What Dimension Counts Tell Us
In a finite-dimensional vector space, the dimension gives a precise limit on how large an independent list can be and how small a spanning list can be. These are useful proof strategies: sometimes the number of vectors alone settles dependence or spanning without requiring a full calculation. But a list with exactly the dimension’s number of vectors is not automatically a basis. It must also be independent or spanning.
Let \(\mathbb{F}\) be \(\mathbb{R}\) or \(\mathbb{C}\), and let \(V\) be a finite-dimensional vector space over \(\mathbb{F}\). A list \((\mathbf{v}_1,\dots,\mathbf{v}_p)\) is linearly independent if the equation \(c_1\mathbf{v}_1+\cdots+c_p\mathbf{v}_p=\mathbf{0}\) forces all \(c_j=0\). It spans \(V\) if every vector in \(V\) is a linear combination of its entries. A basis is a list that is both linearly independent and spanning. The empty list is independent and spans the zero space.
For a list \((\mathbf{v}_1,\dots,\mathbf{v}_p)\), place its coordinate vectors as the columns of \(\mathbf{A}\in\mathbb{F}^{n\times p}\). Then \(\mathbf{A}\mathbf{c}=\mathbf{0}\) is exactly the coordinate form of a zero combination of the list, and \(\mathbf{A}\mathbf{c}=[\mathbf{w}]_{\mathcal{B}}\) asks whether \(\mathbf{w}\) is a combination of the list. Thus independence and spanning questions in \(V\) become questions about columns in \(\mathbb{F}^n\).
The Dimension-Count Theorem
- If the list is linearly independent, then \(p\le n\).
- If the list spans \(V\), then \(p\ge n\).
- If \(p=n\), the list is linearly independent if and only if it spans \(V\); equivalently, it is a basis if either condition holds.
Proof. First suppose \(n>0\), and form the coordinate matrix \(\mathbf{A}\in\mathbb{F}^{n\times p}\) described above. Row operations preserve the solution set of a homogeneous system and are reversible; a matrix in reduced row echelon form has at most \(n\) pivots, with at most one pivot in each column.
If \(p>n\), at least one column has no pivot. In \(\operatorname{rref}(\mathbf{A})\mathbf{c}=\mathbf{0}\), assign \(1\) to one free variable and \(0\) to every other free variable. The pivot variables are then determined by the equations. This gives a nonzero solution \(\mathbf{c}\), so the columns, and hence the original list, are dependent. Therefore an independent list cannot have \(p>n\).
Now suppose the list is independent. Its homogeneous coordinate system has only the zero solution, so every column must have a pivot: otherwise the free-variable construction just given would produce a nonzero solution. There are at most \(n\) pivots, and therefore \(p\le n\).
Next suppose the list spans \(V\). Its coordinate columns span \(\mathbb{F}^n\). If \(p<n\), the row-reduced matrix has at most \(p\) pivots and hence at least one zero row. Apply to the columns the same sequence of row operations used to reduce \(\mathbf{A}\). Each operation is invertible, so it carries a spanning set of \(\mathbb{F}^n\) to another spanning set. But every column of the reduced matrix has a zero in the zero-row position, so its columns cannot span \(\mathbb{F}^n\); for example, they cannot produce the standard basis vector with \(1\) in that position and \(0\) elsewhere. This contradiction proves \(p\ge n\).
Finally let \(p=n\). If the list is independent, every column is a pivot column. There are \(n\) columns and \(n\) rows, so the reduced matrix has a pivot in every row as well. Its reduced row echelon form is \(\mathbf{I}_n\). For any coordinate vector \(\mathbf{d}\in\mathbb{F}^n\), applying the same row operations to the augmented system \(\mathbf{A}\mathbf{c}=\mathbf{d}\) gives \(\mathbf{I}_n\mathbf{c}=\mathbf{d}'\) for some transformed right-hand side \(\mathbf{d}'\). This system has a solution, so the columns span \(\mathbb{F}^n\), and the list spans \(V\).
Conversely, if \(p=n\) and the list spans, its coordinate columns span \(\mathbb{F}^n\). The row-reduced matrix cannot have a zero row, since its columns would then all have zero in that position and could not span \(\mathbb{F}^n\). It therefore has a pivot in each of its \(n\) rows. With \(n\) columns, every column is a pivot column, so the homogeneous system has only the zero solution. The list is independent. If \(n=0\), then \(V=\{\mathbf{0}\}\). The only list of length \(n\) is the empty list, which is independent and spans \(V\); any nonempty list consists of zero vectors and is dependent. The conclusions also hold in this case. \(\square\)
The proof uses only coordinates and the pivot structure of row reduction. It does not require determinants or a rank-nullity argument. Its key point is that \(n\) coordinates provide at most \(n\) pivot positions: too many columns leave a free variable, while \(n\) independent columns leave none and can produce every coordinate vector.
Too Many Vectors Force Dependence
Worked Example: Three Vectors in a Two-Dimensional Space
Consider \(\mathbf{v}_1=(1,0)^\top\), \(\mathbf{v}_2=(0,1)^\top\), and \(\mathbf{v}_3=(1,1)^\top\) in \(\mathbb{R}^2\). Since \(3>2\), the theorem guarantees dependence. The coordinate matrix and its reduction show a specific relation:
Columns 1 and 2 are pivot columns, and column 3 has no pivot. The homogeneous equations are \(c_1+c_3=0\) and \(c_2+c_3=0\). Set \(c_3=1\); then \(c_1=-1\) and \(c_2=-1\). Thus \(-\mathbf{v}_1-\mathbf{v}_2+\mathbf{v}_3=\mathbf{0}\).
Check: \(-(1,0)^\top-(0,1)^\top+(1,1)^\top=(-1+0+1,\;0-1+1)^\top=(0,0)^\top\). The nonzero coefficients certify dependence, as the dimension count predicts.
In a proof, the count \(p>n\) establishes that some dependence relation exists, but it does not identify the relation. Row reduction can provide that extra information. The count is most useful when the desired conclusion is simply that a list cannot be independent.
An Independent List of the Right Size Spans
Worked Example: Independence Gives Spanning in \(\mathbb{R}^3\)
Take \(\mathbf{v}_1=(1,0,1)^\top\), \(\mathbf{v}_2=(0,1,1)^\top\), and \(\mathbf{v}_3=(1,1,0)^\top\). Their coordinate matrix row-reduces to the identity, so all three columns are pivots:
Scale row 3 by \(-\tfrac12\), then clear the entries above its pivot:
The identity form shows that the columns are independent. There are exactly \(3=\dim\mathbb{R}^3\) of them, so the theorem says they also span \(\mathbb{R}^3\). To check this conclusion on a particular vector, take \(\mathbf{b}=(2,-1,3)^\top\). Solving \(\mathbf{A}\mathbf{c}=\mathbf{b}\) by applying the same row operations to the right-hand side gives:
Thus \((c_1,c_2,c_3)=(3,0,-1)\). Check: \(3\mathbf{v}_1+0\mathbf{v}_2-\mathbf{v}_3=(3,0,3)^\top-(1,1,0)^\top=(2,-1,3)^\top=\mathbf{b}\). This verifies the coefficients for the chosen vector; the theorem establishes spanning for every vector.
Why the Number Alone Is Not Enough
The equivalence in the theorem is conditional: when the list has exactly \(n\) vectors, independence and spanning each imply the other. It does not say that any list of \(n\) vectors is a basis. A list can have the right size and still fail both properties.
Worked Example: Two Vectors Do Not Automatically Form a Basis
Consider \(\mathbf{u}_1=(1,0)^\top\) and \(\mathbf{u}_2=(2,0)^\top\) in \(\mathbb{R}^2\). The list has exactly \(2=\dim\mathbb{R}^2\) vectors, but its coordinate matrix reduces as follows:
There is no pivot in column 2, and indeed \(-2\mathbf{u}_1+\mathbf{u}_2=\mathbf{0}\), so the list is dependent. It also fails to span: every combination is \((c_1+2c_2,0)^\top\), whose second coordinate is \(0\). In particular, \((0,1)^\top\) is not a combination of the list.
Check: \(-2(1,0)^\top+(2,0)^\top=(-2+2,0)^\top=(0,0)^\top\). Having the dimension’s number of vectors does not itself supply either independence or spanning; one of those properties must be established.
A Proof Strategy and a Common Pitfall
When dimension counts are useful, first identify the space and its dimension, then count the vectors in the list and match the count to the conclusion you need. A standard strategy is:
Fix a basis of the finite-dimensional space and write \(n=\dim V\).
For \(p>n\), conclude that the list is dependent. For an independent list with \(p=n\), conclude that it spans.
If \(p=n\) but neither independence nor spanning is known, the count alone does not establish a basis.
Common mistake: treating the right size as proof of a basis. A wrong argument is: “There are two vectors in \(\mathbb{R}^2\), so they form a basis.” For the vectors \(\mathbf{u}_1=(1,0)^\top\) and \(\mathbf{u}_2=(2,0)^\top\), that conclusion fails: \(\mathbf{u}_2=2\mathbf{u}_1\), and no combination has a nonzero second coordinate. The correction is to establish independence or spanning as well as count the vectors. With exactly \(n\) entries, either property implies the other; the count by itself implies neither.
There is also a useful direction check. “More vectors than the dimension implies dependence” is not the same claim as “fewer vectors than the dimension implies dependence.” In fact, fewer vectors can be independent, as \((1,0)^\top\) is in \(\mathbb{R}^2\). The complementary count applies to spanning: fewer than \(n\) vectors cannot span an \(n\)-dimensional space.
| List condition in an \(n\)-dimensional space | What follows |
|---|---|
| \(p>n\) | The list is dependent. |
| \(p<n\) | The list cannot span the space. |
| \(p=n\) and the list is independent | The list spans, so it is a basis. |
| \(p=n\) and the list spans | The list is independent, so it is a basis. |
| \(p=n\) with no further information | No conclusion that it is a basis follows from the count alone. |
These dimension arguments turn a numerical comparison into a proof, but only after the relevant hypothesis—independence or spanning—has been matched to the right conclusion. When a count settles the question, there is no need to compute every combination; when it does not, coordinates and row reduction can supply the missing evidence.
Check Your Understanding
For each question, identify the dimension and the list size before deciding what the count proves.
- Can a list of \(6\) vectors in a \(4\)-dimensional space be linearly independent? State the dimension-count reason.
- A list of \(3\) vectors in a \(3\)-dimensional space is independent. What can you conclude about spanning and basis?
- Can \(4\) vectors span a \(5\)-dimensional space? Explain using the theorem.
- Give an example of two vectors in \(\mathbb{R}^2\) that have the correct list size but do not form a basis. Verify one failed property.
- In the zero space \(V=\{\mathbf{0}\}\), what does the theorem say about a nonempty list of vectors in \(V\)?
