Linear Transformations

Note: In these solutions, key strategies are marked with â .

 
a
1. Decide whether T : R2 → P1 defined by T = a2 t + b2 is a linear transformation.
b
Solution. Note: Before you decide whether T is a linear transformation, it’s often helpful to do an
example ofevaluating
 T to make sure the definition and notation of T are clear. For example, here
2
we have T = 4t + 9. From this, we can see that we’re writing elements of P1 in terms of the
3
variable t.
As explained in Prep Video 28, it’s a good habit to first
  check whether T (zero element of domain) =
0
zero element of codomain. The zero element of R2 is while the zero element of P1 is 0t + 0, and it
  0
0
is indeed true that T = 0t + 0 = 0. But this is not enough to assure us that T is linear, so now
0
we need to determine whether T preserves addition and scalar multiplication.
       
a1 a a1 a2
ˆ Does T preserve addition? That is, it is true that T + 2 =T +T for
    b1 b2 b1 b2
a a
all 1 , 2 in R2 ? Let’s check:
b1 b2
     
a1 a a1 + a2
T + 2 =T
b1 b2 b1 + b2
= (a1 + a2 )2 t + (b1 + b2 )2 by definition of T

On the other hand,

   
a1 a2
T +T = (a21 t + b21 ) + (a22 t + b22 )
b1 b2
       
a1 1 a 1
In general, it is not true that these are equal. A concrete example is = and 2 = :
    b 1 0 b2 0
   
1 1 1 1
T + = 4t, while T +T = 2t.
0 0 0 0

â Notice that a single example is enough to show us that T is not closed under addition (and
the same is true for showing that T is not closed under scalar multiplication).

So, T is not a linear transformation (because we have shown that it is not closed under addition).
 
2×2 a b
2. Let T : R → P4 be the linear transformation defined by T = (a+b)x4 +(a−b+d)x3 −(a+b).
c d
Find a basis of ker T and a basis of im T .
   
a b a b
Solution. ker T consists of all matrices in R2×2 such that T = 0 (remember
c d c d
that the 0 here refers to the zero element of the codomain, which is the degree 4 zero polynomial

1
© Harvard Math 21b
0x4 + 0x3 + 0x2 + 0x + 0). That is, ker T consists of
 
a b
all such that (a + b)x4 + (a − b + d)x3 − (a + b) = 0x4 + 0x3 + 0x2 + 0x + 0.
c d

We can rewrite this as the set of

 
a b
all such that a + b = 0 and a − b + d = 0.
c d

We can think of the equations a + b = 0 and a − b + d = 0 as forming a linear system (with variables
a, b, c, d, and we can solve this linear system using Gauss-Jordan or by inspection. If we use inspection,
the first equation says b = −a; plugging this into the second equation gives 2a + d = 0, so d = −2a.
Thus, a and c can be anything, and then b = −a and d = −2a. So, ker T consists of
 
a −a
all matrices of the form .
c −2a

We can rearrange this to say that ker T consists of

   
1 −1 0 0
all matrices of the form a +c ,
0 −2 1 0
       
1 −1 0 0 1 −1 0 0
which exactly says that ker T = span , . It’s clear that and are
0 −2 1 0 0 −2 1 0
   
1 −1 0 0
linearly independent, so one basis of ker T is , .
0 −2 1 0

To find im T , we can use either of two approaches:

ˆ Approach 1: By inspection.

im T is the set of all possible outputs of T ; in other words, it is the set of all polynomials
in P4 which can be written as (a + b)x4 + (a − b + d)x3 − (a + b). Such a polynomial can
be rewritten as a(x4 + x3 − 1) + b(x4 − x3 − 1) + d(x3 ), so the polynomials of this form are
exactly linear combinations of x4 + x3 − 1, x4 − x3 − 1, and x3 . By inspection, we can see that
the latter two are linearly independent while the first is a linear combination of the other two:
x4 + x3 − 1 = (x4 − x3 − 1) + 2(x3 ), so one basis of im T is x4 − x3 − 1, x3 .

ˆ Approach 2: Use a basis of the domain.

       
1 0 0 1 0 0 0 0
The matrices M1 = , M2 = , M3 = , and M4 = form a basis of R2×2 ,
0 0 0 0 1 0 0 1
so im T = span(T (M1 ), T (M2 ), T (M3 ), T (M4 )). We just need to look for linear relations among
T (M1 ), T (M2 ), T (M3 ), T (M4 ), which is exactly what ker T tells us. We’ve already determined
that one basis of ker T is (M1 − M2 − 2M4 , M3 ); this says:

T (M1 − M2 − 2M4 ) = 0, which can be rewritten as T (M1 ) − T (M2 ) − 2T (M4 ) = 0

T (M3 ) = 0

So, one basis of im T is (T (M1 ), T (M2 )) (we can see from the linear relations above that both
T (M3 ) and T (M4 ) are in span(T (M1 ), T (M2 ))). Using the definition of T , T (M1 ) = x4 + x3 − 1
and T (M2 ) = x4 − x3 − 1, so a basis of im T is x4 + x3 − 1, x4 − x3 − 1 .

2
© Harvard Math 21b
 3 Let’s just make sure our answers agree with the rank-nullity theorem: we’ve found that dim(ker T ) =
2 and dim(im T ) = 2, and these do sum to dim(R2×2 ) = 4.

3. Show that T : C ∞ → C ∞ defined by T (f ) = f 00 − 4f 0 + 3f is a linear transformation. Find a basis of

ker T .
Solution. T is linear:

ˆ For all f, g ∈ C ∞ , T (f + g) = (f + g)00 − 4(f + g)0 + 3(f + g) = (f 00 − 4f 0 + 3f ) + (g 00 − 4g 0 + 3g) =

T (f ) + T (g). So, T preserves addition.

ˆ For all f ∈ C ∞ and all scalars c, T (cf ) = (cf )00 − 4(cf )0 + 3(cf ) = cf 00 − 4cf 0 + 3cf = c(f 00 −
4f 0 + 3f ) = cT (f ). So, T preserves scalar multiplication.

The kernel of T is the set of solutions of f 00 − 4f 0 + 3f = 0. This differential equation has characteristic
polynomial λ2 − 4λ + 3 = (λ − 1)(λ − 3), and this polynomial has roots 1, 3. So, ker T is the set of
functions c1 et + c2 e3t . Therefore, a basis of ker T is (et , e3t ) .

4. (Weekly PSet 11, #1) Find a basis for ker T , im T , and verify the Rank-Nullity Theorem for the
linear transformation T : R2×2 → P2 defined by
 
a b
T = (a + c)x2 + (b + c)x + (a − b).
c d

5. (Weekly PSet 11, #2) Consider the linear transformation D : C ∞ → C ∞ defined by D(f ) = f 0 .
It turns out that every real number is an eigenvalue of D. If λ is a real number, find a basis of the
λ-eigenspace of D.

6. Find the kernel and image of Dn : C ∞ → C ∞ . (Here, D : C ∞ → C ∞ is the derivative.)

Solution. The kernel of Dn is the set of polynomials of degree < n. The image of D is C ∞ . (Any
function f in C ∞ can be anti-differentiated n times to get a function g with g (n) = f . So, any function
f in C ∞ can be written as Dn g for some g, which means it is in the image of Dn .)

7. Are the following transformations linear? If so, find a basis of the kernel and image.

(a) T : R2×2 → R defined by T (M ) = rank M .

Solution. As explained in Prep Video 28, it’s a good habit to first check
 whether
 T (zero element in domain) =
0 0
zero element in codomain. The zero element in the domain R2×2 is , while the zero element
0 0  
0 0
in the codomain R is simply the number 0. So, we want to check whether T = 0. That
0 0
 
0 0
is, is the rank of equal to 0? Yes!
0 0
But this is not enough to assure us that T is linear, so now we need to determine whether T
preserves addition and scalar multiplication.

ˆ Does T preserve addition? That is, it is true that rank(M1 + M2 ) = rank(M1 ) + rank(M2 ) for
all 2 × 2 matrices M1 , M2 ? There doesn’t seem to be any reason this must be true, and in fact

3
© Harvard Math 21b
 we can come up with a counterexample. For example, if M1 and M2 are both the identity
1 0
matrix , then rank(M1 ), rank(M2 ), and rank(M1 + M2 ) are all 2, so rank(M1 + M2 ) 6=
0 1
rank(M1 ) + rank(M2 ). So, T does not preserve addition.

Therefore, T is not a linear transformation .

(b) T : R2×2 → R defined by T (M ) = tr M .

Solution. As explained in Prep Video 28, it’s a good habit to first check
 whether
 T (zero element in domain) =
0 0
zero element in codomain. The zero element in the domain R2×2 is , while the zero element
0 0  
0 0
in the codomain R is simply the number 0. So, we want to check whether T = 0. That
  0 0
0 0
is, is the trace of equal to 0? Yes!
0 0
But this is not enough to assure us that T is linear, so now we need to determine whether T
preserves addition and scalar multiplication.

ˆ Does T preserve addition? That is, it is true that

       
a1 b1 a2 b2 a1 b1 a2 b2
T + =T +T
c1 d1 c2 d 2 c1 d1 c2 d2
   
a1 b1 a b2
for all , 2 in R2×2 ? Let’s check:
c1 d1 c2 d2
     
a1 b1 a b2 a1 + a2 b1 + b2
T + 2 =T = (a1 + a2 ) + (d1 + d2 )
c1 d1 c2 d2 c1 + c2 d1 + d2

while
   
a1 b1 a2 b2
T +T = (a1 + d1 ) + (a2 + d2 )
c1 d1 c2 d2

These are indeed equal, so T preserves addition.

    
a b a b
ˆ Does T preserve scalar multiplication? That is, is it true that T k = kT
 c d c d

a b
for all in R2×2 and all scalars k? Let’s check:
c d
    
a b ka kb
T k =T = ka + kd
c d kc kd

while
 
a b
kT = k(a + d)
c d

These are indeed equal, so T preserves scalar multiplication.

Thus, T is a linear transformation .

4
© Harvard Math 21b
  
a b
The kernel of T is the set of all matrices with trace 0; by Daily Problem Set 27, #W1(b),
c d
     
1 0 0 1 0 0
a basis of this space is , , .
0 −1 0 0 1 0

The image of T is the set of all possible traces, which is all real numbers (if you think of any
number, there is a matrix whose trace is that number). All real numbers are scalar multiples of
the number 1, so a basis of im T is (1) .

3 It’s a good habit to use the rank-nullity theorem to check that your kernel and image are
reasonable; here, dim(ker T ) + dim(im T ) = 3 + 1 = 4 = dim(R2×2 ).
 
1
(c) T : R2×2 → R2 defined by T (M ) = M .
2

Solution. Using the same strategy as in the previous part, we find that T is linear :
 
1
ˆ It preserves addition: if we let ~v = , then T (M1 + M2 ) = (M1 + M2 )~v = M1~v + M2~v =
2
T (M1 ) + T (M2 ).

ˆ It preserves scalar multiplication: T (kM ) = kM~v = k(M~v ) = kT (M ).

      
a b a b 1 0
The kernel of T is the set of all matrices such that = . We can rewrite
c d c d 2 0
   
a + 2b 0
this equation as = . This is true exactly when a = −2b and c = −2d, so the kernel
c + 2d 0      
−2b b −2 1 0 0
of T consists of all matrices of the form =b +d . Therefore, a basis
    −2d d 0 0 −2 1
−2 1 0 0
of ker T is , . (We know this is a basis because every element of ker T can be
0 0 −2 1
expressed as a linear combination of these two matrices in exactly one way.)
By the rank-nullity theorem, dim(ker T )+dim(im T ) = dim(R2×2 ), so dim(im T ) = 4−dim(ker T ) =
4 − 2 = 2. Since im T is a subspace of R2 and the only 2-dimensional subspace of R2 is R2 itself,
   
2 1 0
im T must be all of R . Thus, one basis of im T is , .
0 1

5
© Harvard Math 21b