
single peak. The gain increase shown is due to the line C. Rossetti et ai.,“Spectroscopic moleculaire avec source laser. Mo-
overlap effect and is of the orderof a factor of 10 from the mentde
transition
de
vibration-rotation
pour la transition
00”1-10”0de CO,,” C. R . Acad. Sci., vol. 262B, pp. 1684-1686,
smallestpressure tothe highestconsidered.Enormous 1966.
energy output-to-volume ratios are possible ( 104J/l) for C. Cousin, C. Rossetti, and C. Meyer, “Optique moleculair spec-
favorable excitation conditions. New methods of excita- troscopic avec sourcelaser.Etudedestransitions de vibration-
tion, such as a traveling-wave tube exciter, are, of course,
+
rotation us - u,, u3 - 2uz et (u3 + u,) - (v, v,) de I’anhydride car-
bonique,” C. R. Acad. Sci., vol. 268B, pp. 1640-1643, June 1969.
necessary to obtain high powers for large devices. Small T. K. McCubbin, Jr., and T. R. Mooney, “A study of the strength
devices would be used for continuously tunable masteros- and width of lines in the 9.4 and 10.4 pm COz bands,” J . Quant.
Spectrosc.Radiat.Transfer, vol. 8, pp. 1255-1264, 1968.
cillators.2 A. D. Devir and V. P. Oppenheim, “Line width determinationin the
-
9.4and 10.4wm bandsofCOzusingaCO,laser,”Appl. - . Opt.,vol.8,pp.
. ~.

REFERENCES
[l] R. Breene, The Shift and Shape of Spectral Lines. New York:
Pergamon, 1961.
[2] C. Fenstermacher et ai., paper presented at the 24tb Annu. Gaseous
Electronics Conf., Gainesville, Fla., Oct. 1971.
[3] J. A. Beaulieu, “High peak power gas lasers,” Proc. IEEE, vol. 59,
pp. 667-674, Apr. 1971.
[4] D. E. Burch and D. A. Gryvnak, “Absorption of infrared radiant
energy by CO, and H,O, V, Absorption by CO, between 1100 and
1835 cm-’ (0.1-5.5 pm),” J. Opt. Soc. Amer., vol. 61, pp. 499-503,
1971.
a Additional examples of the gain curve may be obtained from the
authors.
2121-2123, i w .
W. H. Christiansen. G. J. Mullanev. and A. Herzberg. “Absorption
in CO, at 10.6 pm with rotationafline overlap,” A&. Phys. i e t t . ,
vol. 18, pp. 385-387, May 1971.
J.Fahrenfort,“Theinfraredabsorptionspectrum of compressed
CO, and C0,-containing gas mixtures,” Ph.D. dissertation, Univer-
siteit van Amsterdam, Amsterdam, The Netherlands, 1955.
T. Y.Chang and 0. R. Wood, “Optically pumped 33 atmosphere
CO, laser,” presented at the 1973 Conf. on Laser Engineering and
Application, Washington, D. C., May 30-June 1, 1973.
A. J. Alcock, K. Leopold, and M. C. Richardson, “A high pressure
CO, laser with UV photo-preionization,” presented at the1973Conf.
on Laser Engineering and Application, Washington, D. ,C., May
30-June 1, 1973.
N. G . Basov et al., “Electron beam ionization lasers,” Zh. Eksp.
Theor. Fiz., vol. 64, p. 108, 1973.

Laser-Induced Electric Breakdown in Solids

NICOLAASBLOEMBERGEN, FELLOW, IEEE

Abstract-A review is given of recent experimental results on laser-
induced electric breakdown in transparent optical solid materials. A fun-
damental breakdown threshold exists characteristic for each material. The
threshold is determined bythe same physical process as dc breakdown,
namely, avalanche ionization. The dependence of the threshold on laser
pulse duration and frequency is consistent with this process. The implication
of this breakdown mechanism for laser bulk and surface damage to optical
components is discussed. Italso determines physical properties of self-
focused filaments.
I. INTRODUCTION
T HE history of laser-inducedelectric
plasmas and for the propagation characteristics of high-
power laser beams through solids, liquids, and gases was
quickly recognized. The subject of electric breakdown in
transparent optical solids, including laser materials, win-
dows,andotheropticalcomponents,remained,until
recently, largely an empirical or engineeringscience.
Although a vast amount of theoretical and experimental
effort was expended in the economically and technically
important problem of
opticaldamage,
quantitative
reproducible breakdown thresholds with
unambiguous

1963 Maker et al. [l] reporteddamage t o transparent

breakdown
is almost as old as the history of lasers itself. Early in theoretical
the last two
interpretations
years. The
havebeen
situation
was
obtained only during
somewhat analogous
dielectrics and the production of a spark in air by focusing to the development of our understanding of the problem
apulsedruby laser beam.Theimportance of these of dc breakdown in electrical insulators. There, too, the
phenomena for theproduction of laser-induceddense field developedlargely by engineeringtrial anderror.
Basic quantitative understanding was not achieved until
reproducibleexperimentalresults
on
well-defined
Manuscript received September 17,1973.The&orkwassupportedin part
by the Joint Services ElectronicsProgram,NASA,andtheAdvanced materialswereobtained[2].Thedifficulties in dc
Research Projects Agency. This review is based on a lecture presented at thebreakdown experiments were manifold: the influence of
2nd International Meeting on Lasersand their Applications, Dresden, Ger-the shape of electrodes, the occurrence of space charges,
many, June 1973.
The authoris with the Division of Engineeringand AppliedPhysics, Har- the rise time of the voltage pulse, the effects of heating due
vard University, Cambridge, Mass. 021 38. to prebreakdown currents, the origin of the first few elec-
376 I E E E J O U R N A L OF QUANTUM ELECTRONICS, MARCH 1974

trons to start the avalanche, etc. The experiments by von thresholdfor self-focusing ratherthan a flux density
Hippel [3] on alkali-halide crystals, 35 years ago, provided threshold for
electric
breakdown. In quantitative
the first reliable breakdown thresholds, which were breakdown experiments self-focusing and even incipient
characteristic for each alkali-halide compound. The data deformation of the laserbeam duetothe intensity-
could be explainedbytheoriesofavalancheionization. dependent index of refraction must be avoided.
Among the various competing theoretical versions [2] of The theories for the initial stages of electric breakdown
avalanche ionization and breakdown in solids, the model by lighthave been extensively developed, especially for
of Seitz [4] appearsto be in closeagreement with ex- gases. Raizer has shown in a good early review article [8]
perimental observations. The few exceptional electrons in that the two fundamental processes in the initial stages of
the high-energy tail of the electron distribution are respon-plasmaformation in transparent gasesare multiphoton
sible for thebuildup. Seitz estimatesthat 30 to 40 ionization and cascade or avalanche ionization. The first
generations, corresponding to anamplificationfactor of electrons are formed by frequency-dependent tunneling.
about io8 to 1O'O in theelectrondensity is requiredfor Keldysh [9] hasshown thatthetunneling mechanism
electric breakdown. We shall postpone further discussion becomes equivalent to multiphoton ionizationin the high-
of this modelto Section 111, where it will be shown that the frequency limit.
same model is capable of explaining laser-induced electric Free electrons may gain energy in the alternating elec-
breakdown. tric field. In the low-frequency limit this process, in con-
Theexperimental difficulties in laserbreakdown in- junction with electronic collisions, leads to joule resistive
vestigations,' which often prevented the establishment of heating.Inthehigh-frequencylimitthesametype of
reproducible thresholds, were the following. process is described as inverse bremsstrahlung [8]. Either
1) Irreproducibility in thetemporalandspatial description will give the possibility of creating some elec-
behavior of the laserfield, caused by multimode operation trons with a sufficiently high energy that they may cause
of high-power lasers, often prevented the establishment of ionization in the next collision, creating a new generation
reproducible thresholds. This problem can be eliminated in the avalanche.
by careful transverse and longitudinal mode control. The same two types of processes occur in dense matter.
2) The influence of absorbing inclusions often The different arrangement of energy levels, in valence and
prevented the establishment of reproducible thresholds. conduction bands, as well as the much shorter collision
Linear absorption in high-intensity laser fields causes such time of electrons in solids [lo], [II], will lead to quan-
a high rate of energy deposition that the temperature of titative differences from the phenomenain gases. The mul-
the absorbing inclusion may rise by thousands of degrees tiphoton absorption process creating transitions from the
andcause localmelting andevaporation.Thethermal valence band to the conduction band has been rather per-
stresses may cause damage to the materialin which the in- sistently invoked as a nonlinear mechanism important for
clusion is imbedded. The heat conduction andmechanical laser damage. It should be noted, however, that in this
stress problem can be solved by classical methods [6]. The case rather abrupt changes in damage threshold should oc-
situation depends of course on the size of the absorbing in- cur as a functionof laser frequency, when this frequency is
clusion, on the laser pulse duration, andon whether thein- raised so that the forbidden gap may be crossed by three
clusion is optically thick orthin. In practice, sub- or by two photons, respectively. If multiphoton absorp-
microscopicabsorbingparticlesoften set thedamage tion processes were important, they should also be clearly
threshold for optical components. The occurrence of Pt observable before catastrophic breakdown occurs.
particles in Nd-glass laser rods is well documented. Such In
this
paper recent
conclusive
experimental and
inclusions can be eliminated. They are, of course, not a theoretical evidence will be presented that laser-induced
fundamental physical property of the material, and the electronavalancheionization is usually the mechanism
damage threshold of the ideally pure solid may be con- thatdeterminesthebreakdownthreshold in pure
siderably higher. In laser-induced breakdown experiments transparent crystalline or amorphous solids and liquids.
it is importantto eithereliminate or else discriminate The threshold is, however, very sensitive to physical and
against the effects of absorption in nominally transparent chemical imperfections. These havein practice determined
materials. the damage threshold in many experimental situations.
3) I n ideally transparent materials the phenomenon of
self-focusing often determines the damage threshold (see 11. EXPERIMENTS ON LASER-INDUCED
[7] for comprehensive references to
the
extensive AVALANCHE IONIZATION IN SOLIDS
literature). The characteristicfilamentary fossil damage An experimental arrangement which positively avoids
tracks in opticalglass components are awell-known exam- self-focusing, excludes multiphoton absorption, and dis-
ple. Although electric breakdown and avalanche ioniza-
tinguishes between inclusion and intrinsic lattice damage
tionmayindeedoccur in the self-focused region, the
has been devised by Yablonovitch [12]. He used a TEA
observedthreshold is in suchacaseacriticalpower
CO, laser putting out a diffraction-limited (TEM,,-mode)
pulse train.The pulse duration is about s, andthe
No attempt is made in this paper to compile a complete bibliography peakpower is less than 1 MW. This is twoordersof
on laser damage investigations. The reader who wishes to delve into the
history of thissubjectmayprofitably start by consultingthegeneral magnitude smallerthan the criticalpowerrequiredfor
references in [ 5 ] . self-focusing at 10.6-pmwavelength for NaCl and other
BLOEMBERGEN:LASER-INDUCEDELECTRICBREAKDOWN
alkali-halidecrystals.Theself-focusingthresholdpower
may be estimated from the known electronic nonlinearity
contributing to the intensity-dependent index of refrac-
tion. Self-focusing by electrostrictive effects can also be es-
timated in these short pulses. For a Gaussian beamVariablethe
critical power for self-focusing [ 131 PC maybe expressed in
+
terms of the intensity-dependent index n = no n21El by
the relations,
with
CX2
PC,= -
32~~12
Even if the power is not sufficient to form a self-focus,
there will be some beam distortion. The radiusof the focal
spot a, produced by an external lens for low-power beams
is altered to a smaller spot size according to [14]
a, = (1 - P/Pcr)1/zuo. (2)
Experimentally,thiscorrection factor waskeptsmaller
than 10 percent. It waschecked that achange of total
beam power with a concomitant change in a,, by varying
of the focal length of the externallens or the incident beam
in diameter,always yielded thesamethreshold field
strength for breakdown.
Multiphotonabsorption in theselargeband-gapin-
sulators would require about 100 quanta at 10.6 pm. The
threshold for this process becomes essentially the same as
for dctunnelingaccording tothe Keldyshtheory [9].
Breakdown occurred a t field strengths of about 2 X lo6
V/cm, an order of magnitude belowthoserequired for
tunneling.
The field strengthnecessary forbreakdown in the
relatively low-powerlaserpulses used was obtained by
focusing with anexternal lens. This lens mustbecare-
fully correctedforsphericalaberrations.TheGaussian
beamfromthefundamentallaser modethenleads to
a geometrically well-defined focal spot inside the crystal.
The focal spotdiameter was about 10 X x cm. The
simple geometry is schematized in Fig.1. It was found that
the nominal power flux density of about 10" W/cm2 was
sufficient to produce breakdown in NaCl by the 10.6-pm
pulses of 10-T-s duration.
The occurrenceofbreakdownmaybeestablished by
one or more of the following observations.
1) There may be a sudden decrease in the transmitted
laserpower, as light is absorbed by the denseplasma
created by the breakdown in the focal region. An illustra-
tion of this phenomenon is shown in Fig. 6. The
transmitted intensity should drop to a very low level for
intrinsic breakdown, as essentially the whole beam is in-
tercepted. By contrast, inclusion damagemayintercept
only a fraction of the incident power.
2) The creation of a hotplasma in thebreakdown
region may be established by any plasma diagnostic tool,
377
L
Y
Laser
Attenuator Lens Sample
Fig. 1. Schematic experimental arrangement to determine the intrinsic
breakdown threshold. L : The laser operated in a single TEM,, mode.
It is @switched or mode locked to give pulse durations between
and 10-12s. Its peak power output is well below the critical power for
self-focusing. N , : Rotatablepolarizing prism. N,: Fixedpolarizing
prism. The prism N, transmits light polarized parallel to the laser field
output. When N , is rotated through an angle 8, the transmitted inten-
sity is proportional tocos40 . L:This lens is well corrected for spherical
aberration. F: Geometry of focal spot is well known. A typical size
would be 10 X (X = 10.6 pm, 1.06 pm, or 0.69 pm).
Fig. 2. Bulk intrinsic damage bubble in NaCl with no inclusions. 10.6-
pm focused radiation was incident from the left. Straight lines repre-
sent cleavage. (After Yablonovitch [12].)
forexample, the scatteringfrom asecondlaser probe
pulse. The observation of the emission of hot quasi-black-
body radiation from the breakdownregion is, of course, a
very simple form of evidence. It is not very specific and
does not discriminate easily against damage from absorb-
ing inclusions.
3) The microscopic examination of the fossile damage
track, after the sample has been exposed to thelaser pulse,
is only applicable in solids. This method is also capable of
distinguishing between intrinsic and inclusion damage.
The former should, of course, start at the geometrical
focal spot where the flux density is highest. The region of
breakdownshouldthenextendbackwardtowardthe
direction of the incident laser beam, as the laser power
behind the focus is decreased by absorption. The typical
geometry of the damaged region resulting from intrinsic
breakdown in the front part of the local volume is shown
in Fig. 2. In Fig. 3, where an alkali-halide crystal with a
higher density of inclusions was used, both fundamental
damage (funnel shape) and inclusion damage (spherical
spots)may be seen. It is clear that the damage by in-
clusions well in front of the focus occurs at a much lower
flux density.While thedamagefrom inclusions has a
threshold, which varies from sample to sample andis even
variable with the characteristics of each particular inclu-
sion inside the same sample, the funnel-shaped inclusions
at the focal spot defined by theexternal lens geometry
have a reproducible threshold characteristic for each alkali
halide. For a given compound the threshold field strength
is reproducible toabout 10 percent, andthereis a
systematicvariationfordifferentcrystalsthroughthe
378 I E E E J O U R N A L OF QUANTUMELECTRONICS, MARCH 1974
1 mm

TABLE I
DIELECTRIC
STRENGTH
( I N UNITSOF loe V/CM) FOR. RMS LASERFIELD
STRENGTH AT X = 10.6 p~ AND AT DC"

NaI NaBr NaCl NaF

dc 2.40 I .50 0.83 0.69
10.6prn 1.95 0.91 0.79
KF KC1 KBr KI
dc 1.80 1 .oo 0.69 0.57
10.6 p m 2.40 1.39 0.94 0.72
RbF RbCl RbBr RbI
__
dc 0.83
. .. 0.5R 0.49
10.6pm 0.93 0.78 0.63
Fig. 3. (a) Inclusion damage along a focused laser beam track. (b) In-
a The 10.6-prn data are taken from Yablonovitch [12]; the dc data are
trinsic damage in the focal region; inclusiondamage only in front of the
focus.Out-of-focus damage tracks should beignored. Straight lines taken from von Hippel [3].
represent cleavage. (After Yablonovitch [ 121.)
TABLE I1
periodic system. In fact, Yablonovitch found that the rms ABSOLUTEDIELECTRIC BREAKDOWN THRESHOLDS IN NACL IN VOLTS
PERCENTIMETER
field strength for breakdown at 10.6 ym has not only the
same systematic variation with the periodic system as the Threshold
dc breakdown strength, shown in Table I, but the two are
equalwithin theuncertainties of the experiments. The
average ratio of 1.25 between the two quantities shown in
Table I is not significant and could bedue entirely to
systematic uncertainties in the geometry of the dc experi-
ment. T
These results made it desirable to extend the technique
to shorter wavelengths. The experimental arrangement of
Fig. 1 can be scaled down to smaller wavelengths. It is true
f
that the critical power for self-focusing scales as X2. The
total laser power used at higherfrequenciesmustbe
reduced correspondingly. The diameter of the laser beam
may, however, be chosen proportional to the wavelength,
and with an external lens of the samefocal length one may
readily focus again to a spot size 10 X. The power density
in the smaller focal spot will then be the sameas before. In I I I
NaF
this manner breakdown experiments have been carried out NaBr NoCl

with a diffraction-limited Nd-YAG laser [15] at 1.06 ym (a)

and a diffraction-limited ruby laser [16] at 0.69 pm. It was
found that the breakdown threshold in the alkali halides
was nearly independent of the light frequency, as shown
by the data for NaCl in Table 11.
Again the same systematic variation of the breakdown
threshold is observed at the near infrared and visible laser
frequencies as one moves through the periodic table of
elements among the alkali-halide compounds. The trend
relative to NaCl is shown in Fig. 4.Note that the power
breakdown threshold for RbIis nearly 10 times lower than
the corresponding threshold for NaCl. The reproducibility
of the measurement, involving the geometry and intensity
of thelaser pulses, is between 5 and 10 percent.Un-
mistakably an intrinsic propertyof the crystalline latticeis
observed in these experiments. The threshold appears to
be essentially independent of frequency in these wide-gap
insulators. If there is any systematic variation,it is toward
Fig. 4. RMS electric field strengths required forbreakdown in nine
a slightly higher threshold at thevisible (ruby) wavelength. alkalihalidesrelative to NaCI, at dc, 1.06 prn, and 0.69 wn. (After
The dependence of breakdown threshold on laser pulse Fradin and Bass [ 161.)
duration was also studied [ 171 at 1.06 ym.A mode-locked
BLOEMBERGEN: LASER-INDUCED ELECTRIC BREAKDOWN 379

RMS OPTICAL ELECTRIC FIELD (MV/crn)

1.0 10 100
I I I

Fig. 6.' Ruby laser pulsestransmittedthroughNaCl.A TEM,,-mode

ruby laser with total pulse energy of 0.3 mJ was focused with a 14-mm
focal length lens. (a) Damagepeak occurs at pulse maximumE,,,, = 1.
1 I
1x)
I
10 1
(b)Damage pulse occursbeforepulsemaximum, Edamage/Epeak =
0.896. (c) Damage peak occurs after pulse maximum, Edamage/Epeak =
D.C. ELECTRIC (MV/crn)
0.954. (d) Three consecutive pulses without damage. Peak maximum
was held constant E,,,, = 1 within 5 percent. (After F'radin et al. [15].)
Fig. 5. Variation of breakdown threshold at 1.06 wm with pulse dura-
tion (after Fradin et al. 1171). Experimental points are taken from 1151
and [17]. Drawncurvesaretakenfrom [I91 andare based ondc zero to unity in an interval of electric field values. The
breakdown data in samples with varying thickness.Uppercurve is breakdown threshold may then be defined as thatvalue for
valid if the hot electron mobility is independent of E,,, i.e., the drift which breakdownoccurs in 50 percent of the pulses.
velocity is proportional to Edc. Lowercurve is for a constant drift
velocity. (Consult [I71 and [I91 for a more detailed discussion.) Ioniza- Under the experimental conditions for which the alkali-
tion rate is related to the pulse duration by ( 5 ) . halide data were taken, the statistical spread is so narrow
that, within the limits of accuracy of the experiment, one
Nd-YAG laser was used to get pulses with aduration of 2 may speak of a sharp breakdown treshold.
X lo-" s in acontrolledgeometry. The experimental
points are shown in Fig. 5 , together with a semiempirical 111. INTERPRETATIONAND DISCUSSION
curve derived from dc data [19],2 which will be discussed OF THE RESULTS
in thenextsection.It is seen that the breakdown field The nearlyidenticalbehavior of the laser-induced
strength at 1.06 pm in NaCl changes from 2 X lo8 V/cm breakdownanddcbreakdownthreshold in the alkali
for t, = s to 2 X IO' V/cm at t, = lo-" s. The process halidessuggests that thesamephysicalmechanism is
is neitherstrictlypower flux density nor strictlyenergy operative [19]. The mechanismfor dc breakdown is
fluence density dependent. I t is known that pulses as short generallyaccepted to beavalancheionization.Whena
as IO-'' s can still producebreakdown,butno reliable sufficiently denseplasmahas been created,thejoule
quantitative threshold has been determined. heating losses in the strongly colliding
electron
gas
The rapid buildup of the plasma density is also evident become so large that internally localized melting and/or
from the sudden cutoff in the transmitted laser intensity, evaporation takes place. The surrounding matrix may be
as shown by the oscilloscope traces in Fig. 6. The traces subsequently damaged by shock waves induced by thermal
cannot be used to measure the buildup timeof the plasma stresses.
density, but they clearly reveal statistical fluctuations in A detaileddescription of theavalancheprocess by
the time required for breakdown. In the bottom trace the means of the Boltzmann transport equationwould involve
same peak power in the pulse produced no breakdown in a complete knowledge of all electron collisional processes
three consecutiveshots. Inthethreeupper traces the including the energy dependence of all atomicand
breakdown occurs always near the pulse maximum, but it molecularexcitationandionizationcrosssections.
clearly occurs sometimes slightly earlier or slightly later. Remarkably detailed understanding has been attained in
Bass and Barrett [ 181 have studied thestatistical character gaseous breakdown, e.g., of the atmosphere [20], but in
of breakdownin more detail. They show that there is not a condensedmatterthere is still considerableuncertainty
sharp threshold, but that a probability distribution canbe aboutthedetails ofelectrondistribution in very high
defined with a probability for breakdown increasing from fields. It is nevertheless possible to understand thegeneral
characteristics of thebreakdownphenomenon by the
In [19] the notation of q and a was inverted from the general usage. recognition [4] that the electron density N ( t ) will increase
380 IEEE JOURNAL OF QUANTUMEL.ECTRONICS,MARCH 1974

because a few electrons will be accelerated to an energy

larger than the band gap. These few electrons have a high
probability to
create
additional
carriers,but
the
probability per electron per unit timeto undergo an ioniz-
ing collision is small.
Seitz [4] cautions that the accuracyof the theoretical es-
timatesoftheavalanchethreshold is not high. His
theoreticalestimateforNaCl is about 0.3 X lo6 V/cm,
and he comments, “The difference between this and the
observedvalue of 1.5 X lo6 V/cmindicates

the breakdown threshold based on Frohlich’s theory are

generallyafactor
theap-
proximate character of the calculations’’[4]. Estimates for

3 to 10 higher than the experimental

T-[L- oc. def. pot. sc.

value [2]. This has led to thesuggestion that theavalanche

mayalsoproceed via theionization of deepimpurity
levels. Hellwarth [21] has examined the role of impurities
for light-induced breakdown in polar crystals. Sizable im-
purity densities, of about 10l8 ~ m - would ~ , be required,
but thepossibility of deep-lying impurity statesin this con-
centration cannot be entirely excluded.
It is beyond the scope of this paper to analyze in detail
the merits of various theoretical approaches to avalanche
ionization. The important point to be emphasized hereis
that the correct approach should give similarresults in
large band-gap materials for dc-induced breakdown and
for breakdown induced by laser beams with frequencies up
to the visible range. The recent experimental data further-
more suggest that an intrinsic breakdown mechanism, in-
sensitive to impurity content, can be identified.
The collision time for slow electrons near the bottom of
the conduction band is, of course, very sensitive to the
band structure and impurity content. In some semicon-
ductorsthis collision time 7 is on theorderofto
10-12 s. In ioniccrystals slow electronsarestrongly
coupled to longitudinalopticalphonons.The collision
time willbe shortand is knownfromtheobserved
mobility of photoelectrons in alkali halides, 7 3 X
s in NaC1. Seitz [4] has,however,emphasized
- thatthe
collision time for hot electrons, i.e., electrons with energies
between 3 and 10 eV, is short in any material. These elec-
trons collide mostly via the deformation potential interac-
tion, creating acoustical phonons with large momentum.
Theyhaveacollisiontime ontheorder of s. This
time may be understood as follows. In pure metals elec-
trons at the Fermi surface with energies between 2 and 5
eV have collision times of about s when the k vector
is nearthemiddle of the Brillouinzone, and the time
generallydecreases to 10-l6 s as the larger momentum
space available near the Brillouin zone boundary is ap-
proached. The deformation potential mechanism is not
sensitive to details of the band structure. Fig. 7, which is
based on the discussion of Seitz [4], shows the trend of the
inversecollisiontimeasafunction of electronenergy.
Electrons with momenta near the boundary of the first
Brillouin zone have similar shortcollision times in all con-
densed materials, including metals, ionic crystals, valence
semiconductors, molecular crystals, and liquids. This fact
agreeswith the short mean free path of about one in-
polor mode sc.
1
Fig. 7. Scatteringprobability(inversecollisiontime) as afunction of
electron energy. (After Seitz141.)In nonpolar materials the scatteringis
due to the deformation potential interactionwith acoustic modes. The
maximum occurs at the electron energy, where its momentum is at the
edge of the first Brillouinzone. In ioniccrystalsthe effect of the
polaron coupling with optical phonons mustbe added to the deforma-
tion potential scattering. This leads to a very short .t.ollision time also
for low-energy electrons in these materials.
teratomic distance for electrons with energies between 3
and 15eV in mostmaterials.Theconcept of collision
probability tends to lose its meaning for such short colli-
sion times when h7-’ becomes a significant fraction of the
electron energy. The concept is, however, helpful in un-
derstanding some of the salient experimental features of
avalanche breakdown.
While electrical conductivity varies over many ordersof
magnitude,the dc avalanchebreakdownthresholdsare
confined to a relatively narrow range forsuch diverse sub-
stances as ionic crystal,glass, molecular fluids, water, ice,
liquified noble gases, etc., provided the dc pulse is short
enough and the electrode configuration is chosen in such a
mannerthat ionic conduction,electrontrapping,and
space-charge phenomena are eliminated. The systematic
variation of the threshold characteristics depends largely
on the deformation potential, band gap, and phonon den-
sity of states. These lead to about one orderof magnitude
variation in thebreakdown threshold over the alkali-
halide series [ 3 ] .
The short collision times are also essential for the un-
derstanding of laser-induced breakdown. It is intuitively
clear that the breakdownmechanism and the threshold are
essentially the same as for the dc case, as long as at least
one collision occurs, on the average, during a light cycle,
i.e.,for UTeff < 1.
The rate of increase in the electron density N during
low-frequency breakdown is determined by
The first term on the right-handside represents the cascade
multiplication, where g(E) is the probability per unit time
BLOEMBERGEN:LASER-INDUCED ELECTRIC BREAKDOWN 38 1

for an electron to undergo an ionizing collision. It is an This dependence has been verified not only in gases, but
averageover the electronenergydistribution.This dis- alsofor dc breakdown in alkali-halidecrystals. The
tribution is, in turn, a function of the applied field E(t). thickness of the crystals must be chosen small enough for
This is nearly equivalent to theprobability for an electron losses during the buildup to be negligible. In this manner
to reach the ionization energy in the applied field. The se- an empiricalrelationbetween dcbreakdownthreshold
cond term represents the rate of electron production per and electrode spacing was established [22], given by the
unit volume due to frequency-dependent tunneling,which lower solid curve in Fig. 5. The equivalence between ( 5 )
becomes synonymous with multiphoton ionization in the and (6) is established by therelationship t , = d / U d r i f t ,
high-frequency limit. If no electrons are initially present, which leads to v ( E ) = a ( E ) U d r i f t ( E ) . If the driftvelocity for
N(t = 0) = 0, the first electrons must be created by this the hot electrons is assumed to be independent of E, the
mechanism. Under the conditions of most breakdown ex- same curve except for a change of units would determine
periments in solids carried out to this date, the second thethresholddependenceon laserpulse duration.The
term is usually small compared to the first term. data in Fig. 5 appear to be in closer agreement with the
The lasttermontheright-hand siderepresentsa assumption of a constant mobility. When u d r i f t is propor-
decrease in the number of carriers due to diffusion out of tionalto E, theupperdrawncurve in Fig. 5 maybe
the discharge region, due to recombination, due to trap- derivedfromthedcdata. It is possible,however, that
ping with exciton formation,or due toelectron capture at tunneling starts to play a role at the shortest pulse dura-
crystalline imperfections. It is an advantage of breakdown tion.
experiments with short laser pulses ( t p < lo-* s) that the An intriguing question is whether the initial number of
losses areoften negligible duringtheformation of the conduction electrons No could not vanish, and how the
breakdownplasma.Undertheseconditionstheplasma first electrons would be formed in that case. This question
densitybuilds up exponentially, and the solution of (3) is of paramount importance for the breakdown in gases at
becomes simply relatively low pressures. The breakdown of air by CO,
laser pulses appears to be initiated by submicroscopic ab-
sorbing particles in the aerosol [23]. Sometimes the first
electrons may be provided by photoelectric effects from
the wall or by multiphoton-ionization processes. In solids
Breakdown may be defined, somewhat arbitrarily, to oc- at room temperature it appears safe to assume that thein-
cur when N exceeds 10'" electron/cm3. At this density the itialcarrierdensity is at least lo8 electron/cm3, due to
energydeposition rate by absorption of laser light photoionization or thermal ionization of shallow donor
becomes so high that the temperature rise of the lattice is levels. Forultrapurecrystalsat low temperature, well
significant even duringpicosecond
a pulse.With an shielded from radiation and light, the creation of the in-
assumedinitialelectrondensity No ,.,
lo8 electron/cm3, itial carriers could be a separate problem.
caused,forexample,bythermalionizationofshallow Even for an initial density of IO8,the actual number of
traps, this corresponds to a multiplication factor of lolo. carriers in the focalvolume is quite small. Inthe ex-
This estimate corresponds well with the estimate of Seitz periments at X = 1.06 pm and 0.69 pm, the focal region
[4] of 30-40 generations in the avalanchebuildup,cor- had a cross section of about loeEcm2 and a length of 0.1
responding to a multiplication of 230 or 2'O. It should be cm.Inthe focalvolume of cm3there wouldinitially
emphasized that any uncertainty in this factor leads to a be only a small number of carriers. This is entirely in ac-
logarithmicallysmalleruncertainty in thebreakdown cord with the observed statistical fluctuations [18] in the
threshold. buildup of the breakdown. One must wait for the chance
that one of the available electrons reaches the energy re-
q(E)t, In N / N o = 18. ( 5 ) quired for further ionization, and so on. It is estimated
from thecollision time and meanfree path that an electron
In dc breakdown experiments, especially those in gases, it in the solid would diffuse over a distance considerablyless
is customary to describe the growth of the avalanche in than 1 pmin IO-" s. Loss ofcarriers by diffusionmay
space as the electrons drift from the cathode to anode. In therefore be ignored for laser pulses of lo-"+ duration or
the absence of losses the gain is then describedby the shorter. Recombination is also negligible before the
Townsend coefficient a, plasma reaches a high density, and trapping may also be
ignored for pure crystals on this time scale. The spatial
and temporal scale of the laser breakdown experiments is
dN
dz
= a(E)N.
such that many of the difficulties that have plagued dc
breakdownexperimentsareavoided.There is nojoule
In theabsence oflosses thebreakdown field strength heating by prebreakdown currents, no space-charge
becomes a function of the cathode-anode distance d ac- buildup, no electron loss rates which are difficult to deter-
cording to mine quantitatively, no electrode effects at the surface of
the specimen. The small bulk volumeinvolved in the laser
experiments is, of course, also essential in eliminating the
382 I E E E J O U R N A L OF QUANTUM ELECTRONICS, MARCH 1974

influence of inclusions. The chance for an inclusion to oc-

cur in the focal volume is small.
Itremains to examinein more detailthefrequency
dependence ofthe threshold and to compare the avalanche
mechanismmorequantitatively with the other intriqic
mechanism,thefrequency-dependenttunneling, or mul-
tiphoton ionization.
In gaseous breakdown widespread use has been made of
a similarity principle (see [24] and the references therein)
that describes the dependence of ionization rate 9 on elec- \

tric field strength, pressure p , and frequency w , ,

-
‘40P

h=G: hv
Fig. 8. Schematic frequencydependence of thebreakdown threshold
field. The avalancheionization curve will move upward with decreasing
where f is a characteristicfunctionforeachmolecular pulse duration.Thepoints T , ~ ~and
- ~ egBp are indicative for alkali
halides. The crossover is likely to occur.somewhere in the visible region
species and 7 is themomentumtransfer collisiontime. of the spectrum.
Thissimilarityrelationmay be derivedfrom the Boltz-
mann transport equation, on the assumption thatcolli- the
halide crystals in the low-frequency limit. The threshold
sion time r is independent of electron energy, andthe inci-
for this process would also depend only logarithmetically
dentfrequencydoesnotproduce significantionization
onthepulseduration.Furthermore,noobservable
from excited bound states. Yablonovitch 1241 has shown
statistical fluctuations would be expected for breakdown
that the observed breakdown threshold of 10.6 pm in the
by tunneling. The magnitude of the observed threshold
gasatatmosphericpressure,for which wr = 103, still
andthe observed dependenceon frequency and pulse
agrees remarkably well with this scaling principle.
duration, as well as the statistical fluctuations, are all con-
In solids the energy dependence of r is quite marked, as
sistent with the avalanche ionization mechanism under the
shown by the curves in Fig. 7. Nevertheless, onemay
experimental conditions used. One may, however, expect
postulate a kind of average collision time Teff for the elec-
that
frequency-dependent tunneling or multiphoton
trons in ionic crystals. This collision time is quite short
ionization would become dominant at high frequencies;
(- s) forboththermalelectrons (-0.1eV) and for
forexample,thoseexceedinghalfthebandgap. The
electronsapproachingtheionizationenergy.It is’
crossover in Fig. 8 is, of course, schematic. It would be of
somewhat longerat
intermediate energies. We may
great interest to extend the breakdown measurements in
therefore postulate that the ionization rate, least at in ionic
the alkali halides to higher frequencies.
solids, obeys a scaling law similar to those forgases at con-
It is also to be expected that the breakdown level for
stant pressure. This leads to a relationship 1191 between
subpicosecond pulses will be set by tunneling. The curves
thebreakdownthresholdatfrequency w andthecor-
of Fig. 5, if extrapolated to pulse duration of less than
responding dc breakdown threshold
10-l2 s, would predict avalanche thresholdlevels exceeding
2 X lo7 V/cm, corresponding to power flux densities of
over 10lz W/cm2. This is the regime in which tunneling
becomes important. Even though initially theelectron
It is assumedthatbothdcand laserexperimentsare
density would increase predominantly by this mechanism,
conducted in such a manner that loss mechanisms can be
which therefore would determine the threshold, it remains
ignored during the pulse duration or during the transit
possible for avalanche multiplicationto become dominant
time of the electron between the dc electrodes.
in the final stages of the plasma density buildup. This is
Experimentally, the threshold appears to remain about
evidentfrom (3), butnodetailedcalculationsforthis
the same from dc to X = 1 pm, and there is some indica-
regime of field strengths appear to be available.
tion of a small increasein threshold at the ruby wavelength
X = 0.69 pm. This would imply that
light, or that reff = 0.3 X
-
w ‘ T ~ ~1~for ruby
s in ioniccrystals.This
The transition from avalanche ionization to the tunnel-
ing mechanism is well established in semiconductor diode
physics [25]. As thedimensionofthebarrierdepletion
should be compared with the crude theoretical estimates
of T - s. Clearly,more precisetheoreticalestimates,
layerbecomes very small,thediodecharacteristicsare
determined by the Zener-Keldysh tunneling mechanism.
based on complex solutions of the Boltzmann transport For larger thicknesses, impact ionization occurs within the
equation over the entire range of electron energies, are depletion layer, and one has avalanche diodes.
required, as well as accurate experimental data on many
other materials, before a quantitative agreement between
Iv. LASER-INDUCED BREAKDOWN
AT SURFACES
theory and experiment can be obtained.
The qualitative trendof the breakdown thresholdversus It is agenerallyobserved phenomenonthatsurfaces
frequency is shown in Fig. X forthetwogenerally tend to have a lower threshold for damage than the bulk
recognizedintrinsicprocesses,avalancheionization and material. It is especially difficult to make coated surfaces
frequency-dependent tunneling. The latter process has a damage resistant. For these reasom it was a widely held
theoretical
threshold of a b o u t 3 X IO’ V/cm in alkali- belief that the mechanism for electric breakdown at sur-
BLOEMBERGEN:LASER-INDUCEDELECTRICBREAKDOWE
faces would be a distinct1.y different process from that in
thebulk.The recentevidencepresented in thissection
suggests, however, that the material at the surface breaks
down by avalancheionizationandadenseplasma
formed by the same mechanism as in the bulk.
is
Inamajorityofcasestheapparentlylowersurface
damage thresholds were due tochemical contamination of
thesurface(dust particles,greasespots).Chemically
cleanedsurfacescan be obtainedformostmaterials, if
necessary by argon ion beam bombardment and evapora-
tion in vacuumofsurfacelayers. The thresholdfor
breakdown can thus be raised significantly [26]. Even the
chemicallycleansurfacestend,however,toexhibit
breakdown thresholds that arelower than the correspond-
ing bulk values, typically by a factor of 2 to 5 in power.
This can be explained by the presence of physical im-
perfectionsatthesurfaceintheformofgrooves,
scratches,cracks, or pores [27]. Typicaldimensions of
these are smaller than a wavelength, since the surfaces of
the components have been prepared and selected for op-
tical quality. For such small dimensions the electric field
configuration in the neighborhood of these physical im-
perfections,illustrated in Fig.9,maybecalculated by
means of electrostatics. There will be a concentration of
the lines of force near the edges. Electric breakdown will
be initiated at these locations. If a crack is mathematically
represented by adisk-shapedellipsoidalvoid, with the
lines of forces in the bulk normal to the disk, the electric
field will be concentrated near the edge by a factor E = n2.
Consequently, the breakdown threshold will be reached at
that location for nominal bulk power flux densities which
are lower by afactor c z = n4. In this manner a surface
threshold that is lower by a factor of 4 than the bulk can
be explained in materials such as glass, quartz, etc. The
correctness of this explanation was proven in a series of
experiments on superpolished samples of quartz andglass
[28]. The nominal damage threshold at the surface was in-
creasedandcouldbe madeequaltothe bulk damage
threshold by extremelyfinepolishing. One need not
eliminatecracksandscratches with criticaldimensions
smaller thanaboutcm. Sinceelectrostaticcontinuum
theory is still applicable in solids for dimensions down to
about 30 nm, there would still be a concentration of the
linesofforce. The region of increased field strength is,
however, so limited thattheavalanchecannot fully
develop. Because the diffusion time of the carriers out of
the edge region is shorter than the duration of the laser
pulse,the breakdownthreshold is increased, compen-
sating the decrease of a factor of n4 due to field concen-
tration. The arguments of the preceding section regarding
the threshold dependenceon electrode spacing thusyield a
criterion as to how well thesurface polishing must be
carried out.
In optical coatings there is a larger probability for the
simultaneouspresence of the chemicalinclusionsand
physical voids. It should be emphasized that the various
factors to lower the damage threshold are cumulative. If
thedamage is initiated by absorbinginclusions, rather
than avalancheionization, the presence of cracks will
again lower the damage threshold because the concentra-
383
L
incident loser beam
Fig. 9. Representativegeometries for electric field enhancement by
physical imperfections. Typical dimensions could be r = 0.1 pm, c =
0.02 ym, and a = 0.5 pm (from [26]).
tion of field lines and absorbinginclusions are likely to oc-
cur in the same spot.
Consider the case of a reflective coating in front of a
metallic mirror to improve the reflectivity. There will be a
standing wave pattern in the coating. At3 X in front of the
metalsurfacethere will be an electric field amplitude
which is twice the nominal value of the incident traveling
wave. If there are voids in the coating at this distance, the
power breakdown threshold will be lowered by a total fac-
tor of 4 n4. The damage threshold may be lowered by an
additional factor, if absorbing inclusions are also present.
The simultaneous presence of concentrations of electric
field lines, inclusions, and stress fields all tend to lower the
damagethreshold of surfacesandcoatings.Chemically
and physically clean surfaces and coatings should have the
same threshold as the bulk materials, except for standing
wave patterns set up by Fresnel reflections [29]. Chemical
or physical imperfections with at least one linear dimen-
sion smaller than cm areprobably not
harmful.
Neither absorption by specific surface states nor absorp-
tion by atomicimpurities lowersthe damagethreshold
significantly. The power absorbed in centers of atomic or
molecular size is .quicklydissipated.Physicalimperfec-
tions such as dislocations donotappearto lower the
threshold appreciably. The role of absorbing inclusions in
coatings and the relationship between their size and dis-
tribution with the laser pulse duration and focal spot area
has been experimentally demonstrated in some recent
papers [30].
v. AVALANCHE
IONIZATIONIN
SELF-FOCUSED
FILAMENTS
The onset of avalanche ionization at field strengths of
about lo7 V/cm for picosecond light pulses has been ex-
[ 171.
perimentally established in the absence of self-focusing
Withthisunderstanding we may turn back to the self-
focusingproblemandobserve thatthemaximum field
384 IEEE JOURNAL OF QUANTUM ELECTRONICS,
MARCH 1974

strength, or power flux density, in the filament cannot exceed TABLE 111
this breakdown threshold [ 191.Occasionally power fluxden- CHARACTERISTICS OF SELF-TRAPPED FILAMENTS~
sities for filaments in condensed matter of 10Is or even I O I 4 Peak Electric
W/cm2 havebeen mentioned in the literature. Such flux den- Diameter Energy Powerb Field"
Filament (PI (erg) (kW) (IO'V/cm)
sities are unphysical, exceptin a plasma whereall the valence
electrons have been stripped off the ions, or in low-density Nitrobenzene CS2 5.9 f 1.3 1.3
130f 0.6 2.0 f 0 . 6
5.8 f 0.8 0.9 f 900.4 1.6 f 0.4
gases during pulse durations shorter than thetimeofflight of 7.9CCI, f 0.2 1.2 f 0.8 120 1.3 f 0 . 4
electrons between molecules. As a rule thumb, of avalanche Toluene
Ether, ethanol,
10.1 f 1.0 0.6 f 0.2 60 0.7 f 0 . 1

ionization limits the power density to 5 X 10" W/cm2 fora toluene

(23°C) 7.6 f 1.4 1.2 f120
1.5
0.6 f 0.4
10-ps pulse, and tunneling would limit it to about 2 X 1 0 I 2 Ether, ethanol,
toluene (- 140'C) 8.9 f 1.3 3.8 f 1.9 380 2.0 f 0.5
W/cmz. Itis conceivable that in certain materialsand under Ether, ethanol,
toluene(-165"C) 17.7 f 1.7 13.1 f 10.2
1310 2.0 f 1.0
some experimental conditions otherphysical mechanisms,
a Liquid temperature is 23°C unless otherwise noted (from [32]).
such as stimulated Raman scattering E311 or hindered orien- Obtainedfromthe preceding columnassumingapulseduration
tational orderingof molecules [32], may limit the degree of At = 1 ps.
Obtainedfromthe peak intensity of themicrodensitometer trace,
self-focusing at a lowerflux density in the filament. It should cnEP2At/16~ erg/cm*, assuming' At = 1 ps. This, therefore, is not the
be stressed, however,that thefilaments occurin all kindsof local field but the field of the transmitted beam.
materials.Theyhavebeenobserved in the absence of
stimulated Raman scattering and crystals, in glasses, and li-
quefied monatomic gases where molecular reorientation tinguish between a dissipative and a nondissipative way in
effects should beminimal[33]. In most substances the sizeof which theplasmacan limit the filamentdiameter.The
the filaments appears tolie between 3 and 20 pm. Further- former is associated with the imaginary part of the plasma
more, estimatesof the maximumfield strength alwayslie in dielectric constant. The absorption in the plasma limits the
the neighborhoodof lo7 V/cm.In Table I11 someearly data laser power in the self-focused region. These points were
of Brewer and Lee [34] are reproduced. In particular, it also emphasized by Hellwarth [2 11. The plasma formation
should be noted that the mixture of ether, ethanol, and also causes adecrease in the real part oftheindex of
toluene is extremely viscous, nearly glass likeat thelowest refraction. The small positive contribution of a bound
temperature of - 165°C. Although molecular reorientation valence electron to the index is replaced by thelarger
is suppressed and the filamentary size changes, the peak elec- negative contribution of a quasi-free electron. Both effects
tric field remains at a constant value near the threshold for may be expressed in terms of a strongly nonlinear (ex-
electric breakdown. The durationof this high intensity in a ponential) contribution to the complex index of refraction,
fixed volume elementof matter in the filament always lies in
the picosecond domain, typically between and 3 X
10-l2 s. If the initial pulse is long, the filamentis really the
track of two moving focal spots.sFrom the spot size and the
velocity of these spotsthe duration ofthe high-intensity field Theexponential field dependence is contained in the
may be estimated. If theinitial pulseis shorterthan the domi- cascade buildup of the avalanche. In the preceding sec-
nant mechanism for the intensity-dependent index of refrac- tions we have demonstrated that the avalanche process is
tion, the self-focusing may proceed by transient molecular sufficiently fast, i.e., T e f f , sufficiently short, and the plasma

reorientation. In thatcase a quasi-stationary funnel-shaped densitymaybuild upto electrondensities exceeding

radial profile is developed. The high-intensity region at the lO1*/cm3 in a fraction of a picosecond.
trailing edge again lasts about 1 ps [35].3 The radial variation of the plasma density is also ex-
It is thus possible to establish a lower limit for the fila- tremely rapid through the radial variation of I E(r)l. It may
ment radius urnin.The minimum power contained in the be estimated from (10) that the real part of the index is
filament is equal to the critical power necessary for self- about - for aplasma density of 10'"/cmS; the absorp-
focusing PC.The maximum supportable field strength is tion depth in the plasma is about 1 cm for a wavelength of
the breakdown threshold fora picosecond.pulse [ 191. One 1 pm and Teff
radius may be stabilized
-
lo-''. It is estimated that the filamentary
undertheseconditions [19],
thus finds
because the effect of an infinitesimaldecrease in radius
produces comparable changes, but with opposite sign, in
n21El and (An),,. The situation is schematically displayed
in Fig. 10.
where PC is given by (la) and (Ib), and where a is the The observationthatthe filamentary fossile damage
radius at which the intensity has dropped to l/e of its on- tracks in glass have a typical diameter of 1 pm, whereas
axisvaluefor an assumedradialGaussian intensity the light is trapped in a cylinder with a diameter five times
profile. larger, can also be explained. The electron density is only
In many instances the observed filament size is close to high in the immediate vicinity of the axis of the filament.
this lower limit, suggesting that avalanche ionizationis in- Only here the rate of absorption is high enough to cause
deedthe limitingphysicalmechanism. One may dis- local melting of the glass. It is estimated that an electron
density exceeding 10Ig cm-s is necessary to cause melting
See the extensive literature quoted in [7]. during a picosecond interval.
BLOEMBERGEN:LASER-INDUCEDELECTRICBREAKDOWN 385

\ Light Intensity
Electron

0
Virtual Change in An
diameter
Fig. 11. Picosecondself-focusinginamediumwithafastelectronic
Kerr effect. Two short moving focal spots are created. The schematic
temporalbehavior at a fixed volumeelement is shown for three
physical quantities (from [35]). Intensity: dashed line. Plasma density:
dotted line. Change in index of refraction: solid line.
Fig. IO. Profile of light intensity and plasma density in a light filament.
The electrondensitychanges very rapidlywithsmallradialpertur-
ful picosecond pulses at 1.06-pm or 0.53-km wavelength
bations.
pass through such diverse materials as water, liquid argon,
glass, or transparent ionic crystals. These short ultraviolet
Theplasmaabsorptionalsoprovidesaconvenient
pulses are
already
beingusedto
investigate fast
mechanism for a cutoff in computer calculations of the
photochemical reactions. The frequency shifts areso large
transient self-focusing by reorientation in molecular fluids.
and the pertinent time durations so short that linear dis-
Shimizu [35] introduced, without physical justification, a
persion will haveaprofound effect ontheobserved
mathematicalabsorptiontermproportionalto IEl l‘ in
phenomenon.
his calculations of filament formation. It is again the ex-
The plasmadensitydecaysrelativelyslowly dueto
ponentialabsorptiondependenceinherent in avalanche
recombination and electron trapping after the self-focused
ionization, rather than a high-order multiphoton process,
spothaspassed.Thisimpliesthatunfocused or in-
that provides a physical mechanism for such a cutoff. The
completely focused light arriving after a self-focused spot
computedsolutions of theself-focusingequations by
may undergo defocusing. The plasma will cause light to
Shimizushowthatsuchamechanismindeedhasthe
leak away from filamentary regions. Such effects have also
desired limiting effect on the filament diameter.is clearly
It
been widely observed, as the term “leaky light waveguide”
of interest to extend such numerical calculations to the
indicates. It is even possible that avalanche ionization is
case where a partial differential equation to describe the
responsible for the onset of four-wave parametric down
spatial and temporal variations of the plasma density, and
conversion. The plasma would generate the light at the
its contribution to the index of refraction, is added to the
new frequencies and in new directions [36]. Due to linear
system of equations governing the temporal and spatial
dispersion the newly formed light waves could mix with
variations of the laserfield and the intensity-dependentin-
otherlaserlight. Theparametric mixing process,with
dex of refraction.
preference for characteristic phase matching conditions,
Therapidtemporalvariation of theplasmadensity
wouldbeprimed by theplasma,and itwouldnotbe
gives rise to an enormous negative rate of change of the
necessary to build it up by a stimulated gain process from
refractiveindex.Whenanintenseincidentpicosecond
the very weak spontaneous emission noise.
pulse is self-focused by the fast electronic Kerreffect, two
moving focal spots arise each with an estimated duration
VI. CONCLUSION
of about 10-13 s, as shown in Fig. 11. The present author
suggested [36] that the phenomenon of superbroadening Intransparentdielectrics,freefromabsorbingin-
[7], [31] is caused by the self-phase modulationof the light clusions,avalancheionizationhasbeenidentifiedasa
traveling along the axisin near synchronism with the mov- crucial mechanismfor
electric
breakdown. This
ing focal spots. A large anti-Stokes shift results from the mechanism is consistent with the observed dependence of
rapidly increasing plasma density. Although for the very breakdownthresholdonpulsedurationandfrequency,
short periods of peak intensity encountered in these focal and with its statistical character. The threshold field for
spots the electron density is likely to grow at first by the breakdown for wide-gap insulators is about the same as
tunneling mechanism, the avalanche term will eventually the dc threshold for frequencies as high as the ruby laser
prevail on the right-hand sideof (3), as the plasma density frequency and is acharacteristicproperty of thepure
has inereased sufficiently.Most of the anti-Stokes broaden- material. There is a (small) systematic tendency for the
ing takes place in a trailing fraction of the subpicosecond breakdown threshold to increase as the frequency is raised
pulse [36]. The creation of extremely short light pulses by into the visible.Typicalvaluesfor the threshold in the
this mechanism is a distinct possibility. alkalihalidesareabout 10’ V/cmforpulses of 10-8-s
The observed anti-Stokes continuum hasbeen observed duration and lo7 V/cm for pulse durations of lo-” s.
to stretch into the ultraviolet beyond 0.3 Fm, when power- For shorter pulses, or for high light frequencies which
386 I E E E JOURNAL OF QUANTUM ELECTRONICS,MARCH 1974

are a significant fraction of the band gap, the frequency- solid dielectrics” (in English), Sou. Phys.-JETP, vol. 30, pp.
6 16-62 1, 1970.
dependent tunnelingmechanismshouldbecome the in- [15] D. W. Fradin, E. Yablonovitch, and M. Bass, “Confirmation of an
itiating mechanism for breakdown. It is of interest to ex- electron avalanche causing laser-induced bulk damage at 1.06 bm,”
tend the breakdown measurements to shorter pulses and Appl.Opt., vol. 12, pp. 700-709, 1973.
[I61 D. W. Fradin andM. Bass, “Electron avalanche breakdown induced
higher frequencies. by ruby laser light,” Appl. Phys. Lett., vol. 22, pp. 206-208, 1973.
It is important to realize that practical damage [I71 D. W. Fradin, N. Bloembergen, and J. P. Letellier, “Dependenceof
thresholds in large sample volumes will usually be deter- laser-induced breakdown field strength on pulse duration,” Appl.
Phys. Left., vol. 22, pp. 635-637, 1973.
mined by the presence of absorbing inclusions or by the [18] M. Bass and H. H. Barrett, “Laser-induced damage probability at
formation ofself-focusedfilaments. Thefundamental 1.06 bm and 0.69 l m , ” Appl. Opt., vol. 12, pp. 690-699, 1973; also
breakdown mechanisms of
avalanche
ionization, or “Avalanchebreakdownandtheprobabilisticnature of laser-
induceddamage,” IEEEJ. Quantum Electron., vol. QE-8,pp.
frequency-dependent tunneling,
set,
however,
fun-
a 338-343, Mar. 1972; also M. Bass and D. W. Fradin, “Surface and
damental limit to the maximum power flux density that a bulk laser-damagestatisticsandtheidentification of intrinsic
dense medium can sustain without being converted to a breakdown processes,” IEEE J. Quantum Electron., vol. QE-9, pp.
890-896, Sept. 1973.
dense hot plasma. Thesemechanisms, therefore, also put a [19] E. Yablonovitch and N. Bloembergen, “Avalanche ionization and
limitingminimumdiameter on self-focusedfilaments. thelimitingdiameter of filaments induced by light pulses in
Avalancheionizationproducesa complexindex non- transparent media,” Phys. Rev. Lett., vol. 29, pp. 907-910, 1972.
[20] N. M. Kroll and K. M. Watson, “Theoretical study of ionization of
linearitywhichdependsexponentiallyonthe field air by intense laser pulses,’’ Phys. Reu. A , vol. A5, pp. 1883-1905,
strength. It is important to incorporate the temporal and 1972.
spatial variations of this nonlinearity fully in calculations [21] R. W. Hellwarth. “Role of ohoto-electrons in oDtical damage.” in - ;

Damage in LaserMaterials: NBS Special Pubi. 341, pp. 67-71,

on self-focusing,self-frequencybroadening,and other 1970.
physical properties of light filaments. A. A. Vorob’ev, G. A. Vorob’ev, and L. T. Murashko, “Anisotropy
Surfaces and thin-film coatings should have fundamen- of the electric strength of rocksalt,” Sou. Phys.-Solid State, vol. 4,
p. 1441, 1963 (transl. from Fiz. Tuerd. Tela., vol. 4, p. 1967,1962).
tal damage thresholds comparable to the bulk thresholds R. T. Brown and D. C. Smith, “Laserinduced gas breakdown in the
of the same materials. In practice, the damage threshold presence of preionization,” Appl. Phys. Lett., vol. 22, pp. 245-247,
will often be significantly lower, even if chemicalin- 1973; also P. J. Berger and D. C. Smith, “Gas breakdown in the
laser as the limitation of pulsed high-pressure CO, lasers,” Appl.
clusions are eliminated. The concentrationof electric field Phys. Lett., vol. 21, pp. 167-170,1972.
linesnearmicropores andcracks is responsibleforthe E. Yablonovitch, “Similarity
principles
for
laser-induced
lowering of the nominal power threshold. breakdown in gases,” Appl. Phys. Lett., vol. 23, pp. 121-122, 1973,
and references quoted therein.
E. M. Conwell, “High field transport in semi-conductors,’’ i n Solid
State Physics (Suppl. 9), F. Seitz, D. Turnbull, and H. Ehrenreich,
REFERENCES Eds. New York:Academic, 1967.
C. R. Giuliano, “Laser induced damage in transparent dielectrics:
[I] P. D. Maker, R. W. Terhune, and C. M.Savage, “Optical third har- Ionbeam polishing as ameans of increasing surfacedamage
monic generation,” in Proc. 3rd Int. ConJ Quantum Electronics, P. thresholds,” Appl. Phys. Lett., vol. 21, pp. 39-41, 1972; also “Laser-
GrivetandN. Bloembergen, Ed. Paris: Dunod, 1964. pp. 1559-1576. induced damage in transparent dielectrics: The relationshipbetween
J. J. O’Dwyer, The Theory of Dielectric Breakdown in Solids. New surface damage and surface plasmas,” IEEE J. Quantum Electron.,
York: Oxford, 1964. vol. QE-8, pp. 749-754, Sept. 1972.
A. von Hippel, “Electric breakdown of solid and liquid insulators,” N. Bloembergen, “Roles of cracks, pores, and absorbing inclusions
J. Appl. Phys., vol. 8, pp. 815-832,1937; also “Electronic conduc- on laser induced damage threshold at surfacesof transparent dielec-
tion in insulating crystals undervery high field strength,” Phys. Reu., trics,” Appl.Opt., vol. 12, pp. 661-664,1973.
vol. 54, pp. 1096-1102,1938. D. W. Fradin and M. Bass, “Comparison of laser-induced surface
F. Seitz, “On the theoryof electron multiplicationin crystals,” Phys. and bulk damage,” Appl. Phys. Lett., vol. 22, pp. 157-159, 1973.
Rev., vol. 76, pp. 1376-1393, 1949. N. L. Boling, M. D. Crisp, and G. Dub&, “Laser induced surface
A. J. Glass and A. H. Guenther, “Laser induced damage of optical damage,” Appl. Opt., vol. 12, pp. 650-660, 1973.
elements-A status report,” Appl. Opt., vol. 12, pp. 637-649, 1973. Proc. 1973 Laser Damage Con$, Boulder, Colo.; also D. Milam, R.
R.W.HopperandD.R.Uhlmann,“Mechanism of inclusion A. Bradbury, and M. Bass, “The role of inclusions and linear ab-
damage in laser glass,” J. Appl. Phys., vol. 41, pp. 4023-4037, 1970. sorption in laser damage todielectric mirrors,” to bepublished; also
S. A. Akhmanov, R. V. Khokhlov, and A . P. Sukhorukov, “Self- L. G . de Shazer, B. E. Newman, and K. M. Leung, “The role of
focusing, self-defocusing and self-modulation of laser beams,’’ in coating defects inlaser induced damagetothin films,” to be
LaserHandbook, vol. 2, F.T. Arecchi and E. 0. Schulz- published.
DuBois,Ed.Amsterdam,TheNetherlands:NorthHolland, 1972, 0. Rahn and M. Maier, “Explanation of limiting diameters of the
ch. E3, pp. 1151-1224. self-focusing of light,” Phys. Reu. Lett., vol. 29, pp. 558-564, 1972.
Yu. P. Raizer, “Breakdown and heatingof gases under the influence A. H. P’iekara and T. K. Gustavson, “The interaction of vibrational
of a laser beam,” Sou. Phys.-Usp., vol. 8, pp. 650-673, 1966 (transl. enhancement and theorientational Kerr effects in liquids,” Opt. Com-
from Usp. Fiz. Nauk, vol. 87, p. 29, 1965). mun., vol. 7, pp. 197-200, 1973.
L. V. Keldysh,“Concerningthetheory of impactionization in R. R. Alfano and S. L. Shapiro, “Emission in the region 4000 to
semiconductors,” Sou. Phys.-JETP, vol. 21, pp. 1135-1144,1965 7000 A via four-photon coupling in glass,” Phys. Rev. Lett., vol. 24,
(transl.from Zh.Eksp.Teor.Fiz., vol. 47, p. 1945, 1964). pp. 584-587, 1970; also “Observation of self-phase modulation and
A.Wasserman,“A mechanism fordamagein solids by intense small-scale incrystalsand glasses,” Phys. Rev. Lett., vol. 24, pp.
light,” Appl. Phys. Lett., vol. 10, pp. 132-133, 1967. 592-594, 1970; also “Direct distortion of electronic clouds of rare-
G. M. Zverev, T.N. Mikhailova, V. A. Pashkov,and N. M. gas atoms in intense electric fields,” Phys. Rev. Lett., vol. 24, pp.
Solov’eva, “Mechanisms of destruction of ruby and leucosapphire 1217-1220, 1970.
crystals by powerful laser
radiation”
(in
English), Sou. R. G. Brewer and C. H. Lee, “Self-trapping with picosecond light
Phys.-JETP, vol:26, pp. 1053-1057, 1968. pulses,” Phys. Reu. Lett., vol. 21, pp. 267-270,1968.
E.
Yablonovitch, “Optical dielectric strength of alkali-halide F. Shimizu, “Numerical calculation of self-focusing and trapping of
crystals obtained by laser-induced breakdown,” Appl. Phys. Lett., a short light pulse in Kerr liquids,” IBM J. Res. Develop., vol. 17,
VOI.19, pp. 495-497, 1971. pp, 286-298,1973, and the references quoted therein.
E. L. Dawesand J. H . Marburger,“Computerstudiesin self- N. Bloembergen, “The influence of electron plasma formation on
focusing,” Phys. Reu., vol. 179, pp. 862-868, 1969. super-broadening in light filaments,” Opt.Commun., vol. 8, pp.
G . M. Zverev and V. A . Pashkov, “Self-focusing of laser radiation in 285-288, 1973.