logo

직합의 성질 📂Linear Algebra

직합의 성질

Theorem1

Let W1,W2,,WkW_{1}, W_{2}, \dots, W_{k} be subspaces of a finite-dimensional vector space VV. The following propositions are equivalent:

  1. V=W1W2WkV = W_{1} \oplus W_{2} \oplus \cdots \oplus W_{k}
  2. It is V=i=1kWiV = \sum\limits_{i=1}^{k}W_{i}, and for any vectors viWi(1ik)v_{i} \in W_{i}(1 \le i \le k), if v1+vk=0v_{1} + \cdots v_{k} = 0, then for all ii, vi=0v_{i} = 0 holds.
  3. All vVv \in V can be uniquely expressed in the form of v=v1++vk(viWi)v = v_{1} + \cdots + v_{k} (v_{i} \in W_{i}).
  4. If γi\gamma_{i} is an ordered basis of WiW_{i}, then γ1γk\gamma_{1} \cup \cdots \cup \gamma_{k} is an ordered basis of VV.
  5. There exists an ordered basis γi\gamma_{i} of WiW_{i} such that γ1γk\gamma_{1} \cup \cdots \cup \gamma_{k} becomes an ordered basis of VV.

Explanation

For two subspaces W1,W2W_{1}, W_{2} of VV,

  • W1+W2W_{1} + W_{2} is the sum of W1,W2W_{1}, W_{2}.
  • W1W2W_{1} \oplus W_{2} is the direct sum of W1,W2W_{1}, W_{2}.

Proof

  • 1.    2.1. \implies 2.

    Assume 1. Then it is V=i=1kWiV = \sum\limits_{i=1}^{k}W_{i}. Let vi+vk=0(viWi)v_{i} + \cdots v_{k} = 0 (v_{i} \in W_{i}). Then for some jj, vj=ijviijWi -v_{j} = \sum\limits_{i\ne j}v_{i} \in \sum\limits_{i\ne j}W_{i} Since vjWjv_{j} \in W_{j}, we obtain the following: vjWjijWi={0} -v_{j} \in W_{j} \cap \sum\limits_{i\ne j}W_{i} = \left\{ 0 \right\} Therefore, for all ii, it is vi=0v_{i} = 0.

  • 2.    3.2. \implies 3.

    Assume 2. Let it be vVv \in V. Then there exist some viWiv_{i} \in W_{i}, which are v=v1++vkv = v_{1} + \cdots + v_{k} by assumption. Suppose there exists another expression v=w1++wk (wiWi)v = w_{1} + \cdots + w_{k}\ (w_{i} \in W_{i}). Then we obtain: 0=vv=(v1w1)++(vkwk) 0 = v - v = (v_{1} - w_{1}) + \cdots + (v_{k} - w_{k}) Therefore, it is viwiWiv_{i} - w_{i} \in W_{i} and since we assumed 2., for all ii, it is viwi=0v_{i} - w_{i} = 0. Thus, v=v1++vkv = v_{1} + \cdots + v_{k} is a unique representation.

  • 3.    4.3. \implies 4.

    Assume 3. Let γi\gamma_{i} be an ordered basis of WiW_{i}. By assumption, it is V=i=1kWiV = \sum\limits_{i=1}^{k}W_{i}. This means γ1γk\gamma_{1} \cup \cdots \cup \gamma_{k} generates VV. To show this set is linearly independent, assume vijγjv_{ij} \in \gamma_{j} and for scalars aija_{ij}, that i,jaijvij=0\sum\limits_{i,j} a_{ij} v_{ij} = 0 is (j=1,,mi, i=1,,k)(j = 1,\dots,m_{i},\ i=1,\dots,k). For each ii, wi=j=1miaijvj w_{i} = \sum_{j=1}^{m_{i}}a_{ij}v_{j} Suppose it holds true for all ii, then it is 0=0++0=w1++wk0 = 0 + \cdots + 0 = w_{1} + \cdots + w_{k}. By assumption, for all ii, it is wi=0w_{i} = 0. 0=wi=j=1miaijvj 0 = w_{i} = \sum_{j=1}^{m_{i}}a_{ij}v_{j} Since each γi\gamma_{i} is a basis, it is linearly independent, and therefore for all i,ji,j, it is aij=0a_{ij} = 0. Therefore, γ1γk\gamma_{1} \cup \cdots \cup \gamma_{k} is linearly independent and is a basis of VV.

  • 4.    5.4. \implies 5.

    It is trivial.

  • 5.    1.5. \implies 1.

    Assume 5. For each ii, let γi\gamma_{i} be an ordered basis of WiW_{i} such that γ1γk\gamma_{1} \cup \cdots \cup \gamma_{k} becomes an ordered basis of VV. Then, the sum of the generating sets equals the generating set of the sum, V=span(γ1γk)=span(γ1)++span(γk)=i=1kWi \begin{align*} V &= \span(\gamma_{1} \cup \cdots \cup \gamma_{k}) \\ &= \span(\gamma_{1}) + \cdots + \span(\gamma_{k}) \\ &= \sum\limits_{i=1}^{k}W_{i} \end{align*} Now, to prove exclusivity, fix j(1jk)j (1 \le j \le k) and for a non-zero vVv \in V, assume the following: vWjijWi v \in W_{j} \cap \sum\limits_{i \ne j} W_{i} By the definition of the intersection, the following holds: vWj=spanγj,vijWi=span(ijγi) v \in W_{j} = \span \gamma_{j}, \qquad v \in \sum\limits_{i \ne j} W_{i} = \span(\bigcup_{i \ne j} \gamma_{i}) Then vv has two different linear combinations in the basis γ1γk\gamma_{1} \cup \cdots \cup \gamma_{k} of VV. This contradicts the uniqueness of linear combinations in a basis, thus the assumption is incorrect. Therefore, non-zero elements of VV do not belong to WjijWiW_{j} \cap \sum\limits_{i \ne j} W_{i}. WjijWi={0} W_{j} \cap \sum\limits_{i \ne j} W_{i} = \left\{ 0 \right\}


  1. Stephen H. Friedberg, Linear Algebra (4th Edition, 2002), p276 ↩︎