ENGR 1405 Engineering Mathematics 1

Faculty of Engineering and Applied Science
1999 Fall


Problem Set 7
Solutions

  1. Find the Taylor or Maclaurin series for each of the following, and discuss convergence

    1.   f (x) = sin (sqrt(x)) ;   in powers of "x"

      f (x) = sin Öx f (0) = 0
      f ' (x) = cos Öx / (2 Öx) f ' (0)   is undefined

      Therefore the Maclaurin series in integer powers of "x" itself does not exist.
      However, the Maclaurin series for sin x is a standard, quotable power series
      sin x = x - x^3/3! + x^5/5! - x^7/7! + ...

      Replacing "x" by "Öx" wherever it occurs generates the power series expression
      sin(sqrt{x}) = sqrt{x} Sum{(-1)^n*x^n/(2n+1)}
      = sqrt{x}*(1 - x/3! + x^2/5! - x^3/7! + ...)

      The interval of convergence is   I = [0, ¥)


    2.   f (x) = cosh (2x) = ½(e2x + e-2x) ;   in powers of "x"

      f (x) = cosh (2x) f (0) = cosh 0 = 1
      f ' (x) = 2 sinh (2x) f ' (0) = 2 sinh 0 = 0
      f '' (x) = 4 cosh (2x) f '' (0) = 4 cosh 0 = 4

      :
      :

      :
      :

      f (n) (x) = 2n cosh (2x) if n is even;
      f (n) (x) = 2n sinh (2x) if n is odd
      f (n) (0) = 2n if n is even;
      f (n) (0) = 0 if n is odd.

      The general term is   an xn , where an = 2n / (n !) if n is even and 0 if n is odd.
      The Maclaurin series expansion for cosh 2x is therefore
      cosh(2x) = 1 + (2x)^2/2! + (2x)^4/4! + (2x)^6/6! + ...

      Alternatively, the standard Maclaurin series expansion for ex leads directly to

      (1/2)*(Sum{(2x)^n/n!} + Sum{(-2x)^n/n!})

      = (2/2)*Sum{(2x)^(2k)/(2k)!} = ...

      Or use the relationship   cosh x = cos jx together with the standard Maclaurin series expansion of cos x and replace x by 2 jx.

      From any of the above, it then follows that the series converges absolutely for all x.


    3.   f(x) = 1 / sqrt{8-x^3};   in powers of "x"

      Re-express f (x) in the form (1 + X)n .
      The binomial expansion then gives

      f(x) = sqrt{2}/4 * (1 + (-x/2)^3)^(-1/2)
      [details]
      = sqrt{2}/4 * (1 + Sum{(1*3*5*...*(2r-1)/(2^r*r!)*(x/2)^(3r)}

      This converges absolutely for |(-x/2)^3| < 1 Þ -2 < x < +2
      and converges conditionally at   x = -2 (because -1 < n < 0).

      Therefore R = 2 and I = [-2, 2) .


    4.   f(x) = 1 / (x-6)^2;   in powers of "x - 3"

      We cannot use the binomial expansion because the centre c = 3 ¹ 0.
      [However, see the alternative solution below.]

      f (x) = (x - 6)-2 f (3) = (3 - 6)-2 = +1 / 32
      f ' (x) = (- 2) (x - 6)-3 f ' (3) = +2 / 33
      f '' (x) = (-2) (-3) (x - 6)-3 f '' (3) = +3 ! / 34
      ::
      ::
      f (r)(x) = (-2) (-3) ... (-(r+1)) (x - 6)-(r + 2) f (r) (3) = +(r+1) ! / 3r+2

      Þ     ar = f (r) (3) / r ! = (r+1) ! / (r ! 3r+2)
      = (r+1) / 3r+2

      Therefore the Taylor series is
      1/(x-6)^2 = Sum{(r+1)*(x-3)^r/(3^(r+2))}

      = (1/9)*Sum{(r+1)*((x-3)/3)^r}

      Ratio test:

      u_n+1/u_n = (n+2)/(n+1) * (x-3)/3
      lim{|u_n+1/u_n|} = |x-3| / 3

      The series converges absolutely if   | x - 3 | < 3     Þ     -3 < x - 3 < +3
      Þ     3 - 3 < x < 3 + 3
                c - R < x < c + R
      Þ     R = 3.

      At the lower endpoint x = 0, the series is (1/9)*Sum{(n+1)*(-1)^n} and | un | ® ¥ as n ® ¥ .

      At the upper endpoint x = 6, the series is (1/9)*Sum{(n+1)} and un ® ¥ as n ® ¥ .

      The series therefore diverges at both endpoints (by the divergence test).

      Therefore I = (0, 6) .


      or

      Use the linear transformation   y = x - 3 , then
      1/(x-6)^2 = ... = (1/9)*(1+(-y/3))^(-2)
      The Taylor series about   x = 3   becomes the Taylor series about   y = 3 - 3 = 0, (that is, the Maclaurin series), so that the binomial expansion shortcut can now be used.

      1/(x-6)^2 = (1/9)*
         (1 + Sum_r=1^oo {(-2)(-3)(-4)...(-r-1)/r! * (-y/3)^r}
      ... = (1/9)*Sum{(r+1)*((x-3)/3)^r}

      Furthermore, the ratio test is no longer needed to find the radius of convergence.
      The binomial expansion is valid for |-y/3| < 1  ==>  | (x-3)/3 | < 1.
      The remainder of the solution is as before.


    5.   f (x) = ln (2x 3 ) ;   in powers of "x - 4"

      f (x) = ln (2x3) = ln 2 + ln x3 = ln 2 + 3 ln x.

      The Taylor series about the centre c = 4 is required:

      f (x) = ln 2 + 3 ln x f (4) = ln 2 + 3 ln 4 = ln 2 + 6 ln 2 = 7 ln 2
      f ' (x) = 0 + 3 x-1 f ' (4) = 3 / 4
      f '' (x) = 3 (-1) x-2 f '' (4) = -3 / 42
      f ''' (x) = 3 (-1) (-2) x-3 f ''' (4) = +3 × 2! / 43
      f (4) (x) = 3 (-1) (-2) (-3) x-4 f (4) (4) = - 3 × 3! / 44
      ::
      ::
      f (r) (x) = 3 (-1)r-1 (r-1)! x-r f (r) (4) = 3 (-1)r-1 (r-1)! / 44       (r > 0)

      Þ     ur = 3 (-1)r-1 (r-1)! (x-4) r / (r! 44) = 3 (-1)r-1 ((x-4) / 4) r / r       (r > 0)

      Therefore the Taylor series for   f (x) = ln (2x3) is

      7 ln 2 + 3 Sum{(-1)^r/r * ((x-4)/4)^r}

      This series converges absolutely for   | (x-4) / 4 | < 1
      Þ     -4 < x - 4 < +4
      Þ     4 - 4 < x < 4 + 4
                c - R < x < c + R
      Þ     R = 4.

      At the lower endpoint x = 0 , f (0) = ln 0 which is undefined (not finite).
      Therefore the series is invalid at x = 0.
      or
      at x = 0 u_n = ... = -1/n (harmonic series), which diverges.

      At the upper endpoint x = 8, u_n = ... = -(-1)^n / n (alternating harmonic series), which converges conditionally.

      Therefore I = (0, 8] .


    6.   f (x) = Arctan(x) = Integ_t=0^t=x { 1/(1+t^2) } dt ;   in powers of "x"

      f(x) = Arctan x = Int_0^x{dt/(1+t^2)},     c = 0
      Note that (d/dx)Int_0^x{f(t) dt} = f(x)

      f (x) = Arctan x f (0) = 0
      f ' (x) = (1 + x2)-1 f ' (0) = 1

      One could continue to find the higher order derivatives of f (x), but this would become very tedious very quickly.   The quotient (or product) rule will be needed from the second derivative onwards.

      Instead, note that   f ' (x) = (1 + x2)-1   can be expressed in its own right as a Maclaurin series, whose terms can be found quickly from a binomial expansion.

      f'(x) = 1 + Sum{(-1)*(-2)*...*(-r)*(x^2)^r / r!}
      f'(x) = Sum{(-x^2)^r}    (|-x^2| < 1)
      Anti-differentiating term by term, we obtain
      f(x) = Arctan x = Sum{(-1)^r/(2r+1) * x^(2r+1)} + C
      But   f (0) = 0       Þ     C = 0.

      Therefore the Maclaurin series for Arctan x is
      Sum{(-1)^r/(2r+1) * x^(2r+1)}
      x - x^3/3 + x^5/5 - x^7/7 + ...

      This series converges absolutely for -1 < x < +1 .
      At both endpoints (x = ±1), the series becomes
      +- Sum{(-1)^r / (2r+1)}
      This is a relative of the alternating p series with p = 1, which is conditionally convergent.   Convergence can be established quickly by the alternating series test, as a_n = 1 / (2r+1) clearly forms a sequence that decreases to zero.
      The limit comparison test, with bn = 1 / n , will establish the divergence of Sum{1 / (2r+1)}

      Therefore the series converges conditionally at both endpoints and I = [-1, 1] .


  1. Determine the first five (5) non-zero terms of the Maclaurin series expansion for

    1. f (x) = tan (x)

      First note that the function   tan x   is an odd function of x.   The first five non-zero terms of the Maclaurin series expansion of f (x) = tan x will therefore take us up to the term in x9.

      f (x) = tan x f (0) = 0
      f ' (x) = sec2x f ' (0) = 1
      f '' (x) = 2 sec2x tan x f '' (0) = 0
      f ''' (x) = 2 (2 sec2x tan2x + sec4x) f ''' (0) = 2
            = 2 (3 sec4x - 2 sec2x)
      f (4)(x) = 2 (12 sec4x tan x - 4 sec2x tan x) f (4) (0) = 0
            = 8 (3 sec4x - sec2x) tan x
      f (5)(x) = 8 (12 sec4x tan x - 2 sec2x tan x) tan x + 8 (3 sec4x - sec2x) sec2x
            = 8 (15 sec6x - 15 sec4x + 2 sec2x) f (5) (0) = 16
      f (6)(x) = 8 (90 sec6x - 60 sec4x + 4 sec2x) tan x f (6) (0) = 0
            = 16 (45 sec6x - 30 sec4x + 2 sec2x) tan x
      f (7)(x) = 16 (270 sec6x - 120 sec4x + 4 sec2x) tan2x
            + 16 (45 sec6x - 30 sec4x + 2 sec2x) sec2x
            = 16 (315 sec8x - 420 sec6x + 126 sec4x - 4 sec2x)
      f (7) (0) = 272
      f (8)(x) = 16 (2520 sec8x - 2520 sec6x + 504 sec4x - 8 sec2x) tan x
      f (8) (0) = 0
      f (9)(x) = 16 (20160 sec8x - 15120 sec6x + 2016 sec4x - 16 sec2x) tan2x
            + 16 (2520 sec8x - 2520 sec6x + 504 sec4x - 8 sec2x) sec2x
      f (9) (0) = 7936

      Therefore
      T_9(x) = Sum_0^9{f^(n)(0) / n! * x^n}

      = 1/1!*x + 2/3!*x^3 + 16/5!*x^5 + 272/7!*x^7 + 7936/9!*x^9

      = x + (1/3)x^3 + (2/15)x^5 + (17/315)x^7 + (62/2835)x^9

      [Note:   this answer can be verified using a good symbolic differentiation program such as Maple.]


    2. f ( x ) = ln [cos (x)]

      f (x) = ln(cos x)
      ==>  f'(x) = -sin x / cos x = - tan x

      But we have just evaluated the first five non-zero terms in the Maclaurin series expansion of f (x) = tan x.   Therefore
      f(x) = -Integ{x + (1/3)x^3 + (2/15)x^5 + (17/315)x^7 + (62/2835)x^9}
      = -(x^2/2 + x^4/4 + x^6/45 + 17x^8/2520 + 31x^10/14175) + C
      But   f (0) = ln(cos 0) = ln 1 = 0     Þ     C = 0
      Therefore

      ln(cos x) = -(x^2/2 + x^4/4 + x^6/45 + 17x^8/2520 + 31x^10/14175)