on computing gauss-kronrod quadrature formulae* · kautsky and elhay [15] and elhay and kautsky...

19
mathematics of computation volume 47, number 176 october 1986,pages 639-650 On Computing Gauss-Kronrod Quadrature Formulae* By Franca Calió, Walter Gautschi, and Elena Marchetti Abstract. We discuss the use of Newton's method for computing Gauss-Kronrod quadrature formulae from modified moments. The underlying nonlinear maps are analyzed from the point of view of numerical condition. A method is indicated of computing the polynomial whose zeros are the Kronrod nodes. Examples include Gauss-Kronrod formulae for integrals with a logarithmic and algebraic singularity at one endpoint. Pertinent numerical results are tabulated in the supplements section at the end of this issue. 1. Introduction. Given a positive measure da(t) on the real line, whose support contains infinitely many points and all of whose moments exist, we call (1.1) / f(t)do(t) = i ajir,) + £a;/(r;) + Rn(f) ■'» v=l ft-1 a Gauss-Kronrod quadrature formula if t„ = T„(n) are the Gaussian nodes associated with do and the nodes t* and weights av, a* are chosen so as to maximize the degree of exactness of (1.1). Since there are 3n + 2 unknowns, we can achieve degree of exactness d = 3n + 1, i.e., R„(f) - 0 whenever / = P3n+1. It is well known, in fact, that the nodes t* must be the zeros of w„*+1, the (monic) polynomial of degree n + 1 satisfying the orthogonality property (1.2) / wn*+x(t)irn(t)tido(t) = 0, 1 = 0,1,...,«, •'r where tt„(-) = "n„(-;do) is the nth degree orthogonal polynomial belonging to the measure da. Note that the measure da*(t) = ir„(t; da)da(t), with respect to which v*+i *s orthogonal, changes sign. We cannot expect, therefore, that 7Tn*+1 has all its zeros necessarily real. We are interested here only in Gauss-Kronrod formulae (1.1) with real nodes t*, all contained in [a, b]—the smallest interval containing the support of da—and with positive weights a*. Then the interlacing property holds (Monegato [17, Theorem 1]), (1.3) a < t„*+1< r„ < r* < t„_, < ••• < T, < T* < b. Our concern is with the actual computation of these nodes and the corresponding weights (provided they exist), given the integer n > 1 and the positive measure da. Received May 7, 1985; revised September 24, 1985. 1980 Mathematics Subject Classification. Primary 65D32, 65A05; Secondary 65H10. * The work of the first and third author was supported by the Ministero della Pubblica Istruzione and by the Consiglio Nazionale delle Ricerche, while the work of the second author was supported by the National Science Foundation under grant DCR-8320561. 639 ©1986 American Mathematical Society 0025-5718/86 $1.00 + $.25 per page

Upload: others

Post on 09-May-2020

7 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

mathematics of computationvolume 47, number 176october 1986, pages 639-650

On Computing Gauss-Kronrod Quadrature Formulae*

By Franca Calió, Walter Gautschi, and Elena Marchetti

Abstract. We discuss the use of Newton's method for computing Gauss-Kronrod quadrature

formulae from modified moments. The underlying nonlinear maps are analyzed from the

point of view of numerical condition. A method is indicated of computing the polynomial

whose zeros are the Kronrod nodes. Examples include Gauss-Kronrod formulae for integrals

with a logarithmic and algebraic singularity at one endpoint. Pertinent numerical results are

tabulated in the supplements section at the end of this issue.

1. Introduction. Given a positive measure da(t) on the real line, whose support

contains infinitely many points and all of whose moments exist, we call

(1.1) / f(t)do(t) = i ajir,) + £a;/(r;) + Rn(f)■'» v=l ft-1

a Gauss-Kronrod quadrature formula if t„ = T„(n) are the Gaussian nodes associated

with do and the nodes t* and weights av, a* are chosen so as to maximize the

degree of exactness of (1.1). Since there are 3n + 2 unknowns, we can achieve degree

of exactness d = 3n + 1, i.e., R„(f) - 0 whenever / = P3n+1. It is well known, in

fact, that the nodes t* must be the zeros of w„*+1, the (monic) polynomial of degree

n + 1 satisfying the orthogonality property

(1.2) / wn*+x(t)irn(t)tido(t) = 0, 1 = 0,1,...,«,•'r

where tt„(-) = "n„(-;do) is the nth degree orthogonal polynomial belonging to the

measure da. Note that the measure da*(t) = ir„(t; da)da(t), with respect to which

v*+i *s orthogonal, changes sign. We cannot expect, therefore, that 7Tn*+1 has all its

zeros necessarily real.

We are interested here only in Gauss-Kronrod formulae (1.1) with real nodes t*,

all contained in [a, b]—the smallest interval containing the support of da—and

with positive weights a*. Then the interlacing property holds (Monegato [17,

Theorem 1]),

(1.3) a < t„*+1 < r„ < r* < t„_, < ••• < T, < T* < b.

Our concern is with the actual computation of these nodes and the corresponding

weights (provided they exist), given the integer n > 1 and the positive measure da.

Received May 7, 1985; revised September 24, 1985.

1980 Mathematics Subject Classification. Primary 65D32, 65A05; Secondary 65H10.

* The work of the first and third author was supported by the Ministero della Pubblica Istruzione and by

the Consiglio Nazionale delle Ricerche, while the work of the second author was supported by the National

Science Foundation under grant DCR-8320561.

639

©1986 American Mathematical Society

0025-5718/86 $1.00 + $.25 per page

Page 2: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

640 FRANCA CALIÓ, WALTER GAUTSCHI, AND ELENA MARCHETTI

Constructive methods for this problem have been considered before. For the

Legendre measure da(t) = dt on [-1,1], Kronrod [16] in his original work obtains

7T„*+i in power form by solving (1.2). Patterson [22] expands n*+x in Legendre

polynomials, while Piessens and Branders [23] use Chebyshev polynomials, the nodes

t* each time being computed by an appropriate rootfinding procedure. The weights

can be obtained by interpolation. Monegato [17], [19], Baratella [1], and Dagnino

and Florentino [3] use similar procedures to compute tr*+ x for Gegenbauer measures

and other classical measures. Kahaner et al. [13], in the case of the Laguerre

measure, and assuming * > n + 1 Kronrod nodes t* in (1.1), expand Lnmk* in

Laguerre polynomials L . All these methods require three distinct phases to obtain

(1.1): The computation of the appropriate polynomial, say 7Tn*+1, finding the zeros of

tt*+x, and computing the weights ar, a*. An entirely different approach is taken by

Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as

eigenvalues of a certain matrix derived by matrix decomposition methods—an

approach which extends the well-known method of Golub and Welsch [12] for

ordinary Gaussian quadratures and its extension by Golub and Kautsky [11]. The

weights are then obtained using general methods for constructing interpolatory

quadrature rules [14], [6].

Here we propose to compute (1.1) directly by solving a set of nonlinear equations

that express the exactness of (1.1) for a given set of 3« + 2 polynomials. These

polynomials are chosen so as to generate a well-conditioned problem. The respective

system of nonlinear equations is solved by Newton's method, careful attention being

given to the choice of initial approximations and to monitoring the progress of the

iteration. We make no claims for our method to be superior in any way to other

methods, but merely demonstrate its feasibility and stability. It would indeed be

interesting to undertake a detailed comparative study of the various methods now in

use.

While aiming directly at the unknown quantities is certainly a virtue if our method

is successful, it is less than satisfactory otherwise, since no information about ir*+ x is

generated when the method fails. Nevertheless, if one wishes to examine the

polynomial it*+x, one can express it in terms of the orthogonal polynomials

{•nk(-;da)} and obtain the coefficients by solving a triangular system of linear

equations.

The implementation of Newton's method for computing the Gauss-Kronrod rule

(1.1) is discussed in Section 2. In Section 3 we study the condition of the underlying

problem. Section 4 deals with the computation of the polynomial it*+x. Examples

will be given in Section 5, and numerical results for da(t) = taln(l/t)dt on [0,1],

a = 0, + j, are tabulated in the supplements section at the end of this issue.

For additional questions related to Gauss-Kronrod quadrature we refer the reader

to surveys by Monegato [20], [21].

2. The Computation of Gauss-Kronrod Rules by Newton's Method. The Gaussian

nodes t„ for the measure da can be computed by well-known methods; see, e.g., [8,

p. 290]. We assume therefore that they are available. For a given system of monic

polynomials (pk(-)}, with degpk = k, k = 0,1,2,..., we further assume that the

Page 3: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

GAUSS-KRONROD QUADRATURE FORMULAE 641

first 3« + 2 modified moments of da,

(2.1) mk= f pk(t)da(t), k = 0,l,...,3n + l,JR

are known. The system of nonlinear equations defining the Gauss-Kronrod formula

(1.1) then is

« « + i

(2-2) L^PkU)+ L%*Pk{r;) = mk, k = 0,1,...,3« + 1.p = i ii=i

We propose to solve this system for the 3n + 2 unknowns r*, av, a* by Newton's

method. A number of practical issues need to be addressed.

First the choice of polynomials pk. We assume that they satisfy a three-term

recurrence relation,

(23) P-M = 0, /»„(/)-1,

/»* + i(0 = it-ak)pkit)-bkPk_x(t), ¿ = 0,1,2,...,

with known coefficients ak, bk. This makes the evaluation of these polynomials, and

of their derivatives, which is required in the computation of the left-hand side of the

system (2.2) and its Jacobian, easy and straightforward. If, in addition, the poly-

nomials {pk} are orthogonal with respect to some given measure ds(t), then this

computation is also numerically stable. If da is one of the classical measures, the

simplest choice is pk(-) = irk(-; da), the corresponding recursion coefficients ak =

ak(da), bk = ßk(da) then being explicitly known; the modified moments become

(2.4) mk= f nk(t;da)da(t) = ß080k, k = 0,1,2,...,•'R

where ß0 = jKda(t) and 80 k is the Kronecker delta, 8QQ = 1, 80 k = 0 if k > 0.

For nonclassical measures da, the choice of pk is usually dictated by the necessity of

being able to compute the modified moments (2.1).

The next important issue is the computation of the initial approximations for the

Kronrod nodes t* and for the weights a„, a*. Since we are interested only in

Gauss-Kronrod formulae with nodes in [a, b] and with positive weights a*, we can

assume the interlacing property (1.3). This suggests as initial approximations t*,

2 < ft < n, the midpoints

(2-5) t; = Kvi + T,)' f» = 2,3,...,«,

between the Gaussian nodes t , and •y The choice of the initial approximations tx*

and t*+ x is less obvious. If the endpoint b is finite, we select a number of trial values

tx* equally spaced between tx and b. If a is also finite, we do the same for t„*+1 in

the interval (a, rn) and let both f„*+1 and if move inward simultaneously, or, if

necessary, let them move independently from one another. If an endpoint, say b, is

infinite, we select t* = rx + (tx ' — r2), or try a number of values rx* equally spaced

between tx and, say, t, + 2(tj - t2). For each set of initial approximations f * we

compute corresponding approximations r)„, à* to the weights by solving the first

2n + 1 equations in (2.2), where t* is replaced by r*. Since the matrix of this

system forms part of the Jacobian matrix used in the first Newton step, the only

overhead of this computation is the solution of the linear system.

Page 4: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

642 FRANCA CALIÓ, WALTER GAUTSCHI, AND ELENA MARCHETTI

Each Newton step is monitored as to the location of the iterates of t*. If at any

stage one of these iterates falls outside the interval [a, b] (if either a or b is finite),

the iteration is terminated and restarted with a new set of initial approximations.

The same action is taken if the number of Newton iterates exceeds a preset limit. If

none of the initial approximations leads to a convergent process, the attempt of

computing (1.1) is declared a failure. This is usually an indication (not a proof!) that

the desired Gauss-Kronrod rule does not exist.

If da(t) = da(-t) is an even measure and the support of da is symmetric with

respect to the origin, then both the Gaussian nodes t„ and the Kronrod nodes t * are

located symmetrically with respect to the origin, and weights corresponding to

symmetric nodes are equal. As a result, (2.2) is trivially true if pk(t) is an odd

polynomial. If we choose for { pk} a system of polynomials satisfying

(2.6) pk(-t) = i-l)kpkit), k = 0,1,2,...,

the system (2.2), therefore, is equivalent to the system

«/2 «/2 l

E °„Plki%) + E ^Plki^*) + 9°(*/2) + l/>2*(0) - 9 ™2*>(2.7e) .=i n=i ¿ ¿

k = 0,1,..., 3n/2,

if n is even, and to the system

[n/2] 1 l«/2] + l j

£ o,/>2*(T,) +2a[«/21 + i^(0)+ E o*p2k(r*) = -^m2k,

(2.7o) " = 1 "=1

fc = 0,l,...,(3« + l)/2,

if n is odd. This in effect reduces the size of the problem by a factor of 2.

3. The Condition of the Underlying Problem. Let mT = [m0, mx,..., m3n+x] be the

vector of modified moments, and yT = [ax,.. .,an,ax*,...,a*+x,tx*,.. . ,t„*+1] the

vector of the weights and Kronrod nodes of the Gauss-Kronrod formula (1.1). The

procedure of Section 2 is an attempt of carrying out the nonlinear map

G„: R3"+2 - R3n+2 m -> y,

where the Gauss-Kronrod formula is assumed to have real nodes. We now wish to

examine the sensitivity of the map Gn to perturbations in the modified moments.

The development parallels the treatments given for Gaussian formulae in Gautschi

[8], [9]; see also Gautschi [10, Section 5].

We assume that the polynomials {pk} defining the modified moments are

orthogonal on the real line with respect to some measure ds. The support of this

measure normally coincides with the support of da, but does not have to. We define

normalized modified moments by

(3.1) mk = d:lmk, d2=( p2(t)ds(t),JR

and consider, in place of G„, the map

(3.2) G„: R3n+2 -> R3n+2 m - y,

where mT = [m0,mx,...,m3n+x\. We analyze the sensitivity of Gn by computing the

Frobenius norm of the Jacobian matrix, J¿ , of the map G„.

Page 5: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

GAUSS-KRONROD QUADRATURE FORMULAE 643

The basic equations can be written in the form

(3.3) 0(y) = m,

where

(3.4) *k{y) = d-A E ovPk(rv) + E^(t;)),\v=l ii-1 )

k = 0,1,2,. ..,3n + 1.

Since the map G„ amounts to solving (3.3) for y, the Jacobian of Gn is the inverse of

the Jacobian 90/tfy of 0,

(3.5) Jôm = (30/3Y)-1.

An elementary computation shows that

(3.6) 90/3y = DXP2*,

where D = diag(d0, dx.d3n+x), 2* = diag(l,..., 1,1,..., 1, af,..., a*+1), and

(3.7)

/>o(Ti) ■■■ PoM Po(t*) ••• />o(TAt) Po(T*) •• Poitf+i)

Pi(T\) ■■■ PiM pAt*) •■• PiiT*+i) /i(t*) ••■ p'iiiï+i)

P3»+l(n) "■ />3» + l(T,.) />3„ + l(T*) •'■ P3»+i(t„*+i) />3„ +1 ( V ) •" P3»+l(T»*+l)

Therefore,

(3.8) Jô> = (S*)'1?"1/).

For the inversion of P, define g„, /i^, A:^ to be the elementary Hermite interpola-

tion polynomials of degree 3« + 1, belonging to the nodes t„ and r*, defined by

£,(tx) = S„x, X = l,2,...,n \

g,(V) = °> «:(V)-°' M = l,2,...,« + l/'

*- 1,2,...,«,

MTx)-°. a = 1,2,...,«

MT*)"V h'Á<) = ̂ r-l,2,...,n + l

MTx)-°' À = 1,2,...,«

M1--*)"0' *;(T*)-V r-l,2,...,» + lj

/x = 1,2,...,« + 1.

Writing

3« + 2 3n + 2 3« + 2

«,(0 = I Vp-i(')- MO- E ¿Vp-i(0, MO- I Vp-i(').p=l p = l p=l

(3.9)

it is easily seen that

(3.10) Px = , A = [aJ, 5 =[*>,„], C=[cJ.

Page 6: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

644 FRANCA CALIÓ, WALTER GAUTSCHI, AND ELENA MARCHETTI

By a computation similar to the one in [8, pp. 303-304] one finds from (3.8) and

(3.10) that the Frobenius norm of J¿ is given by

H+l

Es,2(0+ E M0 + 4ïW1/2

ds(t)(3.11) i/a- /R

Its magnitude, therefore, is critically influenced by the magnitude of the polynomial

(3.12) /„(o= tg,2(0 + "lÍM0 + 4iM09-1 H-l\ %

on the support of ds. The degree of /„ is 6« + 2. The integral in (3.11) can therefore

be computed exactly (up to rounding errors) by a (3n + 2)-point Gaussian quadra-

ture rule belonging to the measure ds.

Explicit forms of the polynomials g„, h^, k^ can easily be given in terms of the

fundamental Lagrange polynomials

1 + 1

ao-n tA = l

/;(') = n tX = l V

belonging to the nodes t„ and t*, respectively. One obtains

?,(') =',(0%\lit)

«tf+iK)

(3.13) .(0-^[W]a(i-(*-v)

y = 1,2,...,«,

+ wM

*n(0

jtt = 1,2,...,« + 1,

,2,...,« + 1,

where tt„(-) = "t„(-; da) and n*+x(t) = nx=i(í _ t*)- We also note the following

properties of /„, which follow directly from (3.9) and (3.12),

/„(/)>0 for all r g R,

(3.14) /„(Ü = l, /„'(t,) = 2g'v(jv), v=l,2,...,n,

/.(tf)-i, /«(t;) = °> f»-i.2,...,» + i.

These conditions, of course, are not sufficient to determine the polynomial /„. There

are 2« + 1 degrees of freedom left, which allow/, considerable room for movement.

4. Computation of the Polynomial m*+,. Expressing the polynomial tt*+ x in terms

of the orthogonal polynomials iTk(-) = trk(-; da),

(4.1) 77*+1(/) = W„ + X(t) + C0Í7„(0 + Cxir„_x(t) + ■■■ +C„770(f),

and replacing the powers t' in (1.2) by the polynomials 7T,(i), one obtains the

conditions

/.Vn(0 + E cktr„_k(t)

k=0

trn(t)ni(t)da(t) = 0, i = 0,1,...,«,

Page 7: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

GAUSS-KRONROD QUADRATURE FORMULAE 645

hence the linear system

«

(4.2) ¿Zalkck = b„ 1-0,1,...,«,k = 0

for the coefficients ck, where

a/* = / */(0*',,-*(0M0<M0. i,fc = 0,l,...,n,(4.3)

*,■= - ( »,-(0»„+i(0M0<M0. / = o,i,...,«.•'R

By orthogonality, aik = 0 if i < k, so that the matrix A = [aik] is lower triangular,

and ah■ = jRir2(t)da(t) > 0, so that A is nonsingular. The solution of (4.2),

therefore, can be effected by forward substitution.

The coefficients aik and b¡ = -a¡ _x satisfy a two-dimensional recursion relation,

which could be used for their computation (see [2]). Noting, however, that the

integrands in (4.3) are polynomials of degree at most equal to 3« + 1, we can also

use w-point Gauss-Christoffel quadrature relative to the measure da, with m =

[(3« + 3)/2], to compute aik and b¡. This might be preferable, since it requires

nothing beyond standard software.

While this procedure of generating tt*+ x is similar to Kronrod's original method,

it is significantly more stable, since the use of powers as a polynomial basis is

completely avoided.

5. Examples. It is known that for da(t) = (1 - t2)x~1/2dt on [-1,1] all Gauss-

Kronrod formulae exist if 0 < X < 2. Furthermore, the interlacing property (1.3)

holds and a* > 0 for ju = 1,2,..., « + 1 (Monegato [17]). If 0 < X < 1, one has in

addition o„ > 0 for v = 1,2,..., n (Monegato [18]). Our first example deals with the

case X = \, i.e., with the Legendre measure. All computations reported were done

on the CDC 6500 computer in single precision (machine precision * 3.55 X 10~15),

unless noted otherwise.

Example 5.1. da(t) = dt on [-1,1].

Taking advantage of symmetry, we apply Newton's method to the system (2.7),

using for p2k the (monic) Legendre polynomials of degree 2k. The required modified

moments are then given by (2.4). We had no difficulty with convergence. By

symmetry, only one "trial" initial approximation, f*, is needed, which was pro-

grammed to move in nine equal steps of length h = (1 - t,)/10 from 1 - h to

t, + h. Convergence was invariably achieved for the first choice of tx*. Moreover,

the problem, suitably scaled, turns out to be extremely stable. In the first four

columns of Table 5.1 we report on the number of iterations required for 12 decimal

place accuracy, the maximum of fn(t) (cf. Eq. (3.12)) on [-1,1] and the value of

\\Jg)\f (cf- Eq- (3.11)) for n = 5, 10, 20, 40, 80. The maximum of f„(t)~an even

function of /—is typically assumed between t2 and t2*, if « is even, there being a

couple of smaller maxima on either side of it. If n is odd, the maximum seems to

occur at / = 1. Through most of the interval (-1,1), however, /„ remains < 1.

Page 8: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

646 FRANCA CALIÓ, WALTER GAUTSCHI, AND ELENA MARCHETTI

Table 5.1

Performance and stability characteristics of Newton's method for generating the

(2n + l)-point Gauss-Kronrod formula with da(t) = dt on [-1,1].

#iter. H/, H. ||/*cond, cond2

(scaled) (scaled)

5 6 1.126 1.293 1.4(3) 3.3(4) 1.2(1) 1.4(1)

10 6 2.456 1.397 1.5(6) 5.2(8) 1.8(1) 2.3(1)20 6 2.520 1.333 1.5(12) 5.3(17) 4.1(1) 4.8(1)40 6 2.535 1.310 1.6(24) 5.8(35) 6.6(1) 7.7(1)

80 6 2.540 1.302 — 1.3(2) 1.7(2)

The linear systems of equations for determining the initial approximations r}„, a*

for the weights, as well as the Jacobian matrices in Newton's method, appear to

become rapidly ill-conditioned as n increases. Typical condition number estimates

(furnished by the LINPACK routine SGECO; cf. [4, Chapter 1]) for the former are

shown in the fifth column of Table 5.1, while those for the latter are shown in the

sixth column. (Numbers in parentheses indicate decimal exponents.) In spite of the

large condition numbers, numerical difficulties were not observed, except in the case

« = 80, when the computation was aborted due to an arithmetic error. We believe

that the apparent ill-conditioning is caused by the use of monic polynomials p2k in

(2.7); their L2-norm goes to zero rather quickly,

22k(2k)\2 I 2 /-„ „^^^-JÄkyTiÄkTl-^2 as/c-oo,

thereby introducing a systematic diminution of the rows down the matrices. If the

row involving p2k is scaled by dividing it by 22/c(2/c)!2/(4A:)!, the condition numbers

indeed become much more reasonable (the solutions remaining the same); they are

shown in the last two columns of Table 5.1.

Example 5.2. da(t) = ln(l/i) dt on [0,1].

It appears that this measure also admits Gauss-Kronrod formulae for all n,

satisfying the interlacing property (1.3) and having all weights positive. A summary

of our numerical experience with Newton's method is given in Table 5.2; it contains

information analogous to the one given in Table 5.1 for Example 5.1.

Table 5.2

Newton 's method for Gauss-Kronrod formulae with da(t) = ln(l/r) dt on [0,1].

#iter. \\fn\\x \\Jà\\F cond, cond-.cond¡ cond2

(scaled) (scaled)

5 6 2.90(4) 12.14 2.0(6) 6.9(9) 5.8(1) 2.7(3)10 6 4.09(5) 24.35 2.2(12) 7.5(18) 2.8(1) 1.6(4)

20 6 5.77(6) 47.43 2.4(24) 8.2(36) 9.8(2) 1.0(5)40 6 8.86(7) 94.59 — 3.8(3) 6.4(5)

Page 9: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

GAUSS-KRONROD QUADRATURE FORMULAE 647

We have used modified moments with respect to the (monic) shifted Legendre

polynomials, pk(t) = [k\2/(2k)\]Pk*(t); they are known to be (cf., e.g., Gautschi

[7])

-o=l, ^-r^OMl/OA-^TT^TT. *-1.2,....

Row scaling of the matrices was performed through division by k\2/(2k)\ (of the

row numbered k + 1).

The initial approximations f,* and fn*+1 were programmed to first move inward

symmetrically in nine steps of length h = (1 - t^/IO and W = t„/10, respectively,

and if this did not work, tx* was varied independently over the same set of points for

each fixed t*+x. Convergence, in general, was achieved for the very first choice of tx*

and f,*+1, i.e., for fx* = 1 — h and f„*+1 = h', except when n is small, for example,

« = 1 and « = 3, in which cases convergence was realized when fn*+1 = h' and

f * = 1 - Ah (for « = 1), t* = 1 - 2A (for « = 3).

The polynomial /„(f) invariably assumes its global maximum at t = 1, has a few

much smaller relative maxima between the first few neighboring nodes t* and t„,

and then settles down to magnitudes around 1 for the remaining portion of the

interval [0,1]. The condition of the problem, though slightly worse than in Example

5.1, is still remarkably good.

Numerical results for the nodes and weights of the (2n + l)-point Gauss-Kronrod

formula for « = 5(5)25 can be found in Table S.l of the supplements section at the

end of this issue. They have been computed in double precision to an accuracy of 25

decimal places after the decimal point. As the results are displayed in D-format,

some of the end figures may not be reliable in those numbers that are much smaller

than 1.

We used Example 5.2 to further experiment with alternative choices of initial

approximations. In particular, we examined how inaccuracies in individual initial

approximations affect the speed of convergence. To obtain a basis for meaningful

comparison, we first obtained "reference" values for the number of iterations

required when all initial approximations are at the same level of accuracy. This was

achieved by imposing on the "exact" results for av, a*, t* (computed to 12 decimal

digits) a random perturbation at level e, i.e., by taking 6„ = a„(l + rve), à* =

a*(l + r*e), t* = ^(1 + s*e), where rv, r*, s* are random numbers from [-1,1].

The results for e = 10"2, 10-5, 10"8, and 10"11 are shown in Table 5.3. (For

e = 10"2 and « = 40, Newton's method did not converge within 20 iterations.)

Table 5.3

The number of iterations required for initial approximations at accuracy level e.

e n = 5 « = 10 « = 20 « = 40

10"2 6 6 5

10"5 3 3 3 4

10"8 2 2 2 2

10-" 2 2 2 2

Page 10: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

648 FRANCA CALIÓ, WALTER GAUTSCHI, AND ELENA MARCHETTI

Table 5.4

The number of iterations required when two pairs of initial approximations

(with indices n andn + 2 — ¡i) are inaccurate.

: 5 „ = 10 „ = 20 « = 40 n = 5ji = n/4

10 n = 20 n = 40 n = 511 = n/2

10 „ = 20 ■ 4»

io-2io-5

IO-8

io-11

We now contrast this with the case in which all initial approximations are at the

accuracy level ¿10~12 of the initially computed results, except for two pairs of

Kronrod nodes and weights (situated symmetrically with respect to the midpoint of

the interval [0,1]), which are randomly perturbed at level e. Choosing the inaccurate

pairs i*, a* to be those corresponding to ju. = 1, \i = n/4, and /x ~ n/2 (and to the

symmetric indices n + 2 - ¡i), we observed the results shown in Table 5.4.

Comparing Tables 5.3 and 5.4, we note some improvement, particularly for

/* == n/2, in the case e = IO"2, when only two pairs of initial approximations are

inaccurate (though for « = 20 there are two instances of deterioration), but in all

other cases the performance of Newton's iteration is practically the same. This seems

to suggest that it is the maximum relative error in the initial approximations (usually

associated with a Kronrod or Gauss node near the end of the interval) which

determines the speed of convergence.

The choice of initial approximations proposed in Section 2 leads to initial

(relative) errors that are reasonably small for "interior" nodes, but comparatively

larger near the "boundary". Specifically, if we regard the three Kronrod nodes

nearest to each of the endpoints of [0,1], and the two Gauss nodes between them, as

belonging to the "boundary", and all others to the "interior", then the relative

errors of the initial approximations in our scheme range for « = 10, 20, 40

respectively from 2.0(-3) to 9.5(-2), 2.4(-5) to 5.5(-2), and 4.3(-6) to 2.9(-2) in the

interior, and from 4.3(-3) to 5.4(-l), 1.3(-3) to 5.6(-l), and 3.3(-4) to 5.7(-l),

respectively, at the boundary. The relatively large number of 6 iterations reported in

Table 5.2 appears to be due to the large maximum error of the initial approxima-

tions in the boundary zones.

Attempts to improve the initial approximations for a„ by using, for example,

o„ = \a¡¡n),** where a„(n) are the Christoffel numbers for da, and by obtaining the

remaining initial approximations a* from a reduced system (2.2) of n + 1 linear

equations, do not speed up Newton's iteration (in fact, require 7 iterations, instead

of 6, when « = 5 and n = 10), precisely because of the initial approximations in the

boundary zones remaining at the same low accuracy level.

Example 5.3. da(t) = taln(l/t)dt on [0,1], a = ± \.

Modified moments with respect to the shifted Legendre polynomials are again

available (Gautschi [7]) and suggest the same scaling as in Example 5.2. Results for

**This approximation was proposed to us by the referee.

Page 11: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

gauss-kronrod quadrature formulae 649

Table 5.5

Newton 's method for Gauss-Kronrod formulae with da(t) = tx/1 ln(l//) dt on [0,1].

#iter. ||/. H. ||.fecondj cond2

(scaled) (scaled)

5 6 1.03(4) 7.542 3.8(3) 7.8(4)10 6 1.35(5) 14.64 7.3(4) 1.0(6)20 6 2.06(6) 29.60 5.2(5) 1.4(7)

40 6 3.40(7) 61.04 1.4(7) 3.0(8)

Table 5.6

Newton ys method for Gauss-Kronrod formulae with da(t) = /~1/2 ln(l/f) dt on [0,1].

#iter. II/IL \\JÔ\\Fcond! cond2

(scaled) (scaled)

4 7 9.78(6) 272.21 1.3(2) 9.0(5)

8 9 3.66(8) 889.01 2.0(3) 1.3(7)16 9 1.52(10) 2961.5 6.3(3) 2.0(8)32 — 6.56(11) 9907.2 8.8(3) 3.1(9)

a = { are summarized in Table 5.5. (The computation was done in double precision,

the number of iterations referring again to 12 decimal place accuracy.) The perfor-

mance of Newton's method is similar as in Example 5.2, except for the (scaled)

matrices now being more ill-conditioned. The ill-conditioning of the Jacobian

matrices is still worse in the case a = - \, as can be seen from Table 5.6. For this

value of a the independent variation of the starting approximations ff, t*+ x proved

to be rather essential, since convergence was never achieved for the first choices of

these starting values. This is in contrast to the case a = \, where the first choice of

t* and t*+x always worked.

When a = — \, Newton's method could not be made to converge for « = 32,

using the implementation described in Section 2. The difficulty, we believe, is caused

by the smallest Kronrod node being almost equal to zero. Computing (in double

precision) ir3*3 by the procedure of Section 4, and applying Newton's method to tr3*3,

we find t3*3 = 3.05867... XlO"9. Using all zeros of it3*3 computed in this way as

initial approximations to Newton's method, and lowering the accuracy requirement

to 20 decimal places, indeed restores convergence and yields the data for « = 32 in

Table 5.6 after 1 iteration.

Noncon vergence (in the case a= - \) was also observed for odd values of «, this

time because of the presence of negative Kronrod nodes. When « = 1, for example,

one computes directly ir2*(t) = t2 - (198/343)/ - (3671/117649), which has the

zeros tx* = .627023... and t2* = - .0497636.... We verified that for all odd

« < 32 the polynomial w*+ x has exactly one negative zero, while all other zeros are

between 0 and 1.

Numerical results for a = \-, n = 5(5)25, computed in double precision, are given

in Table S.2 and for a = - \, n = 4(4)24, in Table S.3 of the supplements section

at the end of this issue.

Page 12: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

650 FRANCA CALIÓ, WALTER GAUTSCHI, AND ELENA MARCHETTI

Dipartimento di Matemática

Politécnico Milano

Piazza Leonardo da Vinci, 32

20133 Milano, Italy

Department of Computer Sciences

Purdue University

West Lafayette, Indiana 47907

Dipartimento di Matemática

Politécnico Milano

Piazza Leonardo da Vinci, 32

20133 Milano, Italy

1. P.Baratella, "Un' estensione ottimale della formula di quadratura di Radau," Rend. Sem. Mat.

Univ. Politec. Torino, v. 37,1979, pp. 147-158.2. F. Calió, E. Marchetti & G. Pizzi, "Valutazione numérica di alcuni integrali con singolarità di

tipo logarítmico," Rend. Sem. Fae. Sei. Univ. Cagliari, v. 54, 1984, pp. 31-40.

3. C. Dagnino & C. Fiorentino, "Computation of nodes and weights of extended Gaussian rules,"

Computing, v. 32,1984, pp. 271-278.

4. J. J. Dongarra et al., UNPACK Users' Guide, SIAM, Philadelphia, Pa., 1979.

5. S. Elhay & J. Kautsky, "A method for computing quadratures of the Kronrod Patterson type,"

Proc. 7th Australian Computer Science Conf., Austral. Comput. Sei. Comm., v. 6, no. 1, February 1984,

pp. 15.1-15.9. Department of Computer Science, University of Adelaide, Adelaide, South Australia.

6. S. Elhay & J. Kautsky, IQPACK: Fortran Subroutines for the Weights of Interpolator Quadra-

tures, School of Mathematical Sciences, The Flinders University of South Australia, April 1985.

7. W. Gautschi, "On the preceding paper 'A Legendre polynomial integral' by James L. Blue," Math.

Comp., v. 33,1979, pp. 742-743.

8. W. Gautschi, "On generating orthogonal polynomials," SIAM J. Sei. Statist. Comput., v. 3, 1982.

pp. 289-317.9. W. Gautschi, "On the sensitivity of orthogonal polynomials to perturbations in the moments,"

Numer. Math., v. 48,1986, pp. 369-382.10. W. Gautschi, "Questions of numerical condition related to polynomials," in MAA Studies in

Numerical Analysis (G. H. Golub, ed.), Math. Assoc. America, Washington, D. C, 1984, pp. 140-177.

11. G. H. Golub & J.Kautsky, "Calculation of Gauss quadratures with multiple free and fixed

knots." Numer. Math., v. 41,1983, pp. 147-163.

12. G. H. Golub & J. H. Welsch, "Calculation of Gauss quadrature rules," Math. Comp., v. 23,1969,

pp. 221-230.13. D. K. Kahaner, J. Waldvogel & L. W. Fullerton, "Addition of points to Gauss-Laguerre

quadrature formulas," SIAM J. Sei. Statist. Comput., v. 5,1984, pp. 42-55.

14. J. Kautsky & S. Elhay, "Calculation of the weights of interpolatory quadratures," Numer. Math.,

v. 40,1982, pp. 407-422.

15. J. Kautsky & S. Elhay, "Gauss quadratures and Jacobi matrices for weight functions not of one

sign," Math. Comp., v. 43, 1984, pp. 543-550.

16. A. S. KRONROD, Nodes and Weights for Quadrature Formulae. Sixteen-place Tables, Izdat.

"Nauka", Moscow, 1964. (In Russian.) [English Translation: Consultants Bureau, New York, 1965.]

17. G. Monegato, "A note on extended Gaussian quadrature rules," Math. Comp., v. 30, 1976, pp.

812-817.

18. G. Monegato, "Positivity of the weights of extended Gauss-Legendre quadrature rules," Math.

Comp., v. 32, 1978, pp. 243-245.

19. G. Monegato. "Some remarks on the construction of extended Gaussian quadrature rules," Math.

Comp., v. 32,1978, pp. 247-252.

20. G. Monegato, "An overview of results and questions related to Kronrod schemes," in Numerische

Integration (G Hämmerlin, ed.), Internat. Ser. Numer. Math., vol. 45, Birkhäuser, Basel, 1979, pp.

231-240.

21. G. Monegato, "Stieltjes polynomials and related quadrature rules," SIAM Rev., v. 24, 1982, pp.

137-158.

22. T. N. L. Patterson, "The optimum addition of points to quadrature formulae," Math. Comp., v.

22, 1968, pp. 847-856. Loose microfiche suppl. cl-cll. [Errata: ibid., v. 23,1969, p. 892.]23. R. Piessens & M. Branders, "A note on the optimal addition of abscissas to quadrature formulas

of Gauss and Lobatto type," Math. Comp., v. 28, 1974, pp. 135-139. Suppl., ibid., pp. 344-347.

Page 13: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

mathematics of computationvolume 47, number 176october 1986, pages s57-s63

Supplement to

On Computing Gauss-Kronrod Quadrature Formulae

By Franca Calió, Walter Gautschi, and Elena Marchetti

Table S.l. Nodes and weights for (2n+l)-point Gauss-Kronrod quadrature

with respect to the weight function ln(l/t) on [0,1], n=5(5)25.

nodes

9.6 70317011313184162899526d-Oi8.9477136103100828363888620-017.946457674432365150600413 0-01

6.77314174582 8203807018027O-015.4742415722 72107646167938a-014.11702520284 90204317493190-012.8339097129822170420424410-011.739772133208976287011397 0-018.7 93120246479704019290778 0-022.9134472151972 05330372676 0-023.0 55453450374047608522276O-03

weights

1.8176 74998509004844923280 0-039.8205147104594648494Ü1820O-032.52 81432798508611737501200-02

4.8359486624192653084139940-028.Ill987879942632270559929O-021.192 8870631016403584921680-01

1.515254284520667107440880a-011.71520396732757562B055522O-011.785532970704892670281547O-011.5186243751630582212233880-016.0850745987120544229083 960-02

nooes

9.9040145663466103992281040-01 19.688479887186335393915045O-01 8

9.375650909249013043766844O-01 28.9816109121900353816695320-01 4

8.5030626998372514460951420-01 8

7.941904160119662173585093O-01 17.3198728660817158247497020-01 26.657752055164245972223813O-01 2

5.9611711699600020923177600-01 35.2379231797184320116116380-01 44.510940209787955261216721O-01 53.8021253960933233397234130-01 63.1193271934195560332374330-01 72.470524162871598242225454O-01 8

1.875727i30810663812513927a-01 91.3531182463925O7748702321O-01 9

9.068916719091722376416700O-02 95.397126622250062950420123O-02 9

2.640564624946856773849341O-02 89.042630962199650636946628O-03 5

1.283562335805386114259342O-03 2

weights

533 569184207643623003683O-04487068240706659310736070O-0428455683274 5782346462B22O-036 71211724894484356 713219 0-03

4672 73 36990677 9154553426O-03

373507422471821554874215a-020098742972641137012981600-0276432 52 82485899850930559O-026865953 5228178105859621BO-027183 82784133641587909658 0-0273461316917431556072 5749O-027334 9195447 9115487650944O-02784758916696079582 379503O-02748258852663594360502 868O-0238182 691070552 392 23561890-026B5544 515771952 988656812O-02

79444646162853 683897 9348O-02456636029965041400178269O-02

19090031692 6699549095493O-02774125174 710110186 821323 0-02

52020199163 792 4753 93 4347O-02

S57

Page 14: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

S58 SUPPLEMENT

i i iD O O■1N>-t p* OiXI XI oo -m 01T CM OCM -| 2)-I X cmn p- h■O cN 1p* h xoi co p-

Xl X =T

-i xi enX P» CM33 x rMn co XO h encm xi hh p» :NOi t Oih m 31—« rs) -0

H X"

a o-m X

ü-i sT

« Xp- -M

4 D0 »Oi r~x> -oif X)01 -o■n or- xip- TiH -n

-M XI

N TiX) sT_n "M

n m n -o no o o o oI I I I Ia O D O oh xi p- f» nO O 1 i .10 «A «f 3> í)01 H O XI -0T *M ("* ̂ en

n -4 xi T ox xi t h p*rn <# Oí n r,fl O D Is -nh »honxi ai -i r\i p-Ol T ,-0 X> 31O X 01 XI Xx x p- xi jixi x» en r\| rMp- x x *t rtrM —i r* irt —iai o t x> enX> X> p* x TT T h ■£ p-X O ."O H "OS í"t /l if>o x n ín m

□ ax XIxi p-p» t■n o

n oxi r-■n on cm-i r-X) 01

h tai xi01 01

X PJX> CMft fM« -HX) r-O OO f]Xi O

I I IDDO<* «t 3-h Oí r^

<í X H

T m n£i h r»xi ai r*-rM O Hun p» p»cm r- xiT OÍ 01h p» .-oX -i r^H O X)T T ^í N T)01 01 TíJl CN XIp» xi hn X> HJi p» x

N CN CNO O OI I Io a dcm r^ xiO^om oí or*> o xim p* xio h oCN -v] p»n Oí <*T p- X)p- -t X»m p~ -tX T XI

cn p- ^x p- r-O <T Hx h o<j p~ ncm n -Oxi x p-p- en 'Xioí -o r,p- "N ooíj1 r^ oí

rM i) r\)O O OI I Ia o a=j ai -iT m -tíf 01 XiXI XI x>Xt ^ XI-i XI -I

H P* O»£> .O Hxi in ¿>V ■* r)

O O OrM r* xifM Ol p*r» "O X>.n -o m-< O r-oí XI oXI X) oN N HH J) J)osrji

> rM i N

0i - Ol

j0 CO O»r- £> ■■oOiN H"O XI X)-0 XI ,-M

xi xi -n

^ -i ofN XI X»

r^ 01H íN 31fM íT rMr- -n OfM 01 O*o rg r-,o « mrM "o o»■o ̂ i mH 01 oHOTo :o xi

N rM rM CMO O O Olilio o a d-i fM rM rnXI <J3 O 01r- tM r> r»>o r- xi o0 m 01 -HX» rM ift OXI T 01 oO» ro P* r^£> ■£> rN 0101 -i T tMCO <T 01 cM0 XI -i 01£> -I -t CM01 m O r-t «r x» cmr*l T O -ir- m >-o cor-- -i rn oH O Xt COiû rM CO kD0 <!■ H CO01 rn XI 00n -i T t01 rM "* sO

N rM

O OI IO D"0 "M

r* nr- mr* «tr- TN O.■M Xi<J X)-I XI

r- COXI iT\XI -j

n xiO r-t- Or- XiOi p*X) rM-n X)-n xO XIO f-XI H

0 o-I ^J-

01 Hoí -nX» H** .-oO X)X) X)O D-n xiT -nr-- oio TH "OD XiX" -oXi -Mr- XI-M H

X) r~-X> XIX) XIr- O

N ¡Il NNo o o oliliÖ D O D-t CO O rMNNO01Hin J)Hxi r^ -h mn -i -o x»x) co m x>oí oo -n mH IN M *Oí B1HCM <î J)ri 01 O 01r* -M r- aixi o r*- oNnMíi0 Oi Oi cmT pv r- co01 o N r-rs 01 Oi rjoT co xi oim í Jioi oi co r»o xi :m nXI 01 -4 ÍM-I O 0% i£i

-M IN -N CM "H

0 O O O OI I I I ID O D D atirno írn -h m co -on o co r- xirM cm X) r- xi-0 XI CO CM <*

01 r- P- H X»^ Xi rM XI X»lO X) X) y) XI

T O O O O<t X) r- oi xixí —i oi «a- xiIM ro M) (MH H ro r- C0P» CM O -O X)C0 XI Ol X) cmO rM O» m COT cm r- p» p-co rn t co r-rM P* p» T OT cm O X) ncm 01 O sf X)p* r- T oi p-p* o oi pv r>¡n co o o "i

| ,n X» -i -H CM ,n f Xi P» 3D -H - HHHrMfMiNn.ilrnmií^í 'JXlXlXiXIXl^rT «j.-n ^cmco

-I X)

XI <T-i (MD Xix> x>M CMr- p*H 01

o o o oI I I Io a o o

O«* .111CO P* -0 HX> P- CO X)P» <J P* ■£>X> 00 X» CMX) r0 01 O-t p* cm mH f> -* X)T Ol -) p~XI t X)

■vT r-'XI cor- -nr- TsT Hr- co-i OX) rno -osf XIp- -iJi Ol

P- rn p* -,01 X> rM 01-i CM P* Op- xi O 01O P» H rM0 O 01 -iH h m Olp* «T CO <3-01 P- XI p-01 CM rn 010 CM CM \001 X) o mCM H X) HDP- ifl >t

o oI IO D

XI XiO Xh rnm TO -nO "M^1 CM

n oiXI XI

x r-0 -Im <*x TXI XI

m xrM oxi X01 X)X 01h r-CM X■N OCM O

O O O Olilio o o oCM XI P" XIXï P- CM OlX> X) -I 00X -I P- X>m cm p- m0 CM p- «j*X) fM P- P-m p- p* p*Í INiÛHCO P* rn co01 X> X XIrn cm X XH H XI O0 01 Xi p»01 X CM XCM P* CM 010 CM 01 H-i 01 -O P»CM X * XI01 O Í 01*f X CM OCM CM en P-r- h en mp» XI CM 01

i ia ooi «p- xi

XI oOi H

I Ia dx -iXi X«* 0iXi oO -J

o o o oliliO Dx> ox> n

H Ol"M T

XI -M■* .-o

O DO Oírn xi=T X>

T X)xi incm rji

O T-i r*O X[*) x>-i 01 cPJ CM0 Oloí n01 XXI o■M T

rM X

CM 01X> rM

X TX) 01CM OX 01rM !Tn -t-3- mp« mXI -I

n <r:m nx> r-.-M oX) oOí x>

p- 01X OíOi o

~n -i-o OlX> TX Txi Ooi ^rx XI io* o »en p- ;CM X '

H rnx> r-XI x>-t 01^J Xn r»^r x»

o o o olilid a o oO cm 0i r*.í) p- s mxi m oí xCM -O X| oíJl X» -M P»x x r*- nO T XI XIX) X XI Trn H y O<f rn P- 01X> X X xXi X» X XIn p- n xx o n p-T oi oi mP- O P- XT O XI 01xi t x> rg

X P- O P*cm x» x> r-X> 01 P» cmt <& & np- m o\ x>

o oi iO Dxi n-o o0 xXI .01

n h01 Ol

X) oisT CM

x m«r x■3* XI

x TOl XlX -Ixi qX) OlO On x>X Ol

X Xixi p*p- CMCM Ol

o oI ID O0 HX Xi<? 0101 CMXI HP- X>x .ncm mCM On xitf CMT om cmh oCM X>

-t ."M01 P-X Xxi r*>x mh xH 0!01 X»XI CM

0 oI Ia dH Ol--M nx xCM CMrM Xn -)-t p-X XI01 X

4 X>r- oi-n rMx> m-I XI

H P-m -m« XIO xiCM Xr- -nO T

X> P-0i jrj

O OI t

D DN OX <rrP- 01■* XX XX» Xï=T XI

H On xiXI -(T enx mp- Xi-t p-<* xx TXI xX rMXI XI

-n oP* Xn tO Xl

CM CM CM CM

0 O O OliliD o a om xi cm p»-I XI P- oír» r» en p-HÍHÍx oi m xf ^f 1"Cnrn en en in3} <T r* !T\X» X> O P»CM O -O CMfM 01 en rnxi m h xiCM P- CM X>01 X T HX> Xi X> CM_) _t Xi P»X> X T enm m o oX« —4 H XIm xi oí xiXl H 01 X)rM X T rnp- H O xiCM CN «r X

CM CM en en To o o o oI I I I Id a D o oP* iß 01 iû Oio oi xi m oix m o p- o»X> XI Xi CM 01cm x) x m xio o h en xien cm cm x nXI P» 0i CM ^x> h T <* mr*- m m xi p»Xi rM o Xl CM-4 o xi oí oo xi -n -4 n0i XI x CM 01«J* 01 XI 01 xi-* ^ run oh r} x» xi xin cm o oi x»X X) O p- TH X) T CM Tx oi n m "ïo o Xi x TX> CM X X 01XI XI 1" ¡ft x>

Oi0i3ioioioiai0ixa)xr-»p»p~x».ox>xixixi ^f q1 rom rncMcMCM h- i oí p* xi en cm - p* cm m

i <? sri o o

i ii o a. x rnI X *fi q> xiI CM O' T «■i cm ni xi x1 X» 0i

l <T O4 Oi P»I XI p»» H rn) -4 n• ft xiI P- Xl' X CM

I X» ol o oi O xi4 01 P-) CM CM. P» CM

I X) P*I 3) O

■n no o

il*f 10n th xO XXI OX CM'XI 01

X CMr^ p-x xx r-oi .-nO CM

Xl HX HOl O

n <rX o01 -IO -r>xi rîO 0i

rn enO OI Ia a

X -*X oj0 P-01 X■d- oiX) oXI Xi<T Oim xien X-M X

If CM0 HP^ Hxi <rx ox r»01 Tx -nx oo Oioi r-

i iD Om xT r-X -Mm p*O CMT "TT XH -in -*X Ol

0 X01 XiXI CMT r-X T■Xl 'Xlh np- p~

cm cmO OI ID Doi r~n <rp- orM Xi

O XiT 01O cmX) OXI oCM Ol

n tX Tp- n^T Ox r*X) j-lCM <fX TCM XP* T0i oCM XCM xi

CM CM0 OI I

D OCM -4p* m01 -4X Xxi en-3" -lCM -4CM 01XI H

X XICM OiO HXI X»x *r<3- Xt Oim nP- HX <*CM XIoi .n

4 CM cm CM> o o olili3 D o ai CM XI 01

) Oi P- 011 X" Xi <tH -4 n XIH Oi 01 p-5 -4 XI X? 01 O H* m xi oi■> x x) m■4 O -4 P-3H.NO- p» h r»T X O e»1 H r- cmg X XI o•J CM H XI3 O O P-sl H <T Oi? 01 O 013 XI ^ XIT rn P* rM3 oi n xH P- XI X)0 CM P* CM

a aH P»p* *rH 01P- CMCM P*Oi X>X X0 HH XI<i r-X) rg01 CMX» 01X Xx mXI p-

m xrM X01 p»H XIxi m01 xp* .nP* N

O DO r-^1 T-m XlrM ^Jp- OiO oO CNXI Ol-M 'M0i rMH 01n xiX "Mr- xn rMT Xoo•N nP- XXI xO Xen P-XI xX O

I Ia dOl 01

x -nr- Oxi xX TP- XOi oCN XIOTH XX XX 01O -IX 01n xiX XO XIX =JO OlT Xo Xcm -n.-n m^ x

■M "M

OOI Ia dx <rx P-01 xX <Txi oO XIrM 0iO -IN Op- noi xt nx enx -H

."M P-H XXi O01 Xr- tCM XIr* p»

^J CM0 o

¿AP~ 'XIP- XX XIn xixi x-n x01 oX X

T oH OO 01XI -H

O HX XP- XT XP~ OiX 01H T

CM CM CM CM0 O O OliliD a O Den rn Xl CMH P- Xl CMXl 01 X P-m h r- -4n >o xi xioi o xi oip- T h o-n t h p»01 xi t XI0i O H X)01 O 33 010 H P* CMX O CM Xx en m oíX X 01 CM01 p«. T oCM X H OX 01 XI enrn Í H P-o T -O Xh xi o mo X> T Xlx o en 01xi r- .n cm

m-4Xi-tHcn<ïXf0i. i-irNeNcnrn*í,^,x!xixxxxx>xx>xiq,en-

0 o o1 1 'D t) O01 T enr- CN HT X> HX 01 tn \0 p*T CM <*X» Oi P-X X 0iO m CNm oi ho h mrN m p*o H xh rg «i(N rM XcN -i mm x p*o X» XCN <* XX çf CN0 x» oxi .n xiXI XI o01 X) p*

H X01 cmn xiX Oixi r*N P*P* X-M 01

0 X01 P-01 oO 01T Xin x

P- 4*

x xX -nH CMx XO HXI oH XXi ~g

0 O O Olilia a a oT T 01 xn <í cm tX CN H OX rn H <foi xi p* mCN X> XI 01m r* cm oT rn 01 XX O X HOíiflOX T 01 H01 P» x xx m h oo h x mCN P- X 01H H ex p.o Oi m xixi o "n oO H O XiCN 33 X CNx rn t r-X X XI 0ioi p* cN moi x en 01

o o

iiP- en

X Xp- nx p-p- nx noi <rp* oXI -ix oiP- XIx P*p- om oin hx oix Tx -Hp- m

o o o OI I I IO D D Om -4 cN n(N m m HT oi x cmoi x* h moi xi xi tX O CM X3m xi t xH CM X» XCM 01 O Xtj* rn n xX 01 P» 0ixi m h x»T 01 X H,-N p- H XI-4 O Ol CN

0 -0 P- P-oi h m aih m p- hÎKJ "Í HrN h n coX H P* CN01 T T 01

. H4f ilifl) X -H X H

o o

iiXl Xr- -nH P*-n cNX 01P- XX -Ncm nX CMXI P*X CN0 P-p- r-«í OX X01 -

0 o oolilia o d oxi pj m xCM X t p-P- cm O Om o m oX P» «J" oH n 0i XX H m CNen x T rnT Xi CMP-oi m o r-CM Xl fM H1" «f Ol1P- CM H Tf01 01 X P-m o o 01

X O P- H X <Tcm m rn mH X» 3> 01xi en cm XH X P-P*o X P- Xroi -4 x mx rM p- en

CMm ó

CM p- ■XI T <X P* <

a ax xmmH 01 .X oCM Xix r-01 XIXI xH Th aip- xO CN0 nTOp- oen xCM OX rN■M -i

XI -i01 XIn oin xiai m

oo

iiXI XIX TH Ox rnOl tCM CM

n xn xio sx aih nno.n xn hp- ai0 xT ain x01 xH CNO X

CN CN CM cmO O O Olili0 o o -aen 01 Oi 01XI XI P- 01fM P- Xi rMT rM rn P»Hdl.lHT Ñ* CM rnXI T o coCM X H Xxi en X TT rM h nrM X X» 01X» O O Hh xi n xim -h m oirM x xi nH ■* O 01T XI H XIXi O X x<01 P» 01 P-O X =0 XX x» ut XIoi x m xXI T 01 xxi en xi fM

iin cmx .nn cm01 P»t p-x .nt ain ox P-n .nO xi

XI xp- -i

X) P*P* XH P»O 01H XH 01n tx t■n xi

013i010l0lXCXXP-P»P*XX)XlXl T^1 ncncMPM H-H H01X»*r CNH TXI

Page 15: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

SUPPLEMENT S59

3 O O) O P-J 01 P»i O CMi X) "*1 p- XJ X 014 O X»4 X O1 X P-' O en) n 31) 01 TI X 01• X XI T enI O n">> n T■ X <T4 01 P*I O 01> CM 01) X o

rM CN CN CN

0 O O OI I I Ia a o am cm ai o01 rn -M Xo m cm p*-t ~t r¡ -t0 p* x HX) X "M xiCM T H XX H XI p-_t CM T X01 en XI XIX X X T-M o X o01 XI T P*T H CM XIO ^ CM enX P~ X Tcm O rn Of> 01 X CNx -n xi p-n J> o o*xi x x xirM -n p* cm31 X> 31 OV en Ji P-

Ia O D DH T 01 X0 XI xi oXI P» 01 Hp~ n oí O01 en 0i xi01 CN en TX «t -t XP* CM CM TP- T O cm<T X T OCM H -I 01CM XI n H01 CM X P*m X t x01 XI N P-H O T CM•f cm n OiH X 01 TT CN X TH H X Xi01 O -i om xi oj Nen xi oi "M.n oi xi -4

I i

Hp-Hn^XP-P^XlcnH

D Dn cm t TXi Xl 01 or- P* H '■**ai o xi mt n o cmn o x ^01 rM X Hx en o enX X» -4 OrN en n t>n rn O O0 T X Xt n h oX ^1 X CMX XI p. XXI T T CNxi <t o r*»x n oi xX H P- 0101 H XI XT o X P-■^ ft T ri01 en "M X>m X O O H -

-t P* CM í P- ■

n cmo o

iiH Txi X»P* CNX OiX Hp» XX XX -Nm xiX XIn cmX XiX) XXI T

XI -I

H Xix nT Xn XX N

rN :mo oI I0 on nXl «0i 01P- -NP- o

Oi 01X -MXI CM

01 T"M O

X r¡T p.01 01

*M "M

o oI I

D OO XIT XX P-T XXI -H

n hoi r>rji XX XT Hn .n-M P>

rM 0i0 Hn TX XXI 4f01 Xn TXI X

O P-XI TX HX N

-M CM N rN0 o o o1 ' I 'a o o dX X X CNcm x r- xi-4 XI X H-I H X H

01 Oi Xl X)T Í P* -op. 01 P- CMp- cm rM XIOl X T Xn x f oX P* H P-p- .n cm cNn t~~ m -OX X> X op» X rN 0ien x t aiO O tf oCM 3D X HX O 01 Tn h xi oiX Í O 01x o .n xXI 31 CN P-X 01 CN CN

CM CM CM CM0 o o o1 I I Io d o a0 CM 01 H01 X H Hn tfi x p*ai t p- TH Is- Ol T

XI XI X CNT H X X»rN XI x rNCM P- CM OX T O HONJirflOi X Xl P"X T X o01 XI p- -NX oi m enX 01 H Tp- n rN nXI CN T 01P- x o Xen H o oX O 01 Tcm x n xiX X XI XO X H T

h cm cm m en en t t t en .n cn .

cm en no o oI I Io □ o

XI 01 xoi n nCM H P-CM XI TH X 01n p- nT X p^X XI Oixi en "»X X CNXI x o0 Oi enT X HXl O XXI O CMX X 01T XI x01 r, oiH P- 3301 01 -4XI o oXI XI TXI P* -4XXX

X cm

0 o oI 1 ID O 001 X oX rM nxi n enT O 01n t TO T XX» p- oX 01 01■n X o-i r) -in oi p-T X rM

oi ó .nh nom x xXI p- X0 01 CMX T P*X XI o-^ f) -t

X 01 H01 rN OX O H

0 o o o1 J. ' 'O D D DX X 01 XIXI X X CMxi n h oiX P" XI 010 P- H X01 P- X XT O H t.-n o p- o-4 01 X X)xi T T X"en CN T O-4 cn X enn n xxT O T x».n n oi oiH X H Xlai xi m p*rN O P- Op» p. en xirN X CM OT O H 01-4 01 P- O

O X X HO P- "T CN

0 oI Ia dH XIX Tx r*X T01 Xcn nn xCM XI

H H31 01CM XIP- Xr- Xh .nT XIX o01 TN p~N p*CN P-XI XX n01 P*O -t

"M CM

o o

iiXI p»o xX HO xirM rMT "Mn nX XIH O0 Ol

T Xn aiXI 01P- -Mx> xiXI 01

n Oi01 01H XO H

p- «ÍT XOi o

ooI I

' o o

r- XXI OX XIH enxi m0 X-n h01 XH O0 TXI 01XI Ol

CN X01 XT o-N "M

X CMo oOi P-

X rM

H T.71 XX CNH>J">no01 XIx nrM rNr- xin 'XiXI CMn xXI X)

XI Tx xoi n-M XI

O CMT O

0 o

ii•M O

<r p»■n "M01 T31 OX P»01 XH OO ^1X »p- XX NX 01O Hp- XP- rM

O <*X HX XXi Oo xX XX rMXI O

0 o

iiXl XCM -H

X P-T XIp- p.H OXI NXl P~

n t-i nX P»X HX o01 X-M p*

X Oin xX -MO nn oiP- XIp- p--N X

iiO "MO 01Xi Or- hX X

T Xp- p.X XXI xrN xiX HO xiH P-O NP- XCM Xru nn Xn xip- oH XH OH T

o o o oI I I IO D D 0n oi h p*X CN XI Xo p- h -np» H H XIp- X" en oen n t p-CM en fN oO xi x Ot x n xO P* XI rn.n T x oi-4 T 01 01oi T n Ten h Xi p»m *tf cn TX X X HH P~ H P-0 H m ox en xi oiX O T CMT <í O Tp» 01 CM T01 Ol CN P»X 01 en x

I.d a o o D D oO X P* 01 X H X)■» H X H <T O H» en T X X rN h> T H X CN X CN-» H CM m* P- XI X

01 X 01 T H01 T o ifï oX O X Ol oXI X X X 01n XI CM CM XIX en <* <r nP- 31 m H ox n H H 31O "T X P- CMO en X 31 p-O 01 X X CM0 P" -n O P-X T cn Xi Xp» X O XI cm01 H T X CNP» X X H P*O H O XI -IP- X -I P« fNO p» P- XI X

X rN

~v| 01

n p-."M T-h -n01 On T

r- 010i p»x xO xi.n r-N Xn nxi OT Xp- nO xi

OifjiXp-XiTcnrNHTH aïOi oioi xcop-x xxiTcn n cm cm h -4X nn n

X XI XIO OOI I I

0 D O> X p»r» x> rnn enxiX H TT XI XI■*}-**?m xi xP- H OCN Xl M"-4 X XiCM n CNXI HOP- N Xh np-P* P* CNT P-Oxi OOp- xi mXI O TX T Xoi n o01 P» 01cn en 01P- X -4

O oI Io an xirn nT oIf 01xi x•N XH OO X>P- -M0 xiXl OXI X01 XX XX P»X XH TP- oX i/iO P--t mrM <To Xxi r^-

0 D."M 01P- o

ai otf 71CM XI

H Xi ■H 01x orM .nX r,X Xx n-H XI

31 XO xiX nn -in OX om nX o

n en en eno o o oI I I I

O D D DOi CM O P-O Xl p» rMxi xi x .nm oi en x»x m T ent m x x0 T H 31T r- h xiH O H XP- O -N CMXI T X CMX H O CNp- x T p»01 T T 01PHPÍCN X P- CNP- O en 01T CN 31 p.31 CM P- Ori co r, £>X H <T XX X -M 31O O 01 TO f 33 rf

m n n nO O O oI I I Io d o aoi x xi no oi p- .nn oi n "in oi p- xi31 X X HCN X 01 XO rn 01 X31 H -t Xri -i xi -iCM P" Xi —("Í H fN Txi xi n p*x m p- TX X HPp* m m oT XI O -In XI h CN0 rn O TrN 01 Xl p»x n t oT X» Xi rn>ÛP Uil01 X T 01O X P- X

n no oI Io oXl xxi n-i ait ai

T "NX r-X OP- 01"M N

N rnP- XX) X

T XIXI -I71 Noi mO P»O XT CMp- ai

n CM CN CM0 o o oI I I Id d o a01 en X XlT O 01 01CM X CN Trt p- o T■f & r~ OT o o oP- X T X-t ~1 £> ̂ on oi x cn<* Xl X ■*H X T X»ri h X 01en O 31 CNXI XI O XirM CM P* H0 Xl X P-01 CN H XIXI XI 01 Xrn X 01 p*m o x> XX O O XlT P- O P*CM XI O XICM o (m m

^1 CM

o oI I

D Ox -na r%]n P*x n-M 01

0 o-i r-X XP- XIX X)p- T

x of>- XX TX Oi01 on nX "0X P*X XIm ocm aiXI x

CM CM CM CMo o o oI I I Id d o aÍNOcjip- o x nT o n p*0 O X XICM O X TH fN O Om p- x cn31 X X <tXl p» X Xrn Xi CM p-01 X CM XlX CM T Ox m h p-H O 01 fNT O O P-X O O X01 O XI 31XI O H P-rN T P- xiX H 01 TT rM xi nCM P- X 01*Î'lHX O CN T

N CM fM CMO O o oI I I Id a D Dcm xi h nri -i -i o\rn X 'XI vO0 <f <rt -tXl O X -1X X X <tp- T p» cmX 01 Xi 0101 Xi CM -iX O T Xi31 01 31 rnOi - 4 Xp. o «r trn T en oO CN T XtXi 31 X CMen *f Xl XH P- X HH CM CM r-'P- O cm Tt~~ ¿} D Ch m rn p*O 01 X XiX p- 01 -t

N CMo oI I

D DP* XO Xn oiXl OX P»X 01

X on on XX X01 X)P- CM

N X01 -Mn XOl fMen P-■M X

X p.en nfN ain t

n nn TT XIx> nO XiP- 01r- xP- CM

Xl XX XI•n -iN P-p- TX o0 XIn TX XH O01 Tm xiXI XX r-

fM CM cm "Mo o o oI I I IO D D O•n p. o enri t P- enx p» p- mX X X HX X 01 01XI T H Ot co en xX X CN P-H H CN P-x m o hP* —4 Xi XIO 33 cn OX CM X> XiXI Xi —( XIX 01 X X0 O en XX p» CM Xp» en h oip« n xi 3iX Ol O TX X X TX 01 01 Xi01 O 01 TX o O -4

N CM N CM0 O O OI I I Iou oo01 *ï P- P~CM H X XCM P» CM H^r oi p* oiO XI H XXt X X HX O Xl íX O P* CMXI O H P-m ai n p*O <f O 01^ P- X xiN X XI Hr*- n 33 &•O rt x Xm h p- xiOi-i P 1H H P- P»X P* Xl p.oí m o en<r h oi <ro rn oi cnx x fM cnH H H O

rM -NO o1 I0 DxoX oX 31X ^1^ r-H rM

O CMX XIp- <JP- H« Ol

CM X

X 31XOx r>

CM CN cm eno o o oI I I IO O O ÜX X X TfX X XI oX P- X XX P- Xi xX X <ï <frXI «J -4 ^X 0i CM XIH P<- O OX O 01 OCM O H Ol

H H O Hm o X orM -Hi H ^O lÛ H xtX Ol X XX T P- CMtMJlCO33 H CM XH •» P- fXI -h X CMCN H rM CMT rn O OH O O OP* ru Xi x

-irN.nen<fxi XP-oiHH' HCMcNfM cNfNCNm,nrn m rn en ^ í*» ^"í ^ ^ .-n m ^ m ni in h ^

0 o oI I Id o aP* CMfM

N np-> 01 cmH P- 01X 01 HX P- XH tOX fM P»01 xio01 P* XIX CM Op- 01 01X Xi tp. *r iCN X X)X P- XX 01 -401 xoX »P-(NtO-t 01 p-m -ta33 TCOCJl 01 X

o o

ii0 enn xioi <r01 CM^r x-4 <rfXI Ol

P- XX HOl XI4* nXI 01pj *r01 X01 OlXl p.rM rMP- XfM X

m aicm r-X XI-4 CM

X P~

-t ~^ -t ~i0 o o o1 I I ID O O OrM 01 X p-CM O H CN<ji x X 01X x O rno m rn XIX fN ^ XP- -4 31 fN<* H H en-, CN CO Xr~ 33 -t &Xi -4 XI P-P» rn o xim cn x *rP- <rf X XlX r^ oí sj>oi r^ cn oH 31 «* XX P- X Xcn <r cm <r<t ai h x»CN X 01 31X 01 P- 01H X ^ XX <* en -t

0 o o oI I I Ia d o oxi oi -n en01 p- Oi 31X O H XlO ex Xl 01X O Xi Xio o t*- oH X H O^ O X -4O X XI <ïx m x -i31 X XI 01en <r -4 p*H X) CM PsP» ^ P- CMX XI XI CMXI X 01 XP- X O rMXI X X «ÍH XI H OX xi T 01CN P- CN CMxi xi en p*H CN CN OO X X ^

o oI Io or> nXI XX 01O 01X Olp- nh r,rf nX CN0i 71P- CMXI XIrM nP- X<T Xi

m cmT -Hn Or-OX Xx -nP- Xp» m-4 Ol

oo

iiX H"M -H

HP

0 Xp- «01 P-01 oH CMx ^rp- OlOl "MX oO -n-m <rxi ain h^T Xm xc^ rMN -^pj -tXI *X CMX •?

o o o oI I I ID O O O0 xi Oi hX X> xi xiX CN XI XIen -h X XH P- 01 P»-4 CN 01 hh <r cm <ïn <* o oiXI <ï X X01 O X 31n oi ^f oiX X Xl orM T sf xi*f X 01 oP» X Xl 01xi xi m xO XI X XP- rM X HM* en X OX CM X" X-4 m «* p*en «ï 31 XXI p. X 31h X xi rN

X) Xi

0 CMP* XIn nP- X'M X

N X01 Xi0i XX OQx nx nX 01<T 01

rN rMO OO P-

o d o aO h ejipx cm .n cnh x x sr0 CM <tf XIX CM XI fNn rt «r xi01 X H 010 X 01 01X en X p-X X XI on oi h hXI X P- oP- H O fajtfiooi01 x o oCM 01 CM fNH O X» P-H H CM XX 01 XI xm rn >n hX p- xi rMo p- n oO 01 01 01^ O P <t

iii

OOOOOO OOOOOOI I I I I I I IODODoaaa0iXiOH01p»0iX3,'3'XrNcncN33(NX fNr^ cmX r,

xi ain cm01 01XI XI

H HP- X71 XX OiX 01ai p»«f XP- HX oTf XX 4»P~ CMoi n■n O01 CMX 01H X

p- n71 enP~ XIP- OiXi -HXI XIX XlO -IX X-M Xlx aiOi oo oXI _4* OX X)0 Xlxi xS o0! CMxop- Oi01 oXI T)

Oi wX XH P»ai n.n p-N 01^r h0 en-n p-X o-i aiP- "T01 XI

n <rrX) p-<T XX P-X P71 -

o oI I□ o0 -I•« X)

n N01 X01 -H CNxi nn o01 N<T 71xi rMx o-t pP- HP- oXi <fCM CN

r> 33r* oo -ii co

x m■XI X

X cm3- CM

P- CMCM XiOP-

o o o oI I I IO D U O

0 rM P- hCM rN 0i Hoi cm r- <r01 CN CM XXI fN X 01cn P* en XX sf XI XiX Xt CM fM<T T CN P*X X 01 «3*O X X «TXI CM O 01CM XI X ïfX en X> 01O O XI XIm <r x p-O X P* x«T H CM X)

CM O H enp» m 01 rMX Xl XI Xl

H X CM HX sf CM H31 p- xi m

O OI Ia o« p.oi ^rX <ïrN aiXI XI

x <rXI p-O X■T H

O <Tx <rO XIT XP- 01

« oX 01x nX «m cmX 01X ;rfT -4H CM

CM N

o oI Io oX XIXion oCM X»X rMX 31CN Xm hH 31O r\P- XÍ0 M*Xi X

n p-XI rfX XlH P-O ^-Í H

CM fM

OOI I0 ox nH aiXI Xl01 oH XX X*mX rnn oiXI H

O X*r->J p»

XI "Mr-- XX 01O XX nxoCN OXI XIXI -t

rN rM

o oI I

D On -tP- rnp- nP- XI-M r>

liliD O D DXI -4 O XIP- O 31 OX Xl h HXI X P- OlX X» X enX rM P* H«t xi m mOi oi m rip* 3i m oCM X ^" XIOi p* cn mPCOCOPm xi X p»XI X X XX cm 31 xX» Xl h OH M* O P*01 X CM X31 X P 01xt m o hxi ai o o"ÍOiOT01 X H P-01 Ol P- CM

01 OiOi OiOi OlOiOlOi OiX XOOXP-r^Pr^x XX xxiinxiijq' ^■mrncnmcNCNCN -4H -t -i H0ip»x^tcncNHOi<y _t ín

Page 16: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

S60 SUPPLEMENT

o oI IO öp* p-fN. oX 71T XIXI OlXI P-O XI

01 N tf*J 31 H■C 01 Ñ*rji o xih n P-OJ O «T3 no

XI 01XI X

0 4*01 o«* Htf »X Hxi OX NO 01

4* 4*

o o

iiX XlP- H0 tf30 f*01 HX CMm oiCM CMX 33p- r-H *-H X)xi nCN X»XI 31Xi enr-xXI 31CN 01X Hp*r-XP-O CMx n

iin xiX OXI CM

O NX H0 oN HP- O

3 CMTi TN XiP- H01 TX fO XH 01h r>X p-XI fr- ai0i nO X

n n0 o

iin oip- *rp- nh xi3- tfn p-P* 01h m4f On cmxi ox mx xipk X

rN rN« nX otf O!X XIH =001 N■g- X

O CM

m n en mo o o oi i i i0 O D Of o o fH xi x HXl -cf X riP- CM X XH cm n X01 O X 01CM O X P-O CM X P»X XI p» xip* X Xl Xx tf m xXI tf -t Hr\ P» m XlX N H Hm xi x x«X en XI P-x en p» xO CM P« On ft t ft0 x* H xiCN CN tf X.n t x p-H tf x o01 X X o

iiXI p-x nXl To o.n <f

XI Hp- ntf CMXI Xi01 01x n-i -frN Or, XH P*çf 31X XX H

CM CNO OI ID OH On n

H CNX X0 OCN -Oxi nr- mr, <foi ox om co01 CMx r-x cnx PX XX XCN ex

4 XIen P- O H

CM cn cn CN0 O O O1 ' I 'DODOxi m n tfX O P- X■o^ f o33 33 ri 33ri XI h -NXI H O OP- CM X HX H XI CMX O Xi Hp~ CN XI H0\ -i -i -^—I P* p» XI

0 01 X Hp» -n o r*-01 X 01 Hn h x rin h xi r»XI P» X" oim «f p» xh en f xx -n h co.n co h xiH X -4 <Tm <f x p-

iiOi XO XIp~ X0 01N X!

p- X.n x-m oXi xtf xm Th htrf 01

H X01 XlN "Ci01 X)01 tf-n aN Nn 0iX Xrs oiX 01

Ia

xi <fH P"n r-^ tfx P>H xo x.XI CMOi "Or- Ox -nr- oO r-o O-o no nOi fN

rM cm0 O

ii01 XIXI XXI HX X)X 33P- Xm inx Tx n

N CMO o

iiO 0iH en0 r-n xXI CM01 CMXI oai xx o

h p--m tp- -nx en

H XJ X H H O

01 O ;X O 1X X" .

<r -hx -n31 XO «X HX tf

4 X

x mp» 01enP-o <r^r enn p-x cm4* H

» <fX <fX oCM COtf ON rM

tf XO XIXi 01X CMP- XX oX Hp- XXI p-oi mP- CMrM P-Xl X

rN CN CM CM CMO O O O OI I I I Id a o o d-i f xi n nJ3 ifliM HTrtf H O f- HX 01 -4 CM CMP» X CN X» P-n h o r** xix xi p" m xim í in 4f bX CM H 01 CMtf rt x rt 330 CM cn O NXI CM X 01 rnx h p» -n xp. X h CM XI-4 If J) ft -toi xi p. x mH X H XI XIp- xi n -o xx en n p enn oi xi o 31xi p- n x» xm CM xi xi htf 31 CM m CM01 f> X ""f CM

■n n0 o

ii.71 Xx nx Pn rNX X01 :mH CM

2 *X) xX XIX XIT XX hXI p.

o mXI 01

CM 71

x tfen p-x oip- ftai r>

-o n m ■■*o o o oi i i i0 0 D Utf cm X 31m x p- x01 X X) enX CM X •*X H en hX 01 XI p.xi o f- rtX 01 P- ftX CM CO CM-tf X «tf 01XI X CM CMH P» X XX CM CM P«H XI 01 XXI X O XI-n en h xm <f p* tfX X 33 31o x xi r>■J f >0 ilio xi en enCM H T P»33 H T Xft XI P- 01

-I ft HfnxHHrN'CMCn'3'X)p»X hHhH-j-HhH CN CM cn CM CM CN CMCMHH_|HH3lp»X».n"

o o o o o oI I6 ó

Xl 01ai xrf (fp- r*P- Xn x4f P-01 X71 O-M P-

X Hr- hn xiN "Mp- nn r>N 7icm nO P>X Xx nVf p*

I-ODD

X XI XI O3i Xi p- X31 Xi -i o0 p xi r-X 01 01 f01 CN O 01en oi n oP» cm 01 fX H XI H-■ X m rMîf ^" _) XI,n xi x pm x p. cmo XI p. xix "-f oi xo "•er --r ox H p» oHl"- tf Hp- X XI P»O 01 x CMHP- 01 CMen tf 01 P»n CM X CMX P- XI tf

o o

iip- r*n aiX) xx xH XIoi -nX CN;f Xlx oXI x-i r-N XlP- nP» 4-fO 'XiH 3>X XXI 01P- Xm no n

0 o o o1 I 1 '0 D D OX <? CM O01 X XI Xcm en X 01X P- X fcn x x- enCM CN O P»P* P- CM t

CM CM X X-H XI O XI

X CO -4 33X 01 CM CMCM CM CM P-CM *"f CM OX O CM X>P- .n X 01tf T tf Hen 01 p» XP- O XI XICM 31 x tfH 01 CM OO HI fN fNO tf P- XX Xi CM 31

o o

iiXI p--I X

7i nh r,n ai4» X

H O■N HXI CMXI XIP- CNXI -H

N XO fP HXI 4fO tfx r-en p.n o■m -iXi CMP- XI

x n

0 o o oI I I Ia D O Dh h n -ftf m p* hXl x p* oCN «T "-f H01 31 en enXl X 01 HH p» p. 31p- X X XI■f Í31J3-4 X X XI«g* f xi cnX O «t P-X O 31 <fXi -| CM 01-4 X X) OfM 31 ri XlCM O ""T XCM 31 X XIen X t X>ti HBO CM CM XX O XI -4H p. H XtO X en 31

0 o o o1 I I I0 D D O31 «-f xi H31 CM XI XIT "t H OH N P- P-M* P X P*p» ft x r-X X O OCM P- ** OXI f -4 H3l-ef ÜTP- O «T O<-f X tf tf33 -i <T ̂O ixl H enX X CM P-p« h n xi<rr oi p- oixi x xi xx x o on n xi .oo f P- 01m cm o xx H f xXI CN X tf

o o o oI I I I6 D O DX» oi p- -np" n p« oiH CN p- Hh ■* o en•J H XI X<f Xi h hX X> XI Om p xi xXI O X op. en ■* o31 CM sf 01m h o opD rjiifrN en x nrM m H 01xi -n cm oX 01 o x>-J-ef cooP- O 01 CMX P» "-f P-CM P" en oO P CM XIO rt 33 -nH P- en O

a o"N XI

r- cmen cmcm OXf-p. XT Xm oX 01-4 CM

01 01X Xx x-7 Ht* Oh n01 P-X x01 XI■tf xx nx X)01 P»x n

o o

iiXI XX P»rM 4*n OO 01xi OrN XlP- enX HO 01H XO CMen or- sj-

H P-r*. hr» xiai P-O On oiXI HX XiX P** iO P- i

CM CN cn en mo o o o o

OO¿OOO0O0O04 XI -

rn HO enHXl

h tfrM X0 Ocm O01 Op- mn oix oai n<r P-0 Tp- o-i xiH CM01 X ÍO»X XI ■

X C001 4*X XO Hp-r-31 r¡P- -Jx nx oiX oX 01<f Hp. CMXI enXi fX x>Xi o

-3- X

7i «■X *ai r*n oP^ HN HXi 01

H XT XX P*m h

I o xH HO "HX p-X XI

i OP-i m xi H ^

p- P- X r> cmCM tf 01 O H0 33 h H XIPflMN-fl01 X» P* O P-CN 01 XI CN XCM O C0 X XX fj" CN 01 enP- H X t CMX CN X B --fCM ifl P» O X).n cn ---f O XP- X» H H CMIB CNiflOtcf XI XI 33 enCM en (N O "-fO CM CM m P»XI H X CM Hp» x» oi co mX Xi O O -efX XI O H CN■f H OiP P31 XI X Xi XIO 01 O Xl o

0101 3131 3101 3101 xX) XP-P-P-P-X XXIXIXI ^f -tf.n ri n rM=N CM- H SO X ** ."O H H** H

O Ö

01 O 33CM O fJ"X X O01 x oH 01 Orn m oCM O XXI en 33h n oí■""P en 01CM X XIn .n -fj-H H H33 t^ rio n enx 01 xx p- xH x XIx h r-■tf -cf Oih r- x

4j nO o

iiX Xh rnoi oai xi■<r xCM P»XI x■-f nX 33H P-n -ftf fXI 01X XICM fNH X*XI enX Xn 33m oXI Xp* -n<3- XI

x P-

■^i r¡ n riO O o O

0 O 0 OX O H Np. XI X XP» >cf CM XI•■f 01 p- HO f X Htf O P» Xcm t n fx cm n enp- x 01 <fONflPcm p* cm ntf H O 01n o n cmfN O H HX 33 01 Otf XI m 01■-f tf CM Ptf P- XI Hr¡ H 01 XICM O m 3DO XI CN -lcn O h -nXI O ri enX CM X X

iiiX 01CM H

X CM« -a-X CMrN "Mçj- XI01 01XI 01f aiH OX XIn cmm "if Ho o-g -fX x>n tfx aiX HO P* ;H «f :O N

Ia

oi x■f nm fx xCM XX X0 fen P*n f01 H■M H

Ol P-en OO CM-M "*

0i trrP- HCM en-M ex« p-O P-

CM CM CM CMO O O O

DODOP- P» P» XCM CM p* enX SO p» fx cn cm p*O 30 H "fH P* XI OO O CM 01«J P« XI HX X CM Htf X 01 XIH Xi Xi HH 01 Xi en33 CM 01 entf p* tf .ntf O H HH CM X f*cm r*- x oiP- X H XIX 01 cm enx -n xi oitj" P» XI 01X N tf 01CM X* p* XtO CN tf X

o xX CMtf p»H NX XP- On cm-M "T

P> -HX P-O XiX f-3- p.

tf 0iai otf o■XI X

h nn p*-H X

.n p-rM OX 01

CM CM CN NO O O O' ' ' i,a o o ofN en X) O31 f H <fO en P- HO H CM X»tf tf XI Xp- o n cm31 O P- XIcm m rt p»X XI XI 31Xi 31 CM Xi31 O O 31O CM CM O0 CM CN mCN O tf tfp- P- tf HCM XI Xl 31■f en nt^n P- CM Hai n tf p-H m XI OXI Xi XI CMtf tj- x x>n H CN XI01 01 x x*

a an xN OX X-f HrN XP- XIX Xp. X31 P-4* O

H -fH X01 X)01 NX "Mo »nP- XXI XN P»X XIp- Xm xiN tff H

-M N

O O

iif Hn tf4f fX 0iO "Mn n0 tf-f nai ■n fXI tf

f OXI «frM O01 cmXl XIX Xlen hx HrM CMX 01xi -gH "fX tf

cm en en tfO O o oI I I I"0 O D O

X 01 tf XO ""f f HX P- 01 XH Cfl PP

0 en X CMxi -n -h oX 01 X XI01 O P- tf■f tf XI [***P- CM X Ox en xi cnX -^ 31 X31 H X COo r- x oiX» r- .-n XlCN "n CN CMH N N 010 33 ^ XCN en X Hp. Oi in p*01 P- Ol CNX CM O COXI X 01 P»O X) tf 01

n h -J rjlH U4f if|P _|H _| H (MfH CMCM CMCMCMCN.NCNrMCNHH-tXl .*) Oi

O O O1 1 '0 D OX P- Ocn h r-H 31 XCM XI O01 X P-

X CM HCN cm XO CM XXI CM CMH tj enCM p» enO «f CMtf P» Xtf N X0 O tfXi o Xtm xi mP- tf enP» O XIrt n <tIhOX X tfXI XI -401 X P-

o o

ii

r- Xeji nX rN

■Xi en01 oXI Xtf -nCM XXI xtf XXI tfr- p-Oi H

OO

iiX oXI 01XI fr- xin hxi •*x n■M P-

0 mrt tfn -iH enH XJtf -Itf o-t r>31 OH HP-P-.01 XXI NX P*rN -IO P"

01 xiH f0 -nH*X Xi* Hp-X01 71XI -rM XXI tf01 Xtf -nnxi01 xio xiN XX XIHO :OP-O rntf X ■P- 31 <en 31 i

a o"f OXi Xp- tfxi O-N N

CMI--O 71x -•■

X) O-M X

x nx 01

oo

üen tfO fMCM f

XI tf

p- p"-t 0iH CNP- OP- fXI Op- 01^ X)X tf

i O tfi H tfI P» CM

O o

iim xio xeM cnX O71 XlP- -I

P- .Nn mX X)O tftf tfo xX Nr-1fM !NXI XP- O0 H01 en4-f XX P-p- ^J

iiCN tfO tfn xX X;f p.P- XIX N■-f XXI CM71 XItf -Moi n^f XIX XX H-n XP- Xlxi nO r,CM Ol

XI X

oo

iiCM OX exP- mm fx xtf tfn cnCM CM01 tftf HX fp» XIm oiCM tfo oio xx x

i P» 01cm XIx -n

0 o

ii01 X^J p-

o r*>xi n-n hx xo -fxi r-.-m ofN XO P-O on oiXi otf xo nn xXI -,

^r xi^r aip- nx x0 P»01 tf

o o o oI I I IO 0 O Dtf ^f «f Xn xi oi tfn oi p« xixi x> CM 01-| Xi CM Xi

tf o r*« <fx oi x HH H O X»XI X> en 33-t p» X X0 H X H-t N XI Xtf 01 H 0101 H XI X33?" ¿> -Ix cm oi m-f o ootf eM cn Op. xi n mcn X» H OX tf 01 oo h f np- oi f ,no xi en o

CN CM CM No o o oI'll0 O O Dx n o oiCN P> f XXI O CM 0133 n 33 Op. 00 tf XIn do xi rgP- P* <cf X>CM X tf 01H XI XI 31Xl <ff x otf ^ p- XIo x x r*CM 01 CM fP» x x -ßX P- 01 fMp- -4 n xm x -f ocm X) O tfX X H f-O CN 01 XIHP* o enX X XI tfP- CM O Hxi n n co

n no o

iiXI H

tf HX X)P- enx no nX X)o o0 enp- n« tf01 31X r>O p-

31 Xi01 XICM P-tf HXI NrN X31 XlX P-

oi oioi aiaioixxp-p-p- xxtxixi q-f ,-n-ncNCNcNHHHP-xi n-

Page 17: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

SUPPLEMENT S61

n n no o oI I ID O On n 01f XI CMf N O31 o P-r^ x nn 31 cmN p» tf

m p» -j xiu x o «fC 01 XI XIOi X XI o■H H O XI

41 01 ^> 01"l P- XI H

-* r) r-Xl p* cmXI o =oo x o0 x mf 31 01tf X XIH OO01 p. P»X P- XI

o o

ii

Oi tfp. f>

-o nX -NH n

-t no oin r-^ h1Oi oX eji

x nx p-O Oin oXI oP- H

H X»n r-rf Ol

Xl X

H H O HO o O O< 1 HI- I0 D O DX tf N Oi01 X x NX XI n COr>j P- XI XX Xt X xip- X X X-I X X CMm en rN x*h p- x -nX Oi ft x-f rn p- 01H'*n tf Ox co x OP- P- f "Cfp- 01 t Hp- P- 01 fn f* o h01 H -4 x-n *r -n -*X 01 trf xoi *n oi en31 en 3i nn f 3i oih m o oi

r -n n cmï o o o

III3 O D OH f O P»1 p- tf x4 XI O X1 H cn H3 B p- n3 x n p-ï X 01 tf4 xi n 3i1 H O CNi m x p-3 tf b P-4 en n h1 P» Q fr f cm xtp N N XI3 en rN XIp Xl p. 131I H -, tf1 P" q- H4 H 31 011 H Ol en3 cm m fg O CN XI1 N O N

iirf CN

X XX tfO HH XI

N NO O

iio fn oix nx "M

p- x-, N

T ON P-XI 01

-t "f31 P-Ji 7101 HX 01P- tf17 271 tfr- Jx 01n nn oi

x xixi ox NXI Xo oP- H

XI Nx xX eji

h 01tr p-N Nn r*tf oH "fO NO Xt

o o

iiO HXI Xi

O 3Dn cnH NX on hf- nx xir- tfp- no -f0 2x fn xj01 nxi xn xiX eMtf »X X■H 33

0 o

iiX XN Xn xi-M -f-f 7lH XXI "M01 x>-f nh P-

01 Xl

X HX nx »IN OXI P-

p, 01P- rN

X "H71 CMn NN fCM X

O O O O tI I I IDODOXI p» 01 XI.n 01 xi oP* 01 XI CMx 31 n en 1-4 n cm tf ■o p* P- Xi 'm ex rt mx -n x o .tf «f p. Xip- o x cm .H 33 x tf (cm 01 m "n :rt O P» X» 1f X h X :H xi x enO en b r» 1-f cn tf o (cn h p* n31 xi r*. x (H-OPt& -4 n -oXI X O O :CM H <-f X Ir- o x O ¡

"frMP--|eNxiB_,0l nCMXH CM en X B h H CM N n XI xtX x

O oOI I IDDOX 31 P*X X XIp» ,n r-

O CN Hxi n p*m en xO N 01H 01 tfX XI CMo p* cnX XI fCN O P*o o 01X B 01B X tfrt p- nCM en OB x fp* n cmX CM Of .n x

IID OX 01x ntf Xn ng- xp- r>0 xH XI■f 31X Ol

Oi XX r-H P-H HX XIX> tf01 CMX 01o o

X BX Htf 01

H CM CM fO O o OI i I J.D O 0 DP- tf X OH CM -4 01XI B o OX O 01 P"P* O XI XH -0 01O■f b n n31 O X XIB CM p- fx .n h xtf .n rn p»x o n p-X xi r*. xxi f cm ejiBP* f HO O X XXI B XI XO O CM nXI B H 01<-» h xi m0 o m cm01 X» Xl XXI «f O "M01 x x o

o o o o

D D D DCM P* X XIP» "f -4 X-f CN Xi XI0 Xt f p*

01 01.n cmH BX X

CN Xirt CNH 01

CN B01 N01 OXI XIo r-o HH XIP« XI-f oB Xi

N N"f Hx ntf Oi■N 0i

X H0 Htf X01 Xi

O H■XI fxi Xx 01X XP- o31 01T P-

3 °O tf

00

iiX B-M B

p- Oin no fo r*H -rxr-x xN XIX Oh nnoX nf trx "ftr -fp- tfXI Xn xH BX 31X XlX X

no o31 tfX HH 01H Or^ XO "fH OC- "fN OX *M

X OH 31X X0 tfH Oxi -nXI -fx oÑf O!x x-M 31

P- X01 o

o o o oI I I ID O D O0 B X 01H P* H BX B f entf XI n 3331 tf tf CMp» X X B01 -n x fB f 01 Hp. tf [». p*H CM XI P*n ni -f 01f en p. BP* O X Btf -n x tftf en h 01cn f n bCM O O 0101 cn r> tf-tri x nb h xi -n-f XI p» BB B 01 CNX P» f CNH CM f P-

H CN cn en tfO O o O O

0 O Ô 0 DP- O p. B 01O 01 B N CMP* CM n B enH *f P- B BB CN CM H XiXI f O XI OP- "f B B XIXl 31 n XI Btf f X rn Nr- f tf tf rxm 01 ^t 01 xiCN H 31 H XIXI H 01 tf BB CM B m BH *f CM Xi xtf 01 x H 31CM H tf CM -4B en xi f xiB P- O XI xiH r» n tf xiH B <f -7 <fB B O Oi BO CM CM P* CNH O f O f

01 xBxien h P- H B 3i0i0iXp*Bx|Xl tf m rN - H B cm Xl -,

B Xl XI tf fOOOOOI j I I I0 0 O 0 O01 xi n ft tfB H "f B Xxi o r- h Ocm cn rn H XiP* 01 01 7 B■f P- -f 33 01tf en xi h x■n sr b m enCO h -f h Hm «f xi n 33cm x B P- tfO Xl B CM 01B X> X 01 01h -n tf b r»B 01 B B OB H H Xi CMCM tf p* h CMXi h O XI 01■o 3} t n r*O -, 31 h OH 31 p- p. XB O 01 CM enf en B en tfXI m B f B

I I0 oX r»X rtH p-0 P-n oXI CM

H HXI H-M B

71 nf p-p* tfXI XIX oN -ICN H

CM P*01 HX o

X CMOi Xr- Oin xi

-f no o

ii71 X-f oo o0 Xn hn en01 tfO HCM OX tfX BP* XXI XIx 01n cmX P-CM P>n xiX Bm -■nnp. 0131 01 rn H•n cm p* cm

ii3- -n01 x0 "N01 Ol

01 o■01 CM

X Bcm BN B31 P»B P»0 BN XI01 XI-1 P--, BXi 01cn en« fr- -nCM BrM 01

m en n m00001 • 1 iD D 0 Oo cm B -nN CN CM XlP* en CM B0 01 01 CMtf B p- CNrn 01 B B«f CM O P*01 P* B B•M CM n OB CM XI P-B XI x CMB B H HB en CM CM0 O tf Htf tf .n XI01 H O 01-4 f O Bp* f n p-fM B X) 'XtO P- O Hf H -t Xxi *f xi mP* CN «f OPn tf -I 01

n rt0 o

iiX HX tfn p-x fxi on xN tfX "M« X01 of Hn fx cm31 ox om H.-n oH 01o 01

DDDODD0OOX N CM XI01 B tf B.n xi -4 btf CM CM Op* B H XCM B P- Om o *n "«f01 XI X 0101 •* n 31X B O cmm xi en bH CM tf O31 B o P-O O 01 CM7 P» p- Btf -t CM eno o m cmO B -4 OCM P- H XIX 31 -, Hcm B 01 BX B Xi BXI 31 XI tfXI f M- O

N CM0 O

ii01 XX XI0 nh 01XI 01tf 31rM xiÍN rih -nn rMn xixi o01 -nO NrM xiO tfx nP- -4tf enX tfX P-P* Nf tfH rM

Xi p-31 OlP- XX P-tf CM0 ON X)01 B01 CMX xiH Bm o"M O

Ol ocm B01 P-en XIO N■7 XO XlXI XItf X.n cnn -f

CM CM CM CMOOOO

IIIw o o oB Xl O "fp- -n o mXl CM f 010 en 01 BH 01 X Btf H P* XICM O O CMB -f 01 CMB O P- OX XI xi «fXi O Xl fH B n XI01 en o cmXI B H CM31 P* CM f33 ^> -t ^ûh h m xio oi m cm-f m m cmp- f B BX X X 01CM B H HH B XI HXI Xi B P-

N N0 O

iin xix 01X -M01 enO B

■n Hn eji■n o-1 orN P-nr-01 01XI 01n op 01X OiP- CMX X

CN CN CM CNOOOO

Ô D 0 DP- H N CMXl CN O XtCM en 01 CMB tf .n BB 01 N OH CM P» P-B 31 HPm 01 f B0 r- B B01 tf B rnB H O Bo m h 01CM B CM Bm p. b tfN O B enCM XI O HO xi cm enB P* B enO B Xl tfcn Xt h BO H O CM31 P* O NO H H BB B B P-

ooooooooooX XX XItf O0 n01 H

0 BXI "fX 01X X

SíX O« XX om bX XIn tftf o01 Hx 01X enX fn BP- X

01 O 01 HX P* x 0101 tf 01 o01 31 H OH 01 CM B01 O "M "tfO P* B tfPP tjP0 P* O P*xt .*n o oxt en 3- cmh xt 01 Xi7 o XI "f31 CM B CMpj p* B mX 01 Xi XB P* n CMm O B BH B rn 01tf h n cm01 CM X CMm f o coH CM CM Ox in f m

tf 31H ÍStf 01O Xl« x,O XIx nx ftf Otf Bp- BO BB 01 ,H H fX B .BP- tri B .B f e

I0

O 'XIXI 01X 31n hxoX BXi o-4^3

O BX Ol

O HH XIf XiN 01B f

iin xiX Hp- p--4 tfh r-

o tfXI P-x nXI tf

o oB Hx <fCM Oen xH en

> m cm4 Oi N4 B 33

n n n tOOOO1 1 1 16000m h f 010 ^ cn 01X XI XI H01 B CM Om O p* r-O h 01 mh xi x Oft O H Xl■n b p» oH H B Oxt "Oi rM cmHif) <f OB ^ XI rMn *r *r -*b o o mO CN en BO) Xi -4 31r^ ~i 33 ^0CM 01 01 tfcm P» en cmXI H O f-f\ft n -o31 XI h encm O O B

1 cm B H eM -f B 31 H H'NcNentf--fxtBr^B0lHH-' ,H "JiP ^ f ,f)N Hill

O o 0I I I0 D OH 01 oCM O tfO h Xtf O Of BPP* p* XB tf HB B CMO Tf B0 Xt CMxi m xt01 O 01n f enxt 01 xtB B fM0 -n OH Xt CMtf p* 3101 xt Nf 01 rMtf 01 Ben x> oB tf 0101 01 co

dU

tf 31XI enB HO Bxi ftf -tf> CM0 BXI tfX CMp- rtHOl01 Htf XIcm 01n xr> 0101 oh nx P*

1D

X O0 xi01 cmtf OiXI CMx 01CM 01P- XH XIXI OX OO HB XXI p*r> 33P- Xtf XB BO OX Xtf CNn aiCM 01X tf

00

iiXI P-01 XH XH CMtf 'XIO XIO tfCM tfcm nX CMCM -M

H HO Ben tftf CMn cmot>B XN P»B XIO 01O tfX Ort cn

HHhHhHHHOO0O0O0O

èooooODO" 31 P* XI H

H B P- BXI ̂ tf otf X f X31 b xt inen tf 01 cnX> 31 O XIxt n B 31O B CM XIXI f -Jf CMm CM P* BS H O XI•7 h m cmO tf Xt P*31 o n 01m b xi enn Xt B BO CM CM xtP* XI XI Bf 01 01 01HP* XI TfO CM Xi BH P» CM BfM 01 r- tf

tf 01xi o-M P-

n nXX)XI Xn hO BX enXI CMX "Cl

XI XIXI enX BX Bx -nm cm.nr-cn fO X

(N O :B O r01 m 1.-n ̂ 1-cf en ;n o i01 01 ;B B :CM P* '

CM -•-f 01 :0 B C01 cn rOO '•M 31 ■33 XI c

01 Hx 01r* nX P"XI CMX Br>r-X 31H XCM CMtf CN31 BX tftf -Mp- p»N 31O nX 01HON CMX B

o o o oI I I IDODO31 H X HB *T en tf31 CN Xl O0i O H HXi en h 01P- Xl o en■4 -t -D 33CM B H OB 01 X CMB XI XI Xto <M h r»P* 01 CM «fO B B BB B B BB CM 01 01n h ri mXI B O BB B fM Bm p* f bB XI CM 01B 01 -îf OXi B B Otf Xi B P-B en O P-

iiin xCM O0 mtf XH BOi XCM Htf mX B01 P-tf p-Oi X0i tfX Xtf Xtf Xn r»XI XIn tfn oip^ x

o o o o1 1 '

_ O D O-îf O XI XO CM b mn f tf bP» en CM B,n p* o xiXI B B HO r*- B eMP* B B Btf B tf mn ft xi cmO 01 H 010 O 31 P*01 CM B BO 01 B P«f B h riH 01 H B31 H P* HP* B CM Hm 31 H "Mtf if 01 CMtt p- cm XiO 01 H inp» B P- P-X XI CM 01

00000000 _

OOOOOO OOOp- ntf 01XI tfP- Xl

01 01X P-n xn 01O NX "fP- P»O Nh -nxi Np- N

x n0 *nai 01B P-01 rNrN xiX Ol

X n

O Hn aiH XI-f Xx n-m mx 01h oP- OX -4.n 01-M CM

X r,N CMXI tf0 xiCM P-CN BX XI0i 3101 XIn 01H mH X

-M n31 xXI XIX CM° 2p. XIX Xf 31n xin aiP- CM

H OiX fHOOi tftf cm

X P-O 31O P>X tfcm -fP- Hxi rt

Xi 01Xi X0i P*H CM,-M CMO entf 01r- -fxi p*x xxi enx p-n 01P-OP- Nxi rnX 01X CMO X

o o 0 oI 1 '

w O D OB B tf 01X B 31 31O 01 B fOOP*B en B oHB enfNO B B 01B 01 h in0 P* Tf Htf O Xi P*tf m b hB B b r**n o -f hXI B 01 H01 B -i^H CM H XIX B 01 OH XI Tf en01 xi r- hB O f XtH f O <în 01 01 cmO B BOB n h o

N "N CM CM0 o o o

è' I Io ô o01 XI B B01 N 31 CMB XI O BXI N tf 31XI N -^ CMB 0i 33 XIB en XI -nB H p- <fpj xt r* 01p- xi x xiPHifl-efB B B Ntf CM XI mo o n enf <f r> h0 B tf HCM B B XIrM XI çf 3i

01 CM fM^1O f B fO B O XtO CM HXin p- eno

iiX "fO BX eMCM 01B P-0 XIX fx oB OrM ftf CMH fn x<f xtf xX) tf01 o0 01x XIp- xiN P"*n 01x n01 o

cm en n tf0 o o ol'ii000031 01 CM XIB 01 P- XI0\ ^> ^O rt01 N O fNO en cm HB h r> oCM xt xt .-nH H p» Bn b p. hai 'Xt n 01B B tf p-o o O 01xi 01 b *nb xi n nB H CM OIs- 01 CM H

B XI B Bp» B .n XlP" B H HXI O Xt "cfXI B -f XIP- CN XI TfB P* H P*CM O O 01

01cji3lOiOi3lOl0lfJifJiBBBBP*r^'^P*BBBBXlXtXVtftrtjrn.nen.nfNcNC^

Page 18: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

S62 SUPPLEMENT

i i0 oX> NX Hn -O■n -fx p-71 n71 Xtf nn B01 OX XN XIX X)H X0 Xn p-XI tfN OX B01 HX Xix noi xp~ xi

tf Tfo o

iiXI fn o0 Hm r>rN 3101 XItf CNtf HX -ncoi XrM 31h XX Htf tNX P-O tfH entf tfH Btf HX P-*oi np- 'XICN 01

en en n enooooliliO D D O0 H tf oítf O O Hcm "7 01 cm01 B XI Btf O P* CMB o n tfCN 31 Oi Ooi b «r i>B tf H N33 r** m B31 -n 31 XICM CM tf HB B X CM•7 H O ■*tf H 01 OH O 01 tftf P- O Otf CM O Tfrn en CM BCM O B P-B Xt xi P»p* B tf r»B e3l B 'XIB P" CM H

n rMo O

ii0 0101 OX tfXI 31B "M01 CMX BX P-H n0 B31 X01 O

-f XX fp. XIX "f01 H-N 01

-f XX P-

•N rM0 O

iiN N01 0171 XIO XI71 X

f 01P- NTi OX 01n O71 tf

n r,0 tf01 Xlcm -n"M B

x n4 31

CM CM CM CMOOOO1 A > XDODOHO ejiPo cm p- nH P* O CMen B Tf xicm tf h enoi n oi B-4 -t r*. r*r- CM o eMO O B Hp* tf n enf> N B 01H O p- 010 h en xiH XI B 01m o h tftf Tf CN tfxi cn f o■Xt b o enCOP r-jJiH en B P^n b x hXl "M h 0101 H 01 Xttf h r- xt

CM CM N No o o oI I I ID D O OH B "f CMB en cm xiB B X Htf O P- tfB B -M HH CN O OCM H B r-f h tf mB P* P* Hxt m b oiCM B X BP- H CM Htf Tf rN P-U B P B0i CN O Op» O X XIO "tf B 01■f B p* P»-t 33 f \0¿O B H BB P- X O^o -o n -op- CN Tf P-tf XI B X

O o

ii-1 p»

tf oX oXI cmr- 0iX cmX fXI 01XI "M

tf BX XXI Xx otf X» B31 "MX XX OO Ho intf nxi fn oO N

O o

iixi oN sf-h nN ■>XI tfO H-N P-X CNXi OrN tfX 0iP- 01XI XP- Xip- nXI -I

B BB BB P*01 OP- P*ri ft

iiN DJl P-

P- X0i fn p-X CMXI NX 3iXI H

N XX Xp- nX P-xi XH Hoi .nf Ho p-O Xh n"f xio xiH rOlX O

O O

iiP- ON OO Xl

H HB fn np- fnr-O NX 01P- oo oxi mH 0171 P-O XIn nn cmX P-O HXiOp- tfxon p-

o o o o oI I I I I0 O O D 0X H B cm Oift p- n cn jiP- B 3 en Tfxi en n o f01 01 31 CM en■n B H CM HH B 31 tf p-.-n b h o o,N X "N f Hifl HP* H

H O p» P- tftf CM B H Bm b tf "> xH O O H CMO XI h XI tfoi x .n tf xiXI B p- B hn B O CM 31B H ri Oi 31tf rn h CN BXI B B P" 01B XI CM tf tfB en n cm x-i 33 JJ *r r^

i tf B HCMtfBXHH" rNrn,netfxtBp.X HHHHHeMcMCMenrntfXi n

0 o

iiX P-31 01n p-xi n01 x0 xi01 3101 -4p- tfr- Hxi aixi xX XI

0 Otf 7101 CMP- Xtf 3101 tftf "Mtf "Mn r--M 3D

X P-O Bxi XX X

OOoOqOoO

b h n H Xi

B B O Hn tf h oiB 01 H 01.n h cn cnp- H ri r)cm B cm Otf B B P"»B m fN CMO H B BH tf tf HXI O H (OlOi xt b nrt oi .no

H Otf OH Xo oO XIn h0 xt31 xXl X01 X

o inxi r>X XItf x-H xn tfX "Mo r-en XIP- XI

Oi P« 01 Xicm p* x r-*3 B O 0iCN tf CM HCM O m CMCM H tf P»m tf o hcn tf o m«f X 31 B■n o p- tf

iii

X f3 BP- X

tf -Ntf Otf HX XIN Xln oi01 Ol

P- XOi XX rtCN CM

tf tf01 Xl

3 7101 H

IO

O ncm tf-n oip- xio enP- CM-M CM

X enO rNtf XrM r-P- -MP- BXI H

n oitf CMX tfp- nB Xp- nxi xO H

.n b

o o

iiP- -4CM tftf -MCM H

Xl HXI tfX tfX tfCM X

0 BX cmrM 3xi Xl■Xi nXI Xl31 XlP- OP- tfH 71H X01 'XIXI tfn r-tf 01

O O O OI I I IO O D O«f cm m r-H B n rMXl B tf 01Tf P- XI PIen oi m tfCM B 01 O0 XI H P-fi|P ooCM B CM On H B oB B tf 01n CM B B31 CM P* Hn 33 nocm tf rt P-P. B O tf0t 01 o enB P* 01 O01 P- 01 enCM Xi CM XIP- tf H Htf en cm CMO en B 01xi O xi o

oo

iiXl 31B HXl 3m tfp- p-m bB XICM 01p. x01 p»tf BH HP* BB enB enOl CMn tf31 tftf CMP- oXI tfO tfn x

o o o i

B P-tf b■tf B

p. 01 ■p" o ;tf n :31 H H H n H

-, CM

o oI Io oO Bp* p*fN 01O HX Otf p-H XI

p- 01tf Btf BP» enB enP* OO XCM tfH Bp. BB 01r*- xitf cmXI tfp* m

n enXI P-CN tfn ooi nh p»71 P--i r-71 X

3 tfn p-XI Xl.n xitf XXI 01Ol XI

noXi Hx oXI H.n ai-, p-p- tfX P»X Ol0 XIN Xx on O01 tfX CN.n p-

CN CN

3 3

iiH 01XI HXI oX 31P* XH Ol

tf Xo X-N P»

o o

CN cn m en XIo o o o o

0 o o o oCM H 01 CM BXi Xt X CM X

P* 33 h CM 31m en b o h31 B X H BP» B tf en cmp- 01 B P** 3301 CM r> Xi B.n oi m r*" ho tf oí oi mb oi p- n xiB f- 31 CM 31B P* O O tfm oí tf o mrn m rn xi oíB Tf H B PJXI O CM B np- B tf p. HCM «f B XI BfN O XI B P*H tf B P* enO tf P* XI oH 01 O B B-4 f) 3\ *f C*~

01O131O1 0101 xx 33 r^ r*- ■& xxixixi tftf ri n cncn -^ -i noip-tf rn h b -

XI tf rn000I I IODDen h xiri cm Btf B BB tf CMONOOl B Hp. n h

go -n 01 o4J O CM 01c o m oQ1 XI X HH XI h H0) X B CM3 tf P- P-

31 tf XiX Xi 31r- xi oB 01 HXI CM tfn cm .nP* H enb n inxt n oitf P» -N

m m m -m0000I I I IOOOOH B tf BO B O 31H Xt B Bn oí b tfo o xt »tf 'XI P* Htf O tf Brg B »tf Bn r* cm hXI 01 p- P-tf h m cmB P- CM tfr*- rN .n enCN tf Xt CMH B O Om B o BX P- tf Bs> -i ^t ^hpb-în p* p* bp* O n Hb n tf oO "tf B 01p» O tf CM

CM CM CM CMOOOO■ ■ri0000Xl 01 o tfCM B 31 tf^ (Ti r^ 33tf CM f» "tf31 01 B BX CM H XIXi CM O Hrt h xi en-i -i 33 fn 33 33 -O.n xi cm hO en xi xitf P* B Xxi r-» n p-B B 01 01n tf tf BB tf n 01CM tf P" 01tf P" B BP- O P- P-XI X CM XIO O tf tfB cm r- p*B B Xt B

tf tfX 01N XH XI0 Bn b71 31XI CMx r*XI XX tf71 tfX Btf XP- Xtf -ntf tftf OP- 7iX B01 XI

> O OI I

I D DI H fM1 X B1 31 CM

O oxi r--M 3

tf OX tf-M XI

* 5p- XX oX OlX nXi Xtf -Nn xien oX X

o o

iiN B31 tfP- XO P-(N nN HXI CMX P-X CNXI XI

x oo on aip- r*O NO B

t N tf» C31 H1 h en> o oI tf P*

OOOOI I I IOOOO

p- B b hO r* m -nB P* 01 31tf tf CM HOi H B tf■JiOpP

O B o tfcn h p- p-H <M B P"B B o Btf O O 3D31 Xt p» tfB O N tfaBpoiO B XI HX B CM XIx r- B r--rt p- tf OP- CM p» CMift P» o XIB tf H CMp* -n 33 Bb <n p- bO tf B tf

0000I I I I000001 B N Btf tf n h0 .n n b01 fM B XIXt en B H01 O X BO tf 01 01H B p- Btf H 33 HX H o 01N B tf tfB P- H Hm o* o tfX B en xitf en cn Hm 01 n P"Xt B P- HB H O OCM P* B CMB H O P-cn O xi Bh rt 33 01p- tf tf XCi f <^ ^>

BtfHfNXtxH h CM m tf XlP^OlHHHrNCMtN.entfXtBtf

I I I0 O D

Xi 01 B*tf CM tfCM B enxi r*. 33O 01 BXI CM CMP- 01 NCM 01 H0 Xi P-■n o o33 J) -*iG r^ -t01 n p-0 B CM01 m tf0 CN CMCM XI Hh xi mB B 01cm b -n33 xi BB B tfCM B en01 P- XI

00

iitf n0 01CM 01.n oiN Nn htf -Itf p»01 XCM XH XIP- BN Oxi no nn Bn P»xi nX XI-M B

tf -MX XI■n p-CN X

'iD Oo on cmCM XIB Bri o

P* CMX) n

3 xi3 OXI X01 XH HH P-P- enXl OP- otf Htf Xtf 01

o o o oI I 1 I0 O D D01 H tf Xp. XI o en33 ri ^ ~ixi h xi mH en XI BB p- n xiXI B X H01 m tf xtCM B O tfm p- xi bO B O cmB B Xt tfO H tf OO tf H Ob p- m 01XI CM O Br> P* H CM^t ^t S3 riCM rn XI P-31 01 XI tfXi CM CM BP- B O tfrt r^ 33 -Ûtf B CM B

00

ii01 oen o-1 CM

p- p-

O P»-, XI

p- nn xH Hai .nX oo r»r> tfX X-M XI

n xX XItf XH tf-n xXI XI

tf CNO tf

o o o oI I I I

O O D DB XI O OCN B 01 Xen P- O oB O O OB O O "^en tf p- BH Xt CM 01Xl H B CNOiO B>CM B B CMXi O B BXt P- B HB B P* Oi0 B CM tfB B CM CMtf n o hCM H 0>Xl B fM CNen p* o mH B 33riroi CM P- tf01 en o xtO H Xi CM33 C) ^i -t

oo

iiX P»X XhOXI X01 Xcm Xlp- tfXI "M01 NXI P>XI -I-(XI

H XI0 nn xiN CN01 XIn 3X rM

-M CN CM CM

0 o o o■ I > ■OOOOh n xi bcoi B en P-n 01 o b01 O» B O.n m B bp- B cm Btf Xi tf BCM O CM Bo tf 01 op- p- tf p-CM B B XICM H Xi no o n Bç^ -4 33 33H Xi h OO cm m CNp» B fN P-tf CM tf cncm B XI 01B Xi Xi BCM 01 Xt Btf Ol O OB H 01 P-m m b h

rt ft

'i0 DCM XI

H O-4 5n cnx xiX inx r-p» xix on OOi XiXI XIH O-n ox P-P» Xr-1XP-p- xiN CMP* Bn ai01 Xtf tf

OlOiOiOlB BP-P-B BXI Xitf rnencMCM h- X Xi cm H cn tf

Page 19: On Computing Gauss-Kronrod Quadrature Formulae* · Kautsky and Elhay [15] and Elhay and Kautsky [5], who compute the nodes t* as eigenvalues of a certain matrix derived by matrix

SUPPLEMENT S63

O XIo o

iiO Btf encm P*B Btf Xn XP. BX tfm On p*o cmO HBP*P- "M-M tfXI B'Xi P-

X Xtf Xoi n"N OlXi XP- NP- N

ii71 P-N X-n p-01 3XI XH Htf tf"M CM

01 CNtf XCM XIO eN71 Xn Oirt tfO BH 01

î ¿Ti 01X Nn xi-n oin tfx x

tf tf tf nOOOOi i t idodoO CM B 31O B XI tfP- H B Htf P* H Bn B B P-31 N Xt P-r- B CM XICM H p» rnB B 31 tfO H H BP» tf N 01XI CM H OH en B B& a\ r~ -tB O CM Htf n b r-H tf CM P*H en B XIxt x» o enH H OPXi P* H XI

tf H XI Ntf P- o P*tf P- B CM

O O 0 o o ■H tf tf 01XI Xt P* X)o o h -nn oi tf tf4) -4 S3 'S}P- B H BXt B O 01m h tf bH H 0i tftf XI N N0 01 B OCM H N ON CM O Bp» 01 O P»tf oí m oB tf CM rnp- H CM enn o h oiCM H tf BB H tf 01XI CM Xi tf01 CM h P*B en B BP- tf H O

.n xiH Btf 3N Hn cmH CM .

n no o

iio o? aX 71-n OX 31XI X

tf Xi

I XI HI X B1 n oii n oi' or-> xi tfI CM P-

I tf P-

!S3• 0i tf' 01 01i 01 P-■ p. n

N NO 3

iin -nx HXI HN Xx n.n o

X OlO r,tf XX n0 inX NH Btf Htf Xx -nXI P-

f tfx p-01 DCM HCM PO

N NO O

iin 'Xix P-O P*n -iH 71tf p.Oi OiN 01x r-n cmo p-0 n31 b0 on cm01 71x no Jdn xiO On cmX tfxi OXI X

N NO o

iitf XIX 71X tfn -ntf x

n bp- tf-N tfXi 71tf 3N Op- O71 tfX X.n xih tfo nH O

x nx oO tf

N "Mo o

iitf CNp- tftf P»N ON Xp- rip- P*p- tfxi tfN N0 oX 01-n X01 CM01 Xo XO tfO HH Boi nm p-■xi xin 7iP- O

N NO OI lD O7i -n0i tfX) tf

n p-n xH P-m n-i p-CM O

X BX 01p- XN NN P-tf tfXI HX) tfO XiXI Ol

N N N CMOOOOI ' 1 I0 D 0 O

P*- 31 P* OXl 01 O BB rn Xl HXI P- P- CMB B XI Htf B B XIXI CM B 01tf CM P- XItf 0i H OH "tf H tf01 O en OX O H 01p. CM 31 O01 tf 01 XIft H N 01o in p. cmO en B tf31 Xt N BO B H 010 O B OXI tf N B01 01 X BH H B Otf 01 tf H

CM CM CN CMOOOOI I I IDODOh h o -nXi B h 01tf X X OB B tf Bet ri o -iB (N B en■N H xt NO en p. Xn B 31 -mtf tf H 01O P- X tftf Oi p. xoí tf n Boi en r^ ooí o 3i nX tf 01 cmB B h enXI P* 01 Xlcm m tf BO 01 cm Btf P* O enen O h X)tf N O XIP- tf CM O

CM H

3 O

iiX tf-N tftf OH Xn xjXI 71x nn otf oX B01 H71 tfXI 01

-n 3X BXI 71.n hH HH B

ooooI I I IDODO

OP B'Jitf en cm 31o tf xi n0 p- n tfn b p- p*B Xl Xi op- n n rnrN O B CM33 ^ 33 -ip- CM XI Bp- 01 31 tfXI X tf enH xi x en01 01 01 tf

0 B B XI01 tf O tftf n n rnjiB ptr31 Xi B fNB B cm enN tf B Xl31 H m Bh m tf xi

OOOOI I 1 IooooB N Xi HB P- 31 tfh b p- enO O P* xtO H X XI■n tf -, CNH B N enB B O tfH P» 01 Xlxi oi n cnH CM P- BXI XI o CM-4 S3 33 tO 01 -4 Hx -n B Or*- o en cnB tf CM BH 01 p. rnXI 01 CM tfB CN n Nn p- xi bO tf P* tfn N H CMp« ai h m

o o O o o1 I

- D D D DN H r*- 'Xi BN XI tf O P-XI O XI CM XIX "M tf B P-B tf CM en BN en P- H XXI en H O Hn xi o h nP* O CM 01 Hp« xi ift n oB B XI XI h.n p» xi oi oiH B 01 B XB B f^ tf tfB B cm cm P*H O Ol B hH P* N B tf-N rn P» B tfb p- n en xiN 01 p- B Bp. m o en p-xi xi h Oi nn 33 h p* p»B O B H P-

xtmrxH.nxixH h^h .ntfxiBp-oiHH hh rNCNrMcnmrn tftfxiBBP-xOi oih ^ -i -, -* h h cm cm cm <n n tf n

iien Xim on hXI tf-H H

h m

XI X

O P-XI tf71 -iH tfX Xtf 31O CMp- mB Om nh ai

tf tfXl Xx xX ntf HN P-O 71O tfnP-X rN

X3Oi xncMO CMr> x-, X

H XH Hr*fP- 71X P-

o o o o

0 o o oH CN B Oix oi xi n01 O tf CMO tf O Htf O H BXI B H Bh m h -nH CM XI Hcm B r> P*tf O CM B0 XI tf 0101 P* P* Ho b tf r-n b n p*S3 -, rt 33B 01 XI HH B XI Xttf P* tf or- ai m hCM B tf Htf b n tfr~ -O c** -4oi p- n bB XI tf CM

ii iN xiO CMN tfx tfO B71 P-H 71O OO 71

n oix oO nX tfN tfx xN H

O 71

0 o o oIII

_ O o DP- tf CM HXi en 31 oP- tf P- Hcn tf B P-tf r* n r-H O B Xt01 P* CM BCN tf P» Hp» Xt B OB Xt XI XP- B H BB H tf XI0 B h H01 O p. CN0i S3 r^r^-O H B BO 01 p. hB 01 CM Brit^ oi cnn en b oXI O CM O33r*- -, -,O B XI OP*- tf -N O

o o o o

ooooP» X O Xn p» tf ooi B rc- riB 01 CM P"en H B XIH 01 H tfH cm O eni/l P* N 01B tf O CMXi B tf tfXt X 01 CNCM tf P- HXi 01 CM Boi tf n tftf p- n h31 H O 01H H CM CMm CM CM 01o tf oi cnTIP xi enXt O XI ox b n hm B B 01P- tf H X

iiO NXI XP- HX Xitf NXi tftf CMX P»•M 71

X -NrN CMX XI■N X

tf HO P*XI XXI NXI XIp. Ol

tf Hx mO xi71 XX) CM

??D D01 tf0 XJX xi01 tfX tfP- Xo mX NX 'NXI tf0 XIX) P-

X HX Bp~ -n01 H

x -n0 n01 BXI Btf Bp- XI01 X

o o o o

DODOB tf P- tfX XI X BP» rn tf CNo tf x p-n cn x tf31 O tf BB O O Bri B 3i HCM B P- XIXI XI H BB H 01 01B H B Hp» tf p- p-p- r- cm 310 O H HOi 33 r^ 33■ft Oi p- nx en p- m01 P» 01 XIh N h -nXl 01 en 31o h .n tftf CM O Xen o p- m

o o o oI'llOOOON O H XS3 't 33 S3B tf O Hm CM tf Bn o .n cmen H H XItf cn n on 33 r^ -0XI O 01 o0 P- H B31 B H 0131 rn Xt Xtf CN O Hn oí xi xiX CN 01 P-01 Xt H CMCM 01 H XIO B O 01O CM X Xtf 01 31 HH CN XI N31 B CM CMB XI Xt XIO P* tf H

oo c

iiN 01 Cp- tf ;tf Xi ;

0 HX Bai ntf p-m hHCN r

P- BN X)X P-tf ncm r>H Hp- mB cntf p-tf Xi01 X■Xl P-33 ft

X Xx ntf N01 01

i o oI I

l o oI X BI H en' O Hi n tf

01 P-01 H■M (M

XI X

0 XX xiX 3101 tftf .7i■M X■M X

ai r*

P- OO xx n

-i -i O -i S3 -i71 P-

tf XXI XX Xm mr> Xm xrN P-X CMtf B (H HB CN toí cn tB Xi :P* XI .'

X XlCM X

CM tfn p-01 tf01 H

H 01X X

XI XN tftf tfX Oxi n3 tfp- p-XI XXi -NO Ntf B

I XI Hi B XI

p- tfi x Oil 01 P-l B riI cm X|

N CN

oo

iiN Oen tftf XP* oX Bcn enm p»N NH P"n oP- "Np- xiB BH XioorOl X0 oH enB BXI 01XI CMB tfXl tf01 XI

-I 01

-I tf-n nX Xtf oOP-tf p-01 71cm 3O 71X P-O 71tf tf-i X

■n n

CN en m tf S33 3 o O OI I I I ID D D D OB Xi p- 31 Btf CM 31 O OB H tf tf encn p- O cm tfP- 01 CM O tfrN N B B Xlp. 01 tf tf otf CM O O XIH B O B Otf tf B P* tfB CM H XI HO P* CM P- 01O B tf 01 HX O B B X31 33 tf H tfP»01 m O BO B H X tfO P- O B enH tf O COI -MB 31 B XI nn cn h n 3iB X h B 31tf r- tf xi ntf X fN O H

OiOlOlOl 3131 OiOl OlB 33 33 XXr-P-P-B BBxiXI j-| XI tf tf tf m mmcMrNCNCN hH -i -t 01 P- Xltf n CN hP mP n

r tf tf n en) O O O o

ooooooooooo

n en n cmOOOO

I

1 P- H> P- Tf> m xii n oii xt xiI Tf CM1 tf CMI B *>I P- XII B tf' N B

» B CMI x CMI Xt P*> B en» xt rt> O en) h p*' B B

' tf 01• P- 01l Xl p«

p* XI P» HX 01 H O»B 01 O tfo o m x»O Xi p. OiN H tf Xt31 33 tf B-^ S) rt ^H P- P* 01tf O B CMn CM Xl OXI P* O Hx oi n tf31 01 P- 01O B H Ocn O Ol P-3i en B O4J 33 r^ •t33 CM rt CMXI tf h P*O H 0i Bp» B xt enB H P* 31P- O H P-

xi no xn aiai x0 x-M CM

X 7171 X3 P-Xi XH 3X HXI en

P- N01 71P- H

O BO XXI 01

01 P» B XI33 S3 33 nn -n p* hp- en h hB "n O CN31 P* 01 01B X O BO P- tf Htf en Xi CNo >n oi mp- CM p- 01tf H X 01tf H 01 CMX O B Olxi en o XliflOPPtf B tf tfB O en 01X B rn enB H O XIB XI 01 XiXI H n BXI B B tfOi tf CM O

CM CM CM CMooooil i i0 O D 001 31 P- Hn ft xt oirN tf X Btf B Xi B0 B Xi CM01 O B 01B H CM OB XI n tftf p» B Bp. Xi Ol nft oí n tf0 01 XI OiCM B P- Op- tf 31 XIcm B xi nxi p- n tfp. n b p-tf o o p-H 01 B 3101 H tf 01p- CM Oi OXI O H HO 01 H 01rt XI 31 CN

N NO o

iiX Oi0i P»X -MXI H

H XO BOi Xp- p-n otf tftf XItf oo xtf tfX Xl

X B

p- XXI HO enn rto oN XP- -4

N cmOO

iiX P-xi ai71 xioi op. tf_,B ■O Btf xiX XIx xo -jXI tfp- tfX 01H 71CN Xi ■

-M N

3 3

iiai oitf tfx ntf on tf

XI XI

o 71tf Hn X

;N N3 O

iitf CMOl H

x r*oi mp- op- Hn oii xi

Xl BH en :Xl Xtf O71 XrN enX o

4 CM

N NO 3

iiN 3n xix P»n -mn otf nx x

iiiai xOl xxi p-

n np- r,m tf"M tfn tf

I B Oi B H XI h Bp- cm rx in :°2 •X 01 ;p« B .'BP 'B CN C

l B ejil Xl P-i 01 "MI b eni X P*I B eni .n N

X 71ai ain mo nn tfo oi

i B 01i n Oi• P- BI O CM

X HCM OX CMX Bh r*B BB tfX Xlm bn cmx m

ia

XI oB CMtf en3 HCM tfp- nH ejiCN XICM BCM CMtf CMO BH BH H

tf CMai otf XIB P*0 nCM BH BB O01 tfCM tf

O OI I0 D01 OB HB Nen r>01 BB P-X BB tfrn mtf on xi01 HB oiH O.n bB CM01 enB XIH Xm cm0 P-B H01 01XI p-

n P-H CM01 oX XH B71 X71 0iP- 'XI0 -nP- Nm x01 P-P- NX P-tf HXI XXI XIXI "Mx r-»x xn 3rt x"N XI

3 N

H H3 O

iin xn XO B31 P-

H Xtf XX 010 xi01 01■n Xp- p»rM OP- CMn ntf OX -NX Br- XIXI XIX 71X P-X X3 P-XI X

O O O OI I 1 IO D 0 DO XI X Xp* m p- en^ n -, r~-ft x H xiCN O P- tfB B O HCM P- cm cnB XI tf Btf p* tf p-H H -N BX tf m XXI XI CM BXi 31 N CMn xi n m& r* & r~O P" P» P*CM Xi H HXI H xt enp* H B enN H tf 01X XI h XIb o o nH XI h 01tf O P* H

)HcnP*LHHcMcnTfBxH -t-t-tcir^ririiTtfin&v^ xoihh -, -t HHN'NcN'N'.ntf Tf en

O o o1 1 IO 6 Dtf p- tfB O tfB h BB 31 BN B -nXI XI Bb xt nB h 01P" XI CNO P- Btf tf P-P- B 01O tf BCM 31 h3>ft33O Oi tf0 h Hen tf Otf 01 HH 01 Otf CM 01n ri ^tP- H cm01 01 B

iii

p* Xn oip- x

x nx no p*

Xt 01m p- xi bN rM CM Hx tf n p-H O B tfH XI tf tf01 O Xt H

01 XX Xi-M B

Xi X

H XtP- Otf p.tf 31N CMtf P-tf tfr> O

■xi ■h mn oXI xioi nai on cm

o o o o

o o o oXi B CN B0i H Xi Xttf P- P- Bxi xi m P-r^ r^ 43 ri~t^> rt hO O p- Oen f B P"tf p» n xtB -tf -n xtP» xt b OCM P- 01 Htf CM 01 BB H H BP« Xi B NXI O O 30tf B B 31B in CM CM31 S3 H B0 B H XICM XI H tfXi m B CMex tf P- 0101 p- tf H

O O O OI I I I0 D D D

P» P* 31 CMcm CM P« P*31 33 h r>tf en tf On cm h xip- tf tf n01 CM tf CMB B o XIB en B P-P- XI X OO B h Btf cm P- tfXt B O OB 01 r) 01tf r* Xt Bxi cm .n oOi tf h Bxi tf n tfP- CM 01 Bm r- p* bH 01 o CNO O XI B3i p- -n bB Xi CM X

H -NX HCM OXI XI31 X31 tftf X-4 O

p- 0101 Oxi xO XH XI-n aiai oO xiB Xx r-CM tf

nD 001 Xtf nX 71XI P-

-n tfn xixi xai cmxi .ntf tfh xim tfx xN Xai -nX oH Xtf ai

h ain xB B

CM B B OXI h P- tf

O O c

iiHB-M X

O BP- NncNOP-tf p-ai mnoXI 01n xtf x

Î27i OiX HX OiO P-'XI XI

X tfai p-tf "MX XCN tfO X

ID

CM XX P-tf XX X0 xiCM CNX tf01 H-4 X

O OXI XO B0 P-3 en01 -MX NH tfP- Nx nx oxi -nx Hx aieN X

u0 DX o

X tfn bx nop-o tfOi X^j 7iH tf3 71X P-xi OX Xn xitf 71tf "MO CMx aip- xrM 01tf XlfM X>xi -,

o o o oi ' i 4.O O D Or* h o hB B O Hoi m oí enh tf n otf xt XI oCM O H CMXI H Xi CMB O tf OO O CM B0 01 tf Hn o p» enXI en CN HB O B Xitf CM B -NX en XI 01b in tf xi01 B B rirt 33 Ol 01XI XI P« HH tf B rMx tf xt enB O O P-H X X XIX tf H X

X oXI CMxi Xh np- tfX 3ai xXI p.

m xo aix -4o oip- nn rt

tf xtf -Hp- .nxi otf Ntf xtf on tfoi tftf o-4 tfCM P*tf tfH H P» P»

71 P-B Otf XIO 0iX Oi3 n

n xX Xo mp- oxi -n

N CM3 O

iiXI XIoi -nn Bp- xx P-x P-p- oX xiX XlN NN aiX tfXI Ntf o3 Xp- tfn BXI tfp- tfai nn oXI -n-, xtf X

cm eN cm eno o o oI i I J,

0 D D OB XI Xi Xp- H 01 rMp. O h Btf P" B OX P- 01 tfen CN N X

01 B p- OXi O P- P»0 B H P*XI P- H tfXI tf B 01ai n xi xh b n otf H 01 xib tf x ntf n x xiX B P*- XIiii B X N31 P- B rM3i n tf xi01 B en rNCN tf H BCM O H ft

iitf nX htf NN NP- O

H Bx no oO »-n p-ai xai noi xp- n"M 3

31 X^J XX r,x mtf xn -m

0i0i013l0i0l01XBXBp«P'p-B BBxiXlXItf tf.n en en fN CM CM H -4 -t HXXtf en N _,tf roiP»