data (curve 4) under the same reactionconditions. Note that the data converge atlarge thickness or diameter, suggesting thatthe calculations of the LSM surface areas,based on ideal geometry of the layers, arenot significantly in error. Furthermore, thehydrogenation of ethylene (C2H4 + H2 -_C2H6) on the Ni LSMs showed no thick-ness dependence of the specific rate, inagreement with observations made withtraditional supported clusters. These resultsimply that only one dimension need be inthe nanometer scale to produce the sizeeffects. The LSM layer thickness is therelevant parameter. Since the surface-to-volume ratio in the LSMs is orders ofmagnitude smaller and the number of con-nected Ni atoms is orders of magnitudelarger than in traditional supported cata-lysts, that ratio or number cannot be thesource of the size effects.

The origin of the size effects must besought in explanations that are indepen-dent of the total number of connectedatoms in the metal but are dependent onthe configuration of the atoms in at leastone nanometer-scale dimension. Two pos-sible explanations can be proposed, al-though neither can be proven at this time.One is the result of a support effect, becausethe metal support contact area parallel tothe nanoscale dimension is essentially thesame in both LSM edges and clusters. It isknown that Ni and SiO2 do not mix in thebulk, but even if that occurs on the nano-meter scale in the LSMs the same effectwould be present in the clusters due tosimilar postpreparation processing. Thestoichiometry of the silica layers is notknown since SiO, may be present in thevapor-deposited layer. However, if thatstoichiometric variation is the origin of thesupport effect, it must be present also in thecluster supports, which are prepared in avery different manner. Alternatively, thecluster and LSM surface structures could bethe same. Because it is unlikely that theminimum free energy surface structure ofthe supported cluster matches that of theminimum free energy surface structure ofthe LSM edge, the LSM result suggests thatneither are equilibrium structures, thus ex-cluding those derived from the model po-tentials for unsupported clusters with a spe-cific total number of atoms in them. Recentcalculations of the structure of smaller un-supported clusters, made with the use ofmetal interaction potentials more appropri-ate to metals, also suggest that the surfacestructures in real systems could be signifi-cantly different from those previously cal-culated (6, 7). Limited studies of supportedcluster catalyst topography with tunnelingmicroscopy further suggest that verynonideal structures are present (8). Al-though increased rates of catalytic reactions

have been reported for amorphous catalysts(9), the larger maximum absolute ratesreported here, relative to clusters, are morelikely due to the narrower size distributionin the LSM edges. Finally, size-constrainedsurface reconstruction as a result of reactioncould occur both on clusters and LSMs.

REFERENCES AND NOTES

1. M. Boudart, Adv. Catal. 20, 153 (1969).2. M. Che and C. 0. Bennett, ibid. 36, 55 (1989).3. R. S. Averback, J. Bernholc, D. L. Nelson, Clus-

ters and Cluster-Assembled Materials, vol. 206 ofthe Materials Research Society Symposium Pro-ceedings (Materials Research Society, Pitts-burgh, 1991).

4. I. Zuburtikudis, thesis, University of Rochester(1992).

5. T. W. Barbie, SPIEJ. 563, 1 (1985).

6. A. Sachdev and R. Masel, paper presented at theAmerican Institute of Chemical Engineers 1991Annual Meeting, Los Angeles, November 1991.

7. D. G. Vlachos, L. D. Schmidt, R. Aris, J. Chem.Phys. 96, 6880 (1992).

8. P. A. Thomas, W. H. Lee, R. I. Masel, J. Vac. Sci.Technol. A 8, 3653 (1990).

9. A. Molnar, G. V. Smith, M. Bartok, Adv. Catal. 36,329 (1989).

10. D. J. C. Yates, W. F. Taylor, J. H. Sinfelt, J. Am.Chem. Soc. 86, 2296 (1964).

11. A. SArkAny and P. T6tenyi, React. Kinet. Catal.Lett. 12, 297 (1979).

12. J. H. Sinfelt, J. L. Carter, D. J. C. Yates, J. Catal.24, 283 (1972).

13. Sincere thanks to L. Fuller and D. Casilio of theRochester Institute of Technology (Department ofMicroelectronics Engineering) for their decisiverole in the lithography and etching work. Thefinancial support of the Elon Huntington HookerFellowship to l.Z. is gratefully acknowledged.

24 June 1992; accepted 24 September 1992

Viscoelastic Dynamics of Confined Polymer MeltsHsuan-Wei Hu and Steve Granick

The frequency-dependent shear response of an ultrathin polymer melt (polyphenylmeth-ylsiloxane) confined between adsorbing surfaces (parallel plates of mica) is described. Thesinusoidal deformations were sufficiently small to give linear response, implying thatmeasurement did not perturb the film structure. A remarkable transition was observed withdecreasing thickness. When the film thickness was less than five to six times the unper-turbed radius of gyration, there emerged a strong rubber-like elasticity that was notcharacteristic of the bulk samples. This result indicates enhanced entanglement interac-tions in thin polymer films and offers a mechanism to explain the slow mobility of polymersat surfaces.

From biology to tribology, prominent in-terfacial processes involve polymers. Withregard to surface mobility, the sharp incon-sistency between theory and experimentemphasizes the prominence of some kind oftrapped state in this problem (1-3). In thisreport, we are concerned with the dynamicsof molten polymer of thickness comparableto the radius of gyration; other examples ofunexpectedly slow surface mobility concernthe spreading of polymer melts (4) and theadsorption-desorption dynamics of poly-mers in solution (5). Theoretical studies ofthe confining forces needed to maintainthin polymer films show that essentiallyzero force is required at equilibrium (6)down to a film thickness of two to threesegment dimensions (7). Experimentally,strong repulsion is observed starting whenthe films are significantly thicker, severaltimes the unperturbed radius of gyration(8-10). Dynamic experiments also indicatea slowing down of near-surface dynamics(11, 12). However, the forced deformationsin those studies caused nonlinear responsesand the experiments were not designed tomeasure associated elastic effects. The is-

Materials Research Laboratory and Department ofMaterials Science and Engineering, University of Illi-nois, Urbana, IL 61801.

SCIENCE * VOL. 258 * 20 NOVEMBER 1992

sues raised are clearly relevant to manydiverse physical phenomena, especiallypolymer-filler interactions in rubbers andcomposites (13), polymer adhesives, lubri-cation, and problems concerning the wallboundary conditions for transport duringpolymer processing (14).We report on the linear viscoelasticity of

thin polymer films. The significance of lin-ear response (obtained with shear ampli-tudes <2 A) is that measurement did notperturb the film structure. A remarkabletransition was seen with diminishing filmthickness. At a thickness less than five tosix times the unperturbed radius of gyration(5 to 6 RC), the viscoelastic behavior cross-es from viscous relaxation, which is charac-teristic of single chains in the bulk liquid,to a predominantly elastic and rubber-likeresponse. This latter response indicates en-hanced steric constraints on motion, whichmay be construed as enhanced entangle-ment interactions.

The experimental approach has beenreported elsewhere (15, 16). In brief, thepolymer fluids were confined between atom-ically smooth, step-free single crystals ofmuscovite mica that were configured asparallel plates and that were surroundedby a droplet reservoir. To measure shear

1339

on

Aug

ust 1

8, 2

011

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

Table 1. Molecular characteristics of the PPMSsamples.

Code (g Mo-1) MAMnt nnt G§(gmol-1)~(A) &

A 2240 1.08 31 9 _B 4620 1.06 65 13 Q

*Number-average molecular weight. tRatio ofweight-average to number-average molecular weight.If all chains had the same length, this ratio would beunity. tNumber-average number of skeletalbonds. §Unperturbed radius of gyration, estimat-ed from the chain length by a rotational isomeric statescalculation (19).

forces, a sinusoidal time-varying shear forcewas applied and the amplitude and phase ofresponse were measured. The small influ-ence of mechanical deformation within thedevice was accounted for by a model inwhich the calibrated properties of the de-vice [especially the glue that attached themica to the rest of the device (16)] acted inparallel with the thin film. Recent improve-ments have increased the system resolutionand stability so that viscous forces as low as10` N were resolved. Control experimentswith shear amplitudes as low as 0.5 Aconfirmed that the response was linear inamplitude over the frequencies studied,0.03 to 250 Hz.

The polymer fluids, atactic methyl-termi-nated polyphenylmethylsiloxane (PPMS)(17), are comprised of amorphous and flex-ible chains that cannot crystallize becausethey are atactic. The bulk glass transitiontemperature (Tg z -20"C) (18) is wellbelow the temperature of these experiments(260 ± 0.50C), which avoids glassy dynam-ics complications that have been much dis-cussed (1, 2). The contact angle with mica is

00. Control experiments, in which PPMSwas displaced from mica by water, showedthat adsorption was reversible. Molecularcharacteristics of the samples studied aregiven in Table 1. The unperturbed RC wasestimated from the chain length by a rota-tional isomeric states calculation (19). Thepersistence length of atactic PPMS in thebulk, estimated from the characteristic ratio,is about six skeletal bonds (10 A) (19).Although we have no direct measurement ofthe entanglement length of PPMS (becauseof a limited amount of sample), from thecalculated chain stiffness (19) one expectsthat the molecular weight between entangle-ments is Me 12,000 g mol` (20), SO thatchains of the length studied here would betoo short to be entangled in the bulk.

The normal pressures required to squeezefilms of polymer B (65 skeletal bonds) to aspecified thickness are shown in Fig. 1 (21).The mean normal pressure, Pa (normalforce normalized by the area of the parallelmica plates), is plotted against film thick-ness (D). The normal pressure rose mono-

1340

0.6

50D(A)

100

Fig. 1. Static pressure in the normal directionrequired to squeeze a film of polymer B (Mn =

4620 and RG 13 A) to the specified separa-tions. Open circles: equilibration time 20 minper datum. Filled circles: 24 hours per datum.Inset shows the derived bulk longitudinal mod-ulus ratio, M = -BF/BInD, plotted against filmthickness (0) normalized by RG.

tonically at thickness D < 60 to 70 A (5 to6 RC) in approximately an exponentialfashion (decay length close to RC)- Becausethe question of equilibration is an essentialpoint, the equilibration time for polymer Awas varied by a factor of nearly 100 (20 minand 24 hours per datum) with no discern-ible effect.

These data suggest two immediate con-clusions. First, long-range repulsion inthese squeezed polymer films was kineticallystable over the long time scales studied.This result decisively corroborates earliermeasurements taken with shorter equilibra-tion times (8-10). It suggests interpretationof P. (D) in terms of the energy to squeeze afixed adsorbance of tethered polymer. Sec-ond, from the slope of the data in Fig. 1, thebulk longitudinal modulus [M = -(PI8InD)I (13) of these molecularly thin filmscan be calculated. As excess polymer wassqueezed out with diminishing thickness, Mrose as indicated in the inset of Fig. 1. Thisvalue is used below to estimate Poisson'sratio.

Turning to the dynamic oscillatory shearforces, we consider first a film of thicknesssuch that it did not support any normalpressure. The physical picture to imagine isan inhomogeneous system: opposing solidsurfaces, each carrying polymer tethered bymultiple points of contact, with unattachedpolymer in between. This situation issketched schematically in the inset of Fig. 2.

In Fig. 2, G"(w) of a film of polymer Bof thickness 90 + 3 A (7 R.) is plottedagainst radian frequency, o, on logarithmicaxes. The loss modulus, G"(w), is the com-ponent of the oscillatory stress that (follow-ing Newton's law of viscous flow) was inphase with the rate of deformation. Herewe ignore possible edge effects and assumethat the length scale by which to normalizethe shear amplitude is the film thickness.

SCIENCE * VOL. 258 * 20 NOVEMBER 1992

6

6Ad_t

4

2

-1 0 1 2 3 4

Log(X) (rad s-l)Fig. 2. Log-log plot of the loss modulus, G"(w),as a function of angular frequency of measure-ment, w, for a film of polymer A of thickness 90± 3 A (7 RG). Error bars are indicated. Insetshows a schematic illustration of the opposingsolid surfaces, each carrying polymer tetheredby multiple points of contact, with unattachedpolymer in between.

On physical grounds, one expects a region ofpolymer segments near the solid surfaces tohave slower mobility, making the rheologi-cal film thickness somewhat less than themica-mica separation (8-11). However, acorrection for this effect would not changethe relative values of the viscoelastic modu-li, nor their orders of magnitude. Because itis not clear how to correct the raw data, weavoided such fine-tuning of the analysis.

As was expected of bulk viscous flow atlow frequency, where G" must equal on (rjis the constant viscosity) (13), the G"(w)data in Fig. 2 have a slope of unity. Thisresult indicates that the inverse frequencyof the experiment was larger than the long-est relaxation time of the liquid. The effec-tive viscosity [G"(o)Au] was 'neff 100 Pa s.

We distinguish between Jneff and a bulkviscosity because the structure of the systemwas inhomogeneous. Nonetheless, the esti-mate, 100 Pa s, is reasonable for the bulkresponse of PPMS liquid of this chainlength.

The physical picture to imagine, as thepolymer films were squeezed to a thicknessless than 5 RC, is a progressive interdigita-tion of the polymer layers that are tetheredto the opposing solid surfaces.We consider first a film thickness just at

the point of onset of static forces in thenormal direction. In Fig. 3, G'(w) andG"(w) of a film 62 + 1 A thick (4.8 RG) areplotted against X on logarithmic axes. Thestorage modulus, G'(w), is the componentof oscillatory stress that (following Hooke'slaw of elastic deformation) is in phase withthe deformation. The data split into tworegimes: (i) low frequencies, where G'(w)and G"(w) are comparable in magnitude,

m-1111

e * | | |. . .

on

Aug

ust 1

8, 2

011

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

6

U

a.4

32

6

ha.-4

2

-1 0 1 2 3 4 -1 0 1 2 3 4

Log(() (mad s-1)

Fig. 3. Log-log plot of storage and loss shearmoduli, GC(w) and G"(w), as a function of angu-lar frequency of measurement, for a film ofpolymer B of thickness 62 ± 1 A (4.8 R,).

and (ii) higher frequencies, where G'(Q)exceeds G'(o). The frequency at the splitdefines a relaxation time of the system (inthis case -0.1 Hz). This slow relaxationtime can be tentatively identified with thekinetics of attachment and reattachment ofthe adsorbed structures. Qualitatively, thebehavior in Fig. 3 resembles the conven-

tional viscoelastic behavior of high molec-ular weight polymer fluids at the transitionbetween terminal and plateau zones (13).

As the films were made thinner, theirelastic shear response became stronger andthe frequency where G'(w) and G"(w) splitmoved to progressively lower frequencies.In Fig. 4, G'(w) and G"(w) of a film 421 A thick (3.2 R0) are plotted against w on

logarithmic axes. We note that: (i) themoduli showed minimal dependence on thefrequency of deformation, (ii) G'(w) ex-

ceeded G"(w) by nearly an order of magni-tude, and (iii) G'(w) ~ 0.3 MPa. Over therange of film thickness from 30 to 40 A,changes in G' (w) were not detected.

These features would also be typical of a

cross-linked or highly entangled rubber inthe bulk (13). The large strain (-0.1) atonset of a nonlinear response (significantlylarger than for a crystal) would also betypical of those systems. Pursuing the anal-ogy, an estimate of the density of effectivenetwork strands follows in which the clas-sical equation of rubber-like or entangle-ment elasticity, G' (w) = (p/Me) (kBT) isused (p is the density, Me is the molecularweight between effective cross-links, kB isBoltzmann's constant, and T is tempera-ture. The result is that Me 6000 g molt.

This value is less than the estimated MeM12,000 g mol-' in the bulk; the comparisonmust however be treated with caution inview of the uncertainty in estimating Me inthe bulk. The main point is that Me in thethin film obviously is larger than the actualsize of the molecules.

Another suggestive calculation involvescomparing the bulk longitudinal modulus,

Log(w) (ed a-')Fig. 4. Log-log plot of storage and loss shearmoduli, G'(w) and GM(w), as a function ofangular frequency of measurement, for a film ofpolymer B of thickness 42 ± 1 A (3.2 RG).

M (from Fig. 1) and the shear modulus[estimated as G'(1 Hz)l, to estimate Pois-son's ratio (g.) by using standard rheologicalrelations for a continuum system (13). Pois-son's ratio indicates how much a materialcontracts when it is extended. When Mgreatly exceeds the shear modulus, it fol-lows that g. = 1/2 and that no change involume accompanies the deformation of a

material (13). The calculations show a

monotonic rise from g. = 0 (P1 = 0) to thelimiting value expected of a rubber,1/2 (the thinnest films). This we believe toreflect progressive drainage of unattachedchains out of the tethered structures.

What of the dependence on chainlength? Experiments with polymer Ashowed that at D < 4 R0, G'(w) reachedthe same limiting value as was observed forpolymer B at the same film thickness. Thegenerality of the effects described above was

thus confirmed.These are the central experimental re-

sults. Let us now restate the problem to beexplained. The low compressibility of con-

fined polymer melts has been discussed pre-viously (1) but strong shear elastic forcesare a qualitatively new phenomenon.

Should one suppose that polymer mole-cules in this narrow gap actually bridged thetwo solid surfaces, becoming physically at-tached to both at once? This structurewould produce elasticity in a fashion anal-ogous to the effect of cross-links in a bulkrubber (22). Simple calculations suggest,however, that the density of bridging wouldnot be high enough to explain the observedmagnitudes of elasticity.

What we believe to be a more likelymechanism comes from an analogy with bulkpolymers. Uncross-linked, linear polymers ofsufficiently high molecular weight also char-acteristically exhibit a zone on the frequencyscale of relatively slow relaxation, where themagnitude of the elasticity is similar to thelevels of G' measured here (13). The phe-nomenon is believed to reflect caging effects,

SCIENCE * VOL. 258 * 20 NOVEMBER 1992

traditionally referred to as entanglements;polymer chains, as they diffuse, can slide byone another but cannot cut across one anoth-er, and chain motions become highly corre-lated (23, 24). Similarly, in the system stud-ied here, chains tethered to one surface maybe blocked in their motion, over the experi-mental frequency range, because they cannotcut across other polymer chains that are teth-ered to the opposing surface. We suggesttentatively that the elastic effects observedhere can usefully be viewed as entanglementphenomena-although the chains are tooshort to be entangled in the bulk.

The motions and relaxations indicatedhere are thus qualitatively different fromthose that would characterize the bulk poly-mer. The frequency-dependent shear forcesshow the emergence, in these thin films, ofa strong elastic component that appeared toreach a limiting value typical of entangledor cross-linked chains in the bulk. The viewthat entanglement interactions are en-hanced in a restricted geometry finds theo-retical support (24). These observationshave evident bearing on understanding avariety of physical situations where con-forming surfaces separated by molten poly-mers come into close contact-especiallyadhesion, spreading, static friction, andlubrication.

REFERENCES AND NOTES1. P.-G. de Gennes, in Liquids at Interfaces: Pro-

ceedings of the Les Houches Summer School,Session XLVIII, J. Charvolin, J. F. Joanny, J.Zinn-Justin, Eds. (Elsevier, Amsterdam, 1990),pp. 296-326.

2. K. Kremer, J. Phys. (Paris) 47,1269 (1986).3. A. K. Chakraborty, Macromolecules 25, 2470

(1992).4. P. Silberzan and L. Lager, ibid., p. 1267, and

references therein.5. H. E. Johnson and S. Granick, Science 255, 966

(1992), and references therein.6. For a review, see D. Ausser6, J. Phys. France 50,

3021 (1989).7. K. F. Mansfield and D. N. Theodorou, Macromol-

ecules 22, 3143 (1989).8. R. G. Horn and J. N. lsraelachvili, ibid. 21, 2836

(1988).9. J. P. Montfort and G. Hadziioannou, J. Chem.

Phys. 88, 7187 (1988).10. R. G. Horn, S. J. Hirz, G. Hadziiouannou, C. W.

Frank, J. M. Catala, J. Chem. Phys. 90, 6767(1989).

11. J. N. Israelachvili, S. J. Kott, L. J. Fetters, J. Polym.Sci. Polym. Phys. Ed. 27, 489 (1989).

12. J. Van Aisten and S. Granick, Macromolecules 23,4856 (1990); S. Granick, Science 253, 1374(1991).

13. For a review, see J. D. Ferry, Viscoelastic Proper-ties of Polymers (Wiley, New York, ed. 3, 1980).

14. M. Denn, Annu. Rev. Fluid Mech. 22, 13 (1990).15. J. Van Alsten and S. Granick, Phys. Rev. Lett. 61,

2570 (1988).16. J. Peachey, J. Van Alsten, S. Granick, Rev. Sci.

Instrum. 62, 463 (1991).17. S. J. Clarson, K. Dodgson, J. A. Semlyen, Polymer

28, 189 (1987).18. S. J. Clarson, J. A. Semlyen, K. Dodgson, ibid. 32,

2823 (1991).19. J. E. Mark and J. H. Ko, J. Polym. Sci. Polym.

Phys. Ed. 13, 2221 (1975).20. S. M. Aharoni, Macromolecules 16,1722 (1983).

1341

000 000oo 0oOO 000* aia - ..*-O000oo0 000 000 00

0i0i 0.000

lnimamm~ 11111111111111111111111111

* * s 5 9 1

g"cwm..

on

Aug

ust 1

8, 2

011

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

21. Because of the parallel plate (rather than crossedcylinder) geometry, the Derjaguin approximationcommonly used in reporting surface forces mea-surements (4-6) did not apply. For this reason, wereport P_, force normalized by the contact area.

22. J. Klein and P. F. Luckham, Nature 308, 836(1984).

23. M. Doi and S. F. Edwards, The Theory of PolymerDynamics (Oxford Univ. Press, New York, 1986).

24. J. F. Douglas and J. B. Hubbard, Macromolecules24, 3163 (1991).

25. We are indebted to J. F. Douglas, J. D. Ferry, G.Reiter, and K. Schweizer for comments and dis-cussions and to S. J. Clarson for donating thePPMS samples used in this study. H.-W.H. ac-knowledges partial support from grant NSF-MSS-92-02143. We acknowledge primary support ofthe National Science Foundation through the Ma-terials Research Laboratory at the University ofIllinois, grant NSF-DMR-89-20538.

31 July 1992; accepted 5 October 1992

Direct Observation of CIO fromChlorine Nitrate Photolysis

Timothy K. Minton, Christine M. Nelson, Teresa A. Moore,Mitchio Okumura

Chlorine nitrate photolysis has been investigated with the use of a molecular beam tech-nique. Excitation at both 248 and 193 nanometers led to photodissociation by two path-ways, CIONO2 -- C10 + NO2 and CIONO2 -. Cl + NO3, with comparable yields. This

experiment provides a direct measurement of the C10 product channel and consequentlyraises the possibility of an analogous channel in C10 dimer photolysis. Photodissociationof the C10 dimer is a critical step in the catalytic cycle that is presumed to dominate polarstratospheric ozone destruction. A substantial yield of C10 would reduce the efficiency ofthis cycle.

The mechanisms of polar ozone depletionare well established (1). The main catalyticcycle is believed to be (1, 2):

C10 + C10 +M C100C1 +MC100C1 + hv COO + ClClOO +M Cl + 02 +M2(CI + 03 C10 + 02)Net: 203 302

The photochemical step in this cycle playsthe pivotal role of releasing atomic chlorinefrom the C1O dimer. This step requires thatthe Cl-0 bond [enthalpy of reaction AHW(0K) = 84 kj/moll breaks preferentially overthe weaker 0-0 bond [AHW(0 K) = 67kj/mol] (3), and it has been rationalized bythe argument that excitation of an n -- *transition localized on the CIO chro-mophore leads to selective dissociation ofthe CI-0 bond. In this report, we describean investigation of the photodissociation ofchlorine nitrate (CIONO2) that refutes thisargument and thus indirectly raises ques-tions about the photolysis products fromClOOCl.

The notion of selective CI-0 bond fis-sion has evolved from earlier observationsthat CIONO2 dissociates almost exclusively

T. K. Minton, Jet Propulsion Laboratory, Mail Stop67-201, California Institute of Technology, 4800 OakGrove Drive, Pasadena, CA 91109.C. M. Nelson, T. A. Moore, M. Okumura, Arthur AmosNoyes Laboratory of Chemical Physics, Mail Stop127-72, California Institute of Technology, Pasadena,CA 91125.

1342

to Cl and NO3 (2). CIONO2 has beenstudied extensively because of its impor-tance as a reservoir species for active chlo-rine and NO, in the stratosphere (4). Ear-lier studies (5-11) reported the dominantchannels to be

ClONO2 Cl + NO3CIONO2 ClONO + 0

The AH0(0 K) of these reactions are 167kJ/mol and 280 kJ/mol respectively. Whilediscrepancies exist in the relative yields forthese channels, none of the earlier studiesreported dissociation of the weakest bond,

CIONO2 -> ClO + NO2which has a AH0(0 K) of 109 kJ/mol,because the indirect methods used to detectCIO failed to reveal its presence. In anevaluation of data for use in stratosphericmodeling, DeMore et al. (12) preferred theresults of Margitan's experiment (10),which was a direct study of ClONO2 pho-tolysis at both 266 and 355 nm. With theuse of resonance fluorescence to detect Clatoms and 0 atoms in a flow cell, Margitanmeasured Cl and 0 quantum yields of 0.9 ±0.1 and 0.1 + 0.1, respectively. Theseresults have gained broad acceptance.

Although Margitan's experiment lendssupport to the picture of a transition local-ized on the CIO chromophore over anargument based on bond energies, the dis-crepancies among the earlier experimentsand their lack of detection of CIO impartuncertainty to the conclusion that the C1Ochromophore can be viewed as a distinct

SCIENCE * VOL. 258 * 20 NOVEMBER 1992

entity. Nevertheless, the ultraviolet absorp-tion spectrum (13) of CIONO2 shows twooverlapping and structureless absorptionfeatures that are reminiscent of absorptionsin the two chromophores that make up themolecule. The feature with a peak near 215nm may correspond to an n -- a* transitionlocalized on the CIO moiety (2), whereasthe rise near 190 nm may correspond to a wr

7r* transition on the NO2 group (14).Within the confines of this simple picture,each of two wavelengths available with anexcimer laser, 248 and 193 nm, shouldexcite one of the respective transitions.We examined the possibility of bond-

selective photolysis of the CIO chro-mophore in a molecular beam study of thephotodissociation of CIONO2 at 248 and193 nm. With the use of molecular beammethods, we could directly detect the phot-ofragments, identify primary and secondarydissociation channels, and quantitativelydetermine the relative product yields. Wedetected both CIO and Cl photoproducts(but not CIONO), and we determined therelative yields of these two products bycalibrating the detector sensitivity for CIOand Cl with a C12O photolysis experimentat 308 nm.

Our experiment utilized the techniqueof photofragment translational energy spec-troscopy (16) and was performed with auniversal crossed molecular beams machine(17) in which a pulsed excimer laser beamwas substituted for one of the molecularbeams (Fig. 1). The laser beam was focusedto a spot size of -0.1 cm2 where it crosseda molecular beam. The laser fluence wasvaried over one order of magnitude tocheck for a power dependence of the vari-ous signals observed. Only photoproductsthat recoil away from the molecular beamdirection can be detected by the mass spec-trometer detector, which can be rotatedabout the molecular beam-laser interactionzone. The distance from the interaction

/ i" 110-4 X

Fig. 1. Schematic diagram of the molecularbeam photodissociation experiment. The num-bers correspond to pressures in Torr for thevarious regions. 0 is the detector angle.

........ ............ ------------..0-

on

Aug

ust 1

8, 2

011

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om