Page 117 - Vector Analysis
P. 117

§4.3 The Surface Integrals                                                                                                                 113

Theorem 4.39. Let Σ Ď R3 be a regular surface, tV1, ψ1u and tV2, ψ2u be two local C 1-
parameterizations of Σ at a point p P Σ, and U = ψ1(V1) X ψ2(V2) Ď Σ. Then for (i, j) =
(1, 2) and (2, 1), the transition function ψj´1 ˝ ψi : ψi´1(U ) Ñ ψj´1(U ) is of class C 1.

Proof. We first note that ψj´1 ˝ ψi is continuous on ψi´1(U ). Moreover, by the chain rule we

find that B (ψj´1 ˝ ψi) is the unique 2-vector satisfying

                   Bu

[ B ψi            ]         [  B             ψj´1                ]       [                            ][ B         (ψj´1 ˝   ψi) (u,    ]
                v)             Bu                              v)         (Dψj )(ψj´1               v)                Bu              v)
  Bu    (u,          =             (ψj  ˝           ˝  ψi)(u,         =                  ˝  ψi)(u,                                         .

Similarly, B (ψj´1 ˝ ψi) is the unique 2-vector satisfying

                    Bv

[ B ψi  (u,       ]  =      [B     (ψj  ˝    ψj´1   ˝  ψi)(u,    ]  =    [               ˝  ψi)(u,  v   ][      B  (ψj´1  ˝  ψi) (u,    ]  .
                v)                                             v)         (Dψj )(ψj´1                  )                              v)
  Bv                         Bv                                                                                       Bv

Therefore, we obtain that

                            []          =  [           ˝  (ψj´1  ˝     ][[  B  (ψj´1 ˝  ψi) ]...[ B (ψj´1 ˝  ψi) ]]   .                  (4.7)
                             Dψi            (Dψj )                  ψi)           Bu
                                                                                                      Bv

Since  []       has         full  rank,      [      ]T[Dψj     ]  is  an  invertible     2  ˆ  2  matrix        (for  if  ATAx  =     0  then
        Dψj                                   Dψj

}Ax}R2 n = xTATAx = 0 which implies x = 0 since A has full rank); thus (4.7) implies that

[[  B  (ψj´1 ˝  ψi)  ]...[  B  (ψj´1 ˝  ψi)  ]]  =  (([      ]T  [       )  ˝  (ψj´1  ˝     )´1[             )  ˝  (ψj´1  ˝  ψi)]T[Dψi]       ;
          Bu                      Bv                    Dψj       Dψj    ]               ψi) (Dψj

thus the partial derivatives of ψj´1 ˝ ψi exist and are continuous. Theorem 2.30 then implies

that ψj´1 ˝ ψi is of class C 1.                                                                                                                  ˝

    Similar to how the directional derivative is defined, we intend to define the differentia-
bility of f through the differentiability of the function f ˝ ψ : V Ñ Rn, where tV, ψu is a

local parametrization of Σ (at some point). Again, we need to talk about if this definition

depends on the choice of local parameterizations. Nevertheless, if tV1, ψ1u and tV2, ψ2u
are two C 1-local parametrization of Σ at p, and f ˝ ψ1 is differentiable at ψ1´1(p), then
the chain rule and Theorem 4.39 imply that f ˝ ψ2 is also differentiable at ψ2´1(p) since
f ˝ ψ2 = (f ˝ ψ1) ˝ (ψ1´1 ˝ ψ2). This induces the following

Definition 4.40. Let Σ Ď R3 be a C 1-regular surface. A scalar function f : Σ Ñ R is
said to be differentiable at p P Σ if for every parametrization tV, ψu of Σ at p, the function
   112   113   114   115   116   117   118   119   120   121   122