LINEARIZATION METHODS FOR VARIATIONAL INTEGRATORS AND EULER-LAGRANGE EQUATIONS

Size: px
Start display at page:

Download "LINEARIZATION METHODS FOR VARIATIONAL INTEGRATORS AND EULER-LAGRANGE EQUATIONS"

Transcription

1 LINEARIZATION METHODS FOR VARIATIONAL INTEGRATORS AND EULER-LAGRANGE EQUATIONS Tesis by AnilN.Hirani Avisors James R. Arvo, Jerrol E. Marsen In Partial Fulfillment of te Requirements for te Degree of Master of Science in Computer Science California Institute of Tecnology Pasaena, California April 000

2 ii c April 000 Anil N. Hirani All rigts Reserve

3 iii Acknowlegements I like to tank my avisors Prof. Jerrol E. Marsen an Prof. Jim Arvo for teir invaluable elp an avice. Te comments mae by te members of my caniacy committee were also very elpful. Besies my two avisors, te committee members were Professors Peter Scröer, Alan Barr an Tom Hou. I am also tankful to Mattew West an Matieu Desbrun for some very elpful suggestions. In te text I ve cite te specific result were Mattew s suggestions were most important. Matieu was also generous wit is avice an gave me some very goo suggestions. Once, e gave me a table of coefficients for finite ifferences an tat really elpe precipitate te completion of tis work. Discussions wit Antonio Hernanez, Dong-Eui Cang an Sameer Jalnapurkar on some geometric mecanics erivations were also elpful.

4 iv

5 v LINEARIZATION METHODS FOR VARIATIONAL INTEGRATORS AND EULER-LAGRANGE EQUATIONS by AnilN.Hirani Abstract In Partial Fulfillment of te Requirements for te Degree of Master of Science in Computer Science Hamiltonian systems arise in a wie variety of iealize pysical systems an Hamilton s equations often must be solve numerically. In general, traitional finite ifference metos for numerically integrating orinary ifferential equations o not take into account te special structure of Hamilton s equations. One woul like te numerical solution to preserve some or all of te properties of te continuous time solution. Variational integrators, also calle iscrete Euler-Lagrange equations in tis tesis, provie a way to o tat. Tis tesis is about ifferentiating, i.e linearizing Euler-Lagrange equations an iscrete Euler-Lagrange equations by ifferentiating te Lagrangian or iscrete Lagrangian. Te original motivation for ifferentiating te Lagrangian was to evelop iger orer variational integrators. However it was foun tat te resulting intgerators are not iger orer. Noneteless, te pursuit of tis goal turne up several interesting results : i. Differentiating a Lagrangian along te acceleration, i.e along ( q, q) results in te Euler- Lagrange equation 0 = 0 (Lemma, page 0). ii. Differentiating a Lagrangian along a general irection (u, u), etermining te Euler- Lagrange equation an ten setting u = q results in an Euler-Lagrange equation tat is te time erivative of te original Euler-Lagrange equation corresponing to te original Lagrangian (Lemma 3, page 1). iii. A similar operation on te iscrete Lagrangian gives a iscrete time erivative of te iscrete Euler-Lagrange equation corresponing to te original iscrete Lagrangian (Lemma 4, page 3). iv. Te operations escribe in te previous two items are part of a commutative iagram (Teorem 1, page 8). v. Numerical measurements an analytical calculation of orer of accuracy for specific examples using tis approac sow tat iger orer metos are not obtaine in general (Capter 5, especially Section 5.4 an 5.5). vi. Te manifol consisting of triples tat satisfy a iscrete Euler Lagrange equation is a smoot manifol (Proposition 4, page 35). vii. Te algoritm on TQ TQ is symplectic an te algoritm tat upates triples of points on Q is presymplectic (Teorem, page 36). I also give a erivation of a symplectic form on TQ TQ were Q is te configuration manifol. Most proofs given in tis tesis are one using local coorinates. A possible irection for furter stuy is to give intrinsic proofs in orer to gain a eeper unerstaning of te geometric meaning of te operations of tis tesis.

6 vi Table of Contents Acknowlegments iii Abstract v 1 Introuction Variational Integrators Results Preliminaries 7.1 Lagrangian Mecanics Geometry Hamiltonian Mecanics Legenre Transform Algoritms Variational Integrators Discrete Mecanics Invariance Properties Linearization Motivation Differentiating Lagrangian along ( q, q) Continuous Time Linearization Discrete Time Linearization Discrete an Continuous Linearization Summary Invariance Properties Symplectic form on TQ TQ Presymplectic algoritm Examples an Conclusion Falling Particle Continuous Discrete Simple Harmonic Oscillator Continuous Discrete Duffing Oscillator Continuous Discrete Numerical Results Accuracy Calculations

7 vii 5.6 Using more accurate u k Conclusion Bibliograpy 53

8 1 Capter 1 Introuction Te traitional meto of numerically solving te initial value problem (IVP) for orinary ifferential equations (ODEs) is to iscretize erivatives by finite ifferences. For example for q R an te equation q = f(q, t), (1.1) t one can coose a step size for stepping te time iscretely an replace te left an sie of (1.1) by q k+1 q k. Muc of te evelopment of numerical metos for IVPs as concerne itself wit wat to o wit te rigt an sie of equation (1.1) in orer to iscretize te equation completely. Te evelopment as resulte in a well evelope teory of truncation errors (accuracy), stability, an convergence. Te books by Hairer et.al [6], Lambert [11], Gear [3] an Iserles [8] are excellent sources of information about numerical metos for ODEs. Te table of coefficients for finite ifference approximations for various orer erivatives on pages 16 an 17 of Fornberg s book [1] is also quite any. But consier now te following simple system of equations, given a scalar function H : R R R suc tat q = H (1.) p ṗ = H q. Tis is an example of a Hamiltonian system of equations, wic is te form tat many iealize equations in pysical sciences take. Some equations tat may not at first appear to be in tis form are also actually Hamiltonian in a sense efine in Capter an in Capter 5 of te book by Marsen an Ratiu [15]. Some examples of equations tat are Hamiltonian are : equation of a particle in a potential, rigi boy equations, rigi boy wit a fixe point in gravitational fiel, Euler equations for an ieal, incompressible flui, Maxwell-Vlasov equations for a collisionless plasma, Maxwell s equations an many oters. Te stuy an solution of Hamiltonian equations is tus a subject wit a long traition. It can be sown tat for suc an equation like (1.) te function H is constant along solutions,i.e Ḣ(q(t),p(t)) = 0 for all time t if (q(t),p(t)) is a solution of te equations 1..

9 If equation (1.) is iscretize using an arbitrary meto tere is no guarantee tat te iscrete solutions will also satisfy te property of keeping H constant along solutions. Even worse, tere may be a systematic biase error introuce wic will make te compute solution rift from te true solution. Since H is often te energy of te system tis is not acceptable for many situations. Anoter important property of Hamiltonian equations is tat te flow is symplectic, i.e te flow of te vector fiel efine by te equations preserves a certain ifferential form (tis will be efine in Capter ). One consequence of symplectic flow is tat areas in pase space are conserve. An arbitrary iscretization of equation (1.) may not yiel an algoritm tat is symplectic. However, one woul like a numerical solution to ave some of tese properties like energy conservation or symplecticness. For example, Figures 1.1 an 1., wic are taken from Iserles an Zanna [7] sow ow initial conitions evolve uner te forwar Euler an implicit mipoint metos. Te smiley face sown is a clou of initial conitions wic are followe over time. Te equation being solve is tat of a simple armonic oscillator, namely q + q = 0 an so te smiley face soul just rotate rigily troug pase space. But in Figure 1.1 te smiley face enlarges. Figure 1.sows te ieal beavior. Te particular implicit meto use in Figure 1.is symplectic, so it conserves pase space area as one steps troug time. Figure 1.1: Pase flow for armonic oscillator as approximate by forwar Euler meto. Horizontal axis is position an vertical axis is momentum. Te uprigt (rigtmost) smiley face is a collection of initial conitions an oters are te location of te collection at various times. From Iserles an Zanna [7]. To see a comparison of te momentum an energy beavior of fourt orer non symplectic Runge-Kutta versus tat of a symplectic algoritm please refer to Figure 13 on page

10 3 Figure 1.: Pase flow for armonic oscillator as approximate by implicit mipoint rule. Horizontal axis is position an vertical axis is momentum. Te uprigt (rigtmost) smiley face is a collection of initial conitions an oters are te location of te collection at various times. From Iserles an Zanna [7]. 18 of Marsen et.al [1]. It is sown tere tat te non symplectic Runge-Kutta meto oes not preserve momentum wile te symplectic meto oes. Furtermore, te energy compute by symplectic metos appears to stay near te true value wile tat compute by te Runge-Kutta meto oes not. Te beavior isplaye by te symplectic integrators in Figure 13 of Marsen et. al [1] is in some sense te best possible because of a teorem prove in Ge an Marsen [], tat a constant time stepping integrator cannot preserve te symplectic form, momentum, an energy simultaneously. However, Kane et.al [9] sow tat by using aaptive time stepping an an appropriate efinition of symplecticness one can inee get symplectic, momentum an energy preserving algoritms. In eveloping symplectic integrators early researcers use generating functions (wic are symplectic transformations) an Hamilton-Jacobi equation. A more recent approac is to create variational integrators using iscrete Euler-Lagrange equations, a meto wic is escribe in te next section. 1.1 Variational Integrators If one works wit te Hamiltonian, symplectic integrators for Hamiltonian equations can be create using generating functions. In general tis is more complicate tan te meto iscovere by Veselov [19]. Tis meto, wic uses te Lagrangian instea of te Hamiltonian, can be caracterize as te creation of integrators base on iscrete variational principles.

11 4 It is escribe an furter evelope by Wenlant an Marsen [0]. Given any Hamiltonian system of equations, uner some general conitions a corresponing Lagrangian exists an te Hamiltonian equations are equivalent to some variational principle. Tis will be mae clearer in Capter 3. Te meto is to iscretize te continuous Lagrangian an use a iscrete variational principle. At least tis was te current wisom wen I starte working on tis problem. In recent work Kane et.al [10] an West [17] ave sown tat te iscrete Lagrangian soul be consiere to be an approximation for te action integral over one time step. But tis point of view will not be explore in tis tesis. Rougly speaking, te meto use in tis tesis is to replace te position variable by some position an te velocity variable in te Lagrangian by some finite ifference. Finally one replaces a variational principle in wic an integral appears by one in wic a sum appears. Extremizing accoring to te principle yiels a iscrete Euler-Lagrange equations wic form te basis for te algoritm. It can be prove tat te algoritm is symplectic an momentum preserving. 1. Results In tis tesis I present some results regaring ifferentiating, i.e linearizing Lagrangians an iscrete Lagrangians, te resulting Euler-Lagrange an iscrete Euler-Lagrange equations an te connections between all tese. Te original motivation for ifferentiating te Lagrangian was to evelop iger orer variational integrators. However it was foun tat te resulting intgerators are not, in general, iger orer. Tus te iscrete sie results escribe below are not computationally useful. Noneteless, te pursuit of tis goal turne up several interesting results : i. Differentiating a Lagrangian along te acceleration, i.e along ( q, q) results in te Euler- Lagrange equation 0 = 0 (Lemma, page 0). ii. Differentiating a Lagrangian along a general irection (u, u), etermining te Euler- Lagrange equation an ten setting u = q results in an Euler-Lagrange equation tat is te time erivative of te original Euler-Lagrange equation corresponing to te original Lagrangian (Lemma 3, page 1). iii. A similar operation on te iscrete Lagrangian gives a iscrete time erivative of te iscrete Euler-Lagrange equation corresponing to te original iscrete Lagrangian (Lemma 4, page 3). iv. Te operations escribe in te previous two items are part of a commutative iagram (Teorem 1, page 8). v. Numerical measurements an analytical calculation of orer of accuracy for specific examples using tis approac sow tat iger orer metos are not obtaine in general (Capter 5, especially Section 5.4 an 5.5). vi. Te manifol consisting of triples tat satisfy a iscrete Euler Lagrange equation is a smoot manifol (Proposition 4, page 35). vii. Te algoritm on TQ TQ is symplectic an te algoritm tat upates triples of points on Q is presymplectic (Teorem, page 36). I also give a erivation of a symplectic form on TQ TQ were Q is te configuration manifol. Te structure of te rest of tis tesis is as follows. Capter is a quick review of

12 5 continuous time Lagrangian an Hamiltonian mecanics. Capter 3 introuces variational integrators an Capter 4 escribes te linearization results mentione above. Capter 5 gives examples of linearization on bot te continuous an te iscrete sie. In tis capter I also o some erivations of orer of accuracy calculations. Tese are accompanie by results from numerical experiments for measuring orer of accuracy. Bot, te numerical experiments an te analytical calculations sow tat our meto oes not, in general, give iger orer symplectic integrators.

13 6

14 7 Capter Preliminaries Before eveloping te teory an escribing te linearization meto I will give some necessary efinitions, stanar results an examples. Tis section is aapte primarily from Lecture 5 of Marsen [13] an from Capters an 5 of te book by Marsen an Ratiu [15]..1 Lagrangian Mecanics Let Q be a configuration manifol of a mecanical system an L : TQ R be a function (calle a Lagrangian). Let q =(q 1,...,q n ) be local coorinates for Q an let c :[a, b] R TQ be a smoot curve on TQ. Using stanar abuse of notation I will write c(t) (q(t), q(t) ) (q(t), q(t)) t for t [a, b]. Te action integral S is efine as S = b a L(q(t), q(t))t. Ten te variational principle calle Hamilton s Principle states tat δs =0 L t q i L =0fori =1,...,n. (.1) qi Definition 1 Te evolution equations L t q i L =0 (.) qi are calle te Euler-Lagrange equations.. Geometry Definition If f : M R is a smoot function on a ifferentiable manifol M ten it can be ifferentiate at any point m M to obtain a map T m f : T m M T f(m) R. Te tangent space of R can be ientifie wit itself since R is a vector space. Tis gives us a linear map f(m) :T m M R. Tus f(m) T mm, te ual of te vector space T m M.

15 8 Tis map f is calle te ifferential of f. Teirectional erivative in te irection v T m M is efine to be f(m) v. Definition 3 A symplectic form Ω on a vector space Z is a nonegenerate skew symmetric, bilinear form on Z. Te pair (Z, Ω) is calle a symplectic vector space. TeformΩ is sai to be (weakly) nonegenerate if Ω(z 1,z ) = 0 for all z Z implies z 1 =0. It can be sown tat tere is a basis for Z in wic Ω(u, v) =u T J v for all u, v Z an were J = ( 0 I I 0 ) were 0 an I are n n zero an ientity matrices for a symplectic vector space Z of imension n. Tus finite imensional symplectic vector spaces always ave an even imension. For an example of a symplectic vector space consier Z = W W were W is a vector space an W is its ual. Te canonical symplectic form ΩonZ is efine as were w 1,w W an α 1,α W. Ω((w 1,α 1 ), (w,α )) = α (w 1 ) α 1 (w ) Definition 4 A symplectic manifol is a pair (P, Ω), were P is a manifol an Ω is a close (weakly) non egenerate two-form on P. An example of a symplectic manifol is te cotangent bunle of te configurations space. Let te manifol Q be te configuration space of a mecanical system. Ten TQ is te isjoint union (over all points of Q) of te tangent spaces an is calle te tangent bunle. At eac point p Q te tangent space T p Q is a vector space an ence tere is an associate ual vector space Tp Q. Te isjoint union of all tese over all te points of te manifol is te cotangent bunle T Q an it is a symplectic manifol. Definition 5 Let (P 1, Ω 1 )an(p, Ω ) be symplectic manifols. A C -map ϕ : P 1 P is calle symplectic or canonical if ϕ (Ω )=Ω 1. Tis pull-back notation means tat for all q P 1,anallv, w T q P 1 Ω 1p (v, w) =Ω ϕ(p) (T p ϕ v, T p ϕ w) (.3) were Ω 1p means Ω 1 evaluate at point p an T p ϕ is te erivative of ϕ at p. Animportant result (see e.g Capter 5 of Marsen an Ratiu [15]) is Proposition 1 A smoot canonical transformation between symplectic manifols of te same imension is volume preserving an is a local iffeomorpism.

16 .3 Hamiltonian Mecanics 9 Te motivation for tese efinitions above concerning symplectic manifols an maps is te following proposition wic is prove by Marsen an Ratiu in Capter of teir book [15]. But first we nee a efinition of Hamiltonian vector fiel on a symplectic manifol. Definition 6 Let (P, Ω) be a symplectic manifol. A vector fiel X on P is calle Hamiltonian if tere is a function H : P R suc tat for all v T z P Ω z (X(z),v)=H(z) v. In tis case one writes X H for X. Hamilton s equations are te evolution equations ż = X H (z). Proposition Te flow ϕ t of a Hamiltonian vector fiel X H consists of canonical (i.e symplectic)transformations. Conversely if te flow of a vector fiel X consists of canonical transformations, ten it is Hamiltonian..4 Legenre Transform A transform tat allows us to move between Hamiltonian an Lagrangian mecanics is te Legenre Transform efine below. Definition 7 Given a Lagrangian L : TQ R efine a map FL : TQ T Q calle te fiber erivative by FL(v) w = s L(v + sw) s=0 were v, w T q Q. Tus FL(v) w is te erivative of L at v along te fiber T q Q in te irection w. In a local cart U E for TQ were U is open in te moel space E for Q, FL(u, e) =(q, D L(u, e)) were D L is te partial erivative of L wit respect to its secon argument. For a finite imensional manifol Q wit (q i ) te coorinate for Q, an(q i, q i ) te inuce coorinates for TQ te fiber erivative as te local expression ( FL(q i, q i )= q i, L ) q i Notation : In te above equation FL(q i, q i ) soul be rea FL(q 1,...,q n, q 1,..., q n ). A similar convention will be use trougout. Tus FL is given by p i = L q i (.4)

17 10 Tis transformation given by is also calle te Legenre transform. j=1 FL :(q i, q i ) (q i,p i ) In te Euler-Lagrange equations (.) te cain rule can be use to to get n ( ) L q j q i qj + L q j q i qj L =0, i =1,...,n (.5) qi Definition 8 To write q j in equation (.5) as a function of q an q it is require tat te matrix L q j q i be invertible. A Lagrangian for wic tis is invertible is calle a regular Lagrangian. If L is regular ten one can o te cange of variables (q i, q i ) (q i,p i ) at least locally ue to te implicit function teorem. Te Legenre transform allows us to go back an fort between Hamiltonian an Lagrangian mecanics ue te following proposition : Proposition 3 Let L(q i, q i ) be a regular Lagrangian. Define a new function ( n ) H(q i,p i )= p i q i L(q i, q i ) i=1 were p i = L q i Wit tis efinition te Euler-Lagrange equations are equivalent to te Hamilton s equations, i.e.5 Algoritms L t q i L q i =0 { q i t = H p i, p i t = H q i Definition 9 An algoritm on a pase space P is a collection of maps F : P P, usually epening smootly on R for small >0. Avancing te solution z k by one time step can be written as z k+1 = F (z k ). Tis means avancing time by wic is also calle te step size. Hereallz k P. Definition 10 An algoritm F is i. a symplectic integrator if eac F is symplectic, ii. an energy integrator if H F = H, iii. a momentum integrator if J F = J, werej is te momentum map for te action of a Lie group G.

18 11 Remark 1 It as been sown by Ge an Marsen [] tat a fixe step size numerical integrator can only be symplectic an momentum preserving or energy an momentum preserving. In te remainer of tis tesis I will refer to symplectic-momentum integrators as simply symplectic integrators. Remark In tis tesis I will not aress te momentum conservation issue an I will not give te efinitions for te terms use in te efinition of a momentum integrator above. Tese can be foun in Wenlant an Marsen [0] an in te references containe terein.

19 1

20 13 Capter 3 Variational Integrators 3.1 Discrete Mecanics As escribe by Veselov [19] an Wenlant an Marsen [0] one can evelop a iscrete picture corresponing to te continuous mecanics escribe in Section.1. Instea of a Lagrangian L : TQ R an te variational principle S = b a L(q(t), q(t))t one efines a iscrete Lagrangian L : Q Q R. Te action integral above is replace by an action sum S : Q N+1 R efine as S = N 1 k=0 L (q k,q k+1 ) were q k Q. Te iscrete variational principle states tat te evolution equations extremize te action sum given fixe en points, q 0 an q N. Extremizing S over q 1,...,q N 1 leas to te iscrete Euler-Lagrange equations : D 1 L (q k,q k+1 )+D L (q k 1,q k )=0 for k =1,...,N 1 (3.1) were D i is te partial erivative wit respect to te i-t argument. Tese equations can also be written as D 1 L Φ+D L =0 (3.) were Φ : Q Q Q Q is efine implicitly by Φ(q k 1,q k )=(q k,q k+1 ). If D 1 L is invertible ten equation (3.) efines te iscrete map Φ wic flows te system forwar in iscrete time. Ten one can sow various properties like symplecticness, momentum map preservation of Φ as one by Wenlant an Marsen [0]. In orer to obtain an algoritm in terms of te original continuous Lagrangian one must relate L wit L an in Wenlant an Marsen [0] tis is one by efining ( y + x L (x, y) =L, y x ) (3.3) were is te step size. For examples of variational integrators constructe accoring to tis sceme please con-

21 14 sult Section 4 of Wenlant an Marsen [0]. See Capter 5 for some aitional examples. Tese examples sow te construction for variational integrators by te sceme escribe above as well as integrators obtaine by ifferentiating te iscrete Lagrangian accoring te main ieas of tis tesis. 3. Invariance Properties To talk about te symplecticness of te algoritm Φ : Q Q Q Q one as to sow tat some symplectic form is being preserve i.e tat Φ is a symplectic map. But wic symplectic form? Te most natural coice is to pull back te canonical symplectic form from te cotangent bunle T Q. Tis is wat was one by Veselov [19] an wic is explaine by Wenlant an Marsen [0]. To o tis, we efine te fiber erivative suc tat FL : Q Q T Q (3.4) FL :(q k,q k+1 ) (q k, D1L (q k,q k+1 )). Now one can efine te symplectic -form on Q Q by pulling back te canonical -form from T Q by FL.TusifΩ CAN is te canonical -form on T Q ten ω on Q Q is ω = FL (Ω CAN) (3.5) Please refer to Sections 3.1 an 3.of Wenlant an Marsen [0] wic give a proof tat te algoritm Φ preserves te symplectic form, i.e Φ ω = ω. Tere one can also fin a local coorinate expression for ω.

22 15 Capter 4 Linearization In tis capter I will escribe our meto for ifferentiating, i.e linearizing Euler-Lagrange equations an variational integrators by manipulating te Lagrangian an te iscrete Lagrangian. Tis manipulation is suc tat in te continuous case, applying te appropriate variational principle yiels a continuous time erivative of te Euler-Lagrange equation corresponing to te original Lagrangian. In te iscrete case, te iscrete Lagrangian is manipulate an te iscrete variational principle yiels a iscrete time erivative of te iscrete Euler-Lagrange equations corresponing to te original iscrete Lagrangian. If te original iscrete Euler-Lagrange equation is linear, te iscrete time erivative is simply a finite ifference. Of course, to linearize continuous (or iscrete) Euler-Lagrange equations one coul just as well start wit te equation an take its erivative (or finite ifference). But as mentione in Section 1., my original aim wic le to te present results, was to moify te Lagrangian an ten examine te resulting Euler-Lagrange equations. Tis is because te iscretization of te Euler-Lagrange equations is generate from te Lagrangian an not from te continuous time equations. 4.1 Motivation Consier te iscrete Euler-Lagrange equation wic gives te symplectic algoritm as escribe in Capter 3. Tis is repeate ere for convenience : D 1 L Φ+D L =0 (4.1) were Φ : Q Q Q Q is efine implicitly by te above equation an Φ(q k 1,q k )= (q k,q k+1 ). Since L : Q Q R, te resulting algoritm map Φ takes a pair of points (q k 1,q k ) Q an prouces a pair (q k,q k+1 ). It is reasonable to conjecture tat an algoritm tat is able to use more points, say 3 points, migt result in a better truncation error since points allow one to approximate erivative an 3 will allow one to approximate acceleration as well. Tis was te motivation for our subsequent ieas tat le to te linearization metos of tis tesis. Tus, wat is esire is an algoritm map Φ () : Q Q Q Q Q Q. 4. Differentiating Lagrangian along ( q, q) Wit a view to introucing accelerations into te framework, consier wat appens wen te Lagrangian is a function of q, q an q, i.e te Lagrangian is some function L () (q, q, q).

23 16 Hamilton s principle (.1) an simple variational calculus (see for example Capter, page 4, equation of Gelfan an Fomin [4]) ensures tat te solutions q(t) satisfytefourt orer Euler-Lagrange equations L () t q i L () t q i + L() q i =0. (4.) Tis is fourt orer because in general te term L () t q i (q, q, q) (4.3) involves a fourt erivative term. Tus in going from L(q, q) tol () (q, q, q) one goes from a secon orer Euler-Lagrange equation to a fourt orer Euler-Lagrange equation. Tis can be represente by Figure 4.1. On te left someting as yet unspecifie is one to introuce L(q, q)? Hamilton s Principle EL(L) is secon orer? L () (q, q, q) Hamilton s Principle EL(L () ) is fourt orer Figure 4.1: On te left q as been introuce into te Lagrangian but on te rigt te equations go from being secon to fourt orer. acceleration into te Lagrangian. On te rigt one sees tat te Euler-Lagrange equations skip an orer an go from being secon orer to fourt orer. Te question marks inicate an as yet unspecifie operation. Wat is te most obvious operation one can perform on L(q, q) to introuce acceleration? Te answer is ifferentiation. Tus one migt expect te question mark on te rigt own arrow to peraps represent ifferentiation also. However, as it turns out, wen te secon orer Euler-Lagrange equations for L(q, q) namely equation (.) are ifferentiate one gets a tir orer equation as sown by te following calculation (using implie summation convention were necessary). First, by ifferentiating te Euler-Lagrange equations for L(q, q) weget: t But te left an sie of te above equation is ( L t q j q i qj + L q j q i qj ( ) L L (q, q) (q, q) =0. (4.4) t q i qi ) L q j q i qj L q j q i qj, (4.5)

24 17 wic is ( = L q j q i qj + q j 3 L q k q j q i qk + ( q j + q j L... q j q i 3 ) L q k q j q i qk + 3 L q k q j q i qk + 3 ) L q k q j q i qk L q j q i qj L q j q i qj = L q j q i qj + 3 L q k q j q i qk q j + L... q j q j q i + 3 L q k q j q i qk q j + 3 L q k q j q i qk q j + 3 L q k q j q i qk q j L q j q i qj L q j q i qj Terefore te ifferentiation of te Euler-Lagrange equation for L(q, q) gives te following equation : L q j q i qj + 3 L q k q j q i qk q j + L... q j q j q i + 3 L q k q j q i qk q j + 3 L q k q j q i qk q j + 3 L q k q j q i qk q j L q j q i qj L q j q i qj =0. (4.6) Tis equation (4.6) is tir orer. Tus te iagram in Fig. 4.1 will not commute if own arrows are tougt of as ifferentiation. On te rigt one nees to go to a tir orer Euler-Lagrange equation. However if L () (q, q, q) was only linear in q ten te term L () q i in equation (4.) will be a function only of q an q as a result of wic L () t q i will only be a tir orer term. Te Euler-Lagrange for suc a Lagrangian will be tir orer. As te left own arrow was being tougt of as ifferentiation tis migt work out. Tis suggests getting L () (q, q, q) froml(q, q) by simply formally ifferentiating L(q, q) wit respect to time. If L(q, q) is simply formally ifferentiate wrt t ten one gets L () (q, q, q) = L q i qi + L q i qi (4.7) Note tat now L () (q, q, q) is linear in q, as esire. Now te next step ougt to be to try an see if te iagram in Figure 4.commutes. But tat will not be wortwile because

25 18 ifferentiate L(q, q) L () (q, q, q) linear in q Hamilton s Principle EL(L) is secon orer ifferentiate Hamilton s Principle EL(L () ) is tir orer Figure 4.: On te left q as been introuce into te Lagrangian by ifferentiating L. Since L () is linear in q, EL(L () ) soul be tir orer. But tere is a major problem as is explaine in te text. of te two Lemmas tat follow. Lemma 1 below sows tat for Lagrangians L of a certain type, wen L () is obtaine as in equation (4.7) an Hamilton s principle is applie using Lagrangian L () one gets te trivial ientity 0 = 0 for te Euler-Lagrange equations. Te next Lemma removes te restriction on te form of Lagrangian. 1 Lemma 1 Let te L be of te form L(q, q) =L T ( q)+l V (q) (4.8) (Te superscripts T an V are reminers tat te term L T ( q) is te kinetic energy an te term L V (q) is minus te potential energy term). Construct a new Lagrangian L () (q, q, q) as L () (q, q, q) = L q i qi + L q i qi (4.9) ForsucaLagrangianL () te Hamilton s principle gives no information about te evolution of state, i.e te Euler-Lagrange equation for L () is 0=0. Proof Wen te Lagrangian of te form (4.8) is ifferentiate to get L () efine by equation (4.9) L () (q, q, q) = LT q i qi + LV q i q i (4.10) is obtaine since te oter terms are 0. Define L (T ) ( q, q) LT q j qj (4.11) L (V ) (q, q) LV q j qj 1 Te more general result in Lemma was suggeste by Mattew West wen I sowe im Lemma 1.

26 19 Now te Euler-Lagrange equation for L () (q, q, q), namely equation (4.) is ( L (T ) q i L (T ) t q i + L (T ) ) t q i + ( L (V ) q i L (V ) t q i + L (V ) ) t q i =0 (4.1) Substitute for L (T ) an L (V ) from equation (4.11) into equation (4.1) an consier te L (T ) an L (V ) parts separately. Te L (T ) part of equation (4.1) becomes 0 L (T ) t q i + L (T ) t q i = ( ) ( ) L T t q i q j qj + L T t q i q j qj ( L T ) + = t q i q j qj = ( L T t q i q j qj =0 ) + t L T t q i ( L T q j q i qj were te last step follows from te equality of mixe partials. Similarly te L (V ) part of equation (4.1) becomes ) L (V ) q i L (V ) t q i +0 = ( ) L V q i q j qj ( L V t q i q j = L V q i q j qj L V t q i = L V q i q j qj LV q j q i qj = 0 qj ) were again te last step follows from te equality of mixe partials. Tus equation (4.1) becomes te trivial ientity 0 = 0. Remark 3 Te type of LagrangianL assume in Lemma 1 arises commonly. Some common examples woul be : a particle in, say, a cubic potential fiel were or a simple penulum of lengt l for wic L(q, q) =m q + q3 L(θ, θ) = l θ + gl cos θ.

27 0 Te restriction on te form of Lagrangian can be remove as is sown by te following more general result. Lemma Let L be any sufficiently smoot function of q an q. Construct a new Lagrangian L () (q, q, q) as L () (q, q, q) = L q i qi + L q i qi (4.13) For tis Lagrangian L () te Hamilton s principle gives no information about te evolution of state, i.e te Euler-Lagrange equation for L () is 0=0. Proof Te Euler-Lagrange equation corresponing to te Lagrangian L () is L () t q i L () t q i + L() q i = 0 (4.14) for i =1,...,n. Te first term of tis equation is (summation over repeate inices) L () t q i = L t q i = ( ) L t q j q i qj + L q j q i qj ( = L ) L q j q i qj + t q j q i q j + ( L... q j q j q i + t Te secon term of equation (4.14) is L () t q i = ( L t q i q j qj + L ) q i + L q i q j qj ( = L ) L q i q j qj + t q i q j q j + ( + L... q j L q i q j + t q i q j an te last term of equation (4.14) is L q j q i qj + L q j q i L q j q i qj ) q j (4.15) ) q j (4.16) L () q i = L q i q j qj + L q i q j qj (4.17) Substituting te expressions (4.15), (4.16) an (4.17) in equation (4.14) an using equality of mixe partials one fins tat te left an sie of equation (4.14) is 0 tus completing te proof. Tus ifferentiating te Lagrangian along (q, q) apparently oes not make te iagram 4.commute. Section 4.3 below sows ow tis can be remeie by a ifferent approac to make it work. Tis results in continuous time linearization of te Euler-Lagrange equations. Section 4.4 sows a similar iscrete time linearization of te iscrete Euler-Lagrange equations.

28 1 4.3 Continuous Time Linearization In Section 4.te Lagrangian L was ifferentiate along te acceleration vector ( q, q) to get a new Lagrangian L () wic was linear in q. I sowe tat tis new Lagrangian L () yiels an Euler-Lagrange equation wic is a trivial ientity. Now I will generalize tis iea of ifferentiating L to get a new Lagrangian but tis time te new Lagrangian oes yiel a nontrivial Euler-Lagrange equation. Ten I will escribe an prove te relationsip between ifferentiating te Lagrangian an te corresponing Euler-Lagrange equations. Given L : TQ R, anq Q te ifferential of L at (q, q) T q Q is te map L(q, q) : T (q, q) TQ T R = R. See Section., Definition for etails. In Section 4., I wrote L () (q, q, q) =L(q, q) ( q, q) = L q i (q, q) qi + L q i (q, q) qi (4.18) an foun tat te Euler-Lagrange equation for L () is 0 = 0. In equation (4.18) te ifferentiation is along te acceleration vector ( q, q) wic is in TTQ at (q, q). (Actually to be correct one soul really write suc an element of TTQ as (q, q, q) or even better, as (q, q, q, q)). Te obvious generalization of tis ifferentiation is to ifferentiate along an arbitrary vector in TTQ. Wit tis view let v q T q Q an let w vq T vq TQ. Define L () by L () (w vq ) L(v q ) w vq (4.19) In coorinates one writes v q =(q, q) anw vq =(q, q,u, u) wereu T q Q an u is not necessarily equal to v q =(q, q). Tus in local coorinates equation (4.19) is L () (q, q,u, u) L q i (q, q) ui + L q i (q, q) ui (4.0) We now introuce te notation EL(L) to represent te Euler-Lagrange equations corresponing to a Lagrangian L. Te main result of tis section is tat ifferentiating EL(L) results in te same equations tat are obtaine by first ifferentiating L to get L (), fining EL(L () ) an setting u = q. Tis is summarize in te following Lemma. Lemma 3 Te following iagram commutes EL(L) EL(L () ) u= q (4.1) L L () Proof First I will escribe ow to rea te commutative iagram above. Te up arrows mean apply Hamilton s principle an te result is te Euler-Lagrange equations corresponing to te Lagrangian one starts wit at te bottom. Te rigt arrows inicate ifferentiation. Te top rigt arrow is ifferentiation wrt t. Te bottom rigt arrow is ifferentiating as in (4.0). Te notation u= q in te top rigt corner only makes sense if one arrives at tat corner from te bottom. To be specific ere is ow te various pats troug te iagram soul be rea. Start at bottom left corner. Going up an ten rigt means starting wit te Lagrangian L fin te Euler-Lagrange equation corresponing to it an ten ifferentiate tat equation wrt t. On te oter an te pat going rigt an ten up means, ifferentiate L as in (4.0) wic gives an expression for L () in terms of L. Ten fin te Euler-Lagrange equation

29 for tis Lagrangian. Finally substitute u = q. Te resulting equations will be te same as tose obtaine by te up an rigt pat. I will now give a local coorinate proof of te commutativity of (4.1). I ave alreay ifferentiate EL(L) an te result is given in equation (4.6). Tus te result of ifferentiating EL(L) can be consiere to be te EL(L) itself an te linearize EL(L) inequation (4.6). Let us now see wat one gets by writing EL(L () ) in terms of te original Lagrangian L.SinceL () (q, q, u, u) as inepenent variables q an u it follows tat EL(L () )consists of te two equations L () t q i L() q i =0 (4.) L () t u i L() u i =0 (4.3) Te left an sie of equation (4.) wen written in terms of L is equivalent to ( ) L ( ) L (4.4) Now = t = t q i q j uj + L q j uj ( L q i q j uj + L q i q j uj q i ) q j uj + L q j uj ( L q i q j uj + L q i q j uj ) (4.5) L q i q j uj + u j L t q i q j + L q i q j üj + u j L t q i q j L q i q j uj L q i q j uj. (4.6) L t q i q j = L t q i q j = 3 L q k q i q j qk + 3 L q k q i q j qk + 3 L q k q i q j qk (4.7) 3 L q k q i q j qk (4.8) Tus one equation of EL(L () ) obtaine by substituting (4.7) an (4.8) in (4.6) is L q i q j uj + L q i q j üj + 3 L q k q i q j uj q k + 3 L q k q i q j uj q k + L q i q j uj L q i q j uj =0 3 L q k q i q j uj q k + (4.9) 3 L q k q i q j uj q k Wen one sets u = q one equation of EL(L () ) u= q in terms of L is obtaine as L q i q j qj + L... q j q i q j + 3 L q k q i q j qj q k + 3 L q k q i q j qj q k + L q i q j qj L q i q j qj =0 3 L q k q i q j qj q k + (4.30) 3 L q k q i q j qj q k

30 3 wic except for a rearrangement of terms is te same as te linearize EL(L) inequation (4.6). Te oter equation of EL(L () ) is equation (4.3) wic in terms of L is equivalent to t u i wic is equivalent to ( L q j uj + L ) q j uj ( L u i q j uj + L ) q j uj =0, (4.31) L t q i L =0, (4.3) qi wic is te same as te original EL(L). Tis completes te proof of commutativity of (4.1). Tus I ave sown tat te Euler-Lagrange equation corresponing to L, namely EL(L) can be ifferentiate by ifferentiating L along (u, u) first to get a new Lagrangian L () an ten eriving te Euler-Lagrange equations EL(L () ) corresponing to tis Lagrangian an finally setting u = q. 4.4 Discrete Time Linearization Wit a view of linearizing te iscrete Euler-Lagrange equations I will mimic on te iscrete sie wat was one on te continuous sie. Tus I first obtain L () from L by ifferentiating. Let L : Q Q R be te iscrete Lagrangian. By analogy wit te continuous case one first ifferentiates tis to get L (q, r) :TQ TQ R were (q, r) Q Q an (v q,v r ) T (q,r) (Q Q) = T q Q T r Q. Ten one efines In local coorinates L () (v q,v r ) L (q, r) (v q,v r ) (4.33) L () (q k,u k,q k+1,u k+1 )=L (q k,q k+1 ) (u k,u k+1 ) (4.34) Te use of subscripts k an k + 1 is suggestive of te fact tat q k+1 is te position after one iscrete time step, starting from q k. However, te above efinition of L () is vali for any q 1,u 1,q,u in te appropriate spaces. Note also tat ere q k etc. are vectors, i.e q k =(qk 1,...,qn k ) etc. I will assume, as one by Wenlant an Marsen [0] tat Q V were V is a vector space. Tis is essential if one is going to be averaging an subtracting points in Q. Now I will prove a result on te iscrete sie analogous to wat was one in Lemma 3 on te continuous sie. Let DEL(L ) represent te Discrete Euler-Lagrange equations corresponing to te iscrete Lagrangian L as in (3.1). Ten te following is true Lemma 4 Te following iagram commutes : L L () DEL(L ) DEL(L () ) uk = q k+1 q k (4.35)

31 4 Proof First I must escribe ow to rea te iagram. Te top rigt arrow represents ifferentiation efine in (4.34). Te own arrows represent using te iscrete variational principle to get te iscrete Euler-Lagrange equations. Te bottom rigt arrow as to be given a meaning. I will efine an operation of iscrete ifferentiation as escribe below. In te case of linear equations tis is te same as finite ifferencing. First let s ceck carefully wat DEL(L () ) is. Tis is obtaine by extremizing te iscrete action sum, i.e by extremizing N 1 S () = L () (q k,u k,q k+1,u k+1 ) (4.36) k=0 One extremizes S () over q k an u k for k =1,...,N 1 keeping q 0,q N,u 0 an u N fixe an by setting S () q j k =0, S () u j k =0 for j =1,...,n. Tis yiels te two sets of equations tat comprise te iscrete Euler- Lagrange equations for L (). Tese equations, wic I will call DEL(L() )are L () q j k L () u j k (q k 1,u k 1,q k,u k )+ L() q j (q k,u k,q k+1,u k+1 ) = 0 (4.37) k (q k 1,u k 1,q k,u k )+ L() u j (q k,u k,q k+1,u k+1 ) = 0 (4.38) k for k =1,...,N 1anj =1,...,n. Now I can start te proof of te commutativity of (4.35). To o tis one must write DEL(L () )intermsofl. Consier te first set of equations represente by (4.37). Substitute for L () in terms of L using (4.34). Wit tis substitution, equation (4.37) becomes ( ) q j u i L k 1 q i (q k 1, q k )+u i L k k k 1 qk i (q k 1,q k ) + ( ) u i L k qk i (q k, q k+1 )+u i L k+1 qk+1 i (q k,q k+1 ) =0, (4.39) wic implies q j k u i k 1 L q j (q k 1,q k )+u i L k qi k k 1 q j (q k 1,q k )+ k qi k u i L k q j (q k,q k+1 )+u i L k qi k+1 k q j (q k,q k+1 )=0. (4.40) k qi k+1 Now let s substitute u i k = qi k+1 qi k

32 5 etc. to get q i k qi k 1 L q j k qi k 1 (q k 1,q k )+ qi k+1 qi k q i k+1 qi k L q j k qi k Te above equation represents one equation of L q j k qi k (q k 1,q k )+ (q k,q k+1 )+ qi k+ qi k+1 DEL(L () ) uk = q k+1 q k In terms of L te oter equation, i.e (4.38) becomes u j k ( wic is L u i k 1 qk 1 i Now DEL(L )is ) (q k 1, q k )+u i L k qk i (q k 1,q k ) + ( u i L k qk i (q k, q k+1 )+u i k+1 L q j k u j k. L q j k qi k+1 L q i k+1 (q k,q k+1 ) = 0 (4.41) ) (q k,q k+1 ) =0 (4.4) (q k 1,q k )+ L q j (q k, q k+1 ) = 0 (4.43) k L q j (q k 1,q k )+ L k q j (q k,q k+1 ) = 0 (4.44) k wic is te same as (4.43). Te iscrete time erivative of tis is same as (4.41). Tis erivative is taken just like te usual erivative except tat instea of q k 1 one writes q k q k 1 If an expression is linear in its variables, its iscrete time erivative is te same as its finite ifference. Consier for example an expression f(q) linear in q. In particular let f(q) =a 1 q + a 0. Ten by te above escription of iscrete time erivative, te iscrete time erivative of f(q) is f q 1 q 0 q q0 wic is Note tat tis is te same as q 1 q 0 a 1. f(q 1 ) f(q 0 ),

33 6 te finite ifference of te expression f(q). Tus we ave sown tat DEL(L () ) qk+1 q k consists of two sets of equations. Te first set, equation (4.37), after substituting u k = q k q k 1 is te iscrete time erivative of DEL(L ). Te secon set, equation (4.38) is te same as DEL(L ) wic sows tat te iagram in equation (4.35) commutes. Tus te situation is exactly te same as te continuous case in wic EL(L () ) u= q consiste of two sets of equations te first of wic was te erivative of EL(L) ante secon of wic was te same as EL(L). Linearize Algoritm Te linearize algoritm is generate by extremizing te action sum corresponing to L (). Tis yiels two iscrete Euler-Lagrange Equations, te linearize an te original, just like for te continuous Lagrangian L (). I call te two iscrete Euler- Lagrange equations collectively as DEL(L () ). It was seen tat te DEL(L() )aregivenby (4.37) an (4.38). After making te substitutions u i k =(qi k+1 qi k )/ in tese equations, one gets wat I calle DEL(L () ) qk+1 q k namely equation (4.41). Notice tat (4.41) can be solve for q k+ assuming invertibility of te matrix L q j (q k,q k+1 ) (4.45) k qi k+1 appearing in te equation. For a regular Lagrangian, by efinition, te matrix L q j q i (4.46) is invertible. It is known [16] tat at least for Lagrangians of te form Kinetic Energy - Potential Energy, te invertibility of (4.46) implies te invertibility of (4.45) for small step size. Presumably te same migt be true for a general Lagrangian. Tus given a triple of points (q k 1,q k,q k+1 ) one can solve (4.41) for q k+. Tus an algoritm tat takes te triple to te triple (q k 1,q k,q k+1 ) (q k,q k+1,q k+ )

34 7 as been efine. One can also write te algoritm in terms of te original continuous Lagrangian but I will not o tat. 4.5 Discrete an Continuous Now te continuous an iscrete Lagrangians must be relate in some way an tis is summarize in te following result. Lemma 5 Te following iagram commutes L L () L L () (4.47) Proof Te vertical arrows represent implementation ecisions, i.e it is up to us to coose a iscretization tat will ave te rigt properties. Starting from te top left corner of (4.47) an going own, i.e L L soul be rea as writing L in terms of L. Tis is provie by equation (3.3) written ere as ( qk+1 + q k L (q k,q k+1 )=L, q ) k+1 q k (4.48) were, in L(q, q) tepositionq as been approximate by (q k+1 + q k )/an q by (q k+1 q k )/. A similar implementation ecision must be mae in going from L () (q, q, u, u) to (q k,u k,q k+1,u k+1 ) i.e on going own te rigt arrow. Tis is one by setting L () in L ().Now q =(q k+1 + q k )/ q =(q k+1 q k )/ u =(u k+1 + u k )/ u =(u k+1 u k )/ L () (q k,u k,q k+1,u k+1 )= L () ( qk+1 + q k, q k+1 q k, u k+1 + u k, u ) k+1 u k (4.49) L () (q k,u k,q k+1,u k+1 )= ui k+1 + ( ui k L qk+1 + q k q i, q ) k+1 q k + ui k+1 ui k L q i ( qk+1 + q k, q ) k+1 q k (4.50) were te last equality follows from (4.0). Equation (4.50) represents starting from top left, going rigt an ten bottom. Tis must be compare wit te oter pat of going

35 8 own from top left an ten going rigt. To o tis one must use equation (4.34). Here, L (q k,q k+1 )= [ 1 L q 1 1 L q 1,...,1 L q n 1 L q n, 1 L q L q 1,...,1 L q n + 1 ] L q n (4.51) Tus L (q k,q k+1 ) (u k,u k+1 )= u i k ( 1 L q i 1 ) ( L 1 q i + u i L k+1 q i + 1 ) L q i (4.5) were i = 1,...,n wic implies tat L () (q k,u k,q k+1,u k+1 )= ui k+1 + ui k L q i + ui k+1 ui k L q i (4.53) were te partials are evaluate at ((q k+1 + q k )/, (q k+1 q k )/). But (4.53) is te same as (4.50) wic completes te proof of commutativity of (4.47). Te previous tree lemmas can be summarize in one iagram as given by te following teorem. Teorem 1 Te following iagram commutes EL(L) EL(L () ) u= q L L () L L () DEL(L ) DEL(L () ) uk = q k+1 q k (4.54) Proof Tis follows from te previous tree lemmas by pasting te corresponing commutative iagrams togeter. 4.6 Linearization Summary I summarize te linearization process by giving te various Euler-Lagrange equations in one place ere. In te following equations of tis section, i, j =1,...,n an k =1,...,N 1, were n is te imension of te configurationmanifol an 0,...,N represents iscrete time.

36 9 Euler Lagrange equations EL(L) for L(q, q) : L t q i L q i = 0 (4.55) Euler Lagrange equations EL(L () ) for L () (q, q,u, u) : L () t q i L() q i = 0 (4.56) L () t u i L() u i = 0 (4.57) EL(L () ) u= q in terms of L : L q i q j qj + L... q j q i q j + 3 L q k q i q j qj q k + 3 L q k q i q j qj q k + 3 L q k q i q j qj q k + (4.58) 3 L q k q i q j qj q k L q i q j qj L q i q j qj =0 L t q i L q i = 0 (4.59) Equation (4.58) is te erivative of equation (4.55). Discrete Euler-Lagrange equations DEL(L ) for L (q k,q k+1 ): L q j k (q k 1,q k )+ L q j (q k,q k+1 ) = 0 (4.60) k Discrete Euler-Lagrange equations DEL(L () ) for L() (q k,u k,q k+1,u k+1 ) L () q j k L () u j k (q k 1,u k 1,q k,u k )+ L() q j (q k,u k,q k+1,u k+1 ) = 0 (4.61) k (q k 1,u k 1,q k,u k )+ L() u j (q k,u k,q k+1,u k+1 ) = 0 (4.6) k

37 30 DEL(L () ) qk+1 q k in terms of L : qk i qi k 1 L q j k qi k 1 qk+1 i qi k L q j k qi k L q j k (q k 1,q k )+ qi k+1 qi k (q k,q k+1 )+ qi k+ qi k+1 L q j (q k 1,q k )+ (4.63) k qi k L q j k qi k+1 (q k,q k+1 )=0 (q k 1,q k )+ L q j (q k,q k+1 ) = 0 (4.64) k Equation (4.63) is te iscrete time erivative of equation (4.60). 4.7 Invariance Properties To see wy te proofs of invariance properties given by Wenlant an Marsen [0] will still work, it is elpful to consier te following picture. L : TQ R L : Q Q R L () : TTQ R L () : TQ TQ R Te left column sows te usual continuous Lagrangian an te iscrete Lagrangians use by Wenlant an Marsen [0]. It was sown by Wenlant an Marsen [0] tat te algoritm obtaine from L is a symplectic-momentum algoritm. On te rigt column are te Lagrangian an iscrete Lagrangian I ave obtaine. Note tat te rigt column is te same as te left one wit Q replace by TQ. Tus all te invariance proofs of te left column still apply. Tis means tat EL(L () ) as symplectic flow an DEL(L () ) in equations (4.37) an (4.38) is a symplectic integrator. I erive te te symplectic form on TQ TQin subsection below. One can ceck explicitly tat tis symplectic form, given as equation (4.68) below, is preserve by te algoritm DEL(L () ) toug tis also follows witout cecking from te proof in Wenlant an Marsen [0] as inicate above. But we must ceck tat similar invariance properties remain true after we set u = q in EL(L () )oru k =(q k+1 q k )/ in DEL(L () ). Tis will be one in subsection 4.7.for DEL(L () ) Symplectic form on TQ TQ I will efine te symplectic form on TQ TQ by pulling back te canonical -form from T TQ. Te pull back will be one uner a fiber erivative of L () were, as efine earlier, L () : TQ TQ R. Tis fiber erivative will be efine by analogy to te fiber erivative for L : Q Q R, efine by Wenlant an Marsen in [0]. Tat fiber erivative can be efine by taking a parameterize curve (q(t),q 1 ) were te secon component is fixe an were q(0) = q 0 an q(0) = u 0. Tis is a curve in Q {q 1 } Q Q an can be ientifie wit te curve q(t). Tus it represents te vector (q 0,u 0 ) T q0 Q. Te fiber erivative FL : Q Q T Q is ten efine as FL (q 0,q 1 ) (q 0,u 0 )= t L (q(t),q 1 ) (4.65) t=0

38 31 I will efine te fiber erivative FL () : TQ TQ T TQ by analogy to tis. But first I note tat te following steps are neee to efine te symplectic form on TQ TQ: i. Define te fiber erivative FL () : TQ TQ T TQ, ii. Compute te tangent map of FL (),namely T ((q0,u 0),(q 1,u 1))FL () : T (q0,u 0)TQ T (q1,u 1)TQ T () FL ((q0,u0),(q1,u1))t TQ iii. Write te local coorinate expression for te canonical one form Θ on T TQ, iv. Define te symplectic form on TQ TQ as Ω = (FL () Θ) were FL () te pullback uner te fiber erivative FL (). means Step 1 :SinceL () maps TQ TQ R, te fiber erivative FL () maps TQ TQ T TQ. Note tat FL () evaluateatapointintq TQis a point in T TQso FL () can be efine by efining ow it acts on a vector in TTQ. By analogy to te efinition of fiber erivative FL given above, let ((q(t),u(t)), (q 1,u 1 )) be a parameterize curve from an open interval of R incluing 0 into TQ {(q 1,u 1 )} TQ TQ suc tat q(0) = q 0 u(0) = u 0 q(0) = e u(0) = f Te curve ((q(t),u(t)), (q 1,u 1 )) can be ientifie wit te curve (q(t),u(t)) an tis curve ten represents te vector ((q 0,u 0 ),e,f) T (q0,u 0)TQ. Te fiber erivative FL () evaluate at ((q 0,u 0 ), (q 1,u 1 )) TQ TQcanbeefinebyowitsactson((q 0,u 0 ),e,f) T (q0,u 0)TQ as follows: FL () ((q 0,u 0 ), (q 1,u 1 )) ((q 0,u 0 ),e,f)= t L () ((q(t),u(t)), (q 1,u 1 )) t=0 Note te analogy wit te efinition in equation(4.65). Wen te above calculation is carrie out, in matrix notation it results in te following efinition FL () ((q 0,u 0 ), (q 1,u 1 )) = ((q 0,u 0 ), D 1 L () ((q 0,u 0 ), (q 1,u 1 )), D L () ((q 0,u 0 ), (q 1,u 1 ))) (4.66) Note tat FL () maps te point ((q 0,u 0 ), (q 1,u 1 )) TQ TQtoacovectorinT(q TQ 0,u 0) base at (q 0,u 0 ) so te first (q 0,u 0 ) tat appears on te rigt an sie of equation(4.66) is te base point of te covector.

Chapter Primer on Differentiation

Chapter Primer on Differentiation Capter 0.01 Primer on Differentiation After reaing tis capter, you soul be able to: 1. unerstan te basics of ifferentiation,. relate te slopes of te secant line an tangent line to te erivative of a function,.

More information

1 ode.mcd. Find solution to ODE dy/dx=f(x,y). Instructor: Nam Sun Wang

1 ode.mcd. Find solution to ODE dy/dx=f(x,y). Instructor: Nam Sun Wang Fin solution to ODE /=f(). Instructor: Nam Sun Wang oe.mc Backgroun. Wen a sstem canges wit time or wit location, a set of ifferential equations tat contains erivative terms "/" escribe suc a namic sstem.

More information

arxiv: v1 [math.na] 8 Feb 2017

arxiv: v1 [math.na] 8 Feb 2017 Variational iscretization of te nonequilibrium termoynamics of simple systems arxiv:17.594v1 [mat.na] 8 Feb 17 François Gay-Balmaz Hiroaki Yosimura CNRS - LMD - IPSL Scool of Science an Engineering Ecole

More information

Nonholonomic Integrators

Nonholonomic Integrators Nonolonomic Integrators J. Cortés an S. Martínez Laboratory of Dynamical Systems, Mecanics an Control, Instituto e Matemáticas y Física Funamental, CSIC, Serrano 13, 8006 Mari, SPAIN E-mail: j.cortes@imaff.cfmac.csic.es,

More information

Examining Geometric Integration for Propagating Orbit Trajectories with Non-Conservative Forcing

Examining Geometric Integration for Propagating Orbit Trajectories with Non-Conservative Forcing Examining Geometric Integration for Propagating Orbit Trajectories with Non-Conservative Forcing Course Project for CDS 05 - Geometric Mechanics John M. Carson III California Institute of Technology June

More information

SECTION 2.1 BASIC CALCULUS REVIEW

SECTION 2.1 BASIC CALCULUS REVIEW Tis capter covers just te very basics of wat you will nee moving forwar onto te subsequent capters. Tis is a summary capter, an will not cover te concepts in-ept. If you ve never seen calculus before,

More information

f(x + h) f(x) f (x) = lim

f(x + h) f(x) f (x) = lim Introuction 4.3 Some Very Basic Differentiation Formulas If a ifferentiable function f is quite simple, ten it is possible to fin f by using te efinition of erivative irectly: f () 0 f( + ) f() However,

More information

0.1 Differentiation Rules

0.1 Differentiation Rules 0.1 Differentiation Rules From our previous work we ve seen tat it can be quite a task to calculate te erivative of an arbitrary function. Just working wit a secon-orer polynomial tings get pretty complicate

More information

Differentiation Rules c 2002 Donald Kreider and Dwight Lahr

Differentiation Rules c 2002 Donald Kreider and Dwight Lahr Dierentiation Rules c 00 Donal Kreier an Dwigt Lar Te Power Rule is an example o a ierentiation rule. For unctions o te orm x r, were r is a constant real number, we can simply write own te erivative rater

More information

Rules of Differentiation

Rules of Differentiation LECTURE 2 Rules of Differentiation At te en of Capter 2, we finally arrive at te following efinition of te erivative of a function f f x + f x x := x 0 oing so only after an extene iscussion as wat te

More information

Discrete Hamilton Jacobi Theory and Discrete Optimal Control

Discrete Hamilton Jacobi Theory and Discrete Optimal Control 49th IEEE Conference on Decision an Control December 15-17, 2010 Hilton Atlanta Hotel, Atlanta, GA, USA Discrete Hamilton Jacobi Theory an Discrete Optimal Control Tomoi Ohsawa, Anthony M. Bloch, an Melvin

More information

Hy Ex Ez. Ez i+1/2,j+1. Ex i,j+1/2. γ z. Hy i+1/2,j+1/2. Ex i+1,j+1/2. Ez i+1/2,j

Hy Ex Ez. Ez i+1/2,j+1. Ex i,j+1/2. γ z. Hy i+1/2,j+1/2. Ex i+1,j+1/2. Ez i+1/2,j IEEE TRANSACTIONS ON, VOL. XX, NO. Y, MONTH 000 100 Moeling Dielectric Interfaces in te FDTD-Meto: A Comparative Stuy C. H. Teng, A. Ditkowski, J. S. Hestaven Abstract In tis paper, we present special

More information

Differential Calculus: Differentiation (First Principles, Rules) and Sketching Graphs (Grade 12) *

Differential Calculus: Differentiation (First Principles, Rules) and Sketching Graphs (Grade 12) * OpenStax-CNX moule: m39313 1 Differential Calculus: Differentiation (First Principles, Rules) an Sketcing Graps (Grae 12) * Free Hig Scool Science Texts Project Tis work is prouce by OpenStax-CNX an license

More information

160 Chapter 3: Differentiation

160 Chapter 3: Differentiation 3. Differentiation Rules 159 3. Differentiation Rules Tis section introuces a few rules tat allow us to ifferentiate a great variety of functions. By proving tese rules ere, we can ifferentiate functions

More information

Lagrangian and Hamiltonian Mechanics

Lagrangian and Hamiltonian Mechanics Lagrangian an Hamiltonian Mechanics.G. Simpson, Ph.. epartment of Physical Sciences an Engineering Prince George s Community College ecember 5, 007 Introuction In this course we have been stuying classical

More information

Derivatives of trigonometric functions

Derivatives of trigonometric functions Derivatives of trigonometric functions 2 October 207 Introuction Toay we will ten iscuss te erivates of te si stanar trigonometric functions. Of tese, te most important are sine an cosine; te erivatives

More information

does NOT exist. WHAT IF THE NUMBER X APPROACHES CANNOT BE PLUGGED INTO F(X)??????

does NOT exist. WHAT IF THE NUMBER X APPROACHES CANNOT BE PLUGGED INTO F(X)?????? MATH 000 Miterm Review.3 Te it of a function f ( ) L Tis means tat in a given function, f(), as APPROACHES c, a constant, it will equal te value L. Tis is c only true if f( ) f( ) L. Tat means if te verticle

More information

In Leibniz notation, we write this rule as follows. DERIVATIVE OF A CONSTANT FUNCTION. For n 4 we find the derivative of f x x 4 as follows: lim

In Leibniz notation, we write this rule as follows. DERIVATIVE OF A CONSTANT FUNCTION. For n 4 we find the derivative of f x x 4 as follows: lim .1 DERIVATIVES OF POLYNOIALS AND EXPONENTIAL FUNCTIONS c =c slope=0 0 FIGURE 1 Te grap of ƒ=c is te line =c, so fª()=0. In tis section we learn ow to ifferentiate constant functions, power functions, polnomials,

More information

1 Lecture 13: The derivative as a function.

1 Lecture 13: The derivative as a function. 1 Lecture 13: Te erivative as a function. 1.1 Outline Definition of te erivative as a function. efinitions of ifferentiability. Power rule, erivative te exponential function Derivative of a sum an a multiple

More information

Lecture 2 Lagrangian formulation of classical mechanics Mechanics

Lecture 2 Lagrangian formulation of classical mechanics Mechanics Lecture Lagrangian formulation of classical mechanics 70.00 Mechanics Principle of stationary action MATH-GA To specify a motion uniquely in classical mechanics, it suffices to give, at some time t 0,

More information

arxiv: v1 [math.na] 18 Nov 2011

arxiv: v1 [math.na] 18 Nov 2011 MIMETIC FRAMEWORK ON CURVILINEAR QUADRILATERALS OF ARBITRARY ORDER JASPER KREEFT, ARTUR PALHA, AND MARC GERRITSMA arxiv:1111.4304v1 [mat.na] 18 Nov 2011 Abstract. In tis paper iger orer mimetic iscretizations

More information

Function Composition and Chain Rules

Function Composition and Chain Rules Function Composition an Cain Rules James K. Peterson Department of Biological Sciences an Department of Matematical Sciences Clemson University November 2, 2018 Outline Function Composition an Continuity

More information

Continuous measurements: partial selection

Continuous measurements: partial selection 0 May 00 Pysics Letters A 97 00 300 306 wwwelseviercom/locate/pla Continuous measurements: partial selection YuA Rembovsky Department of Pysics, MV Lomonosov State University, Moscow 119899, Russia Receive

More information

(a 1 m. a n m = < a 1/N n

(a 1 m. a n m = < a 1/N n Notes on a an log a Mat 9 Fall 2004 Here is an approac to te eponential an logaritmic functions wic avois any use of integral calculus We use witout proof te eistence of certain limits an assume tat certain

More information

Copyright c 2008 Kevin Long

Copyright c 2008 Kevin Long Lecture 4 Numerical solution of initial value problems Te metods you ve learned so far ave obtained closed-form solutions to initial value problems. A closedform solution is an explicit algebriac formula

More information

Physics 5153 Classical Mechanics. The Virial Theorem and The Poisson Bracket-1

Physics 5153 Classical Mechanics. The Virial Theorem and The Poisson Bracket-1 Physics 5153 Classical Mechanics The Virial Theorem an The Poisson Bracket 1 Introuction In this lecture we will consier two applications of the Hamiltonian. The first, the Virial Theorem, applies to systems

More information

Differential Calculus Definitions, Rules and Theorems

Differential Calculus Definitions, Rules and Theorems Precalculus Review Differential Calculus Definitions, Rules an Teorems Sara Brewer, Alabama Scool of Mat an Science Functions, Domain an Range f: X Y a function f from X to Y assigns to eac x X a unique

More information

Notes on wavefunctions II: momentum wavefunctions

Notes on wavefunctions II: momentum wavefunctions Notes on wavefunctions II: momentum wavefunctions and uncertainty Te state of a particle at any time is described by a wavefunction ψ(x). Tese wavefunction must cange wit time, since we know tat particles

More information

DISCRETE MECHANICS AND OPTIMAL CONTROL: AN ANALYSIS

DISCRETE MECHANICS AND OPTIMAL CONTROL: AN ANALYSIS ESAIM: COCV 17 (211) 322 352 DOI: 1.151/cocv/2112 ESAIM: Control, Optimisation an Calculus of Variations www.esaim-cocv.org DISCRETE MECHANICS AND OPTIMAL CONTROL: AN ANALYSIS Sina Ober-Blöbaum 1, Oliver

More information

30 is close to t 5 = 15.

30 is close to t 5 = 15. Limits Definition 1 (Limit). If te values f(x) of a function f : A B get very close to a specific, unique number L wen x is very close to, but not necessarily equal to, a limit point c, we say te limit

More information

Stability of equilibrium under constraints: Role of. second-order constrained derivatives

Stability of equilibrium under constraints: Role of. second-order constrained derivatives arxiv:78.694 Stability of equilibrium uner constraints: Role of secon-orer constraine erivatives amás Gál Department of eoretical Pysics University of Debrecen 4 Debrecen Hungary Email aress: galt@pys.unieb.u

More information

2.8 The Derivative as a Function

2.8 The Derivative as a Function .8 Te Derivative as a Function Typically, we can find te derivative of a function f at many points of its domain: Definition. Suppose tat f is a function wic is differentiable at every point of an open

More information

Derivatives. if such a limit exists. In this case when such a limit exists, we say that the function f is differentiable.

Derivatives. if such a limit exists. In this case when such a limit exists, we say that the function f is differentiable. Derivatives 3. Derivatives Definition 3. Let f be a function an a < b be numbers. Te average rate of cange of f from a to b is f(b) f(a). b a Remark 3. Te average rate of cange of a function f from a to

More information

Shift Theorem Involving the Exponential of a Sum of Non-Commuting Operators in Path Integrals. Abstract

Shift Theorem Involving the Exponential of a Sum of Non-Commuting Operators in Path Integrals. Abstract YITP-SB-6-4 Sift Teorem Involving te Exponential of a Sum of Non-Commuting Operators in Pat Integrals Fre Cooper 1, an Gouranga C. Nayak 2, 1 Pysics Division, National Science Founation, Arlington VA 2223

More information

Consider a function f we ll specify which assumptions we need to make about it in a minute. Let us reformulate the integral. 1 f(x) dx.

Consider a function f we ll specify which assumptions we need to make about it in a minute. Let us reformulate the integral. 1 f(x) dx. Capter 2 Integrals as sums and derivatives as differences We now switc to te simplest metods for integrating or differentiating a function from its function samples. A careful study of Taylor expansions

More information

Table of Common Derivatives By David Abraham

Table of Common Derivatives By David Abraham Prouct an Quotient Rules: Table of Common Derivatives By Davi Abraham [ f ( g( ] = [ f ( ] g( + f ( [ g( ] f ( = g( [ f ( ] g( g( f ( [ g( ] Trigonometric Functions: sin( = cos( cos( = sin( tan( = sec

More information

Lecture Notes Di erentiating Trigonometric Functions page 1

Lecture Notes Di erentiating Trigonometric Functions page 1 Lecture Notes Di erentiating Trigonometric Functions age (sin ) 7 sin () sin 8 cos 3 (tan ) sec tan + 9 tan + 4 (cot ) csc cot 0 cot + 5 sin (sec ) cos sec tan sec jj 6 (csc ) sin csc cot csc jj c Hiegkuti,

More information

ch (for some fixed positive number c) reaching c

ch (for some fixed positive number c) reaching c GSTF Journal of Matematics Statistics and Operations Researc (JMSOR) Vol. No. September 05 DOI 0.60/s4086-05-000-z Nonlinear Piecewise-defined Difference Equations wit Reciprocal and Cubic Terms Ramadan

More information

Recent Progress in the Integration of Poisson Systems via the Mid Point Rule and Runge Kutta Algorithm

Recent Progress in the Integration of Poisson Systems via the Mid Point Rule and Runge Kutta Algorithm Recent Progress in te Integration of Poisson Systems via te Mid Point Rule and Runge Kutta Algoritm Klaus Bucner, Mircea Craioveanu and Mircea Puta Abstract Some recent progress in te integration of Poisson

More information

TRAJECTORY TRACKING FOR FULLY ACTUATED MECHANICAL SYSTEMS

TRAJECTORY TRACKING FOR FULLY ACTUATED MECHANICAL SYSTEMS TRAJECTORY TRACKING FOR FULLY ACTUATED MECHANICAL SYSTEMS Francesco Bullo Richar M. Murray Control an Dynamical Systems California Institute of Technology Pasaena, CA 91125 Fax : + 1-818-796-8914 email

More information

Exam 1 Review Solutions

Exam 1 Review Solutions Exam Review Solutions Please also review te old quizzes, and be sure tat you understand te omework problems. General notes: () Always give an algebraic reason for your answer (graps are not sufficient),

More information

Math 1B, lecture 8: Integration by parts

Math 1B, lecture 8: Integration by parts Math B, lecture 8: Integration by parts Nathan Pflueger 23 September 2 Introuction Integration by parts, similarly to integration by substitution, reverses a well-known technique of ifferentiation an explores

More information

The Verlet Algorithm for Molecular Dynamics Simulations

The Verlet Algorithm for Molecular Dynamics Simulations Cemistry 380.37 Fall 2015 Dr. Jean M. Standard November 9, 2015 Te Verlet Algoritm for Molecular Dynamics Simulations Equations of motion For a many-body system consisting of N particles, Newton's classical

More information

Differentiation in higher dimensions

Differentiation in higher dimensions Capter 2 Differentiation in iger dimensions 2.1 Te Total Derivative Recall tat if f : R R is a 1-variable function, and a R, we say tat f is differentiable at x = a if and only if te ratio f(a+) f(a) tends

More information

The Effects of Mutual Coupling and Transformer Connection Type on Frequency Response of Unbalanced Three Phase Electrical Distribution System

The Effects of Mutual Coupling and Transformer Connection Type on Frequency Response of Unbalanced Three Phase Electrical Distribution System Energy an Power Engineering, 00,, 8-47 oi:0.46/epe.00.405 Publise Online November 00 (ttp://www.scirp.org/journal/epe) Te Effects of Mutual Coupling an Transformer Connection Type on Frequency Response

More information

Numerical Differentiation

Numerical Differentiation Numerical Differentiation Finite Difference Formulas for te first derivative (Using Taylor Expansion tecnique) (section 8.3.) Suppose tat f() = g() is a function of te variable, and tat as 0 te function

More information

Darboux s theorem and symplectic geometry

Darboux s theorem and symplectic geometry Darboux s theorem an symplectic geometry Liang, Feng May 9, 2014 Abstract Symplectic geometry is a very important branch of ifferential geometry, it is a special case of poisson geometry, an coul also

More information

Switching Time Optimization in Discretized Hybrid Dynamical Systems

Switching Time Optimization in Discretized Hybrid Dynamical Systems Switching Time Optimization in Discretize Hybri Dynamical Systems Kathrin Flaßkamp, To Murphey, an Sina Ober-Blöbaum Abstract Switching time optimization (STO) arises in systems that have a finite set

More information

Preface. Here are a couple of warnings to my students who may be here to get a copy of what happened on a day that you missed.

Preface. Here are a couple of warnings to my students who may be here to get a copy of what happened on a day that you missed. Preface Here are my online notes for my course tat I teac ere at Lamar University. Despite te fact tat tese are my class notes, tey sould be accessible to anyone wanting to learn or needing a refreser

More information

CHAPTER 1 : DIFFERENTIABLE MANIFOLDS. 1.1 The definition of a differentiable manifold

CHAPTER 1 : DIFFERENTIABLE MANIFOLDS. 1.1 The definition of a differentiable manifold CHAPTER 1 : DIFFERENTIABLE MANIFOLDS 1.1 The efinition of a ifferentiable manifol Let M be a topological space. This means that we have a family Ω of open sets efine on M. These satisfy (1), M Ω (2) the

More information

1 The concept of limits (p.217 p.229, p.242 p.249, p.255 p.256) 1.1 Limits Consider the function determined by the formula 3. x since at this point

1 The concept of limits (p.217 p.229, p.242 p.249, p.255 p.256) 1.1 Limits Consider the function determined by the formula 3. x since at this point MA00 Capter 6 Calculus and Basic Linear Algebra I Limits, Continuity and Differentiability Te concept of its (p.7 p.9, p.4 p.49, p.55 p.56). Limits Consider te function determined by te formula f Note

More information

and from it produce the action integral whose variation we set to zero:

and from it produce the action integral whose variation we set to zero: Lagrange Multipliers Monay, 6 September 01 Sometimes it is convenient to use reunant coorinates, an to effect the variation of the action consistent with the constraints via the metho of Lagrange unetermine

More information

Implicit-explicit variational integration of highly oscillatory problems

Implicit-explicit variational integration of highly oscillatory problems Implicit-explicit variational integration of igly oscillatory problems Ari Stern Structured Integrators Worksop April 9, 9 Stern, A., and E. Grinspun. Multiscale Model. Simul., to appear. arxiv:88.39 [mat.na].

More information

Short Intro to Coordinate Transformation

Short Intro to Coordinate Transformation Short Intro to Coorinate Transformation 1 A Vector A vector can basically be seen as an arrow in space pointing in a specific irection with a specific length. The following problem arises: How o we represent

More information

Nöether s Theorem Under the Legendre Transform by Jonathan Herman

Nöether s Theorem Under the Legendre Transform by Jonathan Herman Nöether s Theorem Uner the Legenre Transform by Jonathan Herman A research paper presente to the University of Waterloo in fulfilment of the research paper requirement for the egree of Master of Mathematics

More information

Implicit Differentiation

Implicit Differentiation Implicit Differentiation Thus far, the functions we have been concerne with have been efine explicitly. A function is efine explicitly if the output is given irectly in terms of the input. For instance,

More information

6 General properties of an autonomous system of two first order ODE

6 General properties of an autonomous system of two first order ODE 6 General properties of an autonomous system of two first orer ODE Here we embark on stuying the autonomous system of two first orer ifferential equations of the form ẋ 1 = f 1 (, x 2 ), ẋ 2 = f 2 (, x

More information

7.1 Using Antiderivatives to find Area

7.1 Using Antiderivatives to find Area 7.1 Using Antiderivatives to find Area Introduction finding te area under te grap of a nonnegative, continuous function f In tis section a formula is obtained for finding te area of te region bounded between

More information

Quantum Numbers and Rules

Quantum Numbers and Rules OpenStax-CNX module: m42614 1 Quantum Numbers and Rules OpenStax College Tis work is produced by OpenStax-CNX and licensed under te Creative Commons Attribution License 3.0 Abstract Dene quantum number.

More information

Functional Analysis Techniques as a Mathematical tools in Wavelets Analysis

Functional Analysis Techniques as a Mathematical tools in Wavelets Analysis Saer El-Sater, et. al., JOURNAL OF PHYSICS VOL. NO. Feb. (0) PP. 0-7 Functional Analysis Tecniques as a Matematical tools in Wavelets Analysis Saer El-sater, Ramaan Sale Tantawi an A. Abel-afiez Abstract

More information

5 Ordinary Differential Equations: Finite Difference Methods for Boundary Problems

5 Ordinary Differential Equations: Finite Difference Methods for Boundary Problems 5 Ordinary Differential Equations: Finite Difference Metods for Boundary Problems Read sections 10.1, 10.2, 10.4 Review questions 10.1 10.4, 10.8 10.9, 10.13 5.1 Introduction In te previous capters we

More information

Combining functions: algebraic methods

Combining functions: algebraic methods Combining functions: algebraic metods Functions can be added, subtracted, multiplied, divided, and raised to a power, just like numbers or algebra expressions. If f(x) = x 2 and g(x) = x + 2, clearly f(x)

More information

NUMERICAL DIFFERENTIATION. James T. Smith San Francisco State University. In calculus classes, you compute derivatives algebraically: for example,

NUMERICAL DIFFERENTIATION. James T. Smith San Francisco State University. In calculus classes, you compute derivatives algebraically: for example, NUMERICAL DIFFERENTIATION James T Smit San Francisco State University In calculus classes, you compute derivatives algebraically: for example, f( x) = x + x f ( x) = x x Tis tecnique requires your knowing

More information

Chapter 2 Ising Model for Ferromagnetism

Chapter 2 Ising Model for Ferromagnetism Capter Ising Model for Ferromagnetism Abstract Tis capter presents te Ising model for ferromagnetism, wic is a standard simple model of a pase transition. Using te approximation of mean-field teory, te

More information

This file is /conf/snippets/setheader.pg you can use it as a model for creating files which introduce each problem set.

This file is /conf/snippets/setheader.pg you can use it as a model for creating files which introduce each problem set. Yanimov Almog WeBWorK assignment number Sections 3. 3.2 is ue : 08/3/207 at 03:2pm CDT. Te (* replace wit url for te course ome page *) for te course contains te syllabus, graing policy an oter information.

More information

Math Notes on differentials, the Chain Rule, gradients, directional derivative, and normal vectors

Math Notes on differentials, the Chain Rule, gradients, directional derivative, and normal vectors Math 18.02 Notes on ifferentials, the Chain Rule, graients, irectional erivative, an normal vectors Tangent plane an linear approximation We efine the partial erivatives of f( xy, ) as follows: f f( x+

More information

The derivative function

The derivative function Roberto s Notes on Differential Calculus Capter : Definition of derivative Section Te derivative function Wat you need to know already: f is at a point on its grap and ow to compute it. Wat te derivative

More information

221A Lecture Notes Notes on Classica Mechanics I

221A Lecture Notes Notes on Classica Mechanics I 1A Lecture Notes Notes on Classica Mechanics I 1 Precursor: Fermat s Principle in Geometric Optics In geometric optics, you talk about how light rays go. In homogeneous meiums, the light rays go straight.

More information

Math 242: Principles of Analysis Fall 2016 Homework 7 Part B Solutions

Math 242: Principles of Analysis Fall 2016 Homework 7 Part B Solutions Mat 22: Principles of Analysis Fall 206 Homework 7 Part B Solutions. Sow tat f(x) = x 2 is not uniformly continuous on R. Solution. Te equation is equivalent to f(x) = 0 were f(x) = x 2 sin(x) 3. Since

More information

3.4 Worksheet: Proof of the Chain Rule NAME

3.4 Worksheet: Proof of the Chain Rule NAME Mat 1170 3.4 Workseet: Proof of te Cain Rule NAME Te Cain Rule So far we are able to differentiate all types of functions. For example: polynomials, rational, root, and trigonometric functions. We are

More information

DIFFERENTIAL GEOMETRY, LECTURE 15, JULY 10

DIFFERENTIAL GEOMETRY, LECTURE 15, JULY 10 DIFFERENTIAL GEOMETRY, LECTURE 15, JULY 10 5. Levi-Civita connection From now on we are intereste in connections on the tangent bunle T X of a Riemanninam manifol (X, g). Out main result will be a construction

More information

Auctions, Matching and the Law of Aggregate Demand

Auctions, Matching and the Law of Aggregate Demand Auctions, Matcing an te Law of Aggregate eman Jon William atfiel an Paul Milgrom 1 February 12, 2004 Abstract. We evelop a moel of matcing wit contracts wic incorporates, as special cases, te college amissions

More information

THE IDEA OF DIFFERENTIABILITY FOR FUNCTIONS OF SEVERAL VARIABLES Math 225

THE IDEA OF DIFFERENTIABILITY FOR FUNCTIONS OF SEVERAL VARIABLES Math 225 THE IDEA OF DIFFERENTIABILITY FOR FUNCTIONS OF SEVERAL VARIABLES Mat 225 As we ave seen, te definition of derivative for a Mat 111 function g : R R and for acurveγ : R E n are te same, except for interpretation:

More information

G j dq i + G j. q i. = a jt. and

G j dq i + G j. q i. = a jt. and Lagrange Multipliers Wenesay, 8 September 011 Sometimes it is convenient to use reunant coorinates, an to effect the variation of the action consistent with the constraints via the metho of Lagrange unetermine

More information

Lagrangian and Hamiltonian Dynamics

Lagrangian and Hamiltonian Dynamics Lagrangian an Hamiltonian Dynamics Volker Perlick (Lancaster University) Lecture 1 The Passage from Newtonian to Lagrangian Dynamics (Cockcroft Institute, 22 February 2010) Subjects covere Lecture 2: Discussion

More information

SYMPLECTIC GEOMETRY: LECTURE 3

SYMPLECTIC GEOMETRY: LECTURE 3 SYMPLECTIC GEOMETRY: LECTURE 3 LIAT KESSLER 1. Local forms Vector fiels an the Lie erivative. A vector fiel on a manifol M is a smooth assignment of a vector tangent to M at each point. We think of M as

More information

c 2011 Society for Industrial and Applied Mathematics

c 2011 Society for Industrial and Applied Mathematics SIAM J. CONTROL OPTIM. Vol. 49, No. 4, pp. 1829 1856 c 211 Society for Inustrial an Applie Mathematics DISCRETE HAMILTON JACOBI THEORY TOMOKI OHSAWA, ANTHONY M. BLOCH, AND MELVIN LEOK Abstract. We evelop

More information

4. Important theorems in quantum mechanics

4. Important theorems in quantum mechanics TFY4215 Kjemisk fysikk og kvantemekanikk - Tillegg 4 1 TILLEGG 4 4. Important theorems in quantum mechanics Before attacking three-imensional potentials in the next chapter, we shall in chapter 4 of this

More information

The Principle of Least Action and Designing Fiber Optics

The Principle of Least Action and Designing Fiber Optics University of Southampton Department of Physics & Astronomy Year 2 Theory Labs The Principle of Least Action an Designing Fiber Optics 1 Purpose of this Moule We will be intereste in esigning fiber optic

More information

Introduction to Derivatives

Introduction to Derivatives Introduction to Derivatives 5-Minute Review: Instantaneous Rates and Tangent Slope Recall te analogy tat we developed earlier First we saw tat te secant slope of te line troug te two points (a, f (a))

More information

HOMEWORK HELP 2 FOR MATH 151

HOMEWORK HELP 2 FOR MATH 151 HOMEWORK HELP 2 FOR MATH 151 Here we go; te second round of omework elp. If tere are oters you would like to see, let me know! 2.4, 43 and 44 At wat points are te functions f(x) and g(x) = xf(x)continuous,

More information

Physically Based Modeling: Principles and Practice Implicit Methods for Differential Equations

Physically Based Modeling: Principles and Practice Implicit Methods for Differential Equations Pysically Based Modeling: Principles and Practice Implicit Metods for Differential Equations David Baraff Robotics Institute Carnegie Mellon University Please note: Tis document is 997 by David Baraff

More information

Vectors in two dimensions

Vectors in two dimensions Vectors in two imensions Until now, we have been working in one imension only The main reason for this is to become familiar with the main physical ieas like Newton s secon law, without the aitional complication

More information

SECTION 1.10: DIFFERENCE QUOTIENTS LEARNING OBJECTIVES

SECTION 1.10: DIFFERENCE QUOTIENTS LEARNING OBJECTIVES (Section.0: Difference Quotients).0. SECTION.0: DIFFERENCE QUOTIENTS LEARNING OBJECTIVES Define average rate of cange (and average velocity) algebraically and grapically. Be able to identify, construct,

More information

Chapter 2 Lagrangian Modeling

Chapter 2 Lagrangian Modeling Chapter 2 Lagrangian Moeling The basic laws of physics are use to moel every system whether it is electrical, mechanical, hyraulic, or any other energy omain. In mechanics, Newton s laws of motion provie

More information

Introduction to the Vlasov-Poisson system

Introduction to the Vlasov-Poisson system Introuction to the Vlasov-Poisson system Simone Calogero 1 The Vlasov equation Consier a particle with mass m > 0. Let x(t) R 3 enote the position of the particle at time t R an v(t) = ẋ(t) = x(t)/t its

More information

Euler equations for multiple integrals

Euler equations for multiple integrals Euler equations for multiple integrals January 22, 2013 Contents 1 Reminer of multivariable calculus 2 1.1 Vector ifferentiation......................... 2 1.2 Matrix ifferentiation........................

More information

SECTION 3.2: DERIVATIVE FUNCTIONS and DIFFERENTIABILITY

SECTION 3.2: DERIVATIVE FUNCTIONS and DIFFERENTIABILITY (Section 3.2: Derivative Functions and Differentiability) 3.2.1 SECTION 3.2: DERIVATIVE FUNCTIONS and DIFFERENTIABILITY LEARNING OBJECTIVES Know, understand, and apply te Limit Definition of te Derivative

More information

5.1 We will begin this section with the definition of a rational expression. We

5.1 We will begin this section with the definition of a rational expression. We Basic Properties and Reducing to Lowest Terms 5.1 We will begin tis section wit te definition of a rational epression. We will ten state te two basic properties associated wit rational epressions and go

More information

1 Calculus. 1.1 Gradients and the Derivative. Q f(x+h) f(x)

1 Calculus. 1.1 Gradients and the Derivative. Q f(x+h) f(x) Calculus. Gradients and te Derivative Q f(x+) δy P T δx R f(x) 0 x x+ Let P (x, f(x)) and Q(x+, f(x+)) denote two points on te curve of te function y = f(x) and let R denote te point of intersection of

More information

INTRODUCTION AND MATHEMATICAL CONCEPTS

INTRODUCTION AND MATHEMATICAL CONCEPTS Capter 1 INTRODUCTION ND MTHEMTICL CONCEPTS PREVIEW Tis capter introduces you to te basic matematical tools for doing pysics. You will study units and converting between units, te trigonometric relationsips

More information

The Ehrenfest Theorems

The Ehrenfest Theorems The Ehrenfest Theorems Robert Gilmore Classical Preliminaries A classical system with n egrees of freeom is escribe by n secon orer orinary ifferential equations on the configuration space (n inepenent

More information

AMS 147 Computational Methods and Applications Lecture 09 Copyright by Hongyun Wang, UCSC. Exact value. Effect of round-off error.

AMS 147 Computational Methods and Applications Lecture 09 Copyright by Hongyun Wang, UCSC. Exact value. Effect of round-off error. Lecture 09 Copyrigt by Hongyun Wang, UCSC Recap: Te total error in numerical differentiation fl( f ( x + fl( f ( x E T ( = f ( x Numerical result from a computer Exact value = e + f x+ Discretization error

More information

Computing Exact Confidence Coefficients of Simultaneous Confidence Intervals for Multinomial Proportions and their Functions

Computing Exact Confidence Coefficients of Simultaneous Confidence Intervals for Multinomial Proportions and their Functions Working Paper 2013:5 Department of Statistics Computing Exact Confience Coefficients of Simultaneous Confience Intervals for Multinomial Proportions an their Functions Shaobo Jin Working Paper 2013:5

More information

Lectures - Week 10 Introduction to Ordinary Differential Equations (ODES) First Order Linear ODEs

Lectures - Week 10 Introduction to Ordinary Differential Equations (ODES) First Order Linear ODEs Lectures - Week 10 Introuction to Orinary Differential Equations (ODES) First Orer Linear ODEs When stuying ODEs we are consiering functions of one inepenent variable, e.g., f(x), where x is the inepenent

More information

The Principle of Least Action

The Principle of Least Action Chapter 7. The Principle of Least Action 7.1 Force Methos vs. Energy Methos We have so far stuie two istinct ways of analyzing physics problems: force methos, basically consisting of the application of

More information

Phase space in classical physics

Phase space in classical physics Pase space in classical pysics Quantum mecanically, we can actually COU te number of microstates consistent wit a given macrostate, specified (for example) by te total energy. In general, eac microstate

More information

Lecture 2: Symplectic integrators

Lecture 2: Symplectic integrators Geometric Numerical Integration TU Müncen Ernst Hairer January February 010 Lecture : Symplectic integrators Table of contents 1 Basic symplectic integration scemes 1 Symplectic Runge Kutta metods 4 3

More information

Con gurational forces and variational mesh adaption in solid dynamics

Con gurational forces and variational mesh adaption in solid dynamics Con gurational forces an variational mes aaption in soli ynamics Tesis by Matias G. ielonka In Partial Ful llment of te Requirements for te Degree of Doctor of Pilosopy California Institute of Tecnology

More information

ELEC3114 Control Systems 1

ELEC3114 Control Systems 1 ELEC34 Control Systems Linear Systems - Moelling - Some Issues Session 2, 2007 Introuction Linear systems may be represente in a number of ifferent ways. Figure shows the relationship between various representations.

More information