Sunteți pe pagina 1din 5

Problem 2

The Stieltjes series is defined as S(x) =


n=0 (-x) n! . Choosing a reasonably small value of x = 0.1, we
n

generate the sequence of partial sums for the first 33 terms in the series, then determine the optimal oscillation
region.
In[2]:= SN[N_, x_] := Sum- 1n Factorial[n] xn , {n, 0, N}

In[3]:= Data = Table[{m, SN[m, 0.1]}, {m, 1, 32, 1}];

In[4]:= poi = {{9, "Lower Bound"}, {10, "Upper Bound"}};


annotation[x_, y0_, y1_, txt_] := With[{dy = .0025},
{Line[{{x, y0}, {x, y1}}], Rotate[Text[txt, {x + 0.3, y0 - dy}], - 90 ]}];

In[6]:= plot1 = ListPlot{Data, {{9, 0.9154563200000002`}, {10, 0.9158192000000002`}}},


PlotRange {{0, 20}, {0.911, 0.921}}, Joined False, PlotStyle {{Blue}, {Red}},
BaseStyle {FontSize 16, FontFamily "Helvetica", FontColor Black},
ImageSize 550, AxesLabel {"N", "SN "},
Epilog annotation[#[[1]], Data[[#[[1]]]][[2]], 0.911, #[[2]]] & /@ poi;

In[7]:= plot2 = Plot[Mean[Take[Data[[All, 2]], {1, 20}]], {x, 0, 20}, PlotStyle Dashed];

In[8]:= Labeled[Show[plot1, plot2], "Fig: Sequence of partial sums of the Stieltjes Series.",
LabelStyle {Directive[16], Italic, FontFamily "Helvetica"}]

SN

0.920

0.918

Out[8]=
0.916
Upper Bound
Lower Bound

0.914

0.912
N
5 10 15 20
Fig: Sequence of partial sums of the Stieltjes Series.

Minimum oscillation amplitude occurs between N=9 and N=10 (optimal region). Thus the lower and upper
bounds are given by the partial sums S9 = 0.915456 and S10 = 0.915819 respectively, i.e. 0.915456 < S(0.1) <
0.915819.

Printed by Wolfram Mathematica Student Edition


2 PROBLEM 2,3.nb

Problem 3
e-t dt
The Stieltjes integral is defined as . Evaluating this integral numerically for x = 0.1:
0 1+x t
Exp[- x]
In[9]:= NIntegrate , {x, 0, Infinity}
1 + 0.1 x
Out[9]= 0.915633

So, the exact sum of the Stieltjes series for x=0.1 is asymptotic to S(0.1) = 0.915633 which satisfies the
bounds established in the previous problem.
In[10]:=

Problem 4
Listed below are the first 18 coefficients in the power series for the susceptibility function.
In[11]:= An = {1, 6, 30, 150, 726, 3510, 16 710, 79 494, 375 174, 1 769 686, 8 306 862, 38 975 286,
182 265 822, 852 063 588, 3 973 784 886, 18 527 532 310, 86 228 667 894, 401 225 391 222};
1 An[[n]]
In[12]:= data = Table , , {n, 3, 17};
n An[[n - 1]]

In[13]:= bestFit = LinearModelFit[data, x, x]

Out[13]= FittedModel 4.5731 + 1.41792 x

Printed by Wolfram Mathematica Student Edition


PROBLEM 2,3.nb 3

In[14]:= LabeledShowListPlot[data, PlotStyle {Red}],


Plot[bestFit[x], {x, 0, 0.4}, PlotStyle Directive[Blue]], Frame True,
BaseStyle {FontSize 16, FontFamily "Helvetica", FontColor Black},
1 an
ImageSize 550, FrameLabel " ", "rn = ", GridLines Automatic,
n an-1
1
Epilog TextStyle"rn = 4.5731+1.4179 ", Blue, Bold, Directive[14], {0.295, 4.92},
n
1
"Fig: Ratios of adjacent coefficients rn as a function of .",
n
LabelStyle {Directive[16], Italic, FontFamily "Helvetica"}

5.00

4.95
1
rn = 4.5731+1.4179
n
4.90

4.85
an-1
an
rn =

Out[14]= 4.80

4.75

4.70

4.65
0.10 0.15 0.20 0.25 0.30
1
n
1
Fig: Ratios of adjacent coefficients rn as a function of .
n

1 1
According to D Alemberts ratio method, rn = an
= (-1) + . From linear fit, we obtain:
an-1 x0 n x0

rn = an
= 1.41792 1 +4.5731,
an-1 n
1
x0 = = 0.21867
4.5731
and = 1 + x0 *1.41792 = 1.31006.

Problem 5
(a.) The susceptibility power series is given by (x) = n
n=0 an x =
A
(x-x0 )
near x = x0 , where A is a

constant, x0 is the critical point and is the critical exponent. We define the function F(x) to be the logarithmic
derivative of (x),

Printed by Wolfram Mathematica Student Edition


4 PROBLEM 2,3.nb

' -
F(x)= = ,
(x-x0 )
which has a simple pole at the critical point x = x0 with residue - (and no branch points on the complex plane,
unlike (x) which does have a branch point due to the fact that is not an integer). Let the Pade Approxi-
N An xn
mants of F(x) be given by PN (x) = n=0 .
QM (x) 1+Mn=1 Bn x
n

Since F(x) has a simple pole at the critical point, the Pade approximants can be expected to converge rapidly
to F(x). For this reason, the Pade approximants of the logarithmic derivatives can be used to determine the
critical exponent to a very high degree of accuracy.

Near the critical point, i.e. x x0 ,


PN (x) PN (x0 ) and QM (x) QM (x0 ) + QM ' (x0 ) (x - x0 ) = QM ' (x0 ) (x - x0 ),
where QM (x0 ) = 0 because F(x) has a pole at x = x0 and therefore so must PN (x)
QM (x)
,
- PN (x0 )
F(x) = (x-x0 )

QM ' (x0 ) (x-x0 )
PN (x0 )
- QM ' (x0 )

This is how an approximation of the critical exponent can be extracted from the Pade analysis of the logarith-
mic derivative of the susceptibility (all we do is calculate the residue of the Pade Approximant and set it equal
to the residue of F(x) ).
In[15]:=

In[16]:= [x_] := SumAn[[n]] xn-1 , {n, 1, 18}


D[[x], x]
F[x_] :=
[x]

In[18]:= PadeF[x_, N_, M_] := PadeApproximant[F[x], {x, 0, {N, M}}]

(b.) The following function computes the smallest positive pole of the Pade Approximant:
In[19]:= Qroot[N_, M_] := Module[{roots, x0},
roots := x /. {ToRules[NRoots[Denominator[PadeF[x, N, M]] 0, x]]} // Flatten;
x0 = Min[Select[roots, # > 0 &] // Quiet]]

The following function computes from the Pade Approximant:


In[20]:= [N_, M_] := Module{gamma, x0}, x0 := Qroot[N, M];
- Numerator[PadeF[x, N, M]] /. x x0
gamma =
D[Denominator[PadeF[x, N, M]], x] /. x x0

Printed by Wolfram Mathematica Student Edition


PROBLEM 2,3.nb 5

In[21]:= Labeled[
Grid[Insert[Table[{n, 17 - n, Qroot[n, 17 - n], [n, 17 - n]}, {n, 1, 14}], {N, M, x 0 , }, 1],
Frame All, Background {None, {LightGreen}},
BaseStyle {FontSize 16, FontFamily "Helvetica", FontColor Black}],
"Fig: Values of Critial Point and Critical Exponent from Pade' Analysis.",
LabelStyle {Directive[16], Italic, FontFamily "Helvetica"}]

N M x0
1 16 0.219812 1.41762
2 15 0.218137 1.24971
3 14 0.218136 1.24965
4 13 2.3869 10-6 6.6425 10-22
5 12 0.218134 1.24946
6 11 4.1418 10-7 1.32586 10-22
Out[21]=
7 10 2.20771 10-7 5.2615 10-23
8 9 0.218138 1.2498
9 8 0.218135 1.2496
10 7 0.218135 1.2496
11 6 0.218136 1.24963
12 5 0.21814 1.24996
13 4 0.218137 1.24973
14 3 0.218142 1.25011
Fig: Values of Critial Point and Critical Exponent from Pade' Analysis.

The critical point x0 is determined by finding the location of the pole closest to the origin of the Pade Approxi-
mant. From the sequence of Pade approximants generated for M+N =17, x0 and have been calculated and
listed in the above table. It is evident that the critical point is x0 0.218 to 3 decimal places and 1.25 to
two decimal places. The results obtained using Pade analysis are in good agreement with those obtained
using the ratio method.
(For N= 4,6 and 7, there seems to be some discrepancy in the location of the nearest pole leading to nonsensi-
cal results.)
In[22]:=

Printed by Wolfram Mathematica Student Edition

S-ar putea să vă placă și