On a Deconcatenation Problem

Henry Ibstedt
Abstract: In a recent study of the PrimaIity oj the Smarandache Symmetric
Sequences Sabin and Tatiana Tabirca [1] observed a very high frequency of the
prime factor 333667 in the factorization of the terms of the second order
sequence. The question if this prime factor occurs peridically was raised. The odd
behaviour of this and a few other primefadors of this sequence will be explained
and details of the periodic occurence of this and of several other prime factors
will be given.
Definition: The nth term of the Smarandache symmetric sequence of the second order
is defined by S(n)=123 ... n_n ... 321 which is to be understood as a concatenation I of
the first n natural numbers concatenated with a concatenation in reverse order of the n

Factorization and Patterns of DivisIoility
The first five terms of the sequence are: 11, 1221, 123321, 12344321, 1234554321.
The number of digits D(n) of Sen) is growing rapidly. It can be found from the
formula:
D(n) = 2k(n +1)- 2(lO
t
-I) for n in the (1)
9
In order to study the repeated occurrance of certain prime factors the table of Sen) for
XC:;100 produced in [1] has been extended Tabirca's aim was to factorize the
terms S(n) as far as possible which is more ambitious then the aim of the present
calculation which is to find prime factors which are less than 10
8
The result is shown
in table 1.
The computer file containing table 1 is analysed in various ways. Of the 664579
primes which are smaller than 10
7
only 192 occur in the prime factoriztions of Sen)
for Of these 192 primes 37 ocCur more than once. The record holder is
333667. the 28693th prime, which occurs 45 times for while its neighbours
333647 and 333673 do not even occur once. Obviously there is something to be
explained here. The frequency of the most frequently occurring primes is shown
below ..
Table 2. Most frequently occurring primes .
1 In this article the concatenation of a and b is written a_b. Multiplication ab is often made explicit by
writing a.b. When there is no reason for misunderstanding the signs "_" and "." are omitted. Several
tables contain prime factorizations. Prime factors are given in ascending order, multiplication is
expressed by"." and the last factor is followed by .... " if the factorization is incomplete or by Fxxx
indicating the number of digits of the last factor. To avoid typing errors all tables are electronically
transferred from the calculation program, which is DOS-based, to the wordprocessor. All editing has
been done either with a spreadsheet program or directly with the text editor. Full page tables have been
placed at the end of the article. A non-proportional font has been used to illustrate the placement of
digits when this has been found useful.
107
The distribution of the primes 11,37,41,43,271,9091 and 333667 is shown in table
3. It is seen that the occurance patterns are different in the intervals
and Indeed the last interval is part of the It would
have been very interesting to include part of the interval 1000<0<9999 but as we can
see from (1) already S(looo) has 5786 digits. Partition lines are drawn in the table to
highlight the different intervals. The less frequent primes are listed in table 4 where
primes occurring more than once are partitioned.
From the patterns in table 3 we can formulate the occurance of these primes in the
intervals and l00<-IlQoo, where the formulas for the last interval are
indicative. We note, for example, that 11 is not a factor of any term in the interval
This indicates that the divisibility patterns for the
and further intervals is a completely open question.
Table 5 shows an analysis of the patterns of occurance of the primes in table 1 by
interval. Note that we only have observations up to n=2oo. Nevertheless the interval
is used. This will be justified in the further analysis.
Table 5. Divisibility patterns
Interval p n Range for j

3 2+3j j=O.l._

3j j .. l.2._

11 A1.1 values of n

12+11j j=O,l.
-
.7
20+11j j=O.l.
-
.7

None

37 2+3j j .. 0.l,2
3+3j j=0.1.2

12+3j j .. O,l._.28.29

n2+37j j=0.1._.23
l.36+37j j=O,l._.23

41 4+Sj j=O.l
5

14+5j j=O,l._.197

43 None

1l+2l.j j=O.1.3.4
24+2l.j j=0.1.2.3

100
107+7j j=O,l.._,127

271 4+Sj j=O.l.
5

14+5j j=O,l.._.197

909l. 9+Sj j=0.1._,98

333667 8,9

18+9j j=O.1._.9

102+3j j=O,l._.299
We note that no terms are divisible by 11 for n>100 in the and
that no term is divisible by 43 in the Another remarkable observation
is that the sequence shows exactly the same behaviour for the primes 41 and 271 in
the intervals included in the study. Will they show the same behaviour when
108
Consider
S(n)=12 ... n_n ... 21.
Let P be a divisor of Sen). We will construct a number
N=12 ... n_O .. O_n ... 21 (2)
so that p also divides N. What will be the number of zeros? Before discussing this
let's consider the case p=3.
Case 1. p=3.
In the case p=3 we use the familiar rule that a number is divisible by 3 if and only if
its digit sum is divisible by 3. In this case we can insert as many zeros as we like in
(2) since this does not change the sum of digits. We also note that any integer fonned
by concatenation of three consecutive integers is divisible by 3, cf a_a+Ca+2, digit
sum 3a+3. It follows that also a_a+ l_a+2_a+2_a+ Ca is divisible by 3. For a=n+ 1 we
insert this instead of the appropriate number of zeros in (2). This means that if S(n)=O
(mod 3) then S(n+3)=O (mod 3). We have seen that S(2)=O (mod 3) and S(3)=O (mod
3). By induction it follows that S(2+3j)=O (mod 3) for j=1,2, ... and S(3jF-O (mod 3)
. for j=1,2, ....
We now return to the general case. Sen) is deconcatenated into two numbers l2 ... n
and n ... 21 from which we form the numbers
A=12 .. .nl01+[log
,o
B] andB=n ... 21
We note that this is a different way of writing Sen) since indeed A+B=S(n) and that
A+B=O (mod p). We now form M=AlOs+B where we want to determine s so that
M=O (mod p). We write M in the fonn M=A(IO'-I)+A+B where A+B can be ignored
mod p. We exclude the possibility A=O (mod p) which is not interesting. This leaves
us with the congruence
M=A(lO'-I)=O (mod p)
or
10'-1=0 (mod p)
We are particularly interested in solutions for which
pE {11,37,41,43,271,9091.333667}
By the nature of the problem these solutions are periodic. Only the two fIrst values of
s are given for each prime.
Table 6. (mod p)
We note that the result is independent of n. This means that we can use n as a
parameter when searching for a sequence C=n+l_n+2_ ... n+k_n+k_ ... n+2_n+1 such
that this is also divisible by p and hence can be inserted in place of the zeros to form
S(n+k) which then fIlls the condition S(n+kF-O (mod p). Here k is a multiple of s or
sfl in case s is even. This explains the results which we have already obtained in a
different way as part of the factorization of Sen) for see tables 3 and 5. It
remains to explain the periodicity which as we have seen is different in different
intervals
109
This may be best done by using concrete examples. Let us use the sequences starting
with n=12 for p=37, n=12 and n=20 for p=I1 and n=102 for p=333667. At the same
time we will illustrate what we have done above.
Case 2: n=12, p=37. Period=3. Interval:
S(n)= 123456789101112 .121110987654321
N= 123456789101112000000000000121110987654321
c= 131415151413
S(n+k) =123456789101112131415151413121110987654321
Let's look at C which carries the explanation to the periodicity. We write C in the
form
C=101010101010+30405050403
We know that C=O (mod 37). What about 101O10101O1O? Let's write
101010101010=10+10
3
+10
5
+_+10
11
= (10
12
_1) /9=0 (mod 37)
This congruence mod 37 has already been established io table 6. It follows that also
30405050403=0 (mod 37)
and that
x(101010101010)=0 (mod 37) for x = any integer
Combining these observations we se that
232425252423, 333435353433, _ 939495959493=0 (mod 37)
Hence the periodicity is explained.
Case 3a: n=12, p=Il. Period=Il. Interval:
S(12)=12_ .. _12 12_ . _21
S(23)=12_ .. _121314151617181920212223232221201918171615141312_ .. _21
C= 13141516171819202122232322212019181716151413=
C1= 10101010101010101010101010101010101010101010+
C2= 3040506070809101112131312111009080706050403
From this we form
2C1+C2= 23242526272829303132333332313029282726252423
which is NOT what we wanted, but CI=O (modi 1) and also CIIlO=O (mod 11).
Hence we form
2C1+C1/10+C2=24252627282930313233343433323130292827262524
which is exactly the C-term required to form the next term S(34) of the sequence. For
the next term S(45) the C-term is formed by 3C1+2CIII0+C2 The process is
repeated adding Cl+C11l0 to proceed from a C-term to the next until the last term
<100, i.e. S(89) is reached.
Case 3b: n=20, p=ll. Period=ll. Interval:
This case does not differ much from the case 0=12. We have
S (20) =12_ .. _20 20_ .. _21
S(31)=12_ .. _202122232425262728293031313029282726252423222120_ .. _21
C= 21222324252627282930313130292827262524232221=
C1= 10101010101010101010101010101010101010101010+
C2= 1020304050607080910111110090807060504030201
The C-term for S(42) is
3C1+C1/10+C2=32333435363738394041424241403938373635343332
In general C=xCl+{x-l}CIIIO+C2 for x=3,4,5, .. ,8. For x=8 the last term S(97) of
this sequence is reached.
110
Case 4: n=102, p=333667. period=3. Interval: 1oo:5;n$999.
S(102)=12_ .. _101102 102101_ .. _21
S(105)=12_ .. _101102103104105105104103102101_ .. _21
C= 103104105105104103
C1= 100100100100100100
=0 (mod 333667)
=0 (mod 333667)
C2= 3004005005004003 =0 (mod 333667)
Removing 1 or 2 zeros at the end of C1 does not affect the congruence modulus
333667, we have:
C1'= 10010010010010010
C1"= 1001001001001001
We now form the combinations:
xC1+yC1' +zC1' , +C2=0 (mod. 333667)
=0 (mod 333667)
=0 (mod 333667)
'This, in my mind, is quite remarkable: All 18-digit integers formed by the
concatenation of three consecutive 3-digit integers followed by a concatenation of the
same integers in descending order are divisible by 333667, example
376377378378377376=0 (mod 333667). As far as the C-terms are concerned all Sen)
in the range 100:5;n<999 could be divisible by 333667, but they are not Why? It is
because S(1OO) and S(101) are not divisible by 333667. Consequently n=I00+3k and
101+3k can not be used for insertion of an appropriate C-value as we did in the case
of S(102). 'This completes the explanation of the remarkable fact that every third term
S(I02+3j) in the range l00<-I1$99 is divisible by 333667.
1bese three cases have shown what causes the periodicity of the divisibility of the
Smarandache symmetric sequence of the second order by primes. The mechanism is
the same for the other periodic sequences.
Beyond 1000
We have seen that numbers of the type:
10101010_10, 100100100-100, 10001000_1000, etc
play an important role. Such numbers have been factorized and the occurrence of our
favorite primes 11, 37, ... ,333667 have been listed in table 7. In this table a number
like 100100100100 has been abbreviated 4(100) or q(E), where q and E are listed in
separate columns.
Question 1. Does the sequence of terms Sen) divisible by 333667 continue beyond
WOO?
Although Sen) was partially factorized only up n=2oo we have been able to draw
conclusions on divisibility up n=1ooo. The last term that we have found divisible by
333667 is S(999). Two conditions must be met for there to be a sequence of terms
divisible by p=333667 in the
Condition 1. There must exist a number 10001000 ... 1000 divisible by 333667 to
ensure the periodicity as we have seen in our case studies.
In table 7 we find q=9, E=looo. This means that the periodicity will be 9 - if it exists,
i.e. condition 1 is met
111
Condition 2. There must exist a term S(n) with divisible by 333667 which
will constitute the first term of the sequence.
The last term for n<1000 which is divisible by 333667 is S(999) from which we build
S(108)=12_999_1000 ___ 1008_1008 __ 1000_999--21
where we deconcatenate 100010011002 ... 10081008 ... 10011000 which is divisible by
333667 and provides the C-term (as introduced in the case studies) needed to generate
the sequence, i.e. condition 2 is met.
We conclude that S(I008+9j)=O (mod 333667) for j=O,1,2, ... 999. The last term in
this sequence is S(9999). From table 7 we see that there could be a sequence with the
period 9 in the interval l0000<.-n=s;99999 and a sequence with period 3 in the interval
It is not difficult to verify that the above conditions are fIlled also
in these intervals. This means that we have:
S(1oo8+9j)=O (mod 333667)
S(IOOO8+9j)=O (mod 333667)
S(I00002+3j)=O (mod 333667)
forj=OI,2, ... ,999, i.e.
for j=OI,2, .. ,9999, i.e. 10
4
:5n::;lOS-1
forj=OI,2, ... ,99999, i.e.
It is one of the fascinations with large numbers to find such properties. This
extraordinary property of the prime 333667 in relation to the Smarandache symmetric
sequence probably holds for n>106. It easy to loose contact with reality when plying
with numbers like this. We have S(999999)=O (mod 333667). What does this number
S(999999) look like? Applying (1) we find that the number of digits D(999999) of
S(999999) is
D(999999)=26.10
6
-2(l06-)19=11777778
Let's write this number with 80 digits per line, 60 lines per page, using both sides of
the paper. We will need 1226 sheets of paper - more that 2 reams!
Question 2. Why is there no sequence of S(n) divisible by II in the interval
I ()()<-D:5999?
Condition!' We must have a sequence of the form 100100.: divisible by 11 to ensure
the periodicity. As we can see from table 7 the sequence 100100 fills the condition
and we would have a periodicity equal to 2 if the next condition is met.
Condition 2. There must exist a term S(n) with divisible by II which would
constitute the first term of the sequence. This time let's use a nice property of the
prime 11:
Hf=(-lt (mod 11)
Let's deconcatenate the number a_b corresponding to the concatenation of the
numbers a and b: We have:
( -a+b if l+[loglob] is odd
a_b=a . 1 OI+[Iog'Db] +
l a+b if 1+[Iog1ob] is even
Let's first consider a deconcatenated middle part of S(n) where the concatenation is
done with tbree-digit integers. For convienience I have chosen a concrete example -
the generalization should pose no problem
112
27327427527527427352-7+3-2+7-4+2-7+5-2+7-5+2-7+4-2+7-350 (mod 11)
+-+-+-+-+-+-+-+-+-
It is easy to see that this property holds independent of the length of the sequence
above and whether it start on + or -. It is also easy to understand that equivalent
results are obtained for other primes although factors other than +1 and -1 will enter
into the picture.
We now return to the question of finding the first term of the sequence. We must start
from n=97 since S(97) it the last term for which we know that S(nF-O (mod 11). We
form:
9899100101-n_D-101100999852 (mod III independent of nc1000.
+-+-+-+-+-- --+-+-+-+-+-
This means that S(n)=:2 (mod 11) for 1 ~ 9 9 and explains why there is no
sequence divisible by 11 in this interval.
Question 3. Will there be a sequence divisible by 11 in the interval 1000<-11$9991
Condition 1. A sequence 10001000 ... 1000 divisible by 11 exists and would provide a
period of 11, se table 7.
Condition 2. We need to find one value n ~ 1 for which S(n)=O (mod 11). We
have seen that S(999)=2 (mod 11). We now look at the sequences following S(999).
Since S(999)=2 (mod 9) we need to insert a sequence l0001001..m_m ... l0011()()()=9
(mod 11) so that S(m)=O (mod H).Unfortunately m does not exist as we will see
below
1000100052 (mod 11)
+-+-+-+-
1 1
100010011001100052. (mod 11)
+-+-+-+-+-+-+-+-
111 1
1 1
100010011002100210011000=0 (mod l1l
+-+-+-+-+-+-+-+-+-+-+-+-
1 1 1 1 1 1
1 2 2 1
100010011002100310031002100110005-457 (mod 11)
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
111 1 1 1 1 1
12332 1
Continuing this way we find that the residues form the period 2,2,0,7,1,4,5,4,1,7,0.
We needed a residue to be 9 in order to build sequences divisible by 9. We conclude
that Sen) is not divisible by 11 in the interval l000<-1lS9999.
Trying to do the above analysis with the computer programs used in the early part of
this study causes overflow because the large integers involved. However, changing
the approach and performing calculations modulus 11 posed no problems. The above
method was preferred for clarity of presentation.
113 ~
Epilog
There are many other questions that may be interesting to look into. This is left to the
reader. The author's main interest in this has been to develop means by which it is
possible to identify some properties of large numbers other than the so frequently
asked question as to whether a big number is a prime or not. There are two important
ways to generate large numbers that I found particularly interesting - iteration and
concatenation. In this article the author has drawn on work' done previously,
references below. In both these areas very large numbers may be generated for which
it may be impossible to find any practical use - the methods are often more important
than the results.
References:
1. Tabirca, S. and T., On Primality of the Smarandache Symmetic Sequences, Smarandache Notions
Journal, Vol. 12, No 1-3 Spring 2001,114-121.
2. Smarandache F., Only Problems, Not Solutions, Xiquan Publ., Pheonix-Chicago, 1993.
3. lbstedtH. Surflllg on the Ocean afNumbers, Erhus University Press, Vail, 1997.
4. lbstedt H, Some Sequences of Large Integers, Fibonacci Quarterly, 28(1990), 200-203.
114