Applications of X-Ray Photoabsorption Spectroscopy to the Determination of Local Structure in Organometallic Compounds

Transcription

1 8059 (28) B. Longto, F. Morndini. nd S. Bresdol, J. Orgnomet. Chem., 88, C7 (33) G. S. Arnold,. L. Klotz,. Hlper, nd M. K. DeArmond, Chem. Phys. (1975). Lett., 19, 546 (1973). (29) M. G. Clerici, S. Di Giocchino. F. Mspero, E. Perrotti, nd A. Znobi. J. (34) A. P. Ginsberg nd B. K. Teo, presented t the 169th Americn Chemicl Orgnomet. Chem., 84,379 (1975). Society Ntionl Meeting, Phildelphi, P., April (30) C. T. Enos, G. L. Geoffroy, nd T. H. Risby, Anl. Chem., 48, 990 (35) B. K. Teo, A. P. Ginsberg, nd J. C. Clbrese, J. Am. Chem. Soc., 98,3027 (1976). (1976). (31) For other exmples, see ref 19. (36) A. P. Ginsberg, privte communiction. (32) M. J. Mys, R. N. F. Simpson, nd F. P. Stefnini. J. Chem. SOC. A, 3000 (37) Estimted from dt given by L. Vsk nd M. F. erneke, Trns. A!. Y. Acd. (1970). Sci., 33, 70 (1971). Applictions of XRy Photobsorption Spectroscopy to the Determintion of Locl Structure in Orgnometllic Compounds S. P. Crmer, T. K. Eccles, F. Kutzler, Keith 0. Hodgson,* nd S. Donich* Contribution from the Deprtment of Chemistry nd Deprtment of Applied Physics, Stnford University, Stnford, Cliforni 94305, Received Februry Abstrct: Curve fitting methods hve been developed which promise to be of considerble utility in extrcting structurl informtion from xry bsorption spectr. n principle, the extended xry bsorption fine structure (EXAFS) for given tom cn be nlyzed to revel the rdil distnces nd tomic numbers of those toms surrounding the bsorber. The requisite highqulity xry bsorption spectr cn now be esily nd rpidly obtined on solid or solution smples, using synchrotron rdition from the Stnford electronpositron storge ring SPEAR t the Stnford Liner Accelertor Center. n order to provide the bsis for n understnding of metlloprotein bsorption fine structure, dt for ferrocene, CO(NHJ)~~+, Co(OH&*+, nd vriety of iron porphyrins hve been nlyzed. The procedure used ws to fit the EXAFS of known structures with prmet erized, semiempiricl function: ~,C,kfi~ exp(2( u12)k2) sin (2R,k + (k)); where the sum is over ll neighboring toms, the distnce informtion R, is contined in the phse term, nd the other vribles ccount for the mplitude behvior s described in detil in the text. These prmeters re then used in fits on other known structures, nd the ccurcy of the clculted distnces (R,) is used s test for the vlidity of the method. The results of fits on FelTPP, Fe1"TPP(SC6H5), nd FeTpivPP(NMe1m)Oz indicte tht rdil distnces out to four coordintion shells in iron porphyrins cn be determined to better thn 0.1 A. For cses where the interest is only in chnges in coordintion, difference methods of nlysis hve been developed which llow ccurte determintion of the geometry nd distnces to smll molecules such s dioxygen which cn reversibly bind to the metl ion. For oxy nd deoxy FeTpivPP(NMem), the Fe0 distnces re determined to better thn 0.03 A. At present, the interprettion of the EXAFS from n unknown structure relies hevily on comprison with model compound dt. However, the present results suggest tht librry of trnsferble phse shift nd mplitude functions my be eventully compiled, nd tht curvefitting nlysis of the extended fine structure will be useful for determining the locl structure round specific tomic constituents in metlloproteins nd other complex mterils. The recent vilbility of stble, intense xry source using synchrotron rdition from high energy (GeV) electron storge rings' hs generted renissnce of interest in xry bsorption spectroscopy. Following the bsorption edge, where the photon energy equls the binding energy of core electron, simple models of xry photobsorption predict grdul monotonic decrese in the bsorption coefficient with incresing photon energy.2 Although such behvior is pproximtely observed for noble gs spectr, the presence of other toms round the bsorber cuses the ppernce of periodic modultion in the bsorption coefficient. This modultion contins informtion bout the nture of the surrounding toms nd their distnces to the bsorber. The bsorption coefficient oscilltions hve been termed EXAFS (Extended Xry Absorption Fine Structure) or lterntively Kronig structure, nd until recently they were more theoreticl curiosity thn useful structurl tool3 t is the purpose of this pper to illustrte the ppliction of xry bsorption spectroscopy to the study of the locl environments round specific metl centers in complex molecules. Specificlly, in n erlier communiction4 we suggested tht xry bsorption spectroscopy might be useful for structurl investigtion of molecules such s metlloporphyrins. e consider herein detiled ppliction of this method to the study of the moleculr environment of the iron in porphyrins. hile it is esy to show tht the extended fine structure is relted to the rdil distribution of toms round the bsorbing tom, there is considerble difficulty in interpreting the dt in quntittive mnner. The extended fine structure ws first treted theoreticlly by Kronig in the 1930'~.~ More recent theoreticl tretments of EXAFS by Ashley nd Donich,6 Lee nd Pendry,' nd Syers nd Stern8 hve emphsized the effect of shortrnge photoelectron bckscttering on the trnsition probbilities to the continuum sttes. Multiple scttering effects hve lso been treted by these uthors. hile theoreticl clcultions cnnot s yet produce perfect greement with experimentl results, they provide useful foundtion upon which to bse empiricl methods. Since EXAFS is essentilly result of the photoelectron bckscttering from the surrounding toms, it contins informtion bout the nture of these toms nd their distnces from the bsorbing tom. Unfortuntely, this informtion is veiled by the energy dependence of the electrontom bckscttering function. nelstic processes my lso complicte the fine structure. Recent ttempts to unrvel the structurl informtion from EXAFS hve involved three types of pproches: (1) b initio clcultions, (2) Fourier trnsformtion, nd (3) curvefitting. Using shortrngeorder electrontom bckscttering pproch s strting point, Kincid nd Eisenberger9 hve Crmer, Eccles, Kutzler, Hodgson, Donich / XRy Photobsorption Spectroscopy

2 8060 ttempted to clculte n EXAFS spectrum b initio for Br2 nd GeC14 vpors. Electrontom scttering mplitudes nd phse shifts were clculted by numericlly integrting Schrodinger s eqution. Using the bond lengths nd strengths obtined from vibrtionl spectroscopy nd electron diffrction, they then clculted theoreticl EXAFS spectr nd compred the results with the experimentl dt. Although greement ws obtined to within 5% for the phse of the scttering, quntittive disgreement beyond experimentl error ws observed for the mplitude. The results suggest tht the bsic theory of the extended fine structure is correct s fr s elstic scttering goes, but tht b initio clcultion of spectr will lso hve to include inelstic effects. A Fourier trnsform nlysis procedure hs been used extensively by Syers, Stern, nd Lytle to study the short rnge order of morphous mterils. O Comprison of the extended fine structure of morphous nd crystlline germnium clerly shows the decrese in order beyond the first coordintion sphere in morphous Ge. Similr results re observed with morphous vs. crystlline Ge02. Fourier trnsform nlyses hve been crried out on copper porphyrin.l Syers, Stern, nd Lytle hve lso exploited the temperture dependence of the extended fine structure to obtin men squre devitions in tomtom distnces, nd they use their methods for estimting coordintion numbers, electronscttering mplitudes, nd photoelectron men free pths.] b,c The curve fitting methods reported here represent n lterntive pproch for obtining quntittive informtion from extended fine structure dt. e recognize tht both the phse nd mplitude of the fine structure modultion my be photoelectron energy dependent functions. e use minimiztion curvefitting technique to fit the observed spectr of known compounds to semiempiricl formul, expressed in terms of limited number of prmeters. These prmeterized functions my then be pplied to curve fitting spectr of other compounds, leding ultimtely to the solution of unknown structures. Relted methods of nlysis re being developed independently by Shulmn, Eisenberger, Blumberg nd Stmbugh.2 Using ironsulfur phse shifts derived from model compounds, they fitted dt on lyophilized, oxidized rubredoxin with nineprmeter mplitude function. A sttisticl nlysis of the phse of the fine structure reveled tht the FeS distnces were very close, if not equivlent. The results reported in the present pper suggest tht by using our pproch to compre spectr of closely relted compounds, precise nd relible distnce informtion my be obtined bout the environment of the bsorbing tom, out to distnce of 45 A. Both phseshift nd mplitude functions pper to be trnsferble between similr compounds. Finlly, the fine structure for complicted molecule my be pproximted s the sum of contributions from the individul toms. This dditivity of EXAFS is essentilly neglect of multiple nd inelstic scttering processes, nd such n pproximtion ppers vlid when the bsorber is surrounded by low Z scttering toms, none of which re obscured from direct interction with the bsorber. Experimentl Section Mterils. All mterils were prepred ccording to literture methods. mesotetrphenylporphyriniron(), Fe(TPP), ws prepred by the method of Collmn et.l3 mesotetrphenylporphyrinbis(nmethyl(imidzole))iron(), Fe(TPP)(NMe(imid))z, ws obtined from Fe(TPP) by the method of Collmn nd Reed.14 mesotetrphenylporphyrinbis(imidzole)iron() chloride, Fe(TPP)(imid)fJ, ws prepred by the method of Epstein, Strub, nd Mricondi. j The sulfur lignded porphyrin, mesotetrphenylporphyrin(benzenethiolto)iron(), Fe(TPP)(SC6HS), ws prepred by the method of Collmn, Sorrell, nd Hoffmn.6 Regent grde hemin chloride, Fe(PPX)C, ws purchsed from Fluk AG. The oxy picket fence porphyrin, dioxygen [mononmethylimidzolemeso tetr(,,,0pivlmidophenyl)porphyrin]iron( ), Fe(,,,TpivPP)(NMe(imid))02, ws generous gift from Professor J. P. Collmn of the Stnford Chemistry Deprtment. The deoxygented nlogue of this porphyrin Fe(,,cY,TpivPP)(N Me(imid)) ws prepred s described by C~llmn ~ by deoxygention of the solid oxygented porphyrin t 68 OC for 18 h. The deoxygention rection ws monitored by recording the solution spectr in the uv nd visible regions. Ferrocene * nd Co(NH3)&139 were lso prepred by stndrd procedures. XRy Absorption Mesurements. All xry bsorption spectr were recorded using rdition from the Stnford Positron Electron Accelerting Ring (SPEAR) nd the fcilities of the Stnford Synchrotron Rdition Project, SSRP, ntionl fcility for the exploittion of synchrotron rdition.20 The mesurements were mde using n pprtus originlly constructed by Kincid, Eisenberger, nd Syers which hs been described in detil elsewhere.* n essence, the synchrotron rdition is defined by slits nd monochromted by Si(220) chnnel cut single crystl. The monochromtic 1 X 20 mm xry bem psses through one ion chmber to mesure incident intensity (O), then goes through the smple, nd finlly through second ioniztion detector which mesures trnsmitted intensity (). The rtio /o is recorded s function of monochromtor ngle by mens of n interfced PDP 11/05 computer. Porphyrin smples were mesured s solid pellets pproximtely 0.1 mm thick. Ferrocene, Co(NH3)&3, nd CoCl2.2H20 were run s 1 M solutions, the former in benzene, the ltter two in wter. The porphyrin spectr reported herein were typiclly recorded with 1 s per point integrtion with steps being tken pproximtely every 1 ev beginning round 50 ev below the bsorption edge to round 600 ev bove the edge. Lck of ny detectble smple decomposition ws demonstrted by recording the xry bsorption spectr of the sme smple repetedly during period of severl hours. For exmple, with Fe(,,,TpivPP)(NMe(imid))O2, no detectble chnges were observed in the spectr during repeted mesurements of the sme smple over 4h period.22 The deoxy picket fence porphyrin ws prepred s described bove nd trnsferred into gstight cell with Kpton windows in n inert tmosphere box. The cell ws stored under rgon until just prior to recording the spectrum. Numericl Anlysis Methods. The numericl clcultions described in this pper were performed on PDP 11/45 computer with 28 K of core nd floting point processor. The clcultions routinely involved Fourier trnsforms nd minimiztion progrms which were tilored for use by the smll computer.23 Lest bsolute vlue refinements were used rther thn lest squres becuse they converged much more rpidly. Severl lestsqures refinements were done, however, nd these showed tht the constnts rrived t by ech method for the sme fit were within one stndrd devition of ech other. The only difference in the progrmming between the two methods ws the substitution of bsolute vlue for the vlue squred in the function to be minimized. The errors were estimted from the vrince by estblished procedures. For exmple, the error in thejth prmeter, u,, obtined by the fitting procedure ws estimted by: wheren is the number ofdt points used in the fit, n the number of prmeters djusted, xi nd F, dt point vlue nd its corresponding vlue of the fitting function, nd j,, digonl element of the inverse mtrix of norml equtions weighted by k3. The Fourier trnsform progrm ws used in both Fourier filtering nd in ctul dt nlysis s described in the next section of this pper. The rel nd imginry components of the Fourier trnsform were obtined using the following formuls: &el(nrlnc) = 5 Jk k3x(k) 7, cos (2nR,,,k) dk where n is n integer nd Ri,, is the distnce between points in R spce, nd x(k) is the vlue of the normlized fine structure. The integrtions for the Fourier trnsforms were done numericlly by Journl of the Americn Chemicl Society / 98:25 / December 8, 976

3 806 1 Simpson s rule. The modulus of the Fourier trnsform (which is plotted in the trnsforms in this pper) is obtined by the formul &,d = (4re12 *. The Simpson s rule integrtions lso mde it possible to include qudrtic term in the trnsform, tht is cos (2nRi,,k + Ck*). However, the investigtion of the effect of this extr term ws not pursued extensively. Dt Reduction For the purpose of dt nlysis, it is necessry to obtin normlized fine structure s function of photoelectron wve vector in A (k = d (E Eedge) ev). n order to plot the fine structure s function of wve vector, n edge position must first be defined. To determine this position, the xry bsorption dt re plotted s log (Z0/Z) (bsorbnce) vs. energy, long with the first nd second derivtive curves. Figure 1 shows dt from the oxy picket fence porphyrin, Fe(,,,TpivPP)(NMe(imid))(02), nd the first derivtive of these dt. The derivtive curves re clculted by fitting polynomil to five points on either side of ech individul point nd differentiting the polynomil. Such curves re useful in determining the positions of peks nd inflection points of the edges. For the purposes of this pper the edge energy ws usully defined s the energy of the highest inflection point. Errors of up to 5 ev in edge position s defined in this mnner will fll to within <5% in ssigning the k vlue for k > 3.6 Al. Obvious glitches in the xry bsorption dt cn be removed where necessry by fitting polynomil over certin rnge nd replcing the dt vlues with the polynomil vlues. Such procedures re normlly necessitted by the fct tht rpid chnges in incident xry intensity occur t ngles where the 220 monochromting crystl hs other resonnces. Following this procedure, the dt re converted to bsorbnce vs. wve vector k nd rbitrrily normlized so tht the bsorbnce vlues lwys fll between 0 nd 1. The bsorption dt must then be converted to fine structure. This cn be ccomplished by fitting polynomil (of order from 1 to 3) over some rnge of the bsorption curve, nd subtrcting it from the dt.2s One of the min problems in bckground subtrction is tht this bckground represents tomic bsorption in the bsence of neighbors. Since dt for the bsorption in the bsence of neighbors re not normlly vilble, the correction for bckground remins somewht rbitrry. Another bckground subtrction scheme used Fourier filtering1bsc of the dt with respect to k fter the polynomil subtrction. Fourier filtering is ccomplished by Fourier trnsforming the dt into R spce, selecting the R rnge to be kept, nd retrnsforming to k spce. The Fourier filtering cn be used to remove lrge peks in the R rnge of 0 to 1 A whose position nd height vry s function of the bckground polynomil fitting rnge nd order. Peks in the trnsform bove 1 did not show this vrition, nd thus the peks below 1 A were ttributed to residul bckground. As significnt peks in the Fourier trnsforms generlly come below R of 5 A, high frequency cutoff of 7 A is resonble. Figure 2 shows comprison between filtered nd unfiltered dt for Fe(,c~,,TpivPP(NMe(imid))02 in which the filtered dt were Fourier trnsformed to R spce, cut off below R of 0.5 8, nd bove R of 7.0 A, nd retrnsformed to k spce. n ddition to eliminting low frequency drift of the bseline, reduction in the high frequency noise is clerly visible. Figure 3 shows the ctul bckground line subtrcted by ech method. The possible effects of cutoff errors introduced by the trnsform opertion itself re considered lter by compring numericl results on the sme dt set for both filtered nd unfiltered dt. n essence, t vlues of k 2 4 8, where the fitting is normlly done, the simple polynomil scheme nd unfiltered dt worked very well for the porphyrin dt nlysis XRAY ENERGY EV) Figure 1. XRy bsorption spectrum for Fe(,,,TpivPP)(N Me(imid))Oz in the region of the Fe K bsorption edge. The solid line is plot of the bsorbnce vs. energy while the dotted line is the first derivtive of this curve. N r X, u z m [r w m v PHOTOELECTRON AVE VECTOR k (? Figure 2. A comprison of filtered (upper curve) nd unfiltered (lower curve) extended fine structure spectr for Fe(,,,TpivPP)(N Me(imid))Oz. The filtering ws done by trnsforming the dt into R spce, cutting off t R of 0.5 nd 7 A, nd retrnsforming to k spce. Methods of Anlysis nd Results Numericl Anlysis. As expected from the single scttering theory6 the energy dependent modultion x of the xry bsorption coefficient 1.1 of tom X by n tom Y t distnce Rxy is given semiempiriclly by the generl expression: x(k) E 1.10 = 1 2f(7r,k) sin (2kR,, + (k))e2(u2)k2 krxy where k is the photoelectron wve vector;f(r,k) is the bckscttering mplitude of the outgoing electron from neighbor t distnce Rxy; (k) is the totl phse shift undergone by n electron leving the K shell, bckscttering, nd returning to interfere t the K shell; ( oi2) is the men squre therml motion long R ~Y. This expression is expected to be resonble pproximtion t high photoelectron energies, bove 50 ev, in which cse multiple scttering terms re smll nd scttering mplitudes re smoothly vrying. The crucil functions to know for the present nlysis ref(s,k) nd (k), since we find tht in prctice the dmping due to therml motion cn be treted by single prmeter nd is firly well understood. The clcultions of Kincid26 nd others predict tht for low Crmer, Eccles, Kutzler, Hodgson, Donich / XRy Photobsorption Spectroscopy

4 V z LL % 0.80 N z XRAY ENERGY (EV) Figure 3. A comprison of the totl bckground removed by two different bckground subtrction schemes. The solid line shows the bsorbnce vs. energy spectrum for Fe(,,,TpivPP)(NMe(imid))Oz. The dshed ( ) line is the bckground subtrction curve obtined by fitting third order polynomil over the rnge of The other bckground curve (. e. ) ws obtined by Fourier trnsforming the dt into R spce (fter the thirdorder polynomil removl), cutting t R of 0.5 nd 7.0 A, nd retrnsforming into k spce. Z elements (e.g., C, N, 0) the bckscttering mplitude should pek below 100 ev nd then drop off monotoniclly. t lso ppers tht the totl phse shift will be smoothly vrying function beyond k = 4 Al. e decided to model the totl mplitude behvior by function bsed on the form Ce2(u12)2k2/k. The exponentil fctor pproximtes the dmping due to therml motion. The fctor l/k in this mplitude function includes the 1 / k fctor from the generl expression s well s the bckscttering mplitude f(r,k). A qudrtic vrition is ssumed s the model for the totl phse shift ~ ( k = ) uo + lk + 2k2 This gives rise to the following functionl form for the extended fine structure: 0 z 3 2 m 0.5 n N J 5 z C ENERGY (EV) Figure 4. XRy bsorption spectrum of ferrocene. where the sum over i shells indictes tht equivlent toms t the sme distnce from the bsorber re treted s group. The distnce informtion is included in the liner term (2R; + l;)k during the fitting process, for convenience. The distnce is then extrcted t the end if the liner k dependent term [; is known. To test the vlidity of our functionl form, we used dt from the solution spectr of ferrocene, CO(NH~)~~+, nd Co (OH2)6*+. The results of curve fitting this functionl form to the dt re summrized in Tble. Single Shell Results. The xry bsorption spectrum of ferrocene is shown in Figure 4. The inflection point nerest the top of the steep rise occurs t ev nd this point ws tken s the Fe K bsorption edge. The bckground ws removed by fitting thirdorder polynomil over the rnge ev nd subtrcting it from the dt. The Fourier trnsform of this dt is shown in Figure 5. The dt were Fourier filtered by trnsforming it from the rnge of 2 12,& nd retrnsforming from the R rnge 0.57 A. One shell fits were done on both filtered nd unfiltered dt, nd the results of the fits nd Fourier trnsforms re summrized in Tble. The errors shown in prentheses re the estimted stndrd devitions. Note tht when /3, 2, nd 2( i2) re fixed t 2.0, 0.0, nd respectively, the stndrd devition on the remining constnts is much lower, nd still shows consistent (2R + l) vlue. A single shell fit of the ferrocene dt is shown in Figure 6. The fit nd dt shown in this figure hve been Tble 1. Results of Single Shell Fits on Ferrocene, [Co(NH&J3+ nd [CO(OH&]~+ Dt set k rnge (A) No. of points (2R + ]); A P, 2 2(u,?) Ferrocene (unfiltered) (21) 2.15 (55) (010) (010) (06) 2.0 (0) 0.0 (0) (0) (14) 2.13 (45) (0100) (5) (02) 2.0 (0) 0.0 (0) (0) Ferrocene (filtered) (20) 1.93(51) 0.004(18) (1) (05) 2.0 (0) 0.0 (0) (0) (14) 1.90 (45) (O) (5) (02) 2.0 (0) 0.0 (0) (0) [Co(OH?)& 3+ (unfiltered) (26) 1.71 (91) (21) (12) (04) 2.0 (0) 0.0 (0) (0) [Co(NH3)6l3+(unfiltered) (18) 1.96 (63) (O) (7) (03) 2.0 (0) 0.0 (0) (0) (' The numbers given in prentheses represent the estimted stndrd devition on the numbers in the lst digit or digits s so indicted. A stndrd devition of zero indictes tht the vlue ws fixed in the refinement. Journl of the Americn Chemicl Society / 98:25 / December 8, 1976

5 8063 R dl Figure 5. Fourier trnsform of the ferrocene xry bsorption dt shown in Figure 4 fter the bckground ws removed by fitting thirdorder polynomil over the rnge of ev. The dt were trnsformed over the rnge of k = 212 A'. multiplied by k2 in order to show the goodness of fit in the high k region s well s ner the edge. As expected, ner the edge the fit is poor since the dt were not fitted over this rnge, nd becuse the single scttering theory is not pplicble close to threshold, but it is much better over the k rnge of 412 A1. The verge (2R + l) vlue from the ferrocene one shell fits of 3.31 A grees resonbly well with the vlue of two times R of 3.36 A from the ferrocene Fourier trnsform. This suggests tht pproximting (k) s liner function of k is resonble, t lest over the rnge of k = 412 Al. t is lso seen tht the vlues of (2R + l) re more dependent on the rnge of fit for the unfiltered dt, with the lowest (2R + [) vlue being shifted to considerbly higher vlue for the fit done using unfiltered dt over the rnge of k = 212 A. The estimted stndrd devition of (2R + l) s estimted from the vrince of ll of the different fits of ferrocene in Tble is 0.12 A. This vlue grees well with the vlues estimted from the fitting routine nd together these suggest limit on the ccurcy of the dt from this prticulr nlysis of bout 0.1 A. Using floting exponentil fctors nd qudrtic phse shifts, the power of (1/ k) in the mplitude function ws optimized for ferrocene, Co(NH&?, nd CO(OH&~+. e thus rrived t respective powers of (1 / k) of 2.13, 1.96, nd for P prmeters of crbon, nitrogen, nd oxygen, respectively. The generl trend is for bckscttering mplitudes to die out more slowly s the tomic number of the sctterer increses. This is in ccord with the predictions of Kincid bsed on electrontom scttering clcultions. (These clcultions lso predict pek in the scttering mplitude t low photoelectron energies, but our fits begin fter the monotonic decline in f(k,7r) hs strted.) e cn therefore conclude tht for low Z sctterers on the rnge of k = 412 A, the totl mplitude function cn be pproximted by simple c/k2 fctor times n exponentil term. The phse shift is lso sufficiently liner in this rnge to justify neglect of the qudrtic term. For higher Z scttering toms, more complicted empiricl functions will be needed, in order to ccount for the peking in electrontom scttering mplitude t higher energies. Furthermore, for better fits to the phse on wide rnges of dt, extr nonliner prmeters will hve to be introduced. Also, the trend in exponentil dmping terms (from for ferrocene to for queous Co(1)) clerly shows the effect of bond strength on the fine structure mplitude. The fct tht the therml dmping prmeters increse s the metllignd bond strength decreses gin suggests tht our , PHOTOELECTRON AVE VECTOR k (') Figure 6. Single shell fits on ferrocene (upper curves) nd the [Co (NH,)63+ complex in solution (lower curves). n both cses, the semiempiricl function (described in the section on single shell dt nd the functionl form) ws fit over the rngeof k = 412 A' using unfiltered dt. As expected, the functionl form does not give good greement ner the edge t low k. semiempiricl functions ctully hve physicl foundtion. Experimentlly, it becomes evident tht low temperture dt collection will be essentil to extending the energy rnge of observble fine structure. The filure of our functionl form for k < 4.0 A (Ephotoelectron < 61 ev) is not surprising, for both the phse nd mplitude of the bckscttering function exhibit complicted behvior in the low energy regime. A lrge number of extr prmeters would hve to be introduced to obtin perfect correspondence with the observed fine structure, nd structurl informtion would be lost rther thn gined by extr fitting becuse of prmeter correltion. Correltion of fitting prmeters exists becuse fine structure dt were tken over firly limited rnge in kspce (typiclly discernible out to k = 12 A' with low Z sctterers nd the present signltonoise rtio). The phse nd mplitude functions re not completely independent on such limited rnge. Nevertheless, over the rnge of k = 412 A, the fine structure cn be pproximted by firly simple function. Crbon, nitrogen, nd oxygen pper to hve smoothly decresing bckscttering mplitudes, nd the totl phse shifts re lmost liner for these simple model compounds. This permits significnt reduction in the number of prmeters used, nd it permits the extension of curve fitting methods to multishell fits of xry bsorption fine structure dt for molecules such s metlloporphyrins. FourShell Fits of ron Tetrphenylporphyrins. The iron tetrphenylporphyrins give beutiful fine structure observble out to k = 12 A. The first derivtive curves of the porphyrin dt in the vicinity of the edge show tht there re two inflection points in the steep rise (see Figure 1). The energy of the inflection point nerest the top of the steep rise ws verged from ll the porphyrin dt sets to give vlue of ev, which ws tken s the zero of photoelectron energy. Severl bckground subtrction schemes were tried for the porphyrins, involving fitting polynomils of vrious orders over vrious energy rnges. The nlysis presented here compres the two schemes tht gve the most consistent results. The simplest method to give consistent results ws to subtrct thirdorder polynomil fitted to the log (ZdZ) dt s function of energy over the rnge 7200 ev to the end of the dt set (typiclly 7725 ev). The second method ws to subtrct thirdorder polynomil fitted over the rnge 7325 ev to the end Crmer, Eccles. Kutzler, Hodgson, Donich XRy Photobsorption Spectroscopy

6 8064 Tble 11. Dt Results of FourShell Fit on Fe(TPP) Fine Structure Rdil distnce (2R + l), A Rcry*tb 0, (A) FeN (51) FeC,, 5.01 (15) FeC, 5.86 (15) Fe C,j 7.35 (26) ( The bckground ws removed by fitting thirdorder polynomil over the rnge of ev nd subtrcting. The fit ws done over the rnge of k = 41 2 A. J. P. Collmn, J. L. Hord, N. Kim, G. Lng, nd C. R. Reed, J. Am. Chem. SOC.. 97,2676 (1975). Cl culted by using the crystllogrphic distnce nd the vlues of (2R + l) from the fit. V PHOTOELECTRON AVE VECTOR k (i ) Figure 7. Fourshell fit ( ) over the rnge of k = ,l to the unfiltered fine structure dt of Fe(TPP) (). The fit used fixed vlues for the Debyeller fctor, power on k in the mplitude function, nd qudrtic term, which were obtined from the single shell fits shown in Figure 6. of the dt set, then Fourier trnsform the dt s function of k over the k rnge 212 Al, nd trnsform bck from the R rnge 0.57 A. Becuse of the moleculr symmetry, ech metlloporphyrin structure my be viewed s shells of scttering toms grouped t equl distnce from the centrl iron. The mgnitude of the contribution to the fine structure decreses rpidly with incresing distnce from the bsorbing tom, so the bsorption spectr of compound such s Fe(TPP) my be pproximted by the sum of four wves due to: (1) the coordinted nitrogens, (2) the (Y crbons (nd xil imidzole CY crbons if present), (3) the bridging y crbons, nd (4) the p crbons. Ech of the four wves hs six constnts to be determined: the mplitude C,, the Debyeller fctor 2( i2), /3; (the power on k in the denomintor of the mplitude function), nd the three coefficients in the sine rgument, o,, l;, nd 2i. Refining 24 constnts in fourshell fit would be next to meningless. The results from single shell fits described erlier indicted tht good choices 2( i2), nd 2 would be 2.0, 0.015, nd 0.0, respectively for the rnge over which the fits were to be done. These vlues were held fixed in ll fits of the porphyrin dt thus reducing the number of vribles to three per shell. A number of fourshell fits were done on Fe(TPP) using bout 80 smple dt points in ech fit tken from the k rnge 412 A1. The result of the Fe(TPP) fit is summrized in Tble 11. Figure 7 shows the greement between observed nd clculted spectr for this Fe(TPP) fit. Note tht the Figure 8. Fourier trnsform over the rnge of k = 412 8, of the Fe(TPP) fine structure dt. The first six peks beyond R of 1 8, occur t 1.68, 2.30, 2.88, 3.44, 3.80, nd ,. stndrd devition of the (2R + l) vlues increses substntilly for the outer shells. The 1 vlues, clculted by compring (2R + ]) with the crystllogrphiclly determined R, re consistent for the three crbon shells to within one stndrd devition. This consistency, s well s the overll goodness of fit, strongly suggests tht the functionl form used for the fit is vlid. The strting vlues for C1 nd (2R + ) for ech wve were obtined from reltive pek heights nd pek positions in the Fourier trnsform of the respective dt set. For comprison with the results of the multishell fits, the Fourier trnsform of the Fe(TPP) dt over the rnge of k = 412 A1 is shown in Figure 8. The min problem with the identifiction of peks in the Fourier trnsform with vlues of (2R + ]) is tht subsidiry side lobes my be expected for ech min pek in the trnsform. For weker, more distnt shells there is confusion between shell primry pek nd side lobe ssocited with stronger shell. There is lso the problem tht skewing of the trnsformed envelope due to the nonliner chrcter of the phse nd mplitude my rtificilly shift the vlue of the pek in (2R + l). At this point, it becme pprent tht the present rnge of dt (nd the theoreticl understnding thereof) did not justify the inclusion of more prmeters to simulte the experimentl results. e thus embrked on progrm of fine structure difference nlysis, in which we hoped to correlte chnges in EXAFS with welldefined chnges in porphyrin structures. Difference Fourier Anlysis Using the Fourier trnsform method, we hve tried to detect the presence of xil lignds nd mesure the Fe to lignd distnce in the iron porphyrin complexes. Unfortuntely, with the present rnge of dt, we cnnot lwys directly resolve the xil lignd distnce (for exmple, 02, C1, or SC~HS) from the iron porphyrin distnces. ith dt recently cquired over wider rnge in k spce, this resolution hs become possible. Comprison of dt from two similr compounds, however, cn yield informtion bout the structurl differences between the two, ssuming tht the component bckscttered wves from ech neighboring tom re independent, nd tht the resultnt fine structure is due to simple sum of these wves. n the cse of porphyrins, for exmple, if the Fe(TPP) dt re ppropritely subtrcted from the dt of n iron porphyrin which hs xil lignds, the difference should be the fine structure from the xil lignds lone. But, often the structurl differences between two iron tetrphenylporphyrins include the shift of the iron tom reltive to the plne of the porphyrin, which lengthens the distnce to ech rdil shell. f inplne porphyrin fine structure is subtrcted from the outofplne fine structure, the difference resembles the verge of the two, but phseshifted by bout ~ / (since 2 (sin x sin y ) equls 2 cos /~(x + y ) sin h(x y)). Also contributing to the porphyrinlike ppernce of the difference Fourier is the fct tht it is difficult to scle the two dt sets perfectly when sub Journl of the Americn Chemicl Society / 98:25 / December 8, 1976

7 8065 trcting one from the other. hen the difference in two porphyrins involves xil lignds s well s n outofplne shift, the difference in the fine structure will contin the sum of the two effects. Results from simulted dt hve shown tht these effects my contribute eqully to the difference trnsform. e refer to this Fourier trnsform of the difference of two dt sets s difference Fourier. The results of the difference Fourier for Fe(PP1X)Cl minus Fe(TPP) re shown in Figure 9. The one predominnt pek in the difference Fourier is from the extr wve in the hemin chloride bsorption spectrum which is due to bckscttering by the chloride ion. More complicted difference Fouriers re obtined when there re xil lignds with more complicted structures. As mentioned before, this method is not without some pitflls porphyrinlike (four or five wve) residul will be found in the difference Fourier if () the scling of the two dt sets is incorrect, or (b) if substntil chnge in iron porphyrin geometry occurs. Such residul is observble to some extent in ll the difference Fouriers. A most interesting nd useful difference Fourier (shown in Figure 10) is obtined from the comprison of fine structure of the oxy nd deoxy forms of Fe(,,,TpivPP)(N Me(imid)) dt sets. Although there is considerble bckground structure due to chnge in the iron(1)porphyrin distnces, the primry components in the difference Fourier (first two significnt peks) correspond to the ddition of dioxygen lignd with two different ironoxygen distnces. The pproprite phse shifts (vlues of ~ for ) Fe0 scttering re uncertin, so tht clcultion of bsolute ironoxygen distnces is not possible. However, the difference between the two Fourier ironoxygen pek positions should correspond to the difference between the distnce of the Fe toms to the bound oxygen, 0, nd the unbound oxygen, 0. The difference Fourier gives vlue of (Fe0) (Fe0 ) of 1.0 8,. This difference compres extremely well with the crystllogrphiclly mesured difference of l.04 As expected, it cn be seen tht the peks of 2.9, 3.6,5.1, nd 6.2 8, in the difference trnsform occur t vlues which re within bout 0.1 8, of peks in the stright Fourier trnsform of Fe(TPP) (compre Figures 8 nd 10). Becuse of the possible mbiguities ssocited with overlpping peks in the difference Fourier trnsform nd the possibility of nonliner phse shift effects, fitting methods were used to further quntitte xil lignd distnces nd chnges in ironporphyrin geometry on ligtion. Comprison Fit Anlysis n order to roughly determine the movement of the iron tom s it binds xil lignds, s well s to mesure the iron to xil lignd distnces, procedure clled comprison fitting ws developed. Dt from one of the two molecules being compred re fitted with fourshell function s described erlier. The prmeters from this fit which correspond to the three crbon shells in the porphyrin nd the 0 constnt from the nitrogen shell re held fixed in fitting the dt from the other molecule. Extr wves re dded to ccount for xil lignds. The form of the function used to fit the second molecule is then: x(k) = Ci e0.015k* sin [0 + (2R + l)k] + c2(3 crbon shells) + extr wves C, (2R + l), C2, nd the prmeters in the extr wves re refined, nd 0 nd the prmeters in the three crbon shells, which come from the fit of the first molecule, re held fixed, Two comprison fits were done using Fe(TPP) s the stndrd. Dt from Fe(PP1X)Cl nd Fe(TPP)(SC6H5) were both fit using the 0 from the nitrogen shell nd ll the prmeters 03 t 1 : 0 LL v) z : k : w : R (A, Figure 9. Difference Fourier trnsform over the rnge of k = 4 12 A for Fe(PP1X)C minus Fe(TPP) fine structure dt. The pek t R of 1.78 A which predomintes results from the presence of the C lignd. \ OXY MNUS DEOXY R 6, Figure 10. Fourier trnsforms (ll over the rnge of k = 41 2 A ) of oxy nd deoxy Fe(cu,cu,cu,cuTpivPP)(NMe(imid)) dt. The upper trnsform shows the rdil distribution for the oxy complex, nd the middle trnsform for the deoxy complex. The lower difference trnsform ws obtined by normlizing the dt sets to one nother nd then trnsforming the difference. The peks t 1.5 nd 2.5 A occur becuse of the presence of the symmetriclly bound dioxygen molecule. from the three crbon shells from the fourwve fit of Fe(TPP), nd dding one extr wve in ech cse. Comprison fits were lso done on the Fe(,,,TpivPP)(NMe(imid))02 dt using prmeters from the deoxy fourshell fit nd dding two wves for the presence of the oxygen in the oxy dt. Ech fit ws done on bout 85 points over k rnge of 412 As with the fourshell fits, the Debyeller fctor, (2( ui2)), the power on k, (p), nd the qudrtic coefficient in the phse, (2), were ll held fixed t 0.015, 2, nd 0, respectively. n order to ccurtely mesure n iron outofplne shift of 0.4 A, the procedure must be ble to detect ccurtely difference of , in the nitrogen shell (2R +,) vlue. The estimted stndrd devition of this vlue is on the order of , in the preliminry fourshell fit nd somewht lrger in the comprison fit itself, so tht the clculted chnge in the ironnitrogen distnce using this procedure is more qulittive thn quntittive. Further complicting the ttempt t estimting the iron outofplne shift is the fct tht this procedure ssumes tht the conformtion of the porphyrin remins exctly the sme from compound to compound, which my not Crmer. Eccles, Kutzler, Hodgson, Donich / XRy Photobsorption Spectroscopy

8 8066 Tble 111. Results of Difference nd Derivtive Fits" Type of fit Rdil distnce (2R + i) (A)h Rcqril (A) 01 (A)' Fe(PPX)C Difference FeC (40) 2.218" 0.95 nd Fe(TPP) Derivtive FeC (39) 2.21gd 0.95 Fe(TPP)(SC6Hs) Difference FeS 3.45 (1 1) 2.32'' 1.19 nd Fe(TPP) Derivtive FeS 3.43 (08) 2.32' 1.21 Oxy nd Deoxy Difference Fe (07) Fe(,,,TpivPP) Fe0' 5.05 (17) 2.79n 0.53 (N Me( imid)) (Fe0' Fe0) n Derivtive Fe (07) Fe0' 5.05 (17) 2.79R 0.53 (Fe0' Fe0) n (' Fits were ll done over the rnge of k = 412 A' using unfiltered dt. Estimted stndrd devitions re given in prentheses. 1 clculted for the pproprite rdil distnce using the crystllogrphic distnce nd vlues of (2R +,). D. F. Koenig, Act Crystllogr., 18, 663 ( 965). e Reference 27. f Reference 28. g Clculted from ref 28. be the cse if doming or other porphyrin distortions re involved. The (2R + l) vlues re determined for the xil lignds in the extr wves. t is interesting, however, tht outofplne shifts when clculted this wy come close to the crystllogrphic results. For exmple, the clculted movement of the Fe tom reltive to the porphyrin plne when going from Fe(TPP) to Fe(TPP)(SC6Hs) is 0.40 A nd the crystllo grphic result for similr compound Fe(TPP)(SC6H5 NOZ)~~ is 0.43 A. This procedure ws less promising thn the methods using difference nd derivtive fits described below. Difference nd Derivtive Fits By directly compring the fine structure of one iron porphyrin with tht of similr iron porphyrin contining dditionl xil lignds, the distnce to these dditionl xil lignds cn be ccurtely determined. To do this comprison, the fine structure dt from the first porphyrin is fitted to tht of the second. f the two dt sets re scled to ech other in this mnner, the difference between them is the fine structure due to the dditionl xil lignds. Fitting this difference with the fine structure functionl form yields (2R + ]) vlues from which the distnces to the xil lignds cn be obtined, ssuming vlues of! re known. The functionl form used for this "difference fit" is then: where the sum is over the number of shells in the dditionl xil lignds. Strting vlues for the constnts in the sum re obtined from preliminry difference Fourier which is done without ttempting to scle between the dt sets. After the constnt bl is determined from difference fit, finl difference Fourier, in which the two dt sets re ppropritely scled, gives much better picture of the distnces to xil lignds. Problems sometimes occur with difference fits if the geometric chnges between the two porphyrin molecules include chnge in iron to porphyrin distnce (n outofplne metl tom shift), since prt of the difference is then porphyrinlike spectrum. Since the chnge in metlrdil shell distnces is responsible for this porphyrinlike difference, it would be desirble for the fitting function to ccount for this chnge s well s for the presence of extr xil lignds. An pproprite term to dd to the fitting function would be derived from the first term of the Tylor series expnsion s function of R,. C ARi(dRi) i shells where AR; would be djusted in the fitting procedure. e ssume tht the Debyeller fctor is constnt for smll chnges in R nd tht chnges in inelstic scttering processes re negligible. Of course, the vilble dt re function of k not R. However, dr; is relted to k(dddk) s shown in the following wy. First express x s function of k nd R (leving out the qudrtic term in the sine rgument): X(k,R) = %sin (2R;k lk) kdr, Then compre the prtil derivtives with respect to R; nd k: X 2A; sin (2R;k + 0 Tz=Tm + lk) 2A; +CkP'R,2 COS (2R;k + uo + lk) =bx+ 2Ai cos (2R;k + o + lk) + "(if$ "'cos (2R;k +. + lk) 1 = b'x + Aj(2R; + i) kolrjz COS (2Rjk + uo + lk) An pproprite fitting function tht ccounts for both the chnge in iron to porphyrin distnces s well s the presence of dditionl xil lignds is then: The derivtive with respect to k t ech point of interest cn be obtined numericlly by fitting polynomil to dt round the point nd differentiting the polynomil. Difference nd derivtive fits were done on three different sets of dt: Fe(PP1X)C minus Fe(TPP), Fe(TPP)(SC6H5) minus Fe(TPP), nd oxy minus deoxy Fe(,,,TpivPP) (NMe(imid)). These fits were typiclly done on bout 85 smple points over the k rnge 412.&, nd the results of these fits re shown in Tble 111. There lso ppered to be very little difference in numericl results obtined from derivtive nd difference fits done over the sme rnge using filtered nd unfiltered dt. The derivtive nd difference fits give very consistent nd ccurte results when used for predicting distnces to xil lignds. For exmple, the derivtive fit on the deoxy nd oxy forms of Fe(,,,cYTpivPP)(NMe(imid)) is shown in Figure Journl of the Americn Chemicl Society / / December 8, 1976

9 The qulity of the fit obtined by the ddition of the derivtive term nd the contribution of the oxygen cn be seen by compring the upper two curves (oxy nd deoxy) with the lower two curves in which the oxy dt hve been simulted by dding in the contribution due to the dditionl oxygen nd the derivtive term. The derivtive fit gives AR vlue (RFe0 RF~o~) of 1.01 A with n estimted stndrd devition of 0.03 A for RF~o nd 0.08 A for RF~o~. This AR vlue com pres very well with the crystllogrphic distnce of 1.04 A.27 The rtio of the weights of the two oxygen wves (bl(feo )/bl(feo)) comes out somewht higher (59%) thn is predicted by the inverse squre lw (41%) of the functionl form of EXAFS which ws given erlier. This suggests tht t the current level of semiempiricl representtion the reltive weight my be used to give only rough (40% or so) estimte of coordintion numbers. n principle, it should lso be possible to extrct ccurte informtion bout the chnge in the FeN distnce (s well s the xil lignd distnces) from the derivtive fits. From the erlier discussion, it cn be seen tht the term b&(dddr) represents the second term in the Tylor series expnsion of x s function of R. n some cses, subsequent terms in the expnsion re not insignificnt compred to this term so the mgnitude of b2 is not lwys good mesure of the mgnitude of the outofplne shift. However, in most cses exmined thus fr the sign of b2 correctly predicts the direction of the shift, whether in or out of the plne. ncluding the b2 term when n iron tom shift is involved does substntilly improve the fit over difference fit, nd thereby reduces the stndrd devitions of the prmeters. Discussion nd Conclusions e hve reported the development of semiempiricl functionl form to describe extended xry bsorption fine structure spectr nd hve used this form for the interprettion of spectr from orgnometllic complexes using curve fitting techniques. The results described in this pper support the essentil correctness of this pproch nd estblish tht it is vluble lterntive to the Fourier trnsform method of nlysis. Extrpoltion from the single shell fits to the multishell porphyrin fits shows tht under certin conditions the fine structure contributions from different toms re dditive. Curve fitting showed tht simple model for the fine structure mplitude (1/k@, where vlues of rnged from 1.7 to 2.2) gve vlues somewht lower thn the vlue of 3 expected from the Born pproximtion in the high k limit. Although not necessrily significnt, the flloff of the fine structure mplitude function is slower with oxygen thn with nitrogen, which is in turn slower thn with crbon. hile the totl fine structure mplitude for these light elements cn be successfully modeled by simple function (n inverse power of k times n exponentil), such is not expected to be the cse for high Z elements in which the bckscttering mplitude peks t higher photoelectron energies.29 The shpe of the bckscttering mplitude is chrcteristic of the type of bckscttering tom nd its bonding. Such behvior my be dignostic for the identifiction of specific environments of the bsorbing tom. A bound sulfur tom should be distinguishble from oxygen or nitrogen, nd metlmetl interctions will lso be redily observble (since the mplitude will pek t much higher energy). Thus the mplitude s well s the phse will be importnt for the interprettion of the extended fine structure of metlloproteins. e hve shown tht while it is difficult to get precise structurl informtion (better thn 0.1 A) from single EXAFS spectrum of complex molecule, it is possible to get very ccurte informtion on structurl chnges. e hve 0.2 Jl ,s k Figure 11. The effect of derivtive fits (over the rnge of k = 412 the simultion of the oxy Fe(,,,TpivPP)(NMe(imid)) spectrum from the deoxy spectrum plus oxygen wves nd the derivtive term. The upper two curves compre directly the oxy ( ) nd deoxy () fine structure dt. The middle curve is the contribution to be dded to the deoxy dt (contining the two oxygen wves nd the derivtive) in order to simulte the oxy dt. The lower curves show comprison of the oxy dt ( ) with the dt obtined by combining the deoxy fine structure with the derivtive nd oxygen contribution (middle curve). developed new methods of the difference fitting nlysis which redily provide much more ccurte informtion when looking t structurl chnges round the bsorbing metl ion. From the expression (2R + ]) which is vried in the fitting routines, it hs been shown tht resonbly ccurte distnces cn be obtined if the vlue of l is known. From the results of our fits on vriety of iron porphyrins, we hve determined tht resonble vlues of 1 in these compounds30 re 0.95, 1.2, 0.47, 0.79, nd 1.0 A for S, Cl, 0, N, nd C, respectively. These l vlues pper to be resonbly independent of the chemicl environment, nd trnsferrble from molecule to molec~le.~ For exmple, using the Fourier trnsform vlue of 1.12 A determined by Syers, Stern, nd Herriott3* for FeS scttering in FeS2 nd the (2R + ]) vlue of 3.43 A from the verge of our difference nd derivtive fit results for Fe(TPP)(SC&) Fe(TPP) (see Tble 111), we clculted n FeS distnce in Fe(TPP)(SC&) of 2.31 (4) A. The crystllogrphic distnce in very similr compound Fe(PPX)(SC6HSN02) is 2.32 (1) A.27 t hs lso been shown tht the difference nlysis methods cn distinguish between vrious modes of lignd binding (within the limits of rdil distribution). n the cse of the oxy form of Fe(,,,TpivPP)(NMe(imid)), for exmpfe, the difference Fourier showed two peks (t 1.5 nd 2.5 A) indictive of binding of the Fe to dioxygen in the endon or Puling mnner. Should symmetricl bonding hve been the cse, only one pek, intermedite in distnce between the two, would hve been observed. (Likewise, the derivtive fits would hve required only one dditionl wve with twice the weight to describe the oxy spectrum in terms of the deoxy.) hile, in principle, the mgnitude of the derivtive cn be used to extrct chnges in the distnces from the Fe to the porphyrin toms, the series termintion effects mentioned erlier prevented this. Crmer, Eccles, Kutzler, Hodgson, Donich / XRy Photobsorption Spectroscbpy

10 8068 hile not discussed in this pper, xry photobsorption spectroscopy cn give other kinds of informtion bout the bsorbing metl ion. The bsolute position of the bsorption edge gives informtion bout the effective coordintion chrge on the bsorbing metl ion.33 Our work on edge shpes hs recently suggested tht the shpe of the bsorption edge (s determined by the presence of vrious bound sttes, s well s the ioniztion trnsitions) is sensitive to chnges in the type of ligting tom. t should be possible, for exmple, bsed on edge shpes lone to distinguish n xil sulfur lignd from n xil nitrogen lignd in n iron porphyrin. These results will be discussed in detil in future communiction; however, it cn be seen tht this informtion, long with the structurl results obtinble s described herein, mkes xry photobsorption spectroscopy powerful technique for the study of orgnometllic structure. t thus ppers tht xry photobsorption spectroscopy will provide complimentry technique to other nlysis methods by providing direct distnce informtion resulting from structurl chnges in the locl environment of selected metl ion in complex mcromolecule. The pplictions of such methodology spn the res from study of heterogeneous ctlysts to study of superconducting orgnometllics. Perhps some of the most significnt informtion my come from investigtion of enzymtic nd metlloprotein systems where chnges in substrte metl interctions re criticl to the function of the physiologicl process. ork in these res is being pursued. Acknowledgments. e would like to thnk Tom Sorrell for his generous ssistnce in synthesizing number of the porphyrin smples used in this study, nd Dr. Susn Hyes for her preprtion of the "oxypicket fence". e thnk Dr. B. Kincid for his helpful comments on this work. e would lso like to cknowledge the help of Hermn inick nd the other members of the Stnford Synchrotron Rdition Project Stff for their technicl ssistnce in support of this project. This work ws prtilly supported by the donors of the Petroleum Reserch Fund, dministered by the Americn Chemicl Society, nd by the Stnford Synchrotron Rdition Project supported by the Ntionl Science Foundtion Grnt DMR A02 in coopertion with the Stnford Liner Accelertor Center nd the Energy Reserch nd Development Administrtion. Appendix Numericl Anlysis Methods. Lest bsolute vlue (rther thn lest squres) fits were done by minimizing the function N c = c k31xi pi( i= 1 where N is the number of dt points fit, k is the wve vector, x; is the ith dt point, nd F; is the vlue of the fitting function. The minimiztions were done by generl minimiztion progrm using n lgorithm somewht similr to tht of Ro ~enbrock~~ in which no derivtives re required. Ech mini miztion step consisted of serching out minimum in n directions one t time, n being the number of djustble prmeters. At the end of ech step, the n orthogonl directions of serch were rotted such tht the lst direction of serch in ny step ws in the overll direction tken by the previous step. The directions of serch re given by the row vectors in the mtrix shown in (1) where V. V7, V7,....., V, is thevector of the overll tht the progrm serched in ech direction ws decresed s the fit improved until it reched lo'* nd the progrm hlted. The progrm required initil guesses t the prmeters which were then refined. References nd Notes (1) () A. Robinson, Science, 190, 1074 (1975);(b). Metz nd A. Robinson, ibid., 190, 1166 (1975). (2) L.. Schiff, "Quntum Mechnics", 3d ed, McGrwHill, New York, N.Y., 1966, pp (3) For review of xry bsorption spectroscopy, see L. Azroff nd D. Pese in "Xry Spectroscopy", L. Azroff. Ed., McGrwHill, New York, N.Y., 1974, Chpter 5. For review of the older extended fine structure theories, see L. V. Azroff, Rev. Mod. Phys., 35, 1012 (1963). (4) B. M. Kincid, P. M. Eisenberger, K. 0. Hodgson, nd S. Donich. Proc. Ntl. Acd. Sci. U.S.A., 72, 2340 (1975). (5) R. de i. Kronig, Z. Phys., 70, 317 (1931); 75, 191, 468 (1932). (6) () C. A. Ashley nd S. Donich, Phys. Rev. 6, 11, 1279 (1975); (b) A.. Kostrev, Zh. Eksp. Teor. Fiz., 19, 413 (1949); (c) H. Petersen, Z. Phys., 98, 569 (1936);(d) T. Shiriw, T. shimur, nd M. Swd, J. Phys. SOC. Jpn., 13, 847 (1958);(e) A.. Kozlenkov, Bull. Acd. Sci. USSR, Phys. Sci. (Engl. Trns.) 25, 968 (1961). (7) P. A. Lee nd J. B. Pendry, Phys. Rev. 6, 11, 1279 (1975). (8) () E. A. Stern, Phys. Rev. 6, 10,3027 (1974);(b) D. E. Syers, FhD. Thesis, University of shington, Settle, sh (9) B. Kincid nd P. Eisenberger, Phys. Rev. Lett., 34, 1361 (1975). (10) D. E. Syers, F.. Lytle, nd E. A. Stern, Adv. XRy Anl., 13, 248 (1970). (11) () D. E. Syers, F.. Lytle, M. eissbluth, nd P. Pinett. J. Chem. Phys., 62, 2514 (1975); (b) F.. Lytle, D. E. Syers, nd E. A. Stern, Phys. Rev. 6, 11, 4825 (1975);(c) E. A. Stern, D. E. Syers, nd F.. Lytle. ibid., 11, 4636 (1975). (12) R. G. Shulmn, P. Eisenberger,. E. Blumberg, nd N. A. Stmbugh, hoc. NN. Acd. Sci. U.S.A., 72, 4003 (1975). (13) J. P. Coilmn, J. L. Hord, N. Kim, G. Lnd, nd C. A. Reed, J. Am. Chem. SOC., 97, 2676 (1975). (14) J. P. Collmn nd C. A. Reed, J. Am. Chem. SOC., 95, 2046 (1973). (15) L. M. Epstein, D. K. Strub, nd C. Mricondi. lnorg. Chem., 8, 1720 (1967). (16) J. P. Collmn, T. N. Sorrell, nd B. M. Hoffmn, J. Am. Chem. SOC., 97, 913 (1975). (17) J. P. Collmn, R. R. Ggne, C. A. Reed, T. R. Hlbert, G. Lng, nd. T. Robinson, J. Am. Chem. SOC., 97, 1427 (1975). (18). L. Jolly, "The Synthesis nd dentifiction of norgnic Compounds", PrenticeHll, Englewood Cliffs, N.J., 1970, p 484. (19) J. Bjerrum nd J. P. McReynolds, lnorg. Syn., 2, 217 (1946). (20) H. inick in "Proceedings of the Ninth nterntionl Conference on High Energy Accelertors", Stnford Liner Accelertor Center, Stnford, Clif., 1974, pp (21) B. Kincid, P. Eisenberger, nd D. Syers, Phys. Rev., submitted for publiction. Journl of the Americn Chemicl Society / 98:25 / December 8, 1976

11 8069 (22) Although the incident photon energy is high, the totl number of photons bsorbed is very smll compred to the number of molecules present. Under typicl conditions, t 7 KeV the flux is on the order of lo* photons/s with bndwidth of bout 1 ev. (23) See Appendix section for brief description of the numericl methods. (24) H. H. Rosenbrock, Comput. J., 3, 175 (1960). (25) Actully the extended fine structure is defined s (,u,uo)/,uo, where,u is the observed bsorption coefficient nd,uo is tht of the free tom. However, if PO chnges slowly on the given energy rnge, the subtrction procedure will be correct to within normliztion fctor. (26) B. Kincid, Ph.D. Thesis, Stnford University, Stnford, Clif., (27) J. Koch, S. C. Tng, R. H. Holm, R. B. Frnkel, nd J. A. bers, J. Am. Chem. Soc., 97, 916 (1975). (28) J. P. Collmn, R. R. Ggne, C. A. Reed,. T. Robinson, nd G. A. Rodley, Proc. Ntl. Acd. Sci. U.S.A., 71, 1326 (1974). (29) The mplitude modultion of the observed fine structure is discussed in ref 26, p 89 nd ref 1 lb, p (30) The vlues of, were derived from compounds where Fe ws the bsorbing tom. The numericl fits were done over k rnge of 412 A'. (31) Further test of such trnsferrbility hs recently been mde by B. Kincid, P. Citrin, nd P. Eisenberger, Phys. Rev., submitted for publiction. (32) D. E. Syers, E. A. Stern, nd J. R. Herriott, J. Chem. Phys., in press. (33) S. P. Crmer, T. K. Eccles, F.. Kutzier, K. 0. Hodgson, nd L. E. Mortenson, J. Am. Chem. Soc., 98, 1267 (1976). Crystl Structure, Rmn Spectrum, nd Nucler Mgnetic Resonnce Spectrum of Methoxythionyl pfluorobis( pentfluorontimonte( V)), CH30SOfSb2F1 1 R. J. Gillespie,* F. G. Riddell, nd D. R. Slim Contribution from the Deprtment of Chemistry, McMster Unicersity, Hmilton, Ontrio, Cnd. Received June 18, 1976 Abstrct: The compound CH30SO+SbzF 1 hs been prepred by the rection of methyl fluoride with ntimony pentfluoride in liquid sulfur dioxide. ts structure hs been investigted by 19F spectroscopy of its solutions in SO2, by Rmn spectroscopy of the polycrystlline solid, nd by n xry crystllogrphic exmintion of single crystl. The crystls re triclinic with = (5), 6 ~ (3), c = (4) A, = (4), p = (5), y = (2)". The structure ws refined in the spce group P to finl greement index of for 1405 independent reflections. The structure is mde up of discrete CH30SO+ ctions nd liner SbzFll nions. Solutions of CH3F nd SbFs in sulfur dioxide nd in sulfuryl chlorofluoride hve long been thought to contin one or more CH3FSbF5 complexes for which vrious formultions hve been given.'s2 However, the bsence of HF coupling in the 'H nd 9F NMR spectr of these solutions hs long remined n unexplined difficulty. Very recently Peterson et L3 reported the preprtion of white crystlline compound from the rection of 1:l mixture of SbFS nd CH3F with SO2 in solution in S02CF. Becuse the compound rected with methnol nd ethnol to give dimethyl sulfite nd methyl ethyl sulfite, respectively, rther thn dimethyl ether nd methyl ethyl ether, s reported previously,' Peterson proposed tht the compound ws CH30SO+SbF6 (or SbzFl ), in which, of course, no HF coupling is expected. At the sme time Olh et L4 reinterpreted their erlier NMR dt in terms of the formtion of CH3OSO+ in SO2 solution nd CH30SOClF+ in S02CF solution. He hs lso shown by the observtion of the expected HF coupling in the 'H NMR spectrum in S02F2 s solvent, tht the CH3FSbF5 complex is formed in this solvent which, unlike SO2 nd SOlClF, is not methylted. n this pper we report the isoltion of white crystlline compound from solution of CH3F nd SbFs in S02, which is presumbly the sme compound s ws reported by Peterson et L3 nd lso mentioned briefly by Olh et L4 nvestigtion of this compound by xry crystllogrphy, Rmn spectroscopy, nd 'Hnd 9F NMR spectroscopy hs shown tht it is indeed not CH3FSbF5 complex but is the compound CH30SOfSb2F. Experimentl Section Antimony pentfluoride (OzrkMhoning) ws purified by double distilltion in n ll glss pprtus nd sulfur dioxide (M theson) ws repetedly distilled from, nd kept over, phosphorus pentoxide before use. Methyl fluoride (Peninsulr Chemicl Reserch) ws not further purified. n typicl experiment, SbFs (5 X mol) ws syringed into one rm of rigorously dried Pyrex glss double mpule inside drybox. The pprtus ws connected to the vcuum line nd methyl fluoride (2.5 X mol) nd sulfur dioxide (2 X mol) werecondensed onto the SbF5 t 196 "C. The mixture ws llowed to wrm to 78 "C nd ws stirred t this temperture for bout 30 min. The mixture ws llowed to wrm up to 7 "C nd the solvent ws then removed by cooling the other mpule to 30 "C. hite crystls were left behind nd were pumped to dryness. Suitble crystls for xry studies were mounted in thinwlled qurtz cpillries inside drybox equipped with microscope. Crystl Dt. CH30SO+Sb2F is triclinic with = (5), 6 = (3), c = (4) A, = (4),p = (3), y = ()", V = 576A3,Z = 2, D, = 3.06, fw = 531.4, F(O,O,O) = 488, h(mo KZ) = A, nd h(mo K) = 5 1 cml. The unit cell prmeters were obtined from lestsqures refinement of 15 reflections in the region 20" < 28 < 25'. The strusture ws successfully refined in the centrosymmetric spce group P. (Ci,SNo. 2). XRy ntensity Mesurements. The crystl, which ws n irregulr block of pproximte dimensions 0.28 X 0.16 X 0.09 mm, ws Eled in thinwlled qurtz cpillry nd mounted on Syntex P diffrctometer with its 0.2%" edge, which ws pproximtely perpendiculr to the (1,0,0) fce, lmost coincident with the 4 xis of the diffrctometer. ntensities were mesured using grphite monochromted Mo KE rdition, with 828 scn nd scn rte vrying from 8.0 to 24.0 /min in 28 so tht weker reflections were exmined more slowly to minimize counting errors. Sttionry bckground counts, with time equl to hlf the scn time for ech reflection, were mde t ech end of the scn rnge. Thescn width vried from 2" t low 28 to 2.5" for highngle reflections. One stndrd reflection ws regulrly checked to monitor the stbility of the crystl nd its lignment, but no significnt vrition ws observed. ndependent reflections (1405) within the rnge 28 < 50" were mesured resulting in 1331 reflections with intensities greter thn three times their Gillespie, Riddell, Slim / Crystl Structure nd Spectr of CH~OSO+S~~F