0
$\begingroup$

I'm trying to figure out how to best formalize the following interrelationship in first order logic:

A material has (electric) resistance $r$ and conductance $g$, and the two are related as $r \cdot g = 1$.

The idea is that, if I know either $r$ or $g$ for some material, I should be able to infer the other.

Let the relation $\mathrm{res}(m,r)$ mean that material $m$ has resistance $r$, and similarly $\mathrm{con}(m,g)$ for conductance, and let $\mathrm{mul}(x,y,z)$ denote the multiplication relation $x \cdot y = z$. The following formulas express the semantics of resistance/conductance:

  1. $\mathrm{res}(m,r) \land \mathrm{con}(m,g) \rightarrow \mathrm{mul}(r,g,1)$
  2. $\mathrm{res}(m,r) \land \mathrm{mul}(r,g,1) \rightarrow \mathrm{con}(m,g)$
  3. $\mathrm{con}(m,g) \land \mathrm{mul}(r,g,1) \rightarrow \mathrm{res}(m,r)$

I think this works as expected: for example, if I know $\mathrm{res}(X,10)$ for some specific material $X$ (a constant), then rule (2) together with the fact $\mathrm{mul}(10,0.1,1)$ gives $\mathrm{con}(X,0.1)$. However, it feels like we should be able to express the relation between $r$ and $g$ more compactly$-$do we really need three formulas for this?

Intuitively, I thought there should be a single formula describing the relation, but I can't find one that works. Am I missing something, or is there something lurking here that's beyond the expressivity of first order logic?

$\endgroup$
2
$\begingroup$

Are you somehow not allowing your logical language to contain function symbols?

I would write $$ \forall x(\operatorname{material}(x) \to \operatorname{res}(x)\cdot\operatorname{con}(x) = 1)$$ and be done with it.

(Even though first-order logic can be formulated with no function symbols without losing expressivity, the standard presentations allow any combination of predicate symbols, function symbols and constant symbols).

$\endgroup$
0
$\begingroup$

If I understood you correctly, this relationship could be written just as \begin{equation} res(m,r) \leftrightarrow con(m,\frac{1}{r}) \end{equation}

EDIT (after discussion in the comments): \begin{equation} (res(m,r) \land con(m,g)) \leftrightarrow mul(r,g,1) \end{equation} For a specific material (say $m=X$), we have the following;

  • If $res(X,r)$ holds, $con(X,g)$ must hold for $g$ such that $mul(r,g,1)$ holds.
  • If $con(X,g)$ holds, $res(X,r)$ must hold for $r$ such that $mul(r,g,1)$ holds.

I am aware that this lacks the restriction on $m$, but I figured that this formula is not to be used without knowing either $r$ or $g$ of a specific material $m$.

$\endgroup$
8
  • $\begingroup$ Can you give some explanation? $\endgroup$ – user144410 Aug 19 '18 at 10:16
  • $\begingroup$ Following what @Roland said: "..., if I know either r or g for some material, I should be able to infer the other". So, if we know the resistance of $m$ is $r$, than conductance of $m$ must be $1/r$ and vice-versa. Therefore I put logical equivalence, that covers both ways of implication. $\endgroup$ – Sandro Lovnički Aug 19 '18 at 10:20
  • $\begingroup$ Well, that seems simple, but it doesn't look like first order logic. What exactly is the $1/r$ construct? If it's a function (taking terms to terms) then you cannot prove $res(X, 10)$ from $con(X, 0.1)$, because there's no way to unify $0.1$ with the function to get to the value of $r$. I'm pretty sure the inverse must be handled implicitly via a relation. $\endgroup$ – Roland Aug 19 '18 at 10:37
  • $\begingroup$ You are right @Roland, it cannot be done with less than 3 relations. Can it be done if we add your $mul$ relation to above biconditional formula? What are the problems then? $\endgroup$ – Sandro Lovnički Aug 19 '18 at 11:01
  • 2
    $\begingroup$ @SandroLovnički: Your proposal would seem to say that whenever we can find an $r$ and $g$ that multiply to $1$, those particular $r$ and $g$ will magically be parameters of every $m$ in the universe ... $\endgroup$ – hmakholm left over Monica Aug 19 '18 at 13:19

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.