Sie sind auf Seite 1von 70

MA1506 TUTORIAL 1

1. Solve the following differential equations:

(a) x(x + 1)y = 1

(c)

y = e (x3y)

(b)

(sec(x))y = cos(5x)

(d) (1

+ y)y + (1 2x)y 2 = 0

Use www.graphmatica.com to sketch the functions you found as solutions of [a]-[d], if y = 1/2 at x = 1. Graphmatica can also directly sketch the graph if you insert the equation itself; for example, in [a] you just enter x(x + 1)dy = 1 {1, 1/2}. [curly brackets for the initial conditions, with x followed by y.] Here dy represents y . Use this to check that your answers were correct.

2. Experiments show that the rate of change of the temperature of a small iron ball is proportional

to the difference between its temperature T (t) and that of its environment, T env (which is constant). Write down a differential equation describing this situation. Show that T = T env is a solution.

Does this make sense? The ball is heated to 300 F and then left to cool in a room at 75 F . Its temperature falls to 200 F in half an hour. Show that its temperature will be 81.6 F after 3 hours of cooling.

3. In very dry regions, the phenomenon called Virga is very important because it can endanger

aeroplanes. [See http://en.wikipedia.org/wiki/Virga ]. Virga is rain in air that is so dry that the raindrops evaporate before they can reach the ground. Suppose that the volume of a raindrop is proportional to the 3/2 power of its surface area. [Why is this reasonable? Note: raindrops are not spherical, but let’s assume that they always have the same shape, no matter what their size may be.] Suppose that the rate of reduction of the volume of a raindrop is proportional to its surface area. [Why is this reasonable?] Find a formula for the amount of time it takes for a virga raindrop to evaporate completely, expressed in terms of the constants you introduced and the initial surface area of a raindrop. Check that the units of your formula are correct. Suppose somebody suggests that the rate of reduction of the volume of a raindrop is proportional to the square of the surface area. Argue that this cannot be correct.

4. Moths navigate at night by keeping a fixed angle between their velocity vector and the direction

of the Moon [or some bright star; see http://en.wikipedia.org/wiki/Moth]. A certain moth flies near to a candle and mistakes it for the Moon. What will happen to the moth? Hints: in polar coordinates (r, θ), the formula for the angle ψ between the radius vector and the velocity vector is given by tan(ψ)=r dθ dr . [If you want to derive this formula, remember that the tangential component of a small displacement in polar coordinates (r,θ)(r + dr, θ + dθ) is rdθ, and the radial component is just dr. Now use simple trigonometry.] Use the formula to solve for r as a function of θ. Note that graphmatica can graph in polar coordinates if you spell out θ [as theta!]

5. In Cosmology, the ratio of the sizes of the Universe at two different times is measured by a

function of time called the “scale function”, denoted a(t). What are the units of a(t)? The [first]

1

Friedmann Equation [http://en.wikipedia.org/wiki/Friedmann equations] relates this function to the energy density of the Universe and to its spatial curvature. In a particular cosmological model, the Friedmann equation takes the form L 2 a˙ 2 = a 2 2 2 + 1, where L is a positive constant, the dot denotes time differentiation, and the initial condition is a(0) = 1. What are the units of L? Show, without solving this equation, that the universe described by this model is never smaller than a certain minimum size. Now solve the equation and describe the history of this universe.

a

6. Solve the following equations:

(a)

(c)

y = 12y4x

1+y+2x

(b) y = x+y+1

x+y+3 2

x + y +1+(x + y 3)y = 0

2

MA1506 Tutorial 1 Solutions

(1a)

y '

1

1

1

x y ln x 1 cos 4 x
x
y ln
x 1
cos 4
x

ln

1 3 y dx e 3 1 2 y x y
1 3
y
dx
e
3
1
2
y
x
y

c

y

e

x

1

2

c

x

c

x

(

x

1)

x

x 1

cos

e

x

x

e

x

3

e

1

2

cos 6

x

y

dy

e

x

(1b)

'

1

6

sin 6

x

1

4

sin 4

x

y

cos 5

c

(1c)

dy

dx

(1d)

1 y

3

y

2

y

(2)

dT

dy

(2

x

1)

dx

k(T

Tenv)

where k is a positive constant. (If k were a negative, then hot

dTenv

dt

0

dt objects would get hotter when left to “cool”. That doesn’t happen – if it did, we would

not be here to discuss it.) T = Tenv is obviously a solution since

does make sense because objects do not spontaneously become hotter or colder.

and this

Having settled this case, we can assume T Tenv and so we can write dT

 
 

kdt

so

ln

T

Tenv

kt

c

.

T

Tenv

In the case at hand, T >Tenv so

T

Tenv

T

Tenv

and so T

Tenv

e

At t = 0, T = 300 so 300 75 225 . At

t

1 , T = 200, so
2

kt

.

200

75

225

e

k 2
k
2

k 2 ln

125

1.1756

 

225

Thus

T

(3)

75

225

e

3

(3)

k

81.6 .

The volume V is related to the area A by V = a A (3/2) where a is a positive constant with no units; this is reasonable because volume has units of cubic metres and area has units of square metres. [Of course, ``reasonable” doesn’t mean that it’s always exactly true.] Then dV/dt = (3a/2)(dA/dt) A (1/2) The question tells us that dV/dt = - bA, where b is a positive constant with units of metres/sec. This is reasonable because evaporation takes place at the surface of the drop and so its rate can be expected to depend on the area. So we have dA/ A (1/2) =

- 2bdt/3a. Integrating from A 0 , the initial area, up to zero, we find that the time taken for complete evaporation is 3aA 0 (1/2) /b which does indeed have units of time since the numerator has units of metres while the denominator has units of metres/sec.

If, instead of dV/dt = - bA, we propose that dV/dt = - bA 2 , then we would have obtained dA/ A (3/2) = - 2bdt/3a . When you try to integrate this from A 0 to zero, you will get a divergent integral, meaning that the evaporation would take infinite time and the rain would always reach the ground, contrary to the definition of Virga.

(4)

The moth flies in such a way that the angle remains constant at all times, so we have a differential equation tan ( ) = constant = rd /dr, hence dr/r = d / tan ( ) , thus r = R exp( / tan ( )), where we take it that = 0 when the moth first sees the candle, and that her distance from the candle is R at that time. Remember that is the angle between the radius vector of the moth [pointing outwards] and her velocity. From the point of view of the moth looking towards a candle in front of her, would be an angle greater than 90 degrees. Draw a diagram if this is not obvious! Thus tan ( ) will be negative and r will get steadily smaller as increases. Such a curve is called a spiral. So the unfortunate moth will spiral into the candle with tragic consequences. Of course if her first view of the candle is over her ``shoulder” then tan ( ) will be positive and she will spiral outwards, something that would be a lot less noticeable. Finally if is 90 degrees exactly, the moth will fly along a circle until it drops dead from exhaustion or starvation, whichever comes first.

(5)

Since it is a ratio, a(t) has no units. The units of L are those of time, so that both sides of the given equation have the same units [ie, none]. If a(t) were ever smaller than 1, then 2/a >2, yet 1 + a 2 < 2, so the right side of the equation would become negative, while the left side is never negative. Hence a(t) can never be less than 1. Next, separate the variables in the obvious way and change the variable a to y = a 2 and you dy

will get

2

dt

/

L

y

2

y

2

y = a 2 and you dy will get 2 dt / L y 2 y

. Complete the square and make the substitution

y

1

2

3

2

cosh( )

x

[We know we can do this, because y is never less than 1 so the left

side is never less than 3/2. Remember that cosh is never less than 1.] Do the integral

and you get a

be. The graph of a starts out at a = 1, and then steadily increases at a faster and faster

rate. This describes a Universe that begins with non-zero size and then expands. The expansion is not slowing down --- instead it is getting faster and faster.

2 1

3

cosh(2 /

t

L

)

. The right side is equal to 1 at t = 0, as it should

2 2

(6)

These are examples of ODEs where a change of variable is needed.

(6a)

Let v 2x y

so v' 2 y'

v '

2

1

2 v

1 v

v

(2

x

1

2

v

2

y

)

3

1

x

c

2

(2

x

(6b)

v x y

v

'

3

1 v

y

)

y

2

'

3

x

c

v

'

1

v 1 v 3

2

v '

1

v

2

2

v

1

2

v

2

8

v

10

(

v

3)

2

(

v

3)

2

(

v

3)

2

v

v

2

2

4

4

v

v

5

5

2

v

4

2

v

4 v

5

dv

2

dx

dv

2

dx

v

x

2

v

4

dv

1

v

ln

2

2

v

4

v

4

v

5

y

5

2

4

2 dx

2 x

x

c

4

y

y

ln

x

5

2 x

c

(6c)

Here we need 2 changes of variables: first set x X , y Y so

x y 1 X Y ( 1) , x y 3 X Y ( 3) so if we choose

1 3 0 i.e. 2 , 1 then (i.e. Y y 1,

and this is homogeneous so set

X Y

X

x 2 )

Y

)

dY

dX

(

X

0

V

0

Y X 1 V
Y
X
1
V

(

Y

'

1

V

V

)(

XV

V

'

XV ')

dX 1 V

X 1 V

2

dV

ln

ln

since

X

arctan V

1 ln 2
1
ln
2

1 V

2

c

x 2

arctan

y 1

x 2

V

Y

X

y

1

x 2

x 2

1 ln 1
2

y 1

x 2

2

c

MA1506 TUTORIAL 2

1. Solve the following differential equations:

(a) xy + (1 + x)y = e x ,

x> 0

(b) y (1 + x )y = x + 2,

(c) y + y + x = 0

3

y

y(1) = e 1,

x> 0

(d) 2xyy + (x 1)y 2 = x 2 e x ,

x> 0

2. If a cable is held up at two ends at the same height, then it will sag in the middle, making a

U-shaped curve called a catenary. This is the shape seen in electricity cables suspended between poles, in countries less advanced than Singapore, such as Japan and the US. It can be shown using

simple physics that if the shape is given by a function y(x), then this function satisfies

dy

dx =

μ

T x

0

dy dt

2

+ 1 dt,

where x = 0 at the lowest point of the catenary and y(0) = 0, where μ is the weight per unit length of the cable, and where T is the horizontal component of its tension; this horizontal component is a constant along the cable. Find a formula for the shape of the cable. [Hint: use the Fundamental Theorem of Calculus, and think of the resulting equation as a first-order ODE.]

3. Psychologists talk about something called a Performance Curve. Suppose an MA1506

student is solving mathematics problems. She starts with differential equations. Let P(t) be a non-negative function that measures her performance, that is, her success rate at solving DEs. Her performance increases rapidly at first, but then the rate of increase slows down as she becomes more expert. Let M, a positive constant, be the best possible performance; then one can suppose that P satisfies

dP

= C[M P],

dt where C is a constant. What are the units of this constant? What does this constant measure? Solve this equation assuming that she is completely incompetent at t = 0 [that is, P(0) = 0]. Now the student turns to another kind of problem, say in linear algebra. Again her per- formance is low at first but gets better in accordance with this equation. Now as the years go by, her overall ability to solve mathematics problems gradually gets better, so C, instead of being a constant, is really a slowly increasing function of time. Suppose that C(t) = K tanh(t/T), t 0, where K and T are positive constants. [Is this reasonable? Why? What are the meanings of the constants K and T?] Replacing C with C(t), solve for P, again assuming that P(0) = 0. Sketch P(t), the performance curve. [Choose values of the constants for yourself, and use Graphmatica [www.graphmatica.com] if necessary.]

4. A certain MA1506 student starts a rumour to the effect that one of the lecturers has been seen

dating Tang Wei. The number of students who have heard the rumour, R(t), is given by

dR

dt

= KR[1500 R],

1

where K is a positive constant, and 1500 is the number of students taking MA1506. What is the meaning of K? Is this equation reasonable? [Hint: surely the rumour will spread slowly both when hardly anyone has heard it yet, but also when nearly everyone has already heard it!] By regarding this equation as a Bernoulli equation, find R(t). Note that, in reality, R(t) is actually an integer [whole number]. Comment on this fact in relation to your solution.

2

MA1506 Tutorial 2 Solutions

Question 1.

(1a)

y '

(1

1

) y

1 e

x

x x

Integrating factor is

by exp

P

'

y e

P

exp

Pye

1

(1

x

dx

(

)

d

P

exp(

x

ye

P

)

)

d

dx

yxe

(

x

yxe

x

x

)

c

xe

x

 

1

e

x

)

 

x

y

e

x

cx

x

3ln

x

)

(

x

2)

 

1

3

e

x

 

x

 

e

x

x

3

c

 
 

x

e

c

 

1(

exp(

1

x

e

1 x

x

3

e

So

(1b)

 

exp

(1

3

)

 

x

d

dx

(

y

1

x

3

e

x

y

 

x

e

x

e

 

x

3

2

x

 
 

x

x

 

e

 

e

2

2

x

 

x

 

x

 
 

e

 

c

 
 

2

x

 

y

x

 

3

cx e

x

)

2

since

2

x

y(1) e 1 1 ce c 1

 

3

x

y

x

x e

ln

x

)

(1c)

This kind is called a Bernoulli equation -- set

2

z y

z

'

2

y

ze

y

2

x

z' 2 yy'

x

y

(

x

0

1

2

)

1

2

e

2

z

x

'

z

c

z ' y ' 2 y x 0 z ' 2 z 2 1 2
z '
y '
2 y
x
0
z
'
2
z
2
1 2
2
x
y
x
ce
2

xe

x

d

dx

(in general, y' P(x) y multiply

x

(

e

2

x

z

)

2

xe

2

x

(1d)

Since 2

'

yy

(

2

y

dx d

1

x

x

e Y

2

y

1

xe

x

2

)' we define

Y y

2

e

2

x

1

x

e Y

1

e

2

x 2

cxe

x

,

x

Y

'

c

(1

1

x

)

Y

x

xe

,

Question 2. Define v = dy/dx. Then our equation is

x 2 v v 1 dt . T 0
x
2
v
v
1
dt
.
T
0

exp (1

1

x

)

1

x

e

x

Notice that v(0) = 0. Now by the fundamental theorem of calculus

v

'

2 v 1
2
v
1

T

This is a separable differential equation with initial condition v(0) = 0. Separating the

variables and integrating [remember that Integrating, we get

cosh

2

sinh

2

1

] we find that v = sinh( x/T).

y

T

cosh(

T

x

)

C

,

where C must be – T/ since y(0) = 0 and cosh(0) = 1. Thus we have

y

T

cosh(

T

[Draw it!]

Question 3.

x

)

T

. The shape of the graph of this function is indeed “U-shaped”.

 

The constant C has units of 1/time. Solving the equation [either as linear first order or as a separable equation] we get

P

M

Me

Ct

Clearly P will approach M more rapidly if C is large; that is, C measures how rapidly the student is able to learn. Thus the equation indeed expresses the idea that the student’s

performance improves more slowly as she approaches her maximum possible performance.

As the years go by and the student becomes more familiar with the methods of mastering mathematics, her rate of learning new things might be expected to improve; but surely there is an upper bound to how much she can improve. The tanh function is a simple way of representing this since it always increases but is bounded above. [Remind the students of the shape of tanh(x) if necessary.] Then K represents her maximum possible speed of learning [since tanh is asymptotic to 1], and T measures the amount of time required for her to realise her maximum potential. [Note that K has units of 1/time, while of course T has units of time, so KT is dimensionless.] The equation can now be written as

dP

dt

K tanh(t / T )(P M ) 0

.

It’s convenient now to define Q = P – M, so the equation is

dQ

dt

K tanh(t / T )Q 0.

This is a first-order equation with integrating factor cosh

Q cosh

KT

(t / T )

= A,

KT

(t / T ) , so we have

where A is a constant. Since we are assuming that P(0) = 0, we have Q(0) = - M, so we have A = - M, thus finally

P

M 1

sech

KT

(t

/ T )

.

Students are encouraged to graph examples of such questions using [for example] the software available free at http://www.graphmatica.com/

Question 4.

The constant K measures the rapidity with which the rumour will spread. It depends on how interesting the rumour is, how much the students like to gossip, how gullible they are, etc. The right hand side of the equation is designed to be small both near R = 1 and near R = 1500, when indeed the rumour can be expected to spread slowly either because not enough or too many people have heard it.

We have

dR

dt

1500KR

KR

2

.

This is a Bernoulli equation as discussed in the notes. We solve it by defining Z = 1/R, which transforms the equation into a linear one:

dZ 1500 dt

KZ

with solution

1

1

R

1500

C

K

,

exp( 1500

Kt

) .

The problem says that this highly interesting rumour was started by one student, so R(0) = 1. Thus C = 1499/1500. Hence

1

1

1499

1500

1500

R

exp( 1500

Kt

)

.

Of course as t tends to infinity, R tends to 1500. The function R(t) is not really continuous or differentiable, since it only takes integer values. However, this integer- valued function can be well approximated by a smooth function defined to interpolate between integral values.

OPTIONAL NOTE ON THE EQUATION IN QUESTION 2:

Consider a small element of the cable, with the left end located at a point (x,y), and the right at (x+dx,y+dy). Let tan( ) = dy/dx, and let U(x) denote the tension in the cable. The horizontal forces acting are just Ucos( ) and Ucos( ) +d(Ucos( ) ) at the two ends of the small element, so if these forces are to balance we must have Ucos( ) = constant = T. The vertical forces are Usin( ) and Usin( ) +d(Usin( ) ) and also ds, where ds is the length of the element. For these to balance we must

have d(Usin( ) ) =

0 [where dy/dx = 0] we obtain the stated equation.

ds, or d(Ttan( ) ) =

ds. Integrating this equation from x =

MA1506 TUTORIAL 3

1. Solve the following differential equations:

(a) y + 6y + 9y = 0,

(b) y 2y + (1 + 4π 2 )y = 0,

y(0) = 1,

y (0) = 1

y(0) = 2,

y (0) = 2(3π 1)

2. Find particular solutions of the following:

(a)

(c)

y + 2y + 10y = 25x 2 + 3

y y = 2xsin(x)

(b)

y 6y + 8y = x 2 e 3x

(d) y + 4y = sin 2 (x)

3. Use the method of variation of parameters to find particular solutions of

(a)

y + 4y = sin 2 (x)

(b) y + y = sec(x)

4. One of the most important kinds of nonlinear second-order odes takes the form y = F (y).

[That is, y can be expressed in terms of y only, with no y .] These can be solved with the aid of

the following clever trick. Use the chain rule to prove that

d 2 y dx 2 =

d

dy [ 1 2 y 2 ].

Use this trick to show that any equation of the form y = F (y) can be reduced to a separable first-order ode — and therefore solved, because we know that we can always solve this kind of equation. Suppose that the Earth were to stop moving tomorrow. It would of course immediately begin to fall towards the Sun. How long would we have before reaching the orbit of Venus [by which time we would all have been fried]? Information: the radius of the earth’s orbit is about 150 billion m, that of Venus’ orbit is about 100 billion m [both orbits being approximately circular], the acceleration due to gravity at a distance of r from a central object of mass M is GM/r 2 , where M [the mass of the Sun] is about 2 × 10 30 kg, and G is Newton’s constant = 6.67 × 10 11 in MKS units. So you have to solve the equation r¨ = GM/r 2 . Use the above clever trick and solve the resulting first-order separable equation. You will get a nasty definite integral [which can actually be done]; go to http://wims.unice.fr/wims/ if you would rather let a computer do it. [The computer may not want to do it either [why?]; you may have to adjust your domain of integration very slightly in order to persuade it to cooperate.]

5. Sometimes people wonder why Newton’s second law has a second derivative — why not a

third derivative? Let’s investigate. The method we used to solve second-order linear ODEs with constant coefficients also works for higher-order equations. Find the general solution of the “generalized simple harmonic motion

equation”

d 3 y

dt 3

=

1

w 3 y,

where w is a positive constant [with what units?]. [Hint: there are three different numbers x such that x 3 = -1; two of them are complex. The general solution should therefore contain three terms and have three arbitrary constants.] This is what simple harmonic motion would look like if Newton’s Second Law involved third derivatives instead of second derivatives. Use matlab or graphmatica to graph the solution of

d 3 y

dt 3

= 8y,

where y(0) = 0.0001, y˙(0) = 0, y¨(0) = 0. [Hint: d3y = -8y {0,0.0001,0,0}. You may have to zoom in and out a few times to get graphmatica to work.]

2

MA1506 Tutorial 3 Solutions

(1a)

y

2

y

y

y

(0)

"

6

' 0

y

6

(

1

'

9

A

y

9

0

0

)

B

x

3

y

1

3

'

Bx e

A

1

1

3

A

Be

B

3

Set y

x

3(

A

2

y

e

t

)

Bx e

(1

(1b)

2

y

2

(1

4

cos 2

e

x

A

[

x [

2

A

)

2

0

B

1

]

2

i

]

sin 2

x

sin 2

x

y

'

y

e

(0)

y

y

'(0)

B

2

2(3

3

y

A

2

1)

e

x

[

x

2

B

cos 2

x

2(3

2 cos 2

1)

x

2

3sin 2

2

B

]

x

(2a)

Try y Ax

10

2 ( 2

3

x

"

y

2

A

'

y

2

25

10

A

A

2

y

Bx C

Ax

B

)

10

(

Ax

2

25

5

5

2

/

2

,

x

2

,

B

4

A

-1,

x

10

B

C

0

0

2

y

,

Bx

2

A

C

)

2

B

10

C

3

x

2

x e

)

3

x

3

(2b)

Try y

2

C e

(

Ax

Ax

Bx

x

3(

)

2

y

'

y

9

6

2

y

"

A

A

A

(

C e

x

Ax

)

B e

A

18

3

Bx

3(2

1

Ae

18

A

6

3

A

B

x

2

8

A

3

3(2

Ax

1

A

12

9

C

6

e

3 x

B

B

18

8

C

B

8

9

B

B

2)

3

0

C

)

B e

3

B

0

(2

)

B e

3

)

Ax

3

x

2

3

x

x

x

9(

Ax

0

C

2

2

Bx

)

C e

3

x

(2c)

"

y

if we can solve the complex equation z"-z

Try z

z'

z"

"

z

y

2

x

Im

e

ix

(Ax

ix

(Im

imaginary part)

B)e

ix

Ae

i(Ax

Aie

ix

z

(2

Ai

iAe

ix

Ax

B)e

(

B

ix

Ax

B

Ax

)e

ix

B

)e

ix

2

x

e

ix

2xe

ix

2

A

i

1

2

B

0

z

(

x

i

Im

z

cos

x

)e

ix

x

(2d)

"

4

y

1

2

(1

y

cos 2

1

2

2

iBe

Solve z" 4z

z"

-4Bx

4 A

e

1

2

ix

4

A

1

2

x

)

i

cos

sin

x

y

x

cos

x

[

i

Re(

e

2 ix

A

)

z

"

Try z

4

z

4

cos

x

sin

x

x

x

2

e

ix

4

A

Bx

Bxe

2

ix

ix

1

2

1

2

B

x

sin

x

1 2

e

ix

1

2 8

4 iB

B

1

i

2 8

z

y

1

1

ixe

8

Re z

8

1

8

2

ix

1

8

1

8

(1

x

sin 2

( cos 2

x i

x

x

sin 2

x

))

then Imz satisfies the above.

sin

4

x

]

Bxe

2

ix

4

iBe

2

ix

(3a)

Variation of parameters : first solve y" 4y

Promote A and B to functions A(x), B(x).

0

y

Acos2x

Then A(x)cos(2x)

if A(x) and B(x) are chosen to satisfy

B(x)sin2x is

a

solution of y" 4y

1

2

(1

A'

-[

1

2

(

RHS

)

(1

cos 2 x )]sin 2

x

W

,

B'

[

1

2

(

RHS

)

(1

cos 2 x )]cos 2

x

W

Bsin2x

cos 2

x

)

where W

so

1

A'

(cos2x)x(sin2x)'-(cos2x)'sin2x

1

cos

2

x

sin

2

x

1

sin

2

x

sin 2 x

2

1

sin 4

B

'

1

4

4

cos

2 x

4

1

4

cos

2

(

2

x

)

1

4

4

cos 2

x

1

8

(cos

8

4

x

1

A

1

cos

2

x

1

cos 4

x

8 32

x

)

B

1

sin

2

x

1

sin 4 x

x

so the solution is

8

32

Acos 2x

Bsin2x

1

8

8

2 x

1

4

[(cos

cos

4

)cos

x

1

4

sin

4

2

x

x

sin

(sin

2

x

)

1

32

cos 2

x

which is the same as in (2d) since

1

8

(cos

2

1

8

(

1

2 x