Themethodofmovingcurvesandmovingsurfacesisanew,eectivetoolfor Abstract



Similar documents
1Introduction. identicallywhenthesurfacehasbasepoints{thatis,parametervalues(s0;t0)forwhich

Examples of Physical Quantities



Forecasting Methods. What is forecasting? Why is forecasting important? How can we evaluate a future demand? How do we make mistakes?



Color Code Drug Doses L.A. County Kids


Amadeus Virtual MCO User Guide

Probabilistic user behavior models in online stores for recommender systems

Attention windows of second level fixations. Input image. Attention window of first level fixation

ATT-72/94, SALT SIEVE ANALYSIS. This method describes the procedure for determining the gradation of salt.

Standards. Interactive Media, July 2012, Page 1 of 6


Programs Implementing Management System Elements AT&T Environment, Health and Safety Management System ISO EMS Element 4.1General 4.

Australian Santa Gertrudis Selection Indexes

Tuition Reimbursement Program. Handbook

o Ivy Tech CONT 101 Introduction to Construction CONT 106 Construction Blueprint Reading BCOT 104 Floor and Wall Layout

o Ivy Tech DESN 105- Architectural Design I DESN 113- Intermediate CAD o Vincennes University ARCH 221- Advanced Architectural Software Applications

Power Rating Simulation of the new QNS connector generation

Figure2:Themixtureoffactoranalysisgenerativemodel. j;j z

VALUATION OF DEBT CONTRACTS AND THEIR PRICE VOLATILITY CHARACTERISTICS QUESTIONS See answers below

Fundamentals of Electromagnetic Fields and Waves: I

SUBJECT: Army Information Technology Service Management (ITSM) Policy

A mixture model for random graphs

Computer Tech Support, July 2012, Page 1 of 5

Fall Protection Safe Work Plan Fall Distance of 7.5M or More

Fire Science One-Year Certificate

Wealth Management Formula

Scalars, Vectors and Tensors

Study Kit No 9. Aura Lee (Love Me Tender)

Curriculum Vitae. Wayne Loucas. Education University of South Florida, Tampa, FL Master of Fine Arts in Photography

Matrix Differentiation

Thispapersurveysthecontributionsofvemathematicians Eugenio Beltrami(1835{1899),CamilleJordan(1838{1921),JamesJoseph

Chords and More Chords for DGdg Tenor Banjo By Mirek Patek

AMATH 352 Lecture 3 MATLAB Tutorial Starting MATLAB Entering Variables

... Schema Integration

DIGITAL ALARM CLOCK RADIO

Approved BLS/CPR and ACLS Education for EMS Provider Certification November 2013

Foundation Course. Study Kit No 1. Away In A Manger


CBE 6333, R. Levicky 1. Tensor Notation.

Option Pricing. Chapter 12 - Local volatility models - Stefan Ankirchner. University of Bonn. last update: 13th January 2014

Technology - A Brief Summary

The Heat Equation. Lectures INF2320 p. 1/88

DIPLOMA IN SECURITY MANAGEMENT. Term-End Examination December, 2011 BSEI-007 : TRAIN THE TRAINERS

How To Determine Chords for Mountain Dulcimer Playing (How The Person Who Wrote the Chord Book Figured It Out)

Functions. MATH 160, Precalculus. J. Robert Buchanan. Fall Department of Mathematics. J. Robert Buchanan Functions

You Could Have Invented Spectral Sequences

How To Improvise a Solo A Workshop for Beginners

The Secret Weapon for Bar Chords

Lesson 3. Numerical Integration

Coloured stained glass & bevelled designs

Amplifier for Small Magnetic and Electric Wideband Receiving Antennas (model AAA-1B)

Energy Management System based on the PN-EN ISO implementation in Dzierzoniów

b 9 œ nœ j œ œ œ œ œ J œ j n œ Œ Œ & b b b b b c ÿ œ j œ œ œ Ó œ. & b b b b b œ bœ œ œ œ œ œbœ SANCTICITY

Practical Approaches to Principal Component Analysis in the Presence of Missing Values

b.a.b-technologie gmbh Control W Dokumentation

thek-aryn-cubestructure. 1

Phasors. Phasors. by Prof. Dr. Osman SEVAİOĞLU Electrical and Electronics Engineering Department. ^ V cos (wt + θ) ^ V sin (wt + θ)

Navigating Family Medicine. Urban Residency Training Program Department of Family Medicine University of Calgary

Basic Guitar Chords. By Tomas Michaud

EVOLUTION OF NETWORK AND SERVICE MANAGEMENT

Recognition Based CPIT(Service Management) Peter Miao, Chairman of itsmf (HK Chapter)

Review Questions 1. Why is it important to arrange records by location?

Practical Application Fly Line Recommendations Freshwater

Thrashing: Its causes and prevention

Transcription:

OnaRelationshipbetweentheMovingLineand MovingConicCoecientMatrices DepartmentofComputerScience Houston,Texas77005 mzhang@cs.rice.edu RiceUniversity MingZhang DepartmentofInformationSystemsandComputerScience NationalUniversityofSingapore KentRidge,Singapore119260 chionhew@iscs.nus.edu.sg Eng-WeeChionh DepartmentofComputerScience Houston,Texas77005 RonaldN.Goldman rng@cs.rice.edu RiceUniversity tweenthemovinglinecoecientmatrixandthemovingconiccoecientmatrixfor implicitizingrationalcurvesandsurfaces.hereweinvestigatearelationshipbe- Themethodofmovingcurvesandmovingsurfacesisanew,eectivetoolfor Abstract Keywords:Implicitization;MovingLine;MovingConic rationalcurves.basedonthisrelationship,wepresentanewproofthatthemethod ofmovingconicsalwaysproducestheimplicitequationofarationalcurvewhen therearenolowdegreemovinglinesthatfollowthecurve.

1Introduction Foreachrationalcurve wherex(t);y(t);w(t)aredegreenpolynomialsintandgcd(x(t);y(t);w(t))=1,there existsauniqueirreducibledegreenpolynomialf(x;y)suchthatf(x;y)=0represents thesamecurveas(1).theequationsx=x(t)=w(t),y=y(t)=w(t)arecalledthe parametricformofthecurve,whereastheequationf(x;y)=0iscalledtheimplicit formofthesamecurve.surfacestoohaveparametricandimplicitforms.parametric w(t); w(t); representationsareconvenientforrenderingcurvesandsurfaces,whiletheimplicitforms areusefulforcheckingwhetherornotapointliesonacurveorsurface.implicitization istheprocessofndingtheimplicitrepresentationsforcurvesorsurfacesfromtheir parametricrepresentations. curvesandsurfaceswithbasepoints(sederbergetal1994;sederbergetal1997). ofmovingalgebraiccurvesandsurfacestosolvetheimplicitizationproblemforrational pointof(1)isaparametert0suchthatx(t0)=y(t0)=w(t0)=0.atabasepointthe rationalexpressionsforxandybothhavetheform0.sederbergintroducedthemethod identicallyinthepresenceofbasepoints(chionh1990;manocha&canny1992).abase surfaces(goldmanetal1984;demontaudouin&tiller1984).butresultantsvanish Resultantscanbeappliedtosolvetheimplicitizationproblemforrationalcurvesand Amovinglineofdegreed parametert.similarly,amovingconicofdegreed isaoneparameterfamilyofimplicitlydenedlines,withonelinecorrespondingtoeach Amongthealgebraiccurves,movinglinesandmovingconicsarethemostimportant. dxi=0(aix2+biy2+cixy+dixw+eiyw+fiw2)ti=0 dxi=0(aix+biy+ciw)ti=0 (2) isaoneparameterfamilyofimplicitlydenedconics.amovingline(2)oramoving conic(3)issaidtofollowarationalcurve(1)if dxi=0(aix(t)+biy(t)+ciw(t))ti0; (3) ordxi=0(aix2(t)+biy2(t)+cix(t)y(t)+dix(t)w(t)+eiy(t)w(t)+fiw2(t))ti0:(5) (4) zero,wegeneratealinearsystemwithunknownsfai;bi;ci;dig(orfai;;fig).any BysettingthecoecientsofallmonomialstiinEquation(4)(orEquation(5))to 2

(1).Themethodofmovinglines(movingconics)constructstheimplicitequationofa rationalcurvebytakingthedeterminantofthecoecientmatrixofasetofindependent movinglines(movingconics)thatfollowthecurve.hereindependencemeansnotjust solutionofthislinearsystemisamovingline(orconic)thatfollowstherationalcurve thelinearindependenceofthesolutionsofthelinearsystemgeneratedfromequation example,kmovinglines (4)(orEquation(5)),butratherindependenceofthemovinglines(orconics).Thus,for aresaidtobeindependentifthematrix lk(ak;0x+bk;0y+ck;0w)++(ak;dx+bk;dy+ck;dw)td=0 l1(a1;0x+b1;0y+c1;0w)++(a1;dx+b1;dy+c1;dw)td=0;. 264A1;0x+B1;0y+C1;0wAk;0x+Bk;0y+Ck;0w studied(sederbergetal1997),andcleverapplicationsofmovinglinesarepresentedin equationofarationalcurve(sederbergetal1997).movinglineshavebeenthoroughly isofrankk. Itisknownthatthemethodofmovinglinesalwayssuccessfullyproducestheimplicit A1;dx+B1;dy+C1;dwAk;dx+Bk;dy+Ck;dw375... (Coxetal1998;Chionhetal1998).However,themovinglinemethodrequirescomputing alargedeterminanttogeneratetheimplicitrepresentation.incontrasttothemethod anadvantageoverthemovinglinesmethod,sincethemethodofmovingconicscomputes ofmovinglines,themethodofmovingconicsdoesnotalwayssuccessfullyyieldthe implicitequationofarationalcurve.rather,themovingconicsmethodproducesthe theimplicitequationofarationalcurvebytakingadeterminantofmuchsmallersize followingthecurve(sederbergetal1997).nevertheless,themovingconicsmethodhas thanthedeterminantgeneratedbythemethodofmovinglines. implicitequationofarationalcurveifandonlyiftherearenolowdegreemovinglines ofthemethodofmovingconicsforrationalcurvesisessentialtoextendingboththe movingsurfaces,wherebasepointsplayamorefundamentalrole.aclearunderstanding methodandtheproofstorationalsurfaces(c.f.section4). conics.forarationalcurvex=x(t) Ourultimategoalistogeneralizethemethodofmovingcurvestothemethodof Thispaperpresentsanewperspectiveonthemethodsofmovinglinesandmoving matrix.inparticularweshallshowthatformovinglinesandmovingconicsofdegree aremovinglines(movingconics)thatfollowtherationalcurve.wecallthecoecient matrixofthislinearsystemthemovinglinematrix(movingconicmatrix).thegoalof equatingthecoecientsofallmonomialsinttozero.thesolutionsofthislinearsystem generatedbysubstitutingx(t);y(t);w(t)intothemovingline(4)(movingconic(5))and thispaperistoderivearelationshipbetweenthemovinglinematrixandthemovingconic w(t),y=y(t) w(t)ofdegree2m,considerthelinearsystem m?1,thedeterminantofthemovinglinematrixactuallyfactorsasub-determinantof 3

in(sederbergetal1997),thisnewproofseemstogeneralizenaturallytothemethodof movingquadricsforrationalsurfaces. conicssuccessfullyproducestheimplicitequationoftherationalcurve.unlikethework therearenolowdegreemovinglinesfollowingarationalcurve,themethodofmoving themovingconicmatrix.basedonthisobservation,wepresentanewproofthatwhen matrixnotation,andinsection3weintroducethemovinglineandmovingconicmatrices. Section4establishestherelationshipbetweenthemovinglineandmovingconicmatrices torationalcurvesofodddegrees,andbrieydiscusspossiblegeneralizationstorational movingconicssuccessfullygeneratestheimplicitequationofarationalcurvewhenthere isnolowdegreemovinglinethatfollowsthecurve.insection5,weextendtheseresults forrationalcurvesofevendegrees,andusesthisrelationshiptoprovethatthemethodof Therestofthispaperisorganizedinthefollowingway.InSection2wexsome surfaces. by1;t;;td.thatis, denotethe(d+1)kcoecientmatrixofthepolynomialspi(t)whoserowsareindexed degreedi,andletd=max(d1;d2;;dk).weshallwritehp1(t)p2(t)pk(t)icto 2Notation Forconvenience,weadoptthefollowingnotation:Letpi(t),1ik,bepolynomialsof Forexample,h(1+2t)(1+2t)t3(4+5t+6t2)iC=264104 p1(t);;pk(t)=(1td)hp1(t)p2(t)pk(t)ic: Ifhp1(t)p2(t)pk(t)iCisasquarematrix,wedenoteitsdeterminantby 205 p1(t)pk(t): 006 010 020 3 75: InSections3and4,wewillstudyonlyrationalcurveswithevendegreesandnobase 3MovingLineandMovingConicMatrices points.considerthenadegree2mrationalcurveinhomogeneousformx(t):y(t):w(t), 4

andgcd(x(t);y(t);w(t))=1.thecartesiancoordinatesofpointsonthecurvearegiven where x(t)=2mxi=0aiti;y(t)=2mxi=0biti;w(t)=2mxi=0citi; X=x(t) w(t); Y=y(t) w(t): (6) generatedbyequatingthecoecientsofallmonomials(int)inequation(4)tozero.we Thecoecientmatrixhx;y;w,,tdx;tdy;tdwiCisofsize(2m+d+1)(3d+3). canwritethissystemas Tondadegreedmovinglinethatfollowscurve(6),weconsiderthelinearsystem Similarly,tonddegreedmovingconicsthatfollowcurve(6),weconsiderthelinear systemgeneratedbyequatingthecoecientsofallmonomials(int)inequation(5)to zero.wecanwritethissystemas hx;y;w;;tdx;tdy;tdwic[a0;b0;c0;;ad;bd;cd]t0: hx2;y2;xy;xw;yw;w2;;tdxw;tdyw;tdw2ic[a0;b0;c0;;dd;ed;fd]t0:(8) (7) Thecoecientmatrixhx2;y2;xy;xw;yw;w2,,tdxw;tdyw;tdw2iCisofsize(4m+d+ 1)(6d+6). degreem?1thatfollowthecurveandthengeneratestheimplicitrepresentationfrom inx;y.themethodofmovingconicsndsasetofmindependentmovingconicsof thedeterminantofthecoecientmatrixofthissetofmovingconics.considerthenm independentmovingconicsofdegreem?1thatfollowthecurve: Itiswellknownthattheimplicitformofcurve(6)isapolynomialofdegree2m whereci;j(x;y;w),0i;jm?1,arequadraticinx;y;w.itisknownthatsuch independentconicsexistandthat pm?1(x;y;w;t)=cm?1;0(x;y;w)+cm?1;1(x;y;w)t++cm?1;m?1(x;y;w)tm?1; p0(x;y;w;t)=c0;0(x;y;w)+c0;1(x;y;w)t++c0;m?1(x;y;w)tm?1; c0;0(x;y;w)cm?1;0(x;y;w). istheimplicitequationofcurve(6),whentherearenomovinglinesofdegreem?1 thatfollowthecurve(sederbergetal1997).wewillthereforeconsiderthecasewhere d=m?1inequations(7)and(8).then[x;y;w,,tdx;tdy;tdwicisasquarematrix c0;m?1(x;y;w)cm?1;m?1(x;y;w)=0... 5

oforder3m denotethis3m3mmatrixbyml.thematrixhx2;y2;xy;xw;yw;w2,,tdxw;tdyw;tdw2icisofsize5m6m denotethis5m6mmatrixbymc.our showthatthemethodofmovingconicsworkswhenevertherearenolowdegreemoving linesthatfollowthecurve. goalistondarelationshipbetweenmlandmc,andthentoapplythisrelationshipto MCw=hx2y2xyxwywtm?1x2tm?1y2tm?1xytm?1xwtm?1ywiC5m5m: 5m.Wedenotethis5m5msubmatrixbyMCw.Tosummarize: polynomialstkw2;k=0;;m?1.thisproceduregeneratesasquaresubmatrixoforder ToobtainasquaresubmatrixfromMC,wedeletethecolumnsthatrepresentthe ML=hxywtm?1xtm?1ytm?1wiC3m3m; fewpreliminarylemmas. Lemma1jMLjisirreducibleinthecoecientsofx(t);y(t);w(t)(Sederbergetal1997). 4jMLjFactorsjMCwj WearegoingtoshowthatjMCwj=constantResulant(x;y)jMLj2.Webeginwitha (Macaulay1916). Lemma3IfjMLj=0,thenjMCwj=0. Lemma2TheresultantRx(t);y(t)isirreducibleinthecoecientsofx(t);y(t) arelinearlydependent.therefore,jmcwj=0.2 thepolynomialstkx2,tky2,tkxy,tkxw;tkyw,k=0;;m?1.thusthecolumnsofmcw Proof.IfjMLj=0,thenthecolumnsinMLarelinearlydependent.Thatis,thereexist constantsi;i=1;;3m,suchthat MultiplyingbothsidesofEquation(9)byx(t)ory(t),wegetalinearrelationshipbetween 1x(t)+2y(t)+3w(t)++3m?2tm?1x(t)+3m?1tm?1y(t)+3mtm?1w(t)0:(9) Lemma4IftheresultantRx(t);y(t)=0,thenjMCwj=0. 6

commonroott0.therefore, Proof.WhentheresultantRx(t);y(t)=0,thetwopolynomialsx(t);y(t)havea =[x2(t0)y(t0)w(t0)tm?1 [1t0t5m?1 0]MCw Lemma5IfjMCwj=0,theneitherRx(t);y(t)=0orjMLj=0. ThustherowsofMCwarelinearlydependent,sojMCwj=0.2 =[00]15m: 0x2(t0)tm?1 0y(t0)w(t0)]15m Proof.IfjMCwj=0,thenthereexistconstantsi;i=1;;5msuchthat Collectthecoecientsofx2;y2;xy;xw;ywandrewriteEquation(10)as wherepi(t);i=1;;5arepolynomialsintofdegreem?1.equation(11)canalsobe writtenas 1x2(t)+2y2(t)+3x(t)y(t)++5m?1tm?1x(t)w(t)+5my(t)w(t)0:(10) p1(t)x+p3(t)y+p4(t)wx?p2(t)y?p5(t)wy: p1(t)x2+p2(t)y2+p3(t)xy+p4(t)xw+p5(t)yw0; (11) thanorequaltom?1suchthatq(t)x=?p2(t)y?p5(t)w.thatis, wecanassumedegree(x)=2m.therefore,thereexistsapolynomialq(t)ofdegreeless p1(t)x+p3(t)y+p4(t)w,andxmustdivide?p2(t)y?p5(t)w.withoutloseofgenerality, x(t)andy(t)donothaveacommonroot.therefore,fromequation(12),ymustdivide NowweshallprovethatifRx(t);y(t)6=0,thenjMLj=0.IfRx(t);y(t)6=0,then (12) p1(t);p3(t);p4(t),suchthatp1(t)x+p3(t)y+p4(t)w0 Ifp2(t)0andp5(t)0,thenbyEquation(12)thereexistnon-zeropolynomials MLarelinearlydependent,sincethedegreesofq(t);p2(t);p5(t)areallatmostm?1. Whenp2(t)orp5(t)arenotidenticallyzero,relationship(13)assertsthatthecolumnsof q(t)x+p2(t)y+p5(t)w0: (14) (13) jmlj=0.2 againassertsthatthecolumnsofmlarelinearlydependent.therefore,ineithercase, withthedegreesofp1(t);p3(t);p4(t)alllessthanorequaltom?1.thisrelationship Theorem6jMCwj=cRx(t);y(t)jMLj2,wherecissomenon-zeroconstant. Withallthispreparation,wecannownallyproveourmainresult. 7

Proof.FromLemma5,weconcludethatjMCwjhasonlytwonon-constantfactors: x(t);y(t);w(t).first,considerjmcwj.sinceeachcolumnofmcwcontainsentrieshomogeneousinthecoecientsofx(t);y(t);w(t),thedeterminantjmcwjisahomogeneous wherecissomenon-zeroconstant. polynomialinthesecoecients.specically,jmcwjishomogeneousofdegree4minthe Letusnowexaminethedegreesofeachofthesedeterminantsinthecoecientsof jmcwj=chr(x(t);y(t)ipjmljq: (15) R(x(t);y(t))andjMLj.Therefore,thereexistpositiveintegersp;qsuchthat coecientsofx(t)andy(t),andhomogeneousofdegree2minthecoecientsofw(t). Ontheotherhand,theresultantRx(t);y(t)ishomogeneousofdegree2minthecoecientsofx(t)andy(t),andjMLjishomogeneousofdegreeminthecoecientsof x(t);y(t);w(t). Comparingthehomogeneousdegreesinthecoecientsofx(t);y(t);w(t)onbothsides ofequation(15),wehavethefollowingequalities: Itiseasytoseethattheonlysolutiontotheseequalitiesisp=1andq=2.Theproof isthereforecomplete.2 2mp+mq=4m; mq=2m:(inthecoecientsofw(t)) (inthecoecientsofx(t)) thecolumnsthatrepresentthepolynomialstkx2ortky2,0km?1,wecanobtain NotethatintheoriginalmovingconicmatrixMC(c.f.Equation(8)),ifwediscard (inthecoecientsofy(t)) similarresults: Corollary7Ifadegree2mrationalcurvex(t):y(t):w(t)doesnothavebasepoints, jmcxj=cry(t);w(t)jmlj2; andtherearenodegreem?1movinglinesthatfollowx(t):y(t):w(t),thenthemethod jmcyj=crw(t);x(t)jmlj2: ofmovingconicsalwayssucceedsinproducingtheimplicitequationforx(t):y(t):w(t). rootamongx(t);y(t);w(t).wecanthentranslatethecurvex(t):y(t):w(t)to x(t0)=0;y(t0)=0,weknowthatw(t0)6=0since,byassumption,thereisnocommon thesuccessofthemovingconicsmethod.infact,iftheresultantrx(t);y(t)=0, thenx(t)andy(t)haveatleastonecommonroot.butforanyparametert0suchthat Proof:FirstweobservethatthevanishingoftheresultantRx(t);y(t)doesnotaect 8

resultantrx(t);y(t)isnotzero. theimplicitequationoftheoriginalcurveisequivalenttondingtheimplicitequation ofthetranslatedcurve.thuswecanalwaysassume,withoutloseofgenerality,thatthe implicitequationofthetranslatedcurveisf(x+constant;y)=0.therefore,tond commonroots.iff(x;y)=0istheimplicitequationoftheoriginalcurve,thenthe x(t)+constantw(t):y(t):w(t)sothatx(t)+constantw(t)andy(t)donothave x(t):y(t):w(t),thenjmlj6=0.thereforebytheorem7,jmcwj6=0.writethe linearsystem(8)(whend=m?1)as Second,iftherearenodegreem?1movinglinesthatfollowtherationalcurve MCw2 6 4Am?1 A E0. 0 The5mmmatrix[w2tm?1w2]ContherighthandsideofEquation(16)has Em?1 3 75=?[w2tm?1w2]C264F0 fullrankm,becausethecolumnsofthismatrixarelinearlyindependent.therefore,the Fm?1375:. system(16)hasmlinearlyindependentsolutions.letpi,0im?1,bethesolution ofsystem(16)correspondingtosetting tries.hencethedeterminantjp0pm?1jcontainsthetermw2m.thus,thisdeter- minantdoesnotvanishidentically.sinceeachentryinthisdeterminantisquadraticin Thereforethecoecientmatrix[P0Pm?1]Ccontainsw2onlyinthediagonalen- 0j6=i: Then Pi=tiw2+termswithoutw2; Fj=(1j=i; x;y;w,thetotaldegreeofthisdeterminantisatmost2m.moreover,byconstruction, eachcolumnpi,0im?1,isamovingconicthatfollowsthecurve,soforpoints onthecurve,therowsarelinearlydependent;hencethisdeterminantiszeroforpoints 0im?1: m?1thatfollowsthecurve.2 x(t):y(t):w(t)isrepresentedbyauniqueirreducibledegree2mpolynomialequation. onthecurve.ontheotherhand,theimplicitequationofthedegree2mrationalcurve Therefore,thedeterminantjP0Pm?1jmustbetheimplicitequationoftherational curve,sothemethodofmovingconicssucceedswhenthereisnomovinglineofdegree 9

5GeneralizationsandExtensions 5.1RationalCurveswithOddDegrees InSections3and4,wediscussedarelationshipbetweenthemovinglineandmoving propositionshold. conicmatricesforrationalcurvesofevendegrees.forodddegreerationalcurves,similar isofsize(3m+2)(3m+3).togetasquaresubmatrix,discardthelastcolumnfrom ML,andwritetheresultingsubmatrixas Considerarationalcurvex(t):y(t):w(t)ofdegree2m+1.Themovinglinematrix MLw=[xywtmxtmy]C(3m+2)(3m+2): ML=[xywtmxtmytmw]C km,andtmxw;tmyw.theresultingsquaresubmatrixis isofsize(5m+3)(6m+6).deletethecolumnsthatrepresentthepolynomialstkw2;0 Themovingconicmatrix MC=[x2y2xyxwyww2tmx2tmy2tmxytmxwtmywtmw2]C wherecissomenon-zeroconstant.itfollowsfromthisequationbyanargumentanalogous ByananalysissimilartothatofSection4,wehave MCw=[x2y2xyxwywtmx2tmy2tmxy]C(5m+3)(5m+3): tothatintheproofofcorollary7thatwhentherearenobasepointsandwhenthere curvex(t):y(t):w(t),themethodofmovingconicssuccessfullygeneratestheimplicit existsonlyoneindependentmovinglineofdegreemthatfollowsthedegree2m+1 equationfortherationalcurve. jmcwj=cresultantx(t);y(t)jmlwj2; 5.2RationalSurfaces Forrationaltensorproductsurfaces,movinglinesandmovingconicsgeneralizetomovingplanesandmovingquadrics.Empiricalstudiesandnumericalexperimentsshowthat surface.moreover,usingthismethodtheimplicitequationofasurfaceofbidegree(m;n) themethodofmovingquadricsgenerallyproducestheimplicitequationforarational bytheusualresultantmethods(order2mn).furthermore,whenbasepointsarepresent, isrepresentedbyamuchmorecompactdeterminant(ordermn)thantheonegenerated 10

letmqbethemovingquadriccoecientmatrix.weconjecturethat veryusefultoknowexactlywhenthemethodofmovingquadricsworks. simpliesinthepresenceofbasepoints(sederberg&chen1995).therefore,itwouldbe standardresultantmethodseitherfail(becomeidenticallyzero)orbecomeverycomplicated,whereasthemethodofmovingquadricsstillgenerallysucceedsandindeedoften wheremqwisthesubmatrixofmqobtainedbydeletingthecolumnsrepresentingthe amongthecolumnsofmlgeneratestworelationsamongthecolumnsofmcw(multiplyingbyx(t)ory(t))[c.f.lemma3];hencejmljisadoublefactorofjmcwj.forthe bivariatecase,eachrelationamongthecolumnsofmpgeneratesthreerelationsamong jmqwj=cresultant(x(s;t);y(s;t);z(s;t))jmpj3; polynomialsthataremultiplesofw2.notethatintheunivariatesetting,eachrelation Inanalogywithrationalcurves,letMPbethemovingplanecoecientmatrix,and thecolumnsofmqw(multiplyingbyx(s;t),y(s;t)orz(s;t)).soweexpectjmpjshould surface. deedholdforrationalsurfaces.wehopetoprovethisassertioninafuturepaper,and beatriplefactorofjmqwj.numericalexperimentsshowthatthisrelationshipdoesin- Acknowledgments toapplythisresulttoshowthatthemethodofmovingquadricsalwayssuccessfullyimplicitizesarationalsurfacewhentherearenolowdegreemovingplanesthatfollowthe providedbybyu.mingzhangandrongoldmanarepartiallysupportedbynsfgrant CCR-9712345. BrighamYoungUniversity.Hegreatlyappreciatesthehospitalityandfacilitiesgenerously Eng-WeeChionhissupportedbytheNationalUniversityofSingaporeforresearchat References [1]Chionh,E.W.,Zhang,M.,Goldman,R.N.(1998),ImplicitizationMatricesinthe [3]Goldman,R.N.,Sederberg,T.,Anderson,D.(1984),VectorElimination:ATech- [2]Cox,D.,Sederberg,T.W.,Chen,F.(1998),TheMovingLineIdealBasisforPlanar StyleofSylvesterwiththeOrderofBezout,submittedtoComputerAidedDesign. [4]Macaulay,F.S.(1916),TheAlgebraicTheoryofModularSystems,CambridgeUniversityPress. RationalCurves,ComputerAidedGeometricDesign,toappear. niquefortheimplicitization,inversion,andintersectionofplanarparametricratio- nalpolynomialcurves.computeraidedgeometricdesign1,327-356. 11

[5]Manocha,D.andCanny,J.F.(1992),AlgorithmsforImplicitizingRationalParametricSurfaces.ComputerAidedGeometricDesign9,25{50. [7]Sederberg,T.W.,Saito,T.,Qi,D.,Klimaszewski,K.(1994),CurveImplicitization [6]DeMontaudouin,Y.,Tiller,W.(1984),TheCayleyMethodinComputerAided UsingMovingLines.ComputerAidedGeometricDesign11,687-706. GeometricDesign.ComputerAidedGeometricDesign1,309{326. [8]Sederberg,T.W.,Chen,F.(1995),ImplicitizationUsingMovingCurvesandSurfaces,ProceedingsofSIGGRAPH,301-308. [9]Sederberg,T.W.,Goldman,R.N.,Du,H.(1997),ImplicitizingRationalCurvesby themethodofmovingalgebraiccurves,j.symboliccomp.23,153{175. 12