Let $*:[0,1]\times[0,1]\to[0,1]$ be a continuous $t$-norm i.e.

a) $*$ is continuous,

b) $*$ is commutative and associative,

c) $1*a=a~\forall~a\in[0,1],$

d) $a\le b,c\le d\implies a*c\le b*d.$

How to show that given $\epsilon\in(0,1)~\exists~\delta\in(0,1)$ such that $\delta*\delta*\delta<\epsilon?$

Please help.

  • $\begingroup$ Choose $\delta < \epsilon < 1$ then $(\delta*\delta)*\delta \le (\delta*\delta)*1 =1*(\delta*\delta)=\delta*\delta \le \delta*1 = 1*\delta = \delta < \epsilon$. $\endgroup$ – fleablood Mar 31 '19 at 16:37

You can always choose some $\delta>0$ such that $\delta < \epsilon$. We also know that $\delta < 1$. Hence, according to your composition rule (d) of $*$, we get $\delta * \delta< \epsilon * 1=\epsilon$. Repeat this to get the required result.


More generally, to show that, for any $1 > e > 0$, for any integer $n \ge 2$, ther is a $ d > 0$ such that $ d^n < e$.

Note: $d$ and $e$ are easier to enter than $\delta$ and $\epsilon$.

Let $c = \frac1{e}-1$, so that $e = \frac1{1+c}$. Then $c > 0$ since $0 < e < 1$.

By Bernoulli's inequality, if $n \ge 2$, $(1+\frac{c}{n})^n \gt 1+c$, so $e =\frac1{1+c} \gt \frac1{(1+\frac{c}{n})^n} = (\frac1{1+\frac{c}{n}})^n $.

Therefore $\frac1{1+\frac{c}{n}}$ will work.

Note. To prove Bernoulli's inequality in the form if $x > 0$ and $n \ge 2$ then $(1+x)^n > 1+nx$.

For $n=2$, $(1+x)^2 =1+2x+x^2 \gt 1+2x$.

If true for $n \ge 2$, then

$\begin{array}\\ (1+x)^{n+1} &=(1+x)(1+x)^n\\ &>(1+x)(1+nx)\\ &=1+(n+1)x+nx^2\\ &>1+(n+1)x\\ \end{array} $


Just to be clear: If we define $k^n:= \underbrace{k*k*...*}_{k\text{ times}}$ it's fairly straight forward to prove by induction that if $k \in [0,1]$ and if:

$k^n \le k \le 1$ (which we know is true for $n = 1$)

Then as $k^n \le k^n$ and $k\le 1$ < $k^{n+1} = k^n*k\le k^n*1=1*k^n = k^n\le k \le 1$.

So $1 \ge k \ge k^2 \ge k^3 \ge.....$.

This reduces the problem to the trivial: For any $0< \epsilon < 1$ we can find $0 < \delta < \epsilon$ and $\delta*\delta*\delta \le \delta < \epsilon$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.