1Subspaces and affine sets of functions
Work in the vector space ๐ถ1(โ). For each subset below, decide
whether it is a vector subspace, an affine set that is not a vector
subspace, or neither. Justify your answers using the definitions; when
closure fails, give explicit functions showing the failure. For an
affine set that is not a subspace, write it as ๐๐ +๐ and identify
both the particular function ๐๐ and the vector subspace ๐.
(a){๐ โ๐ถ1(โ) :๐(0) =0}.
Solution
This is a vector subspace. The zero function has value 0 at ๐ก =0,
so it belongs to the set. If ๐(0) =0 and ๐(0) =0, then for any real
numbers ๐,๐,
(๐๐+๐๐)(0)=๐๐(0)+๐๐(0)=๐โ
0+๐โ
0=0,
so every linear combination stays in the set.
(b){๐ โ๐ถ1(โ) :๐(0) =1}.
Solution
This is affine but not a vector subspace. The zero function has value
0 at zero, not 1, so it is not in the set. Addition fails too: the
constant function 1 belongs to the set, but 1 +1 =2 has value 2 at
zero.
Let ๐ ={๐ โ๐ถ1(โ) :๐(0) =0}, the subspace from the previous
part, and let ๐๐ be the constant function 1. If ๐(0) =1, then
๐ โ1 has value 0 at zero, so ๐ =1 +(๐ โ1) lies in 1 +๐. Conversely,
if โ โ๐, then (1 +โ)(0) =1 +0 =1. Thus the set is exactly 1 +๐.
(c){๐ โ๐ถ1(โ) :๐(0)๐(1) =0}.
Solution
This is neither. The set does contain the zero function, and it is closed
under scaling, since (๐๐)(0)(๐๐)(1) =๐2๐(0)๐(1) =0. Addition is what
fails. The function ๐(๐ก) =๐ก belongs to the set because ๐(0) =0, and
๐(๐ก) =1 โ๐ก belongs because ๐(1) =0. But their sum is the constant
function 1, and 1 โ
1 =1 โ 0.
So the set is not a subspace. It cannot be affine either: an affine set
that contains zero is automatically a vector subspace, and we just showed
this one is not. (Geometrically, the set is the union of two subspaces:
functions vanishing at 0 and functions vanishing at 1. A union of two
planes is not a plane.)
2Linear maps on polynomials
Let ๐2 be the space of real polynomials of degree at most two, with
ordered basis (1,๐ก,๐ก2). Consider the map
๐:๐2โถโ2,๐(๐)=(๐(0),๐โฒ(0)).
(a)Show that ๐ is linear. Find its matrix using the given basis of ๐2
and the standard basis of โ2.
Solution
Evaluation and differentiation both respect linear combinations:
๐(๐๐+๐๐)=(๐๐(0)+๐๐(0),ย ๐๐โฒ(0)+๐๐โฒ(0))=๐๐(๐)+๐๐(๐).
For the matrix, apply ๐ to each basis polynomial:
๐(1)=(1,0),๐(๐ก)=(0,1),๐(๐ก2)=(0,0).
These are the columns:
(100010).
As a check, ๐ =๐ +๐๐ก +๐๐ก2 has coordinates (๐,๐,๐), and the matrix
sends these to (๐,๐). Indeed ๐(0) =๐ and ๐โฒ(0) =๐.
(b)Find the kernel and image of ๐, giving a basis and dimension for each.
Is ๐ an isomorphism? Explain what information about a polynomial is
lost when we record only its value and first derivative at zero.
Solution
For ๐ =๐ +๐๐ก +๐๐ก2 we have ๐(๐) =(๐,๐). This is zero exactly when
๐ =๐ =0, so the kernel consists of the polynomials ๐๐ก2. It has basis
๐ก2 and dimension 1.
Every (๐,๐ ) โโ2 is hit, for example by ๐(๐ +๐ ๐ก) =(๐,๐ ). So
the image is all of โ2, with basis (1,0),(0,1) and
dimension 2. The dimensions check out: 3 =1 +2.
๐ is not an isomorphism, because it is not injective: ๐(๐ก2) =(0,0) =๐(0).
The lost information is the coefficient of ๐ก2. The value and slope at
zero tell us the tangent line to the graph there, but not how the graph
bends away from it. The polynomials 0 and ๐ก2 have the same value and
slope at zero, yet they are different.
(c)Now record one more derivative:
๐ฝ:๐2โถโ3,๐ฝ(๐)=(๐(0),๐โฒ(0),๐โณ(0)).
Find the matrix of ๐ฝ in the same polynomial basis and the standard
basis of โ3. Is this map an isomorphism? Justify your answer,
and explain how to recover the polynomial from these three numbers.
Solution
๐ฝ is linear for the same reason ๐ is. On the basis,
๐ฝ(1)=(1,0,0),๐ฝ(๐ก)=(0,1,0),๐ฝ(๐ก2)=(0,0,2),
so its matrix is
โโ
โ
โโ100010002โโ
โ
โโ .
This square matrix has determinant 2 โ 0, so it is invertible and ๐ฝ
is an isomorphism. To recover ๐ =๐ +๐๐ก +๐๐ก2, notice that ๐(0) =๐,
๐โฒ(0) =๐, and ๐โณ(0) =2๐. Therefore
๐(๐ก)=๐(0)+๐โฒ(0)๐ก+๐โณ(0)2๐ก2.
This is the Taylor polynomial of ๐ at zero. The second derivative
supplied exactly the missing bending information from the previous part.
3Linear, homogeneous, or neither?
For each equation below, decide whether it is linear. If it is linear, say
whether it is homogeneous or forced, and identify its coefficient functions.
Write each linear equation as ๐ฟ[๐ข] =๐, giving the domain and codomain of
the differential operator ๐ฟ. Use a suitable space ๐ถ๐(๐ผ), where
๐ผ is an open interval.
If it is nonlinear, point to a term which makes it so.
(a)๐ฆโณ +๐ก2๐ฆโฒ โ(sinโก๐ก) ๐ฆ =0.
Solution
Linear and homogeneous. The coefficients are ๐2(๐ก) =1, ๐1(๐ก) =๐ก2,
and ๐0(๐ก) = โsinโก๐ก, and the forcing is ๐ =0. The coefficients vary
with time, but the unknown ๐ฆ and its derivatives appear only to the
first power, multiplied by known functions. All the coefficients are
continuous on โ, so we can take ๐ผ =โ:
๐ฟ:๐ถ2(โ)โถ๐ถ(โ),๐ฟ[๐ฆ]=๐ฆโณ+๐ก2๐ฆโฒโ(sinโก๐ก)๐ฆ,
and the equation is ๐ฟ[๐ฆ] =0.
(b)๐ฆโฒ +3๐ฆ =๐โ๐ก.
Solution
Linear and forced. The coefficients are ๐1(๐ก) =1 and ๐0(๐ก) =3, and
the forcing is ๐(๐ก) =๐โ๐ก. With ๐ผ =โ,
๐ฟ:๐ถ1(โ)โถ๐ถ(โ),๐ฟ[๐ฆ]=๐ฆโฒ+3๐ฆ,
and the equation is ๐ฟ[๐ฆ] =๐โ๐ก.
(c)๐ฆโณ +sinโก๐ฆ =0.
Solution
Nonlinear, because of the term sinโก๐ฆ. The sine is applied to the
unknown itself, and sine does not respect addition or scaling. For
example, the operator ๐[๐ฆ] =๐ฆโณ +sinโก๐ฆ sends the constant function
๐/2 to sinโก(๐/2) =1, but sends twice that function, ๐, to
sinโก๐ =0 rather than 2. (This is the pendulum equation, which we
meet again at the end of the assignment.)
4Solution sets inside function spaces
Now consider the whole set of solutions, with no initial condition imposed.
In each part, work with functions defined on all of โ.
(a)For ๐ขโณ +4๐ข =0, define
๐ฟ:๐ถ2(โ)โถ๐ถ(โ),๐ฟ[๐ข]=๐ขโณ+4๐ข.
Verify that ๐ฟ is linear. Explain why the solution set is kerโก๐ฟ,
and prove that it is a vector subspace of ๐ถ2(โ). What is
the zero vector in this space? Your argument should not require a formula
for the general solution.
Solution
Differentiation and multiplication by 4 both respect linear
combinations, so
๐ฟ[๐๐ข+๐๐ฃ]=(๐๐ข+๐๐ฃ)โณ+4(๐๐ข+๐๐ฃ)=๐(๐ขโณ+4๐ข)+๐(๐ฃโณ+4๐ฃ)=๐๐ฟ[๐ข]+๐๐ฟ[๐ฃ].
A function ๐ข โ๐ถ2(โ) solves the equation exactly when
๐ฟ[๐ข] =0, and that is the definition of the kernel. So the solution set
is kerโก๐ฟ.
The kernel is a subspace. It contains zero, since ๐ฟ[0] =0. And if
๐ฟ[๐ข] =๐ฟ[๐ฃ] =0, then linearity gives
๐ฟ[๐๐ข+๐๐ฃ]=๐๐ฟ[๐ข]+๐๐ฟ[๐ฃ]=0,
so ๐๐ข +๐๐ฃ is also a solution. The zero vector is the zero function,
๐ข(๐ก) =0 for every ๐ก. For a spring, this is the solution that sits at
rest at equilibrium forever.
(b)For ๐ขโฒ +๐ข =1, define ๐ฟ[๐ข] =๐ขโฒ +๐ข on ๐ถ1(โ) and check
that ๐ฟ is linear. If ๐ข and ๐ฃ are solutions, compute ๐ฟ[๐ข +๐ฃ]
and ๐ฟ[๐ข โ๐ฃ]. Explain why the solution set is not a vector subspace,
even though the equation is linear.
Find a particular solution ๐ข๐ and show that the full solution set is
๐ข๐ +kerโก๐ฟ: prove both that every solution has this form and that
every function of this form is a solution. Describe this affine set by
identifying its translation and its space of directions.
Solution
Linearity works just as before:
๐ฟ[๐๐ข +๐๐ฃ] =(๐๐ข +๐๐ฃ)โฒ +(๐๐ข +๐๐ฃ) =๐๐ฟ[๐ข] +๐๐ฟ[๐ฃ].
If ๐ฟ[๐ข] =1 and ๐ฟ[๐ฃ] =1, then
๐ฟ[๐ข+๐ฃ]=1+1=2,๐ฟ[๐ขโ๐ฃ]=1โ1=0.
So the sum of two solutions is not a solution, while their difference
solves the homogeneous equation โโฒ +โ =0. The solution set also fails to
contain zero, since ๐ฟ[0] =0 โ 1. The operator ๐ฟ is linear, but the
right-hand side 1 is not zero, and that is what keeps the solution set
from being a subspace.
The constant function ๐ข๐ =1 is a particular solution: 0 +1 =1.
Every solution has this form. If ๐ฟ[๐ข] =1, then
๐ฟ[๐ข โ1] =๐ฟ[๐ข] โ๐ฟ[1] =1 โ1 =0, so โ =๐ข โ1 is in kerโก๐ฟ and ๐ข =1 +โ.
Every such function is a solution. If โ โkerโก๐ฟ, then
๐ฟ[1 +โ] =๐ฟ[1] +๐ฟ[โ] =1 +0 =1.
So the solution set is 1 +kerโก๐ฟ. We can also find the kernel exactly.
If โโฒ +โ =0, multiplying by ๐๐ก gives (๐๐กโ)โฒ =๐๐ก(โโฒ +โ) =0, so
๐๐กโ is a constant ๐ถ and โ =๐ถ๐โ๐ก. Thus every solution is
๐ข(๐ก)=1+๐ถ๐โ๐ก.
This is a line in function space. It is translated to pass through the
constant function 1, and its direction is the one-dimensional space
spanned by ๐โ๐ก.
(c)For ๐ขโฒ =๐ข(1โ๐ข10), compare (2๐ข)โฒ with
the right-hand side evaluated at 2๐ข. Show that the only solution ๐ข
for which 2๐ข is also a solution is ๐ข =0. Use a nonzero equilibrium
to conclude that the solution set is not a vector subspace of
๐ถ1(โ). Since it contains the zero function, explain why it
cannot be an affine translate of a vector subspace either.
Solution
Let ๐ข be a solution. Then
(2๐ข)โฒ=2๐ขโฒ=2๐ข(1โ๐ข10),
while the right-hand side evaluated at 2๐ข is
2๐ข(1โ2๐ข10).
For 2๐ข to be a solution, these must agree at every time. Their
difference is
2๐ข(1โ๐ข10)โ2๐ข(1โ2๐ข10)=๐ข25,
which is zero only when ๐ข(๐ก) =0. So 2๐ข is a solution only when ๐ข is
the zero function, and doubling the zero function does give a solution.
For a concrete failure, take the equilibrium ๐ข =10. It is a solution,
since both sides are 0. But 2๐ข =20 has derivative 0, while the
right-hand side is 20(1 โ2) = โ20. The solution set is not closed under
scaling, so it is not a vector subspace.
The zero function is a solution: 0โฒ =0 =0 โ
(1 โ0). An affine set that
contains zero is automatically a vector subspace. Indeed, if
0 =๐๐ +๐ฃ for some ๐ฃ โ๐, then ๐๐ = โ๐ฃ already lies in ๐, and
๐๐ +๐ =๐. Since the solution set contains zero but is not a subspace,
it cannot be affine either. (Checking a midpoint shows the same thing:
halfway between the solutions 0 and 10 is the constant 5, and
5(1 โ12) โ 0.)
5A basis of spring solutions
Let ๐ด and ๐ต be the solutions of
๐ฅโณ+4๐ฅ=0
with initial data (๐ด(0),๐ดโฒ(0)) =(1,0) and (๐ต(0),๐ตโฒ(0)) =(0,1).
(a)Find formulas for ๐ด and ๐ต, and verify that each satisfies the equation
and its initial data.
Solution
The equation asks for a function whose second derivative is โ4 times
itself. For ๐ฅโณ = โ๐ฅ we used cosโก๐ก and sinโก๐ก. Speeding them up to
cosโก2๐ก and sinโก2๐ก makes the chain rule produce the factor
22 =4 we need. Adjusting the sizes to match the initial data gives
๐ด(๐ก)=cosโก2๐ก,๐ต(๐ก)=12sinโก2๐ก.
Check: ๐ดโฒ = โ2sinโก2๐ก and ๐ดโณ = โ4cosโก2๐ก = โ4๐ด. Also ๐ด(0) =1 and
๐ดโฒ(0) =0. Next, ๐ตโฒ =cosโก2๐ก and ๐ตโณ = โ2sinโก2๐ก = โ4๐ต. Also ๐ต(0) =0
and ๐ตโฒ(0) =1.
(b)Let S be the solution space of this equation on โ.
Consider the initial-data map
๐ธ0:Sโถโ2,๐ธ0(๐ฅ)=(๐ฅ(0),๐ฅโฒ(0)).
Rewrite the equation as a first-order system and check the hypotheses
of the global linear existence-and-uniqueness theorem. What does the
theorem guarantee for each initial position and velocity?
Solution
Let ๐ฃ =๐ฅโฒ. The state is ๐ =(๐ฅ,๐ฃ), and
๐ฅโฒ=๐ฃ,๐ฃโฒ=โ4๐ฅ,that is,๐โฒ=(01โ40)๐.
This is a linear system ๐โฒ =๐ด(๐ก)๐ +๐(๐ก) whose matrix entries are
constants and whose forcing is ๐ =0. Constants are continuous on all of
โ, so the hypotheses of the global theorem hold with
๐ผ =โ.
The theorem guarantees that for every initial state (๐ฅ0,๐ฃ0) there
is exactly one solution on all of โ with
๐(0) =(๐ฅ0,๐ฃ0). Through the correspondence between the system and the
scalar equation, this means: for each initial position and each initial
velocity, there is exactly one motion ๐ฅ(๐ก) of the spring, and it lasts
for all time.
(c)Show that ๐ธ0 is linear: for solutions ๐ข,๐ฃ โS and
real numbers ๐,๐, verify that
๐ธ0(๐๐ข+๐๐ฃ)=๐๐ธ0(๐ข)+๐๐ธ0(๐ฃ).
Solution
By superposition, ๐๐ข +๐๐ฃ is again a solution, so ๐ธ0(๐๐ข +๐๐ฃ) makes
sense. Then
๐ธ0(๐๐ข+๐๐ฃ)=((๐๐ข+๐๐ฃ)(0),ย (๐๐ข+๐๐ฃ)โฒ(0))=(๐๐ข(0)+๐๐ฃ(0),ย ๐๐ขโฒ(0)+๐๐ฃโฒ(0))=๐(๐ข(0),๐ขโฒ(0))+๐(๐ฃ(0),๐ฃโฒ(0))=๐๐ธ0(๐ข)+๐๐ธ0(๐ฃ).
(d)Explain how existence and uniqueness give the other two properties we need:
Conclude that ๐ธ0 is a linear isomorphism.
Solution
Surjectivity comes from existence. Given any (๐ฅ0,๐ฃ0), the global
theorem supplies a solution on all of โ with ๐ฅ(0) =๐ฅ0 and
๐ฅโฒ(0) =๐ฃ0. That solution is in S, and ๐ธ0 sends it to
(๐ฅ0,๐ฃ0).
Injectivity comes from uniqueness. If ๐ธ0(๐ข) =๐ธ0(๐ฃ), then ๐ข and
๐ฃ are two solutions with the same initial position and velocity. The
theorem says there is only one such solution, so ๐ข =๐ฃ.
A linear map that is both injective and surjective is a linear
isomorphism. In particular, S is two-dimensional, like
โ2.
(e)Compute ๐ธ0(๐ด) and ๐ธ0(๐ต). Do these vectors form a basis of
โ2? Use the isomorphism ๐ธ0 to explain why ๐ด,๐ต
form a basis of the solution space S.
Solution
By construction, ๐ธ0(๐ด) =(1,0) and ๐ธ0(๐ต) =(0,1). These are the
standard basis vectors of โ2.
An isomorphism carries bases to bases, so ๐ด,๐ต is a basis of
S. We can also see both properties directly.
Independent: if ๐๐ด +๐๐ต is the zero function, applying ๐ธ0 gives
๐(1,0) +๐(0,1) =(0,0), so ๐ =๐ =0.
Spanning: if ๐ฅ is any solution, then ๐ฅ and
๐ฅ(0)๐ด +๐ฅโฒ(0)๐ต have the same initial state. By injectivity they are
the same function.
(f)Without solving another differential equation, write down the solution
๐ฅ(๐ก) with ๐ฅ(0) = โ3 and ๐ฅโฒ(0) =5. Explain how expressing the initial
state in a basis of โ2 determines the whole function in the
basis ๐ด,๐ต.
Solution
Write the initial state in the standard basis:
( โ3,5) = โ3(1,0) +5(0,1). Use the same coefficients on ๐ด and ๐ต:
๐ฅ(๐ก)=โ3๐ด(๐ก)+5๐ต(๐ก)=โ3cosโก2๐ก+52sinโก2๐ก.
This works because ๐ธ0 is linear:
๐ธ0( โ3๐ด +5๐ต) = โ3(1,0) +5(0,1) =( โ3,5). So โ3๐ด +5๐ต is a solution with
the right initial state, and by uniqueness it is the solution. The
coefficients that build the initial state out of (1,0) and (0,1) are
the same coefficients that build the whole function out of ๐ด and ๐ต.
(g)The coefficient of ๐ต in your answer is the initial velocity 5, but the
coefficient of sinโก2๐ก is not. Explain.
Solution
The function sinโก2๐ก is not ๐ต. Its derivative is 2cosโก2๐ก, so its
initial state is (0,2), not (0,1). In other words, sinโก2๐ก =2๐ต.
The factor of 2 comes from the chain rule.
To reach initial velocity 5 we need 5 copies of ๐ต, which is
52 copies of sinโก2๐ก. The functions cosโก2๐ก,sinโก2๐ก form a
perfectly good basis of S, but it is not the one whose initial
states are the standard basis. With that basis, reading coefficients
straight off the initial data no longer works. The basis ๐ด,๐ต was
chosen exactly so that it does.
6From higher order to first order
Convert each second-order equation or system below into a first-order
system: name the state, and write the system as
๐โฒ =๐น(๐ก,๐). When the system is linear and homogeneous, write it as ๐โฒ =๐ด๐
with an explicit matrix.
(a)Two carts slide without friction on a track. The first is attached to a
wall by a spring of stiffness ๐1, and the second is attached to the
first by a spring of stiffness ๐2. If ๐ฅ1 and ๐ฅ2 are their
displacements from rest, Hooke's law and Newton's second law give
๐1๐ฅโณ1=โ๐1๐ฅ1+๐2(๐ฅ2โ๐ฅ1),๐2๐ฅโณ2=โ๐2(๐ฅ2โ๐ฅ1).
Convert this pair to a first-order system and write its matrix. How many
standard solutions reconstruct every motion of the carts? Which theorem
from the notes guarantees that every motion exists for all time?
Solution
Each cart needs a position and a velocity, so introduce ๐ฃ1 =๐ฅโฒ1 and
๐ฃ2 =๐ฅโฒ2. The state is
๐=(๐ฅ1,๐ฅ2,๐ฃ1,๐ฃ2).
Dividing the equations by the masses and collecting terms gives
๐ฅโฒ1=๐ฃ1,๐ฅโฒ2=๐ฃ2,๐ฃโฒ1=โ๐1+๐2๐1๐ฅ1+๐2๐1๐ฅ2,๐ฃโฒ2=๐2๐2๐ฅ1โ๐2๐2๐ฅ2.
This is linear and homogeneous, ๐โฒ =๐ด๐, with
๐ด=โโ
โ
โ
โ
โ
โ
โ
โ
โ
โ
โ
โโ00100001โ๐1+๐2๐1๐2๐100๐2๐2โ๐2๐200โโ
โ
โ
โ
โ
โ
โ
โ
โ
โ
โ
โโ .
The state has four coordinates, so the solution space is
four-dimensional, and four standard solutions reconstruct every
motion. For example, the first starts from ๐1 =(1,0,0,0): cart one
pulled out one unit, cart two at rest in place, and both released from
rest.
The entries of ๐ด are constants, so they are continuous on all of
โ. The global existence-and-uniqueness theorem for linear
systems therefore guarantees that every initial state gives exactly one
motion, defined for all time.
(b)A pendulum of length ๐ฟ swinging through the angle ๐ satisfies
๐โณ=โ๐๐ฟsinโก๐.
Convert this to a first-order system. Is it linear? Check the hypotheses of
the local existence-and-uniqueness theorem in state space, and state what
it guarantees about a pendulum released from any angle with any angular
velocity. Does the global theorem for linear systems apply?
Solution
Let ๐ =๐โฒ be the angular velocity. The state is
๐ =(๐,๐), and
๐โฒ=๐,๐โฒ=โ๐๐ฟsinโก๐.
This is not linear, because of the sinโก๐ term.
The rate functions are ๐น1(๐,๐) =๐ and
๐น2(๐,๐) = โ(๐/๐ฟ)sinโก๐. Since ๐น1 does not depend on
๐ and ๐น2 does not depend on ๐, their partial derivatives
with respect to those variables are zero. The other two partials are
๐๐น1๐๐=1,๐๐น2๐๐=โ๐๐ฟcosโก๐.
All of these are continuous everywhere. So the local theorem applies at
every initial state. However we release the pendulum, from any angle and
with any angular velocity, there is exactly one motion, at least for some
interval of time around the release.
The global theorem for linear systems does not apply, because the system
cannot be written as ๐โฒ =๐ด(๐ก)๐ +๐(๐ก). On its own, the local theorem
does not tell us how long the motion lasts. We certainly expect a real
pendulum to keep swinging forever, and it does, but proving that takes an
argument beyond the theorems we have so far.