Experimental Chemistry II
Experimental Chemistry II CHEM C1260
Popular in Course
verified elite notetaker
Popular in Chemistry
This 177 page Class Notes was uploaded by Kailyn Runolfsson on Saturday September 19, 2015. The Class Notes belongs to CHEM C1260 at Purdue University taught by Staff in Fall. Since its upload, it has received 113 views. For similar materials see /class/208008/chem-c1260-purdue-university in Chemistry at Purdue University.
Reviews for Experimental Chemistry II
Report this Material
What is Karma?
Karma is the currency of StudySoup.
You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!
Date Created: 09/19/15
CHll333 LECTURE 33 111907 FALL 2007 Professor Christine Hrycyna The Pentose Phosphate Pathway Phosphogluconate Pathway An Alternative Pathway for Glucose ilucosc rphusphulc V ilucmc 1 NADPHquot hplhmpllulu V m dchydrngcndsc quot 1 II 6Phosphng1uconnlzlctnnc 1130 ilucnnnlnclnnam r quot11 vPhosphogluconme NADP I lniwlniglucnnuu M w dchylrogcntm 0 Rihulnse iplnuphzne Rihulusu39 R1111th phllllhllk39 5phnsplluli39 Lx39pilut nmg p isnmu usc Xylulmc Rlbmr Sphospliiilu Fphmplmly 1 l 1 lrunskriulusc Au 1 7 Sedoheplulose Ch ccraldellydc 7phosphale phosphmc 1 l nlnsuldnthL E ry lhmse Jphusplmlc 1 Lr L 39Ilunxh luluw 1 ll ceruldeliyde vphosphme Fructose H39uclosc iphnsphatc vphosphnle Pathway has three primary functions xiilinivu stage No mulnnve surge 1 Generate NADPH for reductive biosynthesis needed for synthesis of fatty acids and steroids Similar to NAD NADP has an extra phosphate The pentose phosphate pathway generates almost all cellular NADPH used in reductive biosynthesis In mammals this pathway is prominent in tissues actively carrying out biosynthesis of fatty acids and steroids from small precursor molecules need NADPH Glucose Glucose a 6 Phosphate Fructose Glucose 6 Phosphate k 1Phosphate t 6Phosphogluconate t Glycolysis h Glycogen Pentose phosphates In most tissues 8090 of all glucose oxidation is by glycolysis Rest is Pentose Phosphate Pathway Glucose can enter this pathway after conversion to glucose 6ph0sphate important branch point Secondary pathway for glucose metabolism that leads to specialized products that the body needs No ATP consumed or generated 0 o H H 11 H H H fjCNHg nimtmumldc fj NHJ I 111HL T CH1 0 N I CH2 0 N OP O H H op0 H H 1 PW I H 11 I HO OH NH2 393 HO OH NH 013 0 y N 013 0 N N CH ltN 1 J CH ltN 1 J O 2 O N O 2 O N H H H H H H H H Nicolinamidcadcninc Qp0 IklkliliUIKd 39 39 lhnspi l lL dinucleolide 1 NADH O Nicolinamidc adenine dinucleolide phosphule NADPH Figure 1514 Concepts in Biochemistry 3Ie C 2006 John Wiley 81 Sons 0 Especially prevalent in the mammary gland adipose tissue the adrenal cortex and the liver 0 Other tissues eg skeletal muscle brain have virtually no pentose phosphate activity 233 CHM333 LECTURE 33 111907 FALL 2007 Professor Christine Hrycyna 2 Generate Ribose 5ph0sphate R5P for the biosynthesis of nucleotides nucleic acids and several enzyme cofactors Enzymes of the pathway are located in the cytosol 39239 The pathway is divided into an oxidative and nonoxidative branch 0 H C OH 0 The oxidative branch pr aduces NADPH and ribulose 5ph osph ate l H C OH as G6P 1s ox1d1zed H OH o In the nonoxidative branch depending on cellular conditions the O pentose phosphates are either converted to ribose Sphosphate or 2 converted to the glycolytic intermediates F fructose6phosphate F6P and H 10 0 H H I quot1 me H glyceraldehyde 3phosphate GAP HO ltHm n AH m H0 in H W e w l I Conversion to F6P and GAP provides a an 39 route back to glycolysis or r gluconeogenesis HMS HrHll H1 H 3 Convert pentoses to hexoses feeding into H h r glycolysis Equally important reactions of the PPP are to JUC39D Sphmhoi h39m 435 W j390fm convert dietary 5 carbon sugars into both 6 fructose6 phosphate and 3 glyceraldehyde 3 phosphate carbon WC r sugars which can then be utilized by the pathways of H H u s H Ll glyCOIYSlS no OH H Hif illll H OH H 1 OH quot quot l li lifquot quot n on 1quot 1mm H7 7ll NADP NADPH o OVERALL REACTION L ll ream Cilgon lt 3 glucose6P NADP 9 3C02 6 NADPH 6 H H e S quot ll m HeCrUH 2 fructose6P glyceraldehyde 3 P Hz x H 4 Clipwn cuppa kihulmm phmplmw thlucAKPH plank Figure 1615 Concepts in Biochemistry 32 9 2006 John Wiley 8 Sons Table 167 The reactions of the pentose phosphate pathway Reaction Reaction Number Reaction Enzyme Type 1 Glucose ATP glucose phosphate ADP Hexokinase 2 2 Glucose6phosphate NADP GlucoseGphosphate 6phosphoglucono3 lactone NADPH H dehydrogenase l 3 6Phosphoglucono8lactone HZO 6Phosphogluconolactonase 3 6phosphogluconate 4 6Phosphogluconate NADP 6Phosphogluconate 14 DribuloseSphosphate NADPH H CO dehydrogenase 5 DRibuloseSphosphate riboseSphosphate RibuloseSphosphate isomerase 5 Enzymes are listed by common names bReaction type 1 oxidationrreduction 2 phosphoryl transfer 3 hydrolysis 4 non11ydrolytic cleavage addition or elimination5 isomerizationrrearrangement 6 bond formation coupled to ATP cleavage Table 167 Concepts in Biochemistry 32 2006 John Wiley amp Sons 234 CHlVl333 LECTURE 33 111907 FALL 2007 STEP 1 Hexokinase CHZOH H 0 H H OH H ATP HO OH hexoklnuse H OH Glucose Figure 1515 pan 1 Concepts in Biochemistry 3Ie 2006 John Wiley amp Suns Group transfer reaction Same reaction as we saw in glycolysis STEP 2 Glucose 6Phosphate Dehydrogenase 39vgeaction a H CH20PO 39 H O H N ADP NADPH H OH H glucose6phosphalc H OH dehydrogenase H OH Glucose phosphate Figure 1615 part 2 Concepts in Biochemistry 3Ie 2006 John Wiley 8 Sons OxidationReduction Reaction Uses NADP as the cofactor Forms 6phospho Dgluconolactone Professor Christine Hrycyna 04201903 H 0 H H OH H H OH OH GI ucose phosphale ADP CH20PO3 6 Ph0sphoglucono 6 lactone Lactones are cyclic esters formed when an acid and an alcohol group on the same molecule react and usually require that a 5 or 6membered ring be formed 0 This is the regulated step depends on the of NADPH which is a strong inhibitor of the enzyme product inhibition 39 Glucose 6P DH De ciency I The most common known enzymopathy 0 300 known genetic variants 516 amino acid enzyme 0 Affects about 400 million people predominantly in tropical Africa and Asia part of Middle East Mediterranean amp Papua New Guinea 0 Similar to sickle cell trait distribution 235 CHlV1333 LECTURE 33 111907 FALL 2007 Professor Christine Hrycyna Most asymptomatic People usually don t have a problem until exposed to overwhelming oxidative stress 0 Hemolytic anemia 39 Sulfa drugs antimalarials certain herbicides 39 Certain infections I Eating fava beans divicine Glucose J y 7 39 P L P NADH pwggmaln H 1 L LPff l ff f39g39 V 7 7 7 7 7 EVll Vzlirlyrlvgtoflre lt7 manna seqHI 9 HutHIS 7 In red blood cells the major role of 3 SH NADPH is to reduce the disul de form Glutathme of glutathione to the sulfhydryl form reduced l The reduced glutathione is pertinent V Glu Cys Gly for maintaining the normal structure of red blood cells and for keeping hemoglobin in the ferrous state FeII unnumbered figure pg 503 Cnncepts in Biodmmistry 3e e 2006 John Wiley amp Sons 0 Red blood cells use this pathway to produce needed NADPH to maintain reduced iron and to maintain the concentration reduced glutathione to SmM 0 Reduced glutathione is necessary for the removal of H202 and lipid peroxides generated by radical oxygen species including hydrogen peroxide H202 superoxide free radicals other reactive oxygen species generated as metabolic reactions and by the action of some drugs see above Converts H202 to water 0 In people with G6PDdef1ciency radical oxygen species cause cellular damage since NADPH production is diminished and detoxi cation is inhibited Erythrocyte membrane breakdown and subsequent rupture hemolysis results De ciency also leads to resistance to the malarial parasite Plasmodiumfalciparum among individuals of Mediterranean and African descent The basis for this resistance is the weakening of the red cell membrane the erythrocyte is the host cell for the parasite such that it cannot sustain the parasitic life cycle long enough for productive growth 236 CHll333 LECTURE 33 111907 CHZOPO H 0 H20 H H OH H O 6 phospho H glucono laclonasc H OH 6Phosphoglucono 5actone Figure 1515 pan 3 Cuncepts in nimhemisuy 3le c Zou luhn Wiley 5 Suns 395 e action 9 FALL 2007 Professor Christine Hrycyna O CO STEP 3 Gluconolactonase H OH HO Cl H The rst step is hydrolysis of 6phosphoD H f OH glucolactone to gluconate a sugar acid with a H ICiOH COOH instead of an aldehyde at C1 0120130 6 phosphogluconate 0I O H C OH C0 HO C H NADP NADPH HgCiOH 6 ph0sphogluconale HCOH dehydrogenase CHZOPO 6phosphogluconate Figur21615 pan 4 Concepts in Biochemistry 3e 2006 John Wiley 8 Sons Hydrolysis Reaction STEP 4 6Phosphogluconate Dehydrogenase The next oxidation step releases that carboxyl group as C02 again transferring the electrons to CH2OH NADP CO 2 moles NADPHmole G6P are HTTOH formed in this portion of the H lj OH pathway CHzopog Now have a 5 C sugar pentose OxidationReduction and Nonhydrolytic cleavage addition or elimination RibuloseS phosphate STEP 5 Phosphopentose Isomerase IsomerizationRearrangement Reaction 0 H CHZOH c CO H li OH HCOH ribulose Sphosphate HCOH isomerasc H C OH H C OH CH20P03 CH20P0 Ribulose S phosphate Figure 15 15 pan 5 Concepts in Biochemistry 3le 2006 John Wiley 84 Sons Ribose S phosphate 237 CHM333 LECTURE 33 11 1907 FALL 2007 Professor Christine Hrycyna REGULATION Flux through PPP and rate of NADPH production is controlled by the rate of G6PDH Regulated by NADP concentration 7 high concentration increases activates G6PDH Ribulose 5 ph0sphate is converted to either Ribose S phosphate RSP and01 fructose 6 phosphate F6P and glyceraldehyde 3 ph0sphate GAP 0 Both F6P and GAP can enter glycolysis or gluconeogenesis Glucose Fatty acid synthesis 2NADP 2NADPH Other reactions Glucose v Glutathione reduction 6 h t my a e Ribulose Fructose 5 h h t 6 phosphate p 05p ae ll Nonoxidative GI ceraldehyde Rb g phosphate 5ph39ogr ate 3m NADH 3m ATP 4 Nucleotide biosynthesis Pyruvate Glycolysis or Gluconeogenesis Pentose Phosphate Pathway FOUR CELLULAR POSSIBILITIES 1 More ribose S phosphate than NADPH needed Synthesize R5P without making NADPH by bypassing the oxidative steps Use F6P and GAP from glycolytic pathway 2 Both ribose S phosphate and NADPH needed First four reactions predominate Glucose 6phosphate 9 ribose 5phosphate All the I39ibulose 5 ph0sphate is isomerized to ribose 5 ph0sphate and the pathway is completed 238 CHM333 LECTURE 33 111907 FALL 2007 Professor Christine Hrycyna 3 More NADPH than riboseSphosphate is needed R5P made into F6P and GAP for entry into gluconeogenesis In the case of recycling through gluconeogenesis 1 molecule of glucose6 phosphate can be converted by 6 cycles of the pentose phosphate pathway and gluconeogenesis to 6 C02 and 12 NADPH molecules 4 Need NADPH and ATP but NOT riboseSphosphate RibuloseSphosphate converted to F6P and GAP for glycolysis Make ATP and NADH NADPH generated through oxidative steps of PPP Glucose 6phosphate l Fructose R39b 7 5PhOSPhate 7 5Phlo atel El 2NADP 2NADPH Glucose Ribulose Fr 6phosphate Sphosphate 16 bisphosphate CO2 Dihydroxyacetone A Glyceraldehyde Ribose phosphate quot 3phosphate 539Ph05Phate Mode 4 l d 2 NADP r 2 NADPH 2 NADP 2 NADPH El Glucose Ribulose Glucose Rlbulose 6phosphate 5phosphate 6phosphate 5phosphate co2 CO2 Fructose Ribose Fructose Ribose 6phosp39hate 5phosphate 6ph05phate 39 5phosphate amp Fructose Fructose v 16bisphtf1hate 16blsphosphate Dihydroxyacetone A Dihydroxyacetone I Glyceraldehyde Glyceraldehyde phosphate 3phosphate phosphate 3ltphosphate b 20ATP l Pyruvkatel 239 Mixtures and Multiple Classes Randy Julian Lily Research Laboratories Linear boundaries for more than two classes W m w ambiguous region Less ambiguity Three Class Example Regression oflndividual Responses Linear Decisions for Multiple Classes 7 Bayes Decision Rule Choose most likely class given the features measured and the likelihood of the class 41 Using the discriminant function action agi clzisxi catinn discrim inam functions n Regression with three classes Regression ofResponse Matrix 918 Way to get the plot ex21r p lt data frame Poprll Pop5x2classind POPSYH namespltcquotxlquot quotx2quot quotaquot quotbquot quotcquot xp lt seq60 100l length 100 np lt lengthxp yp lt seq60l 100 length 100 pt lt expandgridx1 xp x2 yp ga lt lm a x1 x2 data p gb lt lm b x1 x2 data p gc lt lm c x1 x2 data p Za lt predictgal pt Zb lt predictgb pt Zc lt predictgcl pt decplotxp YP Zal Zb Zc quotRegression of Response Matrixquot 2 decpot lt function decplot lt functionxp YP Zal Zb Zc t plotPop 1 Pop 2 xlimc6010 ylimc610 type quotnquot xlab quotx1quot ylab quotx2quot titlet foril in 13 set lt Pop5ylevelsPoP5Y il textPopsetl 1 Pop setl els ascharacterPopSYset col 2 il 213 lt a pmaxZb contourxp yp matrixzp up39 add T levels 039 labex 0 p lt Zc pmaxza contourxp yp matrixzp npl add T levels 039 labex 0 invisible n Variance amp Bias k15 nK1 Low Bias b 5 El 5 1D x1 14 K1 ngh Variance k1 k1 f gya 2 quot t 2 m E m m k1 Conclusion multiple classes Using discriminant functions for each class and choosing highest value for a test point is superior to computing classtoclass boundanes Both linear regression and knn approaches work well for multiclass problems Classes containing mixtures of distributions Bayes Optimal Classi er 3 r Main problem with mixtures 11Xlwi Training Error 017 Test Error 022 Bayes Error 021 nTraining Set Means 55ubgroups N20 c2 k15 10 k15 11 l k15 12 3 Conclusion mixtures Q Q Q Test bias is inevitable when classes are overlapped or are mixtures With sufficient degrees of freedom training bias can always be 0 Mixtures result from choice of features In analytical chemistry we have control over features at both measurement and preprocessing stages In low dimension local methods can approximate Bayes boundaries Some problems are still linear 13 Next Alternatives to using data as features Linear Discriminant Analysis LDA Canonical Variates Dimensional problems with knn Local methods in high dimension Kernel methods in general Q Analytically informed dimensional reduction Using chemical knowledge to subset data Using spectral knowledge to process data LDA Preview Compute linear transformation to m an axis system which maximizes separation between classes n References 1 Pattern Classi cation Duda Hart Stork John Wiley Sons 2001 2 Modern Applied Statistics in S WN Venables BD Ripley 4th Ed SpringerVerlag 2002 3 quotThe Elements of Statistical Learning Data Mining Inference and Predictionquot Hastie Tibshirani and Friedman SpringerVerlag 2001 15 BME 5952 CHM 5992 Fall 2005 Physics of Spontaneous Raman Scattering Introduction to Biomedical Optics Electronic excited state Rayleigh Scattering Lecture 17 Stokes AntiStokes Raman Ram an Scattering Scattering Spontaneous Raman Microscopy and Applications to Diagnostic Detection Electronic ground state JiXin Cheng Rayleigh Ram an Scattering Scattering Wave number cm391 Classical Description of Raman Spectroscopy Three components of the induced dipole Incident field E E0 cosZmt y a E cos 27rvt 05EUQE cos 27rv vkt COSZIIV Vkt k Induced dipole source of emission 1 1 7 A A u 7 aEO cos 27rvt 1167 u I W VW NIn 45055 13am2 35 VW leexpthWkT a is Mo39eou39ar po39arizability Rayleigh scattering Raman scattering N is the number of molecules in the initial state m n is the final state 4 Janka V0 an eXp hvmn Iantirstakzs V0 an 4 kT or 050 E 5 jg cos 27ka higher order terms It Raman Tensor and Depolarization Ratio Polarization Raman Spectrum of Polystyrene Spontaneous Raman tensor CHCHgli m an 06a 0e E x CEHE or or or E u y W 2 E Normal Mode Picture of 2 an 05 0 22 z LabRam Jobin Womspexmon Molecular Vibration Excitation 6328 nm The symmetric and asymmetric parts of the Raman tensor 39quot 9ra i quot 39 1quot sewn 3054 cmquot aromatic CH a Grating 1800 grovesmm 0 0 Recorded on Aug 102001 Stremh39ng O O O O 39 39 xx yy 22 2906 cmquot antisymmetric aliphatic CH stretching Intensity 2a z ozniozyyZ aw four a 7092 6ayayzz 0521 2852 cmquot symmetric aliphatic CH stretching Depolarization ratio of a Raman band 7 r a 2 500 1000 1500 2000 2500 3000 1600 Cmquot CC Stremhing It 3a Raman Shiftcm1 p f p is in the range from 0 to 075 1002 cmquot Ring breathlng 1i 45a5 4a Molecular Microanalysis of Pathological Specimens in situ with a LaserRaman Microprobe Abraham etal Science 206 1979 716718 How a MicroRaman Spectrum Is Required Lymph Eyepiece node with and Video I i gt foreign camera Diffraction M cmsmpe Eeamsvlmer 39 V V 39 bodiesof Holographic Pinhole Grating x r 1 Silicone notch filte 39 rubber Microscope ublec ve 10 beam lncu5ing and sample viewing Ellinsoidal collection mirror quot gt Sample substrate Detectnr Recorder Double Electlunics munch mth I Schematic diagram ofthe laserRaman microscope Raman spectroscopy is sensitive the changes in moiecuiar compo and conformation that occur curing carcinogenesis of tissues Water objective lens NA12 660 nm excitation A Cytoplasm B Nuc eus C Chromatin Laser power 5 mW 150 seconds cm Schut et al Anal Chem 2000 72 60106018 aman 1 000 shift cm 00 O 00 0 Wide field Line Scan Single Point Bri gjd Imaging Imaging Mapping From Zhang and BenAmotz inquot M all 4 5 Pics Vic co cal CCD Detector Excitation Objective L Sem lo pv Siage pinhole SM scanning mirror beam splitter notch filter muquot cm J W etat Apphed 0pm 2002 41 Imaging with Inherent Signals from a Molecule No external marker tag or label are required Processed ngl39l Simplify the sample preparation procedure Raman Image Use NearInfrared Laser Source less photodamage to biological samples Biggest Challenge Very Weak Signal Raman cross section 103B cm2molecule Fluorescence cross section 10 1E cm2molecule J Ling etalApplied Optic5200241 6006 Resonance Raman Imaging oiTheraphihal T in Living Cancer Cells Bright eld image Raman image Total TP except aggregates Raman im ge TP aggregates Lm mummy Elzumnl sm Raman image TP bound u M RR spectrum ofT Simulation spectru 3 Based on reference spectra of TP bound to proteins 3 monomeric 4 and aggregated 5 in the cytoplasm m 1 Vibrational modes associated with electronic transition are enhanced roteins Raman image 2 UV excitation Resonance Raman ofp Mo Tp Frum P R Fan Academic Press new Yum 1982 Femanuv et al Eiuphys J 2mm 78 4997512 Surface Enhanced Raman Scattering SERS Power of Conventional Raman scattering Pv Nalfm vl Power of SERS PWW N612 AV12AV2 I01 Mechanisms of SERS d Muiecuie 1 Electromagnetic eld Wm enhancement 2 Chemical enhancem ent Field enhancement factor AVlc s2sn rd Fura review 522 K Kneipp Etai J Phys Cundens MattErZEIEIZ 14 R5977624 Aggregates of 20 nm gold particles coated with dipicolinic aCI A series of SERS spectra from a single spot with an integration time of 200 ms Exam 2 Ave 70 number 0 0 10 20 30 40 50 60 70 80 90100 score MultiIayer Neural Networks Randy Julan LIVy Research Laboratories NonLinear Systems 4 I multiple layers two layer Qt Schematic for K classes Output Yi Derived Features Zi Hidden Layer Hidden Unitsquot Input Xi What is going on Model gkYk Yk Regression Yk gkYk Y Softmax functionquotClassification Zm 0a0m xix m1M Yk z Ok fz k1K k213 3K3 Activation Functionquot 0v 1 1e Parameters Weights Zm 0a0m a2X m 1M Yk 0k kTZ k 1K wowaw m1MMp1 weights 0k k k1K KM1 weights N K Squared Errorquot or R9 y 10gfx k k Crossentropy deviance Gx arg kmaxmxi Classifier Comments about the model e With softmax activation and the crossentropy error function Exactly a linear logistic regression model in the hidden units All parameters estimated by maximum likelihood Large number of parameters Problem is probably over parameterized Don t want a global minimum for R6 it is probably an over fit solution lt11 Some regularization is needed to prevent over fitting Directly Penalty term Indirectly Early Stopping Gradient Descent Multiple error surfaces e Jw 1 a 6kiZml 075 km 05 8 025 5244 a 40 400 Qamp 12447 0 aw girl 20 all S 1111 11 a0 aCml Training Back Propagation Delta Rule gt Two pass method Forward pass 1 weights fixed Yk computed Backward pass 1 error in output layer computed 6ik 2 error in middle layer computed smi Smi 0arixii km6ki Compute the gradieInts and update m r 7 i 3R7 km km r i1 3315 N a m 0 R7 aml am yrZ i1 aafn r training iteration epoch y learning rate Early Stopping learning stopped initial weights W Problems with Early Stopping The issue of when to stop is important we know of no satisfactory rule for this algorithm Folklore suggests that disasters have been saved because convergence is so notoriously slow that users cannot afford the computer time to over fit the training data BD Ripley Pattern Recognition and Neural Networks p 155 Penalty Terms R09 116 Q Weight Decay M El n Weight Elimination 2 2 am 2 2 km1 km m11aml shrinks smaller weights towards zero cross validation used to nd A 46 Effects of weight decay Neural Network 10 Units No Weight Decay Neural Network 10 Units Weight Decay002 Training Error 0100 o 9390 Training Error 0160 o 030 Test Error 0259 Test Error 0223 I Bayes Error 0210 0 Bayes Error 0210 O Hidden Units It Number of hidden units governs complexity of decision boundary degrees of freedom Jn 040 A 035 test 030 025 O 20 train 015 9 13 17 21 25 2 9 33 37 total number I I I I I I I L of weights 23455739 IIH Seecting Learning Rates g J J J n lt nu quotI nap quotan lt 71 lt 2 no I l 77 gt 210quot n w wilt I w w Comparing ExamC NN5 Knn11 P uurm EH 1 1 a n 5E HE 055C 22 Ezzn 5E n1 oEnxmzz Z Ezzn 5E n5 on zz NN TrainC Wm H mm m n NN TrainC 10 HH EC EE ER on zz En EE a 055C 22 NNTminc n M NNTrzinC n W NNTrainC Dpra mm o a V f9 9 EH 9 e I I 39 39 39 M 75 D 5 m 15 Logistic Regression Size0 NN TrainC Si 9 0 Decav 0 01 o 2 0 0000 m 9o two layer g 0 0 0 8 a e o o O o o c o o o x X m o T 10 5 o 5 10 15 x1 libraryMASS lt need this too librarynnet nnet x y weights size Wts mask linout FALSE entropy FALSE softmax FALSE censored FALSE skip FALSE rang 07 decay 0 maxit 100 Hess FALSE trace TRUE MaxNWts 1000 abstol l0e 4 reltol l0e 8 size number of units in the hidden layer Can be zero if there are skip layer units linout switch for linear output units Default logistic output units entropy switch for entropy maximum conditional likelihood fitting Default by least squares softmax switch for softmax log linear model and maximum conditional likelihood fitting linout entropy softmax and censored are mutually exclusive skip switch to add skip layer connections from input to output 13 NNets in R ex51R TrainCltereadtable traincdat names TrainC ltec quotX1 X2 y p lt7 asmatriXTrainC 73 tplteTrainCy tpi C1assindtp lt7 seqltminltTrainCXD maxltTrainCXl length 50 mp lt7 1engthltxpgt lt7 seqltminltTrainCX2 max pt lt7 expandgridxl Xp X2 yp TrainCX2 length 50 setseed1 Znn lt nnetp tpi skipzT maxit1000 Znnt lt predictZnn pt softmaxT sizes decayDecay zpnn lt7 Znnt1eZnnt2 Neural Nets in the News J Chem Inf Cum111 Sci 1998 38 450456 Aqueous Solubility Prediction of Drugs Based on Molecular Topology and Neural Network Modeling Jarmo Huuskonen Marja Sale and lyrki Taskjnenquot Division of Pharmaceutical Chemistry Deparlmenl of Pharmacy PCB 56 FIN00014 University of Helsinki Finland Received November 2 997 A method for predicling the aqueous solubilny of drug compounds was developed based on 0pc ogical indices and arti cial neural nelwurk ANN modeling The aqueous solubility values for 2H drugs and 1 r 39 quot nulral XL Munii u a from the literalure The data set divided ima a training set 1 I60 and a randomly chosen test Sci 1 51 Simerurul pal amelel39s used as inputs in a 2 7571 arti cial neural network included 14 amm type electrolopologica indices and nine olher topological indiees For the tesl set a pl39ediclive 4 n d s 053 log unils were achieve 14 J Chem Inf Comput Sci 2000 40 9477955 947 Neural Network Modeling for Estimation of Partition Coef cient Based on AtomType Electrotopologieal State Indices Janno J Huuskonen David J Livingstone and Igor V TetkoM Division of Pharmaceutical Chemistry Department of Pharmacy PCB 56 University of Helsinki Helsinki F1N700014 Finland ChemQuesl Delamerc House 1 Ro al Crescent Sandown isle of Wi ht P036 8L2 Ul Centre for Molecular Design University of Portsmouth Portsmouth l39lants POI ZEG UK Laboratoire de Neurorlrleuristique Institut de Physiologic Universit tie Lausanne Rue du Bugnon 7 Lausanne CHIOOS Switzerland and Biomedical Department Institute of Bioorganie amp Petroleum Chemistry Murmanskaya l Kiln7660 253660 Ukraine Received May 14 1999 A method for predicting log P values for a diverse set of 1870 organic molecules has been developed based on atotn type electrotopologicalstate Estate indices and neural network modeling An extended set of Estate indices which included specific indices with a more detailed description of amino carbonyl and s w t 39 the current study For the training set of 1754 molecules the squared correlation coef cient and rootmeansquared error were r2 090 and RMSLOO 046 respectively Structural parameters which included molecular weight and 38 atomtype Estate indices were used as the inputs in 3951 arti cial neural networks The results from multilinear regression analysis were I1 087 and RMSLOQ 055l respectively For a test set of 35 nucleosides 12 nucleoside bases 19 drug compounds and 50 general organic compounds II 116 not included in me training set a predictive 1 2 094 and RMS 2 041 were calculated by artificial neural networks The results for the same set by niultilinear regression were 1 2 086 and RMS 072 The improved prediction ability of arti cial neural networks can be attributed to the nonlinear properties of this method that allowed the detection of highorder relationships between Estate indices and the noctanolwater partition coefficient The present approach was found to be an accurate and fast method that can be used for the reliable estimation of log P values for even the most complex structures 826 J Chem Inf Comput Sci 1995 35 826 833 Neural Network Studies 1 Comparison of Over tting and Overtraining Igor V Tetko David J LivingstoneFs and Alexander I Luik39M Biomedical Department Institute of Bioorganic and Petroleum Chemistry Mumtanskaya l Kiev660 253660 Ukraine ChemQuest Cheyney House 1921 Cheyney Street Steeple Morden Hens SGE OLP UK and Centre for Molecular Design University of Portsmouth Portsmouth Hams P01 256 UK Received January 27 1995 The application of feed forward back propagation arti cial neural networks with one hidden layer ANN to perform the equivalent of multiple linear regression MLR has been examined using arti cial structured data sets and real literature data The predictive ability of the networks has been estimated using a training test set protocol The results have shown advantages of ANN over MLR analysis The ANNs do not require 39 li order terms or indicator variables to establish complex structure activity relationships Overfming does not have any in uence on network prediction ability when overtraining is avoided by crossvalidation Application of ANN ensembles has allowed the avoidance of chance correlations and satisfactory predictions of new data have been obtained for a wide range of numbers of neurons in the hidden layer 15 n Validation Q Ideally we would set aside some of the data and use it as a validation set Q There is usually not enough data to do this Q Finesse the problem with K foldquot validation Q Can use Bootstrap Estimation KFold Validation Q Divide data into Kparts K510 etc Q CrossValidation estimate of prediction error CV ZLyn fAiK gtxi Ki iK N for Leave One Out LOO 16 Prediction error estimated bv CV d 10fold CV estimate I m xi E O E a a w t 5 o K actual g 1 0 1 5 0 SubseiSizep This is a Neural Network J F00 21 UgoX31921 f 39 i x I 39 i 39 i 39 i i I xi FX E I 39i39 39i 39i and this is a Support Vector Machine 18 33 Scattering Amplitudes Al Wasserman and Ki Burke 331 Introduction Electrons are constantly colliding with atoms and molecules in chemical reactions in our atmosphere in stars plasmas in a molecular wire car rying a current or when the tip of a scanning tunneling microscope in jects electrons to probe a surface When the collision occurs at low ener gies the calculations become especially dif cult due to correlation effects between the projectile electron and those of the target These boundfree correlations are very important For example it is due to boundfree corre lations that ultraslow electrons can break up RNA molecules Hanel 2003 causing serious genotoxic damage The accurate description of correlation effects when the targets are so complex is a major challenge Existing approaches based on wavefunction methods developed from the birth of uantum mechanics and perfected since then to reach great sophistication Morrison 1983 Burke 1994 Winstead 1996 cannot overcome the exponen tial barrier resulting from the manybody Schr dinger equation when the number of electrons in the target is larger Wavefunction based methods can still provide invaluable insights in such complex cases provided powerful com puters and smart tricks are employed see eg Grandi 2004 for lowenergy electron scattering from uracil but a truly abinitio approach circumventing the exponential barrier would be most welcome The purpose of this chapter is to describe several results relevant to this goal Imagine a slow electron approaching an atom or molecule that has N electrons and is assumed to be in its ground state with energy Egsi The asymptotic kinetic energy of the incoming electron is 5 so the whole system of target plus electron has a total energy of EGS 5 This is an excited state of the N 1electron system and as such it can be described by the linear response formalism of DFT starting from the ground state of the N 1electron systemi We will explain howl The targets we will consider must be able to bind an extra electroni For example take the target to be a positive ion so that the N 1electron system with groundstate energy Eggl is neutral Previous chapters in this book have described how to employ TDDFT to calculate eigi excitation energies corresponding to bound A bound transitions from the ground state However in the scattering situation considered here the excitation energy is A Wasserman and K Burke Scattering Amplitudes Lect Notes Phys 706 4937507 2005 DOI 101007375407354257333 SpringerrVerlag Berlin Heidelberg 2005 494 A Wasserman and K Burke known in advance it is I 5 where I is the rst ionization energy of the N 1system I Egg 7 It is the scattering phase shifts rather than the energies which are of interest in the scattering regime The TDDFT approach to scattering that we are about to discuss Wasserman 2005b is very different from wavefunctionbased methods yet exact in the sense that if the groundstate exchangecorrelation potential um and timedependent exchangecorrelation kernel fxc were known exactly we could then in principle calculate the exact scattering phase shifts for the system of N 1 interacting electronsl Any given approximation to um and fXC leads in turn to de nite predictions for the phase shifts The method involves the following three steps Finding the groundstate KohnSham potential of the N 1electron system vN17 ii Solving a potential seatten39ng problem namely scattering from UK1T and iii Correcting the Kohn Sham scattering phase shifts towards the true ones via linear response TDDFTI We start by reviewing those aspects of the linear response formalism of TDDFT that were introduced in Chap 1 and will be used in the following sec tions We then derive TDDFT equations for onedimensional scattering and work out in detail two simple examples to show how to calculate transmis sion and reflection amplitudes in TDDFTI The discussion is then generalized to three dimensions where we explain how the familiar single pole approxi mation for bound A bound transitions can be continued to describe bound A continuum transitions to get information about scattering states We end with a brief summary and outlook 332 Linear Response for the N 1Electron System For a thorough treatment see Chap 1 Here we only review what will be needed for the following sections The central equation of the linear response formalism of TDDFT is the Dysonlike response equation relating the sus ceptibility XN17 7quotw of a system of interacting electrons with that of its groundstate KohnSham analog leggl 7quotw Petersilka 1996a see 1123 The N 1 superscript was added in order to emphasize that we are going to perturb the groundstate of the N 1electron system where N is the number of electrons of the target In what follows however for notational simplicity the N 1 superscript will be dropped from all quantities We write the spindecomposed susceptibility in the Lehman representation XUUT I w lim clciw A 7w 3311 ngt0 Z M w 1 w i 9i i771r with N1 Fi WGSl aTlkTigt flu Z 5T 7 WWW 332 l1 33 Scattering Amplitudes 495 Where was is the ground state of the N lelectron system Ll7i its 2 excited state and 7307 is the aspin density operator In 331 2 is the GS A L17 transition frequency The term cc w A 7w stands for the complex conjugate of the rst term With w substituted by 7w The sum in 331 should be understood as a sum over the discrete spectrum and an integral over the continuum All excited states labelled by With non zero FiU 7 contribute to the susceptibility In particular the scattering state discussed in the introduction consisting on a free electron of energy 5 and an Nelectron target contributes too How to extract from the susceptibility the scattering information about this single state The question Will be answered in the following sections starting in one dimension 333 One Dimension 3331 Transmission Amplitudes from the Susceptibility Consider large distances Where the N lelectron groundstate density is dominated by the decay of the highest occupied Kohn Sham orbital Katriel 1980 the groundstate wavefunction behaves as Ernzerhof 1996 Mr was 1gs12 IN1gt SGS a 02 O39N1 333 mam N 1 Where gs is the groundstate wavefunction of the target SGS the spin func tion of the Nlelectron groundstate and the Nlelectron ground state density Similarly the asymptotic behavior of the ith excited state is IN1qbfzf 15i0390392u UN1 334 Where is an eigenstate of the target labeled by it 539 is the spin function of the ith excited state of the N l7system and 4ka a oneelectron orbital not to be confused with go notation reserved for Kohn Sham orbitals The contribution to Fi0z 332 from channels Where the target is ex cited vanishes as I A gt0 due to orthogonality We therefore focus on elastic scattering only Inserting 333 and 334 into the lD version of 332 and taking into account the antisymmetry of both was and Ll7i Fi0z 10 160 Z 5555 a UN1SZa UN1gt 335 72UN1 The susceptibility at large distances is then obtained by inserting 335 into the lD version of 331 496 A Wasserman and K Burke xz 1w ZXUUzzw 7 117ioo w 7 9i in 60it65Gs51 ccw 7gt 7w 336 I I X Z W 139 Since only scattering states of the N lelectron optical potential con tribute to the sum in 336 at large distances it becomes an integral over wavenumbers k v25 where 5 is the energy of the projectile electron I 1 1 DO Z m w gt g dk WWI 33 i w 7 217 177 117ioo 27f RUM w 7 016 177 T 39 t 39 f 39 I are boxnormal ed and qka where L 7 gt0 is the length of the box The transition frequency 21 EiN17 Eggl is now simply Qk Egg k22 7 Eg1 k22 I where I is the rst ionization potential of the N lelectron system and Egg and are the ground and 2 excited state energies of the M electron system The subscript RHLl indicates that the integral is over both orbitals satisfying ght and left boundary conditions eiik Tke z 7gt 00 R A tkei1m When I 7gt 700 and z 7x the integral of 337 is dominated by a term that oscillates in space with wavenumber 2V k2 7 2I and amplitude given by the transmission amplitude for spinconserving collisions tk at that wavenumber Denoting this by X05 and setting 5 k2 we obtain 338 z7ioo ixZe 5 hm 6 17z5 I l Hml nzn7zx 339 Therefore in order to extract the transmission amplitude t5 from the sus ceptibility when an electron of energy 5 collides with an Nelectron target in one dimension one should rst construct the groundstate density of the N lelectron system perturb it in the far left with frequency I 5 and then look at the oscillations of the density change in the far right the am plitude of these oscillations ampli ed77 by i n fll is the transmission amplitude t5 see Fig 331 The derivation of 339 does not depend on the interaction between the electrons Therefore the same formula applies to the Kohn Sham system 35 m ixggga am I 3310 quotx Vn1n1 In practice the Kohn Sham transmission amplitudes th5 are obtained by solVing a potential scatten39ng problem ie scattering from the N l electron groundstate KS potential li 17gt 33 Scattering Amplitudes 497 Density change Ground7state density Penmbau39an of N 1 sysmm Fig 331 Cartoon of 339 To extract the transmission amplitude for an electron of energy E scattering from an Nelectron target apply a perturbation of frequency E I on the far left of the N lground state system I is its rst ionization energy and look at how the density changes oscillate on the far right Once am pli ed the amplitude of these oscillations correspond to tEi Reproduction from the roof of the Sistine Chapel with permission from artist Illustration of 3310 For one electron the susceptibility is given by Maitra 2003b XKszz5I xnznz ngzz5 gfltsz 1 75 7 21 3311 Where the Green s function 9K5 z z 5 has a Fourier transform satisfying 9 l 92 71 7 vaz ngzz t 7 t 7161 7 z 6t 7 t 3312 Let s nd the transmission amplitude for an electron scattering from a double deltafunction well van 7Z16z 7 Z26z 7 a The Green s function can be readily obtained in this case as Z291I7 a91a7 1 3313 1 Z291a7 a ngOmv 91Iyr 7 where 91 is the Green s function for a single deltafunction of strength Z1 at the origin It is given by Szabo 1989 3314 1 Zeik rHizi 91171 elk 17quot 1 7 ikZ1 With k Having constructed XKS explicitly application of 3310 yields the correct answer ik Z1 ik i 3315 1 Z291a7 a th 498 A Wasserman and K Burke 3332 TDDFT Equation for Transmission Amplitudes The exact amplitudes t5 of the manybody problem are formally related to the th 5 through 339 3310 and 123 the timedependent response of the N1electron groundstate contains the scattering information and this is accessible via TDDFT A potential scattering problem is solved rst for the N1electron groundstate KS potential and the scattering ampli tudes thus obtained tKS are further corrected by fHXC to account for eg polarization effects Even though 339 is impractical as a basis for computations one can rarely obtain the susceptibility with the desired accuracy in the asymptotic re gions as we did in the previous example it leads to practical approximations The simplest of such approximations is obtained by iterating 123 once sub stituting X by XKS in the righthand side of 123 This leads through 339 and 3310 to the following useful distortedwave Born type approximation for the transmission amplitude 1 i25 In 3316 and from now on the doublebracket notation stands for t5 55 HOMO a foca IHOMO 5 3316 ltltHOMOefoclte Igt1HOMOegtgt dz dz awome mm a Imom 1 In 3317 where HOMO is the highestoccupied molecular orbital of the N1electron system and 013 is the energynormalized scattering orbital of energy 5 satisfying Klboundary conditions see 338 This is reminiscent of the singlepole approximation for excitation energies of bound A bound transi tions 131 Many other possibilities spring to mind for approximate solu tions to 339 3333 A Trivial Example N 0 The method outlined above is valid for any number of particles In particular for the trivial case 0 0 corresponding to potential scattering Consider an electron scattering from a negative deltafunction of strength Z in one dimension Fig 332 The transmission amplitude as a function of 5 is given by see Sect 25 of Griffiths 1995 ik t5 7 m k 7 3318 How would TDDFT get this answer Find the groundstate KS potential of the N1 17electron system The external potential admits one bound 33 Scattering Amplitudes 499 7quot Iquot x tg 2 90 Z 2 Fig 332 Left cartoon of an electron scattering from a negative deltarfunction potential Right cartoon of an electron bound to the same potential the ground state density decays exponentially just as in hydrogenic ions in 3D 2 lnl Fig 333 Left cartoon of an electron scattering from 1DHe1 Right cartoon of A111 Please PYOVide two electrons bound to the delta function in a singlet state gure citation in text for Fig 333 state of energy 7Z22 The groundstate KS potential is given by vaz vex z 1ch1 but 1ch 0 or one electron7 so va z Us I 7Z6z ii Solve the groundstate KS equations for positive energies7 to nd th5 ikZ 116 iii In this case7 foc 07 so X XKS and t th Notice that approximations that are not selfinteraction corrected to guarantee 1ch 0 would give sizable errors in this simple case 3334 A NonTrivial Example7 N 1 Now consider a simple 1D model of an electron scattering from a oneelectron atom of nuclear charge Z Rosenthal 1971 A 1 d2 1 d2 H7 7 7Z6 7Z6 A6 7 3319 2 dz 2 dzg 11 12 11 12 The two electrons interact via a deltafunction repulsion7 scaled by A With A 0 the ground state density is a simple exponential7 analogous to hydro genic atoms in 3D 500 A Wasserman and K Burke i Exact solution in the weak interaction limit First we solve for the exact transmission amplitudes to rst order in A using the static exchange method Bransden 1983 The total energy must be stationary with respect to variations of both the bound 451 and scattering orbitals that form the spatial part of the Slater determinant bzl 5zg i bzgq 5zl where the upper sign corresponds to the singlet and the lower sign to t e triplet case The staticexchange equations are 1 d2 2 5 7l sbrl Z5I 151251 Mbs bs17 3320 where 7 2A for the singlet and 0 for the triplet Thus the triplet transmis sion amplitude is that ofa simple 6function 3318 This can be understood by noting that in the triplet state the Hartree term exactly cancels the ex change the two electrons only interact when they are at the same place but they cannot be at the same place when they have the same spin from Paulils principle The results for triplet tmplet and singlet tsinget scattering are therefore ttriplet 750 7 750 E 3321a t 7 7ik2 1 k7 iZ2kiZ ii TDDFT solution We now show step by step the TDDFT procedure yielding the same result 3321 The rst step is nding the groundstate KS potential for two electrons bound by the 6function The groundstate of the N 1electron system N 1 is given to O by tsanglet to 2M1 3321b 1 Wcs110171202 E GS11 GSI2l601T6021 501150211 7 3322 where the orbital gogsz satis es Lieb 1992 Magyar 2004b 1 d2 75 7 Z6I Alsocsrl2 SDGSI MSDGS1 3323 To rst order in A spasm aw i yum e ZlEl4lel 7 3 3324 87 The bare KS 39 39 A tKS5 character e the r quot behavior of the continuum states of vaz 7Z6z Algogszl2 and can be obtained to O by a distortedwave Born approximation see eg Sect 414 of Friedrich 1991 tKS to Atl 3325 33 Scattering Amplitudes 501 08 r singlet triplet 3 06 r I E k 04 i interacting 02 7 KS 39 Fig 334 Real and imaginary parts of the KS transmission amplitude th and of the interacting singlet and triplet amplitudes for the model system of 3319 Z 2 and A 05 in this plot Reprinted With permission from Wasserman 2005b Copyright 2005 American Institute of Physics The result is plotted in Fig 334 along With the interacting singlet and triplet transmission amplitudes 3321 The quantity Atl is the error of the ground state calculation The interacting problem cannot be reduced to scattering from the N lKS potential but this is certainly a good starting point in this case the KS transmission amplitudes are the exact average of the true singlet and triplet amplitudes compare 3325 With 3321 We now apply 339 to show that the focterm of 123 corrects the th values to their exact singlet and triplet amplitudes The kernel foc is only needed to OM fo UUzzw A6ltI 7 117 600 3326 Where the foc of 123 is given to O by fo fH fx i 200 fo 00 here Equation 3326 yields xzzw XKsI 1w g dz XKszzwxz zwi 3327 Since the ground state of the 2electron system is a spinsinglet the Kronecker delta 65Gssl in 336 implies that only singlet scattering information may be extracted from x Whereas information about triplet scattering requires the magnetic susceptibility M 200 00XUU related to the KS susceptibility by spinTDDFT Petersilka 1996b chw mm m 7 g dz szltzx wgtMltz znwgti 3328 502 A Wasserman and K Burke For either singlet or triplet case since the correction to XKS is multiplied by A the leading correction to th 5 is determined by the same quantity7 gs 96 gs where 135 is the 0ch order approximation to the KS susceptibility ie7 with vaz 1135 7Z6z lts oscillatory part at large distances Maitra 2003b multiplied by nzn7zik see 339 is precisely equal to Atl We then nd through 339 3327 and 3328 that tsinglec th Air 7 ttriplet th A751 3329 in agreement with 3321 The method illustrated in the preceeding example is applicable to any one dimensional scattering problem Equations 339 and 123 provide a way to obtain scattering information for an electron that collides with an Nelectron target entirely from the N 1electron groundstate KS susceptibilty and a given approximation to fxc 334 Three Dimensions 3341 SinglePole Approximation in the Continuum We have yet to prove an analog of 339 for Coulomb repulsion in three dimensions But we can use quantumdefect theory Seaton 1958 to deduce the result at zero energy Consider the l 0 Rydberg series of bound states converging to the rst ionization threshold I of the N lelectron system E 7 Egg I 71 2239 7 ill2 3330 where M is the quantum defect of the ith excited state Let 5i 1 l2 Ksi2l 3331 be the KS orbital energies of that series The true transition frequencies Lul EZ 7 Egg are related through TDDFT to the KS frequencies wKS i 5i 7 EHQMQ where EHOMQ is the HOMO energy Within the singlepole approximation SPA Petersilka 1996a7 applicable to Rydberg excitations according to the criteria of applicability discussed in Appel 2003 w W3 i 2ltltH0Mo zquot focwiH0Mo igtgt 3332 Numerical studies AlSharif 1998 suggest that AM M 7 MKSJ is a small number whenz39 7 gt0 Expanding wi around AM 0 and using I 75HQMO we nd wi wKsi 7 AMn 7 MKS i3 3333 We conclude that7 within the SPA 33 Scattering Amplitudes 503 1 e 08 c F g 06 c 39 Jig a maverigr w V 7a m 04 7 y singlet 02 c 0 l l l l 0 02 04 06 08 1 Ene gy H Fig 335 sphase shifts as a function of energy for electron scattering from He Dashed lines the line labeled KS corresponds to the phase shifts from the exact KS potential of the He atom the other dashed lines correspond to the TDDFT singlet and triplet phase shifts calculated in the present Work according to 3335 Solid lines accurate Wavefunction calculations of electronHeJr scattering from Bhatia 2002 The solid line in the center is the aVerage of singlet and triplet phase shifts Dotted lines Static exchange calculations from Lucchese 1980 The asterisks at zero energy correspond to extrapolating the bound gt bound results of Burke 2002 Reprinted With permission from refWasserman 2005b Copyright 20057 American Institute of Physics AM 722 7 MKsi3ltltHOMO 239 focwZHOMO 2 3334 Letting i A gt07 Seatonls theorem 7rlimiH00 M 65 A 0 Seaton 1958 implies 6a 6mg 7 27rltltHOMO efoce 1HOMO 5 3335 a relation for the phaseshifts 6 in terms of the KS phaseshifts 6K3 applicable When 5 A 0 The factor iiuKS i3 of 3334 gets absorbed into the energy normalization factor of the KS continuum states We illustrate in Fig 335 the remarkable accuracy of 3335 When applied to the case of electron scattering from He For this system7 an essentially exact groundstate potential for the N 2 electron system is known This was found by inverting the KS equation using the groundstate density of an extremely accurate wavefunction calculation of the He atom Umrigar 1994 We calculated the lowenergy KS s phase shifts from this potential7 6K5 5 dashed line in the center7 Fig 3357 and then corrected these phase shifts according to 3335 employing the BPG approximation to foc Burke 2002 which amounts to using the adiabatic local density approximation for the 504 A Wasserman and K Burke antiparallel contribution to foc and exchangeonly approximation for the parallel contributioni We also plot the results of a highly accurate wave function calculation Bhatia 2002 solid and of staticexchange calculations Lucchese 1980 dottedi The results show that phase shifts from the Nl electron groundstate KS potential 6Ks5 are excellent approximations to the average of the true singlettriplet phase shifts for an electron scattering from the Nelectron target just as in the onedimensional model of the pre vious section they also show that TDDFT with existing approximations works very well to correct scattering from the KS potential to the true scat tering phase shifts at least at low energies In fact for the singlet phase shifts TDDFT does better than the computationally more demanding sta tic exchange method and for the triplet case TDDFT does only slightly worse Even though 3335 is strictly speaking only applicable at zero en ergy marked with asterisks in Fig 335 it clearly provides a good descrip tion for nite low energies It is remarkable that the antiparallel spin ker nel which is completely local in space and time and whose value at each point is given by the exchangecorrelation energy density of a uniform elec tron gas evaluated at the groundstate density at that point yields phase shifts for eHeJr scattering with less than 20 error Since a signature of densityfunctional methods is that with the same functional approximations exchangecorrelation effects are often better accounted for in larger systems the present approach holds promise as a practical method for studying large targetsi 334 2 Part ialWave Analysis For the case of spherically symmetric N lelectron ground states useful expressions can be derived for the transition matrix elements tmatrix in the angular momentum representation For example the matrix elements in the usual de nition Gonis 1992 t E ik l exp7ik6l sin 6 are given by tz ifs 4ltltfocgtgtz 7 3336 where the tfs are the Kohn Sham tmatrix elements and ltltfocgtgtz dhanmxdmanrln Xemr1 mr2yf omgomDiff 3337 with Mfrfl E Ylm 70quot In 3337 the go s are Tadz39alKohn Sham orbitals regular at the origin and the 457s are quasiparticle amplitudes determined by the asymptotic behavior of the interacting radial Green s function see Secti 232 of Wasserman 2005ai These are generally difficult to obtain in practice but approximating them by the corresponding Kohn Sham orbitals 33 Scattering Amplitudes 505 yields a simple prediction for the tmatrix elements Furthermore the single pole approximation of 3335 is obtained from 3336 after expanding it to rst order in 6 7 Slei 335 Summary and Outlook Based on the linear response formalism of TDDFT we have discussed a new way of calculating elastic scattering amplitudes for electrons scattering from targets that can bind an extra electroni In one dimension transmission ampli tudes can be extracted from the Nlelectron groundstate susceptibility as indicated by 339 Since the susceptibility of the interacting system is determined by the KohnSham susceptibility within a given approximation to the exchangecorrelation kernel the transmission amplitudes of the interact ing system can be obtained by appropriately correcting the bare KohnSham scattering amplitudes Equation 3316 reminiscent of the singlepole ap proximation for bound A bound transitions provides the simplest approxi mation to such a correction A similar formula for scattering phase shifts near zero energy 3335 was obtained in three dimensions by applying concepts of quantum defect theoryi These constitute rst steps towards the ultimate goal which is to accu rately treat boundfree correlation for lowenergy electron scattering from polyatomic molecules An obvious limitation of the present approach is that it can only be applied to targets that bind an extra electron because the start ing point is always the N lgroundstate Kohn Sham system which may not exist if the Nelectron target is neutral and certainly does not exist if the target is a negative ion In addition to extending the formalism to treat such cases there is much work yet to be done a general proof of principle in three dimensions testing of the accuracy of approximate groundstate KS poten tials developing and testing approximate solutions to the TDDFT Dysonlike equation extending the formalism to inelastic scattering etcr Thus there is a long and winding road connecting the rst steps presented here with the calculations of accurate cross sections for electron scattering from large tar gets when bound free correlations are important The present results show that this road is promising Of course the road goes ever on and on 77 Baggins 1973 but this section looks worthwhile Acknowledgements The origin of this book is two summer schools on TDDFT that were inde pendently organized one in the USA another in Europe The USA summer school took place in Santa Fe New Mexico June 5710 2004 It was organized by Carsten Ar Ullrich Kieron Burke and Giovanni Vignale supported by a generous grant from the Petroleum Research Fund Welcome to a new wonderful year of ChE The year is eady under way and classes aren t making it go any faster First off everyone needs to become a member of hE Bring your form and money to the AIChE of ce in room 6a anytime during the day There are a ton of cool events happening this year and I want everyone to 39oy the fun Proof that even Shakespeare Skipped 88 By Kristen Schebler To skip or not to skipthat is the question Whether tis nobler in the mind to suffer The yawns and boredom of attending class Or rather to seek our own diversions And in nding them rejoice To go to stay No doubtand by this quandary say we end The heartache and the thousand things we should be dorng In our college days tis a peace ofmind Devoutly sought by all To go to stay To stay perchance to nap Ay that s the thing For in that time of dodging class what dreams may When we have sidestepped life at last Must give us pause Ought we to have gone When sunlight washed the grassy lawn But who would bear the boredom of a lecture Th puns of the professor the attentiveness of stu The hunger pangs so prevalent in a luncheon class The pointlessness and sheer redundancy of notes When he himself might his escape make good By simply skipping class wide Tbix lime Meet Your Fall Come rmtttee Headsl Executive Board Members P132 2 r 5 P132 5 81011 112 Brought to you by the October 2005 Who would homework bear To grunt and swear over impossible compositions But that the dread ofC s to come That undiscovered country from whose clutches No student may return does chasten us And makes us rather do the work we have Then toss it to the winds of ckle Fate Thus conscience makes cowards of us all And thus the glory of a grand spring day Is shadowed within the sinister walls of dingy lec And plans of lazy drowsing neath the trees With this realization are set so long aside And lose their ne appeal But still What joy is theirs who ditch today make amends tomorrow UPCOMING EVENTS Flag Football Every Thursday evening at 1030pm So ball Every Friday evening at 600pm BOO at the Z00 Oct 23rd Sunday 14pm set up Oct 26th Wednesday 5745pm or 7 930pm Contact Kalyn if you have any questions kwenzlawpurdue edu Game Corner Answers P132 7 P132 8 Page 2 MEET YOUR COMMITTEE HEADS Sandra Azzano Newsletter Year Junior Summer Work Worked for sales engineers Professor most likely to win Jeopardy Prof Franses Collections Shot Glasses Ifyou nd a statue oneus made of15267 Popsicle sticks where are you Do I want to know What makes you want to dance My special someone Favorite ChE Mem ory All those long nights of home work right Fall Plans Study and nd an internship Kristine Hawkins Philanthropy Year Junior Summer Work O Charley s Professor most likely to win Jeopardy Elias Franses If you nd a statue of Zeus of 15267 pop sicle sticks where are you Somewhere I shouldn t be What makes you want to dance Feel good music Do the chickens have large tal ons Yes Favorite ChE Memory Staying up all night working on thermo homework just to discover I had the right answer at 3 in the a ernoon the day before AllI had to do was zoom in on the graph to see the data points Fall plans Class work eat SLEEP Jenny Farney Industrial Relations Year Sophomore Summer Work Waitress at a Country Club Do you have any collections what is it Iused to collect coins when I was younger but now I just collect periodic table paraphernalia If you nd a statue of Zeus made of 15267 popsicle sticks where are you In some third grader s bed room What makes you want to dance jazz Do the chickens have large talons Yes de nitely Favorite ChE Memory studying with Allison Yates another com mittee head Fall plans study study study study study study study devote a bit of time to my friends study study study Kalyn Wenzlawsh Philanthropy Year Junior Summer Work At a restaurant back home Akron OH Zeus statue Why would someone make a statue of Zeus out of popsicle sticks They must not be engineering students What makes you want to dance Hearing good music or having a guy ask me to dance Do the chickens have large talons Sorry too lazy to look up what a talon is Favorite ChE Memory Hopefully the soon to hap pen Fomey sleepover Fall plans To work my ChE happy butt off and maybe have a little fun mixed in as well j Page 3 Siddharth Sheth aka Sid IR Year Senior Summer Work Worked with a pigment company in France Professor most likely to win Jeopardy Michael Thunder Harris Collections I steal balls from computer mice I cur rently have 1024 of them Zeus statue In a refrigerator somewhere in Rome What makes you want to dance The fact that I can be myself J Chickens have large talons I wouldn t have the faintest idea Favorite ChE memory Working on the ANSYS project for ChE 378 Oh not really Fall Plans Hopefully get ajob Anything else all ChE s should know The principle of zee zermo couple and zee de nition of fuga city Stephanie Dennis Mentor Year Junior Summer Work Air Products Professor most likely to win Jeopardy Beaudoin if he e time Collections Giraffes since the 5th grade very nerdy I am we are If you nd a statue oneus made of 15267 popsicle sticks where are you I honestly wishI knew What makes you want to dance The music in my head Chickens have large talons No Fall Plans Pass all my classes and pretend to have a so cial life Fall plans Find ajob James Seto Finance Year Senior work ResearchiCaruthers Professor most likely to win Jeopardy Beau doin Zeus statue Gifted Kindergarten class Chickens have large talons Ma be Manoj Somasundaram Finance Year Freshman Professor most likely to win Jeopardy Tough one Collections Coins and stamps Zeus statue Greece probably What makes you want to dance Fast loud musrc Chickens have large talons Not according to me Fall plans To work diligently and enthusias tically in every piece of work Anything all ChE s should know They are part of a great organization and are heading towards greater things in future The eld of Chemical Engineering is one of the best Elaine Stewart Finance Year 5th Year Summer Work Studied abroad in Londoni ChE Senior lab Cooped at Uniqema in Chi cago IL Professor most likely to win Jeopardy Delgass What makes you want to dance Loud music Fall Plans Find ajob Page 4 Jeff JJ Dombek Philanthropy Year Sophomore Summer Work Painter at a school district Professor most likely to win Jeopardy not Dr Bob Collections Empty bottles of liquor Zeus statue I m lost What makes you want to dance Music and girls Chickens have large talons Yes Favorite ChE Memory Not class Fall plans Pass my classes Anything else all ChE s should know This is a hard major but an interesting ma I jor Ross Frey ChemECar Year Junior Summer Work Sears Professor most likely to win Jeopardy Professor Hair is Collections Not really anymore but I used to col lect rocks yep rocks Don t act like you didn t Zeus statute Tricky one butI am going to go with IU Its easier when you don t have all of that school work getting in the way Chickens have large talons Yes Favorite ChE memory Meeting with the best study group this side of the Mississippi almost everyday last semester to nish thermoChE 320 or organic chem Fall Plans At Purdue working though junior year rst semester of chemical engi neering Anything else all ChE s should know Its rough good luck though you will need it Randy Fletcher Social Athletic Year Senior Summer work Iworked in Prof Andres lab on the Synthesis of Gold nanoparticles for Selective Targeting of Cancer Cells Professor most likely to win Jeopardy Either Prof Harris Prof Franses or Prof Venkatasubramanian Zeus statute Outside of the Good Humor Fac tory What makes you want to dance I don t dance all I do is this Chickens have large talons Who looks at chicken talons But roosters have large talons Favorite ChE memory 11 Fall plans Class research relax Douglas Kilgore Social and Athletic Year Senior Summer Work Internship at Flint Hills Re sources Professor most likely to win Jeopardy Prof Franses Zeus stature A place where people have way too much time on their What makes you want to dance Rap music Favorite ChE memory The hour break be fore 378 on Friday s last spring Fall Plans Tailgate every Saturday home game Watch Purdue foot ball go to a BCS bowl n Watch the Colts go to the Anything else all ChE s should know Broke Phi Broke We ain t got it Albert Baramuli Alumni Relations Year Senior Summer Work Slacking off back home enjoying the sun and the humid weather Professor most likely to win Jeopardy Prof Zeus statue Some popsicle craving guy s dream What makes you want to dance Lean backLean backLean back Chickens have large talons Let me go and ask Colonel Sanders Favorite ChE memory Favorite BAD ChE mem ory Fall plans Surviving this se mester and not blowing any classes that I m taking Anything else all ChE s should know For seniors Stay fo cused we re almost there juniors Good luck you need it for sophomores TURN BACKKK Paul Basiaga ChECar Year Junior Summer Work Gallivanting around Europe Professor most likely to win Jeopardy Elias Franses Zeus statue The saddest place on Earth Chickens have large talons Yes yes they do Favorite ChE memory There are just so many Anything else all ChE s should know Don t take your free time for granted Page 5 Jessica Wilkes Social Athletic Year Junior Summer Work I stayed at Purdue and did research with Professor Beaudoin andI worked at the mall at American Ea le Professor most likely to win Jeopardy Professor Beaudoin he s very intelligent and upbeat Collections Not particularly I do have a lot of shoes though Zeus statue I have no idea a museum What makes you want to dance Music and being surrounded by people with a fun attitude Do the chickens have large talons I don t know but that is a quote from Napoleon Dynamite Favorite ChE Memory The CMFK science fairs LUM dinners just hanging out with all my fellow CHE s Fall plans Working at the mall still studying hard and going to all the football games Boiler Up Anything else all ChE s should know I m also in volved in Phi Sigma Rho the women in engineering technology sorority on carn pus and I have a little sister Taylor from Big Brothers Big Sisters Rochelle Nicole Chism Publicity Year Junior Summer Work Research Assistant for Professor Beaudoin Disney Store Cast Member Professor most likely to win Jeopardy Professor Beaudoin What makes you want to dance I That Nurna Nurna song hahaha Favorite ChE Memory All Ihave to say is Semibatch haha Fall plans To pass all my classes Page 6 Allison Yates Student Faculty Asmey Leido Student Faculty Year Sophomore Year Semor I Summer work Waitress at a GreekItalian Restaurant summer Work Cheetos IHHOVatIOHS TeaIH Collections Post cards they are usually better pic Fnto LaYgt Inc Dallas TX es that I can take anyway I only collect them from Pr0f9559r mOSt likely to Win Jeopardy PTOf places I ve been BSaHdoln I Zeus statue In a bad dream What makes You want to dance Anything I What makes you want to dance When the Denver want to dance 24 7 Do the chickens have large talons No but your mom has large talons Favorite ChE Memory Pulling an allnighter to nish therrno homework and then going to triple XXX before class Fall plans Working hard and playing harder AICHE intramural sports Broncos score Chickens have large talons Only if they don t give them the proper trimming they require Favorite ChE memory Discovering the answer to life while studying with Jenny Farney Fall Plans A lot ofstudying a lot of laughter a lot of fun and some sleepi may e Anything else all ChE s should know SLACKERS Nate Keen Industrial Relations You can remove caffeine from coffee Rich Chiesa Alumni using super critical CO2 Now why Ryan Busch Chem E Car would you ever want decaf That I Alex Schenck Student Faculty don t know Alex Fung Mentor Andy Mills Webpage And the Executive Board Jessica Moore VP External boss of all things IR Alumni amp Philanthropic Jim Anderson Financial Treasurer Year Junior Summer Work Swi Enterprises Ltd Professor most likely to win Jeopardy Beaudoin Collections Noexcept for my David Hasselhoff Year 4th One Summer Work Explosives Research Professor most likely to win Jeopardy Osman collection with a REAL David Hasselhoff Base Zeus statue In my bedroom COHSCHOHS Manx many scrapbOOks What makes you want to dance Dope Beats Zeus statue At anice cream stand on top of Mt Favorite ChE Memory Leaving class Olympus Fall Plans Try not to fail out of ChE What makes you want to dance Tony Danza Tap AnthlHg all ChEgtS should know 1 know who Mike dance Extravaganza or just Ashlee ones is Simpson Chickens have large talons Yes like 50 of em Favorite ChE memory Lacy Mac Donald s near triumph over a Fuch s Orgo exam Fall plans Have a blast Anything else all ChE s should know Have a great semester join AIChE and SLACKERS Adam Slaboski President Kevin Kickham Vice President 7 V I Internal Jason Corah Secretary get to know your classmates They re probably more fun than you d expect Page 7 Y T L U C A F N E U T s A A U L P H 1 L A N T H R 0 P Y R T N 1 N D U s T R 1 A L U E T H E E s T P N 1 D N P w B 1 w T 0 R w U U C R 1 M 0 E L R E A w P F s B C H E U N J Y A B s 0 C 1 A L A T H L E T 1 C P H M 1 N T 1 E A R A T H U E A R E K A R C A T 1 P M R M M G D N Y R s 1 T L T Y P 0 E E E T T 1 w G T A Y H E C T N H P U 0 L A H Y E D C E R H A C N T R R s T F 1 N A N C E 1 NEWSLETTER PHILANTHROPY FINANCE INDUSTRIAL MENTOR SOCIAL ATHLETIC CHEM E CAR ALUIANI PUBLICITY STUDENT FACULTY WEBPAGE C H M E q u i p Each letter stands for another letter in the alphabet Use the clue and your ingenuity to solve the code You know you re a ChE when FZTUFF NF BRZ RBDV HTRL PDQFFUF GJZ HTRL ZMU HRTPU RB QB QTUQ CLUEFS YOUR HOROSCOPES Solution to Word Search pg 7 Aries March 217April Libra Sept 2370ct 19 23 Exciting sign of the day Cheezits make you abandoned bicycle happy Taurus April 207 Scorpio Oct 247Nov May 20 21 When in doubt just Homework does not re smile place sleep Get to bed Gemini May 217June Sagittarius Nov 227 21 Dec 21 You know that dream When the going gets where you get to class tough no sleep for you and you don t have any clothes onyeahthe Capricorn Dec 22 20th Jan 19 Just remember the side Cancer June 227July walk always wins 22 Ifyou nd a Penny Aquarius Jan 207 DON T pick it up Feb 18 Do not collect just pass Leo July 237Aug 22 go Bad luck nds you wher CVef you gO Pisces Feb 197March 20 Virgo Aug 237Sept I m thinkingANOVA 22 The answer is just over Solutlon to CHMEquip pg 7 the rainbow STRESS IS NOT ONLY FROM CLASSES BUT FROM THE FORCE ON AN AREA Hahahaha AIChE C k us https engineering purdueeduaiche Purdue University ME 687 Lecture 18 LaserInduced Polarization Spectroscopy Prof Robert P Lucht Room 86 Mechanical Engineering Building School of Mechanical Engineering Purdue University West Lafayette Indiana Luchtpurdueedu 7654945623 Phone November 6 2008 School of Mechanical Engineering Purdue University Outline of the Lecture Introduction to LaserInduced Polarization Spectroscopy LIPS Theory of LIPS Radiative Coupling of Zeeman States Applications of LIPS for Concentration Measurements School of Mechanical Engineering Purdue University LaserInduced Polarization Spectroscopy LIPS Tunable Laser System Photomultiplier or Photodiode Excited Zeeman States Trans Laser Intensity Ground Zeeman Laser Frequency M0 M1 states 8 I t t Pump LCP peclnesdseelc ive spafia y Probe LCP reso ve WL e range 0 I Probech SpeCIes co erent Signa complicated physics School of Mechanical Engineering Purdue University LIPS Signal Generation Linear polarization can be regarded as a linear superposition of left and rightcircularly polarized light The LCP RCP amplitudes are equal pump 0 MJII l l I LCP 391 LCP RCP 2 Because the circularly polarized pump beam the LCP and RCP components of the probe beam experience different absorptions and phase shifts as they traverse the medium School of Mechanical Engineering Purdue University 1 Polarization States of the Laser Field For rightcircularly polarized light In terms of Sines and cosines the expression above reduces to a A E0 EU I XECOSUfZwl For leftCircularly polarized light jsinkz al School of Mechanical Engineering Purdue University 1 Polarization States of the Laser Field n J For leftcircularly polarized light In terms of Sines and cosines the expression for LCP light is given by 0 coskZ at isinocz wt J5 The sum of LCP and RCP light is given by w 507 t EcosUcZ wt ismUw wt J5 J5 A E0 A E k t 0 X COS Z a yJ Sll l J5 2 3 coskZ at School of Mechanical Engineering kZ wt Purdue University l Polarization States of the Laser Field I Polarization vectors for linearly polarized light 5r x y J5 J5 5r 2 Figure 82 Linear light School of Mechanical Engineering j Purdue University Coupling of the Laser Field with Electric Dipole Transitions The interaction term Veg J is given by Veg l7eg39EL7J leg we lwggt we ei7lwggt eltwe Flwggt e LeveIE EL g Level G School of Mechanical Engineering Purdue University P branchAJ Je Jg 1 Coupling of the zeeman AM Me Mg 1 ag one JeMe m angMggt 1 E R aeJe 7 ltaeJeMe agJe1Me 1 agJe 1 Je Me 1Je Me 2 El39A AM1 J 8 Me 3 agJe1Me 1gt a N MametWag 1 J08 Me 1Je Me 2 5c 49 NIH AMO 058 J8 Me l l ag J81M8gt uRaeJeagJel Je12 Me 2 A Z School of Mechanical Engineering Purdue University Coupling of the Zeeman P brancthJz J6 Jg 1 AMzMe Mgz l 2ng ag ltae JeMe Ztl angMggt aeJeMe agJe1Me1gt RWeJe agJe 1 Je Me 1Je Me 2 2 1y 3 AeG 380 meg NJ J W FJeJe1Je12Je3 8 g Rae e pang For this transition A s 31 Iueg eLCP Iueg J5 0 Ey30 School of Mechanical Engineering Purdue University Coupling of the Zeeman P branchAJ Je Jg 1 AM Me Mg 1 MezMg1 aJMel lagJe1Me 1gt 8 N wage agJe 1 Je Me 1Je Me 2 2 49 NIH AeG 3 80 h A NJ J S z FJeJe 1 J6 12Je 3 8 g 13 8 18 gtang For this transition s A 4 32 1 eg39eRCPeg 5 10 geg2o School of Mechanical Engineering Coupling of the Zeeman Q brcmcthJ 0 AM 1 ltae Je Me 05g Je Me 1 2 waeragJe Je MeJe Me 1 3H9 AM1 aeJeMe l l agJeMe 1 gunaywagnle xJe MeJe Me 1 f i AM o aeJeMe agJeMegt uRaeJeagJeMe 2 School of Mechanical Engineering Purdue University Coupling of the Zeeman R brancthJ 1 AMz l aeJeMe agJe 1Me1gt RaeJeagJe 1 Je MeJe Me 1 xz AM1 aeJeMe p agJe lMe 1gt uRaeJeagJe 1 JeMeJe Me 1x zy AM 2 O 8 18 Meag le 1Megt luRaeJegtagJe1 VJEZ Me 2 School of Mechanical Engineering l Radiative Coupling of the Zeeman States Laser radiation that is circularly polarized in the xy plane couples very efficiently with superposition states with M J value that differ by 1 or 1 circular dipole SM 0 P M 21 states School of Mechanical Engineering Purdue University Radiative Coupling of the Zeeman States Laser radiation that is linearly polarized in the zdirection couples very efficiently with superposition states with the same MJ value linear dlpole SM0 NM 0 states yx 3 t0 t 711 hAE t hAE t T hAE thAE School of Mechanical Engineering 1 Purdue University School of Mechanical Engineering Modeling of the LIPS Process Level3 upper bath 1 I34 lower bath Level 2 Je 1 r3 Me Me 1 0 1 gt rMe gt3 1 4 gtMg 44 Mg 2 1 0 1 2 Level 1 J9 2 Ema 4 Level4 Purdue University TimeDependent Density Matrix Equations for the Laser Interaction Rate of chanqe of population of state k epicCFat i at 22Ikmpmk pkm mG rkpkk 2 1 mkpmm Time development of coherence between states k and i apit i J pkjl60kj 7EZVkmpmj pkamj Couplinq of laser radiation and dipole moment for states k and m Vkm lL ikm IZlkmEprl79tEpump t School of Mechanical Engineering Purdue University Theoretical Approach Numerical Solution of the TimeDependent Density Matrix Equations Timedependent density matrix equations are manipulated for numerical solution rotating wave approximation used Doppler broadening effects included by dividing the state populations into velocity groups Temporal profile of pump and probe lasers are input to code School of Mechanical Engineering Purdue University Different Absorption for Probe Components The linearly polarized probe beam is the sum of equal quantities of left and rightcircularly polarized light Xly x ly 6 A03 J5 co Epmbezl exp iwl ikZA02 The absorption and phase shift terms for the left and rightcircularly polarized components are different because of anistropic pumping by the pump beam 2 1 a A 2 a2 EOh10chl2wz wOch2elzgng ne u gezl School of Mechanical Engineering If Purdue University Experimental Schematic for Picosecond LIPS I shutterI A pump 1 Ehotolysis wedge eam M4 A WI gt 11 v I bd v El po 5 prObe M4 nds M2 9 dIf nds quot nds E pOI pinhoe monochromator M2 PMT Sf h t d d M4 p o 0 IO es mi 111 w T A Reichardt et al J Chem Phys Vol 113 22632269 2000 School of Mechanical Engineering Purdue University PS Experiments with ShortPulse 120 psec Laser System Tunable laser system with 120psec nearFouriertransformlimited pulses developed as facility laser at Sandia Livermore by Farrow and coworkers Short OH PS experiments performed in a gas cell OH produced by UV photodissociation of H202 PS signal dependence on pressure laser intensity investigated School of Mechanical Engineering Purdue University Experimental Results for Picosecond PS of OH in Gas Cell Unsaturated 03 J pump 3 0 Saturated 170 J pump 10 Perturbative for 100ps pulse Perturbative for 10ns pulse quot39DN results 10 100 Cell Pressure torr PS Signal OH 2 arb units School of Mechanical Engineering Purdue University TimeDependence of PS Signal for 100psec Pulses To 1 PSGC rc 1 nsec I 000012 12 E P EH 512IIIIHIHIIH e 7 3 J21 g pump z 1 9 00001 0 Signal V 1O12 1 Q E 1 m E Yc sec 5 9 3 9 1 7 g g 08 e e 810 a 39a 08 y 10 sec39 339 D Q C c Q 4 D E g E I 06 610395 lt 06 m m m lt C g g 9 u U U 75 2 04 410 g 004 19 o E h c Q 5 2 r E 02 210 539 CL02 8quot 3 5 E 3 3 n 0 n U 0 I I I I I I I 0 010 III III I I IIIIIII 0 50 100 150 200 250 300 350 400 0 200 400 600 800 1000 1200 Time psec Time psec Polarization spectroscopy signal is generated mostly after the laser pulses are gone for TL lt TC School of Mechanical Engineering Purdue University TimeDependence of PS Signal for Saturating 100psec Pulses rc1nsec 15quotquotiquotquot39quot39iquotquotquotquotquotquotquotquotiquot81021 j 71039 Signal1000 1e9 Signa 6e11 61039 y 109 sec391 0 Pump Intensity norm 0 U1 1 8 snun 39qJe Ausuelul Ieu ls Sd Time psec School of Mechanical Engineering XML Randy Julian Lily Research Laboratories XML in 10 Points XML is for structuring data XML looks like HTML XML is text but is not meant to be read XML is designed to be verbose XML is a family of technologies XML is new but it has a history amp a heritage XML turns HTML into XHTML XML is modular XML is licensefree platformindependent and well supported XML is a standard maintained by the W3C 0 Q lt9 Q wwww3orgXM Ll1999XM Lin10 points Intro to XML lt9 Markup Encloses parts of a document within tags Structures tags hierarchically Has attributes to specify tag ID attributes allow unique identification amp reference from other elements XML files are fully tagged text files starting from a root tag lttitlegtanalysisltltitlegt ltsamplegt ltnumbergt1ltlnumbergt ltnamegtwaterltlnamegt ltsamplegt ltperson noquot lquotgt Alex ltpersongt ltSpectroML versionquot1 0quotgt ltSpectroMLgt Intro to XML Schema Defines a document type by specifying names and types of elements amp attributes and their number amp order of appearance Na mespace Applies a pre x to a vocabulary to avoid name collision Ensures that an XML document is valid and well formed Processable by a parser when referred to in the document Permits using different vocabularies within an XML document Intro to XML lt62 Stylesheet I XML Source A stylesheet processor transfers one XML I document into another Transformation Allows conversion between data formats and visualization of XML I data XML Destination Example HTML output in browser Permits secure and D39g39tal S39gnature unaltered transmission Signs an XML document and content validation with a unique value based on its content The Structure of an XML Document Q XML documents consist of three parts The prolog The document body The epilog The prolog is optional and provides information about the document itself n The Structure of an XML Document The document body contains the document s content in a hierarchical tree structure The epilog is also optional and contains any final comments or processing instructions The Structure of an XML Document Creating the Prolog Q The prolog consists of four parts in the following order XML declaration Miscellaneous statements or comments Document type declaration Miscellaneous statements or comments This order has to be followed or the parser will generate an error message None of these four parts is required but it is good form to include them The Structure of an XML Document n The XML Declaration The XML declaration is always the first line of code in an XML document It tells the processor what follows is written using XML It can also provide any information about how the parser should interpret the code The complete syntax is ltxm version version number encoding encoding type standalone yes no gt Q A sample declaration might look like this ltxml version 10 encoding UTF 8 standalone yes gt The Structure of an XML Document Inserting Comments lt gt Comments or miscellaneous statements go after the declaration Comments may appear anywhere after the declaration The syntax for comments is lt comment text gt This is the same syntax for HTML comments n Elements and Attributes Elements are the basic building blocks of XML bs XML supports two types of elements Closed elements and empty elements Elements and Attributes Q A closed element has the following syntax ltelementnamegtContentltelementnamegt Example ltArtistgtMiles DavisltArtistgt n Elements and Attributes Element names are case sensitive lt gt Elements can be nested as follows ltCDgtKind of Blue ltTRACKgtSo What 22ltTRACKgt o ltTRACKgtBlue in Green 537ltTRACKgt ltCDgt Elements and Attributes Q Nested elements are called child elements Q Elements must be nested correctly Child elements must be enclosed within their parent elements n Elements and Attributes All elements must be nested within a single document or root element There can be only one root element Q An open or empty element is an element that contains no content They can be used to mark sections of the document for the XML parser Elements and Attributes Q An attribute is a feature or characteristic of an element Attributes are text strings and must be placed in single or double quotes The syntax is ltelementname attribute value gt ltelementnamegt Elements and Attributes Adding elements to the JazzXML File ltxm1 vgrsion l0quot encud ing 39u39rrrsquot standalonalt yesquot 7 07 Thu docunent cnma39ms data an Jazz warehouse specu39l n ers W gt SPECIALS TITLDMunth39ly SEEia39ls at the JA ai almilszltTITLEgt ltCtrgtgtltind nr 5 u mnzmi e DavisltARTIST 39 R x gt50 Nhalt TRACKgt 39 t gtFredc e i eeloaderltTRACKgt A E r gtE39lue in GreenltTRACKgt quotgt 39l BluESltTRACKgt I F lamenltn ikEtCHESltTRACKgt lt 1 nkin39 nswwn39ies Davisltgt PTIST mum y Funny ValentinaTuck TRACK lang E LIES by F39 eltTRACgtltgt mu lan r gtm39r eginltT CK armcx Tang 03quotgtTuna upltTPACKgt lt Dgt ltC ue Tra39ir o39l Trsr ACK lengr x lt ACK ler 5 Nut evmcm 39nn rrmcro a an imagesTRACK t RACK v 9cm ltSPECIALSgt Character References Special characters such as the symbol for the British pound can be inselted into your XML document by using a character reference The syntax is ampcharacter n Character References Character is a entity reference number or name from the ISOIEC character set Character references in XML are the same as in HTML Working with Namespaces Name collision occurs when elements from two or more documents share the same name Q Name collision isn t a problem if you are not concerned with validation The document content only needs to be wellformed However name collision will keep a document from being validated 10 Name Collision gmquot n Mu mm in way wimm m mme uncumanl Using Namespaces to Avoid Name Collision lt00I8 1l0ngt Bank unmaspucu Mlth nnmaspuce n Creating a Valid Document Some elements and attributes may be optional for example an email address An XML document can be validated using either DTDs Document Type Definitions or schemas Schemas A schema is an XML document that defines the content and structure of one or more XML documents 62 To avoid confusion the XML document containing the content is called the instance document It represents a specific instance of the structure defined in the schema 12 Comparing Schemas and DTDs Schema Dialects There is no single schema form Several schema dialects have been developed in the XML language Support for a particular schema depends on the XML parser being used for validation Schema Dialects e m nxscnl lon 4 Starting a Schema File A schema is always placed in a separate XML document that is referenced by the instance document 14 Schema Types child elements XML Schema recognize two categories of element types complex and simple complex type element has one or more attributes or is the parent to one or more simple type element contains only character data and has no attributes Schema Types camplox mm v ltNamegtcynthin mbbsltmamegt ltAqegt 3ltlligegt lt paneit eliiiimiboiial l vi KIWII mm AgcquotSEquotgtCynLna mbthPaucm simian Wm ltpmemgtcyuuua umbwpmen n Simple Type Elements lt Use the followin syntax to declare a simple type element in XML chema ltelement name name type type gt Here name is the name of the element in the instance document and type is the data type of the e ement If a namespace prefix is used with the XML Schema namespace any XML Schema tags must be qualified with the namespace prefix Understanding Data Types XML Schema supports two data types builtin and userderived Q A builtin data type is part of the XML Schema specifications and is available to all XML Schema authors A userderived data type is created by the XML Schema author for speci c data values in the instance document 16 A Understanding Data Types A primitive data type also called a base type is one of 19 fundamental data types not defined in terms of other types 0 A derived data type is a collection of 25 data types that the XML Schema developers created based on the 19 primitive types D WWW D nun 1mm Understanding Data Types 7 mu mi ozscmmn mums Complex Type Elements lt9 The syntax for complex type elements is lt element name namequotgt lt com plexTypegt compositor element declarations compositor attrbute declarations ltcom pleXTypegt lt eementgt n Complex Type Elements Here name is the name of the element in the instance document element declarations are simple type element declarations for each child element compositors define how the list of elements is to be organized and attribute declarations define any of the attributes of the elements Compositors 4 A compositor is a schema tag that defines how the list will be treated Three types of compositors are supported sequence choice and all Q The sequence compositor forces elements to be entered in the same order as indicated in the schema 19 n Compositors The choice compositor allows any one of the items in the list to be used The all compositor allows any of the items to appear in any order Compositors may be nested inside of one another Declaring an Attribute Q Any element that contains an attribute is also a complex type The syntax to declare an attribute is ltattribute name namequot type typequot use use default defautquot fixed xed gt 20 n User Derived Data New data types fall into three categories List a list of values where each list is derived from a base type Union the combination of two or more data types Restriction a limit placed on the facet of a base type Annotating a Schema Q It is helpful to include comments about a created schema for other XML developers An annotation element stores information about the schema The syntax is ltannotationgt ltdocumentationgt documentation comments ltdocumentationgt ltappinfogt application information ltappinfogt ltannotationgt 21 n Summary Q A schema is an XML document that defines the content and structure of one or more XML documents A schema is always placed in a separate XML document that is referenced by the instance document XML Schema recognize two categories of element types complex and simple Summary Q XML Schema supports two data types built in and userderived By attaching a schema to a namespace each part of the combined document can be associated with a different namespace and can draw upon a different schema for validation 22 Purpose Sample Info Peak Data quotPublishedquot Results Systems Biology a a Collection Search Results SpectroM L q r SpectroML Structure SpectroML Dataset sample sample description propert instrument property x39 measurement property instrument description data property 24 SpectroML Elements Instrument Samgle Measurement identifier identifier identifier manufacturer owner titie odei iocati owner casNumber iaboratoryReference iocation fonnuia measurementExecution instmmentAppiication storageMetnod project software disposaiMetnod timestamp version sampiePreparation operator timestamp measurementType operator operator 5 n e instmmentSetting su iier referencesample re oiution re arationnescriptioi mter iinearDispersion sampieAttribute signalNoise spectralBanth dthRange m ie iarWeignt scanNumbers wavden mekngPoit s anDunmon sorb Ra ge boilingPoint measure en e tion detectorType density quaiificationTimestamp Data Values sourceTypes refractiveindex quaiificationReference instmmentParameter p arameter 39 39 slitwdth state p f cyReference spectralSlitVt dth patnLengtn transmittanceTimestamp b mcnan ei amount rans itt ceReference Asmg39e Sputer sampieHoider pressure waveiengthTimestamp sampiePosition temperature wavelengtnReference M39m39p39e 5pm 3quot5peed Muiti dimensionai data pointSeparation Example SpectroML File ltSpectroML ver 10 4 expenmem type 11 Ilt 0quot measurement UVVls quot language enus quot gt e ltmeasuremengt 1legtsample expenmen1legt data clmescampgt dataProperty dataPropertyId quotdpo quot gt daegt200211o7daegt dacapa meter lt1megt101243lt1megt axlsLabel clmes avelengha 1 pach pa Id x w x 5 Po yquotgt39l39ansm1ancelt axlsgt dataPropezyL1nk 90quot dacacozemnk quotch quotgt lt l1egt lnstrument lt1nsxumemgt sample c chquot values 270 576ltvaluesgt sampleProperty samplePropertyIdquotsp0quot ltV1ues dim Yquotgt10 23ltva1uesgt samplepazamece ltdaacoegt saegtllqu1dsaegt ltdaagt amount un1 m1quotgt5ltamoumgt ltexperlmentgt am eParame SpectroML ltsamp1eProperygt samplegt Generalized Analytical Markup Language GAML Represent Analytical Data from Multiple Spectroscopy amp Chromatography Techniques 30gt Compact Simple Dictionary amp Hierarchy Schema gt Use XML Datatypes amp Hierarchical Structure to Mimic Relationships in Data Sources Avoid Parameter quotMappingquot Problem gt Minimize the Need for Complex Dictionaries Permit Future Expansion Q Keep File Sizes Small 26 ltexpelimentgt data flom si gle instrument quotrunquot ltc ll 39 o ectdategt date at ltpararnetergt relevant lttracegt data fl ltco d va ltXdatagt X axis ltva rrne of measurements instrunent parameter orn a single detector or rnatesgt coordinates fol nD data optional uesgt data values array descriptor luesgt data values ar lav ltaltxdatagt alternate x data descriptor optional ltYdatagt Y axis descriptor va uesgt data values array ltpeaktablegt peak list descriptor optional ltpeakgt individual peak descri or ltpeakaaluegt peak location ltpeakaaluegt peak intensi ltbaselinegt baseline descriptor optional ltstarthaluegt baseline values endxvaluegt ltstalthaIuegt ltendealuegt Inclusion of additional standards ltGAMLgt ltexper1mentgt lttracegt pareht hamespace uses thls proposed schema data from a slngle detector ltCML cml XmlnsCMLquotXsSchemazcml 5 13702 szd39 new ltCMLcmlgt ltcoord1nategt ltpeaktablegt 7 che igt hamespace using the quotCMLquot schema data represented here 15 parsed uslng the dlfferent schema but 15 stored 1h the same fllev end of quotCMLquot hamespace remalnder of data 15 1h pareht hamespace 27 An XMLBased Standard for Analytical Result Data 9 SpectroML and GAML serve as starting points for an XMLbased standard interchange format for molecular spectrometry and chromatography Instrument manufacturers data system amp LIMS developers software developers endusers consensus standards organizations regulatory agencies and other interested parties are invited to participate in this effort I INTERNATIONAL Standards Worldwide Creating an XMLBased Analytical Result Data Standard Creating a new XMLBased Analytical Result Data Standard does NOT mean quotstarting overquot The standard should be built on existing ASTM IUPAC instrument vendor and LIMSdeveloper efforts to define common data dictionaries Once the schemas for the new standard are in place straightforward translators can be written to bridge current datasets to the new standard The standard should be developed in a way that makes it extensible to multiple techniques yet avoids duplication of effort and dictionary entries r 28 Characterizing the Influence of Amino Acid Residues on the Charge Inversion of Deprotonated Peptides via Gas Phase lonlon Reactions Joshua F EmoLy39 Scott A McLuckey Purdue University West Lafayette IN Introduction Using ionion reactions for charge inversion eliminates the need for high collision energies and gives product ions with little or no fragmentation Protons are known to localize on basic amino acid residues in protonated peptidesproteins so basic amino acids should be preferred sites for proton transfer during the charge inversion process The ability to differentiate peptides on the basis of amino acid composition using charge inversion could facilitate peptide screening applications Charge inversion enables ionization ofa peptide in one polarity and analysis in the opposite polarity Overview Polypropylenimine diaminobutane DAB dendrimer generations 4 and 5 are used as negative to positive charge inversion reagents to investigate the effect of the presence or absences of basic residues on the charge inversion of peptide anions of YGGFLX X F H K R Charge inversion is expected to occur via a long lived chemical complex2 between the dendrimer and the anion where multiple proton transfers can occur Peptides with basic residues R K H were selectively charge inverted by proton transfer whereas peptides without basic residues were charge inverted predominantly by complex formation Experimental All experiments were performed with a modified version of the MDS Sciex 4000 QTRAP using a research version of MS Expo software version 37 Negatively charged YGGFLX X F H K and R peptide ions were formed by nanospray from an aqueous 1pM peptide solution containing 2 5 ammonium hydroxide Positive DAB dendrimer ions were created by nanospray in aqueous solution containing 2 acetic acid with a dendrimer concentration of 05mgmL E O O D O h D Q Q 2 0 b 0 I 3 D m 1 o 3 1 N2 CAD Gas Aux AC Positive ions Em eng Skimmer D fl t Electrospray Ormce I e ec or Electrospray Needle a Detector Negative Ions 1 Isolate reagent ion during transport through Q1 in route to the 02 collision cell 2 Store reagent ion in 02 and transfer solate desired analyte ion 39nto 02 3 Trap opposite polarity ions in 02 using auxiliary RF and allow ions to react for 500 ms Pulsed Duel Electrospray Ionization Source B7 lt3 M f K DAB7 war 1nn a J m r at J h t ding 5quotquot 7n S 1 C in 5H 5quotquot DAB 7 YGGFLR DAB 5 2 e7 DAB 6 DAB 4 DAB 3 L L d1 1 3w ann 5Bn sun 7nn nun sun 1nnn 11m 12m 13m mun 15n Scheme 1 Reaction of DABn with YGGFLX391 km W RH1 MH 1RnHn MH R2Hltn1gt kdiss1 M RHquot1 complex MRnHltn1gt e k k dISSZ ll cool 1 2 Excited peptide dendrimer complex NH39H RH quot39 MRnHltn1 I I I v 1 Dissociation of the complex Stabilized peptidedendrimer complex Prodl ccejsba pe t39det Chige the complexes observed in the spectra mver e y pro on rams er and a charge reduced dendrimer Scheme 2 Energy Diagram The activation barrier represents the relative energy difference for dissociation of each peptide and dendrimer complex DABquot YGGFLX1n1 XRKFH Energy YGGFLH DABltquot2gt YGGFLFH DABltn2gt YGGFLHH DABltquot2gt YGGFLKH DABltn2gt Reaction Coordinate YGGFLRH DABltquot2gt Reaction of GGFL wit AB Gen 5 9 30e7 DAB9 x 17 DAB YGGFL 8 YGGFL 5e7 DAB 2 YGGFL 7 DAB 2 YGGFL 5 DAB YGGFL 7 DAB7 DAB YGGFL 6 DAB8 DAB 3 YGGFL 6 M M 5 o 600 700 300 900 1000 mix 1100 1200 1300 1400 1500 1600 Reactl 40e7 DAB 7 DAB YGGFLK 6 YGGFLK Not present DAB 5 20e7 DAB 3 D AB 6 DAB 4 A 0 500 600 700 300 900 1000 1100 1200 1300 1400 1500 mix DAB YGGFLF 8 DAB9 DAB YGGFLF 7 DAB 2 YGGFLF 7 90e6 DAB 3 YGGFLF 6 YGGFLF DAB 2 YGGFLF 6 DAB 4 YGGFLF 5 DAB7 DAB 3 YGGFLF 5 DAB8 U hi 6 O 750 850 950 1050 1150 1250 1350 1450 1550 1650 1750 1850 1950 2050 YGGFLH 11e7 m DAB YGGFLH 8 DAB YGGFLH 7 DAB7 DAB 3 YGGFLH 6 DAB 2 YGGFLH 7 DAB 4 YGGFLH 5 DAB 2 YGGFLH 6 DAB 3 YGGFLH 5 DAB8 A L A 6 0 700 B 0 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 mix Summary of YGGFLX 1 Peptides Reactions with DAB Gen 4 7 and Gen 5 9 Dendrimers Peptides without basic amino acids in Red Peptides with basic amino acids in Blue YGGFL YGGFLF YGGFLH YGGFLK YGGFLR Complex Complex Proton Proton Proton Formation Formation TranSfer Transfer Transfer Baseline Baseline Com IeX Little Little Proton Proton Formg on Complex Complex Transfer Transfer Formation Formation Comparison of Two Types of Charge Inversion I A Proton Transfer I A Complex Formtion 100 Generation 4 DAB 80 Percent Proton Transfer versus Percent Complex 40 Formation by Peptide Sequence 2 39 o 60 VGGFL VGGFLF VGGFLH VGGFLK VGGFLR I A Proton Transfer I A Complex Formation 100 Generation 5 DAB 80 Percent Proton Transfer versus Percent Complex 40 Formation by Peptide Sequence 2 39 60 fl VGGFL VGGFLF VGGFLH VGGFLK VGGFLR Basic amino acids are thought to lower the energy barrier for dissociation ofthe peptidedendrimer complex see Scheme 2 to yield the protonated peptide In charge inversion by complex formation the most common product is the 1st generation product ion DABn PeptideWM Second and nth generation product ions represented by DABquot m Peptide11W where x 2 2 and 1 S m S 4 result from DAB dendrimer reactions with multiple peptide ions The number and type of complexes that a peptide forms depends on peptide composition as in the case of YGGFLH the size ofthe dendrimer and the ionion reaction time Peptides with basic residues YGGFLK and YGGFLR charge invert 98 by proton transfer except in the case onGGFLH and peptides without basic residues YGGFL and YGGFLF charge invert 98 100 by complex formation When reacted with DAB Gen 5 YGGFLH charge inverts by 52 proton transfer and 48 complex formation The small difference in proton af nity between Histidine H and Lysine K 236 vs 238 kcalmol does not explain the disparity in observed product ions when YGGLFK and YGGFLH are reacted with DAB Gen 5 It is possible that the structure of the peptide in uences the degree to which a peptide charge inverts by proton transfer or complex formation DAB Gen 4 7 Reaction with MethylEsterified YGGFLK and YGGFLF DAB 7 YGGFLKOCHS DAB 5 DAB 6 1697 DAB 4 DAB 3 J LL I 3297 DAB 7 DAB 5 mz 703 YGGFLFOCH mz 717 1597 DAB 6 DAB 4 I l l 40 50 600 700 800 900 ml 1000 1100 1200 1300 1400 1500 a Nterminal Acetylated YGGFLF DAB 7 a H3CCNHpeptlde DAB YGGFLF 6 30e7 I l r 7 400 500 600 ml 700 300 900 L Nterminal and side chain Acetylated YGGFLK Nterminal Acetylated YGGFLR 0e77 20e7 DAB 7 DAB 7 DAB Doubly Acetylated YGGFLK X 6 DAB 5 2 DAB 6 45e7 10e7 YGGFLR DAB 6 I A l 400 500 600 m 700 300 900 4 o 5 o 600 m 7 0 sec 900 Methyl Esterification and Nterminal Acetylation Proton o on No Proton Proton No Proton Transfer Transfer Transfer Transfer Transfer No Complex No Complex Complex No Complex Complex Formation Formation Formation Formation Formation Methyl Esterification of the Cterminus prevents the peptide from forming a stable complex with the dendrimer suggesting that the Cterminal COO plays an important role in peptidedendrimer complex formation gt o D S m 3 3 o 3 D Z L Q 3 3 c n m 3 Q 2 Q D o 3 9 3 o lt C C I1 I X S m o 3 lt 2 D n 3 D o m 2 o Q Q D c m 3 9 Q 639 0 o 3 o E o 3 S to 3 EL 0 91 SP S S b c m 2 o 6 2 Q r b c lt important role in encouraging the peptidesdendrimer complex to dissociate Conclusions Peptides without basic residues charge invert preferentially by complex formation whereas peptides with basic residues charge invert predominantly by proton transfer Charge inversion occurs via formation of a longlived chemical complex peptides without basic amino acids form more stable complexes than peptides with basic amino acids The Cterminal COO39 group promotes complex formation Inactivating the basic residue in YGGFLK by acetylation causes the peptide to charge invert by proton transfer suggesting that the basic residue is necessary for charge inversion by proton transfer References and Acknowledgments 1 M He JF Emory SA McLuckey Anal Chem 77 2005 31733182 2 JM Wells PA Chrisman SA McLuckey J Am Chem 800 125 2003 72387249 3 Y Xia X Liang SA McLuckey J Am Soc Mass Spectrom 16 2005 17501756 Research sponsored by the Office of Basic Energy Sciences Division of Chemical Sciences under Award No DEFGOZOOER15105 Amy Facility at Purdue University Reprinted from Perspectives on new crops and new uses 1999 J Janick ed ASHS Press Alexandria VA Variability in Oil and Vernolic Acid Contents in the New Vernonia galamensis Collection from East Africa Ali 1 Mohamed Tadesse Mebrahtu and Teklu Andebrhan Epoxy oils are important in industry for the manufacture of plastic formulations protective coatings lubricants and other products Current industrial techniques are expensive generate large amounts of chemi cal waste and produce high Viscosity oil A natural lowViscosity epoxy oil is now available from the seeds of Vernonia galamensis Cass Less a herbaceous member of the sun ower family Asteraceae The low Viscosity and polymerizing characteristics of this oil make it especially valuable as a solvent in industrial coatings and paints for environments where fumes from traditional solvents are hazardous or polluting Kaplan 1989 Some of the products that are being developed from Vernonia oil are degradable lubricants and lubri cant additives epoxy resins adhesives insecticides and insect repellants cropoil concentrates and the for mulation of carriers for slowrelease pesticides The development of alternative crops is receiving increased recognition as an answer to some of the problems facing today s agriculture New industrial crops could significantly diversify American agriculture and create markets that are essentially noncompetitive with existing crops They would also provide a reliable domestic source of essential industrial feedstocks such as unique oils many of which are currently imported Aziz et a1 1984 Cunningham 1987 Kaplan 198939 Perdue 1989 Perdue et a1 1986 Thompson et a1 1994abc Establishment of a new industrial crop such as Vernonia can be an answer to problems facing farmers today who need a high cash crop as a primary source of income This is crucial in states where farms consist of relatively small acreages and who are dependent upon a single cash crop Vernonia galamensis is an annual herb and native of Africa Perdue et a1 1989 It grows in areas with as little as 20 cm of seasonal rainfall Plantings in Virginia Arizona and other states showed that Vernonia is extremely resistant to insects and diseases Vernonia seeds contain up to 40 epoxy oil and this oil has up to 80 vernolic acid cis1213epoxyoleic acid Plantings in Eritrea Kenya and Zimbabwe confirmed that V galamensis has an excellent seed retention compared to V althemantica 30 35 A 25 3 30 Ex 20 Mean 873 kgha 5 4 25 Mean 340 grams 5 5 C C 20 g 15 g g E 15 LE 10 LL U I Ln 8 n 0 0 0 0 I I I e e 9 e L L L L L Y 00 79 Q 9Q 7617 63 7 3 639 QL 01 9 76 6 9amp7 Q7 by a 0 3976 630 629 30 039 9 00 3 77 396 egg 79 x97 7 Yield ranges kgha X103 Seed sIze ranges 9103 seeds Fig 1 Frequency of seed yield kgha X 103 distribu Fig 2 Frequency of seed size grams103 seeds tion in Vernonia galamensis accessions collected in distribution in Vernonia galamensis accessions col Eritrea lected in Eritrea The authors are grateful for HBCUU SAID for funding the project and the collaboration of the research and field staff of the Ministry of Agriculture The State of Eritrea The authors also extend their appreciation to Drs David Dierig and Terry Coffelt of the Water Conservation Lab ARSUSDA for their advice and support 272 45 40 Mean 24 3 30 3 35 i 25 30 a Mean 624 g 20 g 15 g 9 I 15 L39 10 10 5 5 0 7 7 9 e 0 L9 6 K70 9 0 ea 96 91 eye 79 7 a em 046 51 5 Oil content ranges VernOIiC 301d ranges Fig 3 Frequency of oil content distribution in Fig 4 Frequency of vemolic acid distribution Vernonia galamensis accessions collected in Eritrea in Vernonia galamensis accessions collected in Eritrea Despite the highly successful progress in domestication of Ver 22 nonia Thompson et al l994abc substantial research is needed 20 to evaluate Vernonia accessions to make new selections for vemolic acid quantity and quality and to determine the effect of environ 18 ment cultural practices and processing on Vernonia oil The main 5 objectives of this research were to determine seed yield and yield 3 14 7 components of the newly collected accessions and to determine oil E 12 7 content and fatty acid pattern 9ch 10 C O GERMPLASM COLLECTION O 8 The existing V galamensis germplasm collection at ARS 67 USDA is limited to 63 accessions During the rst year of the 47 El project new accessions were collected from Eritrea and Ethiopia through a USAIDHBCU grant Collection was done by selecting 2 individual matured inflorescences and planted for seed multiplica quot tion in Eritrea A total of 61 accessions was collected and planted Palmmc Steam Ole Linolem each in a single row for seed multiplication at Halhale research sta Fatty adds tion in Eritrea Accessions with adequate amount of seeds includ Fig 5 Variations in palinitic steaiic ing breeding lines received from the Water Conservation Lab at oleic and linoleic in Vernoniagalamensis ARSUSDA Phoenix AZ were planted in a fourrow plot at the accessions collected in Eritrea same location At maturity each accession was evaluated for seed yield agronomic traits seed oil content and fatty acid pattern Oil and vemolic acid were analyzed using the methods of Ayorinde et al 1990 and Mohamed et al l995ab EVALUATION During the rainy season of 1996 in Eritrea the 61 accessions evaluated for agronomic and chemical parameters demonstrated the existence of wide genetic variability that could give the possibilities for genetic improvement of the crop Signi cant differences for seed yield oil content and vemolic acid were observed among the 61 accessions evaluated The mean seed yield was 873 kgha and ranged from 60 to 2800 kgha Fig 1 The variation in yield was also re ected in seed size and the mean size was 34 grams103 seeds Fig 2 The mean of the total oil was 24 and ranged from 14 to 31 Fig 3 where the majority 46 of the accessions fall within the overall mean The vernolic acid mean of the accessions was 62 and ranged from 38 to 77 Fig 4 and 49 of the accessions had vemolic acid content which exceeded the mean The fatty acid profile of the accessions is given in Fig 5 Emphasis will be given to the accessions that gave 273 highest seed yield vemolic acid and oil A positive correlation r 028 between oil percentage and vernolic acid was found This indicates that breeding Vernon ia for higher oil content will increase vemolic acid percentage A highly signi cant and negative correlation r 7090 4182 4195 and 4196 were found between vemolic acid and palmitic stearic oleic and linoleic acid respectively REFERENCES Ayorinde FO KD Carlson RP Palvic and J Mcvety 1990 Pilot plant extraction of oil from Vernonia galamensis seed J Am Oil Chem Soc 675127519 Aziz R SA Khan and AW Sabin 1984 Experimental cultivation of Vernonia paucz oraia rich source ofvernolic acid Pakistan J Sci Ind Res 272157219 Cunningham I 1987 Zimbabwe and US develop vemonia as a potentially valuable new industrial crop Diversity 1018719 Kaplan KC 1989 Vernonia new industrial oil crop Agr Res 374 10711 Mohamed AI HL Bhardwaj C Paul and AB Thompson 1995a Vemonia lipase and its inactivation by microwave heating Paper presented at 86 American Oil Chemist Society Annual Meeting May 7711 San Antonio TX Mohamed AI HL Bhardwaj AA Hamama and C Webber 1995b Chemical composition of kenaf oil Indust Crops Prod 41577165 Perdue RE Jr 1989 Vernonia galamensis a new industrial oil seed crop for the semiarid tropics and sub tropics US Department of Agriculture Mimeo Perdue RE Jr KD Carlson and MG Gilfert 1986 Vernonia galamensis potential new crop source of epoxy acid Econ Bot 405L68 Thompson AE DA Dierig ER Johnson GH Dahlquit and R Kleiman 1994a Germplasm develop ment of Vernonia galamensis as a new industrial oilseed crop Indust Crops Prod 2 1857200 Thompson AE DA Dierig and R Kleiman 1994b Variation in Vernonia galamensis owering charac teristics seed oil and vernolic acid contents Indust Crops Prod 2 1757183 Thompson AE DA Dierig and R Kleiman 1994c Characterization of Vernonia galamensis germplasm for seed oil content fatty acid composition seed weight and chromosome number Indust Crops Prod 2299305 274 Purdue University ME 687 Lecture 17 TwoPhotonExcited Fluorescence TPEF Prof Robert P Lucht Room 86 Mechanical Engineering Building School of Mechanical Engineering Purdue University West Lafayette Indiana Luchtpurdueedu 7654945623 Phone October 23 2008 School of Mechanical Engineering Purdue University Outline of the Lecture Applications of TPEF Theory Brief of TwoPhoton Excited Fluorescence TPEF School of Mechanical Engineering I Purdue University TPEF Measurements of the HAtom Problem First excited n 2 electronic level for H lies at 82200 cm1 above ground n 1 level Onephoton excitation at 1216 nm is impossible in flames because of VUV absorption LlF cannot be observed except in nearvacuum conditions Hatom energy level diagram SEER shc Number Cm 1 c 10973 n4 102824 I n3 97492 E E 1656 nm W n2 82259 g 8 n1 0 School of Mechanical Engineering Purdue University TPEF Measurements of the HAtom Osvich Trigger PM m as ADC SamIN Pumped Micro t Oscilloscope Dye Lo Compute System vm W W 50 275 um Sputum V PainBron A 50W in F I dl RC beta 39 I m sm m A 205 an Fomm Lens 9 Faun W Van Flat Flam 3mm mum m lung Rotating quotinquot quot0 Amway Photodioch J Scanning Plot Salmon and Laurendeau Combust Flame Vol 74 221 231 1988 7 School of Mechanical Engineering Purdue University I TPEF Measurements of the HAtom h scc Io I900 l39l Hoo 7 5 2 5 39 g m 9 Isoo x A a Fluorescence a t IZOO a 39 6 uPurtIalEuull g 4 E I E 2 I400 lt B Iooo g 3 4 g K It lg r E 1200 5 2 2 0 fluovuunu I mo 5 z z o quotnotquotqu I z I Ferlill Evil I000 o I I I 6w 0 l I I I 0 I0 20 30 40 50 I0 20 30 40 50 HEIGHT ABOVE IURNER Itquot HEIGHT ABOVE BURNElel Flg 4 Vertical number density pro le of atomic hydrogen In Fig 5 Vertical number density pro le of atomic hydmgen In the IITorr CIHJOJAI lllme at 0 Lo Alno shown In the the 72Tnn39 CHIOIAr ame II t U Also shown are the lame Icnlpenture and on number detulty pro les lllne temperature and on number density pro les Salmon and Laurendeau Combust Flame Vol 74 221 231 1988 School of Mechanical Engineering Purdue University TPEF Measurements of the HAtom School of Mechanical Engineering 39f n4 f M3 656 486 nm nm 656 quotm 656 nm quot2 292 nm 205 nm l 4 243 quotm FIG 1 Multiphotonexcited uorescence detec tion schemes used in this study for measuring atomic hydrogen concentration pro les in ames The up 292 nm ward arrows indicate excitation wavelengths and 4 the downward arrows indicate uorescence detec tion wavelengths I A m 205 nm 243 nm 292 nm Goldsmith 22nol Combustion m1 Symposium pp 1403 Two Three Two 1411 1988 Photon Photon Step Purdue University f I TPEF Measurements of the HAtom h 3 P A2 656 quotm Armor 2 S A 243 nm I 243 nm 2 666 nm A39 243 nm 15 SPECTROMETER sur Goldsmith Optics Letters Vol 10 116118 1985 1 School of Mechanical Engineering Purdue University I TPEF Measurements of the HAtom i In 3 5 I M E quot5 1 oil A e g 49 g Q 0 egg u a a 04 a casual 1 a 6636 9 2 a E 3 Bleh eae x an m a a o w c I 8 n a H l 8 9 u 0 us quotSEE V a o ID 10 so no so no 10 to V a 39 1 Iquot 339 IO 5 39I I l lElGHT ABOVE BURNER mm HEIGHT ABOVE BURNER lmm FIG 7 Relative atomic hydrogen concentration pro les in ame F 25 Torr O 20 acetylene oxygen argon ame measured using twophoton excitation triangles threephoton excitation squares and twostep excitation circles The pro le in the inset diamonds plotted to the same scale as the main part of the ilgure represents an inter ference observed using twophoton excitation with the 205mm wavelength detuned slightly from the atomic hydrogen resonance Goldsmith 22nd Combustion Symposium pp 14031411 1988 FIG 2 Atomic hydrogen concentration pro les in ame A 72Torr 1 06 hydrogenoxygenargon llame Symbols Relative prolllel measured using twophoton excitation triangles threephoton ex citation squares and twostep excitation circles Solid curve Absolute prollle calculated using the ame model School of Mechanical Engineering Purdue Universit F V TPEF Measurements of the HAtom in DiamondForming Flames 15 mm Molybdenum Substrate Probe Volume Stokes Beam 7 Pump Beams f CARS Beam Reaction Zone Wi hieldFlllggrogen Premixed Acetylene Oxygen and Hydrogen H2 CARS measurements performed previoust in Bertagnom 9t 339 J the diamondforming ames Appl Phys Vol 83 23152326 1998 School of Mechanical Engineering TPEF Measurements of the HAtom in f Purdue University DiamondForming Flames From L EELSJ39JE 9995 Delay Generator 7 Unlblllz Pm Shutter I v Dlgllal Oscilllscope Lumonlcs Narrowband Hh 39glyalugla b NdYAG pumped narrowband dye laser generates 584 nm light KDP crystal doubles dye laser output to 292 nm Spectrometer used as a narrowband filter Benagnom et al LIF signal detected by a photomultiplier tube sicfrig Highspeed shutter prevents PMT saturation 2326 1998 School of Mechanical Engineering Purdue Universit F V TPEF Measurements of the HAtom in DiamondForming Flames 7 I Bertagnolli et al J Appl Phys Vol 83 23152326 1998 School of Mechanical Engineering a as 656 556 556 486 656 30 2921 ii 205w 243 303T n 29 H 205 308 243 292 n 1 Purdue University TPEF Measurements of the HAtom Calibration in NearAdiabatic Hencken Flame Scatter due to flame instability and possibly some i photochemical dissociation of water vapor School of Mechanical Engineering 2 9 03901 g 39 39 Ealibration 3 I P I t l A 2 o 0001 E E c 0 a 2 Measured LIF Signal u 00001 1 gt39 E Adlabatlc o Equlllbrlum E 3 10395 39 39 39 39 39 39 lt 0 05 1 15 2 25 a 35 Equivalence Ratio 0 Calibration signal measured at I 126 Benagnom et 9 al J Appl Calibratlon pomt uncertainty of 36 A Phys Vol 83 23152326 1998 J Purdue Universit E y TPEF Measurements of the HAtom in DiamondForming Flames Substrate Temperature 1040 K Exit Velocity 40 mls c Substrate A o x 393 01 Ac39li39ii iiquotquot39 E Equl lilb lutin II Concentration Bertagnolli et al 5 03901 39 Threephoton 395 J39 Appl39 Phys excitation LIF VOI39 8339 2315 g 2326 1998 o 3 ii I g 0001 Avegggfng l 3 Region 0 E computed Pro le 3 Meg39lggsw gal 1993 V lt0390001 1 02 08 12 14 16 Distance from Substrate mm Base case measurement nearest to model conditions I J School of Mechanical Engineering Purdue University Photoionization LossControlled Spectroscopy PICLS Hatom Measurements 41 U I Continuum 3 w3i 550 Salmon and A n 3 Laurendeau 032 A32 6536 Appl Opt Vol 2 V 39 n g 2 26 28812891 1987 2051 School of Mechanical Engineering Purdue University Photoionization LossControlled Spectroscopy PICLS Hatom Measurements Salmon and Laurendeau Appl Opt Vol 26 28812891 1987 7 School of Mechanical Engineering HR Dye Nd YAG Laser Laser OfSIrltch Trlgger 5504275 14500 11500 W H39B Prrstns Delay Line 39 LISOO 4 BD Beam Dump DM Dielectric Mirror HB Harmonic Beamsplltter L Lens focal length in mm MA Mirror Assembly Image Rm m39w Vessel PD Phatodiode PM Mquot Metquot PMTPhotarnultlpller RT Reducing Telescope SP Scattering Plate V0 Voltage Offset Control llin Broao Pressure quot Sampling 39 Oscilloscope I nd Toon I Gwenquot I l I39vbl vm V2 To Computer 777 Purdue University Photoionization LossControlled Spectroscopy PICLS Hatom Measurements 6 200 Salmon and A 5 Laurendeau ft Appl Opt Vol 5 30 26 28812891 e 4 1987 s 80 2 8 3 2 600 2 D 2 8 o E I V 400 392 I 0 Conventional 200 J 0 IO 20 3O 4O 39 50 Height Above Burner mm School of Mechanical Engineering i Purdue University Stimulated Emission from TwoPhoton Excited Atomic Hydrogen BVOL61979 19851989 Dana lonize 205nm 13 Fig 1 Energy levels of atomic hydrogen not to scale relevant to 205nm twophotonexcited FL and SE School of Mechanical Engineering Purdue University Stimulated Emission from TwoPhoton Excited Atomic Hydrogen 3 mod 4 6 o a A a b 4 3 6156 A y A A Z a g A e 0 39quot A AA 3 I quot A z A g A u A an A too zoo 60 0 soo soo PULSE ENERGY LlJ F13 4 Intensity dependence of twophotomede tunichy drogen SE triangles Ind FL circles recorded simultaneously in a loan 1ch ratio 06 72Ton39 hydrogenoxng ame Thednhed neisdnwnwithaslopeofz Goldsmith JOSA B Vol 6 19791985 1989 School of Mechanical Engineering Purdue University 10 29 3 0 39 3 3quot 869 nm Schemes for so 523 s 223 Detection of 335 Atomic Oxygen and s so Atomic Nitrogen g mquot I 40 226 nm 211 nrn 20 Bischel et al 0 2o 3 7 2 5 Appl Opt VOI OXYGEN NITROGEN 211419 1429 1982 Fig 1 States and wavelengths involved in the detection scheme Fine state splittings are too small to show on this scale School of Mechanical Engineering Appl or 2a 1419 Schoo wsvel engchs 39I he in f Purdue University TPEF Scheme for Detection of CO B39z I 39 I vquot 22301 nm 4 A quot I Alden et al I Appl Phys B Vol 33 205 l 208 1984 l x z I v 0 Fig 2 A schematic energylevel diagram showing the two pholon absorption and uorescence wavelengths 39 1 School of Mechanical Engineering f Purdue University TPEF Scheme for Detection of CO i a 1mm I1 mm E A Alden et al Appl Phys B Vol 33205 208 1984 E 9 mm I 13 mm E t I mm 3 mm Fig I The spllial diurihuliom ul scullercd M lighl imensities from C0 molecules at dill39er cm height Iboyc Ihc burner in u L Huir k le Di W MIN 4 ame School of Mechanical Engineering Purdue University TPEF Signal Levels School of Mechanical Engineering 4 4 C ch A v 28 b 03 i Hba0 Hac on i v a Schematic diagram of the threestate system and the twophoton absorption process for the case Where the energy of state c is much higher than the energy of state b R P Lucht S Roy and J R Gord Nonperturbative Modeling of Two Photon Absorption in a ThreeState System Journal of Chemical Physics 121 98209829 2004 Purdue University TPEF Signal Levels y 1213 w m W 3 adj 39 who a Aia 7222a 112Iszal 7525071512 013 w quota web F who a Aia 7222a 112132m I 68 h22Fb7baacba2 0 sat A2 A2 Iucbe Iuace Lucht et al J Chem Phys 2004 School of Mechanical Engineering Purdue University TPEF Signal Levels where mca 8c 8ah wba 2 5b 5a me 80 817h Abel wba 2a Pb rate coefficient for population transfer from state b yab dephasing rate coefficient for state b Lucht et al J Chem Phys 2004 School of Mechanical Engineering Unsupervised Learning Clustering Randy Julian Lily Research Laboratories Various approaches Parameteric I Parameteric I E Density Based Approaches Geometric Approach Clusters All shapes and sizes 1 I 39 39 l39u 39 3953 qCluster Analysis Methods Q Hierarchical Agglomerative bottomup o Start at bottom and merge a selected pair of clusters into a single cluster Pair to merge is selected as having smallest intergroup dissimilarity Divisive topdown Start at the top and split one cluster into two new clusters b Split is chosen to produce two new groups with largest betweengroup dissimilarity HCA Hierarchical Clustering 4 5 a N k w x4 x5 x6 x7 x3 III II III kw gtgtgtr gt Ii Ii Dow LIIk LNN N Ii Ii II n Dissimiarity Metrics Construct pairwise dissimilarities between observations F F DxxxxiZwdxyxxi 2w 1 11 11 2 dxy xxij xv xxij similarin scale n880 Dimensional Data IR of FA s 8 8 O MO 8 g 839 3 m 8 8 o 500 1000 1500 2000 2500 3000 3500 4000 1cm I Compared to IR from PAH s 8 8 o O o a 3 m 8 8 o 500 1000 1500 2000 2500 3000 3500 4000 1cm n E 3 fx T m g M y T 39 g g y x I I El IUD WEI BUD BUD Z c H a y f x 4 I g y f x a Am J S 39 r I 2 y f x Q I I I I I Z H a y f x v 1 absorb 8000 4000 0 1000 1500 2000 2500 3000 3500 1cm 4000 absorb 8000 I I 4000 0 I I I I I I I I 1000 1500 2000 2500 3000 3500 1cm 4000 Distance pa pah2 2702315 pahB 2568599 pah4 2486033 pah5 2760584 pah6 2845282 fa1 4426444 fa2 4313335 fa3 4471794 fa4 4531017 fa5 4537492 fa6 4823245 fa7 4454473 fa8 4150464 fa9 3887052 fa10 3454672 Matrix dist pahZ pah3 pah4 pahS pah6 1971036 2494317 1818368 2347877 2119962 2413258 2754771 2797613 2823649 3053637 4541467 4620404 4351926 4893411 4695286 4450498 4502327 4225091 4806470 4594168 4598107 4652637 4398755 4954208 4744459 4666401 4707196 4468385 5015572 4805539 4683965 4717447 4492176 5006274 4818537 4978743 4998115 4792104 5280891 5090600 4628315 4639072 4386058 4973228 4744904 4340210 4340254 4080935 4690219 4455597 4130974 4095681 3778830 4515098 4226529 3712111 3671306 3360970 4083236 3820474 fa10 fa1 771836 1271151 6030209 1739206 10959837 2166988 15835046 2625921 20632722 2687712 21420239 2821763 22991564 2941073 24437826 3360762 29550562 fa2 nDistances between FA s fa3 fa4 5506861 10797357 15285311 16708467 18919696 21179732 27283120 6588852 10337937 12226365 14971922 17623957 24656561 fa5 fa6 6678778 9002815 696594 11467954 1047253 14914461 1434551 21620286 2185591 Hewght 30000 40000 20000 10000 Tam Cluster Dendrogram 1 515pe rage d1 0 hc1ust ave pahS pahA I R code for hclust quotV librarymva plltereadtablequotpahleirtxtout ElltereadtablequotfaleirtxtoutU namesplltec wn a namesflltec wn quot3 specltedata frame plSap2Sap35ap4ap5ap6ar flSa f2af3af453f5af6af7a EBSa f9aflOSa names spec lt7c quotpahlquot quotpah2quot quotpah3 quotpah4quot quotpahS quotpah6quot Hfaln I Hfa2n I quotfa3 uI quotfa4uI quotfagquot I quotfagquot I quotfa7uI HfagnI quotfagquot I Hfalon tlt7tl l havhclust i d t Speck i v quotavequotl plothc Aggomerative Methods Q Single Linkage Takes the intergroup dissimilarity as the closest least dissimilar pair lt9gt Complete Linkage Takes the intergroup dissimilarity as the furthest most dissimilar pair Q Average Linkage Average dissimilarity between groups Compromise between Single and Complete Cluster Dendrogram Cluster Dendrogram I 40000 50000 Height 10000 20000 30000 Height 10000 15000 20000 25000 30000 5000 O isl5pecl d15l5pec t ompiele hciust a singie d hciu5l Kmeans clustering Iterative descent cluster method Assumes squaredEuclidian distance is used as the dissimilarity measure N observations are assigned to the K clusters in such a way that within each cluster the average dissimilarity of observations from the cluster mean is minimized lt lt0gt Cluster assignment C K 2 c mmz Zia x C k1Cik m means of current xs 2 argmmz quot ieS clusters 2 lxx ml 5 quot 1 TV 3 I u 35 l34 5 Method 1 For a given cluster assignment C the total cluster variance is minimized with respect to the means of the currently assigned clusters Given a current set of means total cluster variance is minimized by assigning each observation to the closest current cluster mean 3 Repeat 12 until assignments in 2 don t change N 10 Stochastic HillClimbing x L lmlim CeankJs Iniual Penman Maraliun Number 20 Guess k2 from dendogram 27 10000 15000 I I FCZ 5000 I 1 3L 40000 I 45000 I I I I I I 720000 710000 0 10000 20000 30000 R code for PCAKmeans example librarymva librarycluster spect lt7 ts ec hclt7hclust dist spect quotavequot 1lt gt1Sp5quot 3901 spec C I mean dimnames initial ltflist NULLdimnames spect 2 k 7 kmeansispecrt initial specpcalt eprincompspect specpxltepredictspecpca dimnames kmScenterS 2 lt7dimnamesspect 2 speccentersltepredict specpcakmcenters plot specpx 12 type n xlab PCl ylab PC2 text spec px 12 cexl labelskmcluster points spec centers lz2 pch3 cex3 12 Wrong Guess at K cutreehc3 4 snnn lEIEIDD 15000 I I l E 2471 2 1 sum 1 Dunn 45000 I I I I I I I I rznnun rlEIEIEIEI u lEIEIEIEI 2mm anunn Practical Issues for k means Must select a number for K Q Must select an initialization Can use withincluster dissimilarity or Gap 2 2 a A 2 6 6 Number uf Clusters Number of Clusters g Multidimensional Scaling Attempt to find a lower dimensional approximation of the data so as to preserve the pairwise distances as well as possible Classical Minimize stress functionquot x1x2x3xN 6 EXP Observations in pdimensions k Transform to kdimensions 212223ZN 6 5 12 SDzlzzz3zNZd TY i 392 Z I I by minimizing SD JDS is R librarymva spect lt tspec loclt cmdscaledistspectt Xlt loc ylt loc2 plotxytypequotnquot mainquotClassical Multidimensional Scalingquot textxynamesspec cex08 14 Classical M ultidimensional Scaling IaI o o c m faZ o o 8 faS o o o m fa4 gt pah5pah2 o pa as o O 839 as fa7 fa8 tag 0 o Ian o g I I I I I30000 20000 10000 0 10000 20000 A look ahead IR data from NIST is in JCAMPDX format ach TYPEINFRARED SPECTRUM Research Labs Under US EPA Contract Standard Reference Data program REGISTRY NO10779276 000052179 0 033238 Y 4500 637 63B 621 624 63B 665 702 723 741 798 15 nJCAMP Data section XYDATAXYY 4500 637 638 621 624 638 665 702 723 741 798 4900 839 862 888 935 928 967 1021 1000 996 1098 5300 1133 1222 1328 1382 1488 1531 1571 1558 1559 1570 END Need a xy pair for use in R Per to the rescue just a fragment of the code line ltSRCFILEgt while 1ine l END tokens split 1ine for index 0 index lt tokens index if index X tokensindex else printf OUTFILE quot 4 lft5 lfnquot Xindexelde1taxtokensindex line ltSRCFILEgt 16 Structures MOL FILES SMILES SI 5 03230321232D 00000 000000 0000000 000000 000000 000000 000000 000000 000000 000000 SMILES CCCCOO 000000 000000 000000 17 TEGRATION OF CHRONOSTRATIGRAPHIC DATABASES FOR THE 21ST CENTURY SUMMARIES 0F NSFisPONSORED WORKSHOP AMHERST MA NOV 3710 2001 CHRONOS STEERJNG COMMITTEE RECOMMENDATIONS FOR SeYEAR PROGRAM Cnntents 1 Exeeuuve Summary 11 Mrssror Statement armor Goals and Applreauous 111 Summary of Chmnax Database System IV Chrouostraugrapme Data Is suese Types Staudards1uterpretatrorr Compaubruty Data Usage VUVh Utw r quot TH wdTvawnwth VI 5eYearP1arr audBudget VII Appeudrees quotParticipants Workshop 1ssuesAgerrda and Se1eeted Current s I Executive Summary Workshop Goals and Themes olWorksliop A r i H 74 Assembling integrating and distributing datarelewntto geologic time processes foi social bene t mee major 39 39 39 39 39 39 39 n lur m Coordination ofdatabase standards and compatibility of global and regional stratigraphic data mnemly disuibuted depositories A concepts Major Decisions and Recommendau39ons 1 39 i in anal 1 quot2 mail a r39 r 39 lmi n r biodiversity
Are you sure you want to buy this material for
You're already Subscribed!
Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'