5

How could I derive by modus ponens the formula $$[A\rightarrow(B\rightarrow C)]\rightarrow[(A\rightarrow B)\rightarrow(A\rightarrow C)]$$ from, and just from, the following axiom schemata?

  1. $(A\lor A)\rightarrow A$.
  2. $A\rightarrow(A\lor B)$.
  3. $(A\lor B)\rightarrow(B\lor A)$.
  4. $(A\rightarrow B)\rightarrow[(A\lor C)\rightarrow(B\lor C)]$.
  5. If $A\rightarrow B$ and $B\rightarrow C$, then $A\rightarrow C$.
  6. $A\rightarrow A$.
  7. $A\leftrightarrow \neg(\neg A)$.
  8. $(A\rightarrow B)\leftrightarrow(\neg B\rightarrow \neg A)$.
  9. $(A\land B)\rightarrow A$ and $(A\land B)\rightarrow B$.
  • In an axiomatic system you cannot derive it at all (you need to add it as an axiom) , but if you have a natural deduction system you can derive it from nothing at all (ex nihilo) – Willemien Feb 20 '14 at 00:22
  • Aha! Could you go in depth please? How come is it independent of the other schemata? – Fitzcarraldo Feb 20 '14 at 00:56
  • 1
    @Willemien - sorry: see our prevoius post. The axioms system (1-4) with modus ponens, used by Bourbaki, was introduced in PM (1910), with one additonal axiom; then simplified by Bernays (1918) who proved the fifth axiom to be derivable from the others four. It is complete for propositional calculus with modus ponens; so the above formula is derivable. – Mauro ALLEGRANZA Feb 20 '14 at 08:02
  • You can see Alfred North Whitehead & Bertrand Russell, Principia Mathematica to n°56 (2nd ed - 1927): Part I: Mathematical Logic [page 85-on; you can skip the first 84 pages: it is the Intro to the 2nd ed, of historical and philosophical interset, but does not add nothing to the propositional calculus]. Pages 96-97 contain the axioms (axiom 1.5 has been proved not necessary); at page 108 you will find n°2.77 : $\vdash [p \supset (q \supset r)] \supset [(p \supset q) \supset (p \supset r)]$. All derivations in PM use only modus ponens and substitution: no Deduction Theorem. – Mauro ALLEGRANZA Feb 20 '14 at 08:54
  • @MauroALLEGRANZA sorry i had to think about it but you are wrong (and i can even prove it see answer) – Willemien Feb 21 '14 at 10:11

3 Answers3

5

In an axiomatic system you cannot derive it at all (you need to add it as an axiom) , but if you have a natural deduction system you can derive it from nothing at all (ex nihilo)

1) It is easely proved in a natural deduction system:

1 | |______ P -> (Q -> R))                             Assumption
2 | | |____ P -> Q                                     Assumption
3 | | | |__ P                                          Assumption 
4 | | | |   Q                                          2,3 Modus ponens
5 | | | |   Q -> R                                     1,3 Modus ponens
6 | | | |   R                                          4,5 Modus ponens
. | | | <------------------------------------------------ end subproof 
7 | | |     P -> R                                     3-6 Conditional proof
. | | <-------------------------------------------------- end subproof 
8 | |      (P -> Q) -> (P  -> R)                       2-7 Conditional proof
. | | <-------------------------------------------------- end subproof 
9 | |      (P -> (Q -> R)) -> ((P -> Q) -> (P  -> R))  1-8 Conditional proof
================================
  QED

But still in an axiomatic proof it is not provable.

A bit metalogic:

If an axiomsystem is consistent (truth preserving) then you cannot proof in any way an untrue sentence.

If there is an interpretation of a system (any interpretation at all) where in all models (valuations) all axioms are always true, then all theorems are also always true.

if in an interpretation where in all models(valuations) the axioms are true but there is a model (valuation) where a formula is false then that formula is not a theorem.

Now $ (P \to (Q \to R)) \to ((P \to Q) \to (P \to R)) $ fails in Lukasiewisc 3 valued logic while all other axioms are theorems of Lukasiewisc 3 valued logic.

So the axioms are true in all valuations , but $ (P \to (Q \to R)) \to ((P \to Q) \to (P \to R)) $ is not true in some particular valuation , therefore $ (P \to (Q \to R)) \to ((P \to Q) \to (P \to R)) $ cannot be a theorem of the system.

The proof

Lukasiewicz 3 valued logic consists of the following truthtables:

       ||     ||  A ->  B || A & B || A v B || 
       || Des ||  1 2 3   || 1 2 3 || 1 2 3 || ~A  
 ----- ||-----||----------||-------||-------||--- 
 A = 1 ||  *  ||  1 2 3   || 1 2 3 || 1 1 1 || 3   
 A = 2 ||     ||  1 1 2   || 2 2 3 || 1 2 2 || 2 
 A = 3 ||     ||  1 1 1   || 3 3 3 || 1 2 3 || 1 

A formula is a tautology or theorem in this logic if and only iff in every possible valuation the formula evaluates to 1 (the true or designated value)

All given axioms do this. (check this yourself)

But $ (P \to (Q \to R)) \to ((P \to Q) \to (P \to R)) $ where $ P = Q = 2$ and $ R = 3 $ fails, it evaluates to 2

(P -> (Q -> R)) -> ((P -> Q) -> (P  -> R) )
(2 -> (2 -> 3)) -> ((2 -> 2) -> (2  -> 3) )
(2 -> (  2   )) -> ((  1   ) -> (   2   ) )
(  1          ) -> (         2            )
(               2                         )

therefore $ (P \to (Q \to R)) \to ((P \to Q) \to (P \to R)) $ is not a theorem of this logic , and therefore it can also not be a theorem of the syatem cannot be a theorem of the system.

QED

missing steps in this proof that you will think about yourself

  • are the axioms always true. just a lot of work , instead of a 8 line truthtable you get a 27 line truthtable

  • is the system truthpreserving? this proof is based on a system that uses modus ponens ($ \vdash A, \vdash A \to B $ then $ \vdash B $ ) as inference rule, the system doesn't have this rule it has the hypothetical syllogism inference rule ($ \vdash A \to B , \vdash B \to C $ then $ \vdash A \to C $ ) to me it looks that the hypothectical syllogism inference rule is equivalent or a bit weaker than the modus ponens inference rule, but this is in need of a proof is $ A \to (( A \to B ) \to B) $ provable a theorem in the system

  • what is the differece between an interpretation and a model? (an interpretation is a logic or logical system itself, a model is a particular valuation)

  • but these are things the op may think about

two posters that disagree with me, while i can proof they are wrong.

Willemien
  • 6,730
  • @MauroALLEGRANZA As far as I can tell, that variant is also true in 3 valued logic, so Willemien's argument is still valid. – Hagen von Eitzen Feb 21 '14 at 16:32
  • @Hagen von Eitzen - Thanks; of course, due to the properties of $\lor$, we expect that the two variants are equivalent. Still the problem remains... – Mauro ALLEGRANZA Feb 21 '14 at 16:42
  • @MauroALLEGRANZA It is all a bit the curse of the bi-value that logicism and the Principia Mathematica has put on us, axiomatic logic is free from most assumptions also this one.

    In Hilbert and Ackerman you have skipped page 29 the axioms for the $ \lnot \to $ system, two sets are given and loo and behold $ (P \to ( Q \to R)) \to (( P \to Q) \to (P \to R )) $ is part of both systems.

    – Willemien Feb 21 '14 at 18:00
  • I don't know whether or not this answer is correct, but I'm giving it +1 for effort. – goblin GONE Feb 25 '14 at 10:33
  • Axiom schema 5:) If A→B and B→C, then A→C )also fails in Lukasiewicz 3-valued logic. A different version of the conditional: A => B, using 3 instead of 2 in the cases where A=2,B=3 and A=1, B=2 does succeed. But with this version, so does the formula [A→(B→C)]→[(A→B)→(A→C)]. – Confutus Mar 28 '14 at 22:08
  • 1
    @Confutus under which valuation of A,B, and C does it fail? I just checked all 27 but in my test all evaluations give 1 as answers, (just do something similar as what i did to disaprove the formula but now only with $ (A \to B ) \to ((B \to C ) \to ( A \to C)) $ and just give one valuations where it doesn't elaluate to 1 – Willemien Mar 29 '14 at 12:05
  • Modus ponens itself (A & (A→B )) → B fails as a tautology with a value of 2 when A=2 and B=3. ((A→B ) & (B→C)) → ( A→C) evaluates as 2 when A=1, B=2, and C=3 using the Lukasiewicz conditional. The reason is that it allows doubtful conditionals, when (A→B ) may have a value of 2. If it is not certain that A implies B, unrestricted use of rules such as Modus ponens, the transitive rule, and (A→B)→((B→C)→(A→C)) may allow false conclusions from true premises. Valid rules of inference should of course forbid this. – Confutus Apr 08 '14 at 18:12
  • 1
    @Confutus This will sound crazy completely absurd but you have the wrong modus ponens(yep there are such things different modus ponens. the inference rule modus ponens ($ \vdash A \to B, \vdash A \implies \vdash B $) is a rule not a formula, and because only formula's can be tautologies so it cannot even be a tautology. Also even if you want a formula that "has the idea of modus ponens" try $ ( A \to ((A \to B ) \to B)) $ (aka assertion) which is a tautology. – Willemien Apr 09 '14 at 09:54
  • @Confutus In addition to Willemien's point about inference rules, and Willemien's reference to "assertion", a substitution instance of the law of identity (A→A) "has the idea of modus ponens". If you don't know what I'm referring to, commute the antecedents of the assertion formula. – Doug Spoonwood Apr 27 '14 at 22:37
  • This also implies, given Mauro's comments as correct, that you can't metalogically prove what we call "the deduction theorem" for propositional logic for the definition-free version of the propositional logic system of Principia Mathematica. – Doug Spoonwood Apr 27 '14 at 22:41
1

In my answer below, we can find a "modern" reconstruction of PM's system of propositional logic.

Here we have, with the help of Wiki Principia Mathematica propositional logic , the original proof of PM's

*2.77 --- $(p \rightarrow (q \rightarrow r)) \rightarrow ((p \rightarrow q) \rightarrow (p \rightarrow r))$.


We establish first some useful theorems :

*(2.05) --- $\vdash (q \rightarrow r) \rightarrow [(p \rightarrow q) \rightarrow (p \rightarrow r)]$ --- from Sum with the substitution : $\lnot p/p$.

*(2.06) --- $\vdash (p \rightarrow q) \rightarrow [(q \rightarrow r) \rightarrow (p \rightarrow r)]$ --- from Comm with the substitution : $q \rightarrow p/p; p \rightarrow q/q; p \rightarrow r/r$ and (*2.05) by modus ponens

We call the above theorems : Syll

*(2.36) --- $\vdash (q \rightarrow r) \rightarrow [(p \lor q) \rightarrow (r \lor p)]$ - from Syll with the subst $p \lor q/p; p \lor r/q; r \lor p/r$ and Perm by mp, followed by Sum and Syll

*(2.37) --- $\vdash (q \rightarrow r) \rightarrow [(q \lor p) \rightarrow (p \lor r)]$ - from Sum, Perm and Syll

*(2.38) --- $\vdash (q \rightarrow r) \rightarrow [(q \lor r) \rightarrow (r \lor p)]$ - from Sum, Perm and Syll


Now we write the derivation of (*2.77) :

*2.53 --- $\vdash (p \lor q) \rightarrow (\lnot p \rightarrow q)$.

From (*2.12) [i.e. $\vdash p \rightarrow \lnot \lnot p$] and Syll (*2.38), with the substitution $q/p; p/q; \lnot \lnot p/r$, and mp and definition of $\rightarrow$.

*2.6 --- $\vdash (\lnot p \rightarrow q) \rightarrow ((p \rightarrow q) \rightarrow q)$.

Proof

$\vdash (\lnot p \rightarrow q) \rightarrow ((\lnot p \lor q) \rightarrow (q \lor q))$ --- from Syll (*2.38) with the subst $q/p; \lnot p/q; q/r$

$\vdash [(\lnot p \lor q) \rightarrow (q \lor q)] \rightarrow [(\lnot p \lor q) \rightarrow q]$ --- from Taut, Sum and mp

$\vdash (\lnot p \rightarrow q) \rightarrow ((\lnot p \lor q) \rightarrow q)$ --- from the two formulas above by Syll (*2.06). Using the abbreviation of $\rightarrow$ we arrive at the final version of the theorem.

*2.62 --- $\vdash (p \lor q) \rightarrow ((p \rightarrow q) \rightarrow q)$ --- from (*2.53) and (*2.6) by Syll.

*2.621 --- $\vdash (p \rightarrow q) \rightarrow ((p \lor q) \rightarrow q)$ --- from (*2.62) with Comm and mp.

*2.73 --- $\vdash (p \rightarrow q) \rightarrow [((p \lor q) \lor r) \rightarrow (q \lor r)]$ --- from (*2.621) and (*2.38), with the substitution : $p \lor q/q; q/r; r/p$ and Syll.

*2.74 --- $\vdash (q \rightarrow p) \rightarrow [((q \lor p) \lor r) \rightarrow (p \lor r)]$ --- from (*2.73) with the substitution : $q/p; p/q$.

Note. We need several steps of Perm and Assoc to get from : $((q \lor p) \lor r)$ to $(p \lor (q \lor r))$ inside the formula. We write the first one.

From : $(q \rightarrow p) \rightarrow [((q \lor p) \lor r) \rightarrow (p \lor r)]$ we apply Comm to get : $((q \lor p) \lor r) \rightarrow [(q \rightarrow p) \rightarrow (p \lor r)]$; then we use Assoc and Syll to get : $(r \lor (q \lor p)) \rightarrow [(q \rightarrow p) \rightarrow (p \lor r)]$; and so on, until we arrive at : $(p \lor (q \lor r)) \rightarrow [(q \rightarrow p) \rightarrow (p \lor r)]$.

At the end of the transformations, we apply again Comm :

*2.74 --- $\vdash (q \rightarrow p) \rightarrow [(p \lor (q \lor r)) \rightarrow (p \lor r)]$.

From (*2.74) with the substitution $\lnot q/q$ and (*2.53), with Syll, we get :

$\vdash (q \lor p) \rightarrow [((p \lor (\lnot q \lor r)) \rightarrow (p \lor r)]$.

Using Perm and Syll and the definition of $\rightarrow$ we arrive at the final version :

*2.75 --- $\vdash (p \lor q) \rightarrow [(p \lor (q \rightarrow r)) \rightarrow (p \lor r)]$.

From *2.75 by Comm we derive :

*2.76 --- $\vdash (p \lor (q \rightarrow r)) \rightarrow ((p \lor q) \rightarrow (p \lor r))$.

From *2.76, with the substitution : $\lnot p/p; q/q; r/r$ we get :

*2.77 --- $\vdash (p \rightarrow (q \rightarrow r)) \rightarrow ((p \rightarrow q) \rightarrow (p \rightarrow r))$.


Appendix. We prove the “missing” axiom : Assoc (*1.05) that has been proved to be derivable from the others PM’s axioms by Bernays (in addition to the other four axioms, we will use only Syll) :

$\vdash (p \lor (q \lor r)) \rightarrow ((p \lor q) \lor r))$;

Proof

$\vdash r \rightarrow (p \lor r)$ --- Add (*1.3)

$\vdash r \rightarrow (p \lor r) \rightarrow [q \lor r \rightarrow q \lor (p \lor r)]$ --- from Sum (*1.6) with subst $r/q;p \lor r/r; q/p$

$\vdash q \lor r \rightarrow q \lor (p \lor r)$ --- modus ponens

$\vdash p \lor (q \lor r) \rightarrow p \lor [q \lor (p \lor r)]$ --- from Sum with the subst $p/p; q \lor r/q; q \lor (p \lor r)/r$ and mp

(A) --- $\vdash p \lor (q \lor r) \rightarrow [q \lor (p \lor r)] \lor p$ --- by Perm (*1.4) and Syll (*2.05)

$\vdash p \rightarrow (r \lor p)$ --- Add

$\vdash p \lor r \rightarrow q \lor (p \lor r)$ --- from above with substitution : $p \lor r/p; q/r$

$\vdash p \rightarrow (p \lor r)$ --- from Add (*1.3) and Perm (*1.4) by Syll (*2.05)

$\vdash p \rightarrow q \lor (p \lor r)$ --- from the last two by Syll

$\vdash [q \lor (p \lor r)] \lor p \rightarrow ([q \lor (p \lor r)] \lor [q \lor (p \lor r)])$ --- from the last by Syll

(B) --- $\vdash [q \lor (p \lor r)] \lor p \rightarrow [q \lor (p \lor r)]$ --- from the last using Taut and Syll.

Finally, from (A) and (B) by Syll :

$\vdash p \lor (q \lor r) \rightarrow (q \lor (p \lor r))$.

Having used only (*2.05) in the proof of Assoc, we can use Assoc to prove Comm (*2.04) :

$\vdash (p \rightarrow (q \rightarrow r)) \rightarrow (q \rightarrow (p \rightarrow r))$ --- from Assoc with the substitution : $\lnot p/p; \lnot q/q; r/r$.

Now, having proved Comm, we can go on from (*2.06).

  • 2
    I checked where a disagreement with Willemien's argument creeps in: The step from 2.76 to 2.77 seems to use $\neg p\lor q \iff p\to q$. In 3 valued logic this fails for the case $p=q=2$. – Hagen von Eitzen Feb 21 '14 at 16:42
  • 1
    @MauroALLEGRANZA it is pure axiomatic logic that prevents the proof, it has nothing to do with the deduction theorem, if you have the deduction theorem the first proof i gave does work what do you think of that??? – Willemien Feb 21 '14 at 18:04
  • @MauroALLEGRANZA what do you mean by "in the "frame" of classical logic, " in axiomatic logic there is no "frame" . also you have to take the "sound and complete" in relation to ND (natural deduction), without axioms you should not be able to prove anything (not even $ P \to P$ ). In ND certain axioms are "built in"/implicit in the system, one of them is the formula we are discussing,but what when a system that doesn't have it as theorem? you just cannot use ND for it. I like the $ \to, \lnot $ system more ,than PM, logic/ reasoning is all about implications, so don't get rid of them. – Willemien Feb 21 '14 at 22:13
  • Both are not part of axiomatic logic, definitions need to be replaced by axioms , and tautologies, we are doing syntaxis here not semantics, things are true because of inference rules, tautologies are only useful to proof countermodels – Willemien Feb 21 '14 at 22:24
  • 1
    @MauroALLEGRANZA Even with that definition, there exists a definition-free version of the Principia system. Willemien's argument (which I haven't checked entirely, but I think correct) implies that in the definition free version system you can't derive self-distribution. – Doug Spoonwood Apr 27 '14 at 13:41
  • @DougSpoonwood - I think you are right; without W&R definition of $\rightarrow$ in terms of their primitives ($\lor$ and $\lnot$) it is possible that we cannot derive it; see Hagen von Eitzen's first comment above. – Mauro ALLEGRANZA Apr 27 '14 at 17:47
  • 1
    @DougSpoonwood - see also the very interesting "revisiting" of PM in "modern terms" : Peter Andrews, An Introduction to Mathematical Logic and Type Theory To Truth Through Proof (1986). He uses PM's definition of $\rightarrow$; thus he can prove self-distribution and, with it, the Deduction Th. – Mauro ALLEGRANZA Apr 27 '14 at 17:52
  • @MauroALLEGRANZA The definition-free version of the Principia system does not have the deduction theorem. The definition-free version of the Principa system is also NOT complete. I don't recall what Post's proof says exactly (Hebrand's proof the "deduction theorem" actually says something different than what we would call the deduction theorem for propositional logic), but Willemien's proof implies that the definition-free (before they messed around with the Sheffer stroke) version of the Principia system does not have the deduction theorem and is not complete. – Doug Spoonwood Apr 27 '14 at 22:55
  • @MauroALLEGRANZA Also, the equivalence isn't quite so easy to use. Maybe it allows you to replace sub-formulas with logically equivalent subformulas in the definition-free Principia system, maybe not. But, then we would have an axiom which is an equivalence. I think it ends up easier to use a variable function "$\delta$" and have an axiom (in Polish notation) 1: C $\delta$ANpq $\delta$ Cpq. Then 1 q/Cqr [q with Cqr] yields 2 C $\delta$ANpCqr $\delta$ CpCqr. 1 q/r yields 3 C $\delta$ANpr $\delta$ Cpr. 2 $\delta$/C'CANpqANpr yields 4 CCANpCqrCANpqANprCCpCqrCANpqANpr. 1/2 – Doug Spoonwood Apr 27 '14 at 23:09
  • 76 is CANpCqrCANpqANpr. Detachment of 76 and 4 yields 5 CCpCqrCANpqANpr. 1 $\delta$/CCpCqrC'ANpr yields 6 CCCpCqrCANpqANprCCpCqrCCpqANpr. Detachment of 6 and 5, yields 7 CCpCqrCCpqANpr. 3 $\delta$/CCpCqrCCpq' yields 8 CCCpCqrCCpqANprCCpCqrCCpqCpr. Detachment of 8 and 7 yields CCpCqrCCpqCpr. The " ' " in say $\delta$/C'r means that we substitute the argument that follows $\delta$ in place of the apostrophe " ' " symbol. 2/2 – Doug Spoonwood Apr 27 '14 at 23:14
  • 1
    @DougSpoonwood - only some "historical" comment. In PM, $\supset$ is defined [1.01, page 94]. The "equivalence" $p \supset q \equiv \lnot p \lor q$ is proved in 4.6 [page 120] from 4.2 : $p \equiv p$. Of course, PM has an (implicit) rule of substitution, which is largely used to play back-and-forth with $p \supset q$ and $\lnot p \lor q$. But we must remember that also $\equiv$ is defined in 4.01 as $(p \supset q) \land (q \supset p)$. – Mauro ALLEGRANZA Apr 28 '14 at 07:06
  • 1
    @DougSpoonwood - Post's proof is relative to PM system, i.e. only $\lnot$ and $\lor$ are primitive (e.g. modus ponens is :"$\vdash P$ and $\vdash \lnot P \lor Q$ produce $\vdash Q$". [see Jean van Heijenoort (editor), From Frege to Godel A Source Book in Mathematical Logic (1967), page 267]). Now, the substituion rule is explicitly stated (see page 267 and footnote); $\supset$ and $\equiv$ are defined [see page 270]. – Mauro ALLEGRANZA Apr 28 '14 at 07:13
1

According to Willemien's answer above, it is not possible to prove it from the proposed axiom system.

It is provable in Principia Mathematica system (which uses the first four axioms and modus ponens) due to the introduction of the abbreviation :

$p \supset q =_{def} \lnot p \lor q$.

This is a "simplified" proof [see (T9) below], following Peter Andrews, An Introduction to Mathematical Logic and Type Theory : To Truth Through Proof (1986).

Preliminaries

I will call the four Axioms as (Ax1)-(Ax4) [see Russell-Bernays axiom system with negation and disjunction] and the five following results as (R5)-(R9).

I will introduce the definition :

$A \rightarrow B$ stands for $\lnot A \lor B$. [*1.01]

We can have also : $A \land B$ stands for $\lnot (\lnot A \lor \lnot B)$ [*3.01], and $A \leftrightarrow B$ stands for $(A \rightarrow B) \land (B \rightarrow A)$ [*4.01].

The axioms will be stated in the primitive form as :

(Ax1’). $\lnot (A\lor A) \lor A$ --- (Taut) [*1.2]

(Ax2’). $\lnot A \lor (A\lor B)$ --- (Add) [*1.3]

(Ax3’). $\lnot (A\lor B) \lor (B\lor A)$ --- (Perm) [*1.4]

(Ax4’). $\lnot (\lnot A \lor B) \lor [\lnot (A\lor C) \lor (B\lor C)]$ --- (Sum) [*1.5]

The original versions will be derived using the definition of $\rightarrow$.

As inference rules, we will use only modus ponens [*1.1], formulated using the definition of $\rightarrow$ as :

from $A$ and $A \rightarrow B$, $B$ may be inferred;

and the substitution rule :

for a propositional variable any formula may be substituted.

I will call (R5) as Syll [*2.06].

About (R7 - Double Negation) : $\vdash A\leftrightarrow \neg(\neg A)$, we are able to prove it as (T2) and (T3), and we really need only : $\vdash A\rightarrow \neg(\neg A)$.

About (R8) and (R9), they will not be used in the proofs below.

From (Ax4) we derive the rule :

(R10). If $\vdash (A\rightarrow B)$ and $\vdash (A\lor C)$, then $\vdash (B\lor C)$ .

From (Ax3) we derive the rule :

(R11). If $\vdash (A\lor B)$, then $\vdash (B\lor A)$.


Initial theorems

(T1). $\vdash A \lor \lnot A$. [*2.11]

Proof

$\vdash ((A \lor A) \rightarrow A) \rightarrow [((A \lor A) \lor \lnot A) \rightarrow (A \lor \lnot A)]$ --- from (Ax4) with the subst: $A \lor A/A; A/B; \lnot A/C$

$\vdash (A \lor A) \rightarrow A$ --- (Ax1)

$\vdash ((A \lor A) \lor \lnot A) \rightarrow (A \lor \lnot A)$ --- by MP

$\vdash (A \lor A) \lor \lnot A $ --- from (Ax2’) with the subst $A/A; A/B$ using (Ax3) and Syll

$\vdash A \lor \lnot A$ --- by MP.

(T2). $\vdash A \rightarrow \lnot \lnot A$. [*2.12]

Proof

From (T1) with the subst : $\lnot A/A$ and the definition of $\rightarrow$.

(T3). $\vdash \lnot \lnot A \rightarrow A$. [*2.14]

Proof

$\vdash (\lnot A \rightarrow \lnot \lnot \lnot A) \rightarrow ((\lnot A \lor A) \rightarrow (\lnot \lnot \lnot A \lor A))$ --- (Ax4)

$\vdash \lnot A \lor A \rightarrow (\lnot \lnot \lnot A \lor A)$ --- from (T2) with the subst $\lnot A/A$ by MP

$\vdash \lnot \lnot A \rightarrow A$ --- with (Ax3) and Syll by MP, and definition of $\rightarrow$.

(T4). If $\vdash A \rightarrow C$ and $\vdash B \rightarrow C$, then $\vdash (A \lor B) \rightarrow C$.

Proof

$\vdash B \rightarrow C$ --- assumed

$\vdash B \lor A \rightarrow C \lor A$ --- (Ax4) and MP

$\vdash B \lor A \rightarrow A \lor C$ --- (Ax3) and Syll

$\vdash A \lor B \rightarrow A \lor C$ --- (Ax3) and Syll

$\vdash A \rightarrow C$ --- assumed

$\vdash A \lor C \rightarrow C \lor C$ --- (Ax4) and MP

$\vdash A \lor B \rightarrow C \lor C$ --- Syll

$\vdash C \lor C \rightarrow C$ --- (Ax1)

$\vdash A \lor B \rightarrow C$ --- Syll

(T5). If $\vdash A \rightarrow C$ and $\vdash \lnot A \rightarrow C$, then $\vdash C$. [*2.61]

Proof

By (T4) with the subst : $\lnot A/B$ and (T1) and MP.

(T6). If $\vdash A \rightarrow C$, then $\vdash A \rightarrow (C \lor B)$ and $\vdash A \rightarrow (B \lor C)$.

Proof

$\vdash A \rightarrow B$ --- assumed

$\vdash A \rightarrow (C \lor B)$ --- (Ax2) and Syll

$\vdash A \rightarrow (B \lor C)$ --- (Ax3) and Syll

(T7). $\vdash [(A \lor B) \lor C] \rightarrow [A \lor (B \lor C)]$. [*2.32]

Proof

1) $\vdash A \rightarrow [A \lor (B \lor C)]$ --- from (R6) and (T6)

2) $\vdash B \rightarrow (B \lor C)$ --- from (R6) and (T6)

3) $\vdash B \rightarrow [A \lor (B \lor C)]$ --- from 2) by (T6)

4) $\vdash (A \lor B) \rightarrow [A \lor (B \lor C)]$ --- from 1) and 3) by (T4)

5) $\vdash C \rightarrow (B \lor C)$ --- from (R6) and (T6)

6) $\vdash C \rightarrow [A \lor (B \lor C)]$ --- from 5) by (T6)

7) $\vdash [(A \lor B) \lor C] \rightarrow [A \lor (B \lor C)]$ --- from 4) and 6) by (T4)

(T8). If $\vdash (A \lor B) \lor C$, then $\vdash A \lor (B \lor C)$.

Proof

From (T7).


Main theorem

(T9). If $\vdash (A \rightarrow B)$ and $\vdash A \rightarrow (B \rightarrow C)$, then $\vdash A \rightarrow C$.

Proof

1) $\vdash A \rightarrow B$ --- assumed

2) $\vdash A \rightarrow (B \rightarrow C)$ --- assumed

3) $\vdash \lnot A \lor (\lnot B \lor C)$ --- by definition of $\rightarrow$

4) $\vdash (\lnot B \lor C) \lor \lnot A$ --- from 3) by (R11)

5) $\vdash \lnot B \lor (C \lor \lnot A)$ --- by (T8)

6) $\vdash B \rightarrow (C \lor \lnot A)$ --- by definition of $\rightarrow$

7) $\vdash A \rightarrow (C \lor \lnot A)$ --- from 1) and 6) by Syll

8) $\vdash \lnot A \lor (C \lor \lnot A)$ --- by definition of $\rightarrow$

9) $\vdash (C \lor \lnot A) \lor \lnot A$ --- by (R11)

10) $\vdash C \lor (\lnot A \lor \lnot A)$ --- by (T8)

11) $\vdash (\lnot A \lor \lnot A) \lor C$ --- from 10) by (Ax3)

12) $\vdash \lnot A \lor C$ --- from (Ax1) (i.e.$\vdash (\lnot A \lor \lnot A) \rightarrow \lnot A$) and (11) by (R10)

13) $\vdash A \rightarrow C$ --- by definition of $\rightarrow$.

For the original version of (*2.77), see the other answer (for most of applications, the present version suffices).


Additional results

(T10). $\vdash A \rightarrow (\lnot A \rightarrow B)$. [*2.24]

Proof

From (T2) and (T6) by MP.

(T11). $\vdash (A \lor B) \rightarrow (\lnot A \rightarrow B)$. [*2.53]

Proof

$\vdash (A \rightarrow \lnot \lnot A) \rightarrow ((A \lor B) \rightarrow (\lnot \lnot A \lor B))$ --- (Ax4)

$\vdash (A \lor B) \rightarrow (\lnot \lnot A \lor B)$ --- by (T2) with MP

$\vdash (A \lor B) \rightarrow (\lnot A \rightarrow B)$ --- by definition of $\rightarrow$.

(T12). $\vdash (\lnot A \rightarrow B) \rightarrow (A \lor B)$. [*2.54]

Proof

As (T11) from (T3).