Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Group Theory

54 views
Skip to first unread message

Richard Carr

unread,
Jul 5, 2000, 3:00:00 AM7/5/00
to

Consider F=the free group on {a,b,c,d,...,z}. Let E be the equivalence
relation generated by ~ (where x~y iff x and y are English words (having
an entry in the OED, say- possibly obsolete or a variant form) which are
anagrams of each other- so for example, altitude~latitude and
clean~lance). Let Q be the quotient group F/E.
The problem is to show that Q is the free Abelian group on
{a,b,c,d,...,z}. (This problem was given to me several years ago,
together with another one (same F, E was with words that sound the same-
homophones, I guess- and the object was to get that Q was trivial, but
that was much easier.)
Anyway, there are 325 relations to get and I've got 302 of them. Any clues
on the missing ones? Here they are:

cj,fj,fq,fx,gq,gx,hz,jq,jw,jx,jz,kq,kx,kz,qv,qw,qx,qy,vx,wx,wz,xz.

Richard Carr

PS It's more of a dictionary puzzle than a mathematical one.
P^2S I didn't use any words with accents- because they wouldn't really be
in F, so I am trying to avoid them.

Hugo van der Sanden

unread,
Jul 5, 2000, 3:00:00 AM7/5/00
to
Richard Carr wrote:
>
> Consider F=the free group on {a,b,c,d,...,z}. Let E be the equivalence
> relation generated by ~ (where x~y iff x and y are English words (having
> an entry in the OED, say- possibly obsolete or a variant form) which are
> anagrams of each other- so for example, altitude~latitude and
> clean~lance). Let Q be the quotient group F/E.
> The problem is to show that Q is the free Abelian group on
> {a,b,c,d,...,z}. (This problem was given to me several years ago,
> together with another one (same F, E was with words that sound the same-
> homophones, I guess- and the object was to get that Q was trivial, but
> that was much easier.)
> Anyway, there are 325 relations to get and I've got 302 of them. Any clues
> on the missing ones? Here they are:
>
> cj,fj,fq,fx,gq,gx,hz,jq,jw,jx,jz,kq,kx,kz,qv,qw,qx,qy,vx,wx,wz,xz.

I don't understand enough of the maths to get what is required here. Can
you give some examples of the relations you already have, that might
clarify?

Hugo

Gareth Taylor

unread,
Jul 5, 2000, 3:00:00 AM7/5/00
to
In article <Pine.LNX.4.21.000705...@cpw.math.columbia.edu>,
Richard Carr <ca...@cpw.math.columbia.edu> wrote:

> Consider F=the free group on {a,b,c,d,...,z}. Let E be the equivalence
> relation generated by ~ (where x~y iff x and y are English words (having
> an entry in the OED, say- possibly obsolete or a variant form) which are
> anagrams of each other- so for example, altitude~latitude and
> clean~lance). Let Q be the quotient group F/E.
> The problem is to show that Q is the free Abelian group on
> {a,b,c,d,...,z}.

Hang on. The OED clause is worrying. With that restriction, then
"clean/lance" and "nclea" are distinct elements in F/E, as "nclea" isn't a
word in the OED, so isn't in the same class as "clean". However, they
_are_ the same element in the free abelian group.

If you drop the "having an entry in the OED" bit and confine yourself to
"x~y iff x is a permutation of y" then it's easy.

Gareth

Gareth Taylor

unread,
Jul 5, 2000, 3:00:00 AM7/5/00
to
In article <tEg*ng...@news.chiark.greenend.org.uk>,
Gareth Taylor <gta...@chiark.greenend.org.uk> wrote:

...something which someone has added to. A friend of mine sent me the
following:

> Oh! *blinks* _Now_ I see. That weasel word `generated' was important.
>
> The equivalence relation _generated by_ ~. That is, not the equivalence
> relation ~ itself (which is a perfectly sensible equivalence relation),
> but the relation that also contains any other equivalences that can be
> deduced from ~ and group theory. Thus:
>
> cat ~ act (defn of ~)
> => cat = act (if it's true in ~, it must be true in generated
> relation =)
> => ca = ac (= is a relation which preserves group-theoretic
> properties)
>
> _That's_ what he's after. A sequence of proofs like that which
> demonstrate Abelianness pairwise between letters of the alphabet.

Is _that_ what you meant?

Gareth

John Rickard

unread,
Jul 5, 2000, 3:00:00 AM7/5/00
to
Hugo van der Sanden <h...@crypt0.demon.co.uk> wrote:
: Richard Carr wrote:
[...]
: > Anyway, there are 325 relations to get and I've got 302 of them. Any clues

: > on the missing ones? Here they are:
: >
: > cj,fj,fq,fx,gq,gx,hz,jq,jw,jx,jz,kq,kx,kz,qv,qw,qx,qy,vx,wx,wz,xz.
:
: I don't understand enough of the maths to get what is required here.

Roughly, for each of those letter pairs, find a pair of single-word
anagrams containing the two letters in which they occur in a different
order in the two words. For example, for the first pair "cj", if
"jolace" and "jenticed" were words then cajole/jolace would work, but
injected/jenticed would not, because the "j" is before the "c" in both
"injected" and "jenticed".

I say "roughly" because Richard Carr's requirements are more stringent
than that: anagrams with the property I've described are necessary,
but not sufficient. However, I think that a solution to the problem
I've described is unlikely not to solve the original problem as well.

--
John Rickard

Richard Carr

unread,
Jul 5, 2000, 3:00:00 AM7/5/00
to
On 5 Jul 2000, Gareth Taylor wrote:

:Date: 05 Jul 2000 11:07:35 +0100 (BST)
:From: Gareth Taylor <gta...@chiark.greenend.org.uk>
:Newsgroups: rec.puzzles
:Subject: Re: Group Theory
:
:In article <tEg*ng...@news.chiark.greenend.org.uk>,

Yes- like this. If I already have ar=ra (say, from art=rat) then I can use
that jar=raj and ajar=raja to get

jar=raj=rajaa^{-1}=ajara^{-1}=ajraa^{-1}=ajr and hence ja=aj.

The main problem is coming up with the anagrams.

:
:Gareth
:


Richard Carr

unread,
Jul 5, 2000, 3:00:00 AM7/5/00
to
On 5 Jul 2000, John Rickard wrote:

:Date: 05 Jul 2000 13:44:38 +0100 (BST)
:From: John Rickard <John.R...@virata.com>


:Newsgroups: rec.puzzles
:Subject: Re: Group Theory
:

:Hugo van der Sanden <h...@crypt0.demon.co.uk> wrote:
:: Richard Carr wrote:
:[...]
:: > Anyway, there are 325 relations to get and I've got 302 of them. Any clues
:: > on the missing ones? Here they are:
:: >
:: > cj,fj,fq,fx,gq,gx,hz,jq,jw,jx,jz,kq,kx,kz,qv,qw,qx,qy,vx,wx,wz,xz.
::
:: I don't understand enough of the maths to get what is required here.
:
:Roughly, for each of those letter pairs, find a pair of single-word
:anagrams containing the two letters in which they occur in a different
:order in the two words. For example, for the first pair "cj", if
:"jolace" and "jenticed" were words then cajole/jolace would work, but
:injected/jenticed would not, because the "j" is before the "c" in both
:"injected" and "jenticed".

:

Right.

:I say "roughly" because Richard Carr's requirements are more stringent


:than that: anagrams with the property I've described are necessary,
:but not sufficient. However, I think that a solution to the problem
:I've described is unlikely not to solve the original problem as well.

:

Indeed, at this stage, with so few relations left, it is more than likely
that this approach would work too- the only problem being
that all the relations involve j, q, x or z so that there is
a chance that finding an anagram may not work simply because it has at
has, for instance, c, f and j in. From caftajack to jaftacack (both of
which I just made up for illustrative purposes) one couldn't conclude
cj=jc of fj=jf because the j has to get past both the f and the c. The
best you could conclude is that jc=cj if and only if fj=jf (which would be
useful in the case of finding some other relation that settled one case or
the other).


:--
:John Rickard
:


Richard Carr

unread,
Jul 5, 2000, 3:00:00 AM7/5/00
to
On Wed, 5 Jul 2000, David Beckman wrote:

:Date: Wed, 5 Jul 2000 13:04:47 -0700 (PDT)
:From: David Beckman <bec...@its.caltech.edu>
:To: ca...@cpw.math.columbia.edu


:Newsgroups: rec.puzzles
:Subject: Re: Group Theory
:

:In rec.puzzles you write:
:
:
:>Consider F=the free group on {a,b,c,d,...,z}. Let E be the equivalence


:>relation generated by ~ (where x~y iff x and y are English words (having
:>an entry in the OED, say- possibly obsolete or a variant form) which are
:>anagrams of each other- so for example, altitude~latitude and
:>clean~lance). Let Q be the quotient group F/E.
:>The problem is to show that Q is the free Abelian group on

:>{a,b,c,d,...,z}. (This problem was given to me several years ago,


:>together with another one (same F, E was with words that sound the same-
:> homophones, I guess- and the object was to get that Q was trivial, but
:>that was much easier.)

:>Anyway, there are 325 relations to get and I've got 302 of them. Any clues


:>on the missing ones? Here they are:
:>
:>cj,fj,fq,fx,gq,gx,hz,jq,jw,jx,jz,kq,kx,kz,qv,qw,qx,qy,vx,wx,wz,xz.
:>

:>Richard Carr
:
:Hi Richard,
:
:I used a word list from some Merriam-Webster dictionary, which has some
:questionable words, and ran it through some anagram programs I had around.
:I checked with OED online and found some of the words. The pairs below
:were in OED as main entries except where noted. (Some were only found in
:a full-text search; some weren't there at all and seem to be variants or
:specialized terminology.) I don't know how much you want to stretch the
:OED words....

Thanks- that was fast. I'm not sure about the status of capitalized words
(I'll have to check with the guy who asked me the puzzle). Also work-box
seems to be hyphenated (whereas words in F are not).

Anyway, angelique and equiangle checked as did philozoic and
zoophilic. Hejazi and jeziah did too (so I'm going to check if proper
words are OK). I found myxofibroma but not fibromyxoma and cajoler but not
jecoral. I couldn't find several of the later words in the OED and don't
have a M-W to check but I'll see if I can get access to one.
I did find both the words for qv- what worries me here is that one is
capitalized (and part of a compound sort of expression) and the other is
accented. Anyway, I thank you for your help and look forward to verifying
all of this.

:
:cj:
:cajoler
:jecoral (jecorary)
:
:fx:
:fibromyxoma (fibroma, myxoma)
:myxofibroma (fibroma, myxoma)
:
:fibromyxosarcoma (fibroma, myxoma, sarcoma)
:myxofibrosarcoma (fibroma, myxoma, sarcoma)
:
:gq:
:angelique
:equiangle
:
:aqueoigneous
:igneoaqueous
:
:hz:
:Hejazi (found in OED full text, main entry Iraqi)
:jeziah
:
:Phytozoa (found in full text)
:Zoophyta (found in full text)
:zoopathy (zoopathology)
:
:philozoic (found in full text)
:zoophilic
:
:philozoist (found in full text)
:zoophilist
:
:Rhizoctonia
:Zonotrichia (found in full text)
:chorization
:
:jz:
:Hejazi (found in full text)
:jeziah
:
:kx:
:boxwork (not found in OED, but a common term for a type of cave formation:
: http://www.goodearthgraphics.com/virtcave/boxwork/boxwork.html)
:workbox
:
:hexakisoctahedron
:octakishexahedron
:
:hexakistetrahedron
:tetrakishexahedron
:
:kz:
:alkalizer (alkalize)
:lazarlike (lazar)
:
:qv:
:eveque (Pont L'Eveque, cheese variety)
:quev'ee (acute accent over second e)
:
:wx:
:boxwork
:workbox
:
:xz:
:azoxine (var. of azoxy? A web search turns up some matches, but not in OED)
:oxazine
:
:
: Dave Beckman (bec...@its.caltech.edu)
:


Hugo van der Sanden

unread,
Jul 6, 2000, 3:00:00 AM7/6/00
to
John Rickard wrote:
> Hugo van der Sanden <h...@crypt0.demon.co.uk> wrote:
> : Richard Carr wrote:
> [...]
> : > Anyway, there are 325 relations to get and I've got 302 of them. Any clues

> : > on the missing ones? Here they are:
> : >
> : > cj,fj,fq,fx,gq,gx,hz,jq,jw,jx,jz,kq,kx,kz,qv,qw,qx,qy,vx,wx,wz,xz.
> :
> : I don't understand enough of the maths to get what is required here.
>
> Roughly, for each of those letter pairs, find a pair of single-word
> anagrams containing the two letters in which they occur in a different
> order in the two words. For example, for the first pair "cj", if
> "jolace" and "jenticed" were words then cajole/jolace would work, but
> injected/jenticed would not, because the "j" is before the "c" in both
> "injected" and "jenticed".
>
> I say "roughly" because Richard Carr's requirements are more stringent
> than that: anagrams with the property I've described are necessary,
> but not sufficient. However, I think that a solution to the problem
> I've described is unlikely not to solve the original problem as well.

Right, along with the extra explained in Richard's reply to Gareth
Taylor I now understand, and was able to confirm that nothing in my own
wordlist will help. :(

Thanks,

Hugo

David Moews

unread,
Jul 6, 2000, 3:00:00 AM7/6/00
to
|Consider F=the free group on {a,b,c,d,...,z}. Let E be the equivalence
|relation generated by ~ (where x~y iff x and y are English words (having
|an entry in the OED, say- possibly obsolete or a variant form) which are
|anagrams of each other- so for example, altitude~latitude and
|clean~lance)...The problem is to show that [F/E] is the free Abelian group on
|{a,b,c,d,...,z}. ...

|Anyway, there are 325 relations to get and I've got 302 of them. Any clues
|on the missing ones? Here they are:
|
|cj,fj,fq,fx,gq,gx,hz,jq,jw,jx,jz,kq,kx,kz,qv,qw,qx,qy,vx,wx,wz,xz.

What anagrams make j and v commute?

Conway, Feit, and Sahi have written a paper on this. Here are some anagrams
from their paper, together with some Chris Cole, Chris Long, Murray Pearce,
and I found in 1993:

(ni2 or 3 = merriam-webster's second or third new international dictionary;
9c = merriam-webster's collegiate dictionary, 9th ed.)

fj fijf jiff (oed,9c) [murray pearce]
fq foursquare quarefours (oed) [chris long]
fx expressionful superflexions (ni2)
fx loxify foxily [conway-feit-sahi]
gx exerting genetrix (ni2, ni3 or oed2) [chris cole]
jw jackscrew screwjack [conway-feit-sahi]
jz jiz zij (oed2, funk & wagnall new standard) [chris long]
qv eveque queeve (ni2)
wx boxwood woodbox [conway-feit-sahi]
wz waz zaw (oed2) [chris long]

Also, AZOXINE is in NI2.

Together with the anagrams found by you and David Beckman, these cover all
pairs except for (j,q), (j,x), (k,q), (q,w), (q,x), (q,y), and (v,x).
--
David Moews dmo...@xraysgi.ims.uconn.edu

Richard Carr

unread,
Jul 6, 2000, 3:00:00 AM7/6/00
to
On 6 Jul 2000, David Moews wrote:

:Date: 6 Jul 2000 11:39:50 -0700
:From: David Moews <dmo...@xraysgi.ims.uconn.edu>


:Newsgroups: rec.puzzles
:Subject: Re: Group Theory
:

:In article <Pine.LNX.4.21.000705...@cpw.math.columbia.edu>,


:Richard Carr <ca...@cpw.math.columbia.edu> wrote:
:|Consider F=the free group on {a,b,c,d,...,z}. Let E be the equivalence
:|relation generated by ~ (where x~y iff x and y are English words (having
:|an entry in the OED, say- possibly obsolete or a variant form) which are
:|anagrams of each other- so for example, altitude~latitude and
:|clean~lance)...The problem is to show that [F/E] is the free Abelian group on
:|{a,b,c,d,...,z}. ...
:|Anyway, there are 325 relations to get and I've got 302 of them. Any clues
:|on the missing ones? Here they are:
:|
:|cj,fj,fq,fx,gq,gx,hz,jq,jw,jx,jz,kq,kx,kz,qv,qw,qx,qy,vx,wx,wz,xz.
:
:What anagrams make j and v commute?

javar and vajra

javar: A swelling in the hollow of the pastern of a horse
vajra: In Hinduism and Buddhism, a thunderbolt or mythical weapon esp. one
wielded by the god Indra.

Progress towards the solution of this problem can be found at

http://www.math.columbia.edu/~carr/Matt.html

(Matt told me about the problem.)

I have put up those of David Beckman's relations (which I have
confirmed) together with those I had already found (his are listed
with "found by David Beckman") and as I confirm the other ones (yours
and his) will add those to the list also with appropriate accredation
(e.g. "found by Chris Cole, shown to me by David Moews"). I have also
confirmed that workbox need not be hyphenated as I first thought. I was
thus able to use boxwood=woodbox to get wx=xw and from that kx=kx using
David's workbox=boxwork.

:
:Conway, Feit, and Sahi have written a paper on this. Here are some anagrams


:from their paper, together with some Chris Cole, Chris Long, Murray Pearce,
:and I found in 1993:
:
:(ni2 or 3 = merriam-webster's second or third new international dictionary;
: 9c = merriam-webster's collegiate dictionary, 9th ed.)
:
:fj fijf jiff (oed,9c) [murray pearce]

I was able to verify this (indeed both are in the OED).

:fq foursquare quarefours (oed) [chris long]

four-square is hyphenated, unfortunately. I was also able to confirm that
capitalized words are not allowed, unfortunately.
Actually, I just looked in the Webster's and foursquare is not hyphenated
there (although it is in the OED).

:fx expressionful superflexions (ni2)

:fx loxify foxily [conway-feit-sahi]
:gx exerting genetrix (ni2, ni3 or oed2) [chris cole]

I was also able to confirm this.

:jw jackscrew screwjack [conway-feit-sahi]

Confirmed jackscrew in Webster's but "screw jack" is listed as 2 words.

:jz jiz zij (oed2, funk & wagnall new standard) [chris long]


:qv eveque queeve (ni2)
:wx boxwood woodbox [conway-feit-sahi]

Confirmed.

:wz waz zaw (oed2) [chris long]


:
:Also, AZOXINE is in NI2.

I need to get or find copies of ni2, ni3 and oed2 evidently. : )

I did find lazarlike in Webster's but not alkalizer and also I found
hexakisoctahedron and hexakistetrahedron but not octakishexahedron or
tetrakishexahedron. I will search for them when I get hold of more
dictionaries.

Richard Carr

PS Thanks by the way- I have been on this problem on and off for a few
years (don't have a good way of searching a dictionary, unfortunately so
mainly I just guess words and look them up to see if they exist- it takes
a lot of time that way- I imagine you guys have better methods).

:
:Together with the anagrams found by you and David Beckman, these cover all

:


John Rickard

unread,
Jul 7, 2000, 3:00:00 AM7/7/00
to
Richard Carr <ca...@cpw.math.columbia.edu> wrote:
: (This problem was given to me several years ago, together with

: another one (same F, E was with words that sound the same-
: homophones, I guess- and the object was to get that Q was trivial,
: but that was much easier.)

I've been thinking (not all that hard) about that one; it does indeed
seem to be much easier. I can show that every letter except M and V
is the identity, using only (what I consider to be) common words and,
so far as I know, using homophones that don't depend on the variety of
English one speaks. I can get M in British English, but I don't know
whether my example works in American English.

What did you use for M and V? (For those less mathematically
inclined, find pairs of homophones where M or V occurs in one word but
not the other, or a different number of times in each word.)

Hah -- just thought of a V as I was writing this, though one of the
words is slang.

--
John Rickard <John.R...@virata.com>

Richard Carr

unread,
Jul 7, 2000, 3:00:00 AM7/7/00
to
On 7 Jul 2000, John Rickard wrote:

:Date: 07 Jul 2000 18:08:25 +0100 (BST)
:From: John Rickard <j...@255.255.255.192>
:Reply-To: John.R...@virata.com


:Newsgroups: rec.puzzles
:Subject: Re: Group Theory
:

:Richard Carr <ca...@cpw.math.columbia.edu> wrote:
:: (This problem was given to me several years ago, together with
:: another one (same F, E was with words that sound the same-
:: homophones, I guess- and the object was to get that Q was trivial,
:: but that was much easier.)
:
:I've been thinking (not all that hard) about that one; it does indeed
:seem to be much easier. I can show that every letter except M and V
:is the identity, using only (what I consider to be) common words and,
:so far as I know, using homophones that don't depend on the variety of
:English one speaks. I can get M in British English, but I don't know
:whether my example works in American English.
:
:What did you use for M and V? (For those less mathematically
:inclined, find pairs of homophones where M or V occurs in one word but
:not the other, or a different number of times in each word.)

:

It's a few years ago now, but I remember that I used bevy=bevvy for v (both
in OED and have same pronunciation- the phonetic spellings are given
and they matched). I can't recall off hand what I used for m- almost
certainly one word had a double m where the other had just a single m. I
used 26 relations on this one.

:Hah -- just thought of a V as I was writing this, though one of the

:


0 new messages