Elliptic integrals, the arithmeticgeometric mean and the BrentSalamin algorithm for π


 Meghan Thomas
 2 years ago
 Views:
Transcription
1 Elliptic integrals, the arithmeticgeometric mean and the BrentSalamin algorithm for π Notes by G.J.O. Jameson Contents. The integrals I(a, b) and K(b). The arithmeticgeometric mean 3. The evaluation of I(a, b): Gauss s theorem 4. Estimates for I(a, ) for large a and I(, b) for small b 5. The integrals J(a, b) and E(b) 6. The integrals L(a, b) 7. Further evaluations in terms of the agm; the BrentSalamin algorithm 8. Further results on K(b) and E(b).. The integrals I(a, b) and K(b) For a, b >, define I(a, b) = An important equivalent form is: π/ (a cos θ + b sin dθ. () θ) /.. We have I(a, b) = dx. () (x + a ) / (x + b ) / Proof. With the substitution x = b tan θ, the stated integral becomes π/ b sec θ (a + b tan θ) / b sec θ dθ = = π/ π/ (a + b tan θ) / cos θ dθ (a cos θ + b sin dθ. θ) / We list a number of facts that follow immediately from () or (): (I) I(a, a) = π a ; (I) (I3) I(b, a) = I(a, b); I(ca, cb) = I(a, b), hence ai(a, b) = I(, b/a); c (I4) I(a, b) decreases with a and with b; (I5) if a b, then π a I(a, b) π b.
2 Note that the substitution x = /y in () gives I(a, b) = dx. (3) ( + a x ) / ( + b x ) /.. We have I(a, b) π (ab) /. Proof. This follows from the integral CauchySchwarz inequality, applied to (). Note further that, by the inequality of the means, π (ab) π ( / 4 a + ). b This upper bound for I(a, b) is also obtained directly by applying the inequality of the means to the integrand in (). Clearly, I(a, ) is undefined, and one would expect that I(a, b) as b +. This is, indeed, true. In fact, the behaviour of I(, b) for b close to is quite interesting; we return to it in section 4. The complete elliptic integral of the first kind is defined by K(b) = π/ ( b sin dθ. (4) θ) / For b, the conjugate b is defined by b = ( b ) /, so that b + b =. Then cos θ + b sin θ = b sin θ, so that I(, b) = K(b ). (An incomplete elliptic integral is one in which the integral is over a general interval [, α]. We do not consider such integrals in these notes.) Clearly, K() = π.3. We have and K(b) increases with b. Another equivalent form is: K(b) = dt. (5) ( t ) / ( b t ) / Proof. Denote this integral by I. Substituting t = sin θ, we have I = π/ cos θ( b sin cos θ dθ = K(b). θ) /.4 PROPOSITION. For b <, π K(b) = d n b n = + 4 b b4 +, (6) n=
3 where d = and d n =.3... (n ).4... (n). where Proof. By the binomial series, ( b sin θ) / = + a n b n sin n θ, n= ( ) a n = ( ) n = n n! ( )( 3) (n (n ) ) = (n) For fixed b <, the series is uniformly convergent for θ π. Now π/ sin n θdθ = πa n: identity (6) follows. In a sense, this series can be regarded as an evaluation of I(a, b), but a more effective one will be described in section 3..5 COROLLARY. For b <, + 4 b π K(b) + b 4( b ). Proof. Just note that + 4 b n= d nb n + 4 (b + b ). Hence π K(b) + b for b. These are convenient inequalities for K(b), but when translated into an inequality for I(, b) = K(b ), the righthand inequality says I(, b) + ( π b )/(4b ) = /(4b ), and it is easily checked that this is weaker than the elementary estimate I(, b) + /b given above. 4 π The special case I(, ) The integral I(, ) equates to a beta integral, and consequently to numerous other integrals. We mention some of them, assuming familiarity with the beta and gamma functions..6 PROPOSITION. We have I(, ) = = dx ( x 4 ) / (7) dx (x 4 ) / (8) = 4 B( 4, ) (9) = Γ( 4 ) 4 π. () 3
4 Proof. For (7), substitute x = sin θ, so that x 4 = cos θ( + sin θ). Then π/ dx = ( x 4 ) / Substituting x = /y, we also have dx = ( x 4 ) / π/ ( + sin dθ = θ) / Substituting x = t /4 in (7), we obtain I(, ) = (cos θ + sin θ) / dθ = I(, ). ( y 4 ) / y dy = dy. (y 4 ) / ( t) / ( 4 t 3/4 ) dt = 4 B( 4, ). Now using the identities B(a, b) = Γ(a)Γ(b)/Γ(a + b) and Γ(a)Γ( a) = π/(sin aπ), we have Γ( ) = π and Γ( 4 )Γ( 3 4 ) = π/(sin π 4 ) = π, hence B( 4, ) = Γ( 4 )Γ( ) Γ( 3 4 ) = πγ( 4 ) π = Γ( 4 ) π..7. We have π/ (sin θ) dθ = I(, ), () / ( x 4 ) / dx = ( x ) /4 dx = 3 I(, ). () Proof. For (), substitute x = (sin θ) / in (7). Denote the two integrals in () by I, I. The substitutions x = t /4 and x = t / give I = B(, 3) and I 4 4 = B(, 5 ). The 4 identity (a + b)b(a, b + ) = bb(a, b) now gives I = I = B(, ) The arithmeticgeometric mean Given positive numbers a, b, the arithmetic mean is a = (a + b) and the geometric mean is b = (ab) /. If a = b, then a = b = a. If a > b, then a > b, since 4(a b ) = (a + b) 4ab = (a b) >. With a > b, consider the iteration given by a = a, b = b and a n+ = (a n + b n ), b n+ = (a n b n ) /. At each stage, the two new numbers are the arithmetic and geometric means of the previous two. The following facts are obvious: 4
5 (M) a n > b n ; (M) a n+ < a n and b n+ > b n ; (M3) a n+ b n+ < a n+ b n = (a n b n ), hence a n b n < n (a b). It follows that (a n ) and (b n ) converge to a common limit, which is called the arithmeticgeometric mean of a and b. We will denote it by M(a, b). Convergence is even more rapid than the estimate implied by (M3), as we will show below. First, we illustrate the procedure by the calculation of M(, ) and M(, ): n b n a n n b n a n Of course, a n and b n never actually become equal, but more decimal places would be required to show the difference at the last stage shown. In fact, in the first example, a 3 b 3 < 4, as we will see shortly. Clearly, M(a, b) = M(a n, b n ) and b n < M(a, b) < a n for all n. Also, M(ca, cb) = cm(a, b) for c >, so M(a, b) = am(, b/a) = bm(a/b, ). For further analysis of the process, define c n by a n b n = c n... With c n defined in this way, we have (M4) c n+ = (a n b n ), (M5) a n = a n+ + c n+, b n = a n+ c n+, (M6) c n = 4a n+ c n+, (M7) c n+ c n 4b. Proof. For (M4), we have c n+ = a n+ b n+ = 4 (a n + b n ) a n b n = 4 (a n b n ). (M5) follows, since a n+ = (a n + b n ), and (M6) is given by c n = (a n + b n )(a n b n ) = (a n+ )(c n+ ). Finally, (M7) follows from (M6), since a n+ > b. (M5) shows how to derive a n and b n from a n+ and b n+, reversing the agm iteration. 5
6 (M7) shows that c n (hence also a n b n ) converges to quadratically, hence very rapidly. We can derive an explicit bound for c n as follows:.. Let R > and let k be such that c k 4b/R. Then c n+k 4b/R n for all n. The same holds with b replaced by b k. Proof. The hypothesis equates to the statement for n =. Assuming it now for a general n, we have by (M7) c n+k+ < (4b) = 4b. 4b R n+ R n+ In the case M(, ), we have c = b =. Taking R = 4, we deduce c n 4/4 n for all n. For a slightly stronger variant, note that c = a a <, hence c n+ 4/(4 n ) = 4/( n+ ) for n, or c n 4/( n ) for n. In particular, c <. We finish this section with a variant of the agm iteration in the form of a single sequence instead of a pair of sequences..3. Let b >. Define a sequence (k n ) by: k = b and Then k n+ = k/ n. + k n M(, b) = n= ( + k n). Proof. Let (a n ), (b n ) be the sequences generated by the iteration for M(, b) (with a =, b = b, whether or not b < ), and let k n = b n /a n. Then k = b and k n+ = b n+ = (a nb n ) / = k/ n. a n+ a n + b n + k n So the sequence (k n ) defined by this iteration is (b n /a n ). Also, so n r= ( + k n) = a n + b n = a n+, a n a n ( + k r) = a n M(, b) as n. Quadratic convergence of (k n ) to is easily established directly: k n+ = k/ n + k n < kn / + k n = ( kn / ) < ( k n ). + k n 6
7 Another way to present this iteration is as follows. Assume now that b <. Recall that for k, the conjugate k is defined by k + k =. Clearly, if t = k / /( + k) (where k ), then t = ( k)/( + k). So we have kn+ = ( k n )/( + k n ), hence + kn+ = /( + k n ), so that M(, b) =. + kn n= 3. The evaluation of I(a, b): Gauss s theorem We now come to the basic theorem relating I(a, b) and M(a, b). It was established by Gauss in 799. There are two distinct steps, which we present as separate theorems. In the notation of section, we show first that I(a, b) = I(a, b ). It is then a straightforward deduction that I(a, b) = π/[m(a, b)]. Early proofs of the first step, including that of Gauss, were far from simple. Here we present a proof due to Newman [New] based on an ingenious substitution. 3. THEOREM. Let a b >, and let a = (a + b), b = (ab) /. Then I(a, b) = I(a, b ). () is even: Substitute Proof. We write I(a, b ) as an integral on (, ), using the fact that the integrand I(a, b ) = dx. (x + a ) / (x + b / ) x = t ab t. Then x as t and x as t +. Also, dx dt = ( + ab/t ) and 4(x + b ) = 4x + 4ab = t + ab + a b = t ( t + ab t ), hence Further, (x + b ) / = ( t + ab ). t 4(x + a ) = 4x + (a + b) = t + a + b + a b t = t (t + a )(t + b ). 7
8 Hence I(a, b ) = ( t (t + a ) / (t + b ) / ab (t + ) + ab ) t t = dt (t + a ) / (t + b ) / = I(a, b). dt Lord [Lo] gives an interesting alternative proof based on the area in polar coordinates. 3. THEOREM. We have I(a, b) = π M(a, b) () Proof. Let (a n ), (b n ) be the sequences generated by the iteration for M(a, b). By Theorem 3., I(a, b) = I(a n, b n ) for all n. Hence π a n I(a, b) for all n. Since both a n and b n converge to M(a, b), the statement follows. We describe some immediate consequences; more will follow later. π b n 3.3. We have π a + b I(a, b) π. (3) (ab) / Proof. This follows from Theorem 3., since π/(a ) I(a, b ) π/(b ). The righthand inequality in (3) repeats., but I do not know an elementary proof of the lefthand inequality. From the expressions equivalent to I(, ) given in.6, we have: 3.4. Let M = M(, ) (.984). Then ( x 4 ) / dx = I(, ) = π M (.39), B( 4, ) = π M ( 5.446), Γ( 4 ) = (π)3/4 M / ( 3.656) Example. From the value M(, ) 6.67 found in section, we have I(, ) 8
9 Finally, we present several ways in which Theorem 3. can be rewritten in terms of K(b). Recall that I(, b) = K(b ), where b + b =, and that if k = ( b)/( + b), then k = k / /( + b) For b <, K(b) = I( + b, b). (4) Proof. By Theorem 3., I( + b, b) = I[, ( b ) / ] = I(, b ) = K(b) For < b <, K(b) = + b K K(b ) = + b K ( b / + b ( b + b ), (5) ). (6) Proof. By (4) and the property I(ca, cb) = I(a, b), c K(b) = I( + b, b) = ( + b I, b ) = + b + b K Also, by Theorem 3., K(b ) = I(, b) = I ( ) + b, b/ = + b I ( b / ) (, b/ = + b + b K + b ). ( ) b. + b Conversely, any of (4), (5), (6) implies () (to show this for (4), start by choosing c and k so that ca = + k and cb = k). For another proof of Theorem 3., based on this observation and the series expansion of ( + ke iθ ) / )( + ke iθ ) /, see [Bor, p. ]. 4. Estimates for I(a, ) for large a and I(, b) for small b In this section, we will show that I(, b) is closely approximated by log(4/b) for b close to. Equivalently, ai(a, ) is approximated by log 4a for large a, and consequently M(a, ) is approximated by F (a) =: π a log 4a. By way of illustration, F () = 6.7 to four d.p., while M(, ) This result has been known for a long time, with varying estimates of the accuracy of the approximation. For example, a version appears in [WW, p. 5], published in 97. 9
10 A highly effective method was introduced by Newman in [New], but only sketched rather briefly, with the statement given in the form ai(a, ) = log 4a+O(/a ). Following [Jam], we develop Newman s method in more detail to establish a more precise estimation. The method requires nothing more than some elementary integrals and the binomial and logarithmic series, and a weaker form of the result is obtained with very little effort. The starting point is: 4.. We have (ab) / dx = (x + a ) / (x + b ) / dx. (ab) (x / + a ) / (x + b ) / Proof. With the substitution x = ab/y, the lefthand integral becomes ab (ab) / ( a b + a y ) / ( a b + b y ) / y dy = dy. (ab) (b / + y ) / (a + y ) / where Hence b / I(, b) = H(x) dx, () H(x) =. () (x + ) / (x + b ) / Now ( + y) / > y for < y <, as is easily seen by multiplying out ( + y)( y), so for < x <, ( x ) (x + b ) < H(x) <. (3) / (x + b ) / Now observe that the substitution x = by gives b / /b / dx = (x + b ) / (y + ) dy = / sinh b. / 4. LEMMA. We have log b < / sinh < log b/ b + b. (4) / 4 Proof. Recall that sinh x = log[x + (x + ) / ], which is clearly greater than log x. Also, since ( + b) / < + b, ( ) / b + / b + = b [ + ( + / b)/ ] < b ( + b) = / b ( + b). / 4 The righthand inequality in (4) now follows from the fact that log( + x) x.
11 where This is already enough to prove the following weak version of our result: 4.3 PROPOSITION. For < b and a, log 4 b b < I(, b) < log 4 b + b, (5) log 4a a < ai(a, ) < log 4a + a. (6) Proof. The two statements are equivalent, because ai(a, ) = I(, ). By () and (3), a Since (x + b ) / > x, we have sinh b / R (b) < I(, b) < sinh b /, R (b) < R (b) = b / b / x dx. (7) (x + b ) / x b / x dx = x dx = b. Both inequalities in (5) now follow from (4), since log b / = log 4 b. This result is sufficient to show that M(a, ) F (a) as a. To see this, write ai(a, ) = log 4a + r(a). Then I(a, ) which tends to as a, since ar(a). a log 4a = ar(a) log 4a[log 4a + r(a)], Some readers may be content with this version, but for those with the appetite for it, we now refine the method to establish closer estimates. All we need to do is improve our estimations by the insertion of further terms. ( + y) / <. Instead, we now use the stronger bound ( + y) / < y y The upper estimate in (3) only used for < y <. These are the first three terms of the binomial expansion, and the stated inequality holds because the terms of the expansion alternate in sign and decrease in magnitude. So we have instead of (3): ( x ) (x + b ) < H(x) < ( / x x4 ). (8) (x + b ) / We also need a further degree of accuracy in the estimates for sinh b / and R (b).
12 4.4 LEMMA. For < b, we have log b + b 3 / 4 3 b < sinh < log b/ b + b / 4 Proof. Again by the binomial series, we have So, as in Lemma 4., 6 b + 3 b3. (9) + b 8 b < ( + b) / < + b 8 b + 6 b3. () ( ) / b + / b + < b ( + b / 8 b + 6 b3 ) = b ( + b / 4 6 b + 3 b3 ). The righthand inequality in (9) now follows from log( + x) x. Also, ( ) / b + / b + > b ( + b / 8 b ) = ( + B), b/ where B = 4 b 6 b (so B < 4 b). By the log series, log( + x) > x x for < x <, so log( + B) > B B > 4 b 6 b 3 b = 4 b 3 3 b. 4.5 LEMMA. For < b, we have: R (b) < b + 4 b b log, () b/ R (b) > b + 4 b b log b / 3 6 b3, () R (b) < b 5 b. (3) Proof: We can evaluate R (b) explicitly by the substitution x = b sinh t: R (b) = c(b) where c(b) = sinh (/b / ). Now c(b) b sinh t dt = b (cosh t ) dt = 4 b sinh c(b) b c(b), sinh c(b) = sinh c(b) cosh c(b) = ( ) / b / b + = b ( + b)/, so R (b) = b( + b)/ b c(b). Statement () now follows from ( + b) / < + b and c(b) > log. Statement () b / follows from the lefthand inequality in () and the righthand one in (4).
13 Unfortunately, (3) does not quite follow from (). We prove it directly from the integral, as follows. Write x /(x + b ) / = g(x). Since g(x) < x, we have b / g(x) dx < b (b b ). For < x < b, we have by the binomial expansion again ) x x g(x) = ( x b( + x /b ) / b b + 3x4, 8b 4 and hence b g(x) dx < ( )b < 3 b. Together, these estimates give R (b) < b 5 b. We can now state the full version of our Theorem. 4.6 THEOREM. For < b and a, we have log 4 b < I(, b) < ( + 4 b ) log 4 b, (4) ( log 4a < ai(a, ) < + ) log 4a. (5) 4a Moreover, the following lower bounds also apply: Hence I(, b) > ( + 4 b ) log 4 b 7 6 b, (6) ( ai(a, ) > + ) log 4a 7 4a 6a. (7) I(, b) = ( + 4 b ) log 4 b + O(b ) for < b <, ( ai(a, ) = + ) log 4a + O(/a ) for a >. 4a Note. As this shows, the estimation log 4a + O(/a ) in [New] is not quite correct. Proof: lower bounds. By (8), (9) and (3), I(, b) > sinh b / R (b) > log 4 b + b 3 6 b ( b 5 b ) > log 4 b, (with a spare term 8 b ). Of course, the key fact is the cancellation of the term b. Also, using () instead of (3), we have I(, b) > log 4 b + b 3 6 b ( b + 4 b 4 b log 4 b ) = ( + 4 b ) log 4 b 7 6 b. 3
14 Upper bound: By (8), I(, b) < sinh b / R (b) R (b), where So by (9) and (), R (b) = b / x 4 b / dx < x 3 dx = (x + b ) / 4 b. I(, b) < log 4 b + b 8 b + 6 b3 ( b + 4 b 4 b log 4 b 3 6 b3 ) b = ( + 4 b ) log 4 b 3 6 b + 4 b3. If b < 3 4, then 3 6 b 4 b3, so I(, b) < ( + 4 b ) log 4 b. For b 3 4, we reason as follows. By., I(, b) π 4 (+ b ). Write h(b) = (+ 4 b ) log 4 b π ( + ). We find that h( 3 ) >. One verifies by differentiation that h(b) 4 b 4 is increasing for < b (we omit the details), hence h(b) > for 3 b. 4 By Gauss s theorem and the inequality /( + x) > x, applied with x = /(4a ), statement (5) translates into the following pair of inequalities for M(a, ): 4.7 COROLLARY. Let F (a) = (πa)/( log 4a). Then for a, ( ) F (a) < M(a, ) < F (a). (8) 4a Since M(a, b) = b M(a/b, ) we can derive the following bounds for M(a, b) (where a > b) in general: π ) a ( b < M(a, b) < π log(4a/b) 4a a log(4a/b). (9) Note: We mention very briefly an older, perhaps better known, approach (cf. [Bow, p. ]). Use the identity I(, b) = K(b ), and note that b when b. Define A(b ) = π/ b sin θ ( b sin dθ θ) / and B(b ) = K(b ) A(b ). One shows by direct integration that A(b ) = log[(+b )/b] and B() = log, so that when b, we have A(b ) log b and hence K(b ) log 4 b. This method can be developed to give an estimation with error term O(b log ), [Bor, p. b ], but the details are distinctly more laborious, and (even with some refinement) the end result does not achieve the accuracy of our Theorem
15 By a different method, we now establish exact upper and lower bounds for I(, b) log(/b) on (, ]. The lower bound amounts to a second proof that I(, b) > log(4/b) The expression I(, b) log(/b) increases with b for < b. Hence log 4 I(, b) log b π for < b ; () π log 4 ai(a, ) log a π for a ; () a log a + π M(a, ) π a for a. () log a + log 4 Proof. We show that K(b) log(/b ) decreases with b, so that I(, b) log(/b) = K(b ) log(/b) increases with b. Once this is proved, the inequalities follow, since I(, b) log(/b) takes the value π/ at b = and (by 4.3) tends to log 4 as b +. By.4, for b <, we have K(b) = π n= d nb n, where d = and d n =.3... (n ).4... (n). Hence K (b) = π n= nd nb n. Meanwhile, log(/b ) = log( b ), so d db log b = b b = b n. By wellknown inequalities for the Wallis product, nπd n <, so K (b) < d db log(/b ). n= Application to the calculation of logarithms and π. Theorem 4.6 has pleasing applications to the calculation of logarithms (assuming π known) and π (assuming logarithms known), exploiting the rapid convergence of the agm iteration. Consider first the problem of calculating log x (where x > ). Choose n, in a way to be discussed below, and let 4a = x n, so that n log x = log 4a, which is approximated by ai(a, ). We calculate M = M(a, ), and hence I(a, ) = π/(m). Then log x is approximated by (a/n)i(a, ). How accurate is this approximation? By Theorem 4.6, log 4a = ai(a, ) r(a), where < r(a) < 4a log 4a. Hence in which log x = a r(a) I(a, ) n n, < r(a) n < log x = 4 log x 4a x. n We choose n so that this is within the desired degree of accuracy. It might seem anomalous that log x, which we are trying to calculate, appears in this error estimation, but a rough estimate for it is always readily available. 5
16 For illustration, we apply this to the calculation of log, taking n =, so that a = = 4. We find that M = M(4, ) , so our approximation is π 4 M , while in fact log = to seven d.p.. The discussion above (just taking log < ) shows that the approximation overestimates log, with error no more than 4/ 4 = /. We now turn to the question of approximating π. The simpleminded method is as follows. Choose a suitably large a. By Theorems 3. and 4.6, where π = M(a, )I(a, ) = M(a, )log 4a a + s(a), s(a) = M(a, ) r(a) 4a < M(a, )log a 4a. 3 So an approximation to π is M(a, ) log 4a, with error no more than s(a). Using our a original example with a =, this gives to five d.p., with the error estimated as no more than 8 5 (the actual error is about ). For a better approximation, one would repeat with a larger a. However, there is a way to generate a sequence of increasingly close approximations from a single agm iteration, which we now describe (cf. [Bor, Proposition 3]). In the original agm iteration, let c n = (a n b n) /. Note that c n converges rapidly π to. By Gauss s theorem, = M(a, c )I(a, c ). To apply Theorem 4.6, we need to equate I(a, c ) to an expression of the form I(d n, e n ), where e n /d n tends to. This is achieved as follows. Recall from. that a n = a n+ + c n+ and c n = 4a n+ c n+, so the arithmetic mean of a n+ and c n+ is a n, while the geometric mean is c n. So by Theorem 3., I(a n, c n ) = I(a n+, c n+ ), and hence I(a, c ) = I( n a n, n c n ) = n a n I (, c ) n. a n Since c n /a n tends to, Theorem 4.6 (or even 4.3) applies to show that I(a, c ) = lim log 4a n. n n a n c n Now take a = and b =. Then c = b, so M(a, b ) = M(a, c ), and (a n ) converges to this value. Since π = M(a, c )I(a, c ), we have established the following: 4.9 PROPOSITION. With a n, b n, c n defined in this way, we have π = lim n log 4a n. n c n 6
17 Denote this approximation by p n. With enough determination, one can derive an error estimation, but we will just illustrate the speed of convergence by performing the first three steps. Values are rounded to six significant figures. To calculate c n, we use c n = c n /(4a n ) rather than (a n b n ), since this would require much greater accuracy in the values of a n and b n. n a n b n c n p n More accurate calculation gives the value p 3 = , showing that it actually agrees with π to eight decimal places. While this expression does indeed converge rapidly to π, it has the disadvantage that it requires the calculation of logarithms at each stage. This is overcome by the GaussBrent Salamin algorithm, which we describe in section The integrals J(a, b) and E(b) For a, b, define J(a, b) = π/ (a cos θ + b sin θ) / dθ. () Note that 4J(a, b) is the perimeter of the ellipse x /a + y /b =. Some immediate facts: (E) J(a, a) = (E) J(a, ) = π/ π/ a dθ = πa, a cos θ dθ = a, (E3) for c >, J(ca, cb) = cj(a, b); (E4) J(b, a) = J(a, b) (substitute θ = π φ); (E5) J(a, b) increases with a and with b; (E6) if a b, then πb J(a, b) πa. (E6) follows from (E) and (E5), since J(b, b) J(a, b) J(a, a). Since x + for x >, we have I(a, b) + J(a, b) π. For < b, we have x J(, b) π I(, b). 7
18 We start with some further inequalities for J(a, b). Given that J(a, b) is the length of the quarterellipse, it would seem geometrically more or less obvious that (a + b ) / J(a, b) a + b, () since this says that the curve is longer than the straightline path between the same points, but shorter than two sides of the rectangle. An analytic proof of the righthand inequality is very easy: since (x + y ) / x + y for positive x, y, we have (a cos θ + b sin θ) / a cos θ + b sin θ for each θ. Integration on [, π/] gives J(a, b) a + b. We will improve this inequality below. For the lefthand inequality in (), and some further estimations, we use the Cauchy Schwarz inequality, in both its discrete and its integral form. 5.. We have J(a, b) (a + b ) /. (3) Proof. By the CauchySchwarz inequality, a cos θ + b sin θ = a(a cos θ) + b(b sin θ) (a + b ) / (a cos θ + b sin θ) /. Integrating on [, π/], we obtain a + b (a + b ) / J(a, b), hence (3). A slight variation of this reasoning gives a second lower bound, improving upon (E6). 5.. We have J(a, b) π (a + b). (4) 4 Proof. By the CauchySchwarz inequality, a cos θ + b sin θ = (a cos θ) cos θ + (b sin θ) sin θ (a cos θ + b sin θ) / (cos θ + sin θ) / = (a cos θ + b sin θ) /. Integration gives (4), since π/ cos θ dθ = π/ sin θ dθ = π 4. Note that equality holds in (3) when b =, and in (4) when b = a. 8
19 5.3. We have Equality holds when a = b. J(a, b) π (a + b ) /. (5) Proof. By the CauchySchwarz inequality for integrals (with both sides squared), ( ) π/ π/ J(a, b) (a cos θ + b sin θ) dθ = π π 4 (a + b ). Since π/( ).7, J(a, b) is actually modelled fairly well by (a + b ) /. For u = (a, b), write u = (a + b ) / : this is the Euclidean norm on R. In this notation, we have J(a, b) = π/ (a cos θ, b sin θ) dθ. The triangle inequality for vectors says that u + u u + u. We show that J(a, b) also satisfies the triangle inequality (so in fact is also a norm on R ) Let u j = (a j, b j ) (j =, ), where a j, b j. Then Proof. For each θ, we have J(u + u ) J(u ) + J(u ). (6) [(a + a ) cos θ, (b + b ) sin θ] (a cos θ, b sin θ) + (a cos θ, b sin θ). Integrating on [, π ], we deduce that J(a + a, b + b ) J(a, b ) + J(a, b ). By suitable choices of u and u, we can read off various inequalities for J(a, b). Firstly, J(a, b) J(a, ) + J(, b) = a + b, as seen in (). Secondly: Second proof of (4). Since (a + b, a + b) = (a, b) + (b, a), we have (a + b) π = J(a + b, a + b) J(a, b) + J(b, a) = J(a, b). Thirdly, we can compare J(a, b) with the linear function of b that agrees with J(a, b) at b = and b = a. The resulting upper bound improves simultaneously on those in (E6) and (). 9
20 5.5. For b a, we have J(a, b) a + Equality holds when b = and when b = a. Proof. Since (a, b) = (a b, ) + (b, b), ( π ) b. (7) J(a, b) J(a b, ) + J(b, b) = (a b) + π b. Inequality (7) reflects the fact that J(a, b) is a convex functon of b for fixed a; this also follows easily from (6). We now consider equivalent and related integrals We have J(a, b) = b (x + a ) / dx. (8) (x + b ) 3/ Proof. Substituting x = b tan θ, we obtain (x + a ) / dx. = (x + b ) 3/ π/ (a + b tan θ) / b sec θ dθ b 3 sec 3 θ = π/ (a + b tan θ) / cos θ dθ b = J(a, b). b The complete elliptic integral of the second kind is E(b) = π/ Since cos θ + b sin θ = b sin θ, we have J(, b) = E(b ). ( b sin θ) / dθ. (9) Clearly, E() = π/, E() = and E(b) decreases with b. Also, K(b) E(b) = π/ ( b sin θ) ( b sin dθ = θ) / π/ b sin θ ( b sin dθ. () θ) / We use the usual mathematical notation E (b) for the derivative d E(b). We alert the db reader to the fact that some of the literature on this subject, including [Bor], writes b where we have b and then uses E (b) to mean E(b ) We have E (b) = [E(b) K(b)] for < b <. () b
21 Proof. This follows at once by differentiating under the integral sign in (9) and comparing with () We have E(b) = ( b t ) / ( t ) / dt. () Proof. Denote this integral by I. The substitution t = sin θ gives I = π/ ( b sin θ) / cos θ cos θ dθ = E(b) For b, b π E(b) 4 b. (3) Proof. For x, we have x ( x) / x. Hence E(b) π/ and similarly for the upper bound. ( b sin θ) dθ = π ( b ), When applied to J(, b) = E(b ), (3) gives weaker inequalities than (4) and (5). We now give the power series expression. Recall from.4 that π K(b) = n= d nb n, where d = and d n =.3... (n ).4... (n). 5. PROPOSITION. For b <, π E(b) = d n n bn = 4 b 3 64 b4 + (4) n= where Proof. By the binomial series, ( b sin θ) / = + ( ) a n = ( ) n = ( )n ( n n! )( ) ( a n b n sin n θ, n=... (n 3) n + ) = (n) For fixed b <, the series is uniformly convergent for θ π. The statement follows, since π/ sin n θ dθ = π (n ) (n)
22 6. The integral L(a, b) Most of the material in this section follows the excellent exposition in [Lo]. Further results on I(a, b) and J(a, b) are most conveniently derived and expressed with the help of another integral. For a > and b, define L(a, b) = π/ cos θ (a cos θ + b sin dθ. () θ) / The first thing to note is that L(b, a) L(a, b). In fact, we have: 6.. For a and b >, L(b, a) = π/ sin θ (a cos θ + b sin dθ. () θ) / Proof. The substitution θ = π φ gives L(b, a) = π/ cos θ π/ sin φ (b cos θ + a sin dθ = θ) / (a cos φ + b sin dφ. φ) / Hence, immediately: 6.. For a, b >, L(a, b) + L(b, a) = I(a, b), (3) a L(a, b) + b L(b, a) = J(a, b). (4) 6.3 COROLLARY. For a, b >, (a b )L(a, b) = J(a, b) b I(a, b). (5) Some immediate facts: (L) L(a, ) = ; L(, b) is undefined; a (L) L(a, a) = π 4a ; (L3) L(ca, cb) = L(a, b) for c > ; c (L4) L(a, b) decreases with a and with b; (L5) L(a, b) a for all b; (L6) if a b, then π 4a L(a, b) π, and similarly for L(b, a) (by (L5) and (L)). 4b
23 6.4. We have J(a, b) = bl(b, a). (6) b Proof. Differentiate under the integral sign in the expression for J(a, b). π/4. So if we write J (b) for J(, b), then J (b) = bl(b, ). In particular, J () = L(, ) = 6.5. We have L(a, b) = L(b, a) = b dx, (7) (x + a ) / (x + b ) 3/ x dx, (8) (x + a ) / (x + b ) 3/ Proof. With the substitution x = b tan θ, the integral in (7) becomes π/ b 3 sec θ (a + b tan θ) / b 3 sec 3 θ dθ = = π/ π/ cos θ dθ (a + b tan θ) / cos θ (a cos θ + b sin dθ. θ) / The integral in (8) now follows from (3) and the integral (.) for I(a, b) If a > b, then L(a, b) < L(b, a). Proof. By (7), applied to both L(a, b) and L(b, a), L(b, a) L(a, b) = a (x + b ) b (x + a ) (x + a ) 3/ (x + b ) 3/ dx. This is positive, since the numerator is (a b )x We have b I(a, b) = L(a, b), (9) b b(a b ) b I(a, b) = b I(a, b) J(a, b). () Proof. (9) is obtained by differentiating the integral expression (.) for I(a, b) under the integral sign and comparing with (7). (The integral expressions in terms of θ do not deliver this result so readily.) Using (5) to substitute for L(a, b), we obtain (). Hence, writing I (b) for I(, b), we have I () = L(, ) = π/4. There is a corresponding expression for K (b), but we defer it to section 8. 3
24 Identities (9) and () are an example of a relationship that is expressed more pleasantly in terms of L(a, b) rather than J(a, b). More such examples will be encountered below. The analogue of the functions K(b) and E(b) is the pair of integrals L(, b ) = π/ cos θ ( b sin θ) / dθ, L(b, ) = π/ Corresponding results apply. For example, the substitution t = sin θ gives L(, b ) = ( t ) / dt ( b t ) / sin θ ( b sin dθ. θ) / (compare.3 and 5.8), and power series in terms of b are obtained by modifying.5. However, we will not introduce a special notation for these functions. Recall Theorem 4.6, which stated: log 4 < I(, b) < ( + b 4 b ) log 4 for < b <. We b now derive a corresponding estimation for J(, b) (actually using the upper bound from 4.8). 6.8 LEMMA. For < b <, where c = log 4, c = π/4. log b + c L(b, ) log b + c, () Proof. By 4.8, we have log b + log 4 I(, b) log b + π. Also, π 4 Statement () follows, by (3). L(, b). 6.9 PROPOSITION. For < b <, J(, b) = + b log b + r(b), () where c 3 b r(b) < c 4 b, with c 3 = log 4 and c 4 = 4 + π 8. Proof. Write J (t) = J(, t). By (6), J (t) = tl(t, ), so J (b) = b tl(t, ) dt. Now b ( t log t) dt = [ t log t ] b b + t dt = b log b + 4 b. By (), a lower bound for J(, b) is obtained by adding b c t dt = c b, and an upper bound by adding c b. Since J(a, ) = aj(, ), we can derive the following estimation for a >, a where c 3 /a r (a) c 4 /a. J(a, ) = a + log a a 4 + r (a),
THREE DIMENSIONAL GEOMETRY
Chapter 8 THREE DIMENSIONAL GEOMETRY 8.1 Introduction In this chapter we present a vector algebra approach to three dimensional geometry. The aim is to present standard properties of lines and planes,
More informationContinued Fractions and the Euclidean Algorithm
Continued Fractions and the Euclidean Algorithm Lecture notes prepared for MATH 326, Spring 997 Department of Mathematics and Statistics University at Albany William F Hammond Table of Contents Introduction
More informationThe cyclotomic polynomials
The cyclotomic polynomials Notes by G.J.O. Jameson 1. The definition and general results We use the notation e(t) = e 2πit. Note that e(n) = 1 for integers n, e(s + t) = e(s)e(t) for all s, t. e( 1 ) =
More informationSome Notes on Taylor Polynomials and Taylor Series
Some Notes on Taylor Polynomials and Taylor Series Mark MacLean October 3, 27 UBC s courses MATH /8 and MATH introduce students to the ideas of Taylor polynomials and Taylor series in a fairly limited
More informationGeneral Theory of Differential Equations Sections 2.8, 3.13.2, 4.1
A B I L E N E C H R I S T I A N U N I V E R S I T Y Department of Mathematics General Theory of Differential Equations Sections 2.8, 3.13.2, 4.1 Dr. John Ehrke Department of Mathematics Fall 2012 Questions
More information1.5. Factorisation. Introduction. Prerequisites. Learning Outcomes. Learning Style
Factorisation 1.5 Introduction In Block 4 we showed the way in which brackets were removed from algebraic expressions. Factorisation, which can be considered as the reverse of this process, is dealt with
More informationSolutions to Homework 10
Solutions to Homework 1 Section 7., exercise # 1 (b,d): (b) Compute the value of R f dv, where f(x, y) = y/x and R = [1, 3] [, 4]. Solution: Since f is continuous over R, f is integrable over R. Let x
More informationAdaptive Online Gradient Descent
Adaptive Online Gradient Descent Peter L Bartlett Division of Computer Science Department of Statistics UC Berkeley Berkeley, CA 94709 bartlett@csberkeleyedu Elad Hazan IBM Almaden Research Center 650
More informationFACTORING POLYNOMIALS IN THE RING OF FORMAL POWER SERIES OVER Z
FACTORING POLYNOMIALS IN THE RING OF FORMAL POWER SERIES OVER Z DANIEL BIRMAJER, JUAN B GIL, AND MICHAEL WEINER Abstract We consider polynomials with integer coefficients and discuss their factorization
More informationInner Product Spaces
Math 571 Inner Product Spaces 1. Preliminaries An inner product space is a vector space V along with a function, called an inner product which associates each pair of vectors u, v with a scalar u, v, and
More informationMOP 2007 Black Group Integer Polynomials Yufei Zhao. Integer Polynomials. June 29, 2007 Yufei Zhao yufeiz@mit.edu
Integer Polynomials June 9, 007 Yufei Zhao yufeiz@mit.edu We will use Z[x] to denote the ring of polynomials with integer coefficients. We begin by summarizing some of the common approaches used in dealing
More informationMath Review. for the Quantitative Reasoning Measure of the GRE revised General Test
Math Review for the Quantitative Reasoning Measure of the GRE revised General Test www.ets.org Overview This Math Review will familiarize you with the mathematical skills and concepts that are important
More informationDefinition of Vertical Asymptote The line x = a is called a vertical asymptote of f (x) if at least one of the following is true: f (x) =
Vertical Asymptotes Definition of Vertical Asymptote The line x = a is called a vertical asymptote of f (x) if at least one of the following is true: lim f (x) = x a lim f (x) = lim x a lim f (x) = x a
More informationCITY UNIVERSITY LONDON. BEng Degree in Computer Systems Engineering Part II BSc Degree in Computer Systems Engineering Part III PART 2 EXAMINATION
No: CITY UNIVERSITY LONDON BEng Degree in Computer Systems Engineering Part II BSc Degree in Computer Systems Engineering Part III PART 2 EXAMINATION ENGINEERING MATHEMATICS 2 (resit) EX2005 Date: August
More informationMetric Spaces. Chapter 7. 7.1. Metrics
Chapter 7 Metric Spaces A metric space is a set X that has a notion of the distance d(x, y) between every pair of points x, y X. The purpose of this chapter is to introduce metric spaces and give some
More informationCHAPTER II THE LIMIT OF A SEQUENCE OF NUMBERS DEFINITION OF THE NUMBER e.
CHAPTER II THE LIMIT OF A SEQUENCE OF NUMBERS DEFINITION OF THE NUMBER e. This chapter contains the beginnings of the most important, and probably the most subtle, notion in mathematical analysis, i.e.,
More informationUndergraduate Notes in Mathematics. Arkansas Tech University Department of Mathematics
Undergraduate Notes in Mathematics Arkansas Tech University Department of Mathematics An Introductory Single Variable Real Analysis: A Learning Approach through Problem Solving Marcel B. Finan c All Rights
More informationSection 1.1. Introduction to R n
The Calculus of Functions of Several Variables Section. Introduction to R n Calculus is the study of functional relationships and how related quantities change with each other. In your first exposure to
More informationLimit processes are the basis of calculus. For example, the derivative. f f (x + h) f (x)
SEC. 4.1 TAYLOR SERIES AND CALCULATION OF FUNCTIONS 187 Taylor Series 4.1 Taylor Series and Calculation of Functions Limit processes are the basis of calculus. For example, the derivative f f (x + h) f
More informationSection 3.7. Rolle s Theorem and the Mean Value Theorem. Difference Equations to Differential Equations
Difference Equations to Differential Equations Section.7 Rolle s Theorem and the Mean Value Theorem The two theorems which are at the heart of this section draw connections between the instantaneous rate
More informationDeterminants, Areas and Volumes
Determinants, Areas and Volumes Theodore Voronov Part 2 Areas and Volumes The area of a twodimensional object such as a region of the plane and the volume of a threedimensional object such as a solid
More informationNumerical Analysis Lecture Notes
Numerical Analysis Lecture Notes Peter J. Olver 5. Inner Products and Norms The norm of a vector is a measure of its size. Besides the familiar Euclidean norm based on the dot product, there are a number
More informationThe Ideal Class Group
Chapter 5 The Ideal Class Group We will use Minkowski theory, which belongs to the general area of geometry of numbers, to gain insight into the ideal class group of a number field. We have already mentioned
More informationRoots of Polynomials
Roots of Polynomials (Com S 477/577 Notes) YanBin Jia Sep 24, 2015 A direct corollary of the fundamental theorem of algebra is that p(x) can be factorized over the complex domain into a product a n (x
More information5 =5. Since 5 > 0 Since 4 7 < 0 Since 0 0
a p p e n d i x e ABSOLUTE VALUE ABSOLUTE VALUE E.1 definition. The absolute value or magnitude of a real number a is denoted by a and is defined by { a if a 0 a = a if a
More informationChapter 4, Arithmetic in F [x] Polynomial arithmetic and the division algorithm.
Chapter 4, Arithmetic in F [x] Polynomial arithmetic and the division algorithm. We begin by defining the ring of polynomials with coefficients in a ring R. After some preliminary results, we specialize
More informationMark Scheme (Results) Summer 2013. GCE Core Mathematics 2 (6664/01R)
Mark (Results) Summer 01 GCE Core Mathematics (6664/01R) Edexcel and BTEC Qualifications Edexcel and BTEC qualifications come from Pearson, the world s leading learning company. We provide a wide range
More informationit is easy to see that α = a
21. Polynomial rings Let us now turn out attention to determining the prime elements of a polynomial ring, where the coefficient ring is a field. We already know that such a polynomial ring is a UF. Therefore
More informationSMT 2014 Algebra Test Solutions February 15, 2014
1. Alice and Bob are painting a house. If Alice and Bob do not take any breaks, they will finish painting the house in 20 hours. If, however, Bob stops painting once the house is halffinished, then the
More informationMATH 304 Linear Algebra Lecture 20: Inner product spaces. Orthogonal sets.
MATH 304 Linear Algebra Lecture 20: Inner product spaces. Orthogonal sets. Norm The notion of norm generalizes the notion of length of a vector in R n. Definition. Let V be a vector space. A function α
More informationA matrix over a field F is a rectangular array of elements from F. The symbol
Chapter MATRICES Matrix arithmetic A matrix over a field F is a rectangular array of elements from F The symbol M m n (F) denotes the collection of all m n matrices over F Matrices will usually be denoted
More informationPUTNAM TRAINING POLYNOMIALS. Exercises 1. Find a polynomial with integral coefficients whose zeros include 2 + 5.
PUTNAM TRAINING POLYNOMIALS (Last updated: November 17, 2015) Remark. This is a list of exercises on polynomials. Miguel A. Lerma Exercises 1. Find a polynomial with integral coefficients whose zeros include
More informationNotes on Factoring. MA 206 Kurt Bryan
The General Approach Notes on Factoring MA 26 Kurt Bryan Suppose I hand you n, a 2 digit integer and tell you that n is composite, with smallest prime factor around 5 digits. Finding a nontrivial factor
More informationLecture Notes on Polynomials
Lecture Notes on Polynomials Arne Jensen Department of Mathematical Sciences Aalborg University c 008 Introduction These lecture notes give a very short introduction to polynomials with real and complex
More information3. INNER PRODUCT SPACES
. INNER PRODUCT SPACES.. Definition So far we have studied abstract vector spaces. These are a generalisation of the geometric spaces R and R. But these have more structure than just that of a vector space.
More informationON LIMIT LAWS FOR CENTRAL ORDER STATISTICS UNDER POWER NORMALIZATION. E. I. Pancheva, A. GacovskaBarandovska
Pliska Stud. Math. Bulgar. 22 (2015), STUDIA MATHEMATICA BULGARICA ON LIMIT LAWS FOR CENTRAL ORDER STATISTICS UNDER POWER NORMALIZATION E. I. Pancheva, A. GacovskaBarandovska Smirnov (1949) derived four
More informationMATH 425, PRACTICE FINAL EXAM SOLUTIONS.
MATH 45, PRACTICE FINAL EXAM SOLUTIONS. Exercise. a Is the operator L defined on smooth functions of x, y by L u := u xx + cosu linear? b Does the answer change if we replace the operator L by the operator
More informationExercises with solutions (1)
Exercises with solutions (). Investigate the relationship between independence and correlation. (a) Two random variables X and Y are said to be correlated if and only if their covariance C XY is not equal
More informationLINE INTEGRALS OF VECTOR FUNCTIONS: GREEN S THEOREM. Contents. 2. Green s Theorem 3
LINE INTEGRALS OF VETOR FUNTIONS: GREEN S THEOREM ontents 1. A differential criterion for conservative vector fields 1 2. Green s Theorem 3 1. A differential criterion for conservative vector fields We
More informationTaylor Polynomials and Taylor Series Math 126
Taylor Polynomials and Taylor Series Math 26 In many problems in science and engineering we have a function f(x) which is too complicated to answer the questions we d like to ask. In this chapter, we will
More informationIRREDUCIBLE OPERATOR SEMIGROUPS SUCH THAT AB AND BA ARE PROPORTIONAL. 1. Introduction
IRREDUCIBLE OPERATOR SEMIGROUPS SUCH THAT AB AND BA ARE PROPORTIONAL R. DRNOVŠEK, T. KOŠIR Dedicated to Prof. Heydar Radjavi on the occasion of his seventieth birthday. Abstract. Let S be an irreducible
More informationPrime Numbers and Irreducible Polynomials
Prime Numbers and Irreducible Polynomials M. Ram Murty The similarity between prime numbers and irreducible polynomials has been a dominant theme in the development of number theory and algebraic geometry.
More information1 Lecture: Integration of rational functions by decomposition
Lecture: Integration of rational functions by decomposition into partial fractions Recognize and integrate basic rational functions, except when the denominator is a power of an irreducible quadratic.
More informationShort Programs for functions on Curves
Short Programs for functions on Curves Victor S. Miller Exploratory Computer Science IBM, Thomas J. Watson Research Center Yorktown Heights, NY 10598 May 6, 1986 Abstract The problem of deducing a function
More informationThe Dirichlet Unit Theorem
Chapter 6 The Dirichlet Unit Theorem As usual, we will be working in the ring B of algebraic integers of a number field L. Two factorizations of an element of B are regarded as essentially the same if
More informationa 1 x + a 0 =0. (3) ax 2 + bx + c =0. (4)
ROOTS OF POLYNOMIAL EQUATIONS In this unit we discuss polynomial equations. A polynomial in x of degree n, where n 0 is an integer, is an expression of the form P n (x) =a n x n + a n 1 x n 1 + + a 1 x
More informationt := maxγ ν subject to ν {0,1,2,...} and f(x c +γ ν d) f(x c )+cγ ν f (x c ;d).
1. Line Search Methods Let f : R n R be given and suppose that x c is our current best estimate of a solution to P min x R nf(x). A standard method for improving the estimate x c is to choose a direction
More informationUNCORRECTED PAGE PROOFS
number and and algebra TopIC 17 Polynomials 17.1 Overview Why learn this? Just as number is learned in stages, so too are graphs. You have been building your knowledge of graphs and functions over time.
More informationStudent Performance Q&A:
Student Performance Q&A: AP Calculus AB and Calculus BC FreeResponse Questions The following comments on the freeresponse questions for AP Calculus AB and Calculus BC were written by the Chief Reader,
More informationTaylor and Maclaurin Series
Taylor and Maclaurin Series In the preceding section we were able to find power series representations for a certain restricted class of functions. Here we investigate more general problems: Which functions
More informationCONTROLLABILITY. Chapter 2. 2.1 Reachable Set and Controllability. Suppose we have a linear system described by the state equation
Chapter 2 CONTROLLABILITY 2 Reachable Set and Controllability Suppose we have a linear system described by the state equation ẋ Ax + Bu (2) x() x Consider the following problem For a given vector x in
More informationwww.mathsbox.org.uk ab = c a If the coefficients a,b and c are real then either α and β are real or α and β are complex conjugates
Further Pure Summary Notes. Roots of Quadratic Equations For a quadratic equation ax + bx + c = 0 with roots α and β Sum of the roots Product of roots a + b = b a ab = c a If the coefficients a,b and c
More informationWe call this set an ndimensional parallelogram (with one vertex 0). We also refer to the vectors x 1,..., x n as the edges of P.
Volumes of parallelograms 1 Chapter 8 Volumes of parallelograms In the present short chapter we are going to discuss the elementary geometrical objects which we call parallelograms. These are going to
More informationCHAPTER 3 Numbers and Numeral Systems
CHAPTER 3 Numbers and Numeral Systems Numbers play an important role in almost all areas of mathematics, not least in calculus. Virtually all calculus books contain a thorough description of the natural,
More informationCourse Notes for Math 162: Mathematical Statistics Approximation Methods in Statistics
Course Notes for Math 16: Mathematical Statistics Approximation Methods in Statistics Adam Merberg and Steven J. Miller August 18, 6 Abstract We introduce some of the approximation methods commonly used
More informationDEFINITION 5.1.1 A complex number is a matrix of the form. x y. , y x
Chapter 5 COMPLEX NUMBERS 5.1 Constructing the complex numbers One way of introducing the field C of complex numbers is via the arithmetic of matrices. DEFINITION 5.1.1 A complex number is a matrix of
More informationInverses and powers: Rules of Matrix Arithmetic
Contents 1 Inverses and powers: Rules of Matrix Arithmetic 1.1 What about division of matrices? 1.2 Properties of the Inverse of a Matrix 1.2.1 Theorem (Uniqueness of Inverse) 1.2.2 Inverse Test 1.2.3
More informationSubsets of Euclidean domains possessing a unique division algorithm
Subsets of Euclidean domains possessing a unique division algorithm Andrew D. Lewis 2009/03/16 Abstract Subsets of a Euclidean domain are characterised with the following objectives: (1) ensuring uniqueness
More informationThe Goldberg Rao Algorithm for the Maximum Flow Problem
The Goldberg Rao Algorithm for the Maximum Flow Problem COS 528 class notes October 18, 2006 Scribe: Dávid Papp Main idea: use of the blocking flow paradigm to achieve essentially O(min{m 2/3, n 1/2 }
More informationMathematics Course 111: Algebra I Part IV: Vector Spaces
Mathematics Course 111: Algebra I Part IV: Vector Spaces D. R. Wilkins Academic Year 19967 9 Vector Spaces A vector space over some field K is an algebraic structure consisting of a set V on which are
More informationAn example of a computable
An example of a computable absolutely normal number Verónica Becher Santiago Figueira Abstract The first example of an absolutely normal number was given by Sierpinski in 96, twenty years before the concept
More informationLinear Algebra Notes for Marsden and Tromba Vector Calculus
Linear Algebra Notes for Marsden and Tromba Vector Calculus ndimensional Euclidean Space and Matrices Definition of n space As was learned in Math b, a point in Euclidean three space can be thought of
More informationLies My Calculator and Computer Told Me
Lies My Calculator and Computer Told Me 2 LIES MY CALCULATOR AND COMPUTER TOLD ME Lies My Calculator and Computer Told Me See Section.4 for a discussion of graphing calculators and computers with graphing
More informationLecture 6. Inverse of Matrix
Lecture 6 Inverse of Matrix Recall that any linear system can be written as a matrix equation In one dimension case, ie, A is 1 1, then can be easily solved as A x b Ax b x b A 1 A b A 1 b provided that
More informationRAJALAKSHMI ENGINEERING COLLEGE MA 2161 UNIT I  ORDINARY DIFFERENTIAL EQUATIONS PART A
RAJALAKSHMI ENGINEERING COLLEGE MA 26 UNIT I  ORDINARY DIFFERENTIAL EQUATIONS. Solve (D 2 + D 2)y = 0. 2. Solve (D 2 + 6D + 9)y = 0. PART A 3. Solve (D 4 + 4)x = 0 where D = d dt 4. Find Particular Integral:
More informationx if x 0, x if x < 0.
Chapter 3 Sequences In this chapter, we discuss sequences. We say what it means for a sequence to converge, and define the limit of a convergent sequence. We begin with some preliminary results about the
More informationFactorization Theorems
Chapter 7 Factorization Theorems This chapter highlights a few of the many factorization theorems for matrices While some factorization results are relatively direct, others are iterative While some factorization
More informationDETERMINANTS. b 2. x 2
DETERMINANTS 1 Systems of two equations in two unknowns A system of two equations in two unknowns has the form a 11 x 1 + a 12 x 2 = b 1 a 21 x 1 + a 22 x 2 = b 2 This can be written more concisely in
More informationMAT188H1S Lec0101 Burbulla
Winter 206 Linear Transformations A linear transformation T : R m R n is a function that takes vectors in R m to vectors in R n such that and T (u + v) T (u) + T (v) T (k v) k T (v), for all vectors u
More informationFurther Study on Strong Lagrangian Duality Property for Invex Programs via Penalty Functions 1
Further Study on Strong Lagrangian Duality Property for Invex Programs via Penalty Functions 1 J. Zhang Institute of Applied Mathematics, Chongqing University of Posts and Telecommunications, Chongqing
More informationMathematical Methods of Engineering Analysis
Mathematical Methods of Engineering Analysis Erhan Çinlar Robert J. Vanderbei February 2, 2000 Contents Sets and Functions 1 1 Sets................................... 1 Subsets.............................
More informationDifferentiating under an integral sign
CALIFORNIA INSTITUTE OF TECHNOLOGY Ma 2b KC Border Introduction to Probability and Statistics February 213 Differentiating under an integral sign In the derivation of Maximum Likelihood Estimators, or
More informationSection 4.4 Inner Product Spaces
Section 4.4 Inner Product Spaces In our discussion of vector spaces the specific nature of F as a field, other than the fact that it is a field, has played virtually no role. In this section we no longer
More informationBANACH AND HILBERT SPACE REVIEW
BANACH AND HILBET SPACE EVIEW CHISTOPHE HEIL These notes will briefly review some basic concepts related to the theory of Banach and Hilbert spaces. We are not trying to give a complete development, but
More informationCONTINUED FRACTIONS, PELL S EQUATION, AND TRANSCENDENTAL NUMBERS
CONTINUED FRACTIONS, PELL S EQUATION, AND TRANSCENDENTAL NUMBERS JEREMY BOOHER Continued fractions usually get shortchanged at PROMYS, but they are interesting in their own right and useful in other areas
More informationCONTINUED FRACTIONS AND PELL S EQUATION. Contents 1. Continued Fractions 1 2. Solution to Pell s Equation 9 References 12
CONTINUED FRACTIONS AND PELL S EQUATION SEUNG HYUN YANG Abstract. In this REU paper, I will use some important characteristics of continued fractions to give the complete set of solutions to Pell s equation.
More informationSection 4.2: The Division Algorithm and Greatest Common Divisors
Section 4.2: The Division Algorithm and Greatest Common Divisors The Division Algorithm The Division Algorithm is merely long division restated as an equation. For example, the division 29 r. 20 32 948
More informationModern Optimization Methods for Big Data Problems MATH11146 The University of Edinburgh
Modern Optimization Methods for Big Data Problems MATH11146 The University of Edinburgh Peter Richtárik Week 3 Randomized Coordinate Descent With Arbitrary Sampling January 27, 2016 1 / 30 The Problem
More informationThe Relation between Two Present Value Formulae
James Ciecka, Gary Skoog, and Gerald Martin. 009. The Relation between Two Present Value Formulae. Journal of Legal Economics 15(): pp. 6174. The Relation between Two Present Value Formulae James E. Ciecka,
More informationSolutions for Review Problems
olutions for Review Problems 1. Let be the triangle with vertices A (,, ), B (4,, 1) and C (,, 1). (a) Find the cosine of the angle BAC at vertex A. (b) Find the area of the triangle ABC. (c) Find a vector
More information19.6. Finding a Particular Integral. Introduction. Prerequisites. Learning Outcomes. Learning Style
Finding a Particular Integral 19.6 Introduction We stated in Block 19.5 that the general solution of an inhomogeneous equation is the sum of the complementary function and a particular integral. We have
More informationInequalities of Analysis. Andrejs Treibergs. Fall 2014
USAC Colloquium Inequalities of Analysis Andrejs Treibergs University of Utah Fall 2014 2. USAC Lecture: Inequalities of Analysis The URL for these Beamer Slides: Inequalities of Analysis http://www.math.utah.edu/~treiberg/inequalitiesslides.pdf
More informationSome Polynomial Theorems. John Kennedy Mathematics Department Santa Monica College 1900 Pico Blvd. Santa Monica, CA 90405 rkennedy@ix.netcom.
Some Polynomial Theorems by John Kennedy Mathematics Department Santa Monica College 1900 Pico Blvd. Santa Monica, CA 90405 rkennedy@ix.netcom.com This paper contains a collection of 31 theorems, lemmas,
More informationInformation Theory and Coding Prof. S. N. Merchant Department of Electrical Engineering Indian Institute of Technology, Bombay
Information Theory and Coding Prof. S. N. Merchant Department of Electrical Engineering Indian Institute of Technology, Bombay Lecture  17 ShannonFanoElias Coding and Introduction to Arithmetic Coding
More information2 Complex Functions and the CauchyRiemann Equations
2 Complex Functions and the CauchyRiemann Equations 2.1 Complex functions In onevariable calculus, we study functions f(x) of a real variable x. Likewise, in complex analysis, we study functions f(z)
More informationComplex Numbers and the Complex Exponential
Complex Numbers and the Complex Exponential Frank R. Kschischang The Edward S. Rogers Sr. Department of Electrical and Computer Engineering University of Toronto September 5, 2005 Numbers and Equations
More informationLectures 56: Taylor Series
Math 1d Instructor: Padraic Bartlett Lectures 5: Taylor Series Weeks 5 Caltech 213 1 Taylor Polynomials and Series As we saw in week 4, power series are remarkably nice objects to work with. In particular,
More information11 Ideals. 11.1 Revisiting Z
11 Ideals The presentation here is somewhat different than the text. In particular, the sections do not match up. We have seen issues with the failure of unique factorization already, e.g., Z[ 5] = O Q(
More informationSolutions of Equations in One Variable. FixedPoint Iteration II
Solutions of Equations in One Variable FixedPoint Iteration II Numerical Analysis (9th Edition) R L Burden & J D Faires Beamer Presentation Slides prepared by John Carroll Dublin City University c 2011
More informationProperties of BMO functions whose reciprocals are also BMO
Properties of BMO functions whose reciprocals are also BMO R. L. Johnson and C. J. Neugebauer The main result says that a nonnegative BMOfunction w, whose reciprocal is also in BMO, belongs to p> A p,and
More informationContinued Fractions. Darren C. Collins
Continued Fractions Darren C Collins Abstract In this paper, we discuss continued fractions First, we discuss the definition and notation Second, we discuss the development of the subject throughout history
More information3.3. Solving Polynomial Equations. Introduction. Prerequisites. Learning Outcomes
Solving Polynomial Equations 3.3 Introduction Linear and quadratic equations, dealt within Sections 3.1 and 3.2, are members of a class of equations, called polynomial equations. These have the general
More informationGeometry of Vectors. 1 Cartesian Coordinates. Carlo Tomasi
Geometry of Vectors Carlo Tomasi This note explores the geometric meaning of norm, inner product, orthogonality, and projection for vectors. For vectors in threedimensional space, we also examine the
More informationInner Product Spaces and Orthogonality
Inner Product Spaces and Orthogonality week 34 Fall 2006 Dot product of R n The inner product or dot product of R n is a function, defined by u, v a b + a 2 b 2 + + a n b n for u a, a 2,, a n T, v b,
More informationMATH 132: CALCULUS II SYLLABUS
MATH 32: CALCULUS II SYLLABUS Prerequisites: Successful completion of Math 3 (or its equivalent elsewhere). Math 27 is normally not a sufficient prerequisite for Math 32. Required Text: Calculus: Early
More informationSection 4.4. Using the Fundamental Theorem. Difference Equations to Differential Equations
Difference Equations to Differential Equations Section 4.4 Using the Fundamental Theorem As we saw in Section 4.3, using the Fundamental Theorem of Integral Calculus reduces the problem of evaluating a
More informationON CERTAIN DOUBLY INFINITE SYSTEMS OF CURVES ON A SURFACE
i93 c J SYSTEMS OF CURVES 695 ON CERTAIN DOUBLY INFINITE SYSTEMS OF CURVES ON A SURFACE BY C H. ROWE. Introduction. A system of co 2 curves having been given on a surface, let us consider a variable curvilinear
More informationRoots and Coefficients of a Quadratic Equation Summary
Roots and Coefficients of a Quadratic Equation Summary For a quadratic equation with roots α and β: Sum of roots = α + β = and Product of roots = αβ = Symmetrical functions of α and β include: x = and
More informationPutnam Notes Polynomials and palindromes
Putnam Notes Polynomials and palindromes Polynomials show up one way or another in just about every area of math. You will hardly ever see any math competition without at least one problem explicitly concerning
More informationDecember 4, 2013 MATH 171 BASIC LINEAR ALGEBRA B. KITCHENS
December 4, 2013 MATH 171 BASIC LINEAR ALGEBRA B KITCHENS The equation 1 Lines in twodimensional space (1) 2x y = 3 describes a line in twodimensional space The coefficients of x and y in the equation
More information