Combinatorics on Words Sergey Kitaev

SMSTC Module

Lecture 6. Universal partial words and universal partial cycles

Contents

1 U-p-words and u-p-cycles for sets of words 2

1.1 U-p-words with up to three 3’s . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 U-p-words with any number of 3’s . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 U-p-cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2 Shortening u-cycles and u-words for permutations 8

2.1 Using linear extensions of posets for shortening . . . . . . . . . . . . . . . . 8

3 Problems 12

Bibliography 13

Universal words (u-words) and universal cycles (u-cycles) considered in Lecture 5 are
examples of combinatorial generation methods (celebrated Gray codes, not to be considered
by us, would be another such example), and they provide a compact way to store all objects
in question, possibly to be used in an exhaustive search through all objects in a practical
problem. However, typical u-words/u-cycles appearing in applications are still rather long,
so accessing the data recorded by them takes significant time and storing these words is
expensive. Thus, u-words and u-cycles are not efficient enough for some problems, and
shortening them is of fundamental importance in, e.g. shortgun sequencing, a method in
DNA sequence assembly, where many gigabytes of storage are required for recoding reads
for a single genome.
To go around this issue, one can use the wildcard symbol 3 to shorten u-words and
u-cycles. For example, instead of the u-word 0001011100 of length 10 that contains each
of the binary words of length 3 exactly once, one can use 330111 of length 6, which has
the same properties if one thinks of a 3 as the universal symbol that can be either 0 or
1. In particular, the factor 301 is responsible for covering the words 001 and 101. On the
other hand, 031 is not a u-p-word for {0, 1}2 (10 is not covered, while 01 is covered twice).
Words containing 3(s) are called partial words. U-cycles (resp., u-words) containing 3(s)
are called universal partial cycles (resp., universal partial words) or u-p-cycles (resp., u-p-
words) for brevity. The notions of u-p-words and u-p-cycles for combinatorial structures
were introduced in [4] and studied further in [3, 5, 6].

1
 In various contexts 3 is called a universal symbol, a joker, a wildcard symbol, a
uncertain symbol, and a hole. Instead of 3, one can use, e.g. ? or ∗. Partial words were
introduced by Berstel and Boasson in [1] in 1999, and since then they have been studied
in connection to comparing genes, avoidability/unavoidability, squares, overlap-freeness,
etc; see the book [2].
In this lecture, we will consider shortening de Bruijn sequences and u-words/u-cycles
for permutations. More research in this direction can be found in [3] where u-p-words for
word-patterns and set partitions are considered.

1 U-p-words and u-p-cycles for sets of words

So, our goal is to construct a partial word (or a partial cycle) that covers each word from
a given set exactly once. In such words the 3 symbol can correspond to any symbol in
the alphabet. 3n = 33 | {z· · · 3} is a trivial u-p-word (note that it does not make sense in
n times
the cyclic case for n ≥ 2).
Let u = u1 u2 · · · uN be a u-p-word for {0, 1}n . We let S(u, i, n) denote the set of all
words that are obtained from the factor ui ui+1 · · · ui+n−1 by replacing any occurrences of
3 by a letter in {0, 1}n . The edges of the spanning subgraph H(u, n) of the de Bruijn graph
B(n, 2) are all the edges induced between the pairs of sets S(u, i, n) and S(u, i + 1, n) for
all i. For example, for u = 03011100 (n = 3), we have S(u, 1, n) = {000, 010}, S(u, 2, n) =
{001, 101}, S(u, 3, n) = {011}, S(u, 4, n) = {111}, S(u, 5, n) = {110}, S(u, 6, n) = {100}.
The edges of H(u, n) are the solid lines in the following figure:

Thus, for example, to find a u-p-word u = u1 3u3 u4 u5 u6 u7 u8 , with ui ∈ {0, 1}, we would
be searching for the spanning subgraph H(u, 3) of the form

in the de Bruijn graph B(3, 2), or would be trying to prove that such a subgraph does not
exist. That is how we obtain the u-p-word u = 03011100.

2
 The following lemma was useful in proving several results on u-p-words. There,
0̄ = 1 and 1̄ = 0, so x̄ is taking the complement of x.

Lemma 1.1 ([4]). Let u = u1 u2 · · · uN be a u-p-word for An , A = {0, 1, . . . , α − 1}, n ≥ 2,

such that uk = 3 and uk+n ̸= 3 (we require that k+n ≤ N ). Then for all i = 1, 2, . . . , n−1
we have that if ui ̸= 3, then uk+i = ui . Moreover, we have that if un ̸= 3, then α = 2
and uk+n = u¯n .

A schematic illustration of Lemma 1.1 would be as follows:

a b c··· 3 a b c··· ̸= 3
u=
k k+n

a b c · · · d ̸= 3 3 a b c··· d¯
u= ⇒α=2
n k k+n

1.1 U-p-words with up to three 3’s

The following theorem deals with the case of a single 3 in the first position.

Theorem 1.2 ([4]). For any n ≥ 2, there is a u-p-word for {0, 1}n with a single 3 at the
first position that begins with 30n−1 1.

Proof. Any Hamiltonian cycle C in the de Bruijn graph B(n, 2) has the following subpath
(or else, the node 0n is not reachable):

· · · → 10n−1 → 0n → 0n−1 1 → · · · .

If we start C at 0n−1 1, follow it and do not visit the last two vertices, namely, 10n−1 and
0n , then the obtained path will give us a u-p-word that starts with 30n−1 1.

Let c(w, n) be obtained by concatenating several copies of w, truncating the resulting

word at length n and complementing the last symbol. For example, c(001, 8) = 00100100̄ =
00100101 and c(001, 9) = 001001001̄ = 001001000. The following theorem deals with the
case of a single 3 in at most (n − 1)-th position.

Theorem 1.3 ([4]). For n ≥ 3 and k ∈ {2, . . . , n − 1}, there is a u-p-word w for {0, 1}n
with a single 3 at the k-th position that begins with 01k−2 3c(01k−1 , n).

Now we deal with a single 3 in position n.

Theorem 1.4 ([4]). For any n ≥ 2, there is no u-p-word w for {0, 1}n with a single 3 in
position n.

3
Proof. The case of n = 2 is easy, so assume that n ≥ 3.
Suppose w exists. Then its length is 2n − 1 and it must be of the form

w = b1 b2 · · · bn−1 3a1 a2 · · · an w′ .

Consider u = a1 a2 · · · an−1 ān which must be covered by w. Note that u cannot begin right
after the 3. If xu is a factor of w for x ∈ {0, 1}, where one of u’s letters can be given by 3,
then xa1 a2 · · · an−1 is covered twice by w (since 3 can play the role of x); contradiction.
Thus, u must begin with the leftmost letter of w. But then the role of ān in u will be
played by the 3, so that the word a1 a2 · · · an is covered twice by w; contradiction.

Two more theorems with non-existence results are as follows starting with a single
3 for a non-binary alphabet.

Theorem 1.5 ([4]). For A = {0, 1, . . . , α −1}, α ≥ 3, and any n ≥ 2, there is no u-p-word
for An with exactly one 3.

Theorem 1.6 ([4]). There is no u-p-word for {0, 1}n with a single 3 in position k in the
following three cases: n = 3 and k = 4, and n = 4 and k ∈ {5, 7}.

It was conjectured in [4], and checked for all 2 ≤ n ≤ 13, that the theorems above on
binary alphabets describe the only situations (up to reverse of p-words) of non-existence
of u-p-words the alphabet {0, 1} with a single 3.
Based on the theorems presented above, Table 1 gives examples of u-p-words for
{0, 1}n with a single 3 at a specified position.
One can also construct an infinite family of binary u-p-words with two 3’s.

Theorem 1.7 ([4]). For any n ≥ 4, there is a u-p-word for {0, 1}n with two 3’s that
begins with 30n−1 1n−2 310n−2 1.

Examples of u-p-words for {0, 1}n with two 3’s can be found in Table 2.
Three non-existence results for two 3’s for an arbitrary alphabet are given next.

Theorem 1.8 ([4]). For any n ≥ 2 and α ≥ 2, there is no u-p-word w for An , A =

{0, 1, . . . , α − 1}, that contains two 3’s such that the distance between the 3’s is different
from n and w has the form w = uxv where neither u nor v contain any 3’s and |u| ≥ n
and |v| ≥ n.

Theorem 1.9 ([4]). For any n ≥ 2 and α ≥ 2, there is no u-p-word w for An , A =

{0, 1, . . . , α − 1}, that contains two 3’s and that has the form w = x3y3z with |x| ≥ n,
|y| ≤ n − 2 and |z| ≥ |y| + 1.

Theorem 1.10 ([5]). There does not exist a u-p-word w = u33v over any alphabet α ≥ 2,
when n ≥ 4 and u and v are (possibly empty) words.

4
 n k
1 1 3
2 1 3011
2 —
3 1 300111010
2 03011100
3 —
4 —
4 1 300011110100101100
2 03010011011110000
3 0130111100001010
4 —
5 —
6 011003011110100
7 —
8 001111030010110
5 1 30000111110111001100010110101001000
2 0301011000001101001110111110010001
3 013011000001000111001010111110100
4 01130111110000010100100011010110
5 —
6 0010130010011101111100000110101
7 0100113010000010101101111100011
8 0100110301000001110010111110110
9 0111001030111110110100110000010
10 0100110113010001111100000101011
11 0101000001301011111001110110001
12 0101000001130101101111100010011
13 0010011010113001010000011111011
14 0011101111100300110100010101100
15 0101000001001130101101111100011
16 0010000011010113001010011111011

Table 1: Examples of u-p-words for {0, 1}n with a single 3 at position k

Let us consider a special case of Theorem 1.10 to demonstrate a proving technique.

Theorem 1.11. For n ≥ 2, there is no binary u-p-word w that begins with 033.

Proof. The case of n = 2 is trivial, so we assume that n ≥ 3. Suppose that

w = 0⋄⋄x1 x2 · · · xn−2 w′ .
If n = 3 then 00x1 is covered twice (starting in positions 1 and 2). Thus n ≥ 4. No
x1 x2 · · · xn−2 can occur in x2 x3 · · · xn−2 w′ because otherwise ⋄⋄x1 x2 · · · xn−2 would result

5
 n=2 33
n=3 330111
30010113
n=4 30001131001011
30001011310011
00131103001
n=5 301003101011000001110111110010
300001113100010010101100110111
300001001310001101011111011001
300001001113100011001010110111
300001010111310001101100100111
03001130100010101101111100000
03010110300011101111100100110
03010111030001101100100111110
03010111110300011011001001110
03010110111030001100100111110
0030011300101011011111010000
0130110010111030100000111110
0130110010111110301000001110
01301000001010110001111103110
001301013001110111110000010
011301101001030111110000010
011301101010010003011111000
01130111110001101010000010310
011301101000001111100010010131
01001311103010000011011001

Table 2: Examples of u-p-words for {0, 1}n with two 3’s

in w covering some word (ending with x1 x2 · · · xn−2 ) twice. w must cover all four words
of the form x1 x2 · · · xn−2 ∗ ∗, where ∗ is “we do not care” symbol over {0, 1}.
w = 0 ⋄ ⋄ x1 x2 x3 ··· xn−3 xn−2 ···
x1 x2 x3 x4 x5 x6 ···
x1 x2 x3 x4 x5 ··· ∗ ∗
x1 x2 x3 x4 ··· xn−2 ∗ ∗

This leads to the fact that x1 = x2 = · · · = xn−2 = 0 and thus 0n is covered twice
by w (starting in positions 1 and 2); contradiction.

Some examples of binary u-p-words with three 3’s are listed in Table 3, and two ex-
ample of u-p-words with four 3’s are as follows: 3333011111 and 300103011131001101100003.

6
 n
3 333
4 33301111
33001311010
30013110300
03001311030
5 3001030111310011011000001
3000011131000100110110010131
30000111031000101001101011113
3000010011131000110110010131
30000101110310001101010011113
300001111101310001011001301
30000101011103100011010011113
30000101001110310001101011113
3000011011001113100010010131
30000110101001110310001011113
3000011011001001113100010131
300001001010111110031101300
031100300111110110100010131

Table 3: Examples of u-p-words for {0, 1}n with three 3’s for n = 3, 4, 5

1.2 U-p-words with any number of 3’s

Theorem 1.12 ([5]). For any n ≥ 2, 3n−1 01n is a u-p-word for {0, 1}n .

Theorem 1.13 ([4]). For n ≥ 4 and any 2 ≤ d ≤ n − 2, there is no u-p-word for {0, 1}n
that begins with 3d xd+1 · · · xn+2 with xi ∈ A for all i = d + 1, . . . , n + 2.

Note an arbitrary alphabet A = {1, 2, . . . , α} in the following theorems.

Theorem 1.14 ([5]). Let w be a u-p-word for An with |A| ≥ 3. If wi = 3 then wj = 3

for all i ≡ j mod n. Thus √
each window of size n contains the same number of 3’s. This
number cannot exceed n − n− 7
4 − 1
2 3s.

Theorem 1.15 ([5]). For α ≥ 3, if gcd(α, n) = 1 then there are no nontrivial u-p-words
for An .

Theorem 1.16 ([5]). For α ≥ 2 and 2 ≤ k ≤ n/2, there does not exist a u-p-word
w = u3k v where u and v are (possibly empty) words.

Theorem 1.17 ([4]). For A = {0, 1, . . . , α − 1}, α ≥ 3, and any k ≥ 2, for large enough
n there is no u-p-word for An with exactly k many 3’s.

7
1.3 U-p-cycles

A binary u-p-cycle exists for n = 4, namely, 31103001, but not for n = 2, 3. In fact, [5]
gives a construction of u-p-cycles for any even-size alphabet and n = 4, e.g. for α = 4:

0013110300331123021313030233132320133103203331232213330322333323

For odd-size alphabets u-p-cycles do not exist. Interestingly, the following theorem shows
that for non-binary alphabets, cyclic and non-cyclic cases are essentially equivalent!

Theorem 1.18 ([5]). If w is a u-p-word for An with α ≥ 3, then the first n − 1 letters of
w equal the last n − 1 letters of w.

Theorem 1.19 ([4]). For any n ≥ 2 and α ≥ 2, there is no u-p-cycle w for An , where
A = {0, 1, . . . , α − 1}, that contains a letter from A and a 3 at distance n from each other.

Corollary 1.20 ([4]). Let α ≥ 2 and n ≥ 2. If w is a cyclic u-p-word for An then

|w| = αn−ℓ for some ℓ, 1 ≤ ℓ < n, such that n divides ℓαn−ℓ .

As an immediate corollary of the last result, we can exclude the existence of u-p-
cycles for many combinations of α and n.

Corollary 1.21 ([4]). Let A = {0, 1, . . . , α − 1} and n ≥ 2. If gcd(α, n) = 1, then there

is no u-p-cycle for An . In particular, for α = 2 and odd n, there is no u-p-cycles for An .

Proof. Suppose that such a u-p-cycle u = u1 u2 · · · uN exists. Then, by Corollary 1.20

we have N = αn−d for some d, 1 ≤ d ≤ n − 1, such that n divides dN . However, as
gcd(α, n) = 1, n does not divide N = αn−d , so n must divide d, which is impossible,
yielding a contradiction.

2 Shortening u-cycles and u-words for permutations

This section is based on [6], where the authors approach the problem of shortening u-cycles
and u-words for permutations from two different angles:

• The non-determinism allowing shortening is in using incomparable elements and

considering linear extensions of partially ordered sets (posets), and then compression
possibilities for u-cycles and u-words for permutations can be studied; see Section 2.1.

• The second approach is a plain extension of the studies in [4, 5] (presented in Sec-
tion 1) to the case of permutations. However, using the “wildcard” symbol 3 seems
to be inefficient in the context (it is dominated by non-existence results), so one can
consider its refinement 3D , where D is a subset of the alphabet in question. In any
case, in these lecture notes we omit any discussions of 3’s or 3D ’s in connection to
permutations.

8
2.1 Using linear extensions of posets for shortening

To illustrate the idea here, consider the word 112, which is claimed by us to be a u-cycle for
all permutations of length 3, where we modify the notion of a u-cycle for n-permutations
by allowing equal elements in a factor of length n and declaring them to be incomparable.
Thus, we managed to shorten a “classical” u-cycle for these permutations, say, 145243.
Indeed, we can treat equal elements as incomparable elements, while the relative order of
these incomparable elements to the other elements must be respected. Thus, 112 encodes
all permutations whose last element is the largest one, namely, 123 and 213; starting at the
second position (and reading the word cyclically), we obtain the word 121 encoding the
permutations 132 and 231, and finally, starting at the third position, we (cyclically) read
the word 211 encoding the permutations 312 and 321. More generally, it is clear that the
word 11 · · · 1} 2 = 1n−1 2 encodes all n-permutations and is of length n (instead of length
| {z
n−1 times
n! for earlier defined u-cycles for permutations). However, there are other compression
possibilities creating u-cycles of lengths between n and n!. For example, the word 1232
is also a u-cycle for permutations of length 3. Note that the word of the form 11 · · · 1 is
a (trivial) u-word for all permutations of the respective length (when words are not read
cyclically), while this word is not a u-cycle because the definition of a u-cycle cannot be
applied to it.

Definition 2.1. Two different permutations, π1 · · · πn and σ1 · · · σn , are called twin per-
mutations, or twins, if

• red(π1 · · · πn−1 ) = red(σ1 · · · σn−1 ), and

• |πn − π1 | = |σn − σ1 | = 1.

Examples of twins are 3124 and 4123, 2413 and 3412, and 23451 and 13452.
Recall the notion of clustering of the graph of overlapping permutations where each
cluster collects all n-permutations whose first n − 1 elements form the same pattern, that
is, these elements in each permutation in the cluster are order-isomorphic to the same
(n − 1)-permutation. We refer the Reader to Figures 1 and 2 to check their understanding
of the following four lemmas in the cases of n = 3 and n = 4, respectively.

“12” “21”
123 213
132 312
231 321

Figure 1: Clustering the graph of overlapping permutations of order 3

Lemma 2.2 ([6]). Each cluster has exactly one pair of twins.

9
 1234
1243
“123”
1342
2341
1324 3124
1423 4123
“132” “312”
1432 4132
2431 4231

2314 2134
2413 2143
“231” “213”
3412 3142
3421 3241
4321
4312
“321”
4213
3214

Figure 2: Clustering the graph of overlapping permutations of order 4

Proof. Let the signature (the first n − 1 elements of the permutations in the reduced form)
of a cluster be “x1 · · · xn−1 ”. The only possibilities to create twin permutations are to
−
1 or x1 at the end of x1 · · · xn−1 , and these possibilities always exist.
adjoin x+

By parallel edges between clusters we mean multiple edges oriented in the same way.
In particular, a pair (resp., a triple) of parallel edges is called a double edge (resp., a triple
edge). In what follows, double and triple edges from a cluster X to a cluster Y will be
denoted, respectively, by X Y and X Y .

Lemma 2.3 ([6]). For any cluster X, there is a unique cluster Y such that X Y .

Also, for no clusters X and Y , we have X Y .

Proof. Both of the statements follow from the fact that parallel edges can only be pro-
duced by twins (the last (n − 1) elements in non-twin permutations in a cluster cannot be
isomorphic), but by Lemma 2.2, there is only one such pair in each cluster.

10
Lemma 2.4 ([6]). For any cluster Y , there is a unique cluster X such that X Y .

Proof. Let the signature of Y be “x1 · · · xn−1 ”. Then the only double edge that can come
to Y is given by the permutations x−
n−1 x1 · · · xn−1 and xn−1 x1 · · · xn−1 (both belonging to
+

the same cluster with the signature “xn−1 x1 · · · xn−2 ”).

By Lemmas 2.3 and 2.4, the clustered graph of overlapping permutations can be
partitioned into a disjoint union of cycles formed by double edges.

Lemma 2.5 ([6]). Any of the disjoint cycles formed by the double edges goes through
exactly n − 1 distinct clusters.

Proof. Since double edges are formed by twin permutations, we can assume that any such
cycle is of the form:

x1 x2 · · · xn−1 x+ x2 x3 · · · xn−1 x1 x+ xn−1 x1 · · · xn−2 x+

1 2 ··· n−1
x1 x2 · · · xn−1 x−
1 x2 x3 · · · xn−1 x1 x−
2 xn−1 x1 · · · xn−2 x−
n−1

where the last cluster is linked to the first one by a double edge. Since all xi ’s are distinct,
the cycle must involve exactly n − 1 clusters.

“12” “21”
123 212
121 321

Figure 3: Applying incomparable elements on distance 2 for 3-permutations

Theorem 2.6 ([6]). Using incomparable elements at distance n−1, one can obtain u-words
for n-permutations of lengths n! + n − 1, n!, n! − (n − 1), . . . , n! − (n − 1)! + n − 1.

Proof. We have already seen in Lecture 5 that the clustered graph of overlapping n-
permutations is balanced and strongly connected for any n ≥ 1. There are (n − 1)!
clusters. By Lemma 2.5, there are (n − 2)! = (n − 1)!/(n − 1) disjoint cycles formed
by double edges, and we can decide in which cycles to replace every double edge by a
single edge thus maintaining the property of the graph (whose nodes are clusters) being
balanced. This action will correspond to replacing every double edge of the form

x1 x2 · · · xn−1 x+
1 x2 x3 · · · xn−1 x1 x+
2
x1 x2 · · · xn−1 x−
1 x2 x3 · · · xn−1 x1 x−
2

11
 by

x1 x2 · · · xn−1 x1 x2 x3 · · · xn−1 x1 x2

and thus introducing incomparable elements inside some of clusters. Strong connectivity
in the graph will clearly be maintained as well, since our action is in simply replacing a
pair of equivalent edges by a single edge.
So, by removing double edges in such a way we guarantee the existence of an Eulerian
cycle going through the clusters, which gives the existence of the respective Hamiltonian
cycle (recall that to each word or permutation there corresponds exactly one edge), and
thus the existence of a respective u-word for n-permutations.

See Figure 3 for an illustration of the proof of Theorem 2.6 in the case of n = 2, and
Figure 4 for that in the case of n = 3 when both of the double edge cycles were replaced.
Examples of u-words that can be obtained from Figures 3 and 4, respectively, are 123212
and 123847687657859423123.

1234
1243 “123”
1231

1324 3123
“132” 1423 4132 “312”
1321 4231

2314 2134
“231” 2312 2132 “213”
3421 3241

4321
“321” 4312
3213

Figure 4: Applying incomparable elements at distance 3 for 4-permutations

Conjecture 2.7 ([6]). Using incomparable elements at distance n − 1, one can obtain
u-cycles for n-permutations of lengths n!, n! − (n − 1), n! − 2(n − 1), . . . , n! − (n − 1)!.

12
 Examples of u-cycles supporting Conjecture 2.7, that can be obtained from Figures 3
and 4, respectively, are 1232 and 123847687657859423.

3 Problems
1. Explain why 030133 is not a u-p-word for binary words of length 3.
2. Theorem 1.3 guarantees the existence of a u-p-word for {0, 1}15 with the single 3 is
position 5. What prefix of length 15 such a u-p-word can have?
3. Which u-p-words in Table 2 have the structure of u-p-words described in Theo-
rem 1.7?
4. Prove Theorem 1.12.
5. Let A = {0, 1, . . . , α − 1} and α ≥ 2. Show that no u-p-word for An can contain
exactly one non-3 symbol.
6. Explain why the following table shows, for n ≥ 4, non-existence of a binary u-p-word
of the form w = 0x33x1 x2 · · · xn−2 w′ . [Hint: Only three out of four cases, C1 –C4 ,
hold.]
w 0 x 3 3 x1 x2 ··· xn−6 xn−5 xn−4 xn−3 xn−2 ···
C1 x1 x2 x3 x4 x5 x6 ··· xn−2 ∗ ∗
C2 x1 x2 x3 x4 x5 ··· xn−3 xn−2 ∗ ∗
C3 x1 x2 x3 x4 ··· xn−4 xn−3 xn−2 ∗ ∗
C4 x1 x2 x3 ··· xn−5 xn−4 xn−3 xn−2 ∗ ∗

7. What can you say about the word u = 34321432345234 in connection to Conjec-
ture 2.7?

References
[1] J. Berstel and L. Boasson. Partial words and a theorem of Fine and Wilf. Theoretical
Computer Scence 218 (1999) 1, 135–141.
[2] F. Blanchet-Sadri. Algorithmic combinatorics on partial words, Chapman &
Hall/CRC, 2008.
[3] T. Z. Q. Chen and S. Kitaev. On universal partial words for word-patterns and set
partitions, RAIRO – Theoretical Informatics and Applications 54 (2020), Art 5, 14pp.
[4] T. Z. Q. Chen, S. Kitaev, T. Mutze and B. Y. Sun. On universal partial words,
Discrete Math. & Theoretical Computer Science 19(1) (2017) #16.
[5] B. Goeckner, C. Groothuis, C. Hettle, B. Kelld, P. Kirkpatrick, R. Kirsch, and R.
Solava. Universal partial words over non-binary alphabets, Theoretical Computer Sci-
ence 713 (2018) 56–65.

13
[6] S. Kitaev, V. Potapov and V. Vajnovszki. On shortening u-cycles and u-words for
permutations, Discrete Applied Math. 260 (2019) 203–213.

14