theoretical analysis of the cycloaddition of ethylene

Upload: jose-cortes

Post on 01-Jun-2018

236 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/9/2019 Theoretical Analysis of the Cycloaddition of Ethylene

    1/13

    Chemical Physics 52 (1980) 151-163

    @ North-Holland Publishing Company

    THEORETICAL ANALYSIS OF THE CYCLOADDITION OF ETHYLENE

    E. KASSAB, E.M. EVLETH, J.J. DANNENBERG* and J.C. RAYEZ**

    Cenne de M icanique Ondul aroireApp l i qu& , 75019 Pari s, France,

    and t he Deparhnent of Chemi stry, Cit y Uni versit y f New York, Hmrer Coll ege,

    New York , N- Y. 10021, U SA

    Received 2 January 1980

    Revised manuscript received 1 July 1980

    The cycloaddition of ethylene is theoretically analyzed for portions of the excited singlet and triplet hypersurfaces

    using a combination of semi-empirical and intermediate level ab initio techniques. The semi-empirical UHF calculations

    on the addition of triplet ethylene and methyl radical to ethylene showed that these

    two

    reactions have comparable

    theoretical parameters, including activation energies, spin transfer and spin polarization at the transition state. For the

    ?.S+ZS excited ingletstate surfaces, the results of both the ab initio and semi-empirical calculations are qualitatively the

    same and correspond to the classical ideas generated from orbital symmetry rules. At the ab initio level the results are

    quantitatively poor, partially due to the use of an intermediate level configuration interaction treatment. In particular, it

    was not possible

    t o

    obtain other than a small fraction of the total estimated valence correlation energy in cyclobutane.

    The configuration interaction problem for both ab initio and semi-empirical calculations is discussed in detail.

    1 Introduction

    The goals of this article are two-fold. First,

    we will characterize the theoretical nature of the

    photodimerization of ethylene for portions of

    both the excited singlet and triplet hypersur-

    faces. Second, we will explore some of the

    methodological problems encountered in using

    semi-empirical and ab initio methods to eluci-

    date photochemical reaction mechanisms. We

    are especially interested in developing semi-

    empirical methods in order to investigate larger

    systems on which good quality ab initio cal-

    culations are presently financially impractical.

    The dimerization of ethylene and other

    related olefins to give substituted cyciobutanes

    (or the retroaction) is well studied both

    experimentally [l] and theoretically [2]. The

    orbital symmetry rules for the concerted 2S+2S

    process are a standard pedagogical exercise [3].

    *

    City University of New York, Hunter College, USA.

    Present address: Univeaiti de Bordeaux I, Talence,

    NATO Postodoctoral Fellow, 1975-76.

    While existing theoretical work is in support of

    these rules, there are several important nuances

    with regard to the excited state surfaces. Firstly

    the triplet surfaces have not been explored.

    Secondly, recent work on the analogous Hq

    surface [4] indicates that a conceivable route for

    deactivation of the excited singlet state of the

    ethylene dimer could occur by a crossed (Dzd)

    approach of two ethylenes [4b]. Thirdly, the

    theoretical nature of the so-called doubly

    excited state at large ethylene-ethylene separa-

    tions is not anticipated by the orbital symmetry

    rules [4]. In this article we will explore portions

    of the triplet surface which yield the triplet

    tetramethylene diradicd. For the singlet surface

    we will only explore planar face-to-face

    approaches having rectangular-trapezoidal

    carbon atom configurations.

    We will specifically treat the theoretical

    nature of the doubly excited state at both small

    and large ethyiene+thylene separations. Finally,

    we will discuss the problems encountered in

    using an intermediate level configuration inter-

    action treatment.

  • 8/9/2019 Theoretical Analysis of the Cycloaddition of Ethylene

    2/13

    1JI

    E. Kassab et al. / 77worelicn f

    nnlysis of hc cycloaddirion of erhpkne

    Tlnc major portions of the ab initio cal-

    ~~~~t~~n~ ere performedusing the gaussian

    dique of Whitten and co-workers [S].

    an over contractedvalence basisset,

    olg) for carbon and (5s/24 for hydro-

    md with orbital

    exponents,

    [SJ.

    Further

    studies

    using a

    ~~~~nt~r~d using the intermediate tevel CI

    W&Mrnffnt dl~ubeed below

    and lack of financial

    &a b srrry oul a complete treatment.

    @fesU%ZP alculutionswere performed oh

    dk# @l@#B#hell ground and open shell singlet

    &nd @ipfftt ctlnflgurutians. The CI treatment [6]

    8% 9~ ir,irrlly

    symmetric tates was expanded

    two parents,one being the original

    sh@ll onflgmkm,

    the other

    being the

    nntigtration

    resulting from two

    promofcd from the HOMO to

    Is trentmentyields au after CI

    ground state, So, and the so-

    nclted stnte, P*. By

    symmetry

    @QrMlUnn,he rlagtetand triplet states cal-

    @&MrPjer@, 7 ltndTI,

    have

    B1,, symmetries

    i&f the OJ~ orm of the dimsr, these being

    r&&d to r&Has enerated by cxcitoa inter-

    Hi8dttbof the mr* (%J,,) tutes with the

    $$t%unrd

    18IQ, Thu Cl

    cxpansious for these open

    sr&ll etztes

    WQ~Q

    performed around only one

    %Li arent cznllgurutian. All CI calculations

    WVl@ he result of the threshold terms [?i] of 5 x

    n with tho number of configurations

    1 4OfJ

    n each

    af the

    three

    separate

    i%lpcl salcululions were performed by

    Ereez

    M I oh@

    OWVM 1

    MOs (essentially Is on C), the

    rNiano9 searchbeing done over the next

    8, Fhus he highest 12 virtual orbitals

    %@~i-ot ttxaminad. Since the Ci calculations

    @@IFerfarmr;d nly ORa portion of the

    &M arhitz\lswe decided to estimate the

    a%1 4%lencleurrrelntiunnergies of tiylobutane

    and ethylene. This was done u&g &

    modification cf the Gaussian 70

    m in

    which the size of the individual

    I

    d s-cl.

    3G orbitals is related to the mrrelalion cw?gy

    PI.

    2.2. Semi empir ical

    The method used is the previowfy

    moderately reparameterized CM)0 te&niq~e

    [2fl applied at the medium CI tevei @O-120~ CPI

    at the UHF level. An identical paratneter&&on

    was used which yielded, at the 60x60 level. an

    approximate enthalpy of -20 k&/mole [Zfj*

    (obs.,

    -18) [lc] for the reaction of two ethyl-

    enes to give cyclobutane as well as reasoaabk

    geometries

    for these two molecules (e.g. CC

    distar.aes of 1.35 and 1.56 A, respe&vely). The

    semi-empirical CI treatment is to be contrasted

    with the ab initio one in that both the

    grauhd

    and excited singlet aud triplet Cl states were

    generated from the same set of CN00 SCF

    closed s ~ell molecular basis orbit&. In add&n,

    no z ltomatic confguration selection procedure

    was used in the semi-empirical CI treatment.

    Important configurations were inch&d in the

    treatment as a result of a number of trial c2&

    culations in which the inadequzy of the Cl

    basis set was evident by disco&n&ties of the

    S** and Sf surfaces at the HOMN,UiMO

    inversion geometry (ca. 2.1 A separation

    between the two ethylenes). AIthougb the orig-

    inal calibration of the relative enthalpies of

    ethylene and cyclobutanewas done at the 60 x

    60 CI level, it was found that the excited states

    were better treated at a larger CI level Thus,

    the So and S** states shown here were treated

    at the 102 CI level of which 88 were doubly

    excited including 26 four open shel9 amfigura-

    tions whose importance will be d&cussed later.

    The ST and T1 states were treated

    at

    the 75

    md

    84 CI level, respe-tively.

    At the time this study was done we

    had no

    available scheme for geometry optimization at

    the after-C1 level. We performed a partial

    point-by-point optimization for the S** state al

    t A

    misprinl in ref.

    [2fl

    quotes his valueat 4 kczllmk

  • 8/9/2019 Theoretical Analysis of the Cycloaddition of Ethylene

    3/13

    2.0 A ethylene-ethylene separation. The same

    was done for the ST state at 2.8 A. This after-

    CI optimization was done by varying only the

    CC distances and the CHI ffap angles.

    For the UHF triplet surface a more complete

    optimization was performed [g] using the same

    repatameterized CNDO method [2f]. However,

    exploratory calculations were aiso done using

    the unparameterized INDO as well as

    MINbO/S methods. These latter methods were

    rejected for reasons discussed below,

    3. Reaulls nnld discussion

    3.1.

    Gert eral met hodological onsiderat ior rs

    It is our general view that useful inexpensive

    information on excited state behavior can be

    obtained using semi-empirical methods [2f, 91.

    The disrepute of such methods lies mainly with

    their quantitative uncertainties and seemingly

    ever changing parameterization schemes. It 1s

    probably a general opinion that for small molec-

    ular systems with a large basis set, large CI

    cnlculation can be trusted for a surface cal-

    culation involving some changes in relative cor-

    relstion energies along that surface [lo]. For

    huge molecular systems, financial or technical

    limitations impose smeller basis sets such ih: l t a

    lorgc CI treatment may be difficult or presently

    impossible and the resulting surfaces poor [IO].

    Since both ab initio and semi-empirical cal-

    culntions using similar size basis sets will carry

    the same symmetry information, semi-empi::ical

    CI methods still have the potential of giving

    useful information on large systems, In this case

    the advantages of the small basis set semi-

    empirical CI method over its ab initio counter-

    part lies in its variable parameterization. Th,us,

    the CNDQ/S method

    [l

    11

    will still give be.ter

    estimates of transition energies For a large

    system tharl will a small basis set ab initio

    method. Neither the CNDCI/S nor the original

    CNDQ/INDO methods were originally

    parameterized with the intention of doing s.u-

    face calculations. The main reason for their

    poor estimate of the relative ground state

    energies of different molecules [12] hes been

    established [13] and largely corrected at the

    MCSCF level [14]. Likewise, we have demon-

    strated that our reparameterized CNDQ CI

    method yields adequate appearing ground and

    excited state bond rupture surfaces [2f]. It

    remains to be shown by comparative cal-

    culations, however, whether or not a CNDO CI

    method can generally mimic the excited state

    surface features generated by an ab initio cal-

    culation of similar basis size. Thus, part of what

    .-Jill be explored here is a comparison of such

    calculations and problems encountered.

    3.2.

    The et hyl ene-et hyl ene t ri plet dimeri zati otl

    Available thermodynamic, spectroscopic and

    theoretical data indicate that the reaction of

    twisted triplet ethylene (3E90-, CC = 1.48 a [IS])

    with ground state ethyfene (l$) to give the

    triplet tetra-methylene diradicai (TMDR) is

    about 19 kcal/mole exothermic:

    Eob+ 3Ego.= 3TMDR, AH = - 19 kcal/mole,

    (l

    I$,- -I-I& =

    cyclobutaae,

    AH = -

    18

    kcal/moIe [ Ic],

    ( ;

    cyclobutane = TMDR,

    AH =53 kcal/mofe [ICI,

    13)

    *Eoo hv = Euom, AH = 64 kcal;mole,

    (1)

    This estimate (1) is obtained by combining

    reactions (2), (3), and assuming that TMDR

    and TMDR are nearly isoenergetic.

    Only in reaction (2) do we have a true

    experimental value. For reaction (3) we assume

    that, the activation energy for the thermal

    decomposition of cycfobutane is identical to the

    enthalpy change generating TMDR even

    though it has been theoretically indicated [Zc]

    that there may be several TMDR intermediates

    which are several kcal/mole lower in energy.

    This would make the value shown for (1)

    slightly more negative. Finaflv, the value shmvn

    for reaction (4) is theoretical [15a. b]. Llowevcr.

    theoretically Ego0 nd Ew are nearly iso-

    energetic and thus the energy of reaction (2)

  • 8/9/2019 Theoretical Analysis of the Cycloaddition of Ethylene

    4/13

    154

    E. K assab et al. 1 theoreti cal anal ysis of he cycloaddi ti in of ethy lene

    will approach the known activation energy for

    thermal cis-trans isomerization of ethylene [lc].

    Even though triplet sensitized photo-

    dimerizations of small ring oIefins have been

    commonly observed [le-j] reaction (1) has

    never been fully demonstrated to occur with

    ethylene .of simple acyclic olefins. This obser-

    vation WI

    be conveniently rationalized [lg] by

    examining the kinetic expression for the disap-

    pearance of triplet olefin (3A) by competitive

    unimolecular intersystem crossing

    3A%1A,,

    and bimoIecular dimerization

    (5)

    3A c

    A,,

    2 3DR_ (6)

    The fates of the generated diradical, DR+

    DR, would only be important if for some

    reason the regeneration of starting material

    from DR was highly favored only in the case of

    acyclic olefins. if we assume that only reactions

    (5) and (6) dominate in both types of olefins,

    then:

    d(3A)/dt = - c~(~A) - kX3A)(A,J.

    (7)

    The trajectory calculations of Warshel and

    Karplus 1161, indicate that ki, in ethylene is

    strongly dependent on the 3AS(l,1A90- energy

    gap, a result similar to the known experimental

    and theoretical energy gap dependence for trip-

    let-ground state intersystem crossing in

    aromatics [17]. The near degeneracy of the trip-

    Iet and ground state surfaces in olefins occurs

    onIy near the twisted 90 configuration of the

    p-orbitals comprising the z-bond. While the

    triplet states of small ring olefins will undergo

    some relaxation, it is axiomatic that the T-S

    energy gaps in relaxed cyclic olefin triplets will

    be larger than for acyclic triplets. Thus, it can

    be argued that it is the variation of kk with

    olefin structure which controls the partitioning

    of reactions (5) and (6).

    There is an additional nuance to the

    arguments posed above. If the values for kd are

    less than diffusion controlled ones this implies

    enthalpyas well as entropy of activation effects

    in reactions of the same type as (1). It is com-

    monly assumed that triplet addition reactions

    are radical-like in character. If so, one can

    anticipate radical-like kinetic parameters. Our

    goal will be to estimate activation energy and

    enthalpy for reaction (1) as well as to attempt to

    characterize the radical-like nature of the reac-

    tion pathway using geometry and spin transfer-

    spin polarization criteria_ We will also investigate

    the possibie differences in triplet sensitized

    photodimerization cyclic and acyclic olefins.

    We will only report in detail on the computed

    optimized triplet reaction path for reaction (1)

    generated using our reparameterized CNDO-

    UHF method [2f]. It was initially determined

    that an unreparameterized INDO-UHF method

    gave a hopelessly false enthalpy for reaction (1)

    (ca. -100 kcal/moIe, versus -19 estimated) as

    well as intuitively false geometries for 3TMDR

    and intermediate structures. While the

    MINDOi3 half-electron method gave a

    reasonable estimate of this enthalpy

    (-31 kcal/moIe), the optimized C-C geometry

    for 3E90n (1.36 A) was very far from the best ab

    initio value (1.48) [15]. Likewise, the

    MIND0/3 optimized geometry for trans-

    3TMDR gave 1.44 8, for the -CHz-CHz dis-

    tance, somewhat far from what one would

    expect from a C-C sp3-sp2 hybrid (1.52) [IS]+.

    Our own CNDO-UHF reparameterization

    yielded -27 kcal/mole for the enthalpy of reac-

    tion (l), and 1.49 and 1.52 for the C-C dis-

    tances in 3Egc. and 3TMDR. Since the same

    CNDO reparameterization is used in our

    CI calculations discussed in the next section

    we decided to retain the same overall

    parameterization for both the UHF triplet and

    RHF-CI singlet surface in spite of the fact, as

    will be shown, that the final computed activa-

    tion energy for reaction (1) was not satisfactory.

    It should be stressed, however, that from a

    general methodological point of view the use of

    a UHF single determinate method to compute a

    surface of a bimolecular reaction between what

    are, at a dissociation limit, a closed and open

    shell species, will give rise to a relative cor-

    f

    See

    also

    ref. [k].

  • 8/9/2019 Theoretical Analysis of the Cycloaddition of Ethylene

    5/13

    E. K assab et al . Theoretical nalysisof he cyclooddi ti on f ethy l ene

    155

    relation energy error [19]. In any case, the

    following energetic analysis is given using the

    reparameterized CNDO-UHF approximation.

    First, in order to determine the energetics of

    the reaction of nearly rigid smali ring triplets we

    computed the following UHF geometry opti-

    mized reactions using our modified CNDO

    parameterization:

    3Eoa(CC 1.35) = 3Erp CC = 1.54),

    AH = -28 kcal/mole,

    3Eoe(CC= 1.54) = 3E90. (CC= 1.49),

    (8)

    M = -16 kcal/mole.

    (9)

    Reactions (8) and (9) have previously been

    estimated by Baird and co-workers 1201 at -32

    and -16 kcal/mole, respectively, with virtually

    the same optimized CC distance. As a model

    calculation for the cyclopentene triplet, we per-

    formed geometry optimizations on ethylene in

    which only two cis-hydrogens and the CC dis-

    tance were allowed to vary and found an essen-

    tialIy planar structure with the same CC dis-

    tance and energy as shown for reaction (8).

    Thus, in a certain sense, small ring cyclic olefin

    triplets contain about 16 kcal/mole excess

    energy as compared to relaxed acyclic olefin

    triplets.

    The geometry optimized CNDO-UHF mini-

    mum energy pathway for the reaction of 3Eg0.

    with the ethylene ground state is shown

    in fig.

    1. For comparison purposes, the reaction of the

    methyl radical with ethylene is shown in the

    same figure. The critical coordinate is the CC

    bond distance between reacting carbon centers,

    this being sufficiently large at the transition state

    (ca. 2.5 A) that little activation energy

    difference was found for rotational variants

    about the forming CC bond. The optimized

    geometry of the tfansition state for reaction (l),

    shown in fig. 2, shows little geometry change in

    the reacting moieties as compared with their

    isolated structures. The possibly more critical

    measure of progress along the reaction coor-

    dinate is the spin density. In both reactions

    there is virtually no spin transfer at the tran-

    sition state. Both the geometry and spin density

    transfer criteria are consistent with the inter-

    pretation that the transition state is more reac-

    tant-like than product-like. On the other hand,

    while little spin transfer has occurred, the

    accepting carbon atom in both reactions exhibits

    a large negative spin density, a polarization

    which occurs well before the transition state.

    Even though an unprojected UHF calculation

    will overemphasize such spin polarization, what

    is observed demonstrates the principle that a

    negative spin density at the receiving carbon

    atom in radical reactions is a required pre-

    condition for bond formation [22]. What we had

    not anticipated is that such spin polarization

    would occur at distances very much larger than

    that occurring for the threshold for spin trans-

    fer. In any case, the theoretical profiles for both

    radical and triplet addition to ethylene are

    similar, giving theoretical support to the idea

    [22] that some triplet state reactions are radical-

    like in character.

    As discussed below, the main negative feature

    of the CNDO-UHF calculations presented is

    that they give unreasonably high activation

    energies. However, we decided to use our

    CNDO-UHF

    calcuiation of the methyl radical-

    ethylene reaction to estimate the probable

    activation energy of reaction (1). Reaction (1)

    has a computed activation energy of

    28 kcaljmole while that of a simulated cyclo-

    pentene triplet-cyclopentene is 21 kcal/mole.

    This latter value is comparable with our own

    computed value, 19 kcal/mole, for the methyl

    radical-ethylene reaction which is, in turn,

    about a factor of two, too large (ohs., 8

    kcal/mole) [23]. Thus, it can be argued that the

    olefin tripiet-olefin reaction should have an

    activation energy comparable to normal radical

    addition reactions. Therefore, based on this

    comparative method we predict that the reac-

    tion of cyclic olefin triplets with olefins should

    have activation energies in a region of

    8 kcal/mole. Acyclic olefin triplets may have

    activation energies several kcaljmole higher

    than this value. Thus, it is predicted that the

    low quantum yields of acyclic triplet sensitized

    photodimerizations are due to a combined high

    value for

    kec

    and lower than diffusion controlled

    rate for kd However, in cyclic olefins ka should

  • 8/9/2019 Theoretical Analysis of the Cycloaddition of Ethylene

    6/13

    CARBQN-CARBON BOND DISTANCE, ANGSTROMS

    Flfl. I. Rep~rnmcterizi zd [2f] CNDO4JHF calcul ations o f the additi on of tripl et ethylene to ethyk~ and m&,4 r&k% BD

    cthylct~c. Shown is the minimum geometry optimized pathway for an in-plane carbon amm co&go&o% wi~-irhCfM &T%B~

    kept con stnnl al 1.08 A. Also sh own are the spin transfer from the radical or tripl et speci es to e~I~@oe and the mzgz~~ spz~

    dcnuily n1 the occepling carbon atom along the reaction coordinate.

    . .

    nlso be lower than diffusion controlled rate.

    lhus,

    in a renl sense, ir is the variation in the

    kilir values which differentiate the triplet sensi-

    tized phatodimerization kinetics of acycIic and

    cyclic olcfins. Sin&r effects should occur in

    cyclic and acyclic polyenes. It is known that the

    qunntum yields for sensitized photodimerization

    OFcyclohexadiene (ca. 1) are much higher than

    for butadiene (O.Ol) [22]. On the other nand,

    \ve have found virtually no reported experi-

    mcntul values for activation energies for triplet

    nddition reaction aside from a recent value for

    tht reaction of triplet trimethylenemethane with

    a substituted olefir~ (6 kcal/mole) [24].

    3.3 LVe cth

    yltme-eth ylene excited singlet surfaces

    3.3. I. Qttalitative aspects

    The essential details of the calculations

    presented here are shown in figs. 3 and 4 for

    the semi-empirical and ab initio calculations,

    respectively. The detailed energeti of the

    latter calculations are shown in tablie P. I%e

    geometry choices for the ab in&o ~~~~~ at

    the intermediate geometries (ca. cuLat&~, B-5

    table 1 and fig. 2) were bzsed cm S&e e1?3-

    empirical results, oost amsi&rat&ms preventing

    us from doing a surface sear& fcr the 9*

    minimum. In both the semi-empfal a& &

    initio czlculatians the minimum d time S 51if-

    face is in the 2.0-2.2 A regk~~

    Qualitatively, the ab initio ay.xd e&e

    calculations show the same be

    portions af the H, surface wea

    Michl and co-workers 13

    possible photochemical

    faces has already been

    these workers.

    We will dli ia detail s~mr:d

    the complications invoived in treatig a very

    much larger system, especially in wbzt mzmncr

    OUT alcul2tions are diEerentia@d &om mrE+

    calculations on the same system.

  • 8/9/2019 Theoretical Analysis of the Cycloaddition of Ethylene

    7/13

    E. Kmab er al. f Th@orelfcal at&is

    o lthz

    yclnaddirion

    of thylene

    157

    Fig 2. The geomeby of the uptiMid transition state of the

    calculation shown in fig. 1. A&J shown are the individual

    atomic spin deMies as well I the total amount of spin

    density tidefred of retained on each ethylene udit.

    Both the semi-empirical and ao initio ca -

    culatiors give essentially the same qualitative

    infarmation, namely that there are two excited

    states Sf and S** of the ethylene dimer which

    are involved in the photodimerizadon process.

    Initially, ar large ethylene-ethylene separations,

    tbe ST surface

    is

    lower or nearly isoenergetic to

    the

    S

    state, but dt the HOMQ-LUMP

    avoided state crossing region near 2.0-i-2 8,

    ethylene-ethylene separatian the S** state

    becomes lower in

    energy. A5

    seen in table 2, at

    short ethjrkne-ethylene distances, the

    configurational compasitinns of the SOand s

    states are as expected from orbitel symmetry

    rules. They both consist mainly of plus and

    minus combinations of two major configura-

    tions, one the closed shell SCF solution

    (HOMO doubly occupied) and the other the

    doubly excited configuration (LUMO doubly

    occupied). The relative weights of these two

    configurations change radically for the SOand

    S** states over a Iairly narrow geometry

    region

    o f significence

    singlet excited stare, S,

    and dauhty excited state, S**.

  • 8/9/2019 Theoretical Analysis of the Cycloaddition of Ethylene

    8/13

    158 E. Kasab eta . / ?heoreticaI analysis of the cycloaddition of ethylene

    I I

    Fig. 4. Ab initio CI calculation of the cycloaddition of

    ethy ene to give cylobutane. See table 1 for geometry

    detaiIs. Caicularion used an after contraction 40 orbital basis

    set for cyclobutane plus CI at about the 400x400 level

    using a configuration selection threshold of 5x IOm1hartree.

    Energies for the e:hylene (e) dimer limit (00) are twice the

    energies shown in table 1.

    than it is usually. Likewise, this surface instabil-

    ity will permit the formulation of diabatic

    functions interconnecting the SOand S**.sur-

    faces [257. More simply stated, this geometry

    region is sutliciently theoretically special to

    warrant the speculation that it is where the

    internal conversion from S** and So occurs [4d].

    A detailed discussion is necessary concerning

    the correlation of the S** state at large ethyl-

    ene-ethylene separations [26]. The configura-

    tional behavior (table 2) of the So state is

    simple, it becomes largely monoconfigurational

    at the cyclobutane or two ethylene limit.

    The configurational behavior of the S** state

    is much more complicated. It does not evolve

    towards a S** state of monomeric ethylene as

    might be indicated by a 2 ~2 CI [2b] or a state

    correlation diagram. As in the case of I& the

    S** state correlates with two ethylene triplets

    [4,26]f. This is shown in table 2 where the S**

    state is shown to be neither principally a closed

    shell doubly excited state nor a two open shell

    monoexcited state but a four open shell doubly

    excited state at 3.5 A. The calculation shown in

    table 2 is done under conditions where the

    ethylene dimer is in a trapezoidal geometry

    where the MOs are Iocalized on either one

    ethylene or the other at large ethylene-ethylene

    separations. If the dimer is calctdated under DZh

    symmetry, a set of delocalized MOs is obtained

    which have a correspondance with those anti-

    cipated from orbital symmetry rules. Un-

    fortunately, the configurational composition of

    the S* state at large ethylene-ethylene separa-

    tions depends on whether one is working with

    localized or delocalized MOs In the case of

    ? Note that this point is overlooked in thecorrelation

    diagrams proposed in related types of photocycloaddition

    reactions: see for instance ref. [27].

    Table 1

    AS initio SCF-CI calculations on ethylene-ethylene

    Cal. .9 R, RL R,

    State

    SO

    S**

    SCF

    CI

    CI

    1 IS 1.41 1.415 1.90 -155.562 -155.955

    2 13 1.40 1.425 2.00 -155.640 -155.936

    3 11 1.39 I.435 2.10 -155.700 -155.931

    4 9 1.38 1.440 2.20 -155.747 -155.945

    5 3.8 134 1.500 335 -156.875 -155.078

    6 ethylene -17.952 -78.079

    7 cyclobutaneb -155.961 -156.029

    -155.745

    -155.562 -155.767 -155.736 -155.866

    -155.809 -155.639 -155.789

    -155.757 -155.886

    -155.838 -155.645 -155.792 -155.766 -155.893

    -155.825

    -155.644 - 155.787

    -155.771 -155.889

    -155.733

    nonconvergent -155.723 -155.908

    -77.477 -77.526 -77.672

    -77.808 -77.852

    SCF CI

    T*

    SCF CI

    a)Experimental geometry.

    ) Planar geometry as found in 6-31G* calcuiation of Cremer [32]. Other geometries were taken from semi-empirical cal-

    culations, geometrical parameters not shown were not changed, CH kept constant at 1.09, HCH bond angle at 112. See fig.

    1 for definitions of S,

    R,. R2

    and

    RO_

  • 8/9/2019 Theoretical Analysis of the Cycloaddition of Ethylene

    9/13

    E. Kassab eta . / Theoreti calanal ysi s of he cycloaddi ti on f eth yl ene

    1.59

    Table 2

    Leading configurational terms in the ab initio calculations

    Ri

    Config.

    coeKb =

    (HOMO) (LUMO)

    1.9

    So

    0.215 - 0.919

    s**

    =

    0.900 + 0.200

    2.0 So

    =

    0.379 - 0.860

    .S**

    =

    0.846

    -I-

    0.359

    2.1

    So

    =

    0.662 -

    0.665

    Y*

    =

    0.539 + 0.639

    2.2

    SO

    =

    0.842 - 0.420

    .S**

    =

    0.397

    c

    0.807

    3.5

    So

    =

    0.929 -

    0.151

    S**

    =

    -0.048 - 0.213=j

    AngtrBm units.

    ) Either double occupation of HOMO or LUMO.

    For this distance the major configuration is a four open

    shell having a value of 0.771. this configuration is related

    IO two ethylene triplets, see

    text.

    d For both the S* and T states at each distance, the major

    configuration (0.9.5-0.96) has a single occupation for the

    HOMO and LUMO.

    localized orbitals (table 2) the double triplet

    character of the proper (S2 = 0) four open shell

    configuration is hidden within the poly-

    determinatal structure shown below:

    where the indices 1 and 2 refer to orbitals

    largely localized on either ethylene 1 or 2. The

    energy of this configuration at large ethylene-

    ethylene distances tends to become bitriplet in

    character because only the central two deter-

    minates contribute to the orbital-orbital

    exchange term K. Likewise these central

    determinates are triplet in character because

    one has either both (Y or fl spins in the same

    molecule. This behavior is also found in the

    semi-empirical calculations performed using

    localized orbitals. However, using del6calized

    orbitals the S**state becomes polyconfigura-

    tional at large distances, i.e. HOlMO+

    LUM02; HOMO+ (LUMO + 1); (HOMO -

    l)+ LUM02; (HOMO - 1)2-+ (LUMO -I-1)

    together with the four-bpen-shell determinate

    shown above. The conceptual relationship

    between these five configurations and two triplet

    ethylenes at large ethylene-ethylene separations

    is difficult to see. Our general conclusion is that

    in this and similar types of calculations it is best

    to impose a slight symmetry breaking on the

    system is order to obtain localized orbitals at

    large separations.

    With respect to the ab initio calculations the

    configurational behavior of the SF and T1 states

    is simple, they are both largely purely HOMO+

    LUMO open shell states (over 90% mono-

    configurational). With regard to the semi-

    empirical calculations the lowest energy S* state

    (Bz in C,,) has Z-U* not Z-T* character at large

    ethylene-ethylene separations but undergoes

    change in configurational character at small

    separations and becomes JXT*. This is an arti-

    fact of the CNDO method even under our

    parameterization.

    3.3.2.

    Quant it ati ve aspect s-the correlat ion

    energy problem

    With regard to the computed ground state

    energetics of the cycloaddition of ethylene,

    there are a number of previous calculations at

    the ab initio SCF [2b, 2c] SCF-small CI [2b, 2~1

    as

    well as

    ih

    semi-empirical level [2a, 2d, 2f-

    2i]. At the SCF level, the ground state thermo-

    dynamics are greatly in error using a small ST0

    [2b] or STOJG [2c] basis set (talc. ca.

    -80 kcal/mole, obs.,

    -18) [lc]. Essentially

    experimental values are obtained from 4-31G

    and 6-31G* calculations [28]. With regard to

    the STO-calculations [2b] a simple 2 ~2 CI

    showed essentially no energy lowering for

    cyclobutane but about 0.08 au (ca. 50

    kcal/mole) for two ethylenes. Similarly a srpall

    CI treatment [2b] at the STO-3G level

    shows a much greater energy lowering for

    ethylene than for cyclobutane [2c]. These

    small C&mall basis set treatments essentially

    improve the apparent agreement between

    the calculated (after CI) and observed enthalpies

    of the reaction by lowering the energy of

    ethylene more than cyclobutane. This implies

    that the total valence correlation energy of

    two ethylenes is much different than

  • 8/9/2019 Theoretical Analysis of the Cycloaddition of Ethylene

    10/13

    for cyclobutane which, as we wid show, is not

    the case. Our own calculation at the SCF level

    yielded a computed enthalpy for this ethylene

    cyeloaddition reaction of -44 kcaf/mole (exptl.

    - ,R)

    which

    while better than the STO-3G-ST0

    cefculationa quoted above is worse than a 4-

    1 I G or 6-31G estimate. However, imposition

    af Cf at about the 400x400 level makes this

    value in even worse*(ca. +bO kcal/mole)

    ngrcrmcnt with the experimental value than at

    the SCF level The possi,bfe sohnxon of imptov-

    brg rho SCF treatment using a larger basis set

    will not improve the above agreement except at

    the SCb Icvel. The problem resides in the CI

    technique used. An analysis of this problem

    requires some estimate of the correlation ener-

    gies of both ethylene and cyclobutane in order

    to find cut how much of this energy is not being

    cnlculntcd.

    Shown in tnble 3 are the estimated cor-

    rekition cncrgies for each localized carbon 1s

    ?orc, rmd the CH and CC valence orbitals for

    koth ethylene and cyclobutane as obtained from

    a correlative equation [7] relating the SKI-3Ci

    localized orbital size to the correlation energy.

    In the RISCof ethylene, the computed total

    corrclntion energy, OS4 au, is sufficiently close

    10 the estimated vult~c (0.522) 17,291 that some

    lilhk I

    ~~~~IcIIII~B~nergy AimatCs fat ethylene and cyclobutane

    __Ix_ -.-_..

    h~

  • 8/9/2019 Theoretical Analysis of the Cycloaddition of Ethylene

    11/13

    .G. Karsnb

    et

    al. / Theoreiicqt analpis

    o the

    cycloaddition

    ofethykne

    161

    along

    certain portions

    of the surface than along

    others in addition to that obtained from a mere

    HOMO-LUMO indwcd avoided crossing

    behveen SOand S* (i,e. a 2x 2 CI). In the case

    treated here, the computed barrier for the

    concerted 2S+2S transition state for the ground

    State surface decomposition of cyclobutane to

    two ethylenes is less than 60 kcal/mole. This is

    less than the observed activation energy (+63

    kcal/mole) [lc] for a process which is probably

    biradical-like in character [2c]. Our conclusion

    is that an intermediate level CI treatment

    cannot give adequate results on this surface

    regardlessof

    the irrik ab

    initib

    basis

    set

    With regard to the CNDO-CI calculations it

    must be pointed out that a similar size consis-

    tency problem exists with regard to the CI This

    can be demonstrated by examining fig. 3. Orig-

    inally the overall thermodynamics of the

    cyclobutane-two-ethylene reaction was cali-

    brated at the 60x60 CI level by individual cal-

    culations on cyslobutane and a sblgle ethylene

    to give the experiniental value of about

    20 kcal/mole [2f]. In the surface actually

    generated in fig. 3 the same configurational

    composit,an was maintained along a soxface

    which begins with cyclobutane and terminates at

    4.5 A with an ethyb;e dimer The So surface in

    fig. 3 does not asympotitically approach the

    20 kcal/mole calibrated value at large

    separations. This results from the fact that as

    with the ab initio CI calculation it is easier to

    compute a greater percentage of the correlation

    energy when the system is smaller (e.g. ethyl-

    ene) than larger (ethylene dimet) using the same

    number of configurations. However, unlike the

    ab initio calculations, a semi-empirical tech-

    nique can be recalibrated under size consistent

    conditions (i.e. cyclobutane and ethylene dimer

    at large separations). Likewise, since the semi-

    empirical-after-Cl results are parameter

    dependent, a completc CI is not necessary as

    long as the major correlatively important

    configurations are included. In the case of the

    S** surface at large ethylene-ethylene sepata-

    tions the semi-empirical CI treatment must

    include doubly excited four-open-shell

    configurations. In principle, the inclusion of

    such configurations would involve a large CX

    treatment requiring a Parge search ta obtain rhe

    important ones* In fact, the important

    configuratiuns can be detelm,ined by a few pre-

    liminary calculations at various points along ?he

    reaction surface. Ir this mtmner, the semi-

    empirical Cf treatment can be kept wlthin the

    level 100 x 100. However, it must be pointed

    out that the amount of semi-empirical car-

    rel&ion energy will also depend on the size of

    the

    Cl

    treatment. At the 60

    x

    60 level this

    energy was found to be 0.045 au for the So stntc

    of ethylene, increasing to 0,087 au ai the 1 llx

    110 level, the latter containing all totally sym-

    metric mono- and dr;uble+xcitations. The latter

    energy obviously has no relationship to lhc

    above discussed estimated valeilce correlation

    energy (table 2) of 0.33 au. Likewise, a semi-

    empirical potential eaergy curve calibrated with

    one set of configurations Will be out of calibr;l-

    tion under another set.

    With regard to I~.:: problem treated here, $1

    comparison of the semi-empirical and ab initio

    stirfeces (figs. 3 and 4j shows that there is a

    VISUJ esemblance in the two computations ~II

    the region of avoided crossing at 2.0-2.2 A ill:

    well as along the surface between 2.0 to 4.5 h.

    The semi-empirical calculation gives the same

    correlative information as does the ab initio

    one. Both sets of calculations, however, Jo Ilot

    give us an accurate idea

    as to

    the height of the

    So and S*k tates in the region of thy HOMO-

    LUMQ crossing,

    For the triplet diradical surface repara-

    meter&ted CNbO-UHF calculations indicate

    that the reaction of olefin triplets with olefins

    should involve activation energies similar to

    I-adical-olefin reactions. The theoretical

    profi t.s

    of both types of reaction are similar, as

    measured by spin transfer and spin polarizotiotl

    ;rt the transition state as well as the general

    : ppearance of the reaction diagtan?.

    A comparison of ab initio sod semi-empilicl

    I :I calculations for the 2S+ 2S cyctoaddition

  • 8/9/2019 Theoretical Analysis of the Cycloaddition of Ethylene

    12/13

    162

    E.

    Kamb et al. / Theoretical analysis of the cycloadditionfethylene

    surfaces shows similar qualitative behavior.

    e

    demonstrate that the configurational behavior of

    the S** state at large ethylene+zthylene dis-

    taxes is unlike that prediCted by simple orbital

    symmetry rules. This state correlates with two

    triplets of ethylene. FinAlly, we show -that the

    system cannot be treated using ab initio tech-

    niques at the intermediate CI level A separate

    estimate of the valence correlation energy of

    cyclobutane shows that our calculations only

    obtained about 10% of that energy using a

    configuration selection of 5 X low4 au.

    References

    [l] [a) G. Scacchi, C. Richard and M.H. Back. Intern. J.

    Chem. Kinetics 9 (1977) 513;

    (b) G. Scacchi and M H Back, Intern. J. Chem.

    Kinetics 9 (1977) 525;

    (c) SW. Benson and H.E. ONeal, Kinetics Data on

    Gas Phase Reactions, NSRDS-NBS 21 (U.S. Govert.

    Printing ORice, Washington);

    (d) H. Yamazaki, RJ. Cvetanavic and RS. Irwin, J.

    Am. Chem. Sot. 98 (1976) 2198:

    (ej G.S. Hammond, NJ. burro Ad A. Fischer, J. Am.

    Chem. Sot. 83 (1961) 4674;

    (f) J.P. Chesick, J. Am. Chem. Sot. 85 (1963) 3718;

    (g) PJ. Wagner and G.S. Hammond, Advan. Photo-

    them. 5 (1968) 76;

    (h) H.H. Stechl. Aogew. Chem. 2 (1963) 243;

    i) R. Srinivasan and K.A. Hill, J. Am. Chem. Sot. 88

    (1966) 3765;

    (j) D.R. Arnold and R.M. Morchat. Can. J. Chem. 55

    (1977) 393.

    [2] (a) R. Hoffmann. S. Swaminathan, B.G. Ode11 and R.

    Gleiter. J. Am Chem. Sot 92 (1970) 7091;

    (b) J.W. Wright and L. Salem, J. Am. Chem. Sot. 94

    (1972) 322;

    (c) GA. Segal. J. Am. Chem. Sot. 96 (1974) 7892;

    (d) R.C. Bingharo, MJ.S. Dewar and D.H. Lo, I. Am.

    Chem. Sot. 97 (1976) 1294;

    (e) N.D. Epiotis and S. Shaik. J. Am. Chem. Sot. 100

    (1978) 1, 9;

    (f) E.M. Evleth and E. Kassab, J. Am. Chem. Sot. 100

    (1978) 7859;

    (g) MJS. Dewar sod S. Kirscbner. J. Am. Chem. Sot.

    96 (1974) 5246;

    (h) P. Coffey and K. Jug, Theoret. Chim. Acta 34

    (1974) 213;

    (i) L. &brink, C. Fridh, E. L.indhohn and G. AhIgen,

    Chem. Php. 33 (1978) 195.

    [3] R.B. Woodyard and R. Hoffman, The conservation of

    orbital symmetry (V&g Chemie, Weinheim, 1970)

    pp. 15-22; 73-75.

    [4] (a) W. Gerhartz, RD. Poshust+and J. h&hl, J. &.

    Chem. Sor 98 (1976) 6427;

    (b) W. Gerhartz, RD. Poshata and J. Michl, J. Am.

    Chem. Sot. 99 (1977) 4263;

    (c) J.D. Goddard and LG. Csiiadia, dhem. Phys:

    Letters 43 (1976) 73;

    (d) J.D. Goddard and LG. Csiiadia, Chem. Phys.

    Letters 64 (1979) 219;

    (e) J. Michl. Photochem. Photobiol. 2.5 (1977) 141.

    [S] J.L. Whitten, J. Chem. Phys. 44 (1966) 359.

    [6] (a) J.L Whitten artd M. Hackmeyer, J. Chem. Phys. 51

    (1969) 5584;

    (b) R. Daudel. H. Le Rouxo, R. Cimiraglia and J.

    Tomasi. Intern. J..Quantom Chem. 13 (1978) 537.

    [7] M.R. Peterson, R.A. Poirer, R. Daudel and LG.

    Csiiadia, Intern. J. Quantum. Chem. (1980), to be

    published.

    [8] A. Dargelos, D. Liotard and M. Chaillet, Tetrahedron

    28 (1972) 5596.

    [9] JJ. Dannenberg, Angew. Chemie 15 (1976) 579.

    DO) M.T. Rayez-Meaume. J.J. Dannenberg and J.L. Whit-

    ten, J. Am. Chem Sot. 100 (1978) 747.

    [ll] J. Del Bene and H.H. Jaffe, J. Chem. Phys. 48 (1968)

    1807,405O.

    [12] T.A. Halgran, D.A. Kleier, J.H. Hall Jr., L.D. Brown

    and W.N. Lipscomb. 3. Am. Chem. Sot. 100 (1978)

    6595.

    [13] H. Fischer and H. Ko lmar, Theoret. Chii. Acta 13

    (1969) 213.

    [14] C.W. Eaker and J. Hinze, J. Am. Chem. Sot. 96

    (1974) 4084.

    [15] (a) R.J. Buenker and S.D. Peyerimhoff, Chem. Phys. 9

    (1976) 75;

    (b) RJ. Buenker, S.D. Peyerimhofi and H.L. Hsu,

    Chem.Phys. L.etters 11 (1971) 65;

    (c) B.R. Brooks and H_F. Schaefer III. J. Am. Chem.

    sot. 101 (1979) 307.

    [16] A. Warshel and M. Karplus, Chem. Phys. Letters 32

    (1975) 11.

    [17] J.B. Birks, Photophysics of aromatic molecules (Wiley-

    Interscience, New York, 1970).

    [IS] J.A. Pople and D.L. Beveridge, Approximate molecu-

    lar orbital theory (McGraw-Hi. New York, 1970) pp_

    90-92.

    [19] K. Niblaeux, B.O. Ross and EM. Siegbahn, Chem.

    Phys. 26 (1977) 59.

    [20] N.C. Baird and J.R. Swenson, Chem. Phys.

    Letters

    22

    (1973) 183.

    [21] V. Bona&-Koutecky, J. Koutecky and L. Salem, J.

    Am. Chem. Sot. 99 (1977) 842.

    1221 N.J. Turro, Modem molecular photochemistry (Ben-

    jamin, New York, 1978) ch. 11.

    [23] MJ.S. Dewar and S. OliveUa, J. Am. Chem. Sot. 100

    (1978) 5290.

  • 8/9/2019 Theoretical Analysis of the Cycloaddition of Ethylene

    13/13

    E. Kassab et al . l- heoreti cal

    nalysis

    of

    the cyclonddi ti on

    o

    thylene

    1247

    (a) M.S. Platz and J.A. Berson, J. Am. Chem. Sot. 98

    (c) P.C. Hxiharan and J.A. Pople, Theoret. Chim. Acta

    (1976) 6743;

    28 (1974) 213.

    @J)

    J.A.

    Berson, Accounts Chem. Res. 11 (1978) 446.

    1251 C. Galloy and J.C. Lorquet. J. Chem. Phys. 67 (1977)

    4672.

    [29] J.M. McKelvey and A. Streitwieser, J. Am. Chem. Sot.

    99 (1977) 7121.

    &]

    E M

    Evleth and G. Feler, Chem. Phys. Letters 22

    (1973) 499.

    [30] R-T. Buenker, SD. Peyerimhoff and SK. Shih, Chem.

    Phys. Letters 65, (1980) 7.

    [27] F.D. Lewis. Accounts Chem. Res. 12 (1979) 152.

    [28] W.J. Hehre and J.A. Pople, J. Am. Chem. Sot. 97

    (1975) 6941;

    (b) W.A. Lathan, WJ. Hehre and J.A. Pople, J. Am.

    Chem. Sot. 33 (1971) 808;

    [31] J.A. Pople, POSTHartree-Fock: Configuration Inter-

    action, report on the workshop, NRCC (Lawrence

    Berkeley Laboratory. California, 1978) p. 233.

    [32] D. Cremer, J. Am. Chem. Sac. 99 (1977) 1307.

    163