Page 102 - Vector Analysis
P. 102

98 CHAPTER 4. Vector Calculus

Theorem 4.10. Let C Ď Rn be a curve with C 1-parametrization γ : [a, b] Ñ Rn. Then

                                                           żb
                                                  ℓ(C) = }γ 1(t)}Rn dt .

                                                                   a

Proof. Let ε ą 0 be given. Since γ : [a, b] Ñ Rn is C 1, there exists δ ą 0 such that

               }γ 1(t)  ´     γ 1(s)››Rn    ă        ε     ´  a)    whenever            s, t P [a, b], |s ´ t| ă δ .
                                               4?n(b

By the definition of the length of curves, there exists a partition P = ta = t0 ă t1 ă ¨ ¨ ¨ ă

tk = bu of [a, b] such that

                                    ℓ(C )   ´  ε  ă      k          ´   γ(ti´1)››Rn        ď  ℓ(C) .
                                               4
                                                        ÿ ››γ(ti)

                                                        i=1

W.L.O.G., we can assume that }P} ă δ. For each component γj of γ, the mean value

theorem implies that for some ξi P [ti´1, ti],

                                         γj(ti) ´ γj(ti´1) = γj1(ξi)(ti ´ ti´1) ;

thus for each i P t1, ¨ ¨ ¨ , ku and si P [ti´1, ti],

ˇˇγj (ti )  ´ γj(ti´1) ´      γj1(si)(ti    ´  ti´1)ˇˇ  ď    ˇˇγj1(ξi)  ´  γj1(si)ˇˇ|ti  ´  ti´1|   ă        ε     ´ a) |ti       ´  ti´1| .
                                                                                                       4?n(b

As a consequence, for each i P t1, ¨ ¨ ¨ , ku and si P [ti´1, ti],

ˇ ˇˇ                                                                                                                                        ˇ

ˇ››γ  (ti)  ´  γ(ti´1)››Rn       ´  ››γ  1(si)››Rn |ti  ´  ti´1|ˇˇ  ă   ˇ››γ  (ti)  ´  γ(ti´1)››Rn   ´    ››γ  1(si)(ti  ´     ti´1  )››Rn  ˇ
ˇ                                                                       ˇ                                                                   ˇ

               ››γ(ti)                                     ti´1)››Rn          [   n(          ε                         )2  ]  1
                   ε                                                             ÿ                                 ti´1|       2
                                          γ 1(si)(ti
            ď           ´  γ(ti´1)  ´                   ´               ď               4?n(b    ´  a) |ti   ´

                                                                                 j=1

            ă 4(b ´ a) |ti ´ ti´1|

which further implies that

                           ˇ  k                                      k                             ˇ         ε
                           ˇ                                                                  ti´1|ˇˇ        4
                           ˇ  ÿ  ››γ(ti        γ(ti´1)››Rn          ÿ ››γ 1(si)››Rn|ti

                                         )  ´                 ´                            ´           ă        .

                              i=1 i=1

Therefore, for a = t0 ď s0 ď t1 ď s1 ¨ ¨ ¨ ď sk ď tk = b,

                                            ε      k    ››γ 1(si)››Rn|ti                               ε
                                            2                                                          2
                                                  ÿ

                                 ℓ(C )   ´     ă                           ´     ti´1|  ă  ℓ(C )  +       .

                                                  i=1

Since }γ 1} is Riemann integrable over [a, b], we must have

                                                 żb                                                                                            ˝
              ℓ(C) ´ ε ă L(}γ 1}Rn, P) ď ››γ 1(t)››Rndt ď U (}γ 1}Rn, P) ă ℓ(C) + ε ,

                                                                            a

and the theorem is concluded because ε ą 0 is given arbitrarily.
   97   98   99   100   101   102   103   104   105   106   107