MATH22620 Project
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
QUESTION 1
(a) To show that if both H₁(x) and H₂(x) solve the interpolation problem, then H₁(x) = H₂(x), we can use the uniqueness theorem for interpolating polynomials, which states that there can be only one polynomial of degree at most n that interpolates n + 1 given points.
Suppose that H₁(x) and H₂(x) both solve the interpolation problem, i.e., H₁(xj) =
f(xj) and H’₁(xj) = f’ (xj), and H₂(xj) = f(xj) and H’₂(xj) = f’ (xj), for j = 0, 1 … …. , n.
H(Le)₁(t)(x(d))−=H) )f(h)s p0(o)a(yn)n(o)d(m)d’(ia)xj(of))drH’(ee)xo H’(2n)xj(1)T f(n)’)−=f’ (xj) =
0, for j = 0, 1… …. , n.
Therefore, d(x) ℎas n + 1 distinct roots at x0, x1, . . . , xn. But a polynomial of degree at most 2n + 1 can have at most 2n + 1 roots, so d(x) must be identically zero. Thus, H₁(x) = H₂(x), and the interpolation problem ( ∗ ) has a unique solution.
(b) (i) We have
n H xp ) = ILjxp2Bjxp j=0 |
= Lp xp 2Bp xp + 2 Ljxp Lp xp Bj xp
j≠p
= Bp xp ) + 2
j≠p
f xj − Bp xj)Ljxp
xp〕= |
oxp – xj) xp – xj |
= 0 foT j ≠ p. |
Taking the derivative with respect to x and evaluating at xp, we get
H' xp) = B' xp − 2 f xj − Bp xj))Lj’ xpxj)2
= B’p xp − 2 f xx(j)p−−Bxj(p)xj) Lj’ xpxj)
= B’p xp − 2
j≠p
(f xj) − Bp xj)
L ’p xp〕
We want to find conditions relating f(xp) and f’ (xp) to Bp(xp) and B’p(xp). We can use the given
interpolation conditions to eliminate some of the terms in the expression above. Specifically, we know
that H(xp) = f(xp), which means that Bp(xp) = f(xp), and we also know that H’ (xp) = f’ (xp), which means that B’p(xp) = f’ (xp). Substituting these into the expression above, we get:
H' xp = f’ xp) − 2
j≠p
(f x(xj)p fxj(x)p
L ’p xp)
= f’ xp − 2
j≠p
f xj − f xp)
L ’p xp)
To simplify the expression further, we can use the following property of Lagrange polynomials:
n
Ljx = 1 foT all x
j=0
This means that:
Lp xp = 1, and Ljxp = 0
j≠p
Taking the derivative of the first equation with respect to x and evaluating at xp, we get:
L’p(xp) = 0
Substituting this into the expression for H’ (xp), we get:
H' xp = f’ xp − 2
f x0 − f xp)
L’p xp − 2
n
j=1, j≠p
f xj − f xp)
L ’p xp
= f’ xp) − 2
f x0 − f xp)
(0 − 2
n
j=1, j≠p
f x(xj)j x(f)pxp
0)
= f’ xp) + 2
= f’ xp) + 2
n
j=1, j≠p
n
j=1, j≠p
f xp(xp) x(f)jxj))
(f xj(xp) x(f)pxj))
This is our final expression for H’ (xp).
(ii) To construct the suitable functions Bp(x), we can use the polynomial of the lowest degree that
passes through the nodes xj and satisfies the condition Bp(xp) = f(xp). This polynomial is given by
Bp(x) = σLj(xj)xp(Lj)x
where Lj(x) is tℎej − tℎ Lagrange polynomial. The degree of this polynomial is at most n, and it is
unique since it is determined by the nodes and the function values at those nodes. We can then
substitute Bp(x) and L’p(xp) into the equation for H' (xp) to obtain a formula for the derivative at xp.
(c) (i) To show that F’ (t) has at least 2n + 2 distinct roots in the interval [a, b], we can use the Rolle's theorem repeatedly. Note that F(t) is a polynomial of degree at most 2n + 1. Since F(a) = F(b) = F' (a) = F' (b) = 0, there must be at least n + 1 distinct roots of F' (t) in the interval (a, b) by
Rolle's theorem. Similarly, F'' (a) = F'' (b) = 0, so there must be at least n distinct roots of F'' (t) in the interval (a, b) by Rolle's theorem. Continuing this process, we can show that F2n+2t has at least one root in [a, b].
(ii) From the previous part, we know tℎat F2n+2t has at least one root in [a, b]. Let C be a root of F2n+2t〕in [a, b]. Then by Taylor's theorem, we have
f(x) − H(x) = F(x) + R(x)
where R(x) is the remainder term given by
R x = x − C2n+2 f2n+2ξ
for some ξ between x and C. Since c is a root of F2n+2 t , we have R(x) = 0 wℎen x = C, so
f x − H x = F x + R x = F x) + 0 x − C2n+2)
where the notation 0 x – C〕2n+2) indicates that the error term is of order x – C〕2n+2 or smaller. Therefore, the error in the approximation f(x) ≈ H(x) is of order x – C2n+2 or smaller.
(d) (i) In the limit E → 0, the divided difference table will converge to the derivative of the function at x0, since
f x0, x0 + E = lim f x0 + E − f x0) = f' (x0)
ℎ→0 E
and
f x0, x0 + E, x0 + 2E = lim f x0 + E, x0 + 2E1−f x0, x0 + E) = f''(x0)
ℎ→0 E
and so on for higher-order divided differences.
(ii) To construct the Newton polynomial in the limit E → 0, we can start by writing the divided difference table with two nodes:
x0 f[x0]
x0 + E f[x0, x0 + E]
In the limit as E approaches 0, the divided difference f[x0, x0 + E] approaches f' (x0), and the table converges to:
x0 f[x0]
x0 + E f' (x0)
We can then use the standard formula for the Newton polynomial with two nodes:
P(x) = f[x0] + (x − x0)f[x0, x0 + E]
In the limit as E approaches 0, this becomes:
P(x) = f(x0) + (x − x0)f' (x0)
This is a linear approximation to the function f(x) neaT x0, which is the same as the first-order Taylor approximation.
To show that the Newton polynomial will continue to satisfy the conditions P(x0) =
f(x0) and P’ (x0) = f’ (x0) if further nodes are added, we can use induction. Suppose we have a divided difference table with k + 1 nodes:
x0, x1, . . . , xk
In the limit as E approaches 0, the table will converge to:
x0 f[x0]
x1 f[x0, x1]
. . . . . .
xkf[x0, x1, . . . , xk]
The Newton polynomial with k + 1 nodes is:
P(x) = f[x0] + (x − x0)f[x0, x1] + . . . + (x − x0)(x − x1). . . (x − xk)f[x0, x1, . . . , xk]
In the limit as E appToacℎes 0, the divided differences f[x0, x1], f[x0, x1, x2], . . . , f[x0, x1, . . . , xk] will converge to the corresponding derivatives at x0, and the Newton polynomial will become:
P x = f x0 + x − x0 f ' x0 + … + x − x0x − x1 … x − xk(k)! fkx0
This polynomial satisfies P(x0) = f(x0) and P' (x0) = f' (x0), and we can use induction to show that it will continue to satisfy these conditions if more nodes are added to the table.
QUESTION 2
(a) Let w1, w3, and w5 be the weights for the Gk5 rule, and let t1 = − t5 and t3 = 0 due to
symmetry. Then we have the following system of equations:
w1 + w5 = 2
w1t1 + w3t3 + w5t5 = 0
w1t12 + w3t32 + w5t52 =
Since t1 = − t5 and t3 = 0, we can simplify the system to:
w1t1 −
w1t12 +
w5 = 2
w5t1 = 0
w5t12 1
= —
3 3
(b) Since P2(t) is the Legendre polynomial of degree 2, we have:
P2t) = 3t2 − 1
2
Let kat) = At3 + Bt2 + ct + D. Then we have:
1 P2tkattT dt = 0 foT T = 0, 1, 2.
Substituting P2(t) and expanding the integrand, we get:
( A 1 t5dt + ) 1 t4dt + ) 1 t3dt + ) 1 t2dt = 0
(2
(2
A 1 t4dt +
A 1 t3dt +
1 t3dt +
1 t2dt +
1 t2dt +
1 tdt +
1 tdt = 0
1 1dt = 0
Simplifying and solving for A, B, C, and D, we get:
A = , B = 0, C = , D = 0
Therefore, kat) = t3 – t.
The roots of kat)aTe t1 = − , t3 = 0, and t5 = .
(c) (i) Substituting t1 = − , t3 = 0, and t5 =
get:
into the system of equations from part (a), we
w1 + w5 = 2
− w1 + w5 = 0
(5) (7 w5 1
Solving for w1 and w5, we get:
w1 =
w5
18 2
=
45 5
28
=
45
(ii) To calculate the exact value of the first nonzero coefficient Sp in the error formula for the GK5 rule, we need to apply the error formula to a function that has a nonzero fourth derivative at some point
within each subinterval. A function that satisfies this condition is f x = x5.
We can then compute the coefficients as follows:
Sp = 1 f6t
2023-08-11