SUM OF TWO SUBSPACES OF A VECTOR SPACE

Suppose 𝑊1 and 𝑊2 are subspaces of a vector space 𝑉 over ℝ. Then define

𝑊1 + 𝑊2 = {𝑤1 + 𝑤2 : 𝑤1 ∈ 𝑊1 , 𝑤2 ∈ 𝑊2 }
The set 𝑊1 + 𝑊2 is called the sum of the subspaces 𝑊1 and 𝑊2 .

Theorem 01
The sum of two vector spaces 𝑊1 and 𝑊2 , 𝑊1 + 𝑊2 of a vector space 𝑉 is also a subspace of 𝑉.

Proof
The sum of 𝑊1 and 𝑊2 is defined by 𝑊1 + 𝑊2 = {𝑤1 + 𝑤2 : 𝑤1 ∈ 𝑊1 , 𝑤2 ∈ 𝑊2 }.
Since both of 𝑊1 and 𝑊2 are subspaces of 𝑉, both subspaces contain the zero vector of 𝑉,
i.e. 0 ∈ 𝑊1 and 0 ∈ 𝑊2 which implies that 0 + 0 = 0 ∈ 𝑊1 + 𝑊2 .

Therefore 𝑊1 + 𝑊2 is a non-empty subset of 𝑉.

1
Let 𝑤11 + 𝑤21 , 𝑤12 + 𝑤22 ∈ 𝑊1 + 𝑊2 where 𝑤11 , 𝑤12 ∈ 𝑊1 and 𝑤21 , 𝑤22 ∈ 𝑊2 .
Then consider (𝑤11 + 𝑤21 ) + (𝑤12 + 𝑤22 ) = (𝑤11 + 𝑤12 ) + (𝑤21 + 𝑤22 ) ∈ 𝑊1 + 𝑊2

Vector addition is Definition of sum of subspaces

associative and
commutative

Therefore the vector addition satisfy the closure property.

Let 𝑤11 + 𝑤21 ∈ 𝑊1 + 𝑊2 and 𝑐 ∈ ℝ , then 𝑤11 ∈ 𝑊1 and , 𝑤21 ∈ 𝑊2 .
Since 𝑊1 and 𝑊2 are vector spaces themselves, we have 𝑐𝑤11 ∈ 𝑊1 and c𝑤21 ∈ 𝑊2 .
It gives that 𝑐𝑤11 + c𝑤21 = 𝑐(𝑤11 + 𝑤21 ) ∈ 𝑊1 + 𝑊2 .
Therefore the scalar multiplications satisfy the closure property.
So 𝑊1 + 𝑊2 is a subspace of 𝑉.
Moreover 𝑊1 + 𝑊2 is the smallest subspace of 𝑉 that contains both 𝑊1 and 𝑊2 .

Example
(a) Let 𝑊1 = {(𝑥, 0,0) ∈ ℝ3 , 𝑥 ∈ ℝ} and 𝑊2 = {(0, 𝑦, 0) ∈ ℝ3 , 𝑦 ∈ ℝ} .
Then both of them are subspaces of the vector space ℝ3 .
Now we can find 𝑊1 + 𝑊2 as follows:
𝑊1 + 𝑊2 = {(𝑥, 𝑦, 0): 𝑥, 𝑦 ∈ ℝ}.

(b) Next, let 𝑉 = ℝ4 , 𝑊1 = {(𝑥, 𝑦, 𝑧, 𝑤): 𝑥 + 𝑦 + 𝑧 = 0, 𝑥 + 2𝑦 − 𝑧 = 0},

𝑊2 = {(𝑥, 2𝑥, 3𝑥, 𝑡): 𝑠, 𝑡 ∈ ℝ}
Find a basis for the subspace 𝑊1 + 𝑊2 .

2
The following theorem tells us about the dimension of 𝑊1 + 𝑊2 and the proof of the theorem suggests
how to write its basis.
Theorem
If 𝑊1 and 𝑊2 are subspaces of a vector space 𝑉, then
dim(𝑊1 + 𝑊2 ) = dim(𝑊1 ) + dim(𝑊2 ) − dim(𝑊1 ∩ 𝑊2 )
Proof:
Let 𝑆 be a basis of 𝑊1 ∩ 𝑊2 (If 𝑊1 ∩ 𝑊2 is the zero space then 𝑆 = 𝛷.
For each 𝑖 = 1,2, extend 𝑆 to a basis 𝐵𝑖 of 𝑊𝑖 .

Let 𝑆 = {𝑢1 , 𝑢2 , … , 𝑢𝑟 }, 𝑊1 = {𝑢1 , 𝑢2 , … , 𝑢𝑟 , 𝑣1 , 𝑣2 , … , 𝑣𝑠 } and 𝑊2 = {𝑢1 , 𝑢2 , … , 𝑢𝑟 , 𝑤1 , 𝑤2 , … , 𝑤𝑡 }.

Then dim(𝑊1 ∩ 𝑊2 ) = 𝑟, dim(𝑊1 ) = 𝑟 + 𝑠 and dim(𝑊1 ) = 𝑟 + 𝑡 .

Let 𝐵 = {𝑢1 , 𝑢2 , … , 𝑢𝑟 , 𝑣1 , 𝑣2 , … , 𝑣𝑠 , 𝑤1 , 𝑤2 , … , 𝑤𝑡 }.

It is enough to show that 𝐵 is a basis of 𝑊1 + 𝑊2 , because then we have

dim(𝑊1 + 𝑊2 ) = 𝑟 + 𝑠 + 𝑡 = (𝑟 + 𝑠) + (𝑟 + 𝑡) − 𝑟 = dim(𝑊1 ) + dim(𝑊2 ) − dim(𝑊1 ∩ 𝑊2 )
To show that 𝐵 is linearly independent, ∑𝑟𝑖=1 𝑎𝑖 𝑢𝑖 + ∑𝑠𝑟=1 𝑏𝑟 𝑣𝑟 + ∑𝑡𝑖=1 𝑐𝑡 𝑤𝑡 = 0

Then ∑𝑟𝑖=1 𝑎𝑖 𝑢𝑖 + ∑𝑠𝑗=1 𝑏𝑗 𝑣𝑗 = − ∑𝑡𝑘=1 𝑐𝑘 𝑤𝑘 .

Now the LHS is in 𝑊1 and the RHS is in 𝑊2 . So this element is in 𝑊1 ∩ 𝑊2 .

Thus − ∑𝑡𝑘=1 𝑐𝑘 𝑤𝑘 = ∑𝑟𝑖=1 𝑑𝑖 𝑢𝑖 so that ∑𝑟𝑖=1 𝑑𝑖 𝑢𝑖 +∑𝑡𝑘=1 𝑐𝑘 𝑤𝑘 = 0 which implies that 𝑑𝑖 = 0
and 𝑐𝑘 = 0 for all 𝑖 and 𝑘.(Since 𝐵2 is linearly independent)
Therefore, ∑𝑟𝑖=1 𝑎𝑖 𝑢𝑖 + ∑𝑠𝑗=1 𝑏𝑗 𝑣𝑗 = 0 which implies 𝑎𝑖 = 0 and 𝑏𝑗 = 0 for each 𝑖 and 𝑗.
(Since 𝐵1 is linearly independent)
Thus 𝐵 is linearly independent.

Let 𝑤 ∈ 𝑊1 + 𝑊2. Then 𝑤 = 𝑤1 + 𝑤2 for some 𝑤𝑖 ∈ 𝑊𝑖 for 𝑖 = 1,2.

Then 𝑤1 = ∑𝑟𝑖=1 𝑝𝑖 𝑢𝑖 + ∑𝑠𝑗=1 𝑞𝑗 𝑣𝑗 and 𝑤2 = ∑𝑟𝑖=1 𝑔𝑖 𝑢𝑖 + ∑𝑘=1 ℎ𝑘 𝑤𝑘 .

Now 𝑤 = ∑𝑟𝑖=1(𝑝𝑖 + 𝑔𝑖 )𝑢𝑖 + ∑𝑠𝑗=1 𝑞𝑗 𝑣𝑗 + ∑𝑘=1 ℎ𝑘 𝑤𝑘 which is in 𝑠𝑝𝑎𝑛(𝐵).

Therefore 𝐵 is a basis for 𝑊1 + 𝑊2 and the theorem is proved.

3
The sum 𝑊1 + 𝑊2 is called direct if 𝑊1 ∩ 𝑊2 = {0}.

In particular, a vector space 𝑉 is said to be the direct sum of two subspaces 𝑊1 and 𝑊2 if 𝑉 = 𝑊1 + 𝑊2
and 𝑊1 ∩ 𝑊2 = {0}

When 𝑉 is a direct sum of 𝑊1 and 𝑊2 , we write 𝑉 = 𝑊1 ⨁ 𝑊2 .

Theorem
Suppose 𝑊1 and 𝑊2 are subspaces of a vector space 𝑉 so that 𝑉 = 𝑊1 + 𝑊2 . Then 𝑉 = 𝑊1 ⨁ 𝑊2 if
and only if every vector in 𝑉 can be written in a unique way as 𝑤1 + 𝑤2 where 𝑤𝑖 ∈ 𝑊𝑖 for 𝑖 = 1,2.
Proof:
Since 𝑉 = 𝑊1 + 𝑊2 for every vector in 𝑉 is a sum of a vector in 𝑊1 and a vector in 𝑊2 .
Suppose that for every 𝑣 ∈ 𝑉, there is only one pair (𝑤1 , 𝑤2 ) with 𝑤𝑖 ∈ 𝑊𝑖 such that 𝑣 = 𝑤1 + 𝑤2 . If
𝑊1 ∩ 𝑊2 is non-zero, pick a non-zero vector 𝑢 ∈ 𝑊1 ∩ 𝑊2 .
Then 𝑢 = 𝑢 + 0.
With 𝑢 ∈ 𝑊1 , 0 ∈ 𝑊2 and 𝑢 = 0 + 𝑢 with 0 ∈ 𝑊1 and 𝑢 ∈ 𝑊2 .

This contradicts our uniqueness assumption.

Conversely , suppose 𝑉 = 𝑊1 ⨁ 𝑊2 .Then 𝑉 = 𝑊1 + 𝑊2 and 𝑊1 ∩ 𝑊2 = {0}.

If for 𝑣 ∈ 𝑉, we have 𝑣 = 𝑤1 + 𝑤2 = 𝑤1′ + 𝑤2 ′ for 𝑤1 , 𝑤1′ ∈ 𝑊1 and 𝑤2 , 𝑤2′ ∈ 𝑊2 , then

𝑤1 − 𝑤1′ = 𝑤2 ′ − 𝑤2 .
The LHS is in 𝑊1 and the RHS is in 𝑊2 : therefore , this vector is in 𝑊1 ∩ 𝑊2 . Since by assumption
𝑊1 ∩ 𝑊2 = {0}, we have 𝑤1 − 𝑤1′ = 0 and 𝑤2′ − 𝑤2 = 0 so that 𝑤1 = 𝑤1′ and 𝑤2′ = 𝑤2 .

Example:
Let 𝑉 = ℝ2 , 𝑊1 = {(𝑥, 2𝑥): 𝑥 ∈ ℝ} and 𝑊2 = {(𝑥, 3𝑥): 𝑥 ∈ ℝ}. Show that 𝑉 = 𝑊1 ⨁ 𝑊2.

4
 Preposition 01
Let 𝑊1 and 𝑊2 are subspaces of a vector space 𝑉. Then 𝑉 = 𝑊1 ⊕ 𝑊2 if and only if the following
conditions hold:
1. 𝑉 = 𝑊1 + 𝑊2 ,
2. If 𝑤1 + 𝑤2 = 0 with 𝑤1 ∈ 𝑊1 and 𝑤2 ∈ 𝑊2, then 𝑤1 = 𝑤2 = 0.

Proof
Suppose 𝑉 = 𝑊1 ⊕ 𝑊2 . Then condition 1 hold by definition.
Certainly 0 = 0 + 0, and by uniqueness this is the only way to write 0 ∈ 𝑉, we have 𝑤1 = 𝑤2 = 0.

Suppose condition 1 and 2 hold. By condition 1, we have that , for all 𝑣 ∈ 𝑉, there exist 𝑤1 ∈ 𝑊1 and
𝑤2 ∈ 𝑊2 such that 𝑣 = 𝑤1 + 𝑤2 . Suppose 𝑣 = 𝑢1 + 𝑢2 for 𝑢1 ∈ 𝑊1 and 𝑤2 ∈ 𝑊2 .
Subtracting these two equations , we obtain 0 = (𝑢1 − 𝑤1 ) + (𝑢2 − 𝑤2 ) where 𝑢1 − 𝑤1 ∈ 𝑊1 and
𝑢2 − 𝑤2 ∈ 𝑊2 . By condition 2 implies that 𝑢1 − 𝑤1 = 0 and 𝑢2 − 𝑤2 = 0 or equivalently

𝑢1 = 𝑤1 and 𝑢2 = 𝑤2 as desired.

Preposition 02
Let 𝑊1 and 𝑊2 are subspaces of a vector space 𝑉. Then 𝑉 = 𝑊1 ⊕ 𝑊2 if and only if the following
conditions hold:
1. 𝑉 = 𝑊1 + 𝑊2
2. If 𝑊1 ∩ 𝑊2 = {0}.

Proof
Suppose 𝑉 = 𝑊1 ⊕ 𝑊2 . Then condition 1 hold by definition.
If 𝑢 ∈ 𝑊1 ∩ 𝑊2 , then 0 = 𝑢 + (−𝑢) with 𝑢 ∈ 𝑊1 and −𝑢 ∈ 𝑊2 and by preposition 1, we have 𝑢 =
0 and − 𝑢 = 0 so that 𝑊1 ∩ 𝑊2 = {0}

Suppose condition 1 and 2 hold.

To prove that 𝑉 = 𝑊1 ⊕ 𝑊2 holds, suppose 0 = 𝑤1 + 𝑤2 where 𝑤1 ∈ 𝑊1 and 𝑤2 ∈ 𝑊2 .

By preposition 1, it suffices to show that 𝑤1 = 𝑤2 = 0. It implies that 𝑤1 = −𝑤2 ∈ 𝑊2 . Hence 𝑤1 ∈

𝑊1 ∩ 𝑊2 which in turn implies that 𝑤1 = 0. Then we also have 𝑤2 = 0.

5
Everything in this section can be generalized to 𝑚 subspaces 𝑊1 , 𝑊2 , … , 𝑊𝑚 with the noble exception
preposition2. To see this, consider the following example:
Example
𝑊1 = {(𝑥, 𝑦, 0): 𝑥, 𝑦 ∈ ℝ} , 𝑊2 = {(0,0, 𝑧): 𝑧 ∈ ℝ} and 𝑊3 = {(0, 𝑦, 𝑦): 𝑦 ∈ ℝ} are subspaces of ℝ3 .
but ℝ3 ≠ 𝑊1 + 𝑊2 + 𝑊3 , since , for example

But 𝑊1 ∩ 𝑊2 = 𝑊2 ∩ 𝑊3 = 𝑊1 ∩ 𝑊3 ={0} so that the analog in preposition 2 does not hold.