A sequence of functions can behave well at every individual point and still behave badly as a family. The pointwise Cauchy condition watches each input $x \in E$ separately; it permits the index at which stability begins to depend on $x$. Uniform Cauchyness forbids that dependence. It asks for a single index after which all functions in the tail are close at every point of the domain at once. [example: Pointwise Stability Without Uniform Stability] Let $E=(0,1)$ and define $f_n:(0,1)\to \mathbb{R}$ by $f_n(x)=x^n$. Fix $x \in (0,1)$. Since $0<x<1$, the geometric sequence $(x^k)_{k=1}^{\infty}$ converges to $0$, so for every $\varepsilon>0$ there is $K \in \mathbb{N}$ such that $x^k<\varepsilon/2$ whenever $k \ge K$. If $m,n \ge K$, then $x^m\ge 0$ and $x^n\ge 0$, and the triangle inequality gives \begin{align*} |f_n(x)-f_m(x)|=|x^n-x^m|\le |x^n|+|x^m|=x^n+x^m<\varepsilon/2+\varepsilon/2=\varepsilon. \end{align*} Thus $(f_n(x))_{n=1}^{\infty}$ is Cauchy for each fixed $x\in(0,1)$. The index $K$ may depend on the chosen point $x$, and that dependence is exactly what prevents uniform Cauchyness. To see the failure of uniform Cauchyness, take $\varepsilon=1/5$. Let $N\in\mathbb{N}$ be arbitrary, and compare the two tail terms $f_N$ and $f_{2N}$. For $x\in(0,1)$, we have $0<x^N<1$, so multiplying $x^N<1$ by the positive number $x^N$ gives $x^{2N}<x^N$. Hence \begin{align*} |f_N(x)-f_{2N}(x)|=|x^N-x^{2N}|=x^N-x^{2N}. \end{align*} Choose $x=2^{-1/N}$. Since $2^{-1/N}>0$ and $2^{-1/N}<1$, this point lies in $(0,1)$. Also, \begin{align*} x^N=(2^{-1/N})^N=2^{-1}=1/2. \end{align*} Therefore \begin{align*} x^{2N}=(x^N)^2=(1/2)^2=1/4. \end{align*} At this point $x=2^{-1/N}$, the tail difference is \begin{align*} |f_N(x)-f_{2N}(x)|=x^N-x^{2N}=1/2-1/4=1/4>1/5. \end{align*} Thus for every proposed tail index $N$, the indices $N$ and $2N$ are both at least $N$, but there is a point $x\in(0,1)$ where $|f_N(x)-f_{2N}(x)|>1/5$. No single tail index can control all points of $E$ at once, so the sequence is pointwise Cauchy but not uniformly Cauchy on $(0,1)$. [/example] This example is the basic warning: a limit constructed point by point need not be a limit in the function space one actually wants to use. Uniform Cauchyness is the Cauchy condition appropriate to the sup norm. It is especially useful because it talks only about the sequence itself, not about a candidate limit that may not yet be known. ## Definition The ordinary Cauchy condition for numbers compares $a_n$ and $a_m$ once $m,n$ are large. For functions, there are many comparisons to make, one at each point of the domain. If the tail index may depend on the point, the family can keep moving its bad behaviour around; the opening example does exactly that. The definition below removes that escape route by requiring one tail index to work simultaneously across the domain. [definition: Uniformly Cauchy Sequence] Let $E$ be a set, let $(Y,d_Y)$ be a metric space, and let $(f_n)_{n=1}^{\infty}$ be a sequence of functions $f_n:E \to Y$. The sequence $(f_n)_{n=1}^{\infty}$ is uniformly Cauchy on $E$ if for every $\varepsilon>0$ there exists $N \in \mathbb{N}$ such that, for all $m,n \ge N$ and all $x \in E$, \begin{align*} d_Y(f_n(x),f_m(x))<\varepsilon. \end{align*} [/definition] When $Y=\mathbb{R}$ or $Y=\mathbb{C}$, the metric is usually the absolute value metric, so the condition becomes $|f_n(x)-f_m(x)|<\varepsilon$ for all $x \in E$. The order of the quantifiers is the substance of the definition: $N$ is chosen after $\varepsilon$, but before $m,n$, and before $x$. Throughout the page, a sequence of functions on $E$ means a sequence $(f_n)_{n=1}^{\infty}$ in which every $f_n$ has the same domain $E$ and the same codomain. [example: Shrinking Linear Functions Are Uniformly Cauchy] Let $E=[0,1]$ and define $f_n:[0,1]\to \mathbb{R}$ by $f_n(x)=x/n$. We show that for every $\varepsilon>0$ there is one tail index $N$ that controls $|f_n(x)-f_m(x)|$ for all $x\in[0,1]$ at once. Fix $\varepsilon>0$. Choose $N\in\mathbb{N}$ such that $N>2/\varepsilon$. Let $m,n\ge N$ and let $x\in[0,1]$. Then \begin{align*} |f_n(x)-f_m(x)|=\left|\frac{x}{n}-\frac{x}{m}\right|. \end{align*} Since $\frac{x}{n}=x\cdot\frac{1}{n}$ and $\frac{x}{m}=x\cdot\frac{1}{m}$, distributivity gives \begin{align*} \frac{x}{n}-\frac{x}{m}=x\left(\frac{1}{n}-\frac{1}{m}\right). \end{align*} Therefore, using $|ab|=|a||b|$, \begin{align*} \left|\frac{x}{n}-\frac{x}{m}\right|=\left|x\left(\frac{1}{n}-\frac{1}{m}\right)\right|=|x|\left|\frac{1}{n}-\frac{1}{m}\right|. \end{align*} Because $x\in[0,1]$, we have $|x|=x\le 1$, so \begin{align*} |x|\left|\frac{1}{n}-\frac{1}{m}\right|\le \left|\frac{1}{n}-\frac{1}{m}\right|. \end{align*} By the triangle inequality for the absolute value, \begin{align*} \left|\frac{1}{n}-\frac{1}{m}\right|=\left|\frac{1}{n}+\left(-\frac{1}{m}\right)\right|\le \left|\frac{1}{n}\right|+\left|-\frac{1}{m}\right|. \end{align*} Since $m,n\in\mathbb{N}$, both $1/n$ and $1/m$ are positive, and hence \begin{align*} \left|\frac{1}{n}\right|+\left|-\frac{1}{m}\right|=\frac{1}{n}+\frac{1}{m}. \end{align*} Because $m,n\ge N$, we have $1/n\le 1/N$ and $1/m\le 1/N$. Combining the estimates, \begin{align*} |f_n(x)-f_m(x)|\le \frac{1}{n}+\frac{1}{m}\le \frac{1}{N}+\frac{1}{N}=\frac{2}{N}. \end{align*} The choice $N>2/\varepsilon$ implies $2/N<\varepsilon$, so \begin{align*} |f_n(x)-f_m(x)|<\varepsilon. \end{align*} The index $N$ was chosen using only $\varepsilon$, not $x$, $m$, or $n$ beyond the tail condition $m,n\ge N$. Hence $(f_n)_{n=1}^{\infty}$ is uniformly Cauchy on $[0,1]$. [/example] The computation illustrates the main technique: estimate the difference by a bound independent of $x$. The moment the bound still depends on $x$, the argument has not yet proved uniform Cauchyness. ## Quantifiers and Sup Norm Geometry ### Uniform Quantifiers The definition looks small, but most errors with uniform Cauchyness are quantifier errors. Pointwise Cauchyness permits $N$ to depend on $x$. Uniform Cauchyness does not. This difference is invisible if one tests only finitely many points, but it becomes decisive on infinite domains. ### Sup Norm Viewpoint A sequence of functions may also be viewed as a sequence in a metric space of functions. This viewpoint explains why the definition is not a new kind of convergence but the usual Cauchy condition applied to a stronger metric. A compact way to read uniform Cauchyness is to measure the worst possible distance between two functions over the whole domain. This gives a metric on any class of functions for which that worst-case distance is always finite, and it gives an extended-valued comparison on the larger class of all functions. [definition: Uniform Distance] Let $E$ be a nonempty set, let $(Y,d_Y)$ be a metric space, and let $\mathcal{F} \subset Y^E$ be a set of functions such that \begin{align*} \sup_{x \in E} d_Y(f(x),g(x))<\infty \end{align*} for all $f,g \in \mathcal{F}$. The uniform distance on $\mathcal{F}$ is the map $d_\infty:\mathcal{F}\times \mathcal{F}\to [0,\infty)$ defined by \begin{align*} d_\infty(f,g)=\sup_{x \in E} d_Y(f(x),g(x)). \end{align*} [/definition] With this notation, uniform Cauchyness says exactly that $d_\infty(f_n,f_m) \to 0$ as $m,n \to \infty$, provided the sequence lies in a function class where the relevant uniform distances are finite. The original quantifier definition remains the safer primary definition because it does not first require choosing such a finite-distance function class. [definition: Sup Norm] Let $E$ be a nonempty set and let $B(E,\mathbb{R})$ be the [vector space](/page/Vector%20Space) of bounded functions $f:E \to \mathbb{R}$. The sup norm is the map $\|\cdot\|_\infty:B(E,\mathbb{R})\to [0,\infty)$ defined by \begin{align*} \|f\|_\infty=\sup_{x \in E}|f(x)|. \end{align*} [/definition] For real-valued bounded functions, uniform distance is just the sup norm of the difference: $d_\infty(f,g)=\|f-g\|_\infty$. This observation is useful only if it is made precise: it tells us when a quantified condition over points can be replaced by a single metric condition in a function space. The next theorem gives that translation, so later completeness arguments can use ordinary [Cauchy sequence](/page/Cauchy%20Sequence) language without losing the uniform quantifier. [quotetheorem:9530] The theorem converts a pointwise-looking condition into ordinary metric geometry. It lets all the usual language of Cauchy sequences, complete metric spaces, and Banach spaces apply directly to function sequences. ### Moving Bad Sets A common failure is to prove that for each fixed $x$, the tail is Cauchy, and then stop. The next example shows why the missing uniformity cannot be repaired after the fact. [example: Moving Transition Layer] Let $E=(0,1)$ and, for each $n \in \mathbb{N}$, define $f_n:(0,1)\to \mathbb{R}$ by $f_n(x)=x^n$. For each fixed $x\in(0,1)$, the powers of $x$ tend to $0$: if $\varepsilon>0$, choose $K\in\mathbb{N}$ with $K>\log(\varepsilon)/\log(x)$ when $0<\varepsilon<1$, and take $K=1$ when $\varepsilon\ge 1$. Since $\log(x)<0$, for $k\ge K$ this gives $k\log(x)<\log(\varepsilon)$, hence $x^k=e^{k\log(x)}<\varepsilon$. Thus $(f_n(x))_{n=1}^{\infty}$ is Cauchy for each fixed $x\in(0,1)$. We compute the size of the tail difference between $f_n$ and $f_{2n}$. Fix $n\in\mathbb{N}$ and let $x\in(0,1)$. Since $0<x<1$, multiplying the inequalities $0<x<1$ together $n$ times gives $0<x^n<1$. Multiplying $x^n<1$ by the positive number $x^n$ gives \begin{align*} x^{2n}=(x^n)^2<x^n. \end{align*} Therefore $x^n-x^{2n}>0$, and so \begin{align*} |x^n-x^{2n}|=x^n-x^{2n}. \end{align*} Put $t=x^n$. Then $0<t<1$, and \begin{align*} x^n-x^{2n}=t-t^2. \end{align*} Conversely, if $t\in(0,1)$ and $x=t^{1/n}$, then $x\in(0,1)$ because $0<t<1$, and \begin{align*} x^n=(t^{1/n})^n=t. \end{align*} Thus the values of $x^n$ as $x$ ranges over $(0,1)$ are exactly the values of $t$ in $(0,1)$, so \begin{align*} \sup_{x\in(0,1)}|x^n-x^{2n}|=\sup_{0<t<1}(t-t^2). \end{align*} For $0<t<1$, completing the square gives \begin{align*} \frac{1}{4}-\left(t-\frac{1}{2}\right)^2=\frac{1}{4}-\left(t^2-t+\frac{1}{4}\right)=t-t^2. \end{align*} Since $\left(t-\frac{1}{2}\right)^2\ge 0$, this identity gives \begin{align*} t-t^2=\frac{1}{4}-\left(t-\frac{1}{2}\right)^2\le \frac{1}{4}. \end{align*} At $t=1/2$, which lies in $(0,1)$, \begin{align*} t-t^2=\frac{1}{2}-\left(\frac{1}{2}\right)^2=\frac{1}{2}-\frac{1}{4}=\frac{1}{4}. \end{align*} The corresponding point of the original domain is $x=2^{-1/n}$, since $0<2^{-1/n}<1$ and \begin{align*} (2^{-1/n})^n=2^{-1}=\frac{1}{2}. \end{align*} Hence the upper bound $1/4$ is actually attained, and \begin{align*} \sup_{x\in(0,1)}|x^n-x^{2n}|=\frac{1}{4} \end{align*} for every $n\in\mathbb{N}$. So even though each fixed point has a Cauchy value sequence, the pair $f_n,f_{2n}$ stays separated by $1/4$ somewhere in the domain. The separating point is $x_n=2^{-1/n}=\exp(-(\log 2)/n)$; because $(\log 2)/n\to 0$, we have $-(\log 2)/n\to 0$, hence $x_n\to e^0=1$. The bad location moves toward the endpoint rather than disappearing uniformly. [/example] The pathology is not that the pointwise limit is mysterious; it is the zero function. The pathology is that the tail never becomes uniformly stable across the whole interval. ## Relationship with Uniform Convergence ### Cauchy Before the Limit Uniform convergence is often introduced first, but uniform Cauchyness is more intrinsic. It does not mention the limit, so it is the right condition when the target space is supposed to produce the limit automatically. [definition: Uniform Convergence] Let $E$ be a set, let $(Y,d_Y)$ be a metric space, let $(f_n)_{n=1}^{\infty}$ be a sequence of functions $f_n:E \to Y$, and let $f:E \to Y$ be a function. The sequence $(f_n)_{n=1}^{\infty}$ converges uniformly to $f$ on $E$ if for every $\varepsilon>0$ there exists $N \in \mathbb{N}$ such that, for all $n \ge N$ and all $x \in E$, \begin{align*} d_Y(f_n(x),f(x))<\varepsilon. \end{align*} [/definition] Uniform convergence immediately forces uniform Cauchyness by comparing $f_n$ and $f_m$ through the common limit. The converse requires the codomain to be complete, because the pointwise limit must exist inside $Y$. [quotetheorem:9531] This theorem is the uniform analogue of the elementary fact that every convergent sequence in a metric space is Cauchy. The stronger direction is the one that matters for constructing limits. ### Completeness of the Codomain Uniform Cauchyness can only build a limit if the codomain has enough points to receive all the pointwise limits. The uniform estimate supplies a common tail bound, but it does not by itself manufacture missing values in an incomplete space. This is why the next criterion names completeness of $Y$ as a hypothesis rather than treating it as background. [quotetheorem:9532] Completeness of $Y$ is not decorative. For each $x \in E$, uniform Cauchyness gives a Cauchy sequence $(f_n(x))_{n=1}^{\infty}$ in $Y$. Completeness is what turns those pointwise Cauchy sequences into actual values of a function $f:E \to Y$. [example: Failure Without a Complete Codomain] Let $E=\{0\}$ and take $Y=\mathbb{Q}$ with the usual metric. Choose a sequence $(q_n)_{n=1}^{\infty}$ of rational numbers such that $q_n\to \sqrt{2}$ in $\mathbb{R}$, and define $f_n:\{0\}\to \mathbb{Q}$ by $f_n(0)=q_n$. We first show that $(f_n)_{n=1}^{\infty}$ is uniformly Cauchy on $E$. Let $\varepsilon>0$. Since $q_n\to \sqrt{2}$ in $\mathbb{R}$, there exists $N\in\mathbb{N}$ such that $|q_k-\sqrt{2}|<\varepsilon/2$ whenever $k\ge N$. If $m,n\ge N$, then the only point of $E$ is $0$, and \begin{align*} |f_n(0)-f_m(0)|=|q_n-q_m|. \end{align*} Writing $q_n-q_m=(q_n-\sqrt{2})+(\sqrt{2}-q_m)$ and applying the triangle inequality gives \begin{align*} |q_n-q_m|\le |q_n-\sqrt{2}|+|\sqrt{2}-q_m|. \end{align*} By the choice of $N$, \begin{align*} |q_n-\sqrt{2}|+|\sqrt{2}-q_m|<\varepsilon/2+\varepsilon/2=\varepsilon. \end{align*} Thus $|f_n(0)-f_m(0)|<\varepsilon$. Since $0$ is the only point of $E$, this same $N$ works for every $x\in E$, so $(f_n)_{n=1}^{\infty}$ is uniformly Cauchy on $E$. Now suppose, for contradiction, that $(f_n)_{n=1}^{\infty}$ converges uniformly to some function $f:E\to\mathbb{Q}$. Write $r=f(0)\in\mathbb{Q}$. Uniform convergence at the single point $0$ says that for every $\varepsilon>0$ there exists $N_1\in\mathbb{N}$ such that $|q_n-r|<\varepsilon/2$ whenever $n\ge N_1$. Since $q_n\to\sqrt{2}$ in $\mathbb{R}$, there exists $N_2\in\mathbb{N}$ such that $|q_n-\sqrt{2}|<\varepsilon/2$ whenever $n\ge N_2$. For any $n\ge \max\{N_1,N_2\}$, the triangle inequality gives \begin{align*} |r-\sqrt{2}|=|r-q_n+q_n-\sqrt{2}|\le |r-q_n|+|q_n-\sqrt{2}|. \end{align*} The two tail estimates give \begin{align*} |r-q_n|+|q_n-\sqrt{2}|<\varepsilon/2+\varepsilon/2=\varepsilon. \end{align*} Hence $|r-\sqrt{2}|<\varepsilon$ for every $\varepsilon>0$. If $r\ne\sqrt{2}$, then $\delta=|r-\sqrt{2}|$ is positive; applying the previous sentence with $\varepsilon=\delta$ would give $\delta<\delta$, impossible. Therefore $r=\sqrt{2}$. It remains to recall why this contradicts $r\in\mathbb{Q}$. Suppose $\sqrt{2}\in\mathbb{Q}$. Then $\sqrt{2}=a/b$ for integers $a,b$ with $b\ne0$ and with $a$ and $b$ having no common integer divisor greater than $1$. Squaring gives \begin{align*} 2=\frac{a^2}{b^2}. \end{align*} Multiplying by $b^2$ gives \begin{align*} a^2=2b^2. \end{align*} Thus $a^2$ is even. If $a$ were odd, then $a=2k+1$ for some integer $k$, and \begin{align*} a^2=(2k+1)^2=4k^2+4k+1=2(2k^2+2k)+1, \end{align*} which is odd. Hence $a$ is even, so $a=2c$ for some integer $c$. Substituting into $a^2=2b^2$ gives \begin{align*} (2c)^2=2b^2. \end{align*} Therefore \begin{align*} 4c^2=2b^2. \end{align*} Dividing by $2$ gives \begin{align*} b^2=2c^2. \end{align*} The same parity argument shows that $b$ is even. Thus both $a$ and $b$ are divisible by $2$, contradicting the choice of $a/b$ in lowest terms. Therefore $\sqrt{2}\notin\mathbb{Q}$, and no uniform limit $f:E\to\mathbb{Q}$ exists. This example isolates the role of completeness: the functions are uniformly Cauchy, but the only possible pointwise limit lies outside the codomain. [/example] This is the smallest possible counterexample, and that is why it is useful. The obstruction has nothing to do with complicated domains or oscillating functions; it is purely the incompleteness of the codomain. ## Preservation of Structure ### Continuous Function Spaces Uniform Cauchyness is valuable because uniform limits preserve many pointwise structures. The most famous preservation result concerns continuity: a uniform limit of continuous functions is continuous. The Cauchy formulation lets us state a completeness theorem for spaces of continuous functions. [definition: Bounded Continuous Function Space] Let $X$ be a nonempty [topological space](/page/Topological%20Space). The space $C_b(X)$ is the set of all bounded continuous functions $f:X \to \mathbb{R}$, equipped with the sup norm $\|\cdot\|_\infty:C_b(X)\to [0,\infty)$ defined by \begin{align*} \|f\|_\infty=\sup_{x \in X}|f(x)|. \end{align*} [/definition] The boundedness assumption is what makes the sup norm finite. Continuity is the structure we want to preserve under limits. The next question is whether these two requirements survive a Cauchy construction: if the tail of a sequence is stable in sup norm, does the limiting object still live in the same space, or must we enlarge the space to include it? The completeness theorem answers that question for $C_b(X)$. [quotetheorem:9533] This theorem says that a uniformly Cauchy sequence of bounded continuous real-valued functions has a bounded continuous uniform limit. The result is one of the main reasons uniform Cauchyness appears in analysis: it provides a usable convergence test inside a function space. ### Series as Cauchy Sequences Many uniform convergence arguments enter through series rather than through a named limiting function. In that setting the natural objects are the partial sums, and the right question is whether their tails become uniformly small. The following example records the standard majorant estimate in Cauchy form. [example: Uniformly Cauchy Series of Functions] Let $E$ be a set, and let $g_k:E\to\mathbb{R}$ satisfy $|g_k(x)|\le a_k$ for every $x\in E$ and every $k\in\mathbb{N}$, where $a_k\ge 0$ and the numerical series $\sum_{k=1}^{\infty}a_k$ converges. Define the partial sums $s_n:E\to\mathbb{R}$ by \begin{align*} s_n(x)=\sum_{k=1}^{n}g_k(x). \end{align*} We show that $(s_n)_{n=1}^{\infty}$ is uniformly Cauchy on $E$. Let $\varepsilon>0$. Since $\sum_{k=1}^{\infty}a_k$ converges, its partial sums \begin{align*} A_n=\sum_{k=1}^{n}a_k \end{align*} form a Cauchy sequence in $\mathbb{R}$. Hence there exists $N\in\mathbb{N}$ such that, whenever $p,q\ge N$, \begin{align*} |A_q-A_p|<\varepsilon. \end{align*} Now let $m,n\ge N$ and let $x\in E$. If $m=n$, then $s_m(x)=s_n(x)$, so \begin{align*} |s_m(x)-s_n(x)|=|0|=0<\varepsilon. \end{align*} Suppose next that $m>n$. Using the definitions of $s_m$ and $s_n$, \begin{align*} s_m(x)-s_n(x)=\sum_{k=1}^{m}g_k(x)-\sum_{k=1}^{n}g_k(x). \end{align*} Split the first finite sum at $n$: \begin{align*} \sum_{k=1}^{m}g_k(x)=\sum_{k=1}^{n}g_k(x)+\sum_{k=n+1}^{m}g_k(x). \end{align*} Therefore \begin{align*} s_m(x)-s_n(x)=\left(\sum_{k=1}^{n}g_k(x)+\sum_{k=n+1}^{m}g_k(x)\right)-\sum_{k=1}^{n}g_k(x)=\sum_{k=n+1}^{m}g_k(x). \end{align*} Applying the finite triangle inequality gives \begin{align*} |s_m(x)-s_n(x)|=\left|\sum_{k=n+1}^{m}g_k(x)\right|\le \sum_{k=n+1}^{m}|g_k(x)|. \end{align*} Since $|g_k(x)|\le a_k$ for each $k$, \begin{align*} \sum_{k=n+1}^{m}|g_k(x)|\le \sum_{k=n+1}^{m}a_k. \end{align*} Because $a_k\ge 0$ and $m>n$, \begin{align*} \sum_{k=n+1}^{m}a_k=\left(\sum_{k=1}^{m}a_k\right)-\left(\sum_{k=1}^{n}a_k\right)=A_m-A_n. \end{align*} Also $A_m-A_n\ge 0$, so $A_m-A_n=|A_m-A_n|$. Since $m,n\ge N$, the choice of $N$ gives \begin{align*} |A_m-A_n|<\varepsilon. \end{align*} Combining these estimates, \begin{align*} |s_m(x)-s_n(x)|\le \sum_{k=n+1}^{m}a_k=A_m-A_n=|A_m-A_n|<\varepsilon. \end{align*} If $n>m$, the same argument with $m$ and $n$ interchanged gives \begin{align*} |s_n(x)-s_m(x)|<\varepsilon. \end{align*} Since $|s_m(x)-s_n(x)|=|-(s_n(x)-s_m(x))|=|s_n(x)-s_m(x)|$, it follows that \begin{align*} |s_m(x)-s_n(x)|<\varepsilon. \end{align*} Thus for every $m,n\ge N$ and every $x\in E$, we have $|s_m(x)-s_n(x)|<\varepsilon$. The index $N$ was chosen from $\varepsilon$ alone, so the sequence of partial sums is uniformly Cauchy on $E$. [/example] This is the Cauchy form of the Weierstrass test. Instead of first naming the limiting function, it proves that the partial sums are stable as functions. ## Pointwise Versus Uniform Control ### Finite Domains Uniform Cauchyness is strongest when the domain is large enough that pointwise testing is misleading. On a finite domain the distinction largely disappears, because finitely many pointwise tail estimates can be combined by taking the maximum of finitely many indices. [quotetheorem:9534] The finite-domain theorem identifies where the difficulty enters: infinitely many points require infinitely many tail estimates, and there may be no largest one to choose. [remark: Uniformity Is a Global Tail Condition] Uniform Cauchyness does not require the functions to be continuous, bounded, monotone, or defined on a metric domain. The domain $E$ can be any set. All metric structure is needed only in the codomain, where function values are compared. [/remark] This remark prevents a common misconception. Compactness of the domain, continuity of the functions, and equicontinuity are powerful hypotheses in other theorems, but they are not part of the definition of uniformly Cauchy. ### Infinite Domains A useful mental model is that uniform Cauchyness controls the whole graph at once. Two late functions may still wiggle, but their graphs must remain inside an $\varepsilon$-tube around each other over the entire domain. [example: Uniform Cauchyness Does Not Require Monotonicity] Let $E=\mathbb{R}$ and, for each $n \in \mathbb{N}$, define $f_n:\mathbb{R}\to \mathbb{R}$ by $f_n(x)=(\sin x)/n$. We show that one tail index controls all $x\in\mathbb{R}$ at once. Let $\varepsilon>0$. Choose $N\in\mathbb{N}$ such that $N>2/\varepsilon$. If $m,n\ge N$ and $x\in\mathbb{R}$, then \begin{align*} |f_n(x)-f_m(x)|=\left|\frac{\sin x}{n}-\frac{\sin x}{m}\right|. \end{align*} Since $\frac{\sin x}{n}=(\sin x)(1/n)$ and $\frac{\sin x}{m}=(\sin x)(1/m)$, distributivity gives \begin{align*} \frac{\sin x}{n}-\frac{\sin x}{m}=(\sin x)\left(\frac{1}{n}-\frac{1}{m}\right). \end{align*} Using $|ab|=|a||b|$, we get \begin{align*} \left|\frac{\sin x}{n}-\frac{\sin x}{m}\right|=\left|(\sin x)\left(\frac{1}{n}-\frac{1}{m}\right)\right|=|\sin x|\left|\frac{1}{n}-\frac{1}{m}\right|. \end{align*} For every real $x$, $|\sin x|\le 1$, so multiplication by the nonnegative number $\left|\frac{1}{n}-\frac{1}{m}\right|$ gives \begin{align*} |\sin x|\left|\frac{1}{n}-\frac{1}{m}\right|\le \left|\frac{1}{n}-\frac{1}{m}\right|. \end{align*} By the triangle inequality for the absolute value, \begin{align*} \left|\frac{1}{n}-\frac{1}{m}\right|=\left|\frac{1}{n}+\left(-\frac{1}{m}\right)\right|\le \left|\frac{1}{n}\right|+\left|-\frac{1}{m}\right|. \end{align*} Because $m,n\in\mathbb{N}$, both $1/n$ and $1/m$ are positive, hence \begin{align*} \left|\frac{1}{n}\right|+\left|-\frac{1}{m}\right|=\frac{1}{n}+\frac{1}{m}. \end{align*} Since $m,n\ge N$, we have $1/n\le 1/N$ and $1/m\le 1/N$. Combining the estimates gives \begin{align*} |f_n(x)-f_m(x)|\le \frac{1}{n}+\frac{1}{m}\le \frac{1}{N}+\frac{1}{N}=\frac{2}{N}. \end{align*} The choice $N>2/\varepsilon$ implies $2/N<\varepsilon$, so \begin{align*} |f_n(x)-f_m(x)|<\varepsilon. \end{align*} The index $N$ depends only on $\varepsilon$, not on $x$, so $(f_n)_{n=1}^{\infty}$ is uniformly Cauchy on $\mathbb{R}$. The functions still oscillate, but their amplitudes shrink uniformly across the whole unbounded domain. [/example] The example separates two issues that are often conflated. An unbounded domain is not itself a problem; the problem is lack of a bound independent of the point in the domain. ## Beyond and Connected Topics Uniformly Cauchy sequences sit between elementary sequence theory and the functional-analytic language of complete spaces. The next natural topic is [Uniform Convergence](/page/Uniform%20Convergence), where the limiting function is part of the statement rather than something inferred from a Cauchy criterion. They also belong naturally to [Metric Space](/page/Metric%20Space) theory. In that setting, the page becomes an instance of the general principle that completeness is exactly what makes Cauchy sequences converge. In analysis, uniformly Cauchy sequences appear constantly in function-space arguments, especially when proving that spaces such as $C_b(X)$ or spaces equipped with sup norms are complete. They also prepare the ground for stronger compactness theorems, where a sequence may fail to be uniformly Cauchy but contains a uniformly convergent subsequence after additional hypotheses. The failure examples point toward equicontinuity and compactness. A moving transition layer such as $x^n$ on $(0,1)$ shows that pointwise convergence alone does not control where variation concentrates. ## References Androma, [Metric Space](/page/Metric%20Space). Androma, [Uniform Convergence](/page/Uniform%20Convergence). Androma, [Cambridge IB Analysis and Topology](/page/Cambridge%20IB%20Analysis%20and%20Topology). Androma, [Cambridge IA Analysis Notes](/page/Cambridge%20IA%20Analysis%20Notes). Androma, [Cambridge IB Complex Analysis](/page/Cambridge%20IB%20Complex%20Analysis). Androma, [Cambridge II Analysis of Functions](/page/Cambridge%20II%20Analysis%20of%20Functions). Walter Rudin, *Principles of Mathematical Analysis* (1976). John B. Conway, *A Course in Functional Analysis* (1990).