CCP3SURF ACESCIENCENEWSLETTERCollaborative Computational Project3on Surface ScienceNumber26-January2000ISSN1367-370XDaresbury LaboratoryContents1Editorial1 2Scientific Articles2 3SRRTNet-a new global network124High performance computing164.1Cluster-Computing Developments in the UK (16)4.2New HPC Support Mechanisms (22)5Reports on visits256Meetings,Workshops,Conferences296.1Reports on Bursaries (29)6.2Reports from Meetings (32)6.3Upcoming meetings (34)7Abstracts of forthcoming papers37 8Surface Science Related Jobs40 9Members of the working group43 Contributions to the newsletter from all CCP3members are welcome and should be sent to ccp3@eful Links:CCP3Home Page /Activity/CCP3CCP3Program Library /Activity/CCP3+896 SRRTNet /Activity/SRRTNetDLV /Activity/DLVCRYSTAL /Activity/CRYSTALCASTEP /Activity/UKCPMany useful items of software are available from the UK Distributed Computing Support web site,DISCO /Activity/DISCOEditors:Dr.Klaus Doll and Dr.Adrian Wander,Daresbury Laboratory,Daresbury, Warrington,WA44AD,UK1EditorialThe renewal of CCP3,which is due in the summer,is on all our minds,and this edition of the newsletter reminds us of our aims and achievements.Theflagship project supported by the post-doc is at the heart of the CCP–new programs can be developed which would not get offthe ground otherwise.Over the last three years CCP3has been very lucky to have had Klaus Doll working on the development of analytic gradient methods for the CRYSTAL electronic structure package.This will lead to much more efficient structure optimization,of particular benefit to surface scientists where surface relaxation and reconstructions are so important.Klaus’achievements so far are described in thefirst article,and it is most satisfactory that tests on the CO molecule and bulk MgO have proved successful.The next step is to build in symmetry,to achieve greater computational efficiency,and then the new code can be released.As theflagship in the renewal proposal,the working group has chosen the development of methods to study the electronic structure and physical properties of large clusters.These clusters themselves possess surface-like properties, but at the same time it is proposed to study their interaction with surfaces.Such systems are a topic of active research for several of the members of the working group,both theoretical and experimental,and it is expected that their expertise will contribute greatly to the success of the project.Having Adrian Wander as a permanent member of staffat Daresbury supporting CCP3will lead to very welcome support for the synchrotron radiation community in the development of new surface program packages.In this newsletter he describes developments in SRRTNet,originally an American collaboration for providing sup-port for surface scientists using synchrotron radiation,which is likely to develop into an international collaboration based on the CCP model.Adrian is also in discussion with the Daresbury-based X-ray community with a view to developing new codes for the analysis of near-edge spectra in a wider range of systems than can be tackled at the moment,using the improved self-consistent electron potentials available for complex materials.This work would be based at Daresbury,and will form part of the CCP3collaboration.This issue contains short articles on our visitor programme,and by Ally Chan (Nottingham)and Yu Chen(Birmingham)who received student bursaries for participating in ECOSS-18.Please continue to apply for CCP3support!It is interesting to read in the pieces by Ally and Yu what most impressed them at ECOSS–I was struck by Ally’s comment that surface science has broadened to include nanoparticles and nanowires.Just what we thought in our choice offlagship project next time round.John Inglesfield12Scientific ArticlesAnalytical Hartree-Fock gradients for periodic systemsK.Doll,V.R.Saunders,and N.M.HarrisonCLRC,Daresbury Laboratory,Daresbury,Warrington,WA44AD,UKWe report on the progress of the implementation of analytic gradients in the program package CRYSTAL.The algorithm is briefly summarised and tests illustrate that highly accurate analytic gradients of the Hartree-Fock energy can be obtained for molecules and periodic systems.IntroductionComputational materials science has been a fast growingfield in the last years. This is mainly because methods which were developed earlier(density functional theory,molecular dynamics,Hartree-Fock and correlated quantum chemical meth-ods,Monte Carlo schemes,the GW method,etc)can now be applied to demanding realistic systems due to the increase in computational resources(faster CPUs,par-allelisation,cheaper memory and diskspace).CCP3is a collaboration in the area of surfaces and interfaces where progress de-pends on an interaction between experimental and theoretical approaches.There-fore,codes which provide a better theoretical understanding are important.One of the key issues in surface science is the determination of surface structure and adsorption energetics.From the computational point of view,a fast structural optimisation must be possible.Availability of numerical or analytical gradients facil-itatesfinding a minimum energy structure,and availability of analytical gradients can make optimisation algorithms more efficient.As a rule of thumb,analyticalgradients are about N3times more efficient than numerical gradients(with N beingthe number of variables).Also,for future developments such asfinding transition states,gradients are essential.Analytical gradients in quantum chemistry were pioneered by Pulay who did the first implementation for multicentre basis sets[1].In many molecular codes based on quantum chemistry methods,analytical gradients are now implemented and gradient development has become an important task in quantum chemistry[2,3,4,5].Simi-larly,in solid-state codes such as CASTEP,WIEN,or LMTO,analytic gradients are available.Analytic Hartree-Fock gradients have already been implemented in a code for systems periodic in one dimension[6].CRYSTAL[7,8]was born in Turin and is now jointly developed in Turin and Daresbury.CRYSTAL was initially designed to deal with the exact exchange in and to solve the Hartree-Fock equations for real systems.With the modern versions of the code,density-functional2calculations or calculations using Hybrid functionals such as B3LYP with the ad-mixture of exact exchange are also possible.The target of this project,which began in October1997,was the implementation of analytical gradients in CRYSTAL and in autumn1999,thefirst test calculations on periodic systems were performed.In this article,we try to outline the theory and implementation of analytical gra-dients.We try to keep the mathematics at a minimum;a more formal publication is intended in the near future[9].A very comprehensive summary of the theory underpinning CRYSTAL will appear in the future[10].Total energyFirst,we want to briefly summarise how the total energy is obtained.The total energy consists of•kinetic energy of the electrons•nuclear-electron attraction•electron-electron repulsion•nuclear-nuclear repulsionCRYSTAL,similar to molecular codes such as GAMESS-UK,MOLPRO(Stuttgart and Birmingham),GAUSSIAN,TURBOMOLE,etc,solves the single particle Schr¨o dinger equation and a wavefunction is calculated.The wavefunction is based on crystalline orbitalsΨi( r, k)which are linear combinations of Bloch functionsΨi( r, k)= µaµ,i( k)ψµ( r, k)(1) with the Bloch functions constructed fromψµ( r, k)= gφµ( r− Aµ− g)exp(i k g)(2) g are direct lattice vectors, Aµdenotes the coordinate of the nuclei.φµare the basis functions which are Gaussian type orbitals.For example,an s-type function centred at R a=(X a,Y a,Z a)is expressed asφ(α, r− R a,n=0,l=0,m=0)=φµ( r− R a)= Nexp(−α( r− R a)2).In molecular calculations,no mathematical problem arises from any of the interac-tions.In periodic systems,however,there are several divergent terms which have to3be dealt with:for example,in a one dimensional periodic system with lattice con-stant a and n being an index numerating the cells,the electron-electon interaction per unit cell would have contributions like:∞n=1e2na(3)This sum is divergent(similarly in two and three dimensions).Therefore,an indi-vidual treatment of this term is not possible.Instead,all the charges(nuclei and electrons)are partitioned and a scheme based on the Ewald method is used to sum the interactions[11].The Hartree-Fock equations are solved in terms of Bloch functions because the Hamiltonian becomes block-diagonal(i.e.at each k-point the equations are solved independently).The wavefunction coefficients aµ,i are optimised due to this procedure and the total energy can be evaluated.For the computation of gradients,the dependence of the total energy on the nuclear coordinates must be analysed.There are three dependencies of the total energy on the nuclear coordinates:•nuclear-nuclear repulsion and nuclear-electron attraction:obviously,the coor-dinates of the nuclei enter•wavefunction coefficients(or density matrix):we will obtain a different solution with different density matrix when moving the nuclei•basis functions:the basis functions are centred at the position of the nu-clei and therefore moving the nuclei will change integrals over the basis func-tions.These additional terms are called Pulay forces.They are missing when the Hellmann-Feynman theorem is applied and therefore Hellmann-Feynman forces often differ substantially from energy derivative forces in the case of a local basis set(see[1]and references therein).Density matrix derivatives are difficult to evaluate.However,for the solution of the Hartree-Fock equations,this problem can be circumvented and instead a new term is introduced,the so-called energy-weighted density matrix which is easily evaluated [12].However,this is only strictly correct for the exact Hartree-Fock solution. In practice,convergence is achieved up to a certain numerical threshold(e.g.a convergence of10−6E h of the total energy corresponding to27.2114×10−6eV).For very accurate gradient calculations,it may be necessary to make this threshold even lower.The remaining main problem is to generate all the derivatives of the integrals. In a second step,these derivatives have to be mixed with the density matrix.4Evaluation of integralsIn this section we summarise the types of integral which occur.The simplest type is the overlap integral between two basis functions at two centres:Sµν R k R l= φµ( r− R k)φν( r− R l)d3r(4) Obviously we can shift R k to the origin,and suppressing 0in the notation,we obtain:Sµν R i= φµ( r)φν( r− R i)d3r(5) with R i= R l− R k.A kinetic energy integral has the form:Tµν R i= φµ( r)(−12∆ r)φν( r− R i)d3r(6) the nuclear attraction integral has the form:Nµν R i= φµ( r)Z c| r− A c|φν( r− R i)d3r(7) and the electron-electron interaction has the form:Bµν R iτσ R j = φµ( r)φν( r− R i)φτ( r′)φσ( r′− R j)| r− r′|d3rd3r′(8)These integrals are in principle sufficient to deal with molecules.In the case of periodic systems,new types of integrals appear(e.g.multipolar integrals,integrals over the Ewald potential and its derivatives)[11,13,14].The fast evaluation of integrals is one of the main issues in the development of quan-tum chemistry codes.CRYSTAL uses a McMurchie-Davidson algorithm[15].Its idea is to map a product of two Gaussian type orbitals at two centres in an expan-sion of Hermite polynomials at an intermediate centre.This algorithm has proven to efficiently evaluate integrals,although in recent years progress in this specialised field of quantum chemistry has been made(see for example the introduction in[16] or two recent reviews[17,18]).The expansion[15,14]looks like:5φ(α, r− A,n,l,m)φ(β, r− B,n′,l′,m′)= t,u,v E(n,l,m,n′,l′,m′,t,u,v)Λ(γ, r− P,t,u,v)(9)withγ=α+βand P=α A+β Bα+β.Λis a so-called Hermite Gaussian type function Λ(γ, r− P,t,u,v)= ∂∂P x t ∂∂P y u ∂∂P z v exp(−γ( r− P)2)(10)The start value E(0,0,0,0,0,0,0,0,0)=exp(−αβ( B− A)2)can be verified by inserting it in equation9.It can be derived from the Gaussian product rule[19,20]:exp(−α( r− A)2)exp(−β( r− B)2)=exp −αβα+β( B− A)2 exp −(α+β) r−α A+β Bα+β 2(11) General values E(n,l,m,n′,l′,m′,t,u,v)are obtained from recursion relations[15, 14].The E-coefficients depend on the distance( B− A),but not on P or r.All the integrals can be expressed in terms of E-coefficients[15,14,11,13].Evaluation of gradients of the integralsOne of the issues of the gradient project is to generalise the algorithms used to generate the energy integrals to obtain the gradients of the integrals.This madea new implementation of recursion relations necessary which are used to obtainthe coefficients G in the expansion of the gradients of the integrals in Hermite polynomials.∂Φ(α, r− A,n,l,m)Φ(β, r− B,n′,l′,m′)∂A x= t,u,v G A x(n,l,m,n′,l′,m′,t,u,v)Λ(γ, r− P,t,u,v)(12)Once the coefficients are known,the integration can be performed.The integrationfor the case of gradients of integrals is similar to the case of integrals for the total energy.The only difference is that,instead of the coefficientsE(n,l,m,n′,l′,m′,t,u,v)which enter the energy expression,the gradient coefficientsG A x(n,l,m,n′,l′,m′,t,u,v),G A y,G A z,G B x,G B y,and G B z6are used.The coefficients G B x can efficiently be obtained together with the coeffi-cients G A x[21].For example,the evaluation of the overlap integral is done as follows:Sµν R i = φµ( r)φν( r− R i)d3r=t,u,v E(n,l,m,n′,l′,m′,t,u,v)Λ(γ, r− P,t,u,v)d3r=E(n,l,m,n′,l′,m′,0,0,0)Λ(γ, r− P,0,0,0)d3r=ME(n,l,m,n′,l′,m′,0,0,0)From thefirst line to the second,we have used the McMurchie-Davidson scheme, from the second to the third we exploited a property of the Hermite Gaussian type functions:all the integrals of the type Λ(γ, r− P,t,u,v)d3r with t=0or u=0or v=0vanish because of the orthogonality of these functions.The integration(fromthe third to the fourth line)is trivial.M is a normalisation constant. Calculating the gradient is easy once we know the new expansion:∂Sµν R i ∂A x =∂∂A xφµ( r)φν( r− R i)d3r=∂ t,u,v E(n,l,m,n′,l′,m′,t,u,v)Λ(γ, r− P,t,u,v)∂A x d3r=t,u,v G A x(n,l,m,n′,l′,m′,t,u,v)Λ(γ, r− P,t,u,v)d3r=G A x(n,l,m,n′,l′,m′,0,0,0)Λ(γ, r− P,0,0,0)d3r=MG A x(n,l,m,n′,l′,m′,0,0,0)This way,all the derivatives can be calculated!There are some integrals which involve three centres(for example nuclear attraction)where we exploit translational invariance:∂∂C x =−∂∂A x−∂∂B x(13)because the value of the integral is invariant to a simultaneous uniform translation of the three centres.Four centre integrals can be reduced to a product of two integrals over two centres which makes the calculation of gradients straightforward.As a whole,the calculation of gradients of the integrals is closely related to calcu-lating the integrals itself.This means that most of the subroutines can be used for7the gradient code.One of the main differences is that array dimensions need to be changed-dealing with gradients is similar to increasing the quantum number(a derivative of an s-function is a p-function,and so on).However,the task of adjusting the subroutines should not be underestimated.After obtaining the derivatives of the integrals,we mix them with the density ma-trix just like in the energy calculation.We have to take into account the new term which arose because we did not calculate a density matrix derivative—the energy weighted density matrix.Again,coding this additional term can be done by modi-fying existing subroutines.After this,wefinally obtain the forces on the individual atoms.Results from test calculationsIn this section,we summarise results from test calculations.We have considered the CO molecule which was arranged as a single molecule,as a molecule which is peri-odically reproduced with a periodicity of4˚A in one spatial direction(”polymer”), periodically reproduced with a periodicity of4˚A in two spatial directions(”slab”), and periodically reproduced with a periodicity of4˚A in three spatial directions (”solid”).Because of the large distance of4˚A,the molecules can be considered as nearly independent and the forces are quite similar.Still,the calculation of energy and gradient is completely different and therefore this is an important test of Ewald technique and multipolar expansion.The results are given in table1.The results agree in the best case to at least6digits which is the numerical noise and in the worst case up to4digits.The difference between analytic and numerical gradi-ents in periodic systems mainly originates from an approximation made within the evaluation of the integrals[22]and from the number of k-points which affects the accuracy of the energy-weighted density matrix.In table2,we display results from a MgO solid with one oxygen atom slightly distorted from the symmetrical position.Again,the forces agree well up to5digits with numerical derivatives.Future developments and ConclusionThe present version of the code is able to calculate Hartree-Fock forces for periodic systems up to a precision of4and more digits.There is no extra diskspace needed and the additional memory usage is moderate.This code will certainly be useful for structural optimisation and for future program development towards molecular dynamics or the calculation of response functions.The present version,however,is not yet ready for a release.Instead,the following steps are necessary:Firstly,the usage of symmetry must be implemented.This is of highest importance to make the code fast enough so that it can be used for practical optimisations.We expect8Table1:Force on a CO molecule with a carbon atom located at(0˚A,0˚A,0˚A)and an oxygen atom located at(0.8˚A,0.5˚A,0.4˚A).In the periodic case,the molecule is generated with a periodicity of4˚A.This means,that in one dimension,for example,there would be other molecules with a carbon atom at(n×4˚A,0˚A,0˚A)and an oxygen atom at((n×4+0.8)˚A,0.5˚A,0.4˚A),with n running overall positive and negative integers.Forces are given in E h,with E h=27.2114eVand a0=0.529177˚A.Higher ITOLs means a lower level of approximation in the evaluation of the integrals[22].ITOLs) k-points) numerical force0.3769140.37660(0.37664)0.376310.37566(0.37566) analytical force0.3769130.37663(0.37665)0.376330.37588(0.37578))on the atoms of an MgO solid.The MgO solid was chosen Table2:Forces(in E ha0to have an artificially high lattice constant of6.21˚A to make the calculation faster. Coordinates are given in fractional units,e.g.the second Mg is at0˚A,0.5×6.21˚A,0.5×6.21˚A.A normal fcc lattice would be obtained if the sixth atom(Oxygenat0.53,0,0)was at(0.5,0,0).Moving this atom from its normal position has ledto the nonvanishing forces.Mg(0.00.00.0)-0.03018-0.03019Mg(0.00.50.5)-0.00314-0.00314Mg(0.50.00.5)0.008950.00895Mg(0.50.50.0)0.008950.00895O(0.50.50.5)-0.00379-0.00379O(0.530.00.0)0.004290.00430O(0.00.50.0)0.007460.00746O(0.00.00.5)0.007460.00746that a version of the present code with symmetry will already be fast enough to compete with numerical derivatives.Further developments will be the coding of the bipolar expansion(a method to evaluate the electron-electron repulsion integrals faster),and sp-shells(s and p shells are often chosen to have the same exponentsto make the evaluation of integrals faster).Also,the newly written subroutines arenot yet optimal and they will certainly go through a technical optimisation(moreefficient coding).In later stages,the code should be made applicable to metals (there is an extra term coming from the shape of the Fermi surface[23]which is notyet coded)and to magnetic systems(unrestricted Hartree-Fock gradients).Finally, pseudopotential gradients and density functional gradients should be included. References[1]P.Pulay,Mol.Phys.17,197(1969).[2]P.Pulay,Adv.Chem.Phys.69,241(1987).[3]P.Pulay,in Applications of Electronic Structure Theory,edited by H.F.Schae-fer III,153(Plenum,New York,1977).[4]H.B.Schlegel,Adv.Chem.Phys.67,249(1987).[5]T.Helgaker and P.Jørgensen,Adv.in Quantum Chem.19,183(1988)[6]H.Teramae,T.Yamabe,C.Satoko,A.Imamura,Chem.Phys.Lett.101,149(1983).[7]C.Pisani,R.Dovesi,and C.Roetti,Hartree-Fock Ab Initio Treatment of Crys-talline Systems,edited by G.Berthier et al,Lecture Notes in Chemistry Vol.48(Springer,Berlin,1988).[8]V.R.Saunders,R.Dovesi,C.Roetti,M.Caus`a,N.M.Harrison,R.Orlando,C.M.Zicovich-Wilson crystal98User’s Manual,Theoretical Chemistry Group, University of Torino(1998).[9]K.Doll,V.R.Saunders,N.M.Harrison(in preparation)[10]V.R.Saunders,N.M.Harrison,R.Dovesi,C.Roetti,Electronic StructureTheory:From Molecules to Crystals(in preparation)[11]V.R.Saunders,C.Freyria-Fava,R.Dovesi,L.Salasco,and C.Roetti,Mol.Phys.77,629(1992).[12]S.Bratoˇz,in Calcul des fonctions d’onde mol´e culaire,Colloq.Int.C.N.R.S.82,287(1958).[13]R.Dovesi,C.Pisani,C.Roetti,and V.R.Saunders,Phys.Rev.B28,5781(1983).[14]V.R.Saunders,in Methods in Computational Molecular Physics,edited by G.H.F.Diercksen and S.Wilson,1(Reidel,Dordrecht,Netherlands,1984).[15]L.E.McMurchie and E.R.Davidson,put.Phys.26,218(1978).[16]R.Lindh,Theor.Chim.Acta85,423(1993).[17]T.Helgaker and P.R.Taylor,in Modern Electronic Structure Theory.Part II,World Scientific,Singapore,725(1995)[18]P.M.W.Gill,in Advances in Quantum Chemistry,edited by P.-O.L¨o wdin,141(Academic Press,New York,1994)[19]S.F.Boys,Proc.Roy.Soc.A200,542(1950).[20]R.McWeeny,Nature166,21(1950).[21]T.Helgaker and P.R.Taylor,Theor.Chim.Acta83,177(1992).[22]The integrals Bµν R iτσ R j =Bτσ R jµν R iwhich should have the same value,are notnecessarily evaluated within the same level of approximation—this is nearly inevitable for periodic systems,as enforcing this symmetry would require a much higher computational effort and much more data storage.The derivation of the equations for the analytic gradients,however,relies on these integrals be-ing equivalent.Therefore,the introduced asymmetry will lead to inaccuracies in the gradients.This can be controlled with the ITOL-parameters(tolerances as described in the CRYSTAL manual[8])which control the level of approx-imation.Higher ITOLs lead to a higher accuracy in the forces.However,the defaults appear to give forces with an accuracy up to4digits which should be good enough for most purposes.[23]M.Kertesz,Chem.Phys.Lett.106,443(1984).3SRRTNet-a new global networkFrascati’99-Birth of a NetworkScientific MeetingFrom the23rd to the25th September1999,a workshop on Theory and Computation for Synchrotron Radiation was held at the laboratory in Frascati just outside Rome, Italy.This was the third in an ongoing series of meetings on various aspects of synchrotron radiation,and follows meetings on Theory and Computation for Syn-chrotron Applications held at the Advanced Light Source in Berkeley in October 1997and Needs for a Photon Spectroscopy Theory Center held at the Argonne National Laboratory in August1998.This was an excellent meeting,featuring a variety of high quality scientific pre-sentations from both experimental and theoretical participants.Thefirst day was devoted to presentations concentrating on resonant x-ray processes and orbital or-dering effects,particularly in V2O5.The second day then moved on to discussions of photoemission,photoelectron diffraction and holography,and studies of high T c superconductors.This day was concluded with an excellent conference dinner which finished rather late!Thefinal day then concluded with discussions of EXAFS,and x-ray spectroscopies.The overheads used in all the presentations can be viewed on line at http://wwwsis.lnf.infn.it/talkshow/srrtnet99.htmSRRTNet DiscussionsThe Friday programme also included a two hour session devoted to the idea of forming a global network concentrating on theory for synchrotron radiation re-search based research.The session began with a talk from Michel Van Hove of the Lawrence Berkeley National Laboratory who outlined the purposes and function of the proposed network.This was then followed by presentations by John Rehr of the University of Washington who highlighted moves to extend the synchrotron radia-tion research theory network(SRRTNet)in North America,by Maurizio Benfatto of the INFN Frascati,who presented the European perspective,by Kenji Makoshi of Himeji Institute of technology who discussed the Japanese efforts and by Adrian Wander of the Daresbury Laboratory who presented CCP3as a possible model of how the network could be run.The concept of establishing a global network was received with enthusiasm from all present.OutcomeGiven the support of the meeting for the concept of global network of this sort,it was decided to extend SRRTNet into the global arena.The aims of the network are:•To provide a central repository for information of relevance to synchrotron radiation research•To develop theoretical methods pertaining to the experiments performed on synchrotron facilities•To provide state of the art and user friendly software for the analysis and interpretation of experiments•To provide training in the use of relevant software through workshops and site visits•To host visiting scientists•To hold periodic workshops for the dissemination of new results and method-ologiesThe directors of the network are Michel Van Hove and John Rehr.As afirst step in the development of the network,Daresbury has agreed to host the web pages, and theoretical groups have been contacted and ask to provide input to this central web hub of what will grow into a globe encompassing network.If you are interested in contributing to the network and missed our e-mail announcement,the invitation letter follows;Dear Colleague,You may know of the recently established Synchrotron Radiation Research Theory Network(SRRTNet).We are contacting you to invite you,and all theorists inter-ested in this topic,to actively participate in the next phase of the network. SRRTNet aims to provide theory for experiments that use synchrotron radiation,by means of a global,web-based network linking theoretical and experimental research groups.The driving philosophy is to promote interactions between theory and exper-iment for mutual benefit,by means of web-based information,workshops,exchange of theoretical methods and computer codes,as well as establishing visiting scientist programs.At the last SRRTNet workshop,conducted at Frascati near Rome in September1999, it was decided to strengthen the global character of this network by establishing a cen-tral,web-based source of information.Daresbury Laboratory is hosting this web site with Dr.Adrian Wander acting as editor.It is anticipated that all synchrotron facilities will provide direct links for their users to this web site,and consequently we expect this site to grow into an essential resource for synchrotron radiation re-searchers.An importantfirst function of the web site will be to provide information about theorists’research interests and links to relevant web pages.The network will be all the more valuable as this coverage becomes complete:it will thus allow theorists and experimentalists alike tofind the best sources of information about the various methods for solving specific scientific problems.The purpose of this message is to ask you to provide such information and links about your group.You may visit the new web site/Activity/SRRTNetand see not only an overview of the network in general,but also the beginnings of such information about specific theoretical groups.The idea is to put a list of your research topics on the SRRTNet web site,while providing links to your own web site for more detailed and up-to-date information. If you prefer,the SRRTNet site can itself host a more complete web page covering your activities.The information we wish to present(or link to)includes as many as possible of the following items:•your topics of scientific activity related to synchrotron radiation(directly or by methodology);•your computer codes,with their capabilities and availability;•your publications,such as abstracts,papers,databases and web-presentations;•how to contact you or your group.。