矩阵范数极限证明：如何证明n×n实矩阵满足lim_{V→0}|Tr(V²)|/||V||=0？

Alright, let's break down this proof step by step. The key here relies on two fundamental facts about matrix norms in finite-dimensional spaces and trace inequalities—let's dive in.

Step 1: All matrix norms are equivalent on finite-dimensional spaces #

First, remember that the space of $n \times n$ real matrices is a finite-dimensional vector space (its dimension is $n^2$). A critical property of finite-dimensional normed spaces is that all norms are equivalent. This means for any two matrix norms $|\cdot|$ and $|\cdot|'$, there exist positive constants $C_1, C_2$ such that:
$$C_1|A|' \leq |A| \leq C_2|A|'$$
for every $n \times n$ matrix $A$.

Why does this matter? Because if we can prove the limit holds for one specific norm (like the spectral norm $|\cdot|_2$ or Frobenius norm $|\cdot|_F$), it will automatically hold for any norm—since equivalent norms preserve convergence behavior.

Step 2: Bound $|\text{Tr}(V^2)|$ using the spectral norm #

Let's pick the spectral norm $|\cdot|_2$ (the largest singular value of $V$, equal to the maximum absolute value of its eigenvalues for real matrices). For any $n \times n$ matrix $V$, let $\lambda_1, \lambda_2, ..., \lambda_n$ be its eigenvalues (counted with algebraic multiplicity). By definition:
$$\text{Tr}(V^2) = \lambda_1^2 + \lambda_2^2 + ... + \lambda_n^2$$
Applying the triangle inequality to the absolute value of the trace:
$$|\text{Tr}(V^2)| \leq |\lambda_1|^2 + |\lambda_2|^2 + ... + |\lambda_n|^2$$
Since each $|\lambda_i| \leq |V|_2$ (the spectral norm bounds all eigenvalue magnitudes), we can substitute to get:
$$|\text{Tr}(V^2)| \leq n \cdot |V|_2^2$$
That's our key upper bound for the trace term.

Step 3: Connect to the arbitrary norm and evaluate the limit #

Using norm equivalence, there exists a constant $C > 0$ such that $|V|_2 \leq C|V|$ (since we can relate the spectral norm to our arbitrary norm $|\cdot|$). Substitute this into our bound:
$$|\text{Tr}(V^2)| \leq n \cdot (C|V|)^2 = nC^2 \cdot |V|^2$$
Now divide both sides by $|V|$ (for $V \neq 0$, which is fine since we're taking the limit as $V \to 0$):
$$\frac{|\text{Tr}(V^2)|}{|V|} \leq nC^2 \cdot |V|$$
As $V \to 0$, $|V| \to 0$, so the right-hand side tends to 0. By the squeeze theorem, the left-hand side must also tend to 0.

Alternative approach using the Frobenius norm #

If you prefer working with the Frobenius norm $|\cdot|_F$ (the square root of the sum of squared entries), we can use the Cauchy-Schwarz inequality for the trace inner product. For any matrices $A, B$, $|\text{Tr}(AB)| \leq |A|_F|B|_F$. Applying this to $A = B = V$:
$$|\text{Tr}(V^2)| \leq |V|_F^2$$
Again, using norm equivalence, $|V|_F \leq D|V|$ for some constant $D > 0$, so:
$$\frac{|\text{Tr}(V^2)|}{|V|} \leq \frac{D^2|V|2}{|V|} = D^2|V| \to 0$$
as $V \to 0$. Same conclusion!

内容的提问来源于stack exchange，提问作者Anonmath101