A scalable parallel algorithm for large-scale reactive force-field molecular dynamics simulations

Size: px
Start display at page:

Download "A scalable parallel algorithm for large-scale reactive force-field molecular dynamics simulations"

Transcription

1 Computer Physcs Communcatons A scalable parallel algorthm for large-scale reactve force-feld molecular dynamcs smulatons Ken-ch Nomura Rav K. Kala Achro Nakano Prya Vashshta Collaboratory for Advanced Computng and Smulatons Department of Computer Scence Department of Physcs & Astronomy Department of Chemcal Engneerng & Materals Scence Unversty of Southern Calforna Los Angeles CA USA Receved 18 June 2007; receved n revsed form 16 August 2007; accepted 17 August 2007 Avalable onlne 15 September 2007 Abstract A scalable parallel algorthm has been desgned to perform multmllon-atom molecular dynamcs MD smulatons n whch frst prncplesbased reactve force felds ReaxFF descrbe chemcal reactons. Envronment-dependent bond orders assocated wth atomc pars and ther dervatves are reused extensvely wth the ad of lnked-lst cells to mnmze the computaton assocated wth atomc n-tuple nteractons n 4 explctly and 6 due to chan-rule dfferentaton. These n-tuple computatons are made modular so that they can be reconfgured effectvely wth a multple tme-step ntegrator to further reduce the computaton tme. Atomc charges are updated dynamcally wth an electronegatvty equalzaton method by teratvely mnmzng the electrostatc energy wth the charge-neutralty constrant. The ReaxFF-MD smulaton algorthm has been mplemented on parallel computers based on a spatal decomposton scheme combned wth dstrbuted n-tuple data structures. The measured parallel effcency of the parallel ReaxFF-MD algorthm s on IBM BlueGene/L processors for a 1.01 bllon-atom RDX system Elsever B.V. All rghts reserved. PACS: c; Ns; Db Keywords: Molecular dynamcs; Reactve force feld; Parallel computng 1. Introducton Molecular dynamcs MD s an atomstc smulaton method for studyng a wde class of materals such as metals ceramcs and bomolecules under ambent as well as extreme condtons. Contnuous efforts are beng made to ncorporate chemcal reactons nto MD smulatons so that broad materal processes such as catalyss corroson and detonaton can be studed. Currently a number of software packages are avalable for frst-prncples MD smulatons based on quantum mechancal QM calculatons to descrbe reactve events such as bond breakng and formaton [1]. One of the longstandng problems n materals scence s mechancally nduced chemcal reacton such as stress corroson crackng and shock-nduced detonaton [23] whch may drastcally change materal propertes. A study of chemcal reactons nduced by mechancal loadng s a challengng task because they are often nseparable from geometrcal heterogenety nvolvng mllons of atoms. Ther understandng thus requres the combnaton of vast spatal extent and accurate chemstry. Because of the ntensve computatonal requrement and poor scalablty however the system sze n QM smulatons s usually lmted to 10 2 atoms. One approach to tackle ths problem s the hybrdzaton of dfferent methodologes such as QM/molecular mechancs MM schemes [4 8]. However they requre adaptve selecton of reactve stes and sophstcated treatment of QM/MM boundares whch are nontrval for complex mechanochemcal processes. An alternatve approach s to ft essental chemcal reactons by sem-emprcal force felds [9 11] n whch parameterzed nteratomc potentals are traned to reproduce datasets obtaned wth accurate QM calculatons. Recent advances n reactve force * Correspondng author. E-mal address: anakano@usc.edu A. Nakano /$ see front matter 2007 Elsever B.V. All rghts reserved. do: /.cpc

2 74 K.-. Nomura et al. / Computer Physcs Communcatons felds ReaxFF [12 14] have opened up a possblty to study reactve processes n a wde range of materals such as hydrocarbons [12] slca [13] alumna[15] ntramnes [14] catalysts [16] and polymers [17]. We have recently developed a scalable mplementaton of ReaxFF MD on massvely parallel computers whch provdes us wth the requste couplng of quantummechancally accurate descrpton of chemcal reactons and large length-scale mechancal processes to study mechanochemcal processes under extreme condtons. For example our parallel ReaxFF algorthm has enabled mllon-to-bllon atom MD smulatons on a large number 10 5 of processors to study shock-nduced ntaton of energetc materals wth focuses on the effects of mcrostructures e.g. vods and gran boundares and nanofludcs e.g. nanoets on chemcal reactons [1819]. In ths paper we present the desgn and mplementaton of our parallel ReaxFF algorthm along wth ts scalablty tests on varous parallel computers. Secton 2 descrbes the ReaxFF MD algorthm and ts parallel mplementaton s dscussed n Secton 3. Results of benchmark tests are presented n Secton 4 and Secton 5 contans summary. 2. Reactve force feld molecular dynamcs algorthm In the reactve force feld ReaxFF nteratomc nteractons between N atoms comprse valence nteractons descrbed wth bond order BO as well as noncovalent nteractons [16]: E ReaxFF {r } {r k } {r kl } {q } {BO } = E bond + E lp + E over + E under + E val + E pen + E coa + E tors + E con + E hbond + E vdwaals + E Coulomb 1 n whch the valence nteractons nclude the bondng energy E bond lone-par energy E lp overcoordnaton energy E over undercoordnaton energy E under valence-angle energy E val penalty energy E pen 3-body conugaton energy E coa torson-angle energy E tors 4-body conugaton energy E con and hydrogen bondng energy E hbond.ineq.1 the total energy s a functon of relatve postons of atomc pars r trplets r k and quadruplets r kl as well as atomc charges q and bond orders BO between atomc pars. The noncovalent nteractons comprse van der Waals energy E vdwaals and Coulomb energy E Coulomb whch are screened by a taper functon [12]. ReaxFF also ncorporates a charge transfer mechansm to descrbe realstc chemcal reacton pathways. The envronment-dependent charge dstrbuton s descrbed usng the electronegatvty equalzaton method EEM [20] n whch ndvdual atomc charges vary n tme. The analytcal formulas for these energy terms are gven n the followng subsectons Bond-order calculaton An essental buldng block of ReaxFF s the bond order that descrbes the nature of covalent bonds n terms of three exponental functons correspondng to σπ and double-π bonds. The bond order BO s an attrbute of atomc par where = 1...N wth N beng the number of atoms. The BO calculaton n ReaxFF nvolves two steps. The frst step computes raw BOs BO that are smply functons of the nteratomc dstances of all pars of atoms wthn sngle covalent-bond cutoff dstance r cb typcally 3 4 Å. In the second step each raw BO s corrected as a functon of the valences and of the atoms comprsng the par to obtan the full bond order BO. Many-body effects are mplctly ncluded through the valency terms. We frst compute the raw bond order BO as a sum of σπ and double-π bond contrbutons [ pbo2 ] [ pbo4 ] [ BO = BO σ + BO π ππ r r + BO = exp p bo1 ro σ + exp p bo3 ro π + exp p bo5 r r ππ o pbo6 ] where r = r = r r s the nteratomc dstance between th and th atoms r s the poston of the th atom. These raw bond orders are subsequently corrected to produce correct atomc valences as 2 BO = BO σ + BOπ + BOππ BO σ = BO σ f 1 f4 BO f5 BO BO π = BO π f 1 f1 f4 BO f5 BO BO ππ = BO ππ f 1 f1 f4 BO f5 BO = Val + f 1 = 1 2 n BO Val + f 2 Val + f 2 + f 3 + Val + f 2 Val + f 2 + f 3 f 2 = exp p boc1 + exp p boc1 f 3 = 1 { 1 [ ln exp pboc2 p boc2 2 + exp p boc2 ]} 3a 3b 3c 3d 3e 3f 3g 3h

3 K.-. Nomura et al. / Computer Physcs Communcatons Table 1 Bond-order calculaton algorthm Algorthm BO_calc {r = 1...N}=set of atomc postons where N s the number of atoms {n}=set of neghbor lsts where n s the lst of neghbor atoms of the th atom {BO }=set of full bond orders { }=set of full valences {BO }=set of raw bond orders { }=set of raw valences 1 compute raw bond orders {BO } and valences { } for each atom r for each atom r n BO BO σ r + BO π r + BO ππ r // see Eq. 2 Val // Val s atomc valency n Eq. 3e for each atom r for each atom r n + BO // Eq. 3e 2 compute full bond orders {BO } and valences { } for each atom r for each atom r n BO BO f 1 f 4 f 5 // f 1 f 4 and f 5 are BO correcton functons n Eq. 3 Val for each atom r for each atom r n + BO // Eq. 4 f 4 BO = exp p boc3 p boc4 BO BO boc f 5 BO = exp p boc3 p boc4 BO BO boc boc = Val boc + BO n 3 + p boc5 3 + p boc5 3k where n s the set of neghbor atoms that are wthn sngle covalent-bond cutoff dstance r cb from the th atom and the parameters n Eqs. 2 and 3 are lsted n Table A.1 n Appendx A. After the bond orders BO are computed for all pars the valency of the th atom s computed as = Val + BO. 4 n Table 1 shows the BO calculaton algorthm BO_calc. The algorthm also computes full valences as a functon of the corrected bond orders BO Eq. 4. To avod repeated computaton of BO dervatve values n subsequent energy and force calculatons we compute the dervatve of BOs wth respect to the raw BO and valency here and keep them untl the atomc coordnates are updated Energy and force calculatons The varous nteracton functons n Eq. 1 represent a varety of chemstres whch can be grouped accordng to 1 valence or noncovalent type 2 number of explctly nvolved atoms and 3 degree of the neghbor-atom effect. Here the number of explctly nvolved atoms s the nestng level of summatons that consttute each energy term e.g. 3 for the trply nested summaton n E val n Secton below. The degree of the neghbor-atom effect s defned as the number of mplctly nvolved loops through valency terms n the energy functons. For example the corrected valency term s of second degree snce ts evaluaton ncludes doubly nested sums: BO BO = BO BO BO m BO n. 5 m n

4 76 K.-. Nomura et al. / Computer Physcs Communcatons Table 2 Classfcaton of nteracton functons Number of explct atoms Energy terms 1 E lp 2 E over 3 E under 3 2 E bond 1 E vdwaals 0 a E Coulomb 0 a 3 E val 2 E pen 2 E coa 2 E hbond 2 4 E tors 2 E con 2 a Noncovalent nteractons. The numbers n the parentheses are the greatest degrees of the neghbor atom effect. Fg. 1. Schematc of atomc confguratons n energy terms: a 1-body b 2-body c 3-body d 4-body e hydrogen bondng and f noncovalent nteractons respectvely. A gray sphere represents the poston of the prmary atom n each energy term. Open bars represent covalent bonds whle dotted lnes are noncovalent bonds. In ths subsecton we frst explan the nature of nteracton functons whch s essental to map them onto a parallel computer. We then present actual algorthms. Table 2 classfes the energy terms accordng to the number of explctly nvolved atoms. The noncovalent nteractons are parwse 2-body functons wth two explct atoms whle valence terms nvolve up to four explct atoms. Each explct atom mplctly nvolves effects from surroundng atoms through BO and valency terms. The mplct atoms requre addtonal computatons and ncur ncreased cutoff dstances. For example n force calculaton the corrected valency term causes a doubly nested loop over a neghbor lst of the th atom wth the assocated computatonal cost OM 2 where M s the average number of covalent bonds per atom. The nteracton cutoff length s accordngly ncreased to 2r cb. The level of neghbor lst s thus regarded as the degree of envronment effect. The numbers n the parentheses n Table 2 shows the greatest degree of neghbor-atom effect among explct atoms. Fg. 1 shows the confguratons of atoms nvolved n the computaton of 1-body 2-body 3-body 4-body hydrogen bondng and noncovalent energy functons respectvely. In the fgure a gray sphere represents the poston of a prmary atom scanned by the outermost loop of each energy-term calculaton. The prmary atoms play a pvotal role on our parallel mplementaton n Secton 3. Note that the energy terms n ReaxFF share common computatonal structures. In addton to algorthms to compute ndvdual energy terms we defne two algorthms force_bo and force_ whch consttute the kernel of energy and force calculatons. The force_bo and force_ algorthms are gven n Tables 3 and 4 respectvely. The algorthm force_bo n Table 3 computes force components nvolved n sngle BO term where the A 1 A 2 and A 3 terms are the pre-computed dervatves of BO wth respect to BO and respectvely. The algorthm force_ n Table 4 computes the force contrbutons arsng from the dervatve wth respect to whch s smply a wrapper functon of force_bo loopng over the neghbor lst of the th atom. We now defne sx algorthms 5 valence and 1 noncovalent for force calculatons each of whch contans energy functons sharng the same computatonal characterstcs n Sectons Combned wth a multple tme-step ntegrator ths groupng of the force components facltates effcent tme ntegraton whch wll be dscussed n Secton body energes Among the sx force-calculaton algorthms force_1body s responsble for the 1-body energy calculatons E lp E over and E under whch contan one explct atom see Eqs Here the valency term n the lone-par functon e the number of lone par electrons n lp the coordnaton effect lp and the correcton term lpcorr due to lone-par electrons n the over/under coordnaton energy functon E over and E under are defned n Eq. 9. The parameters n Eqs. 6 9 are lsted n Table A.2 n Appendx A. E lp = p lp2 lp 1 + exp 75 lp 6

5 K.-. Nomura et al. / Computer Physcs Communcatons Table 3 Force calculaton arsng from the dervatve of BO Algorthm force_bo {r }=set of atomc postons n n = lsts of neghbor atoms for the th and th atoms C BO = E/ BO {f }=set of updated atomc forces f f + C BO A 1 r // A 1 = BO / BO BO / r f f + C BO A 1 r for each atom r k n f f + C BO A 2 r k // A 2 = BO / / r k f k f k + C BO A 2 r k for each atom r k n f f + C BO A 3 r k // A 3 = BO / / r k f k f k + C BO A 3 r k E over = n p ovun1de σ BO [ lpcorr lpcorr + Val E under = p ovun5 1 expp ovun6 lpcorr 1 + exp p ovun2 lpcorr e = Vale + BO n e n lp = + exp 2 lp = n lp-opt n lp lpcorr = p lp1 Table 4 Force-calculaton arsng from the dervatve of Algorthm force_ {r }=set of atomc postons n = lst of neghbor atoms for the th atom C = E/ {f }=set of updated atomc forces for each atom r n call force_bor r C 2 + e 2 e expp ovun2 lpcorr lp ] p ovun7 exp[p ovun8 n lp BOπ + BOππ p ovun3 expp ovun4 n lp BOπ + BOππ. 7 8 ] 9a Due to the trply nested neghbor-lst loops n the summaton of the valency see Table 5 force_1body requres the traversal of a deep tree structure emanatng from each prmary atom n force computaton wth a relatvely large nteracton cutoff dstance. In Table 5 E 1body denotes one of E lp E over and E under terms Bondng 2-body energy E bond n Eq. 10 descrbes the covalent bond energy of atomc pars whch ncludes energy contrbutons from the σπ and double-π bond terms. The parameters n the 2-body nteractons are lsted n Table A.3 n Appendx A. E bond = D σ e BO σ exp[ p be1 1 BO σ pbe2 ] D π e BO π Dππ e BO ππ Algorthm force_2body n Table 6 computes E bond. Though Table 6 does not dstngush the σπ and double-π sub-terms for smplcty they are handled smply by nputtng dfferent dervatves n algorthm force_bo. InTable 6 ID denotes a sequental ndex of the th atom. Our parallel MD algorthm mantans global sequental ndces over all atoms across processors.. 9b 9c 9d 10

6 78 K.-. Nomura et al. / Computer Physcs Communcatons Table 5 1-body force calculaton Algorthm force_1body {r }=set of atomc postons {n}=set of neghbor lsts where n s the lst of neghbor atoms of the th atom {f }=set of atomc forces C 1 = E 1body / BO C 2 = E 1body / for each atom r compute C 1 and C 2 for each atom r n call force_bor r C 1 call force_r C 2 Table 6 2-body force calculaton Algorthm force_2body {r }=set of atomc postons {n}=set of neghbor lsts where n s the lst of neghbor atoms of th atom {f }=set of atomc forces C 3 = E bond / BO for each atom r for each atom r n f ID > ID then compute C 3 call force_bor r C body energes The 3-body nteracton functons E val E pen and E coa n Eqs descrbe the stablty of the valence angle Θ k and covalent bonds between and k atoms. In E val the functon f 7 n Eq. 11a guarantees that the energy term goes to zero as the two covalent bonds from the center atom dssocate. Eqs. 11c 11f compute the equlbrum valence angle takng nto account π double-π and lone par electron effects. E pen s the penalty energy functon n atomc trplets. E coa s the conugaton energy among trplet atoms for example an NO 2 group n ntramnes. The parameters n Eqs are lsted n Table A.4 n Appendx A. E val = f 7 BO f 7 BO k f 8 { p val1 p val1 exp [ 2 ]} p val2 Θ0 SBO Θ k 11 k f 7 BO = 1 exp p val3 BO p val4 11a 2 + expp val6 angle f 8 = p val5 p val5 1 11b 1 + expp val6 angle + exp p val7 angle SBO = BO π m + BO ππ m + [1 exp BO 8 ] angle m p val8 n lp 11c angle m n = Val angle + m n BO m m n 0 SBO 0 SBO SBO2 = p val9 0 < SBO SBO p val9 1 < SBO 2 2 SBO > 2 Θ 0 SBO = π Θ 00 { 1 exp [ pval10 2 SBO2 ]} 11d 11e 11f

7 K.-. Nomura et al. / Computer Physcs Communcatons Table 7 3-body force calculaton Algorthm force_3body {r }=set of atomc postons {n}=set of neghbor lsts where n s the lst of neghbor atoms of th atom {f }=set of atomc forces C 4 = E 3body / BO C 4k = E 3body / BO k C 5 = E 3body / C 5 = E 3body / C 5k = E 3body / k for each atom r for each atom r n for each atom r k n f ID > IDk then compute C 4 and C 5 call force_bor r C 4 call force_bor r k C 4k call force_r C 5 call force_r C 5 call force_r k C 5k compute force components from 3-body angle dervatves E pen = p pen1 f 9 exp [ p pen2 BO 2 2] exp [ p pen2 BO k 2 2] k 12 f 9 = 2 + exp p pen3 1 + exp p pen3 + expp pen4 12a E coa = 1 p coa1 [ p 1 + expp k coa2 val exp coa3 BO + 2 ] BO m m n exp [ p coa3 BO k + 2 ] BO km exp [ p coa4 BO 1.5 2] exp [ p coa4 BO k 1.5 2]. m nk Algorthm force_3body n Table 7 computes E val E pen and E coa energes whch have three explct atoms Eqs In force_3body and also n force_4body descrbed below the force arsng from angular dervatves needs to be computed separately from BO dervatves as s descrbed elsewhere [21]. InTable 7 E 3body denotes one of E val E pen and E coa body energes and parameters 4-body dhedral angle energy E tors and 4-body conugaton energy E con are descrbed by Eqs. 14 and 15. The angles Θ k and Θ kl are computed from atom trplets kand kl. The dhedral angle ω kl s the angle between the two planes defned by the two atom trplets. The parameters n the 4-body nteractons are lsted n Table A.5 n Appendx A. E tors = 1 f 10 BO BO k BO kl sn Θ k sn Θ kl 2 k l [ V cos ω kl + V 2 exp { p tor1 2 BO π k f 11 k 2} 1 cos 2ω kl + V cos 3ω kl ] f 10 BO BO k BO kl = [ 1 exp p tor2 BO ][ 1 exp p tor2 BO k ][ 1 exp p tor2 BO kl ] f 11 k = 1 + exp[ p tor3 angle 2 + exp[ p tor3 angle + angle k ] + angle k ]+exp[p tor4 angle + angle k ] E con = p cot1 f 12 BO BO k BO kl [ 1 + cos 2 ω kl 1 ] sn Θ k sn Θ kl k l f 12 BO BO k BO kl = exp [ p cot2 BO 1.5 2] exp [ p cot2 BO k 1.5 2] exp [ p cot2 BO kl 1.5 2]. 14a 14b 15 15a

8 80 K.-. Nomura et al. / Computer Physcs Communcatons Table 8 4-body force calculaton Algorthm force_4body {r }=set of atomc postons {n}=set of neghbor lsts where n s the lst of neghbor atoms of th atom {f }=set of atomc forces C 6 = E 4body / BO C 6k = E 4body / BO k C 6kl = E 4body / BO kl C 7 = E 4body / C 7k = E 4body / k for each atom r for each atom r k n f ID > IDk then for each atom r n f ID IDk then for each atom r l nk f ID IDk and ID IDl and ID IDl then compute C 6 and C 7 call force_bor r C 6 call force_bor r k C 6k call force_bor k r l C 6kl call force_r C 7 call force_r k C 7k compute force components from 4-body angle dervatves Algorthm force_4body n Table 8 computes E tors and E con energes that nvolve four explct atoms Eqs. 14 and 15. force_4body consumes large amount of computaton tme to handle atomc quadruplets at every MD step whch can be reduced by the multple tme-step scheme descrbed n Secton 2.4. InTable 8 E 4body denotes ether E con or E tors Hydrogen bondng energy and parameters The hydrogen bondng energy functon E hbond n Eq. 16 takes a 3-body nteracton form.e. as a functon of the and k bonds as well as the angle Θ k.thee hbond term thus conssts of three explct atoms where the th atom s hydrogen the th atom s chosen from atom s neghbor lst and the kth atom s selected wthn the noncovalent nteracton cutoff radus r cnb 10 Å from atom. The parameters n the hydrogen bondng nteractons are lsted n Table A.6 n Appendx A. E hbond = [ p hb1 1 expphb2 BO ] [ r o exp p hb hb3 + r ] k r k rhb o 2 k sn 8 Θ k 2 To handle the unque form of E hbond we defne a separate algorthm force_hbond n Table 9. Because of ts asymmetrc structure consstng of valence and noncovalent bonds t s necessary that atomc trplets k and k are dealt dfferently. Forces arsng from the dervatve wth respect to noncovalent atoms k are thus computed separately n Table Table 9 Hydrogen bondng force calculaton Algorthm force_hbond {r }=set of atomc postons {n}=set of neghbor lsts where n s the lst of neghbor atoms of th atom {f }=set of atomc forces C 8 = E hbond / BO for each atom r for each atom r n for each atom r k {r k r k r cnb } compute C 8 call force_bor r C 8 compute force components from 3-body angle dervatves compute force components from noncovalent term

9 K.-. Nomura et al. / Computer Physcs Communcatons Table 10 Noncovalent force calculaton Algorthm force_noncov {r }=set of atomc postons {f }=set of atomc forces C 9 = E noncov / r for each atom r for each atom r {r r r cnb } f ID > ID then compute C 9 compute energy and force components Noncovalent energes and parameters The noncovalent nteractons consst of E vdwaals Eq. 17 and E Coulomb Eq. 18 terms. The functons are screened by a taper functon wth a cutoff length r cnb n Eq. 17b where the coeffcents of the polynomal are Tap 7 = 20 Tap 6 = 70 Tap 5 = 84 Tap 4 = 35 Tap 3 = 0 Tap 2 = 0 Tap 1 = 0 and Tap 0 = 1. Noncovalent energy functons n ReaxFF are all-atom nteractons whch do not requre any excluson rule [22] as n some other force felds. E vdwaals = { D Tapr exp [α 1 f 13r r vdw [ f 13 r = r p vdw1 1 pvdw1 ] 1/pvdW1 + γ w 7 α r Tapr = Tap α r cnb α=0 E Coulomb {r } {q } = χ q + 1 q Hr q 2 Tapr Hr = J δ + [r 3 + 1/γ 3 ] 1/3 1 δ { 1 = δ = 0 ] 2exp [ 1 2 α 1 f ]} 13r r vdw where χ s the electronegatvty J s the self Coulomb repulson coeffcent and γ s a parameter for the smeared Coulombc functon. Algorthm force_noncov n Table 10 computes van der Waals energy E vdwaals and Coulomb energy E Coulomb whch are functons of nteratomc dstance wth the cutoff dstance r cnb Eqs. 17 and 18. We use potental tables to compute the noncovalent nteractons n order to reduce the number of floatng-pont operatons. In Table 10 E noncov denotes ether E vdwaals or E Coulomb. Snce the noncovalent nteractons are functons of nteratomc dstance precomputng the functon values along wth a proper nterpolaton scheme sgnfcantly reduce the computaton tme Electronegatvty-equalzaton scheme In ReaxFF atomc charges q are varables that change dynamcally n tme. Whenever after atom coordnates are updated the EEM subroutne updates the charge dstrbuton {q } by mnmzng the Coulomb energy E Coulomb under the charge-neutralty constrant q = 0. Wth the Lagrange-multpler method the constraned energy mnmzaton s equvalent to solvng the electronegatvty equalzaton problem [ ] g E Coulomb = μ 19 q where μ s the electronegatvty. We solve the constraned mnmzaton problem usng the conugate-gradent method [2627] n Table 11 n whch the gradent vector g s orthogonalzed to the unform shft vector e = 1 = 1...N. Because of the electronegatvty equalzaton condton Eq. 19 the force contrbuton from E Coulomb {r } {q } E Coulomb r N k=1 E Coulomb q k q k = E Coulomb μ r r r N k=1 q k = E Coulomb r 17 17a 17b 18 18a 20

10 82 K.-. Nomura et al. / Computer Physcs Communcatons Table 11 Electronegatvty-equalzaton computaton Algorthm EEM {r }=set of atomc postons {q }=set of atomc charges {g }={ E Coulomb / q }=gradent vector {h }=conugate-gradent vector compute the ntal gradent vector {g 0} n 0 do f n = 1 then h 0 = g0 = 1toN else h n gn 1 q n q n 1 g n gn 1 g n 1 g n 2 h n 1 = 1toN + gn 1 g n 1 h n H h n h n = 1toN + gn 1 g n 1 h n H h n H h n = 1toN + gn 1 g n 2 n n + 1 untl the energy s converged does not contan chan-rule terms assocated wth varables q k and thus the noncovalent force calculaton descrbed n Secton remans vald the last equalty n Eq. 20 results from the charge-neutralty condton k q k = 0. It s also noteworthy that the Hessan Hr n Eq. 18 s a constant untl atom coordnates are updated. Thus reusng pre-computed Hr durng the EEM loop n Table 11 sgnfcantly reduces the total computaton tme Tme ntegrator Our parallel ReaxFF algorthm employs the rrespa tme ntegrator [28 30] whch s a tme-reversble algorthm based on symmetrc Trotter decomposton of the tme-propagaton operator. It allows decomposng the energy functons nto groups wth dstnct characterstc tme scales. Accordngly the forces f are decomposed nto stff force components f stff whch need to be updated wth hgher frequency and soft and slowly varyng force components f soft : f = f soft + f stff. Table 12 shows the rrespa algorthm comprsng of a doubly nested tme ntegraton loop. The compute ntensve force calculaton.e. f soft s performed only once per outer loop wth a large tme step t typcally t 1 fs. The atomc veloctes are updated accordng to f soft at the begnnng and the end of the outer loop. The nner loop updates the atomc postons and veloctes wth a fner tme resoluton δt = t/n mts usng the stff force components f stff typcally N mts 5. It s often the case that the soft force components f soft are computatonally ntensve [28]. Thus proper force decomposton sgnfcantly reduces the total computaton tme. For example n dense covalently bonded materals the 4-body force calculaton s most tme consumng. In test calculatons on a damond the 4-body force calculaton accounts for nearly half of the total computaton tme. We have reduced the total smulaton tme by 30% wth the number of multple tme step N mts = 5 see Table 12 by separatng out the 4-body force terms nto f soft.e. by performng the tme-consumng 4-body force calculaton only once n 5 MD steps. 3. Parallelzaton To desgn a scalable ReaxFF-MD smulaton algorthm on massvely parallel computers we employ spatal decomposton [31] combned wth lnked-lst cell and Verlet neghbor-lst methods [21]. In spatal decomposton the smulaton space wth lengths L x L y and L z n the x y and z drectons respectvely s decomposed nto P = P x P y P z rectangular sub-domans Ω 0 Ω P 1 of equal volume whch are then mapped to P processors n a parallel computer. Specfcally atom at poston r = r x r y r z s assgned to processor p [0P 1] where 21 { p = px P y P z + p y P z + p z p α = r α P α /L α α = xyz. 22

11 K.-. Nomura et al. / Computer Physcs Communcatons Table 12 Multple tme-step algorthm Algorthm multple_tme_step {r t}=set of atomc postons at tme t {v t}=set of atomc veloctes at tme t N step = number of total MD steps N mts = number of multple tme steps t = N mts δt = tme dscretzaton unt {r t + N step δt}=set of atomc coordnates at t + N step δt {v t + N step δt}=set of atomc veloctes at t + N step δt {f }=set of atomc forces compute soft force components {f soft } for outer =1toN step /N mts v v + t/2m f soft // m s the mass of the th atom compute stff force components {f stff } for nner = 1toN mts v v + δt/2m f stff r r + δtv compute stff force components {f stff } v v + δt/2m f stff compute soft force components {f soft } v v + t/2m f soft Each processor stores the coordnates r veloctes v and types σ e.g. hydrogen carbon ntrogen or oxygen of all atoms that are assgned to t and s also n charge of updatng the coordnates and veloctes by numercally ntegratng the Newton s equaton of moton. In the lnked-lst cell method each doman s further decomposed nto smaller rectangular cells. The atom nformaton r v σ n each cell s organzed usng a lnked lst whch reduces the computatonal cost of neghbor-lst calculaton to ON/P where N s the total number of atoms. We defne the dmenson of a lnked-lst cell L c to be no less than the sngle covalent-bond cutoff r cb.ther cb s determned as the nteratomc dstance at whch any type of BO becomes neglgble. We use a crteron BO r cb<10 4. At the doman boundares t s necessary to copy suffcent atom nformaton from neghbor domans to compute forces. In 3D smulaton space each doman has 26 adacent domans from whch atom nformaton needs to be cached nto buffer layers that surround the doman shown as a shaded area n Fg. 2. The requste thckness of the buffer layer L b s determned by the greatest cutoff dstance among all nteracton functons. Specfcally L b = maxr cnb l max r cb where l max s the deepest level of covalently bonded trees wth the prmary atom as a root see Fg. 2B. The cachng operaton s mplemented as the exchange of Fg. 2. A Schematc of the spatal decomposton conssts of sub-domans Ω 0 Ω 3. On the sub-doman Ω 0 atom nformaton from neghbor domans Ω 1 Ω 2 and Ω 3 are stored n the buffer layer wth the thckness L b shown as shaded area surroundng Ω 0. The prmary atoms n each cluster gray spheres must be n Ω 0 whle other atoms whte spheres may be coped from neghbor domans. B An example of 4-body atom confguraton on a doman boundary. The explct atoms are shown wth sold-lne spheres and bonds k andl and so are mplct atoms wth dotted-lne spheres and bonds m 1 m 5. Here th atom s the prmary of the cluster. In ths example the deepest level from the prmary atom s l max = 3.

12 84 K.-. Nomura et al. / Computer Physcs Communcatons messages between processors where the messages contan the nformaton of atoms that le wthn L b from the doman boundary. After cachng processor p has the nformaton of a set of atoms S = S resdent S cached ={ r Ω p } { r / Ω p r Ω p <L b } where Ω p s the outer-boundary surface of doman Ω p and r Ω p s the shortest dstance between atom and Ω p.ths allows each processor to compute the forces on ts resdent atoms S resdent wthout tme-consumng nterprocessor communcatons. Our parallel ReaxFF algorthm attempts to mnmze the total computaton tme by avodng duplcated energy/force calculatons among processors. For ths purpose care needs to be taken for the cached atoms stored n the buffer layer. Here we llustrate the energy/force calculatons n the parallel MD framework. Fg. 2B shows a sample case of 4-body calculaton that nvolves four explct atoms denoted as k and l and fve mplct atoms m 1 m 5. At the begnnng of the energy/force calculaton on doman Ω 0 atomslm 1 and m 2 are cached from doman Ω 1 to Ω 0 va communcaton network. We requre that the prmary atom s a resdent atom of doman Ω 0. The computed potental energy of the atomc cluster s stored n the doman of the prmary atom. At the end of energy/force calculaton the reacton forces on the cached atoms are transferred back to the orgnal atoms n doman Ω 1. Subsequently atom coordnates are updated accordng to the tme ntegraton algorthm n Table 12. Resdent atoms mgrate to new domans f they have moved out of the current doman. All nter-doman communcatons follow a 6-way communcaton method [32]. The amount of communcaton s proportonal to the number of atoms on doman surface whch s ON/P 2/3. The parallel ReaxFF program s wrtten n the Fortran 90 language wth the message passng nterface MPI lbrary [33] for communcatons. 4. Performance test The scalablty of parallel ReaxFF program has been tested on varous hgh-end supercomputers ncludng the processor IBM BlueGene/L at the Lawrence Lvermore Natonal Laboratory LLNL the processor SGI Altx 3000 at the NASA Ames Research Center and an AMD dual-core Opteron cluster at the Hgh Performance Computng Center HPCC of the Unversty of Southern Calforna USC. The parallel ReaxFF algorthm maxmally utlzes data localty regardless of actual physcal archtecture and accordngly no performance degradaton has been observed on any of these archtectures. Descrpton of each platform and computatonal resources used for each benchmark are gven below. BlueGene/L The BlueGene/L system at the LLNL comprses of computatonal node CN chps each of whch has two PowerPC 400 processors processors n total wth 700 MHz clock speed. On sngle CN the two processors share 512 MB memory. Each processor has a 32 KB nstructon/data cache a 2 MB L2 cache and a 4 MB L3 cache. The theoretcal peak performance s 2.8 Gflops per processor. Two types of nterconnecton 3-D torus and tree topologes are desgned for dstnct purposes. The 3-D torus network s used mostly for common e.g. pont-to-pont communcatons whle the tree network s optmzed for collectve communcatons. The nterconnecton bandwdths are 175 MB/s and 350 MB/s per lnk respectvely. Altx 3000 The SGI Altx 3000 system named Columba at NASA-Ames conssts of 20 of SGI Altx model 3700 boxes each consstng of 512 Intel 1.5 GHz Itanum2 processors. Each processor has 128 floatng-pont regsters a 32 KB L1 cache a 256 KB L2 cache and a 6 MB L3 cache and ts theoretcal peak performance s 6 Gflops. The SGI NUMAlnk4 nterconnecton provdes 1TB memory globally shared among cluster nodes wthn an Altx box. We have used up to four Altx boxes up to 1920 processors for our benchmark. Opteron USC-HPCC operates 1824-node Lnux cluster wth 15.8 Tflops Lnpack performance. We have used up to 512 nodes of dual-cpu dual-core AMD 2 GHz Opteron processors 2048 processors n total. Each node has 4 GB memory and Myrnet nterconnecton provdes 256 MB/s bandwdth. Sngle Opteron core has 64 KB nstructon/data caches and 1 MB L2 cache. Fg. 3 shows the parallel effcency of the parallel ReaxFF algorthm on the three archtectures. Here the sogranular parallel effcency s defned as the total executon tme dvded by that on P = 1 where the gran sze.e. the number of atoms per processor s kept constant N/P = on the BlueGene/L on Columba and on the Opteron cluster. In all performance tests we smulate energetc molecular crystal RDX 135-trntro-135-trazne C 3 N 6 O 6 H 6 [14]. Overall the parallel ReaxFF algorthm acheves nearly perfect 1 effcency. On BlueGene/L processors the parallel effcency 23

13 K.-. Nomura et al. / Computer Physcs Communcatons Fg. 3. Isogranular parallel effcency of the parallel ReaxFF algorthm as a functon of the number of processors on A IBM Blue Gene/L wth the number of atoms per processor N/P = B SGI Altx 3000 wth N/P = and C AMD dual-core Opteron cluster wth N/P = Fg. 4. Executon tme of the parallel ReaxFF program as a functon of the number of atoms on processor IBM Blue Gene/L 1920-processor SGI Altx 3000 and 2048-core AMD Opteron. s Only excepton s the sudden drop n the parallel effcency on the Opteron cluster whch s lkely due to the non-dedcated testng envronment n whch other users processes share the same network. Fg. 4 shows the executon tme as a functon of the number of atoms whle the total number of processors s fxed. We use IBM BlueGene/L processors 1920 Itenum2 processors of SGI Altx 3000 and 2048 AMD Opteron cores. The largest number of atoms s on the BlueGene/L. The parallel ReaxFF algorthm exhbts perfect lnear scalablty as a functon of the problem sze. 5. Summary For chemcally-reactve molecular dynamcs smulatons we have desgned a scalable parallel algorthm that maxmally exposes data localty. The parallel ReaxFF algorthm ncorporates spatal decomposton and multple tme-steppng for effcent valence/noncovalent force calculatons. Envronment-dependent varable charges are ncorporated wth an electronegatvty equalzaton method. Benchmark tests have exhbted hgh scalablty parallel effcency s on BlueGene/L processors as well as good performance portablty. The parallel ReaxFF algorthm enables unprecedented scales of chemcally reactve smulatons whch pave a way to study longstandng scentfc problems such as mechancally nduced chemcal reactons. Acknowledgements Ths work was partally supported by ARO MURI DOE DTRA and NSF. Performance tests were performed at the Unversty of Southern Calforna usng the 5384-processor Lnux cluster at the Research Computng Faclty and the 2048-processor Lnux

14 86 K.-. Nomura et al. / Computer Physcs Communcatons cluster at the Collaboratory for Advanced Computng and Smulatons. We thank Dr. A.C.T. van Dun and Prof. W.A. Goddard III for valuable dscussons. Appendx A. ReaxFF parameters Appendx A contans Tables A.1 A.7. Table A.1 The parameters n the bond-order functons p bo1 p bo6 ro σ rπ o rππ o Val Val boc p boc1 p boc2 p boc3 p boc5 bond parameters bond radus parameters valences of atom overcoordnaton parameters 1 3 bond order correctons Table A.2 The parameters n the 1-body energy Val e n lp-opt p lp1 p lp2 p ovun1 p ovun4 p ovun5 p ovun6 p ovun8 valency of atom the number of lone par electrons at normal condton lone par parameter lone par energy parameter overcoordnaton parameters undercoordnaton energy parameter undercoordnaton parameters Table A.3 The parameters n the E bond De σ Dπ e Dππ e p be1 p be2 σπ and double π bond energy parameters bondng energy parameters Table A.4 The parameters n the 3-body energy p val1 p val2 p val7 p val8 p val9 p val10 Θ 00 Val angle p pen1 p pen2 p pen3 p pen4 p coa1 p coa2 p coa4 valence angle energy parameter valence angle parameters valency/lone par parameter valence angle parameters equlbrum valence angle atom valency penalty energy parameter double bond/angle parameter double bond/angle parameters for overcoordnaton 3-body conugaton energy parameter valency angle conugaton parameters Table A.5 The parameters n the 4-body energy V 1 V 3 p tor1 p tor2 p tor3 p tor4 p cot1 p cot2 torson energy parameters torson parameter torson/bo parameter torson overcoordnaton parameters 4-body conugaton energy parameter 4-body conugaton parameter Table A.6 The parameters n the hydrogen bondng energy p hb1 p hb2 p hb3 rhb o hydrogen bond energy parameter hydrogen bond parameters hydrogen bond radus

15 K.-. Nomura et al. / Computer Physcs Communcatons Table A.7 The parameters n the noncovalent energy D α r vdw p vdw1 γ w γ van der Waals energy parameter van der Waals parameter van der Waals parameter van der Waals sheldng van der Waals parameter Coulomb parameter References [1] D.G. Truhlar V. McKoy Computng n Scence & Engneerng [2] J.J. Glman Scence [3] J.J. Glman Materals Scence and Technology [4] S. Ogata E. Ldorks F. Shmoo A. Nakano P. Vashshta R.K. Kala Computer Physcs Communcatons [5] S. Dapprch I. Komárom K.S. Byun K. Morokuma M.J. Frsch Journal of Molecular Structures Theochem [6] M.J. Feld P.A. Bash M. Karplus Journal of Computatonal Chemstry [7] U.C. Sngh P.A. Kollman Journal of Computatonal Chemstry [8] A. Warshel M. Levtt Journal of Molecular Bology [9] E.E. Santso K.E. Gubbns Molecular Smulaton [10] J. Tersoff Physcal Revew B [11] D.W. Brenner Physca Status Sold b [12] A.C.T. van Dun S. Dasgupta F. Lorant W.A. Goddard Journal of Physcal Chemstry A [13] A.C.T. van Dun A. Strachan S. Stewman Q.S. Zhang X. Xu W.A. Goddard Journal of Physcal Chemstry A [14] A. Strachan A.C.T. van Dun D. Chakraborty S. Dasgupta W.A. Goddard Physcal Revew Letters [15] Q. Zhang T. Cagn A.C.T. van Dun W.A. Goddard Y. Q L.G. Hector Physcal Revew B [16] K.D. Nelson A.C.T. van Dun J. Oxgaard W.Q. Deng W.A. Goddard Journal of Physcal Chemstry A [17] K. Chenoweth S. Cheung A.C.T. van Dun W.A. Goddard E.M. Kober Journal of the Amercan Chemcal Socety [18] K. Nomura R.K. Kala A. Nakano P. Vashshta A.C.T. van Dun W.A. Goddard Physcal Revew Letters n press. [19] P. Vashshta R.K. Kala A. Nakano B.E. Homan K.L. McNesby Journal of Propulson and Power [20] W.J. Morter S.K. Ghosh S. Shankar Journal of the Amercan Chemcal Socety [21] M.P. Allen D.J. Tldesley Computer Smulaton of Lquds Oxford Unversty Press [22] W.D. Cornell P. Ceplak C.I. Bayly I.R. Gould K.M. Merz D.M. Ferguson D.C. Spellmeyer T. Fox J.W. Caldwell P.A. Kollman Journal of the Amercan Chemcal Socety [23] A.K. Rappe W.A. Goddard Journal of Physcal Chemstry [24] F.H. Stretz J.W. Mntmre Physcal Revew B [25] T.J. Campbell R.K. Kala A. Nakano P. Vashshta S. Ogata S. Rodgers Physcal Revew Letters [26] A. Nakano Computer Physcs Communcatons [27] B. Bayraktar T. Bernas J.P. Robnson B. Rawa Pattern Recognton [28] G.J. Martyna M.E. Tuckerman D.J. Tobas M.L. Klen Molecular Physcs [29] M.E. Tuckerman B.J. Berne G.J. Martyna Journal of Chemcal Physcs [30] M.E. Tuckerman D.A. Yarne S.O. Samuelson A.L. Hughes G.J. Martyna Computer Physcs Communcatons [31] A. Nakano R.K. Kala P. Vashshta T.J. Campbell S. Ogata F. Shmoo S. San Scentfc Programmng [32] A. Nakano R.K. Kala K. Nomura A. Sharma P. Vashshta F. Shmoo A.C.T. van Dun W.A. Goddard R. Bswas D. Srvastava Computatonal Materals Scence [33] W. Gropp E. Lusk A. Skellum Usng MPI second ed. The MIT Press 1999.

ReaxFF Potential Functions

ReaxFF Potential Functions ReaxFF Potental Functons Supportng nformaton for the manuscrpt A ReaxFF Reactve Force Feld for Molecular Dynamcs Smulatons of Hydrocarbon Oxdaton by Kmberly Chenoweth Adr CT van Dun and Wllam A Goddard

More information

Supporting information

Supporting information Supportng nformaton for the manuscrpt Intaton Mechansms and Knetcs for Pyrolyss and Combuston of JP Hydrocarbon Jet Fuel by Kmberly Chenoweth Adr CT van Dun Sddharth Dasgupta and Wllam A Goddard III ReaxFF

More information

ReaxFF potential functions

ReaxFF potential functions ReaxFF otental functons Suortng nformaton for the manuscrt Nelson, K.D., van Dun, A.C.T., Oxgaard, J.., Deng, W. and Goddard III, W.A. Develoment of the ReaxFF reactve force feld for descrbng transton

More information

Kernel Methods and SVMs Extension

Kernel Methods and SVMs Extension Kernel Methods and SVMs Extenson The purpose of ths document s to revew materal covered n Machne Learnng 1 Supervsed Learnng regardng support vector machnes (SVMs). Ths document also provdes a general

More information

Hongyi Miao, College of Science, Nanjing Forestry University, Nanjing ,China. (Received 20 June 2013, accepted 11 March 2014) I)ϕ (k)

Hongyi Miao, College of Science, Nanjing Forestry University, Nanjing ,China. (Received 20 June 2013, accepted 11 March 2014) I)ϕ (k) ISSN 1749-3889 (prnt), 1749-3897 (onlne) Internatonal Journal of Nonlnear Scence Vol.17(2014) No.2,pp.188-192 Modfed Block Jacob-Davdson Method for Solvng Large Sparse Egenproblems Hongy Mao, College of

More information

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module 3 LOSSY IMAGE COMPRESSION SYSTEMS Verson ECE IIT, Kharagpur Lesson 6 Theory of Quantzaton Verson ECE IIT, Kharagpur Instructonal Objectves At the end of ths lesson, the students should be able to:

More information

Army Ants Tunneling for Classical Simulations

Army Ants Tunneling for Classical Simulations Electronc Supplementary Materal (ESI) for Chemcal Scence. Ths journal s The Royal Socety of Chemstry 2014 electronc supplementary nformaton (ESI) for Chemcal Scence Army Ants Tunnelng for Classcal Smulatons

More information

Supplementary Information:

Supplementary Information: Supplementary Informaton: Vsualzaton-based analyss of structural and dynamcal propertes of smulated hydrous slcate melt Bjaya B. Kark 1,2, Dpesh Bhattara 1, Manak Mookherjee 3 and Lars Stxrude 4 1 Department

More information

Note 10. Modeling and Simulation of Dynamic Systems

Note 10. Modeling and Simulation of Dynamic Systems Lecture Notes of ME 475: Introducton to Mechatroncs Note 0 Modelng and Smulaton of Dynamc Systems Department of Mechancal Engneerng, Unversty Of Saskatchewan, 57 Campus Drve, Saskatoon, SK S7N 5A9, Canada

More information

Supporting Information

Supporting Information Supportng Informaton The neural network f n Eq. 1 s gven by: f x l = ReLU W atom x l + b atom, 2 where ReLU s the element-wse rectfed lnear unt, 21.e., ReLUx = max0, x, W atom R d d s the weght matrx to

More information

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE Analytcal soluton s usually not possble when exctaton vares arbtrarly wth tme or f the system s nonlnear. Such problems can be solved by numercal tmesteppng

More information

Resource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud

Resource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud Resource Allocaton wth a Budget Constrant for Computng Independent Tasks n the Cloud Wemng Sh and Bo Hong School of Electrcal and Computer Engneerng Georga Insttute of Technology, USA 2nd IEEE Internatonal

More information

MOLECULAR DYNAMICS ,..., What is it? 2 = i i

MOLECULAR DYNAMICS ,..., What is it? 2 = i i MOLECULAR DYNAMICS What s t? d d x t 2 m 2 = F ( x 1,..., x N ) =1,,N r ( x1 ( t),..., x ( t)) = v = ( x& 1 ( t ),..., x& ( t )) N N What are some uses of molecular smulatons and modelng? Conformatonal

More information

( ) 1/ 2. ( P SO2 )( P O2 ) 1/ 2.

( ) 1/ 2. ( P SO2 )( P O2 ) 1/ 2. Chemstry 360 Dr. Jean M. Standard Problem Set 9 Solutons. The followng chemcal reacton converts sulfur doxde to sulfur troxde. SO ( g) + O ( g) SO 3 ( l). (a.) Wrte the expresson for K eq for ths reacton.

More information

4DVAR, according to the name, is a four-dimensional variational method.

4DVAR, according to the name, is a four-dimensional variational method. 4D-Varatonal Data Assmlaton (4D-Var) 4DVAR, accordng to the name, s a four-dmensonal varatonal method. 4D-Var s actually a drect generalzaton of 3D-Var to handle observatons that are dstrbuted n tme. The

More information

COMPOSITE BEAM WITH WEAK SHEAR CONNECTION SUBJECTED TO THERMAL LOAD

COMPOSITE BEAM WITH WEAK SHEAR CONNECTION SUBJECTED TO THERMAL LOAD COMPOSITE BEAM WITH WEAK SHEAR CONNECTION SUBJECTED TO THERMAL LOAD Ákos Jósef Lengyel, István Ecsed Assstant Lecturer, Professor of Mechancs, Insttute of Appled Mechancs, Unversty of Mskolc, Mskolc-Egyetemváros,

More information

(Online First)A Lattice Boltzmann Scheme for Diffusion Equation in Spherical Coordinate

(Online First)A Lattice Boltzmann Scheme for Diffusion Equation in Spherical Coordinate Internatonal Journal of Mathematcs and Systems Scence (018) Volume 1 do:10.494/jmss.v1.815 (Onlne Frst)A Lattce Boltzmann Scheme for Dffuson Equaton n Sphercal Coordnate Debabrata Datta 1 *, T K Pal 1

More information

PHYS 705: Classical Mechanics. Calculus of Variations II

PHYS 705: Classical Mechanics. Calculus of Variations II 1 PHYS 705: Classcal Mechancs Calculus of Varatons II 2 Calculus of Varatons: Generalzaton (no constrant yet) Suppose now that F depends on several dependent varables : We need to fnd such that has a statonary

More information

DFT with Planewaves pseudopotential accuracy (LDA, PBE) Fast time to solution 1 step in minutes (not hours!!!) to be useful for MD

DFT with Planewaves pseudopotential accuracy (LDA, PBE) Fast time to solution 1 step in minutes (not hours!!!) to be useful for MD LLNL-PRES-673679 Ths work was performed under the auspces of the U.S. Department of Energy by under contract DE-AC52-07NA27344. Lawrence Lvermore Natonal Securty, LLC Sequoa, IBM BGQ, 1,572,864 cores O(N)

More information

An efficient algorithm for multivariate Maclaurin Newton transformation

An efficient algorithm for multivariate Maclaurin Newton transformation Annales UMCS Informatca AI VIII, 2 2008) 5 14 DOI: 10.2478/v10065-008-0020-6 An effcent algorthm for multvarate Maclaurn Newton transformaton Joanna Kapusta Insttute of Mathematcs and Computer Scence,

More information

Appendix B. The Finite Difference Scheme

Appendix B. The Finite Difference Scheme 140 APPENDIXES Appendx B. The Fnte Dfference Scheme In ths appendx we present numercal technques whch are used to approxmate solutons of system 3.1 3.3. A comprehensve treatment of theoretcal and mplementaton

More information

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS Avalable onlne at http://sck.org J. Math. Comput. Sc. 3 (3), No., 6-3 ISSN: 97-537 COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

More information

Tensor Smooth Length for SPH Modelling of High Speed Impact

Tensor Smooth Length for SPH Modelling of High Speed Impact Tensor Smooth Length for SPH Modellng of Hgh Speed Impact Roman Cherepanov and Alexander Gerasmov Insttute of Appled mathematcs and mechancs, Tomsk State Unversty 634050, Lenna av. 36, Tomsk, Russa RCherepanov82@gmal.com,Ger@npmm.tsu.ru

More information

Electrical double layer: revisit based on boundary conditions

Electrical double layer: revisit based on boundary conditions Electrcal double layer: revst based on boundary condtons Jong U. Km Department of Electrcal and Computer Engneerng, Texas A&M Unversty College Staton, TX 77843-318, USA Abstract The electrcal double layer

More information

One-sided finite-difference approximations suitable for use with Richardson extrapolation

One-sided finite-difference approximations suitable for use with Richardson extrapolation Journal of Computatonal Physcs 219 (2006) 13 20 Short note One-sded fnte-dfference approxmatons sutable for use wth Rchardson extrapolaton Kumar Rahul, S.N. Bhattacharyya * Department of Mechancal Engneerng,

More information

Physics 181. Particle Systems

Physics 181. Particle Systems Physcs 181 Partcle Systems Overvew In these notes we dscuss the varables approprate to the descrpton of systems of partcles, ther defntons, ther relatons, and ther conservatons laws. We consder a system

More information

χ x B E (c) Figure 2.1.1: (a) a material particle in a body, (b) a place in space, (c) a configuration of the body

χ x B E (c) Figure 2.1.1: (a) a material particle in a body, (b) a place in space, (c) a configuration of the body Secton.. Moton.. The Materal Body and Moton hyscal materals n the real world are modeled usng an abstract mathematcal entty called a body. Ths body conssts of an nfnte number of materal partcles. Shown

More information

Thermodynamics General

Thermodynamics General Thermodynamcs General Lecture 1 Lecture 1 s devoted to establshng buldng blocks for dscussng thermodynamcs. In addton, the equaton of state wll be establshed. I. Buldng blocks for thermodynamcs A. Dmensons,

More information

Computing Correlated Equilibria in Multi-Player Games

Computing Correlated Equilibria in Multi-Player Games Computng Correlated Equlbra n Mult-Player Games Chrstos H. Papadmtrou Presented by Zhanxang Huang December 7th, 2005 1 The Author Dr. Chrstos H. Papadmtrou CS professor at UC Berkley (taught at Harvard,

More information

Indeterminate pin-jointed frames (trusses)

Indeterminate pin-jointed frames (trusses) Indetermnate pn-jonted frames (trusses) Calculaton of member forces usng force method I. Statcal determnacy. The degree of freedom of any truss can be derved as: w= k d a =, where k s the number of all

More information

International Journal of Pure and Applied Sciences and Technology

International Journal of Pure and Applied Sciences and Technology Int. J. Pure Appl. Sc. Technol., 4() (03), pp. 5-30 Internatonal Journal of Pure and Appled Scences and Technology ISSN 9-607 Avalable onlne at www.jopaasat.n Research Paper Schrödnger State Space Matrx

More information

A Robust Method for Calculating the Correlation Coefficient

A Robust Method for Calculating the Correlation Coefficient A Robust Method for Calculatng the Correlaton Coeffcent E.B. Nven and C. V. Deutsch Relatonshps between prmary and secondary data are frequently quantfed usng the correlaton coeffcent; however, the tradtonal

More information

Some Comments on Accelerating Convergence of Iterative Sequences Using Direct Inversion of the Iterative Subspace (DIIS)

Some Comments on Accelerating Convergence of Iterative Sequences Using Direct Inversion of the Iterative Subspace (DIIS) Some Comments on Acceleratng Convergence of Iteratve Sequences Usng Drect Inverson of the Iteratve Subspace (DIIS) C. Davd Sherrll School of Chemstry and Bochemstry Georga Insttute of Technology May 1998

More information

Modeling of Dynamic Systems

Modeling of Dynamic Systems Modelng of Dynamc Systems Ref: Control System Engneerng Norman Nse : Chapters & 3 Chapter objectves : Revew the Laplace transform Learn how to fnd a mathematcal model, called a transfer functon Learn how

More information

Supplementary Notes for Chapter 9 Mixture Thermodynamics

Supplementary Notes for Chapter 9 Mixture Thermodynamics Supplementary Notes for Chapter 9 Mxture Thermodynamcs Key ponts Nne major topcs of Chapter 9 are revewed below: 1. Notaton and operatonal equatons for mxtures 2. PVTN EOSs for mxtures 3. General effects

More information

Lecture 12: Discrete Laplacian

Lecture 12: Discrete Laplacian Lecture 12: Dscrete Laplacan Scrbe: Tanye Lu Our goal s to come up wth a dscrete verson of Laplacan operator for trangulated surfaces, so that we can use t n practce to solve related problems We are mostly

More information

THE EFFECT OF TORSIONAL RIGIDITY BETWEEN ELEMENTS ON FREE VIBRATIONS OF A TELESCOPIC HYDRAULIC CYLINDER SUBJECTED TO EULER S LOAD

THE EFFECT OF TORSIONAL RIGIDITY BETWEEN ELEMENTS ON FREE VIBRATIONS OF A TELESCOPIC HYDRAULIC CYLINDER SUBJECTED TO EULER S LOAD Journal of Appled Mathematcs and Computatonal Mechancs 7, 6(3), 7- www.amcm.pcz.pl p-issn 99-9965 DOI:.75/jamcm.7.3. e-issn 353-588 THE EFFECT OF TORSIONAL RIGIDITY BETWEEN ELEMENTS ON FREE VIBRATIONS

More information

Least squares cubic splines without B-splines S.K. Lucas

Least squares cubic splines without B-splines S.K. Lucas Least squares cubc splnes wthout B-splnes S.K. Lucas School of Mathematcs and Statstcs, Unversty of South Australa, Mawson Lakes SA 595 e-mal: stephen.lucas@unsa.edu.au Submtted to the Gazette of the Australan

More information

Which Separator? Spring 1

Which Separator? Spring 1 Whch Separator? 6.034 - Sprng 1 Whch Separator? Mamze the margn to closest ponts 6.034 - Sprng Whch Separator? Mamze the margn to closest ponts 6.034 - Sprng 3 Margn of a pont " # y (w $ + b) proportonal

More information

Physics 5153 Classical Mechanics. D Alembert s Principle and The Lagrangian-1

Physics 5153 Classical Mechanics. D Alembert s Principle and The Lagrangian-1 P. Guterrez Physcs 5153 Classcal Mechancs D Alembert s Prncple and The Lagrangan 1 Introducton The prncple of vrtual work provdes a method of solvng problems of statc equlbrum wthout havng to consder the

More information

MMA and GCMMA two methods for nonlinear optimization

MMA and GCMMA two methods for nonlinear optimization MMA and GCMMA two methods for nonlnear optmzaton Krster Svanberg Optmzaton and Systems Theory, KTH, Stockholm, Sweden. krlle@math.kth.se Ths note descrbes the algorthms used n the author s 2007 mplementatons

More information

Canonical transformations

Canonical transformations Canoncal transformatons November 23, 2014 Recall that we have defned a symplectc transformaton to be any lnear transformaton M A B leavng the symplectc form nvarant, Ω AB M A CM B DΩ CD Coordnate transformatons,

More information

Irregular vibrations in multi-mass discrete-continuous systems torsionally deformed

Irregular vibrations in multi-mass discrete-continuous systems torsionally deformed (2) 4 48 Irregular vbratons n mult-mass dscrete-contnuous systems torsonally deformed Abstract In the paper rregular vbratons of dscrete-contnuous systems consstng of an arbtrary number rgd bodes connected

More information

12. The Hamilton-Jacobi Equation Michael Fowler

12. The Hamilton-Jacobi Equation Michael Fowler 1. The Hamlton-Jacob Equaton Mchael Fowler Back to Confguraton Space We ve establshed that the acton, regarded as a functon of ts coordnate endponts and tme, satsfes ( ) ( ) S q, t / t+ H qpt,, = 0, and

More information

Why BP Works STAT 232B

Why BP Works STAT 232B Why BP Works STAT 232B Free Energes Helmholz & Gbbs Free Energes 1 Dstance between Probablstc Models - K-L dvergence b{ KL b{ p{ = b{ ln { } p{ Here, p{ s the eact ont prob. b{ s the appromaton, called

More information

Inductance Calculation for Conductors of Arbitrary Shape

Inductance Calculation for Conductors of Arbitrary Shape CRYO/02/028 Aprl 5, 2002 Inductance Calculaton for Conductors of Arbtrary Shape L. Bottura Dstrbuton: Internal Summary In ths note we descrbe a method for the numercal calculaton of nductances among conductors

More information

Structure and Drive Paul A. Jensen Copyright July 20, 2003

Structure and Drive Paul A. Jensen Copyright July 20, 2003 Structure and Drve Paul A. Jensen Copyrght July 20, 2003 A system s made up of several operatons wth flow passng between them. The structure of the system descrbes the flow paths from nputs to outputs.

More information

829. An adaptive method for inertia force identification in cantilever under moving mass

829. An adaptive method for inertia force identification in cantilever under moving mass 89. An adaptve method for nerta force dentfcaton n cantlever under movng mass Qang Chen 1, Mnzhuo Wang, Hao Yan 3, Haonan Ye 4, Guola Yang 5 1,, 3, 4 Department of Control and System Engneerng, Nanng Unversty,

More information

ELASTIC WAVE PROPAGATION IN A CONTINUOUS MEDIUM

ELASTIC WAVE PROPAGATION IN A CONTINUOUS MEDIUM ELASTIC WAVE PROPAGATION IN A CONTINUOUS MEDIUM An elastc wave s a deformaton of the body that travels throughout the body n all drectons. We can examne the deformaton over a perod of tme by fxng our look

More information

Lecture 14: Forces and Stresses

Lecture 14: Forces and Stresses The Nuts and Bolts of Frst-Prncples Smulaton Lecture 14: Forces and Stresses Durham, 6th-13th December 2001 CASTEP Developers Group wth support from the ESF ψ k Network Overvew of Lecture Why bother? Theoretcal

More information

OPTIMISATION. Introduction Single Variable Unconstrained Optimisation Multivariable Unconstrained Optimisation Linear Programming

OPTIMISATION. Introduction Single Variable Unconstrained Optimisation Multivariable Unconstrained Optimisation Linear Programming OPTIMIATION Introducton ngle Varable Unconstraned Optmsaton Multvarable Unconstraned Optmsaton Lnear Programmng Chapter Optmsaton /. Introducton In an engneerng analss, sometmes etremtes, ether mnmum or

More information

Open Systems: Chemical Potential and Partial Molar Quantities Chemical Potential

Open Systems: Chemical Potential and Partial Molar Quantities Chemical Potential Open Systems: Chemcal Potental and Partal Molar Quanttes Chemcal Potental For closed systems, we have derved the followng relatonshps: du = TdS pdv dh = TdS + Vdp da = SdT pdv dg = VdP SdT For open systems,

More information

Numerical Heat and Mass Transfer

Numerical Heat and Mass Transfer Master degree n Mechancal Engneerng Numercal Heat and Mass Transfer 06-Fnte-Dfference Method (One-dmensonal, steady state heat conducton) Fausto Arpno f.arpno@uncas.t Introducton Why we use models and

More information

Second Order Analysis

Second Order Analysis Second Order Analyss In the prevous classes we looked at a method that determnes the load correspondng to a state of bfurcaton equlbrum of a perfect frame by egenvalye analyss The system was assumed to

More information

Electron-Impact Double Ionization of the H 2

Electron-Impact Double Ionization of the H 2 I R A P 6(), Dec. 5, pp. 9- Electron-Impact Double Ionzaton of the H olecule Internatonal Scence Press ISSN: 9-59 Electron-Impact Double Ionzaton of the H olecule. S. PINDZOLA AND J. COLGAN Department

More information

Introduction to Vapor/Liquid Equilibrium, part 2. Raoult s Law:

Introduction to Vapor/Liquid Equilibrium, part 2. Raoult s Law: CE304, Sprng 2004 Lecture 4 Introducton to Vapor/Lqud Equlbrum, part 2 Raoult s Law: The smplest model that allows us do VLE calculatons s obtaned when we assume that the vapor phase s an deal gas, and

More information

Physics 5153 Classical Mechanics. Principle of Virtual Work-1

Physics 5153 Classical Mechanics. Principle of Virtual Work-1 P. Guterrez 1 Introducton Physcs 5153 Classcal Mechancs Prncple of Vrtual Work The frst varatonal prncple we encounter n mechancs s the prncple of vrtual work. It establshes the equlbrum condton of a mechancal

More information

The equation of motion of a dynamical system is given by a set of differential equations. That is (1)

The equation of motion of a dynamical system is given by a set of differential equations. That is (1) Dynamcal Systems Many engneerng and natural systems are dynamcal systems. For example a pendulum s a dynamcal system. State l The state of the dynamcal system specfes t condtons. For a pendulum n the absence

More information

Difference Equations

Difference Equations Dfference Equatons c Jan Vrbk 1 Bascs Suppose a sequence of numbers, say a 0,a 1,a,a 3,... s defned by a certan general relatonshp between, say, three consecutve values of the sequence, e.g. a + +3a +1

More information

Implicit Integration Henyey Method

Implicit Integration Henyey Method Implct Integraton Henyey Method In realstc stellar evoluton codes nstead of a drect ntegraton usng for example the Runge-Kutta method one employs an teratve mplct technque. Ths s because the structure

More information

Lecture 7: Boltzmann distribution & Thermodynamics of mixing

Lecture 7: Boltzmann distribution & Thermodynamics of mixing Prof. Tbbtt Lecture 7 etworks & Gels Lecture 7: Boltzmann dstrbuton & Thermodynamcs of mxng 1 Suggested readng Prof. Mark W. Tbbtt ETH Zürch 13 März 018 Molecular Drvng Forces Dll and Bromberg: Chapters

More information

Module 1 : The equation of continuity. Lecture 1: Equation of Continuity

Module 1 : The equation of continuity. Lecture 1: Equation of Continuity 1 Module 1 : The equaton of contnuty Lecture 1: Equaton of Contnuty 2 Advanced Heat and Mass Transfer: Modules 1. THE EQUATION OF CONTINUITY : Lectures 1-6 () () () (v) (v) Overall Mass Balance Momentum

More information

ALGORITHM FOR THE CALCULATION OF THE TWO VARIABLES CUBIC SPLINE FUNCTION

ALGORITHM FOR THE CALCULATION OF THE TWO VARIABLES CUBIC SPLINE FUNCTION ANALELE ŞTIINŢIFICE ALE UNIVERSITĂŢII AL.I. CUZA DIN IAŞI (S.N.) MATEMATICĂ, Tomul LIX, 013, f.1 DOI: 10.478/v10157-01-00-y ALGORITHM FOR THE CALCULATION OF THE TWO VARIABLES CUBIC SPLINE FUNCTION BY ION

More information

COSC 4397 Parallel Computation

COSC 4397 Parallel Computation COSC 4397 Solvng the Laplace Equaton wth MPI Sprng Numercal dfferentaton forward dfference formula y From the defnton of dervatves f( x+ f( f ( = lm h h one can derve an approxmaton for the st dervatve

More information

A Hybrid Variational Iteration Method for Blasius Equation

A Hybrid Variational Iteration Method for Blasius Equation Avalable at http://pvamu.edu/aam Appl. Appl. Math. ISSN: 1932-9466 Vol. 10, Issue 1 (June 2015), pp. 223-229 Applcatons and Appled Mathematcs: An Internatonal Journal (AAM) A Hybrd Varatonal Iteraton Method

More information

DO NOT DO HOMEWORK UNTIL IT IS ASSIGNED. THE ASSIGNMENTS MAY CHANGE UNTIL ANNOUNCED.

DO NOT DO HOMEWORK UNTIL IT IS ASSIGNED. THE ASSIGNMENTS MAY CHANGE UNTIL ANNOUNCED. EE 539 Homeworks Sprng 08 Updated: Tuesday, Aprl 7, 08 DO NOT DO HOMEWORK UNTIL IT IS ASSIGNED. THE ASSIGNMENTS MAY CHANGE UNTIL ANNOUNCED. For full credt, show all work. Some problems requre hand calculatons.

More information

Power law and dimension of the maximum value for belief distribution with the max Deng entropy

Power law and dimension of the maximum value for belief distribution with the max Deng entropy Power law and dmenson of the maxmum value for belef dstrbuton wth the max Deng entropy Bngy Kang a, a College of Informaton Engneerng, Northwest A&F Unversty, Yanglng, Shaanx, 712100, Chna. Abstract Deng

More information

Professor Terje Haukaas University of British Columbia, Vancouver The Q4 Element

Professor Terje Haukaas University of British Columbia, Vancouver  The Q4 Element Professor Terje Haukaas Unversty of Brtsh Columba, ancouver www.nrsk.ubc.ca The Q Element Ths document consders fnte elements that carry load only n ther plane. These elements are sometmes referred to

More information

Simulated Power of the Discrete Cramér-von Mises Goodness-of-Fit Tests

Simulated Power of the Discrete Cramér-von Mises Goodness-of-Fit Tests Smulated of the Cramér-von Mses Goodness-of-Ft Tests Steele, M., Chaselng, J. and 3 Hurst, C. School of Mathematcal and Physcal Scences, James Cook Unversty, Australan School of Envronmental Studes, Grffth

More information

A linear imaging system with white additive Gaussian noise on the observed data is modeled as follows:

A linear imaging system with white additive Gaussian noise on the observed data is modeled as follows: Supplementary Note Mathematcal bacground A lnear magng system wth whte addtve Gaussan nose on the observed data s modeled as follows: X = R ϕ V + G, () where X R are the expermental, two-dmensonal proecton

More information

Temperature. Chapter Heat Engine

Temperature. Chapter Heat Engine Chapter 3 Temperature In prevous chapters of these notes we ntroduced the Prncple of Maxmum ntropy as a technque for estmatng probablty dstrbutons consstent wth constrants. In Chapter 9 we dscussed the

More information

2.29 Numerical Fluid Mechanics

2.29 Numerical Fluid Mechanics REVIEW Lecture 10: Sprng 2015 Lecture 11 Classfcaton of Partal Dfferental Equatons PDEs) and eamples wth fnte dfference dscretzatons Parabolc PDEs Ellptc PDEs Hyperbolc PDEs Error Types and Dscretzaton

More information

Mathematical Preparations

Mathematical Preparations 1 Introducton Mathematcal Preparatons The theory of relatvty was developed to explan experments whch studed the propagaton of electromagnetc radaton n movng coordnate systems. Wthn expermental error the

More information

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009 College of Computer & Informaton Scence Fall 2009 Northeastern Unversty 20 October 2009 CS7880: Algorthmc Power Tools Scrbe: Jan Wen and Laura Poplawsk Lecture Outlne: Prmal-dual schema Network Desgn:

More information

Yong Joon Ryang. 1. Introduction Consider the multicommodity transportation problem with convex quadratic cost function. 1 2 (x x0 ) T Q(x x 0 )

Yong Joon Ryang. 1. Introduction Consider the multicommodity transportation problem with convex quadratic cost function. 1 2 (x x0 ) T Q(x x 0 ) Kangweon-Kyungk Math. Jour. 4 1996), No. 1, pp. 7 16 AN ITERATIVE ROW-ACTION METHOD FOR MULTICOMMODITY TRANSPORTATION PROBLEMS Yong Joon Ryang Abstract. The optmzaton problems wth quadratc constrants often

More information

Mathematical Modeling to Support Gamma Radiation Angular Distribution Measurements

Mathematical Modeling to Support Gamma Radiation Angular Distribution Measurements Mathematcal Modelng to Support Gamma Radaton Angular Dstrbuton Measurements V. Baty, O. Stoyanov Insttute for Safety Problems of Nuclear Power Plants, Natonal Academy of Scences of Ukrane Ukrane D. Fedorchenko,

More information

Games of Threats. Elon Kohlberg Abraham Neyman. Working Paper

Games of Threats. Elon Kohlberg Abraham Neyman. Working Paper Games of Threats Elon Kohlberg Abraham Neyman Workng Paper 18-023 Games of Threats Elon Kohlberg Harvard Busness School Abraham Neyman The Hebrew Unversty of Jerusalem Workng Paper 18-023 Copyrght 2017

More information

Appendix B: Resampling Algorithms

Appendix B: Resampling Algorithms 407 Appendx B: Resamplng Algorthms A common problem of all partcle flters s the degeneracy of weghts, whch conssts of the unbounded ncrease of the varance of the mportance weghts ω [ ] of the partcles

More information

coordinates. Then, the position vectors are described by

coordinates. Then, the position vectors are described by Revewng, what we have dscussed so far: Generalzed coordnates Any number of varables (say, n) suffcent to specfy the confguraton of the system at each nstant to tme (need not be the mnmum number). In general,

More information

Electronic Quantum Monte Carlo Calculations of Energies and Atomic Forces for Diatomic and Polyatomic Molecules

Electronic Quantum Monte Carlo Calculations of Energies and Atomic Forces for Diatomic and Polyatomic Molecules RESERVE HIS SPACE Electronc Quantum Monte Carlo Calculatons of Energes and Atomc Forces for Datomc and Polyatomc Molecules Myung Won Lee 1, Massmo Mella 2, and Andrew M. Rappe 1,* 1 he Maknen heoretcal

More information

The Study of Teaching-learning-based Optimization Algorithm

The Study of Teaching-learning-based Optimization Algorithm Advanced Scence and Technology Letters Vol. (AST 06), pp.05- http://dx.do.org/0.57/astl.06. The Study of Teachng-learnng-based Optmzaton Algorthm u Sun, Yan fu, Lele Kong, Haolang Q,, Helongang Insttute

More information

Lecture Note 3. Eshelby s Inclusion II

Lecture Note 3. Eshelby s Inclusion II ME340B Elastcty of Mcroscopc Structures Stanford Unversty Wnter 004 Lecture Note 3. Eshelby s Incluson II Chrs Wenberger and We Ca c All rghts reserved January 6, 004 Contents 1 Incluson energy n an nfnte

More information

An Interactive Optimisation Tool for Allocation Problems

An Interactive Optimisation Tool for Allocation Problems An Interactve Optmsaton ool for Allocaton Problems Fredr Bonäs, Joam Westerlund and apo Westerlund Process Desgn Laboratory, Faculty of echnology, Åbo Aadem Unversty, uru 20500, Fnland hs paper presents

More information

High resolution entropy stable scheme for shallow water equations

High resolution entropy stable scheme for shallow water equations Internatonal Symposum on Computers & Informatcs (ISCI 05) Hgh resoluton entropy stable scheme for shallow water equatons Xaohan Cheng,a, Yufeng Ne,b, Department of Appled Mathematcs, Northwestern Polytechncal

More information

Report on Image warping

Report on Image warping Report on Image warpng Xuan Ne, Dec. 20, 2004 Ths document summarzed the algorthms of our mage warpng soluton for further study, and there s a detaled descrpton about the mplementaton of these algorthms.

More information

First Law: A body at rest remains at rest, a body in motion continues to move at constant velocity, unless acted upon by an external force.

First Law: A body at rest remains at rest, a body in motion continues to move at constant velocity, unless acted upon by an external force. Secton 1. Dynamcs (Newton s Laws of Moton) Two approaches: 1) Gven all the forces actng on a body, predct the subsequent (changes n) moton. 2) Gven the (changes n) moton of a body, nfer what forces act

More information

ECE559VV Project Report

ECE559VV Project Report ECE559VV Project Report (Supplementary Notes Loc Xuan Bu I. MAX SUM-RATE SCHEDULING: THE UPLINK CASE We have seen (n the presentaton that, for downlnk (broadcast channels, the strategy maxmzng the sum-rate

More information

CSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography

CSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography CSc 6974 and ECSE 6966 Math. Tech. for Vson, Graphcs and Robotcs Lecture 21, Aprl 17, 2006 Estmatng A Plane Homography Overvew We contnue wth a dscusson of the major ssues, usng estmaton of plane projectve

More information

Solutions to exam in SF1811 Optimization, Jan 14, 2015

Solutions to exam in SF1811 Optimization, Jan 14, 2015 Solutons to exam n SF8 Optmzaton, Jan 4, 25 3 3 O------O -4 \ / \ / The network: \/ where all lnks go from left to rght. /\ / \ / \ 6 O------O -5 2 4.(a) Let x = ( x 3, x 4, x 23, x 24 ) T, where the varable

More information

A FAST HEURISTIC FOR TASKS ASSIGNMENT IN MANYCORE SYSTEMS WITH VOLTAGE-FREQUENCY ISLANDS

A FAST HEURISTIC FOR TASKS ASSIGNMENT IN MANYCORE SYSTEMS WITH VOLTAGE-FREQUENCY ISLANDS Shervn Haamn A FAST HEURISTIC FOR TASKS ASSIGNMENT IN MANYCORE SYSTEMS WITH VOLTAGE-FREQUENCY ISLANDS INTRODUCTION Increasng computatons n applcatons has led to faster processng. o Use more cores n a chp

More information

Single-Facility Scheduling over Long Time Horizons by Logic-based Benders Decomposition

Single-Facility Scheduling over Long Time Horizons by Logic-based Benders Decomposition Sngle-Faclty Schedulng over Long Tme Horzons by Logc-based Benders Decomposton Elvn Coban and J. N. Hooker Tepper School of Busness, Carnege Mellon Unversty ecoban@andrew.cmu.edu, john@hooker.tepper.cmu.edu

More information

Physics 607 Exam 1. ( ) = 1, Γ( z +1) = zγ( z) x n e x2 dx = 1. e x2

Physics 607 Exam 1. ( ) = 1, Γ( z +1) = zγ( z) x n e x2 dx = 1. e x2 Physcs 607 Exam 1 Please be well-organzed, and show all sgnfcant steps clearly n all problems. You are graded on your wor, so please do not just wrte down answers wth no explanaton! Do all your wor on

More information

The Finite Element Method

The Finite Element Method The Fnte Element Method GENERAL INTRODUCTION Read: Chapters 1 and 2 CONTENTS Engneerng and analyss Smulaton of a physcal process Examples mathematcal model development Approxmate solutons and methods of

More information

NON-CENTRAL 7-POINT FORMULA IN THE METHOD OF LINES FOR PARABOLIC AND BURGERS' EQUATIONS

NON-CENTRAL 7-POINT FORMULA IN THE METHOD OF LINES FOR PARABOLIC AND BURGERS' EQUATIONS IJRRAS 8 (3 September 011 www.arpapress.com/volumes/vol8issue3/ijrras_8_3_08.pdf NON-CENTRAL 7-POINT FORMULA IN THE METHOD OF LINES FOR PARABOLIC AND BURGERS' EQUATIONS H.O. Bakodah Dept. of Mathematc

More information

Nice plotting of proteins II

Nice plotting of proteins II Nce plottng of protens II Fnal remark regardng effcency: It s possble to wrte the Newton representaton n a way that can be computed effcently, usng smlar bracketng that we made for the frst representaton

More information

Stability of Tip in Adhesion Process on Atomic Force Microscopy Studied by Coupling Computational Model

Stability of Tip in Adhesion Process on Atomic Force Microscopy Studied by Coupling Computational Model Appl. Sc. Converg. Technol. 26(1): 6-10 (2017) http://dx.do.org/10.5757/asct.2017.26.1.6 Research Paper Stablty of Tp n Adheson Process on Atomc Force Mcroscopy Studed by Couplng Computatonal Model Yasuhro

More information

In this section is given an overview of the common elasticity models.

In this section is given an overview of the common elasticity models. Secton 4.1 4.1 Elastc Solds In ths secton s gven an overvew of the common elastcty models. 4.1.1 The Lnear Elastc Sold The classcal Lnear Elastc model, or Hooean model, has the followng lnear relatonshp

More information

Computational Fluid Dynamics. Smoothed Particle Hydrodynamics. Simulations. Smoothing Kernels and Basis of SPH

Computational Fluid Dynamics. Smoothed Particle Hydrodynamics. Simulations. Smoothing Kernels and Basis of SPH Computatonal Flud Dynamcs If you want to learn a bt more of the math behnd flud dynamcs, read my prevous post about the Naver- Stokes equatons and Newtonan fluds. The equatons derved n the post are the

More information

Additional Codes using Finite Difference Method. 1 HJB Equation for Consumption-Saving Problem Without Uncertainty

Additional Codes using Finite Difference Method. 1 HJB Equation for Consumption-Saving Problem Without Uncertainty Addtonal Codes usng Fnte Dfference Method Benamn Moll 1 HJB Equaton for Consumpton-Savng Problem Wthout Uncertanty Before consderng the case wth stochastc ncome n http://www.prnceton.edu/~moll/ HACTproect/HACT_Numercal_Appendx.pdf,

More information

RELIABILITY ASSESSMENT

RELIABILITY ASSESSMENT CHAPTER Rsk Analyss n Engneerng and Economcs RELIABILITY ASSESSMENT A. J. Clark School of Engneerng Department of Cvl and Envronmental Engneerng 4a CHAPMAN HALL/CRC Rsk Analyss for Engneerng Department

More information