Let X and Y be normed vector spaces1 and U be an open subsets of X. We say that a function f : U → Y is differentiable at x0 if there exists a bounded linear map T : X → Y such that
khklimX→0
|f (x0+ h) − f (x0) − T (h)kY
khkX = 0.
We denote T by Df (x0). If f is differentiable at every point of U, we say that f is differen- tiable on U.
Example 1.1. If f : (a, b) → R is differentiable at x0, then Df (x0)(h) = f0(x0)h for any h ∈ R.
Suppose that f is differentiable at every points of U. Then Df defines a map Df : U → Lb(X, Y ), x 7→ Df
where Lb(X, Y ) is the space of bounded linear maps from X to Y. We equip Lb(X, Y ) with the operator norm defined by
kT k = max
khkX=1
kT (h)kY for T ∈ L(X, Y ).
Then Lb(X, Y ) is a normed space. Since Lb(X, Y ) is a normed space, we can discuss the continuities and differentiabilities of Df. If Df : U → Y is continuous on U, we say that f is continuously differentiable on U. The set of all continuously differentiable functions from U to Y is denoted by C1(U, Y ). We denote C1(U, R) (or C1(U, C)) by C1(U ).
Definition 1.1. We say that a function f : U → Y is twice differentiable at x0 if there exists δ > 0 such that Df (x) exists at every point of x ∈ B(x0, δ) and the function
Df : B(x0, δ) → Lb(X, Y )
is differentiable at x0. We denote D(Df )(x0) by D2f (x0). If f is twice differentiable at every point of U, we say that f is twice differentiable on U.
Let us look at the example when X = Y = Rn. Suppose that f : (a, b) → R is twice differentiable at x0, i.e. Df : (a, b) → Lb(R, R) is differentiable at x0. We remark that Lb(R, R) coincides with L(R, R). Then there exists a linear map T : R → L(R, R) such that
k→0lim
kDf (x0+ k) − Df (x0) − T (k)kL(R,R)
|k| = 0.
For > 0, there exists δ > 0 so that
kDf (x0+ k) − Df (x0) − T (k)kL(R,R) < |k|
whenever 0 < |h| < δ. By the norm inequality,
|Df (x0+ k)(h) − Df (x0)(h) − T (k)(h)| ≤ kDf (x0+ k) − Df (x0) − T (k)kL(R,R)|h| < |h||k|.
Since Df (x)(h) = f0(x)h for any x ∈ (a, b), we have
Df (x0+ k)(h) − Df (x0)(h) = (f0(x0+ k) − f0(x0))h.
Dividing the above inequality by k, we obtain
f0(x0+ k) − f0(x0)
k h − T (1) (h)
< |h|.
1All the normed vector spaces in this note are assumed to be real or complex.
1
T (1)(h) = lim
k→0
f0(x0+ k) − f0(x0)
k h = f00(x0)h.
By linearity of T, we obtain that T (k)(h) = f00(x0)hk. In other words, (D2f )(x0)(k)(h) = f00(x0)hk.
Let us recall the basic definition of bilinear maps. Let X, Y, Z be normed vector spaces.
A bilinear map T : X × Y → Z is a function such that (1) T (ax1+ bx2, y) = aT (x1, y) + bT (x2, y)
(2) T (x, ay1+ by2) = aT (x, y1) + bT (x, y2)
for any a, b ∈ R (or C) and for any x, x1, x2 ∈ X and for any y, y1, y2 ∈ Y. We say that T is a bounded bilinear map if there exists M > 0 so that kT (x, y)kZ ≤ M kxkXkykY. The space of bounded bilinear maps from X × Y to Z is denoted by bil(X × Y, Z). In general, we consider the n-linear maps
Definition 1.2. Let X1, · · · , Xnand Y be normed vector spaces. We can define the notion of n-linear maps
T : X1× · · · × Xn→ Y.
We say that T is bounded if there exists M > 0 so that kT (x1, · · · , xn)kY ≤ M
n
Y
i=1
kxikXi for any (x1, · · · , xn) ∈ X1× · · · × Xn.
The space of all bounded n-linear maps from X1× · · · × Xn to Y is denoted by mult(X1× · · · × Xn, Y ) = {T : X1× · · · × Xn→ Y : T is bounded n-linear}.
The norm of a bounded n-linear map T is defined to be kT k = sup
kxikXi=1, 1≤i≤n
kT (x1, · · · , xn)kY. Then we have the following results:
Proposition 1.1. mult(X1× · · · × Xn, Y ) is a normed vector space. Furthermore, if Y is a Banach space, so is mult(X1× · · · × Xn, Y ).
Proof. This is left to the reader as an exercise.
Proposition 1.2. Let X, Y, Z be normed vector spaces. We have the following isomorphism of normed vector spaces
Lb(X, Lb(Y, Z)) ∼= bil(X × Y, Z).
Proof. Suppose T : X → Lb(Y, Z) is a linear map. For each x ∈ X, T (x) ∈ Lb(Y, Z), i.e.
T (x) : Y → Z is a bounded linear map for any x ∈ X. We define ϕT : X × Y → Z by ϕT(x, y) = T (x)(y). Then ϕT is bilinear. Now let us prove that ϕT is bounded. Since T is bounded, we can find M > 0 so that kT (x)kLb(Y,Z) ≤ M kxkX for any x ∈ X. Since T (x) ∈ Lb(Y, Z), kT (x)(y)kZ ≤ kT (x)kLb(Y,Z)kykY for any y ∈ Y. For any x ∈ X and Y ∈ Y, one has
kϕT(x, y)kZ= kT (x)(y)kZ ≤ kT (x)kLb(Y,Z)kykY ≤ M kxkXkykY.
This shows that ϕT ∈ bil(X × Y, Z). We obtain a map
ϕ : Lb(X, Lb(Y, Z)) → bil(X × Y, Z), T 7→ ϕT.
We leave it to the reader to check that ϕ is an isomorphism of normed vector spaces. We
complete the proof of our assertion.
If f : U → Y is twice differentiable on U, Df : U → Lb(X, Y ) is differentiable on U.
Therefore D2f defines a map from U into Lb(X, Lb(X, Y )). By Proposition 1.2, Lb(X, Lb(X, Y )) ∼= bil(X × X, Y ).
Hence we obtain a map
D2f : U → bil(X × X, Y ).
If D2f is continuous, we say that f : U → Y is twice continuously differentiable on U. The set of all twice continuously differentiable functions from U into Y is denoted by C2(U, Y ).
The set C2(U, R) (or C2(U, C)) is denoted by C2(U ) for simplicity.
Inductively, we say that f is j-times differentiable on U if Dj−1f exists at every point of U and the map Dj−1f : U → multj−1(Xj−1, Y ) is differentiable on U. We denote Djf (x) = D(Dj−1f )(x) for x ∈ U when f is j-times differentiable on U. If f is k-times differentiable on U, we obtain a function
Djf : U → multj(Xj, Y ).
When Djf is continuous on U, we say that f is j-times continuously differentiable on U. The set of all j-times continuously differentiable functions from U to Y is denoted by Cj(U, Y ).
The set Cj(U, R) (or Cj(U, C)) is simply denoted by Cj(U ).
Definition 1.3. Let U be an open subset of a normed space X. We set C∞(U ) =
∞
\
j=0
Cj(U ).
We call C∞(U ) the space of real valued smooth functions on U ; elements of Cj(U ) are called smooth functions on U.
We can check the following results.
Example 1.2. Let f ∈ Cj(a, b). For each 1 ≤ i ≤ j, the map (Dif )(x) : (a, b) → multi(Ri, R)
is given by (Dif )(x)(h1, · · · , hi) = f(i)(x)h1· · · hi at any point x ∈ (a, b). Furthermore, Dif (x)(h, · · · , h) = fi(x)hi for any h ∈ R. The Taylor Theorem can be rewritten as
f (x + h) =
j−1
X
i=0
(Dif )(x)(h, · · · , h)
i! +Djf (x + ch)(h, · · · , h) j!
where h is a real number so that |h| < δ for some δ > 0 and c ∈ [0, 1].
Example 1.3. Let U be an open subset of Rn and f : U → R be a function. Suppose that f ∈ Cj(U ) when j is sufficiently large (or we may assume that f is smooth). Let us compute D2f : U → L(Rn× Rn, R). Let h = (h1, · · · , hn) and k = (k1, · · · , kn). Then for x ∈ U,
(1.1) (D2f (x))(h, k) =
n
X
i,j=1
∂2f
∂xi∂xj(x)hikj.
f : U → L(R × R × R , R). Assume l = (l (D3f (x))(h, k, l) =
n
X
i,j,s=1
∂3f
∂xi∂xj∂xs
(x)hikjls.
Inductively it is possible for us to write down all the formulas for Dif (x) for any i ≥ 0.
Recall that
Hi(f )(x)(h) = X
|α|=i
i α
(Dαf )(x)hα.
It is easy for us to see that Hi(f )(x)(h) = (Dif )(x)(h, · · · , h) for any h ∈ Rn. Let us prove (1.1). Let T = D2f (x). For any > 0, there exists δ > 0 such that
kDf (x + k) − Df (x) − T (k)kL(Rn,R)< kkk whenever 0 < kkk < δ. By norm inequality,
|Df (x + k)(h) − Df (x)(h) − T (k)(h)| ≤ kDf (x + k) − Df (x) − T (k)kL(Rn,R)khk < khkkkk.
Choose h = ei where {e1, · · · , en} is the standard basis for Rn. We know that Df (x)(ei) = ∂f
∂ei(x).
We find that
∂f
∂xi(x + k) − ∂f
∂xi(x) − T (k)(ei)
< kkk.
Let k = tej; we find that when 0 < |t| < δ,
∂f
∂xi
(x + tej) − ∂f
∂xi
(x) − T (h)(ei)
< |t|.
Dividing the above equation by |t|, we obtain that
fxi(x + tej) − fxi(x)
t − T (ej)(ei)
< whenever 0 < |t| < δ. This shows that
fxixj(x) = lim
t→0
fxi(x + tej) − fxi(x)
t = T (ej)(ei).
If we write h =Pn
i=1hiei and k =Pn
j=1kjej, then T (k)(h) =
n
X
i,j=1
hikjT (ej)(ei) =
n
X
i,j=1
fxixj(x)hikj. This shows that
D2f (x)(h, k) = T (k)(h) =
n
X
i,j=1
fxixj(x)hikj. Inductively, we are able to determine all Djf (x) for any j ≥ 1.
Example 1.4. Let f (x, y) = ex+2y. Compute Df (0, 0), D2f (0, 0) and D3f (0.0).
Answer: let h = (h1, h2) and k = (k1, k2) and l = (l1, l2) in R2 Then Df (x, y)(h) = fx(x, y)h1+ fx(x, y)h2
D2f (x, y)(h, k) = fxx(x, y)h1k1+ fxy(x, y)h1k2+ fyx(x, y)h2k1+ fyy(x, y)h2k2 D3f (x, y)(h, k, l) = fxxx(x, y)h1k1l1+ hyyy(x, y)h2k2l2
+ fxxy(x, y)h1k1l2+ fxyx(x, y)h1k2l1+ fyxx(x, y, z)h2k1l1 + fxyy(x, y)h1k2l2+ fyxy(x, y)h2k1l2+ fyyx(x, y)h2k2l1.
Remark. Let U be an open subset of Rn. We remark that f ∈ Cj(U ) if and only if Dαf exists and continuous on U for any n-dimensional multiindices α with |α| = j. It follows from the fact that any norms on finite dimensional (real) vector spaces are equivalent.
Department of Mathematics, National Cheng Kung University, Taiwan, fjmliou@mail.ncku.edu.tw NCTS, Mathematics