Approved for public release; further dissemination unlimited

UCRL-ID-135590

Producing KDP and DKDPCrystals for the NIF Laser

A. K. Burnham, H. F. Robey, N. P. Zaitseva, J. J. DeYoreo, R. A. Hawley-Fedder, M. Runkel, M. Yan, M.Staggs, S. A. Couture, R. L. Combs, R. C. Montesanti, P.J. Wegner, and L. J. Atherton

September 2, 1999

LawrenceLivermoreNationalLaboratory

U.S. Department of Energy

DISCLAIMER This document was prepared as an account of work sponsored by an agency of the United StatesGovernment. Neither the United States Government nor the University of California nor any of theiremployees, makes any warranty, express or implied, or assumes any legal liability or responsibility forthe accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, orrepresents that its use would not infringe privately owned rights. Reference herein to any specificcommercial product, process, or service by trade name, trademark, manufacturer, or otherwise, does notnecessarily constitute or imply its endorsement, recommendation, or favoring by the United StatesGovernment or the University of California. The views and opinions of authors expressed herein do notnecessarily state or reflect those of the United States Government or the University of California, andshall not be used for advertising or product endorsement purposes. Work performed under the auspices of the U. S. Department of Energy by the University of CaliforniaLawrence Livermore National Laboratory under Contract W-7405-Eng-48.

This report has been reproduced directly from the best available copy.

Available to DOE and DOE contractors from the

Office of Scientific and Technical Information P.O. Box 62, Oak Ridge, TN 37831

Prices available from (423) 576-8401 http://apollo.osti.gov/bridge/

Available to the public from the

National Technical Information Service U.S. Department of Commerce

5285 Port Royal Rd., Springfield, VA 22161 http://www.ntis.gov/

OR

Lawrence Livermore National Laboratory

Technical Information Department’s Digital Library http://www.llnl.gov/tid/Library.html

1

The cost and physics requirements ofthe NIF have established two impor-

tant roles for potassium dihydrogen phos-phate (KDP) crystals.

1. To extract more laser energy per unitof flashlamp light and laser glass, theNIF has adopted a multipass archi-tecture as shown in Figure 1. Light isinjected in the transport spatial filter,first traverses the power amplifiers,and then is directed to main ampli-fiers, where it makes four passes

before being redirected through thepower amplifiers towards the target.To enable the multipass of the mainamplifiers, a KDP-containing Pockelscell rotates the polarization of thebeam to make it either transmitthrough or reflect off a polarizer heldat Brewster’s angle within the mainlaser cavity. If transmitted, the lightreflects off a mirror and makesanother pass through the cavity. Ifreflected, it proceeds through thepower amplifier to the target.

PRODUCING KDP AND DKDP CRYSTALSFOR THE NIF LASER

A. K. Burnham R. A. Hawley-Fedder R. L. Combs

H. F. Robey M. Runkel R. C. Montesanti

N. P. Zaitseva M. Yan P. J. Wegner

J. J. De Yoreo M. Staggs L. J. Atherton

S. A. Couture

Deformable

mirrorCavity spatial

filter lenses

Mainamplifier

Poweramplifier

Pockels cell

window (2)Elbow

mirror

Cavity

mirror

Vacuum

window

Diagnostic

beam splitter

Transport spatial

filter lenses

Polarizer

Focus lens

Debris shield

Frequency

converterDiffractive optics

plates (2)

Switchyard/target area mirrors

Pockels cells Doublers

Triplers

FIGURE 1. Schematic ofthe NIF laser showingthe location of the KDPand DKDP components.Also shown is the orien-tation of the conversioncrystals within theboules.(40-00-0299-0434pb01)

the original seed crystal as the pyramidfaces grow. Unfortunately, this pyramidalgrowth is very slow, and it takes about two years to grow a crystal to NIF size. Toprovide more programmatic flexibility andreduce costs in the long run, we havedeveloped an alternative technology com-monly called rapid growth. Through acombination of higher temperatures andhigher supersaturation of the growth solution, a NIF-size boule can be grown in 1 to 2 months from a small “point” seed.However, growing boules of adequate sizeis not sufficient. Care must be taken to pre-vent inclusions of growth solution andincorporation of atomically substitutedimpurities in the prism growth. Otherissues important for meeting transmittedwavefront quality, absorption, and laserdamage criteria must be addressed also.

During the past year, we made substan-tial progress towards bringing the rapidgrowth technology to the stage needed tosupply most of the KDP and some of theDKDP needed for the NIF. This articlereviews the technical hurdles that wereovercome during this period and outlinessome of the issues yet to be fully resolved.It also covers some of the efforts to trans-fer this technology to two potential ven-dors for NIF crystals—CCI and Inrad—aswell as the development of crystal finish-ing technology.

Crystal GrowthFor several years, the rapid growth pro-

cess has reproducibly grown crystals ofhigh visual quality up to about 20 cm inlinear dimensions in tanks containing lessthan 100 L of solution. These systemsdemonstrated many of the important con-ditions needed to grow crystals of NIFquality, such as solution purity and waysto achieve adequate mass transfer at thegrowing crystal surface. Over the pastyear, this understanding has been appliedat the 1000-L tank scale in order toimprove yields of NIF-size boules to thepoint where rapid growth can become anindustrial production process. This sectionreviews the fundamentals of the growthprocess and the particular problems thataffect the commercial viability of rapidgrowth for NIF-scale boules.

2. Implosions for ICF work far better atshorter wavelengths due to less gener-ation of hot electrons, which preheatthe fuel and make it harder to com-press. Compromising between opticlifetime and implosion efficiency, bothNova and the NIF operate at a tripledfrequency of the 1053-nm fundamen-tal frequency of a neodymium glasslaser. This tripling is accomplished bytwo crystals, one made of KDP andone made of deuterated KDP (DKDP).The first one mixes two 1053-nm photons to make 526-nm light, and the second one combines a residual1053-nm photon with a 526-nm pho-ton to make 351-nm light.

The locations of the Pockels cell and fre-quency conversion crystals in the laser arealso shown in Figure 1, along with the ori-entation of the plates as they are cut fromthe crystal boule. Although all finished crys-tals will be 41 cm square, their different ori-entation with respect to the crystal axes,required in order to accomplish their differ-ent functions, causes different boule sizerequirements for the three types of finishedcrystals. The Pockels cell crystal is the easi-est, since it is cut horizontally with respectto the base of the crystal. As a result a 43-cm square boule is large enough, includ-ing a 1-cm buffer for finishing purposes,and the plates stack efficiently up to or eveninto the pyramidal cap, depending on thesize of the base. Only about 15 of thesesmall boules are needed. The KDP doublercrystal is the most challenging, as it is rotat-ed in two axes with respect to the base. As aresult, the minimum size base for a singledoubler crystal is 51 cm, and a symmetric55-cm-square by 55-cm-high boule will gen-erate only six doublers, so nearly 35 of theseboules will be required. A symmetric DKDPtripler must be a minimum of 55 cm high tobe tall enough for a single tripler, but thatminimum size will yield about 15 triplers.

Boules of both KDP and DKDP meetingNIF size and quality requirements havebeen grown by Cleveland Crystals, Inc.(CCI), by what is often called conventionalgrowth. In this case, impurities in thegrowth solution poison growth of the verti-cal faces (prisms), thereby maintaining across section approximately equal to that of

2

PRODUCING KDP AND DKDP CRYSTALS FOR THE NIF LASER

The fundamental property of natureenabling rapid crystal growth is that KDPcan attain very large and stable supersatu-rations in solution. In other words, KDPwill not spontaneously crystallize fromsolution when a solution is prepared athigh temperature and then cooled so thatthe salt concentration is above its equilibri-um solubility. This is because any pro-tocrystal formed by a statistical fluctuationhas to reach a minimum size beforegrowth is thermodynamically favorable.(This same nucleation criterion occurs inmany aspects of chemistry and physics.)Measurements have shown that stablesupersaturations from 35% at 65oC to 100%at 10oC can be attained if the solution isthoroughly preheated to eliminate anynucleation sites, and stability is not affect-ed by impurities at the level of tens to hun-dreds of parts per million.1 This super-saturation is much higher than the 3–20%required for growing crystal faces at ratesof 10–20 mm/day.

When a seed crystal is introduced intothis supersaturated solution, the crystalimmediately grows at a rate that dependson a variety of chemical kinetic and masstransfer factors. Practical experience overmany years has shown that the best way tostart this process is to first partially dissolvean oriented seed crystal of about 1 cm3 insize above the saturation temperature, thendecrease the temperature until the solutionsupersaturation reaches about 3%, at whichtime the seed “regenerates.”2 Regenerationis a process in which a rectangular base andpyramid form over the partially dissolved,rounded seed crystal. A picture of a regen-erated seed is shown in Figure 2. It isimportant to accomplish this regenerationrapidly (over 1–3 hours) so that each crystalface will have numerous imperfectionscalled dislocations. The crystal grows byadding atoms from solution to a set ofatomic steps that emanate from these dislo-cations. This configuration is called agrowth hillock, and a microscopic picture ofa growth hillock is shown in Figure 3.3 Asgrowth proceeds, the stronger growthhillocks crowd out the weaker ones, andfavorable growth proceeds with one to four hillocks on each crystal face. Crystalscontinue to grow as long as the temperatureis decreased to maintain appropriate

supersaturations. The growing crystal isrotated back and forth on a horizontal plat-form with a washing machine-like action tomaintain good mass transfer.

Even though the growth solution is sta-ble with respect to homogeneous nucle-ation, occasional heterogeneous formationof a single unwanted seed at a variety ofpossible locations can cause the formationof unwanted crystals, which are generallyfirst observed on the bottom of the tank.This crystal, being grown in nonoptimalconditions, usually cracks and subdivides,thereby seeding other parts of the tank.When one of these seeds inevitably landson the product crystal, its quality will bespoiled and the run ruined. Eliminatingthese spurious crystals for the entire twomonths of a growth run continues to beone of the most important challenges forcost-effective production.

Impurities in solution are detrimentalfor a variety of reasons:

• They affect the growth rates of theprism faces.

• They can enhance the formation ofinclusions of growth solution thatreduce optical quality.

• They substitute into the atomic lat-tice in the prism sectors and causeinhomogeneities in the refractiveindex and loss of optical transmis-sion by absorption.

3

PRODUCING KDP AND DKDP CRYSTALS FOR THE NIF LASER

FIGURE 2. Photographof a recently regeneratedseed. The cloudy regen-eration layer is coveredwith about 1 cm of cleargrowth.(40-00-0299-0435pb01)

• They may form particulates thataffect the laser damage threshold ofthe material.

While the chemical structure of the pyra-mid face causes typical ionic impurities tobe rejected from the growing crystal, thechemical structure of the prism face causesthem to be selectively absorbed and incor-porated into the crystal.4 A few examples ofthis selective absorption and rejection areshown in Table 1. Fe is the most importantimpurity in terms of transmittance, becauseFePO4 is highly absorbing at 351 nm, result-ing in a maximum acceptable Fe concentra-tion of 200 ppb for salt from which DKDPtriplers are to be grown.

4

PRODUCING KDP AND DKDP CRYSTALS FOR THE NIF LASER

1

2

3

(100)

(101)

(a)

(b)

(c)

FIGURE 3. (a) Schematic of the crystal surfaces indi-cating where the atomic steps (pictured through anatomic force microscope) emanate from dislocationson the (b) pyramid and (c) prism. The resultingmacroscopic feature is called a growth hillock. Thecrystal grows by adding atoms from solution at theedge of the steps. (40-00-0299-0436pb01)

TABLE 1. Concentrations of typical impurities in the raw material and in the pyramidal and prismatic sectors of KDP crystals.

RawImpurity material* Pyramid* Prism*

B 1000 ND ND

Na 86,000 ND ND

Al 900 200 4400

Si 12,000 <100 390

Ca 3600 ND ND

Cr 2000 490 11,000

Fe 5300 110 12,000

*Units of ng/g KDPND = not detected

Impurity incorporation into the prismface also affects the relationship betweengrowth rate and supersaturation, as shownin Figure 4.5 At low supersaturations,impurities effectively stop prism growth,and this condition is called “the deadzone.” Conventional growth occurs in thisregion, and only pure pyramid material isformed. Unfortunately, this corresponds togrowth rates less than 1 mm/day, whichresults in growth times greater than oneyear for NIF-size boules. In the middletransition zone, relatively small changes ingrowth conditions can have a drastic effecton growth rate. As is typical for any stable,reliable industrial process, this regionshould be avoided. This leaves the high-growth region as the most appropriate fordevelopment.

While the effect of impurities on growthrates is understood quantitatively oversome range of conditions, it is not under-stood quantitatively for mixtures of impuri-ties and over the wider temperature rangeused to grow crystals. Rather than attemptto quantify this parameter space in detail,our approach was instead to reduce impuri-ties to the lowest practical level, which isalso important for meeting other specifica-tions, and use the qualitative principle ofmaintaining the highest possible growthrate to minimize the effect of impurities on growth instability. Once the necessary

purity of the starting salt was attained (<0.5 ppm impurities), the contribution ofimpurities from the Pyrex growth tankswas explored. Though usually inert, there is a finite rate of tank dissolution in the hotKDP solutions used for rapid growth.4

Using measured Pyrex dissolution rates andthe uptake coefficients for various impuri-ties in the prism sector, we have successful-ly modeled the buildup and eventualconsumption of Al and Fe in the growthsolution,6 as shown for Al in Figure 5.Elements such as B and Si are not absorbedin the crystal and continue to build up inthe growth solution, while others such asCa are roughly constant during eachgrowth run but increase after each resatura-tion of the growth solution. Plastic tanks arebeing considered to eliminate this problem.Another important recent advance formaintaining crystal quality is the successfulimplementation of constant filtration toremove particulates that come from movingequipment or precipitation.7

At high growth rates, another problembecomes important. Variations in KDP con-centration on the µm scale at the growingcrystal steps can cause inclusions ofgrowth solution. These inclusions can easi-ly be large enough to cause more obscura-tion by scattering than can be tolerated inthe laser. Figure 6 summarizes much his-torical data on the occurrence of pyramidal

5

PRODUCING KDP AND DKDP CRYSTALS FOR THE NIF LASER

Prismatic growth rate and step speed vs s (Tsat = 50.5¡C)

12

10

8

6

4

2

01 2 3

Supersaturation, s (%)4 5 60

v(10Ð4cm

/s),

R(10Ð6cm

/s)

Step speed nonlinearity

limits s from below

v (data of Rashkovich)

R (data of Rashkovich)*

linear asymptote to v(s)

quadratic fit to R(s)

Growth rate

constraint

limits s from

above

(Rn = 10 mm/day

= 8 ´ 10Ð6cm/s

FIGURE 4. Effect ofimpurities on the stepvelocity and normalgrowth rate of KDP. Theoptimal growth rate isabove the nonlinearregion of step velocityand an upper bounddetermined by the uni-formity of mass transfer.(40-00-0299-0437pb01)

degradation of prismatic KDP that occursat slow growth rates while avoiding thepotential formation of massive pyramidalinclusions at high growth rates.

Microscopic investigations, hydrody-namic modeling, and theoretical modelinghave been combined to provide a goodmechanistic understanding of the

inclusions in 1000-L growth tanks through1998, indicating that for low rotation ratespyramidal inclusions are much moreprevalent at high growth rates. The sametrend occurs for prismatic inclusions. This presents the crystal grower with adilemma—how to avoid the detrimentaleffects of impurity buildup and associated

6

PRODUCING KDP AND DKDP CRYSTALS FOR THE NIF LASER

Day

Al concentration for Run F-4

50403020

CAl estimated from glass

leach rate and prism volume

CAl measured from chemical

analysis of weekly samples

100

1400

1200

1000

800

600

400

200

0Alconcentration(ppb)

FIGURE 5. Comparisonof measured and calcu-lated Al concentration inthe growth solution as afunction of time. The Alconcentration initiallyrises because the rate ofglass tank dissolutiondominates when thecrystal is small; it laterfalls because the rate ofuptake in the prism faceincreases with crystalsize and exceeds theglass dissolution rate,which drops with tem-perature.(40-00-0299-0438pb01)

Verticalgrowth

rate,

R2(m

m/day)

15

20

25

10

00 100 200 600500400300

5

Crystal height (mm)

Rotation rate was inadvertently

slowed from 25 rpm down to 15 rpm

Slow rotation, inclusions

Slow rotation, no inclusions

Rapid rotation, no inclusions

No inclusions@ high rotation

No inclusions@ low rotation

FIGURE 6. Summary of1000-L growth tank datashowing the relationshipbetween pyramidalinclusions and growthrates for slow and fastrotational regimes. Forruns prior to late 1998,the rotational rate wastypically 25 rpm, whilethe high rotation rateduring the recent runrepresented by the filledsquares was at 50 rpm. (40-00-0299-0439pb01)

formation of inclusions. A micrograph ofone type of unstable growth is shown inFigure 7, in which “fingering” is apparenton the advancing growth step.8 Althoughthermodynamics promote the filling of cav-ities between the fingers during slowgrowth, kinetics can cause the fingers togrow catastrophically, thereby surroundingand occluding growth solution. Solutionmoving in the opposite direction of the stepadvance is depleted in concentration as itmoves past the tips of the fingers andtowards the cavities, thereby causing afaster growth rate at the tip of the fingerthan at the base. Rapidly alternating thedirection of flow helps prevent this catas-trophic growth. A second type of instabilityinvolves the bunching and bending of ele-mentary growth steps into macrosteps dueto inadequate stirring.9 When macrostepsfrom two sides of a growth hillock bendaround and approach each other from theopposite direction, a deep valley can beformed. Inclusions also tend to form in thevicinity of this valley. Valley formation canbe minimized by rapid rotation, therebykeeping the solution concentration in thecenter of the crystal face closer in magni-tude to that along the edges.

Through a combination of these fundamen-tal studies and growth experiments at variousscales over the past year, we have shown that better mass transfer by increased accelera-tion and rotation rates can increase the inclu-sion-free growth rate by 20 to 40% over theslow rotation limit shown in Figure 6. Thebest-quality NIF-size KDP boule grown to the

time of this writing, completed in late 1998,followed the solid-square growth trajectoryalso shown in Figure 6, substantially abovethat previously considered safe. An importantaspect of achieving these higher rotation rateswas the design and fabrication of streamlinedAl growth platforms coated with a nonleach-ing Teflon-like coating.

The shape of the product crystal is also apractical problem. When grown rapidlyfrom a pure solution intended to meet quali-ty objectives, DKDP especially tends togrow with a height-to-width (aspect) ratioas low as 0.7. Since our circular growth plat-forms have a diameter of 90 cm, the maxi-mum symmetric-base is 63 cm, whichwould produce a crystal only 44 cm high—too short for triplers. A similar though lesssevere problem occurs for KDP. Improvingaspect ratio by allowing more solutionimpurities to retard prism growth is unde-sirable, especially for DKDP, which is chal-lenged to meet a more difficult 351-nm laserdamage threshold. As a result, we areexploring a variety of ways to improve dou-bler and tripler yields by making more opti-mal shapes. One idea is using an off-centerseed to promote asymmetric growth of thetype shown in Figure 8a, which is appropri-ate for increased doubler yield. A methodfor triplers is to grow the DKDP boule hori-zontally, which increases aspect ratio (nowrotated) by eliminating one growth prismand adding a second pyramid. Again veryrecently, a good-quality horizontal DKDPboule, shown in Figure 8b, was grown forthe first time to NIF size. Successful shapecontrol could increase conversion crystalyields by a factor of two over that fromsymmetric, vertically grown boules.

Laser DamageAt high laser fluences, KDP and DKDP

damages in the bulk by forming pinpointscattering sites a few micrometers in size.10

The basic consideration driving the laserdamage specification is that this scatteringshould not exceed 0.1% of the laser light peroptic traversed at the fluences expected for a 1.8-MJ ignition shot. The relevant fluencesare an average of 12 J/cm2 and 3σ limit of 18 J/cm2 for 1053-nm light over 3 ns on theKDP doubler and an average of 8.7 J/cm2

and 3σ limit of 14.3 J/cm2 for 351-nm light

7

PRODUCING KDP AND DKDP CRYSTALS FOR THE NIF LASER

1 mm

FIGURE 7. Fingering growth of macrosteps on theKDP surface. (40-00-0299-0440pb01)

We made progress during the past year inunderstanding both the relationship of easilymeasured, laser-damage distribution curvesand how to condition the crystals to mini-mize damage on-line. Through a combina-tion of constant filtration during growth andthermal annealing of the plates at 160oC overseveral days, damage of KDP at 1053 nm isnot a serious issue. On the other hand, thedamage characteristics of DKDP at 351 nmare more variable for unknown reasons, andmuch of the material formed under the bestconditions as currently understood and pre-conditioned by low-fluence shots is onlymarginal at NIF fluences.

To understand laser damage studies, onemust first understand the nomenclatureand characteristics of the standard tests:

1/1 a single laser shot of specified fluence

S/1 a set of shots (typically >100) of specified fluence

N/1 a small sequence of shots from lowto high fluence

R/1 a ramp of many shots (typically>100) from low to high fluence

In addition, the damage characteristicsof any material varies from location to loca-tion, so results are ordinarily presented interms of the distribution of fluences atwhich a specified amount of damage occursover many 1-mm beam spots, typically 100.

Figure 9 shows a comparison of the S/1and R/1 damage distributions curves forKDP sample 214, both as grown and afterthermal annealing. Note that the R/1 dam-age curve occurs at higher fluence than thecorresponding S/1 curve, indicating thebeneficial effect of the laser conditioninginherent in an R/1 experiment. Also notethat both the S/1 and R/1 curves shift tohigher fluence upon thermal annealing.These results are typical of all KDP sam-ples grown with constant filtration, indi-cating that thermally annealed KDPnormally damages at fluences above thatrequired for the NIF, especially if the laserenergy is increased to NIF fluences overseveral shots. The acceptability of the R/1threshold is unambiguous, since essential-ly no damage occurs for the maximumexpected NIF fluence, but further clarifica-tion is needed for the S/1 case. Even 10%

over 3 ns on the DKDP tripler. The allow-able damage is equivalent to about 100 pinpoints/mm3 of crystal volume.

In practice, the allowable damage criterionis more complicated. First, the extent of dam-age occurring at any given fluence dependson the previous laser exposure of the optic:low-fluence shots tend to condition bothKDP and DKDP, resulting in less damage athigher fluences than if the material were ini-tially exposed to the high fluences. Second,since much of the damage occurs in areas ofthe beam with above-average fluences, theaccumulated damage over time depends onhow much and how fast the areal fluencedistribution changes in the NIF beam overmany shots. In addition, the amount of dam-age occurring in KDP can be reduced bythermally annealing the material.

8

PRODUCING KDP AND DKDP CRYSTALS FOR THE NIF LASER

(a)

(b)

FIGURE 8. Recentrapid-growth NIF-sizeboules. The top boulewill yield about 11 KDPdoublers, and the bot-tom boule will yieldabout 12 DKDP triplers.(40-00-0299-0441pb01)

failure probability on this scale for theaverage NIF fluence is not a problembecause damage is detected at a level inwhich most of the light is still transmittedand the damage pinpoints do not grow.

The situation for damage of DKDP at 351 nm is less certain because of fewer data,the greater variability of DKDP properties,and the overall proximity to the damagerequirement. Furthermore, enhancement of

the damage threshold by thermal annealingis not practical because either decompositionor recrystallization occurs at temperaturesrequired for annealing. A summary of allR/1 data to date for both conventional- andrapid-growth DKDP are shown in Figure 10.Note that the best conventional-growthDKDP has a damage distribution far greaterthan NIF requirement, indicating that thedesired material can be grown. However, the

9

PRODUCING KDP AND DKDP CRYSTALS FOR THE NIF LASER

Thermal annealing comparison

Fluence (J/cm2, 1w @ 10 ns)

100

90

80

70

60

50

40

30

20

10

00 30 60 90 120 150

Cumulativefailure

probability(%

)

KDP214 S/1

KDP214 S/1 thermal

KDP214 R/1 laser

KDP214 R/1 thermal + laser

FIGURE 9. Percent of 1-mm aperture beamexposures havingdetectable damage for arapidly grown KDPsample exposed to vari-ous preconditioning andfluence schedules. Theband represents the 4σfluence distribution onthe NIF adjusted forpulse length.(40-00-0299-0442pb01)

0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35

Cumulativefailure

probability(%

)

Cumulativefailure

probability(%

)

0

100

Fluence (J/cm2, 355 nm, 7.6 ns)Fluence (J/cm2, 355 nm, 7.6 ns)

Rapid growth DKDP Conventional growth DKDP

80

60

40

20

0

100

80

60

40

20

DKDP11

DKDP367

DKDP586

LL1 70%

LL4 70%

LL3 70%

LL1 80%

LL6 80%

DKDP354DKDP527DKDP529DKDP549

RG8A

FIGURE 10. Summary of all 3ω damage data for rapidly and conventionally grown DKDP. To compare to NIF requirements, these fluencesare scaled to a 3-ns pulse length by dividing by (7.6/3)0.5 = 1.6. (40-00-0299-0443pb01)

1-cm-diam, 3-ns, 351-nm beam with theOptical Science Laser (OSL). Pinpoint den-sity and obscured area for a 1-cm-thickcrystal is given in Figure 10 as a function offluence for an N/1 ramp of 8 shots. Whenthis pinpoint density is convolved with theexpected NIF beam fluence profile asshown in Figure 11, 0.3% obscuration ispredicted by the time the laser reaches a1.8-MJ shot. This is close to the desiredgoal, so the R/1 damage distribution of thissample represents a possible lower accept-able limit.

Crystal RG8A-1 was used to generate351-nm light on one of the Beamlet experi-mental campaigns. The crystal wasramped over 9 shots to a maximum flu-ence of 3.8 J/cm2 over 1.5 ns, which corre-sponds to ~5.5 J/cm2 in 3 ns and 8.6 J/cm2

in 7.6 ns assuming a τ0.5 scaling law.Essentially no bulk damage was observed,which is consistent with being at the verybeginning of damage for the R/1 curve.An estimate of the pinpoint density vs flu-ence distribution, convoluted with themeasured beam profile on Beamlet, leadsto a prediction of about 0.1% obscurationloss. However, the average NIF full flu-ence corresponds approximately to the 80–100% damage level, so RG8A materialis not expected to meet the damagerequirements of NIF for full-fluence shots.Consequently, the ability to grow NIF-sizeDKDP boules of sufficient quality by rapidgrowth is yet to be demonstrated.However, it is encouraging that severalsamples of sufficient quality have beengrown in 20-L tanks.

While conversion crystals initiallyinstalled onto the NIF may see a gradualincrease in fluence, and thereby achievelaser conditioning on line, subsequentreplacements will not have that opportuni-ty. Moreover, uncertain performance duringinitial start-up provides some risk to theinitial conversion crystals. Consequently, itis highly desirable to provide laser condi-tioning of the conversion crystals by rasterscanning prior to installation. Figure 12shows the improvement of the S/1 damagefluence distribution with one and tworaster preconditioning steps. (The two steps are required because attempts toraster-condition crystals in a single scan

damage threshold is quite variable for rea-sons not yet completely understood (dis-cussed below), so highly efficient productionis not yet in hand.

Because of the proximity of the damagedistribution curve of many samples to theNIF fluence distribution, a closer evaluationis appropriate. Sample DKDP11, one of the best by rapid growth, was exposed to a

10

PRODUCING KDP AND DKDP CRYSTALS FOR THE NIF LASER

150.1

1

10

100

Pinpointden

sity

(N/mm3) NIF redline

5 10

Fluence (J/cm2, 351 nm, 3 ns)

1000

3 pp/mm3 Zeus

detection threshold

(a)

10

12

6

0

10

0

Ð10

Distance

(cm)

Ð10 0

Distance (cm)

NIF SW THG spatial profile

(b)

20

100

50

0

Ð10

0

20

10

Distance

(cm)

Number

ofpinpoints

0 10Ð10Ð20Ð20

Distance (cm)

(c)

FIGURE 11. Relation-ships among laser flu-ence, pinpoint bulkdamage, and obscura-tion at NIF fluences.(40-00-0299-0444pb01)

is hampered by jitter of the raster beam,which causes too large of fluence jumps.)The two-step preconditioning raises the S/1damage distribution to near that of the R/1curve, demonstrating that off-line condi-tioning can be an effective way to bringnew optics on line.

Considerable effort has been invested indiscovering the precise mechanism of bulklaser damage so that the damage thresholdcan be improved. Recall that the NIF fluenceis constrained by damage to the crystals.First, a comparison of preexisting laser scat-tering sites with subsequent laser damagelocations reveals no correlation betweenscattering and damage at either 1053 or 355 nm, presumably because the scatteringsites have negligible absorption.11 Similarly,bulk absorption is not related to damage;the 355-nm damage distributions for prismand pyramid material are the same eventhough the content of dissolved iron andresulting absorption coefficient are substan-tially different.12 Simple thermal calcula-tions show that even a few percent of laserabsorption evenly distributed throughoutthe material should not be a problem, butlocalized heating from highly absorbing par-ticles in the range of 10–100 nm can easilycause thermal and acoustic shock damage tosurrounding material. FePO4 has therequired absorption coefficient at 3ω, andaddition of FePO4 powder to the growthsolution causes a major drop in the 3ω dam-age curve. In addition, secondary ion massspectrometric (SIMS) analysis of the material

sputtered from three damage sites showedsignificant amounts of Fe and Cr in additionto major amounts of Ca. Perhaps the Cabuildup from repeated resaturations causesthe formation of mixed cation phosphateprecipitates. Further work is in progress totest this hypothesis.

Finishing DevelopmentProducing finished crystal optics with

an aspect ratio of greater than 40:1 fromlarge boules poses a variety of fabricationchallenges. Many of the primary specifica-tions for NIF crystals are difficult to meet.Three of the more difficult specifications arerelated to crystal surface figure and finishand include: short wavelength surfaceroughness (3.0 nm rms for λsp <0.12 mm),surface waviness (6.4 nm rms and powerspectral density for λsp between 0.12 mmand 33 mm), and 5λ (λ = 633 nm) peak-to-valley surface figure for λsp >33 mm.Perhaps the most difficult specification isproducing finished crystals oriented withrespect to the crystallographic axes suchthat the average phase matching angle for frequency conversion is accurate to ±15 µrad (external angle) for doublers and ± 30 µrad for triplers. Since the phasematching angle is strongly dependent uponuse temperature and wavelength, extremelyprecise finishing machines and metrologytools are required for process control duringcrystal fabrication. In addition to difficulttechnical specifications, aggressive cost and

11

PRODUCING KDP AND DKDP CRYSTALS FOR THE NIF LASER

0

20

40

60

100

80

0 15 20 25 30 35 40105

1.6-mm beam

Stage velocity =

15 µm/shot

Equivalent ramp

~20 shots/site

Fluence (J/cm2at 355 nm, 7.6 ns)

Cumulativefailure

probability(%

)

R/1 virgin

R/1 annealed

S/1 virgin

S/1 annealed

S/1 rastered

FIGURE 12. Raster con-ditioning increases 3ωdamage performance ofKDP. Two successiveraster scans at 10 and 14 J/cm2 increase theS/1 performance to nearthe R/1 level.(40-00-0299-0445pb01)

production schedules necessitated thedevelopment of improved manufacturingprocess and tools.

Early development activities focused ondetermining the requirements for long-leadmachine tools. The Pneumo Final FinishingMachine at CCI was improved in 1995–96,as illustrated in Figure 13. The inspectiondata from subsequent parts was used inlaser propagation modeling, which validat-ed that diamond-turned parts could meetNIF requirements and established the newfinishing machine requirements. However,the Pneumo machine is 20 years old, under-sized, and cannot finish parts in the timerequired. Consequently, a new machinewas designed and fabricated by the MooreTool Company. The resulting machine,shown in Figure 14, is currently undergoingperformance acceptance test at LLNL.

While roughness, waviness, and wedgeare controlled by the final finishing machine,optical figure is largely determined by theblank fabrication process. Again, currentdevelopment activities can be traced backseveral years. The Nova fabrication and fin-ishing process could not reliably producefinished crystals meeting NIF figure require-ments. As a result, CCI modified its propri-etary blank fabrication process in 1995during Beamlet crystal production, and thenew process has been used to produce manylarge-aperture crystals that meet NIF figurerequirements. As shown in Table 2, the pro-cess of producing flat blanks is now well inhand. Again, however, the CCI machines areold, undersized, and too slow to meet NIFproduction schedules.

12

PRODUCING KDP AND DKDP CRYSTALS FOR THE NIF LASER

(a) Before 7/94

LL112R10b

LL112R70h

0.01

104

102

10Ð2

10Ð4

10Ð6

104

102

1

10Ð2

10Ð4

10Ð6

1

10 10010.1Frequency (mmÐ1)

(b) After 3/96

10LL112Re

70LL112Ra

PSD(nm

2 mm)

PSD(nm

2 mm)

FIGURE 13. Comparison of power spectral density(PSD) for the surface roughness of Beamlet crystalsmachined on the Pneumo machine at CCI in 1994 and1996, showing an improvement from 3- to 8-nmroughness to 1- to 2-nm roughness.(40-00-0299-0446pb01)

FIGURE 14. The MooreFinal Finishing Machinehas now been assembledand is undergoingacceptance testing. (40-00-0299-0447pb01)

Consequently, Lawrence LivermoreNational Laboratory (LLNL) started devel-oping a new set of custom machine toolsin 1997: a boule-facing machine, a profilingmill, and three flatness machines. The flat-ness machine is the most challenging ofthe three types. It must produce surfacesthat meet the flatness requirements for λsp >33 mm. A model of it is shown inFigure 15, and the first machine is nearingdry-cut testing. Three z-axis slides provideapproximately 100 mm of vertical traveland up to 0.25 degrees of tip and tilt.Compliance tests performed on the z-axislast summer determined the stiffness of thez-axis—an important metric on precision

machine tools—to be 36 nm/N (6.5 µin./lb). This is actually slightly better than the existing Pneumo machineand leaves open the possibility that thebasic machine design may be good enoughto allow one of the flatness machines, withmodest upgrading to remove roughnessand waviness at scale lengths <33 mm, toserve as a backup final finishing machine.

Precise crystal orientation is perhaps themost difficult finishing specification.Beamlet tests in 1997 revealed spatial varia-tions in frequency conversion efficiency thatrelated to bulk crystal features, making theorientation process even more challenging.At the time, the only off-line method thatexisted for predicting frequency conversionperformance of crystals involved using sub-aperture lasers to measure variations intheir phase matching angles. The only pro-duction tool for this purpose was theCrystal Orientation Measurement System(COMS) at CCI. COMS compares the peakof the tuning curve of a proof crystal and acrystal being tested by comparing theamount of laser energy converted by eachcrystal as both are rocked in parallel beamson a common mount. However, the COMSconfiguration could only be used to accesstwo points on each crystal.

Diagnostic techniques developed duringBeamlet frequency conversion experimentshave provided a breakthrough in thisregard. Variations in frequency conversionefficiency correlate very well with varia-tions in crystal birefringence determinedsimply by subtracting the transmittedwavefront through crystals at two orthogo-nal polarizations corresponding to the “o”and “e” transmission axes.13 This tech-nique, now called “orthogonal polarizationinterferometry” (OPI), provides a valuablealternative to mechanically complex two-dimensional scanning for determining therelative phase matching angles across thefull aperture of crystals. However, OPIdetermines only the relative values acrossthe crystal, not the absolute phase match-ing angle needed for NIF production.

The Crystal Alignment Test System(CATS), now under development, relies onOPI to determine the distribution of phasematching angles across a crystal. Small beamfrequency conversion data from the CATS

13

PRODUCING KDP AND DKDP CRYSTALS FOR THE NIF LASER

TABLE 2. A summary of crystals produced forBeamlet since 1996 shows that the flatness process can meet NIF specifications at high yield.

Surface figureCrystal (waves @ λ = 633 nm)

328-4 8328-5 5328-6 4

RG8B-1 2.2RG8B-2 2.3RG9B 2.1345-1 4.5

70% LL37-1 1.1LL6-11 4.5

FIGURE 15. The LLNL Prototype Flatness Machinedesign is highly integrated to reduce complexity dur-ing machine assembly. (40-00-0299-0448pb01)

temperature between the proof and thecrystal being tested must be controlled.The system is also insensitive to variationsin laser wavelength between the NIF usewavelength and the CATS laser, since boththe proof and crystal being tested see thesame wavelength. There is only a minorerror introduced if the deuteration level ofa tripler crystal being measured is differ-ent from the tripler proof.

ConclusionsRapid growth has now grown numer-

ous boules of NIF size. The technical chal-lenge for KDP is to increase processreliability and doubler yields to minimizecosts, as all performance specificationshave been achieved. Thermal annealing isan important aspect of meeting those spec-ifications. The principal technical chal-lenge for DKDP is to increase 3ω damagethreshold at full size to the NIF require-ment. The best DKDP grown in smalltanks has better 3ω damage propertiesthan needed for the NIF, but that level hasyet to be demonstrated in large boules.

The Moore Final Finishing Machine,flatness machines, and CATS togetherform the primary foundation needed tomeet the most difficult of the crystal fin-ishing specifications. The Moore FinalFinishing Machine is undergoing final testing. Mechanical assembly of the LLNLPrototype Flatness Machine is nearly complete, and compliance tests were very

will supplement the OPI with absolutephase matching angles for a line of pointsalong the crystal. The two data sets can thenbe correlated to provide a map of the abso-lute phase-matching angle at all points onthe crystal and an average phase matchingangle that is used to correct crystal orienta-tion during final finishing operations.

The CATS design, shown schematically inFigure 16, is conceptually very similar to theexisting COMS in that each measurement isreferenced to a precisely oriented proof crys-tal. Two parallel beams pass through theproof and test crystals, which are mountedon a single diamond-turned chuck. Penta-prisms are used to produce two parallelbeam paths, thereby greatly simplifying themechanical and controls requirements of thesystem. Because frequency conversion crys-tals are insensitive to out-of-plane rotationerrors in the pentaprisms, there is no need tocompensate for slide straightness errors oradjust beam alignment at each point in thescan line. During operation, one beam samples the proof while the other samplesthe crystal to be tested. The mount is rocked,and converted power is measured in bothchannels. The resulting tuning curves arecurve-fit to determine the phase matchingangle offset between the two crystals.

The use of calibrated proof crystals inCATS greatly reduces the sensitivity of themeasurements to systematic errors, includ-ing temperature, laser wavelength, point-ing jitter, etc. For example, by comparingcrystals to a known proof, only the relative

14

PRODUCING KDP AND DKDP CRYSTALS FOR THE NIF LASER

FIGURE 16. The concep-tual layout of CrystalAlignment Test Systemreveals many similaritiesto COMS, its predecessor.(40-00-0299-0449pb01)

Receiving translation stage

Glan/laser

polarizer

Spatial

filter Optical

isolator Shutter

Integrating

sphere

Corner cube

reflector

CCD

Fast diode

Lens

Translation stage

for pentaprisms

Laser

successful. A new CATS design takesadvantage of new metrology techniques,and construction of several subsystems areready to begin.

AcknowledgmentsThis work represents the efforts of

dozens of people including Leslie Carman,Igor Smolsky, Sergey Potapenko, MartinDeHaven, Randy Floyd, Dick Spears, MikeMartin, and Warren Bell in the growtharea, and Mary Norton, Jerry Auerbach,Larry Morris, Sue Locke, Gale Johnson,Sam Thompson, Gregg Wilkinson, DaveKennedy, Stan Locke, Frank Gonzales,Dannie Johnson, Charles Cass, BlaineBeith, Mercedes Dickerson, GeorgeWeinert, Timm Wulff, Richard Arney, DanSchumann, Inge Fine, Wayne Olund, andChris Steffani in fabrication support.

Notes and References1 N. P. Zaitseva, J. J. De Yoreo, M. R. DeHaven, R.

L. Vital, L. M. Carman, and H. R. Spears, J.Crystal Growth 180, 255–262 (1997).

2. N. P. Zaitseva, I. L. Smolsky, and L. N.Rashkovich, Krystallografiya 36, 198 (1991).

3. J. J. De Yoreo, T. A. Land, L. N. Rashkovich, T. A.Onischenko, J. D. Lee, O. V. Monovskii, and N. P.Zaitseva, J. Crystal Growth 182, 442–460 (1997).

4. N. P. Zaitseva, L. Carman, I. Smolsky, R. Torres,and M. Yan, “The effect of impurities and super-saturation on the rapid growth of KDP crystals,”submitted to J. Crystal Growth.

5. L. N. Rashkovich and N. V Kronsky, J. CrystalGrowth 182, 434–441 (1997).

6. H. Robey, R. Floyd, R. Torres, and A. Burnham,“Impurity leaching rates of 1000 liter growthtanks,” Lawrence Livermore NationalLaboratory, Livermore, CA, UCRL-ID-133365.

7. N. Zaitseva, J. Atherton, R. Rozsa, L. Carman, I.Smolsky, M. Runkel, R. Ryan, and L. James,“Connection between continuous filtration anddislocation structure of KDP crystals,” submittedto J. Crystal Growth.

8. H. F. Robey and S. Potapenko, “Ex-situ micro-scopic observation of the lateral instability ofmacrosteps on the surface of rapidly grownKH2PO4 crystals,” submitted to J. Crystal Growth.

9. H. F. Robey, S. Potapenko, and K. Summerhays,“Bending of steps on rapidly grown KH2PO4crystals due to an inhomogeneous surface super-saturation field,” submitted to J. Crystal Growth.

10. M. Runkel, J. De Yoreo, W. Sell, and D. Milam,“Laser Induced Damage in Optical Materials:1997,” in Proc. SPIE, vol. 3244, pp. 51–63.

11. B. Woods, M. Runkel, M. Yan, M. Staggs, N.Zaitseva, M. Kozlowski, and J. De Yoreo, “LaserInduced Damage in Optical Materials: 1996,” inProc. SPIE, vol. 3244, pp. 20–31.

12. M. Runkel, M. Yan, J. De Yoreo, and N. Zaitseva,“Laser Induced Damage in Optical Materials:1997,” in Proc. SPIE, vol. 3244, pp. 211–222.

13. P. J. Wegner et al., Frequency ConverterDevelopment for the National Ignition Facility,Lawrence Livermore National Laboratory,Livermore, CA, UCRL-JC-129725. Prepared forProc. 3rd Intl Conf on Solid State Lasers for ICF.

15

PRODUCING KDP AND DKDP CRYSTALS FOR THE NIF LASER