You are on page 1of 15
ecture 6 Multivariable Calculus Multivariate like univariate calculus has to do with the local approximation of functions by affine ones. Definition: Let U be an open subset of RY and let f: U = R™, The function f is differentiable at c € U if there exists a linear transformation Df(c): RY ~ RY, called the derivative of f at c, such that for every positive number e, there exists a positive number 8 such that Ifo ~ f(o) ~ DA(o)(x =o < ell ~ all if [lk - cl] <8. That is, tim LfGo = fle) — Df(e) (x = ¢) us Ix= al This definition says that the affine function f(c) + Di(e)(x - ¢) = [f(e) ~ Df(e}o] + DI(x)(x) approximates the function f locally at c. Remark: It should be clear that if f(x) = a + T(x), where ac RM and T: RY ~ RM is linear, then Df(x) = T, for all x. That is, the derivative of an affine function is its linear part, Lemma 6.1: A function f: U- RS, where U is an open subset of RY, has at most ono derivative at a point. Proof: Suppose that S: RY ~ RM and T: RY ~ RM are linear and satisfy the definition of a derivative of f at ce U. If S #7, then there is a v € RY such that |jvl] = 1 and |[S(v) ~ TI > 0. Let e > 0 and 5 > 0 be such that [ff(x) ~ f(c) ~ S{x - o)]] < ellx ~ ol] and |If(x) ~ f(c) = T(x = o)|] < el} — cl), if jx - o| $8. Let t be a non-zero number such that |t{ < 8 and let x = ctw, Then 0 < |fx—ol] <8 and < |t] IS) = TIM] = [Sty - TivIll IIH) - Mey] + Sex = 0) + [10x - fe) - TH ~ o)fl (If) — f(0) ~ Sx = el] + IFO) ~ fle) — TH = eI) 2elx — ol] = 2eljtvil = 2eltl. Dividing by [i], we see that 0 < |[S(v) ~ T(v)[] < 2e, for all positive numbers e, which is impossible, . The next lemma implies that every linear transformation is continuous. Lemma 6.2: If T: RY + R™ is a linear transformation, then there exists a positive number b such that |[T(v) — T(w)|| < bllv — wl], for all v and w in RX. Proof: Let A be the MxN matrix representing T with respect to the standard bases of A" andRMandleta= max |a |. If y = T(x), then by the Cauchy-Schwarz inequality, temem t= [Lax] stall. for m ..,M, where @_ is the mih row of A. Because lath= [be s Ina = adn it follows that lv. sad Nix and so liyll = 4 S92 = diary txt? = ad MN IIx Let b = a{MN. Then |[T(v) — T(w)Il = |Tv — w)I] < bllv ~ wil, for all v and w in A, as was to be proved. . Lemma 6.3: Let f: U ~ R¥, where U is an open subset of A". If fis differentiable at ce U, then there exist positive numbers 6 and B such that |[f(x) - f(0)|| < Bllx ~ cll, if {lk cl] <8. In particular, fis continuous at ©. Proof: The proof simply says that because f locally approximates at ¢ a continuous function, it is itself continuous there. Since f is differentiable at o, there exists a positive number 8 such that IIx) - f(e) - Df(e)x = ©)]l < [lk ~ ell, if Ie ol] $8. By lemma 6.2, there is a positive number b such that UIDH(e)(x ~ c)]} < bl}x ~ off Therefore {IFO = f(c)]] < IIE) — H(c) ~ Dfle)(x - e)l] + [IDFo)x - DI] s (1 + b)llx ~ cll, if |x ol] <8. LetB=1 +b. . | next relate the derivative to directional derivatives and partial derivatives. Definition: Let f: U~ R, where U is an open subset of RY, and let v ¢ RN. The number Y {(c) is the directional derivative of f at c in the direction v if, for every positive number e, there is a positive number such that if 0 < |t] <6, then f(o+ W) -e) —¥ f(o)| {(x), if [Ik - cl]

You might also like