This preview shows pages 1–5. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: MATH 237 Assignment 5 Winter 2007
Solutions 4.A3 a) Given f(a:,y) = 6395—29 we have fa:(x, : 36301—2?! fy($,xy) = _263$—2y
fx$($9y) : gen—2y fang/(33,29) : —6€32—2y fyy(m,y) = 4e3x—2y. so that Vf(2,3) = (fm(2,3),fy(2,3)) = (3, —2). and
_ fmcc(2a3) fwy(2a3) _ 9 "'6
W“) ‘ ( mm) was) > ‘ l —6 4 b) The linear approximation at (2, 3) is Lemmy) = f(2, 3) + M2, 3) (Iv  2) + M2, 3) (y — 3)
= 1+3(:c—2)—2(y—3). The second—order Taylor polynomial is ' ($_29y_3) + %(x_2ay_3)Hf(2i3) ( >
1 1+ 3(m — 2) — 2(y — 3) + §[9(x — 2)2 —12(:v — 2)(y — 3) +4(y  3)2]. P2,(2,3) (13: y) C) We ﬁnd Vf(:z:,y) = (aw—29$, —2), which is always parallel to (3 — 2). Thus the level
curves of f, which are everywhere orthogonal to the gradient (and so to (3, —2)) are straight
lines. This is clear from the form of f as the level curves are deﬁned by 3:73 — 2y = constant. 4.A5 1) With f(:1:,y) = (:c — y) sin(:1: + y), we ﬁnd fm(a:, y) : sin(a: + y) + (a: — y) cos(m + y)
y) fylﬂo = — $11017 + y) + (33 — y) COS($ + ’9) fax, .9) = 00805 + y) + 603(33 + y)  (a: — y) sinCI + y)
fmytr, y) = 608(w + y) — 008(56 + y) — (9:  y) sinks + y)
fyy(:1:,y)= — cos(a: + y) — cos(:c + y) — (:1: — y) sin(:r + Evaluating at (7T, 7r), we have mm) = 0, MW?) = 0, mm) = 0
fxx(7ra 7r) = 2: fxylnv”) : 0, Joya/(Wu?) : *2 Then the second order Taylor polynomial at (7r, 7r) is Iaaaew>= ﬂmwhthme—ﬂ+JMmﬂwWt+
g(tamwxm—wr+2tnmwxm—ee—wu+tnmwxyma = e—«V—e—«V ii) Since f and its partial and second partial derivatives are continuous (that is, f is a
02 function), Taylor’s Theorem tells us that P2,(,,‘,r)(:r, y) provides a good approximation to
f (:12, 3/) near (7T, 7r). Thus the level curves of f near (7r, 7r) are well approxmiated by the level
curves of P2, (7r, 7r) near (7r, 7r) These curves 0 = (:0 — 7r)2 — (y — 7r)2 are hyperbolae centred
at (71', 7r), as shown below. We conclude that the function f looks like a saddle near (7r,7r). 4.B2 a) Let u(m,t) = Ae‘($”“)2. Then ux(a:, t) = —2A(:c — ewe—(“0”2 and
we = We — avemew _ gig—W
while
“453, t) = 2Ac(x — Ct)e—($—ct)2
and WW”: 15) = 4A262(x — ct)2e_(1"c‘)2 _ 2AC2e—(m—ct)2 so cgum (m, t) 2 2mm, t) as required. b) me the graphs of it against :1: for various values of t, we see that the wave moves with speed 1% = c to the right. 0) Observing that the combination a: — ct lead to a rightward motion, we deduce that :1: + at
will lead to a leftward motion. A leftward moving wave of the same shape is described by
u($, t) = Ae‘(z+0‘)2. d) Provided f has continuous second partials (so that we may safely apply the chain rule in
What follows), we let u(:1:, t) = f (:1: — ct) and ﬁnd I/ ux(a:,t) = fl(x— ct) and um(m,t) = f (:3 —ct) while // 21.L($,t) = —cfl(ar — ct) and utt(x,t) = 02f (at — ct), 2 which conﬁrms c 113:9;(1‘, t) = uu(m,t). 4B7 We suppose f : R2 —> R has continuous second partial derivatives which, for (32,31) in
a neighbourhood of (a, b), are all bounded by some constant .M. Consider the error Rum)
made by using the linear approximation to estimate f at a point (33, in this neighbourhood.
Then, from Taylor’s Theorem (page 93 in course notes) there is some point C on the line
segment joining (a, b) and (:r,y) such that R1.(a,b) = g _ (1)2 + 2f1y(x _ (IX?! _ b) + fyy(C(y _ (7)2]  Then
mm s glMo—aero—axy—b)+M(y—b)2
= g[($—a)2+2(m—a)[y—b)+(y—b)2
s g(m—a>2+<y—b)2+MI<m—a)(y—b)l 3 s %(m,y)—<a,biil2+ = M(ﬂv,y)(aab)l2 M 31l(m,y)— (mm? as required. The second last line follows from — a)(y — b) S — (1)2 + (y — b)2, which
follows from expanding — a) + (y — (1)]2 2 0. 5.A1(iii) a) With f(:r,y) = (a:2 + 3/2 — 1)y we ﬁnd fx(x,y) = 2:62; Many) = 9:2 + 3y2 — 1 The critical points are those (:13, y) for which f3; = fy = 0. The ﬁrst condition speciﬁes m = 0 or y = 0. If :r = 0, then fy = 0 implies 3y2 = 1 or y = if?“ while if y = 0 then fy = 0 implies 3:2 = 1 or m = :i:1. We conclude there are four critical points: (0, ii) and (i1, 0).
To classify these critical points, we ﬁnd the second derivatives fmx($ay) : 2y fmy($ay) = 2m fyy<xay) : 631‘ Then ;_ a 0 whose determinant fmfW — fy is positive and whose ﬁrst term is positive. We conclude
1 that (0, W) is a local minimum. We next ﬁnd Hf(0 —i)— _% O
1 V3 _ O _\/i§ Whose determinant fmfm, — $21, is positive and whose ﬁrst term is negative. We conclude that (0, is a local maximum. Treating the last two points together we have Hf(:i:1,0)= (i3 In both cases the determinant fmfyy — 12.3] is negative. These are saddle points. b) From the above analysis we have a classiﬁcation of the nature of the level curves near
these critical points (i.e. ellipses and saddles). A more precise picture can be reached by
considering the second order Taylor polynomials: while Putting these together, we have the following: c<<2., mil ll Extra 8. The function f (at, y) = 1 + 3:2 + y2 has derivatives 1: 1/
firm: = _ f 17,? =——————
( y) ,————1+m2+y2 A J) ,———1+x2+y2 —a:y 1 + 51:2 fm($ay) = W fwy($,yl = W fyy(33,y) = W We note that f” is everywhere positive, and the determinant of the Hessian is ll (1_+3:21TyZ)3/_2 [(1 + y2)(1 + x2) — (_my)2] 1
(1+ on? + MW2 2
fmfyy _ my [1 + y2 + 3:2 + yzmg — y2m2] > 0. We conclude that the function is convex. ...
View
Full
Document
This note was uploaded on 01/17/2011 for the course MATH 237 taught by Professor Wolczuk during the Spring '08 term at Waterloo.
 Spring '08
 WOLCZUK

Click to edit the document details