Componentwise Continuity

Why Coordinates Can Simplify Continuity

The previous tutorial defined continuity by comparing the distance between two inputs with the distance between their outputs. When the output lies in \(\mathbb{R}^n\), that output distance combines changes in all \(n\) coordinates. It is often simpler to study those scalar changes separately. Because there are only finitely many output coordinates, control of every coordinate is equivalent to control of the full vector.

Let \(D\subseteq\mathbb{R}^m\), and write a function \(f:D\to\mathbb{R}^n\) in coordinates as \(f(x)=(f_1(x),\ldots,f_n(x))\), where each \(f_i:D\to\mathbb{R}\). The functions \(f_i\) are the component functions of \(f\). Continuity of a component is understood using the Euclidean metric on the domain and the usual absolute value on \(\mathbb{R}\).

The key estimates come from the Euclidean norm. For any \(v=(v_1,\ldots,v_n)\in\mathbb{R}^n\), each coordinate satisfies \(|v_i|\leq\|v\|_2\), while the sum of the squared coordinates gives \(\|v\|_2^2=\sum_{i=1}^n|v_i|^2\). These estimates let us pass in both directions between coordinate errors and the total vector error.

The Componentwise Criterion

Proof. Suppose first that \(f\) is continuous at \(a\). Fix an index \(i\) and let \(\varepsilon>0\). By continuity of \(f\), there is a \(\delta>0\) such that, for every \(x\in D\), \[ \|x-a\|_2<\delta \quad\Longrightarrow\quad \|f(x)-f(a)\|_2<\varepsilon. \] The \(i\)-th coordinate difference is bounded by the Euclidean norm of the whole difference vector: \[ |f_i(x)-f_i(a)|\leq\|f(x)-f(a)\|_2. \] Thus \(\|x-a\|_2<\delta\) implies \(|f_i(x)-f_i(a)|<\varepsilon\). This proves that \(f_i\) is continuous at \(a\). Since \(i\) was arbitrary, every component is continuous there.

Conversely, suppose each component \(f_i\) is continuous at \(a\), and let \(\varepsilon>0\). For each \(i\in\{1,\ldots,n\}\), continuity supplies a number \(\delta_i>0\) such that, for every \(x\in D\), \[ \|x-a\|_2<\delta_i \quad\Longrightarrow\quad |f_i(x)-f_i(a)|<\frac{\varepsilon}{\sqrt{n}}. \] There are finitely many components, so the minimum \(\delta=\min\{\delta_1,\ldots,\delta_n\}\) is positive. If \(x\in D\) and \(\|x-a\|_2<\delta\), then all \(n\) coordinate estimates hold. Consequently, \[ \|f(x)-f(a)\|_2^2 =\sum_{i=1}^{n}|f_i(x)-f_i(a)|^2 <n\left(\frac{\varepsilon}{\sqrt{n}}\right)^2 =\varepsilon^2. \] Both sides are nonnegative, so \(\|f(x)-f(a)\|_2<\varepsilon\). Hence \(f\) is continuous at \(a\). \(\square\)

The proof uses the fact that the output has finitely many coordinates: it takes the minimum of finitely many positive radii. It also shows why the factor \(\sqrt n\) is convenient. Requiring each coordinate error to be less than \(\varepsilon/\sqrt n\) makes the sum of their squares less than \(\varepsilon^2\).

Proof. Apply the Componentwise Criterion at each \(a\in D\). The statement on the whole domain means exactly that the pointwise condition holds for every such \(a\). \(\square\)

Coordinate Projections and Component Functions

The component functions can also be obtained by composing \(f\) with coordinate projections. For \(i\in\{1,\ldots,n\}\), define \(\pi_i:\mathbb{R}^n\to\mathbb{R}\) by \(\pi_i(y_1,\ldots,y_n)=y_i\). Then \(f_i=\pi_i\circ f\). The projections themselves are continuous.

Proof. Fix \(y\in\mathbb{R}^n\) and let \(\varepsilon>0\). If \(\|z-y\|_2<\varepsilon\), then \[ |\pi_i(z)-\pi_i(y)|=|z_i-y_i|\leq\|z-y\|_2<\varepsilon. \] Thus \(\delta=\varepsilon\) works in the definition of continuity at \(y\). Since \(y\) was arbitrary, \(\pi_i\) is continuous everywhere. \(\square\)

This observation connects the componentwise criterion with the Composition of Continuous Functions theorem from the previous tutorial. If \(f\) is continuous, then \(\pi_i\circ f\) is continuous, so each component is continuous. Conversely, the Componentwise Criterion assembles the continuous scalar component functions into a continuous vector-valued function.

A Common Pitfall: Output Coordinates Are Not Input Directions

“Componentwise” refers to the coordinates of the output, not merely to testing the function along selected directions in the input space. Checking a function on the coordinate axes, or on a few lines through a point, does not establish continuity there. Those tests examine only some input paths; continuity must control every domain point sufficiently close to the point in question.

Using the Criterion Carefully

For a map into \(\mathbb{R}^n\), the theorem reduces a vector-valued continuity problem to finitely many scalar ones. It is especially useful when the formulas for the component functions are simpler than a direct estimate of the full Euclidean output distance. One may prove each component continuous by an estimate, by a limit argument at an accumulation point, or by composing simpler continuous functions when the hypotheses of the Composition of Continuous Functions theorem apply.

The finiteness of the output dimension matters to the proof: the minimum of finitely many positive radii is positive. The result here concerns maps into \(\mathbb{R}^n\), with finite \(n\), and should not be transferred automatically to an infinite-coordinate setting. Also, continuity of a component means continuity as a function on all of \(D\) near the point, not merely continuity of its restrictions to a few curves or coordinate lines.

When using the theorem at a particular \(a\), verify every component at that same point. When proving continuity on all of \(D\), verify each component at every point of \(D\). The criterion is an equivalence: it provides both a method for proving vector-valued continuity and a way to disprove it, since one discontinuous component is enough to rule out continuity of the full map.