On the Partitions of the R-Pyramid , being the first class of R-gonous X-edra .
By the Rev. THOMAS P. KiRKMAN , A.A. , F.R.S. , Rector of Croft with Southworth .
Received October 14 , -Read November 26 , 1857 .
I. As the partitions of an r-gon are made by drawing diagonals , so the partitions of an r-ace are made by drawing diapeds , each a line in two faces non-contiguous about the r-ace .
A partitioned r-ace standing on a partitioned r-gon is a partitioned pyramid of r-gonal base and vertex .
I am about to determine the number of such partitions of this r-pyramid , that can be made with K diapeds and k diagonals , so that no two partitions shall be syntypous ; i. e. one the repetition or the reflected image of the other .
I have proved in a memoir " On Autopolar Polyedra " in the Transactions of the Royal Society for 1857 , that the problem of the polyedra reduces itself to the determination of the x-edra generable from the r-pyramid .
Such an xi-edron is r-gonous .
JDef .
An r-gonous x-edral y-acron is one that by vanescence of convanescible and evanescible edges can be reduced to the r-pyramid , and cannot be so reduced to an ampler pyramid .
The definition of convanescible and evanescible edges is found at Article II .
of the above-mentioned memoir , as follows : An edge AB is said to be convanescible , when neither A nor B is a triangle , and AB joins two summits which have not , besides A and B , two faces , one in each summit , collateral , nor covertical .
An edge ab is said to be evanescible , when neither a nor b is a triace , and the two faces about ab are not , one in each , in two summits , besides a and 6 , collateral , nor in one face .
II .
Theorem .
No r-gonous polyedron has an ( r+l)-gon among its faces , nor an ( r+1)-ace among its summits .
For if it has an ( r+1)-gonal face , and no vanescible lines out of that face , it is an ( r+1)-pyramid , contrary to hypothesis ; and if it has such a face and such lines out of that face , these can be made to vanish , until the figure is an ( r+1)-pyramid , i. e. it is ( r+1)-gonous ; contrary to hypothesis .
In the same way the theorem is proved if the figure has an ( r+1)-ace .
Cor. No r-gonous polyedron can be reduced by vanescences ( i. e. disappearances of convanescible and evanescible edges ) , to one having an ( r+l)-gon or an ( r+ l)-ace ; i. e. no r-gonous polyedron contains an ( r+1)-gony .
III .
The first family of r-gonous polyedra are those arising from partitioning the base and vertex of the r-pyramid , or , which is the same thing , from laying a partitioned r-ace upon a partitioned r-gon .
If we draw k diagonals in the base , none crossing another , and suppose the base to be a system of k+1 faces intersecting in those lines , we have a ( r+ 1)-acral ( r+k+1)-edron .
If we next draw K diaipeds , i. e. edges each in two non-contiguous faces about the r-ace , none of them enclosing a space , we have K new summits , and have before us a ( r+K+1)-acral ( r+k+1)-edron , of the class now to be considered .
Another family arises from partitioning the faces which intersect in the K diapeds , and the summits joined by the k diagonals ; and a third family from the partitioning of the faces and summits upon the diapeds and diagonals which constitute the second .
The first class alone are here to be enumerated , and .
the question before -as is to determine how many different r-gonous ( r+K+1)-acral ( r+k+1)-edra can be made , by laying a ( 1 +K)-partitioned r-ace upon a ( 1 +k)-partitioned r-gon , the r rays passing through the r angles .
IV .
Theorem .
Every ( r+K+1)-acral ( r+k+1)-edron Q , made by laying a ( 1+K)partitioned r-ace A on a ( 1+k)-partitioned r-gon G , the r rays upon the r angles , is r-gonous .
For let it be , supposed that Q is ( r+l)-gonous : it is then a ( r+1+K)-acral ( r+1 +k)-edron having K-1 diapeds and k-I diagonals , the vanescence of which will reduce it to the ( r+1)-pyramid .
These K-1 diapeds cannot be any K-1 of the K diapeds of A ; for of these all the K must vanish to form an r-ace , much less can K-1 vanish to form a ( r+1)-ace .
Nor can these diapeds be all of them edges and diagonals of G , for if G contains an ( r+1)-gonous system of convanescibles , one at least of them must be-a diagonal d , which may be made to vanish last of the K-1 , and must give rise to an ( r+1)-ace .
But d vanishing can bring together only four edges of the r-gon ; it must then bring together r3 terminations of different diagonals ; but if G has r3 diagonals , it is reduced to triangles , in which no line is convanescible ; which is contrary to hypothesis .
Therefore this ( r+1)-gonous system of convanescibles must contain at least one ray 0 of the r-ace A ; and as convanescibles may vanish in any order , one by one , as may diagonals or evanescibles , this 0 can be made to vanish last of the K-1 .
It must therefore at last carry at one end an ( r+l-e)-ace , and at the other a ( 2+e)-ace .
Let the base summit on 4 , be the ( r+ 1e)-ace : this summit having only two edges of the r-gon G and but one ray 0 , has r-e-2 diagonals of G terminating in it ; wherefore 1+2+r-e-2=r+1 -e summits of G are occupied by r+1 -e edges of Q that meet at P , one of those summits .
The other extremity a of d carries a ( 2+e)-ace , the edges of which are first , 4 ; secondly , two diapeds of A , for if d at a meet only rays of A on one or both hands , it would be in one or two triangles , and would be not convanescible ; thirdly , e-1 rays of A , because 2+e 1-2=he-1- , which e1 rays terminate at e -1 summits of the r-gon G , all different from the r+ 1e summits above mentioned ; for if an edge of the ( e+-2)-ace meet one of the ( r+1-e)-ace , D would be not convanescible , by definition .
We have still an account to give of the rays of A meeting in the other extremities of the two diapeds joining A in the ( r+1 -e)-ace .
There must be at least a set of two rays at each of those extremities .
One ray in each set will be in each non-triangular face about D , and therefore in a face about the ( r+1e)-ace ; another ray in each set , that most remote from 4 , will be in a face about the ( 2+e)-ace , and cannot therefore meet any edge of the ( r+1-e)-ace , because d is convanescible ; each ray must therefore pass to a summit of the r-gon not occupied by the r-e-2 diagonals above mentioned , and being rays of A , neither can meet any other ray on the r-gon .
But r+l -e summits of G have been shown to be occupied by the edges of the ( r+1 -e)-ace , and e-1 more by the e1 rays meeting in the ( 2 +e)-ace ; therefore there are no summits of the r-gon remaining to which the ' two last considered 'rays can pass from separate extremities of the two diapeds meeting 4 .
Q. E. A. Therefore Q is not ( r+1)-gonous .
In the same way it can be proved acfortiori that Q is not ( r+l+r')-gonous .
V. This theorem being established , at a cost of words of which I feel ashamed , our problem is reduced to the enumeration of the different figures obtainable by laying any ( 1+K)-partitioned r-ace A on any ( 1+k)-partitioned r-gon G. But we are to exclude from our reckoning any figure P ' which is the reflected image of another , P ; for P ' , being only P turned inside out through some face supposed open , has the same arrangements and ranks of summits and faces with P , i. e. is syntypous with P. What follows will be intelligible to the reader who has before him my memoir " 1 On the'k-partitions of the r-gon and r-ace , " in the Transactions of the Royal Society for 1857 .
These partitions can be found by formulae there given .
VI .
Let a ( 1 +k)-partition A of the r-ace , having j axial planes of reversion , be laid on a ( 1+k)-partition G of the r-gon , having i axes of reversion ( vide the above memoir , Art .
LXXIII .
and Theorems A , B , C , &amp;c. ) .
I call the intersection of an axial plane with the r-gon a trace , and by an axis I mean always an axis of reversion 'of the r-gon G. If we can lay a trace upon an axis , the r rays of A passing through the r summits of G , we shall see , among the angles &gt; 0 at which the remaining traces are inclined to the axes , a certain least angle 0 .
If x be the number of half-edges of the r-gon between two adjacent traces , and y that between two adjacent axes , and z that in the angle 9 , o is the least positive value of z in ax= by ?
z , a and b being numbers of these intervals x and y measured in the same direction from the united trace and axis .
When x is prime to y , it is well known that z=1 ; and when x and y have n for their greatest common measure , we have ax by im max= byl~m .
x2 Here m is the least possible value of z ; for if we say ax= by+r-n , we shall have rn-n divisible by m. Q. E. A. The number x is 2r:2j , and y is 2r:2i , that of the half-edges of the r-gon divided by the number 2j or 2z of intervals between traces or axes .
If we put r for the greatest common measure of rj and r:i , this A is the number of half-edges in 0 , the least interval &gt; 0 between a trace and an axis .
Now let 0 be diminished by an entire edge , every ray taking the place of that preceding it in the direction of revolution of the r-ace .
We shall thus step by step diminish 0 either to zero or to half an edge , as un is even or odd .
The combined configuration will be different at every step , because the least angle between a trace and axis is always diminished , and the configuration C carried by either end of the revolving trace is brought at each step to stand over a different configuration in the r-gon ; for mqu the least interval of the two x=r~j and y=r:i ; that is , no trace is made to cross an axis by this process of diminishing 0 .
VII .
When 0 has its greatest value , it is either rJ or r:i , containing the whole interval , reckoned in half-edges , between two adjacent traces , or adjacent axes .
If it is even , it can be reduced to zero , and the last position as well as the first will show a trace coincident with an axis ; but whether these positions show a trace t over two adjacent axes , or an axis a under two adjacent traces , the combined configurations will be different , because the alternate axial configurations read at the terminations of either traces or axes are always different .
Hence , when 0 is reduced to zero , the figure is not a repetition of a previous one .
But if we diminish 0 below zero , we shall repeat in reversed order the configurations seen when 0 was &gt; 0 0e=e on one side of the united trace and axis giving the reflected image of the figure seen in 0= -e ; as everything is reversible about the united trace and axis , when 0= 0 .
Hence a trace is never to cross an axis in our process .
VIII .
If then we can lay a trace upon an axis for a first position for 0 undiminished , we shall obtain as many additional figures by diminishing 0 by an edge at a step as there are entire edges in 0 .
That is , if 0 is even , we get 1+ &gt; m=l+l J~={2+ } different figures ; and if 0 is odd , 12(- ) 2{ +j~ different figures .
Or if we put " SN =the greatest integer in IN , BE3EINGf THE FIRST CLASS OF R-GONOIJS X-EDRA .
149 we obtain both for 0 even and 0 odd , !
i{2 + , } different figures , whenever a trace can be laid upon an axis , 0 even being finally reduced to zero , and O odd to half an edge .
, But if A has only diachorial traces and G only diagonal axes , or if A has only achorial traces and G only agonal axes , we can lay no trace upon an axis , if the r rays pass through the r angles of G. Our first position will therefore show the 0 reckoned as above diminished by half an edge , and our last position will show 0 diminished to half an edge .
Wherefore 0 must be even , as is also evident from the consideration that the interval between either two traces or two axes in these cases is , an even number of half-edges , so that their greatest common measure is even .
Hence the number offigures obtainable by the reduction of 0 is one less than that obtained above for 0 even , i. e. r is the number of different 'figures attainable , when no trace can be laid upon an axis , by laying A upon G in every possible way .
Now we can always lay a trace upon an axis , unless either the traces are all achorial and the axes all agonal , or the traces are all diachorial and the axes are .
all diagonal .
IX .
Hence we have the Theorem : The number of ( r+K+I)-acral ( r+k+l)-edra that can be made by laying A one of RiDi(r , K ) on G one of Idi(r , I ) , or A one of RJAc(r , K ) on G one of RTh(r , k ) , isi ' and the number of them obtainable in every other case by laying ajply reversible ( 1 +k)partition A of the r-ace on an i-ly reversible ( 1 +k)-partition G of the r-gon is r.~~~~~~~~~~~ where3Z is the greatest common measure of rj and r:i , and !
'N is the greatest integer in JN .
X. Now let A , a ( I+K)-partition of the r-ace reversible aboutj axial planes , be laid on G , an i-ly irreversible ( 1+k)-partition of the r-gon .
If we lay A in any way on , G , so that the r rays pass through the r angles of G , we see at p , the termination of a certain trace , a configuration G ' in G under the axial configuration A ' in A. This A ' is seen in A ] times , at half the 2j terminations of traces ; and G ' is seen in G( i times , at the first point of each of the i equal irreversible sequences that are read round the r-gon , beginning at p. Let O ' be the number of entire edges between that A ' and G ' that are nearest to each other without coincidence , observed in our first position of A on G. Then , if we reduce 0 ' by an edge at once , till it is only a single edge , we shall obtain 0 ' different combined configurations ; because no two of them show the same distance measured in the same direction between the A ' and G ' nearest each other .
If we reduce 0 ' to zero , we see again G ' under A ' , as in our first position , and begin here a series of steps that reproduce the previous figures .
Hence the exact number of different results is 0 ' , the least value of z in the numbers a and b of intervals rj and r:i of whole edges from A ' to A ' and from G ' to G ' being measured from p. This z is the greatest common measure of rji and r:i , i. e. -fr whole edges .
Wherefore this is the number of different figures obtained by laying one of Ri(r , K ) on one of J(r , k ) , and the same exactly is ' that obtained by laying one of F(r , K ) on one of Ri(r , k ) .
No new figures can be obtained by reversing the irreversible G , or by turning inside out the irreversible A , because A in the first , and G in the second , of these cases being reversible , we shall merely obtain by that reversal a reflected image of the figure before the reversal .
XI .
It remains that we lay A one of I(r , K ) on G one of Ii(r , k ) , an irreversible on an irreversible .
A being laid in any way on G , we see at a certain point p the configuration A " over the configuration G " .
A " is found in Aj times , and G " in Gi times .
The least interval in whole edges &gt; 0 between an A " and a G " in our first position is the least value of z in rra +Z , r~j being the number of rays in one irreversible sequence in A , and r:i that of summits in one irreversible sequence in G. And we have r -jli for the number of different figures , each showing a different distance measured in one direction , between the nearest A " and G " .
If we now either reverse G , or turn A inside out , we can double thisnumber of results , for A in the first case and G in the second being irreversible , we obtain figures which are not reflected images of preceding ones .
Wherefore the entire number of figures obtainable is 21 jyoy bylayingr ainy one 'A of V( , r , K ) upoinany oneG Ao I Z( , c ) Collecting our results , we have shown that there are !
-{2 +~j } ways of laying A one of RJDi2r , K ) on G one of + Ria(r , k ) or of Riadi(r , I ) , or A one of RjA`(r , K ) on G one of Riag(r , k ) or of Riagdi(r , k ) , or A one of RjAcDi(r , K ) on G one of IRiag(r , k ) , Ridi(r , k ) or Riagdi(r , k ) , or A one of RjMo(r , K ) on G one of Rimo(r , k ) .
And there are 4r different ways of laying A one of ]EtDi(r , K ) on G one of RidI(r , k ) , or A one of R'Ac(r , K ) on G one of Riag(r , k ) .
Also there are j7i different ways of laying A any one of Rj(r , K ) on G any one of IV(r , k ) , or A any one of I(r , K ) on G any one of RI(r , k ) .
And there are 2.different ways of laying A any one of I(r , K ) on G any one of IP(r , k ) .
XII .
It is certain that , in every one P of these figures , the configuration with respect to the r-gon 12 3. .
r , is different from that seen with respect to the r-gon 12 3. .
r upon any other figure P ' .
But it remains to be considered whether there may not be on P another closed r-gon , whose summits are not 12 3. .
r , about which is seen that configuration which we read on P ' about the r-gon 12 3..r .
If this be so , the ( r+K+1)-acral ( r+k+l)-edron P may be merely the ( r+K+1)-acral ( r+k+l)-edron P ' .
That is , P may be reducible by vanescence to two ( r+l)-edral pyramids not having the same signatures , and may be considered as A laid upon G , or as A ' , a differently partitioned r-ace from A , laid on G ' , a differently partitioned r-gon from G. If P has this double character , I call it a bigenerate r-gonous ( 1+K)-acral ( +Ik)edron ; and if it can be made by laying A on G , and by laying A ' on G ' , and by laying A " on G " , &amp;c. , I call it a multigenerate .
XIII .
We are thus compelled to inquire , how many multigenerates of thefirst clas can be made by laying a ( 1 +K)-partitioned r-ace on a ( 1 +k)-partitioned r-gon ; for as we are enumerating only those of the first class , we have no repetitions to fear out of it .
It is true that some of these ( r+K+1)-acral ( r+k+1)-edra of the first r-gonous class , are also ( r+K+1)-acral ( r+k+I)-edra of the second ; as A ' having K-e diapeds , and laid on G ' having k diagonals , and also e diapeds of the second class , may give the same polyedron with A having K diapeds , laid on G having k diagonals .
This will perplex the discussion of the second class , bAt does not trouble us here .
Let Q be such a multigenerate , made by laying A on G , and by A ' on G ' , &amp;c. As A is not A ' , the r rays of A ' will not comprise all those of A ; nor can they be all diapeds of A , for A cannot have more than r3 diapeds ; nor can they all be found among the sides and diagonals of G , for it is impossible to make above rIedges and diagonals by any arrangement or.convanescences to meet at one summit of G ; while A ' is reducible by convanescence to an r-ace .
Wherefore g( &gt; 0 , &lt; r ) of the rays of A must be also rays of A ' .
Let these 6 common rays be aa , , bb , , cc , , dd , , ... , the points abc d ... being summits of G ' , and ab c , d , ... summits of G. If the K diapeds of A convanesce , A becomes a simple r-ace R standing on G the r-gon a , b , c , d , ... , wherefore ab , be , ed , ... are diapeds of A ; and in like manner a1b , , bc , , ... are diapeds of A ' .
When A is reduced to the simple r-ace R , there are no summits of the figure on the side of the r-gon a , b , c d , d ... remote from R , since A is laid on the partitioned r-gon a , b , c , d , ... ; and if A ' be reduced to the simple r-ace R ' , there are no summits in the figure on the side of the r-gon abc d. remote from R ' .
Hence Q , the figure A upon G , is of this form , for the case r-10 .
Here G is the lO-gon a , b , c , d , efg hij , G ' is the lO-gon abd of gh ; The g common rays are aa , b6 , cc , dd , ; ( fo4 ) The diapeds of A are ab , be , cd ; The diapeds of A ' are a , bc cfd , ; The diagonals of G are api , alh , by c , f , all rays of A ' ; The diagonals of G ' are a , ch , cg , df , all rays of A. I R ' IfIII / / I/ / lI/ / -11 C~~~~ The number of these diagonals , along with the f common rays and the lines a ) and de , must make up in A ' the r rays , t. e. k=r-g-2 , of which one must pass through each of the r- ?
-2 summits of CI fg hi , which are also summits of G through which pass the rays of A , at , ch , cg , df .
The number of diapeds in either A or A ' is K=g-l=r-k-3 .
XIV .
The condition to be fulfilled in drawing the diagonals of G is , that all the -1 lines ab , be , cd shall be diapeds , i. e. none of them in a triangle , and the diagonals of G ' must be so drawn that all the lines a1p , ic , cid , shall be also convanescible .
If these conditions are fulfilled , the figure is bigenerate , otherwise it is not .
This condition is fulfilled , if the k rays of A from ih if be drawn in any manner not crossing one another to one or more of the g summits abe d , and if the k rays of A ' from ih gf are drawn in any manner not crossing one another to any of the summits a 6 , / d , .
For example , the rays of A , or rather the diagonals of G ' , may be drawn all to a , viz. ai , ah , ag , af ; the lines ab , be , ad will be convanescible , being three sides of an open 6-gon ab cd of ; and thus it is easily seen that these k diagonals of G ' through ihg and f may be distributed in any manner upon the summits abc d. Such a figure Q can only be bigenerate ; for suppose that it was trigenerate , made by laying A on G , or A ' on G ' , or A " on G " .
Then it is proved that A and A " have g-r-k-2 common rays , and that the only diapeds of A are ab , bc , od ... ; hence the rays common to A and A " must be ai , bl , eg , df ... , and the figure must be of the form But this is not properly multigenerate , the three 3 eh a C ' lO-gons jihgfedcba , jikqfed~cbja , , and jabcdedjcjba , , being all the same partition G , surmounted by the same partitioned 10-ace A. A 4-generate Q is here entirely out of question ; for the figure must reduce , by the convanescence of the r-k2 diapeds of A , to a simple r-ace standing on G having k diagonals .
Wherefore the Q to be considered here is bigenerate only .
XV .
We shall know the number of bigenerates , if we determine in how many ditfrerent ways these k diagonals drawn from k consecutive summits of the r-gon can terminate at the rk2 summits abc d. But if G and G ' had the same arrangement of their k diagonals , or if one were merely the reflected image of the other , we should have a figure generated by A upon G only , i. e. a figure that we have constructed and counted only once among those made by laying ( K+1)-partitioned r-aces upon ( k+1)-partitioned r-gons .
For example , neither of the two figures following has been twice enumerated ; for whether we take ab , be , and ed , or a,1 , bec , and cod , for the diapeds in either , we find indeed A on G and A ' on G ' , but A differs in no respect from , A ' , nor G from G ' , nor is there any difference in the way of applying the A to the G. Either of them is made by laying in one way only the 4-partitioned 10-ace A on the 5-partitioned lO-gon G , here given separately .
The ( 1+k)-partitions now to be enumerated may be either reversible or irreversible .
If k is odd and e=r-k.-2 is odd also , that is , if r is even and k is odd , one of the k diagonals may be a diameter about which the partition is reversible , If A is even , and i , i. e. r , is either even or odd , , the partition may be reversible about either an agonal or a monogonal axis ; 12k diagonals being similarly drawn on either side of it .
But if k is odd and r-k-2 is even , i. e. if r be odd , no axis but a monogonal can be drawn through the central one of the k summits i , A , g , &amp;c. , which axis cannot be a diagonal .
Hence if k is even and r is odd , the partition cannot be reversible .
XVI .
First , let k be odd and r even .
We can draw a diameter b through the central point of i , Al g , &amp;c. for one diagonal , and can draw I(k-1 ) on one side of w to terminate in one or more of 2(r--k-1 ) summits , viz. the central point of a be ... and those on one side of it , in .
I(k ? ?
1 ) different ways , and the figure can be completed symmetrically about b in so many ways into a reversible ( 1 +k)-partitiortIf k is even and r is even , we can draw lk of the diagonals to one or more of the corresponding half of the rk-2 summits i , A , g , &amp;c. in different ways , and complete the figure by drawing the remaining 2k into a ( 1 +k)-partition reversible about an agonal axis having on each side of it 1k diagonals .
If k is even and r is odd , we can draw.1k of the diagonals to the Y(r-k-1 ) points consisting of the central point of the r-k-2 , and the rest on one side of it , and the figure can be completed symmetrically about the mon-goonal axis through that central point into a reversible ( 1+ ?
k)-partition , in different ways .
4(k ? ?
2 ) Wherefore the entire number of reversible ways of drawing k diagonals from k consecutive summits i , A , g ... to one or more of the ( r-k-2 ) summits a , b , c ... , is NI.k ... 2r{k~ ' -Mr-k))()1-'2(1(r-k2 ) ) 11 ?
2r2____________ Nr k=2r{2k-(2 _ ) +2kf -g _ ) ) +2 -l1t.2( Y)kr k-1k~~ [ i ( 2 ) ) r I(k+2 211(k ?
1)t+2 The entire number of ways in which k diagonals can be drawn from the k points i , A , g , ... to one or more of the r-k --2 points ab c. .
, is ( r-k-2)ill among these every reversible partition comes once only , but every irreversible Q occurs twice ; for Q and its reflected image both occur .
Hence the number of irreversible ( 1+ k)-partitions is r , k=--)rk } XVII .
Of these Nr .
k reversible modes of partition any one can be combined in G with any other in G ' , giving o k(N ' , k 1 ) pairs , and as many bigenerates , which have been each twice constructed and counted .
Of these N ' modes an neay be combined in G with any of the irreversibles in G ' , giving N , k N , k , and as many bigenerates .
Of the N " k irreversible partitions any one in 43 may be combined with another in G ' , and again with that reversed , so that each of the I k(Nr k-1 ) pairs gives two bigenerates .
For example , the two bigenerates have each the same irreversible partition4 G on the lower side in the same position , and each the same irreversible partition G ' above , but in positions one the reverse of the other .
Hence the entire number of bigenerates is Nr , k =7 r , k(Nr , k1)+ Nrk Nr , k+Nr , k(N7 , k-i ) , where N ' and NM have the values just found .
And this number l ; r , ~k is to be subtracted from the results of our enumeration by preceding formulae , being the exact number of ( r+K+1)-acral ( r+k+1)-edra that were twice constructed , each one Q being made both by laying A upon G , and A ' upon G ' .
XVIII .
Thus we have completely determined all the partitions of the ( r+l)-edral pyramid .
All that is necessary is to give to k every value from k= 0 to kl=r3 in the formulae of Art .
XI .
and the same range of values to K ; then to give to k in the two preceding articles every value from k=1 to k=r4 ; for here since neither k nor K &lt; 1 , k=r-K3:r-4 .
And we have at once the number of r-gonous ( r+K+1)-acral ( r+lc+1)-edra of the first class , by giving to k and K their values in ( XI .
) and ( XVII .
) .
That is , if r K , k be this number , K , , K , [ { RjADi(r , K)(IRadi(r , k)+Riag(r , k)+Rd(r , k ) ) +RiAC(r , K)(Riagdi(r , k ) +IRidi(r , k ) ) +PRDi(r , K ) ( [ Riagdi(r , k ) + Riag(r , k ) ) +Ri f(r , K ) .
Rimo(r , k ) X !
2++ { RiAC(r , K ) .
Riag(r , k ) +IDi(r , K ) .
Ridi(r , I ) } X 1 .
i+ { I(r ; K ) .
Ri(r , k)+R'(r , K).I(r , k ) } J- , +21i(r , K).I(r , I ) .
J ] ( rk ) 2 Nk(Nr , h1 , +ror r , k-1 ) rr where is the greatest common measure of an , and !
A is the greatest integer in 1A , and where Ri(r , k ) denotes the entire number of j-ly reversibles about all axes , the rest of the notation here used being that of my paper referred to in Art .
V. ; and where ( rkl))______ 2__ ( p(r-k-2))k 1 21 +2.2( 1 N ' k_ r{2k1 ( ~(~7[i ( k-+ 1 ) +k-(k+ 2 ) io r fi 2 ) and 98_l(rk.-2)kll _N Nr , k 1{(r $A i Nr , k } .
And the entire number Dr of partitions of the ( r+l)-edral pyramid , made by partitioning both the vertex and the base by diapeds and diagonals , is Dr= 1K~kHr , ,K , key the double integral being taken for every value both of K and k from zero to r 3. .
The quantity Nr , k ) subtracted in Ir , K , k , vanishes for k=O and for k=r-3 .
XIX .
Thus far the theory of the polyedra has been opened and discussed without descending to any classification according to the ranks of the faces and summits .
We have had nothing but an r-ace and an r-gon to partition .
But I do not see how the second and higher classes of r-gonous x-edra can be enumerated without such a classification .
This will , I fear , introduce a boundless complexity , and go far to deprive the investigation of all claim to scientific generality .
Yet others may find out a more practicable method of attacking this interesting problem , and I may live to see the remaining cases of it discussed within a reasonable compass .
I wish the analyst joy of his task who shall undertake to complete what I have had the good fortune to begin .
XX .
I have no doubt that the number of r-gonous x-edra is always limited ; but the maximum number of their edges is no simple function of r. It is worth while to write out all the 4-gonous polyecdra .
These are , first and second,3144 245 3244 333 3342 433 -4423545 35423631 3644331 , 35313145 36423344 32333445 , a 6-acral 5-edron , and its polar syntyp,31443245 32443s33 33423433 34423545 35423631 36443131 353i314.5 36423344 32333445 .
The heavy type expresses the faces , and the lighter the summits .
Both are thus represented in one paradigm , by the method explained in my paper " On the Representation .
of Polyedra , " in the Philosophical Transactions for 1856 , using the circles 123456 and 12345 , and writing a at ( 1 , 4 ) and ( 2 5 ) , &amp;c. : e o~~~~~ de h , 1l33'yb 6 &lt; a4I~~~ 3f Both these are of the first class of 4-gonous polyedra , the only one besides of this class being the autopolar 6-acral 6-edron .
This is,41343235 32434335 4 323436 34414136 3543 641 41343235 32 343 5 43 23436 34414136 35433641 , which is also thus represented , using the circles 123456 , 123456 , and putting A for the gamic of a , &amp;c. , D. da __ cA B. b B. eCEa bD d. E c. .
which reduces by the gamic pair Ee to the pyramid on 1234 .
The fourth and fifth are a sympolar pair , a 6-acral 7-edron , and a 7-acral 6-edron , of the second class of 4-gonous x-edra , made by partitioning a 4-gon or a 4-ace in the preceding autopolar so as to introduce no pentace or pentagon .
Thus , 41323241 3243 341 33443441 3444 536 35443645 364s3745 37324145 37433232 33433644 41413436 41363545 , 41323241 32433341 3344344 , 34443536 35443645 36433745 37324145 37433232 33433644 41413436 413635451 both which are seen in the paradigm made by the circles 1234567 and 123456 , as follows:d e73 t ?
.fh1 " .
, e X~~~~~~~~~ 1 The sixth and seventh are the regular 6-edron and its polar syntyp the regular 8-edron , both of the second class , the first represented by writing under 3434 the 12 quadruplets 1426 2435 3443 4153 5263 6275 7286 8116 1144 8251 3365 7625 , and the second by writing under 3434 the same 12 quadruplets .
Both are exhibited thus by using the circles 12345678 , the closed , octagon through the summits of the 6-edron , and 123456 , the hexagon on the faces : 2eg3 3\ R. ; b if 4d 1 i_ 2I This arises from partitioning the only tesserace in the preceding 7-acron , and is reduced to that by the convanescence of any one of its twelve edges .
It is also of the second class .
The eighth is an autopolar heptaedron,41354236 42443333 44423537 36443741 37354i.2334436 35423641 44333342 42353744 44374136 35414i37 33443642 __ __ e ... E Ad fbB\ / . .
BacA F IL 37 b26 F.d C* &lt; D a. IX~ ~~~E .
DIc It is not possible to reduce any of these eight polyedra to the 5-based pyramid ; nor can any diagonal or diaped be drawn in any of them which shall not either produce another of the eight , or introduce a pentagonq , , i. e. a 5-gonous system of vanescibles .
There are no 4-gonous x-edra of a class beyond the second .
XXI .
There is no difficulty in finding tentatively the number of 5-gonous polyedra , by partitioning the faces and summits of the first class of them , taking care to introduce no hexagony .
The partitions of the 5-based pyramid , that is , the whole of this first class , are given by the formula of XVIII .
, in which N5'-0 N-0 1-1 N',2=1 , N"2 ?
z. e. N5 , ,=0=N5,2 .
And the only partitions of the 5-gon are R5mo(5 , 0)=1 , Rmo(5 , 1)_1 Rm0(5 2)=1 .
Wherefore { { RM0(5 , 1).R5mo(5,0)+IR5Mo(5 , O).TMo(5 , 1 ) ?
+Rmo(5 , 2 ) R 5mo(5 , 0 ) + Rmo(5,2 ) R5l ` ( 5 , 0 ) } X !
l 2+ 5 !
+ { Rmo(5 , 1 ) iRmo(5 , 1 ) 4-RM0(5 , 2 ) Rmo(5 1)+Rmo(5 , 1 ) RnW(5 , 2)++Rmo(5 , 2 ) Rmo10(5 , 2 ) } X !
i(2+\ !
)Ns l-Ns\1 2 1/ N5,1 N5,2 { 1.1+1.1+1.1+1.1}1 ?
{ 1.1+1.1+1.1+1.1 }3 16 .
For D6 , the partitions of the 6-pyramid , we have the following partitions of the hexgon : IR6agdi(6 , 0)-I IR2agdi:(6 , 1 ) 1Rdi(6 , 1 ) -1 R2agdi ( 6 , 2)= 1 , Rdi(6 , 2 ) =1 , I(6 , 2 ) = Rdi(6 , 3 ) =1 , 12(6 , 3)=l , Rldi ( 6 , 3)=1 ; wherefore the formula of XVIII .
becomes , E6=198 , thus : 96 ={16AcDi(6 , 0).IRdi(6 , 1+6 , 2 ?
+6 3)+R6agdi(6 , 0).RDi(6 , 1+6 , 2+6 , 3 ) } !
1(2+ 6 ) [ = 6 ] +~ ~~~ ~~~~~~~~~~ ) { Ri(61 I6 16 1gd(6 , O)+R~gi,1+6 2 ) 66A1i 62+ 6 , )= ] +{R2ACDi(6 , 1+6 , 2).jRd(6 , 1)6 , I2 ( 6 , 1+ 6 , 1 ( 6 , 0 6 , ) } . !
S(2+ ) [ =24 ] +{IR2AC(6 , 1+6 , 2)7R agd(6 , 1+6 , 2 ) } !
IS(2 ?
+-2 ) [ = 8 ] 16~~~~~64 +{RDi(6 , 1+6,2+6 , 3).IRd(6 , 1+6 , ( 2+6 , 3)}1 1 !
[ =27 ] + 1R6AcD1(6 , 0 ) .
IRldi(6 , 3)+Rt6agdi(6 , 0 ) .
IRIDi(6 , 3 ) } . !
( 2+613 ) R[=2 ] + { ]R2AM i6 1 +61 2 ) .
R[di(6 1+6,2+6 3 ) + " g12a6 , 1 6 , 2.R+6,23 ) 6 , 1 -(2+.612 { = 64 ] +{IR~~~i(6,162+6,3).TRsdi(6,3)+IRdi(6 , 1+6 , 2+6 , 3).Th ( 6,3 ) } .
[ = 6]3 +R3'( , 3.Rdi6 , ) 3 !
+I6,2)1(,2..llI(6 , 3)12(6 , 3).2 .
[ - !
L19 ] + ]a2AcDi(6 , OI(+6 2 ) .
R~agdi(6 , OI(+6 , 2 ) } G ! .
2+8+ IR6Ac~(6~ 0)12(6 , 3 ) +R6agdiQ5 , 0)I2(6 , 3)}.*6 !
[ =2 +{R2AcDi(6 , 1+6,2)1(6 , 2)+R2agdi(6 , 1+6 , 2).I(6,2)}2 !
[ =12 ] +{R2ACDi(6 1+6 2)12(6 , 3)iR2ajdi(6 , 1+6 , 2)12(6 , 3)}.2-2 [ =12 ] -{IRADi(6,1+6,2+6 , 3).I(6,2)+Iid(6,1+6,2+6,3)I(6,2)}41 !
[ =36 ] + { RDi(6 , 1+6 , 2+6,3)I2(6 , 3)+Rdi(6 , 1+6,2+6 , 3).I2(6 , 3 ) [ 18 ] 6~~~~~~~~~~~~ + { R , ( 61 3 ) I(61 2 ) +1(6 , ~ 2 ) : Wd(61 3)1 +P , +Di ( 6 , 3)12(6 , 3)+12 ( 6 , 3 ) .
R &amp; I(6 , 3 ) } a !
2 A= 6 ] + { 1(6 , 2 ) .
12(6 , 3 ) +I2(6 , 3).I 6,2}.2* [ = 12 ] -(N 6 , 1+N6 , 2+N 6 , 3 ) [ =-2 ] But I see no advantage that can arise from the construction of these partitions of the 5-based and 6-based pyramids , which is , however , to be effected with the greatest ease .
XXII .
It may be worth while to observe , that the regular 12-edron , and of course its polar syntyp the regular 20-edron , are 8-gonous polyedra of the second class .
The 12-edron is made by laying a 5-partitioned 8-ace having only two triangular faces on an 8-gon in which four lines are drawn each cutting two edges , and passing through no summit .
Two of these cutting lines terminate in the base of each of the triangles just named .
This sympolar pair can be exhibited in a paradigm thus ; for closed polygons can be drawn through the 20 summits and on the 12 faces of the 12-edron : *5 14a~t a z. .
t de ... .6 ... .
z\ X 6 ... . .
k y ... 1 .
, ( ..'e'OPT ' / S X\.x 'e aa IMX~~~~~~~~~ ~ ... Jmn '6 n o , ' ... ~~~~~~~~~~~~~~~~~~~~~~4 3 Here the 30 edges ab ... z 0 .
zcy are all 3535 , carrying the subindex quadruplets l12 2 , 2832 , 374 2 , 435 2 , 536 12 , 647 12 , 74811 , 85911 , 951010 , 105119 , 115126 , 124136 , 133146 , 147156 , 157169 , 168179 , 1781810 , 181:1910 , 1912011 , 201112 , 18821 , 16738 , 12584 , 13463 , 159116 , 12512 , 2012 711 , 1911910 , 1710109 , 47143 .
The closed polygon 123 ... 20 is drawn through the 20 summits , and the closed polygon 123..12 through the 12 faces .
If any five continuous edges , of which no three are in a pentagon , be made to convanesce , as the five xsrpg , the 8-ace is restored , 6 triaces thus uniting to form it .
The disappearance of the remaining vanescibles will restore the 8-gon , Every x-edron of the first class i. e. every partition of a pyramid , can be thus exhibited in a paradigm , and the greater number of those of the higher classes .
The partitions of the r-pyramid have all this property , that each contains a discrete r-gony , i. e. an r-gonous system of vanescibles of which no diaped meets a diagonal ; or each contains an r-gony of the first class .
Some of them , however , contain also a mixed r-gony , on which are one or more angles made by a diaped and a diagonal .
If the figure contain a discrete r-gony , it is an r-gonous polyedron of the first class ; if not , it is a polyedron of a higher class .
The diapeds of a discrete r-gony form a continuous line of convanescibles ; if a diagonal be drawn in a face about one of these , the r-gony is no longer discrete , but mixed .
The problems that are next to be solved towards the completion of the theory of the polyedra , are the following ; and I have little hope of their solution , in terms of r. How many partitions can be made of the summits of a partitioned r-gon , so that no ( r+1)-gon shall be introduced ?
How many partitions can be made of the faces of a partitioned r-ace , so that no ( r+1)-ace shall be introduced ?
In how many ways can a partitioned partition of the r-ace be laid on a partitioned partition of the r-gon , so that no ( r+e)-gony , nor ( r+e)-gon , or ( r+e)-ace shall be introduced ?
MDGCCCLVIII .

