You are on page 1of 12

Mark Scheme

Final Mark Scheme


1

2625/01

June 2004

(i) Sketch and description of Euler's method


Sketch and description of modified Euler
Sketch and/or description of how the above are used as a predictorcorrecter
pair.
Iteration.

[E1G1]
[E1G1]
[E1E1]

[subtotal 6]
(ii)

ypr

0.1
0.1
0.1
.
0.1
0.1
0.1

x
0
0.1
0.2
.
1.3
1.4
1.5

0
0.0952872
0.1818417
.
1.2295711
1.5137525
1.9931114

0.1
0.1858618
0.2643761
.
1.4533641
1.8592147

yc1
0.0952918
0.1818565
0.2611982
.
1.5016561
1.9568911

yc2
0.0952872
0.1818417
0.2611732
.
1.5116454
1.9848468

yc3
0.0952872
0.1818417
0.2611730
.
1.5137525
1.9931114

0.1
0.05
0.025
0.0125

1.5
1.5
1.5
1.5

1.9931114
diffs
ratio
values
1.9420570 -0.0510543
analysis
1.9295985 -0.0124585 0.2440258
1.9265070 -0.0030914 0.2481367 approx 0.25 so 2nd order

1.5
1.6
1.7
1.8
1.9

y
1.9931114
3.0159137
7.1504452
1233.0004
2.27E+36

[M4]

[A1]
[A2]
[M1A1]

Eg:

(iii)
h

x
0.1
0.1
0.1
0.1
0.1

ypr
2.6112984
4.4914210
15.860621
274885.21
9.84E+71

yc1
yc2
yc3
2.8578100 2.9656660 3.0159137
5.4774949 6.3130524 7.1504452
34.154134 116.49924 1233.0004
71785167 4.89E+18 2.27E+36
9.6E+142 9.4E+284

Solution curve appears to approach an asymptote / gradient tends to infinity /


gradient becomes exteremely large. Step-by-step methods cannot track
such a solution.

[B1]

[E1E1]
[subtotal 3]
[TOTAL 20]

Final Mark Scheme

2 (i) set up RHS as


f(x) = 1:
f(x) = x:
f(x) = x2
f(x) = x3

2625/01

June 2004

a(f(-a) + f(a))
(by symmetry)
2h = 2a
hence a = h
0=0
2h3/3 = 2aa2
hence a = h/sqrt(3)
0 = 0 (not strictly necessary)

[M1A1]
[M1A1]
[A1]
[A1]

G2PR is a two point rule: like trapezium rule in number of function evaluations
but G2PR is accurate up to cubics, like Simpson's rule. However, G2PR
evaluates the function at inconvenient points.

(ii)

h
m
m-h/sqrt3
m+h/sqrt3 f(m-h/sqrt3) f(m+h/sqrt3)
integral
1.0471975 1.5707963 0.9661965 2.1753961 0.9070444 0.9070444 1.8997094

[E1]
[E1]
[E1]
[subtotal 9]
diffs

[M1A1]

0.5235987 1.0471975 0.7448976 1.3494974 0.8233447 0.9877313


0.5235987 2.0943951 1.7920952 2.3966949 0.9877313 0.8233447 1.8965544 -0.0031550

[M2A1]

0.2617993
0.2617993
0.2617993
0.2617993

0.7853981
1.3089969
1.8325957
2.3561944

0.6342482
1.1578469
1.6814457
2.2050445

0.9365481
1.4601468
1.9837456
2.5073444

0.7697870
0.9570481
0.9969376
0.8975062

0.8975062
0.9969376
0.9570481
0.7697870 1.8960972 -4.57154E-

[M1A1]

0.130900
0.130900
0.130900
0.130900
0.130900
0.130900
0.130900
0.130900

0.6544984
0.9162978
1.1780972
1.4398966
1.7016960
1.9634954
2.2252947
2.4870941

0.5789234
0.8407228
1.1025222
1.3643216
1.6261210
1.887920
2.149720
2.4115192

0.7300734
0.9918728
1.2536722
1.5154716
1.7772709
2.039070
2.300870
2.5626691

0.7396777
0.8632064
0.9446420
0.9893228
0.9992346
0.9747492
0.9149055
0.8166543

0.8166543
0.9149055
0.9747492
0.9992346
0.9893228
0.9446420
0.8632064
0.7396777 1.8960540 -4.32215E-

[M1A1]

Differences reducing by a factor of about 10 (theory indicates 16) so 1.89605 is safe.

[M1A1]
[subtotal 11]
[TOTAL 20]

Final Mark Scheme

(i)
4
2
-5

2625/01

matrix
7
0
1
-3.5
9.75

-1
3
6
3.5
4.75
14.5

Determinant = 4 x -3.5 x 14.5

June 2004

identity
1
0
0
1
0
0
-0.5
1
1.25
0
-0.1428571 2.7857142

0
0
1
0
1
1

inverse
0.0147783 0.2118226 -0.1034482

0.1330049 -0.0935960 0.0689655


-0.0098522 0.1921182 0.0689655

-203
setup as M I and attempt to
solve
Gaussian elimination without pivoting

[M1M1]

[M1A1M
1A1M1A
1]
elements of inverse (-1 per error)
[A2]
determinant
[M1A1]
[subtotal 12]

matrix
0.3333333
0.25
0.25
0.2
0.2
0.1666666
0.0125
0.0166666

0.2
0.1666666
0.1428571
0.0166666
0.0228571
6.34921E-

1
0
0
-0.75
-0.6
0.4

Determinant =

matrix
0.33

0.25
0.2

0.25

identity
0
1
0
1
0
-1.3333333

0
0
1
0
1
1

2.64550E-

identity
0.2

0.2
0.17
0
1
0.17
0.14
0
0
0.0106060 0.0184848 -0.7575757
1
0.0184848 0.0187878 -0.6060606
0
-0.0134285 0.7142857 -1.7428571
Determinant =

-0.000047

300

inverse
-900

630

-900
630

2880
-2100

-2100
1575

inverse
determinant

[M1A1]
[A1]

inverse
0

0
1
0
1
1

19.148936 21.27659
5

53.19148
9

21.276595 -131.91489 129.78723


-53.191489 129.78723 -74.468085

inverse
determinant

Problem is highly ill conditioned (or small changes in elements of A lead to large changes
in the inverse).
Determinant is very small in relation to size of elements.

[M1A1]
[A1]

[E1]
[E1]
[subtotal 8]
[TOTAL 20]

Final Mark Scheme

(ii)

(iii)

2625/01

June 2004

(i) By any permitted method:


C1 = -0.75
C2 = 3
C3 = 0.75

[M1]
[A1]
[A1]
[A1]
[subtotal 4]

Cubic splines give a smooth fit to a set of data points without the unwanted
additional turning points often obtained when fitting a polynomial

r
0
1
2
3
4

xr
-2
-1
0
1
2

yr
1
0
-1
2
7

-1
-1
3
5

2
0
4
2
*

Cr
0
-0.75
3
0.75
0

Dr
-0.25
1.25
-0.75
-0.25
*

[E1]
[E1]
[subtotal 2]
Br
-0.75
-1.5
0.75
4.5
*
*[M1M1M1]

Hence f0(x) as required


f1(x) = -1.5 (x + 1) - 0.75 (x + 1)2 + 1.25 (x+1)3

[A1]

f2(x) = -1 + 0.75x + 3x2 - 0.75x3

[A1]

f3(x) = 2 + 4.5 (x - 1) + 0.75 (x-1)2 - 0.25 (x-1)3

[A1]

x
-2
-1.75
-1.5
-1.25
-1
-0.75
-0.5
-0.25
0
0.25
0.5
0.75
1
1.25

spline
1
0.8085937
0.59375
0.3320312
0
-0.4023437
-0.78125
-1.0195312
-1
-0.6367187
0.03125
0.9335937
2
3.1679687

1.5
1.75
2

4.40625
5.6914062
7
data table
graph

[M2A4]
[G2]

Examiners Report

Report on the Units taken in June 2004

2625 Numerical Computation


General Comments
There were only 19 candidates for this paper. The best of them offered nearly perfect
solutions, but a couple showed very few signs of having prepared for the examination. All
candidates showed complete familiarity with Excel; any weaknesses were in the
understanding of the fundamental methods of numerical computation.

Comments on Individual Questions


Q.1

A number of candidates seemed to misunderstand the algorithm in this question.


They understood Eulers method and the modified Euler method, but they did not
manage to combine them correctly into the predictor-corrector method. However, the
numerical differences between the correct solution and that offered were often slight,
so the nature of the solution was not obscured. In part (iii), the solution curve
approaches an asymptote and the numerical method over-shoots it. This point was
often not appreciated by candidates.

Q.2

The derivation of the rule was generally done well. However the numerical
application was often less good. Many candidates had difficulty translating the
formula for an interval (h, h) to an interval (a, b).

Q.3

The inverses were generally found accurately and efficiently. Several candidates,
however, mis-remembered how to calculate a determinant.

Q.4

This question was very well done, with most candidates having no difficulty with the
algorithm, the calculations or the graph.

You might also like