Every graph is local antimagic total and its application to local

antimagic (total) chromatic numbers

Gee-Choon Lau
Faculty of Computer & Mathematical Sciences,
Universiti Teknologi MARA (Segamat Campus),
85000, Johor, Malaysia.
[email protected]
arXiv:1906.10332v3 [math.CO] 26 Nov 2020

Abstract
A graph G = (V, E) of order p and size q is said to be local antimagic if there exists a
+
bijection g : E(G) → {1, P2, . . . , q} such that for any pair of adjacent vertices u and v, g (u) 6=
+ +
g (v), where g (u) = uv∈E(G) g(uv) is the induced vertex color of u under g. We also say
G is local antimagic total if there exists a bijection f : V ∪ E → {1, 2, . . . , p +P q} such that for
any pair of adjacent vertices u and v, w(u) 6= w(v), where w(u) = f (u) + uv∈E(G) f (uv)
is the induced vertex weight of u under f . The local antimagic (and local antimagic total)
chromatic number of G, denoted χla (G) (and χlat (G)), is the minimum number of distinct
induced vertex colors (and weights) over all local antimagic (and local antimagic total) la-
belings of G. We also say a local antimagic total labeling is local super antimagic total if
f (v) ∈ {1, 2, . . . , p} for each v ∈ V (G). In [Proof of a local antimagic conjecture, Discrete
Math. Theor. Comp. Sc., 20(1) (2018), #18], the author proved that every connected
graph of order at least 3 is local antimagic. Using this result, we provide a very short proof
that every graph is local antimagic total. As an application, we showed that there exists
close relationship between χla (G ∨ K1 ) and χlat (G). A sufficient condition is also given for
the corresponding local super antimagic total labeling. Sharp bounds of χlat (G) and close
relationships between χlat (G) and χla (G ∨ K1 ) are found. Bounds of χlat (G − e) in terms of
χlat (G) for a graph G with an edge e deleted are also obtained. These relationships are used
to determine the exact values of χlat (G), χlat (G − e), χla (G ∨ K1 ) and χla ((G − e) ∨ K1 ) for
many dense and sparse graphs G. The work of this paper also provides many open problems
on χlat (G). We also conjecture that each graph G of order at least 3 has χlat (G) ≤ χla (G).

Keywords: Local antimagic (total) labeling, Local antimagic (total) chromatic number, Join
of graphs, Dense and sparse graphs

2010 AMS Subject Classifications: 05C78; 05C15.

1 Introduction
Consider a (p, q)-graph G(V, E) of order p and size q. If e is an edge of G, denote by G − e
the graph G with the edge e deleted. Let g : E(G) → [1, q] be P a bijective edge labeling that
+
induces a vertex labeling g : V (G) → N such that g (v) = +
uv∈E(G) g(uv). We say g is
a local antimagic labeling of G if g+ (u) 6= g+ (v) for each uv ∈ E(G) [1, 2]. The minimum
number of distinct vertex labels induces by g is called the local antimagic chromatic number of
G, denoted χla (G) [1]. Clearly, χla (G) ≥ χ(G). Let G ∨ H be the join of G and H with vertex
set V (G) ∪ V (H) and edge set E(G) ∪ E(H) ∪ {uv : u ∈ V (G), v ∈ V (H)}. Let G + H be
the disjoint union of G and H with vertex set V (G) ∪ V (H) and edge set E(G) ∪ E(H). For
convenience, nG is the disjoint union of n ≥ 1 copies of G, and nK1 = On . For positive integers
a

a and b, assume [a, b] = {a, a + 1, . . . , b} and b = 0 if a < b.

1
Let f : V (G) ∪ E(G) → [1, p + q] be a bijective total labeling that induces a vertex labeling
w : V (G) → N such that X
wf (v) = f (v) + f (uv)
uv∈E(G)

and is called the weight of v for each vertex v ∈ V (G). We say f is a local antimagic total labeling
of G (and G is local antimagic total) if w(u) 6= w(v) for each uv ∈ E(G). Clearly, w corresponds
to a proper vertex coloring of G if each vertex v is assigned the color w(v). Let w(f ) be the num-
ber of distinct vertex weights induces by f . The number min{w(f ) | f is a local antimagic total labeling of G}
is called the local antimagic total chromatic number of G, denoted χlat (G). Clearly, χlat (G) ≥
χ(G). It is well known that determining the chromatic number of a graph G is NP-hard [11].
Thus, in general, it is very difficult to determine χla (G) and χlat (G).
In [3], the author proved that every connected graph of order at least 3 is local antimagic.
Using this result, we provide in Section 2 a very short proof that every graph is local antimagic
total. As an application, we showed that there exists close relationship between χla (G ∨ K1 ) and
χlat (G). A sufficient condition is also given for the corresponding local super antimagic total
labeling. Sharp bounds of χlat (G) and close relationships between χlat (G) and χla (G ∨ K1 ) are
found. Bounds of χlat (G − e) in terms of χlat (G) are also obtained. These relationships are then
used to determine the exact value of χlat (G), χlat (G − e), χla (G ∨ K1 ) and χla ((G − e) ∨ K1 ) for
many dense and sparse graphs G in Section 3. We also conjecture that each graph G of order
at least 3 has χlat (G) ≤ χla (G).

2 Preliminary results
By definition, χlat (On ) = χlat (Kn ) = n. Moreover, if G is a graph with n isolated vertices, then
χlat (G) ≥ n. In what follows, we only consider nonempty graphs.

Theorem 2.1: Every graph G is local antimagic total.

Proof. For G∨ K1 and p ≥ 1, we let G be of order p and size q such that V (G) = {vi | 1 ≤ i ≤ p},
V (K1 ) = {v} and E(G ∨ K1 ) = E(G) ∪ {vvi | i = 1, 2, . . . , p}. It is obvious that each graph
G of order p ≤ 3 are local antimagic total. We now assume G is of order p ≥ 4. In [3],
the author proved that every graph without isolated edges (by definition, necessarily without
isolated vertices) admits a local antimagic labeling. Thus, G ∨ K1 is local antimagic. Let g be
a local antimagic labeling of G ∨ K1 . Define a total labeling f : V (G) ∪ E(G) → [1, p + q] of
G such that f (e) = g(e) for each edge e ∈ E(G) and f (vi ) = g(vvi ). Clearly, w(vi ) = g+ (vi ).
Thus, w(vi ) 6= w(vj ) if vi vj ∈ E(G) and f is a local antimagic total labeling of G.

In [10], the authors extended the concept of local antimagic total labelings to local super
antimagic total labelings (and chromatic number, denoted χlsat ) that assign only integers in
[1, |V (G)|] to the vertices of G. Since a local super antimagic total labeling is also a local an-
timagic labeling, but the converse may not hold, we then have χ(G) ≤ χlat (G) ≤ χlsat (G). By
an argument similar to that of Theorem 2.1, we get a sufficient condition for a graph G to admit
a local super antimagic total labeling.

Theorem 2.2: If G ∨ K1 admits a local antimagic labeling that assigns only integer(s) in
[1, |V (G)|] to the edge(s) not belong to G, then G is local super antimagic total.

2
The following three theorems follows directly from definition and Theorems 1, 7, 8 in [10]
respectively.

Theorem 2.3: If T is a tree on n ≥ 2 vertices with k pendants, then χlat (T ) ≤ n − k + 1.

For graphs G and H, the corona product of G and H, denoted G ⊙ H, is obtained from G and
|V (G)| ≥ 2 copies of H by joining the i-th vertex of G to every vertex of the i-th copy of H.

Theorem 2.4: If H is a regular local super antimagic total graph of order n ≥ 2 and Om is a
null graph of order m ≥ 2, then χlat (H ⊙ Om ) ≤ χlsat (H) + 1, where (m, n) 6= (odd, even).

Theorem 2.5: Suppose H is a regular graph of order n ≥ 2 and G is obtained from H ∨ K1 by

joining each vertex of H to m pendant vertices for m ≥ 1. If G is local super antimagic total,
then

(i) χlat (G) ≤ χlsat (H ∨ K1 ) + 1, for m ≥ 2 and (m, n) 6= (odd, even),

(ii) χlat (G) ≤ χlsat (H) + 2, for (m, n) = (odd, even),

(iii) χlat (G) ≤ χlsat (H) + 2, for m = 1.

The next theorem shows that χlat (G) can be arbitrarily large for a graph G with small χ(G).

Theorem 2.6: If G = K2 + On , n ≥ 1, then

(
2 for n = 1, 2,
χlat (G) =
n otherwise.

Proof. Let V (G) = {u1 , u2 , vi | 1 ≤ i ≤ n}. For n ≥ 1, define f (ui ) = i, f (u1 u2 ) = 3 and
f (vi ) = i + 3. We now have w(u1 ) = 4, w(u2 ) = 5 and w(vi ) = i + 3. Thus, χlat (G) ≤ 2 for
n = 1, 2, and χlat (G) ≤ n for n ≥ 3. By definition, χlat (G) ≥ χ(G) = 2 and since all the isolated
vertices must have distinct weights, this implies that χlat (G) ≥ n. So, the theorem holds.

Theorem 2.7: Let G ∨ K1 be as defined under Theorem 2.1.

(a) χlat (G) ≤ χla (G ∨ K1 ) − 1 and the equality holds if χ(G) = χla (G ∨ K1 ) − 1.

(b) P
Suppose χlat (G) = χ(G ∨ K1 ) − 1 with a corresponding local antimagic total labeling f . If
p
i=1 f (vi ) 6= w(vj ), 1 ≤ j ≤ p, then χla (G ∨ K1 ) = χ(G ∨ K1 ).

Proof.

(a) Suppose χla (G∨K1 ) = c. From the proof of Theorem 2.1, we know that every local antimagic
labeling of G∨ K1 that induces c distinct vertex labels corresponds to a local antimagic total
labeling of G that induces c− 1 distinct vertex weights. Thus, χlat (G) ≤ c− 1. By definition,
χlat (G) ≥ χ(G). Thus, χlat (G) = c − 1 if χ(G) = c − 1.

(b) Suppose χ(G ∨ K1 ) = c so that χlat (G) = c − 1. Let g : E(G ∨ K1 ) → [1, p + q]Psuch that
g(e) = f (e) if e ∈ E(G), and g(vvi ) = f (vi ) for each vi ∈ V (G). Clearly, w(v) = pi=1 f (vi ),
and w(vi ) = f + (vi ) 6= f + (vj ) = w(vj ) if vi vj ∈ E(G). Since w(v) 6= w(vj ), g is a local
antimagic labeling that induces c distinct vertex weights and χla (G ∨ K1 ) ≤ c. Since
χla (G ∨ K1 ) ≥ χ(G ∨ K1 ) = c, we have χla (G ∨ K1 ) = c.

3
By an argument similar to that for Theorem 2.7, we have the following theorem.

Theorem 2.8: Let G ∨ K1 be as defined under Theorem 2.1. Suppose G ∨ K1 admits a local
antimagic labeling that assigns only integers in [1, p] to edges not belong to G.

(a) χlsat (G) ≤ χla (G ∨ K1 ) − 1 and the equality holds if χ(G) = χla (G ∨ K1 ) − 1.

(b) Suppose χlsat (G) = χ(G ∨ K1 ) − 1 with a corresponding local super antimagic total labeling
p+1

f . If 2 6= w(vj ), 1 ≤ j ≤ p, then χla (G ∨ K1 ) = χ(G ∨ K1 ).

Corollary 2.9: For any graph G of order at least 2,

(1) χ(G) ≤ χlat (G) ≤ χla (G ∨ K1 ) − 1;

(2) if G ∨ K1 admits a local antimagic labeling that assigns only integers in [1, p] to edges not
belong to G, then χ(G) ≤ χlsat (G) ≤ χla (G ∨ K1 ) − 1.

For 1 ≤ i ≤ p and 1 ≤ j ≤ m, let G be a graph of order p ≥ 2 and size q ≥ 1 with V (G ∨ mK1 ) =

{vi , uj } and E(G ∨ mK1 ) = E(G) ∪ {ui vj }.

Theorem 2.10: Suppose g is a local antimagic labeling of G ∨ 2K1 that induces a minimum
number of vertex labels and 2p + q + 1 + g + (u1 ) 6= g+ (vi ), 1 ≤ i ≤ p, then
(
χla (G ∨ 2K1 ) if g + (u1 ) = g+ (u2 );
χlat (G ∨ K1 ) ≤
χla (G ∨ 2K1 ) − 1 if g + (u1 ) 6= g+ (u2 ).

Moreover, equality holds if χ(G ∨ K1 ) equals the respective upper bound.

Proof. For 1 ≤ i < j ≤ p, define a bijection f : V (G ∨ K1 ) ∪ E(G ∨ K1 ) → [1, 2p + q + 1] such

that for 1 ≤ i < j ≤ p, f (vi vj ) = g(vi vj ) if vi vj ∈ E(G), f (u1 vi ) = g(u1 vi ), f (vi ) = g(u2 vi ),
and f (u1 ) = 2p + q + 1. Clearly, w(vi ) = g+ (vi ) and w(u1 ) = 2p + q + 1 + g+ (u1 ). Since
2p + q + 1 + g + (u1 ) 6= g + (vi ) for 1 ≤ i ≤ p, f is a local antimagic total labeling of G ∨ K1 that
induces χla (G ∨ 2K1 ) distinct vertex weights if g+ (u1 ) = g + (u2 ), and induces χla (G ∨ 2K1 ) − 1
distinct vertex weights if g+ (u1 ) 6= g + (u2 ). Moreover, equality holds if χ(G ∨ K1 ) equals the
respective upper bound.

The following lemmas are analogous to Lemmas 2.2 – 2.5 in [7].

Lemma 2.11: Suppose G is a d-regular graph of order p and size q with an edge e. If f is a
local antimagic total labeling of G, then g = p + q + 1 − f is also a local antimagic total labeling
of G with w(g) = w(f ). Moreover, suppose w(f ) = χlat (G) and f (e) = 1 or f (e) = p + q. If
χ(G − e) = χlat (G), then χlat (G − e) = χlat (G). Otherwise, χlat (G − e) ≤ χlat (G).

Proof. Let x, y ∈ V (G). Here, wg (x) = (d + 1)(p + q + 1) − wf (x) and wg (y) = (d + 1)(p +
q + 1) − wf (y). Therefore, wf (x) = wf (y) if and only if wg (x) = wg (y). Thus, g is also a local
antimagic total labeling of G with w(g) = w(f ).
If f (e) = p + q, then we may consider g = p + q + 1 − f . So without loss of generality, we may
assume that f (e) = 1. Define h : V (G − e) ∪ E(G − e) → [1, p + q − 1] such that h(x) = f (x) − 1
and h(xy) = f (xy) − 1 for xy 6= e. So, wh (x) = wf (x) − d − 1 for each vertex x of G − e.

4
Therefore, wf (x) = wf (y) if and only if wh (x) = wh (y). Thus, h is also a local antimagic total
labeling of G with w(h) = w(f ). Consequently, χ(G − e) ≤ χlat (G − e) ≤ χlat (G). The theorem
holds.

Note that if G is a regular edge-transitive graph, then χlat (G − e) ≤ χlat (G).

Lemma 2.12: Suppose G is a graph of order p and size q and f is a local antimagic total
labeling of G. For any x, y ∈ V (G), if
(i) wf (x) = wf (y) implies that deg(x) = deg(y), and
(ii) wf (x) 6= wf (y) implies that (p + q + 1)(deg(x) − deg(y)) 6= wf (x) − wf (y),
then g = p + q + 1 − f is also a local antimagic total labeling of G with w(g) = w(f ).

Proof. For any x, y ∈ V (G), we have wg (x) = (deg(x) + 1)(p + q + 1) − wf (x) and wg (y) =
(deg(y) + 1)(p + q + 1) − wf (y). If wf (x) = wf (y), then condition (i) implies that wg (x) = wg (y).
If wf (x) 6= wf (y), then condition (ii) implies that wg (x) 6= wg (y). Thus, g is also a local
antimagic total labeling of G with w(g) = w(f ).

For t ≥ 2, consider the following conditions for a graph G.

(i) χlat (G) =St and f is a local antimagic total labeling of G that induces a t-independent
partition ti=1 Vi of V (G).
(ii) For each x ∈ Vk , 1 ≤ k ≤ t, deg(x) = dk satisfying wf (x) − da 6= wf (y) − db , where x ∈ Va
and y ∈ Vb for 1 ≤ a < b ≤ t.
(iii) There exist two non-adjacent vertices u, v with u ∈ Vi , v ∈ Vj for some 1 ≤ i 6= j ≤ t such
that
(a) |Vi | = |Vj | = 1 and deg(x) = dk for x ∈ Vk , 1 ≤ k ≤ t; or
(b) |Vi | = 1, |Vj | ≥ 2 and deg(x) = dk for x ∈ Vk , 1 ≤ k ≤ t except that deg(v) = dj − 1; or
(c) |vi |, |Vj | ≥ 2 and deg(x) = dk for x ∈ Vk , 1 ≤ k ≤ t except that deg(u) = di − 1,
deg(v) = dj − 1,
each satisfying wf (x) + da 6= wf (y) + db , where x ∈ Va and y ∈ Vb for 1 ≤ a 6= b ≤ t.

Lemma 2.13: Let H be obtained from G with an edge e deleted. If G satisfies conditions (i)
and (ii) and f (e) = 1, then χ(H) ≤ χlat (H) ≤ t.

Proof. By definition, we have the lower bound. Define g : E(H) → [1, |E(H)|] such that
g(e′ ) = f (e′ ) − 1 for each e′ ∈ E(H). Observe that g is a bijection with wg (x) = wf (x) − dk − 1
for each x ∈ Vk , 1 ≤ k ≤ t. Thus, wg (x) = wg (y) if and only if x, y ∈ Vk , 1 ≤ k ≤ t. Therefore, g
is a local antimagic total labeling of H with w(g) = w(f ). Thus, χlat (H) ≤ t.

Lemma 2.14: Suppose uv ∈ E(G). Let H be obtained from G with an edge uv added. If G
satisfies conditions (i) and (iii), then χ(H) ≤ χlat (H) ≤ t.

Proof. By definition, we have the lower bound. Define g : E(H) → [1, |E(H)|] such that g(uv) =
1 and g(e) = f (e) + 1 for e ∈ E(G). Observe that g is a bijection with wg (x) = wf (x) + dk + 1
for each x ∈ Vk , 1 ≤ k ≤ t. Thus, wg (x) = wg (y) if and only if x, y ∈ Vk , 1 ≤ k ≤ t. Therefore, g
is a local antimagic total labeling of H with w(g) = w(f ). Thus, χlat (H) ≤ t.

5
3 The exact local antimagic total chromatic number
We first consider some dense graphs. An amalgamation of two graphs G1 and G2 over a fixed
graph H is the simple graph obtained by identifying the vertices of two induced subgraphs
isomorphic to H, one of G1 and the other of G2 . Suppose G is a graph with a Kr , r ≥ 1,
subgraph. Let A(mG, Kr ) be the amalgamation of m ≥ 2 copies of G of order at least r + 1
along Kr . When r = 1, the graph is also known as one-point union of graphs. For m ≥ 2
and n > r ≥ 1, let V (A(mKn , Kr )) = {vi,j | 1 ≤ i ≤ m, 1 ≤ j ≤ n} and E(A(mKn , Kr )) =
{vi,j vi,k | 1 ≤ i ≤ m, 1 ≤ j < k ≤ n} where v1,j = · · · = vm,j for each 1 ≤ j ≤ r. Note that
A(mK2 , K1 ) ∼ = K1,m and A(mK3 , K1 ) is the friendship graph fm , m ≥ 2.

Theorem 3.1: For m ≥ 3, n ≥ r + 1 ≥ 3, if 3 ≤ m ≤ 4 and 2 ≤ r ≤ 3, or if (m − 1)(r − 1) <

r+1

2 , then χlat (A(mKn , Kr )) = n.

Proof. Let V (A(mKn , Kr )) = {vi,j | 1 ≤ i ≤ m, 1 ≤ j ≤ n} and E(A(mKn , Kr )) = {vi,j vi,k | 1 ≤

i ≤ m, 1 ≤ j < k ≤ n} such that {vi,j | 1 ≤ j ≤ r} is the set of vertices common to each Kn .
Thus, v1,j
= v2,j = · · · =
vm,j for 1 ≤ j ≤ r. Define f : V (A(mKn , Kr )) ∪ E(A(mKn , Kr )) →
[1, m n+1
2 − (m − 1) r+1
2 ] as follows.
h i
(i) f (vi,j vi,k ) = m k−1 r



2 − 2 + j − 1 + i for 1 ≤ i ≤ m, 1 ≤ j < k ≤ n, r + 1 ≤ k ≤ n;
n r



(ii) f (vi,j ) = m( 2 − 2 ) + (m + 1 − i)(n − 1) + 2 − j for 1 ≤ i ≤ m, r + 1 ≤ j ≤ n.

Let t = m( n2 − 2r ). We now consider the labeling of the remaining r vertices and 2r edges

that belong to the Kr .

Case (1). Suppose m = 4. If r = 2, we are left with t + n − 1, t + 2n − 2, t + 3n − 3 to label
the vertices and edge of the unlabeled K2 subgraph. Let f (vi,2 ) = t + n − 1, f (vi,1 ) = t + 2n − 2
and f (vi,1 vi,2 ) = t + 3n − 3. We now have the followings.

4 X
n
(  )
h k − 1
X i
w(vi,1 ) = 4 − 1 + i + 2t + 5n − 5
2
i=1 k=3
"  # "  #
n n
= 16 − n + 2 + 10(n − 2) + 8 − 1 + 5n − 5
3 2
8n3 n
= − 4n2 + − 1
3 3
and
4 X
n
(   )
X k−1
w(vi,2 ) = 4 + i + 2t + 4n − 4
2
i=1 k=3
  "  #
n n
= 16 + 10(n − 2) + 8 − 1 + 4n − 4
3 2
8n3 46n
= − 4n2 + − 32.
3 3

6
For 3 ≤ j ≤ n,
j−1
X n
X
w(vi,j ) = f (vi,j ) + f (vi,j vi,k ) + f (vi,j vi,k )
k=1 k=j+1
j−1
  "   ! #
n X j −1
= 4 − 4 + (5 − i)(n − 1) + 2 − j + 4 +k−2 +i +
2 2
k=1
n
"   ! #
X k−1
4 +j −2 +i
2
k=j+1
  "  #  
n j−1 j
= 4 − 4 + 5(n − 1) + 2 − j + 4(j − 1) −2 +4 +
2 2 2
   !
n j
4(n − j)(j − 2) + 4 −
3 3
2n3 11n 4j 3 17j
= + 4nj − + − 8j 2 + − 3.
3 3 3 3
which is an increasing function. Moreover,

!
8n3 n 2n3 11n 4n 3 17n
w(vi,1 ) − w(vi,n ) = − 4n2 + − 1 − + 4n2 − + − 8n2 + −3
3 3 3 3 3 3
2n3 5n
= − + 2 > 0.
3 3
Since w(vi,3 ) < · · · < w(vi,n ) < w(vi,1 ) < w(vi,2 ), we have f is a local antimagic total labeling
that induces n distinct vertex weights so that χla (A(mKn , Kr )) ≤ n.
If r = 3, we are left with t + n − 2, t + n − 1, t + 2n − 3, t + 2n − 2, t + 3n − 4, t + 3n − 3 to label
the vertices and edges of the unlabeled K3 subgraph. Let f (vi,3 ) = t + n − 2, f (vi,2 ) = t + n − 1,
f (vi,2 vi,3 ) = t + 2n − 3, f (vi,1 ) = t + 2n − 2, f (vi,1 v1,2 ) = t + 3n − 4 and f (vi,1 v1,3 ) = t + 3n − 3.
We now have the followings.

4 X
n
(  )
X h k − 1 i
w(vi,1 ) = 4 − 3 + i + 3t + 8n − 9
2
i=1 k=4
8n3 92n
= − 2n2 − + 53,
3 3

4 X
n
(  )
X h k − 1 i
w(vi,2 ) = 4 − 2 + i + 3t + 6n − 8
2
i=1 k=4
8n3 50n
= − 2n2 − +6
3 3

7
 and
4 X
n
(  )
X h k − 1 i
w(vi,3 ) = 4 − 1 + i + 3t + 6n − 8
2
i=1 k=4
8n3 2n
= − 2n2 − − 42.
3 3
For 4 ≤ j ≤ n,
j−1
X n
X
w(vi,j ) = f (vi,j ) + f (vi,j vi,k ) + f (vi,j vi,k )
k=1 k=j+1
2n3 35n 4j 3 17j
= + 4nj − + − 8j 2 + −3
3 3 3 3
which is an increasing function. Moreover,

8n3 92n  2n3 35n 4n3 17n 

w(vi,1 ) − w(vi,n ) = − 2n2 − + 53 − + 4n2 − + − 8n2 + −3
3 3 3 3 3 3
2n3 74n
= + 2n2 − + 56 > 0.
3 3
Since w(vi,4 ) < · · · < w(vi,n ) < w(vi,1 ) < w(vi,2 ) < w(vi,3 ), we have f is a local antimagic total
labeling that induces n distinct vertex weights so that χla (A(mKn , Kr )) ≤ n.
Case (2). Suppose m = 3. The proof is similar to Case (1) and is thus omitted. We note
2 2
that when r = 2, we can get w(vi,1 ) = 3n (n−1) 2 + 2n − 5, w(vi,2 ) = 3n (n−1)
2 + 10n − 22, and
n3 5n 3 2
for 3 ≤ j ≤, w(v1,j ) = 2 + 3nj − 2 + j − 6j + 4j − 2, which is an increasing function with
3 3n3
w(vi,1 ) − w(vi,n ) = 2n3 − 5n 29n
3 + 2 > 0. For r = 3, we can get w(vi,1 ) = 2 − 2 + 18, w(vi,2 ) =
3n3 15n 3n3 3n n3 17n 3 2
2 − 2 −8, w(vi,3 ) = 2 + 2 −35, and for 4 ≤ j ≤ n, w(vi,j ) = 2 +3nj − 2 +j −6j +4j −2,
2
which is an increasing function with w(vi,1 ) − w(vi,n ) = 3n − 10n + 20 > 0.
Case (3). Suppose (m − 1)(r − 1) < r+1


2 . We
may assume (m, r) 6= (4, 2), (4, 3), (3, 2), (3, 3).
Note that the r+1 n+1 r+1



2 unused integers in [1, m 2 − (m − 1) 2 ] are as in

(i) ai = {t + in − r + 2 − i, t + in − r + 3 − i, t + in − r + 4 − i, . . . , t + in − i} that has r − 1

integers for 1 ≤ i ≤ m, and

(ii) am+1 = {t+mn−m+1, t+mn−m+2, . . . , t+mn+ 2r −r(m−1)} that has 2r +m−r(m−1)

integers.

Let us list the above integers in increasing order to get

A(1) = t + n − r + 1, A(2) = t + n − r + 2, A(3) = t + n − r + 3, . . . , A(r − 1) = t + n − 1,
A(r) = t + 2n − r, A(r + 1) = t + 2n − r + 1, A(r + 2) = t + 2n − r + 2, . . . , A(2r − 2) = t + 2n − 2,
A(2r − 1) = t + 3n − r − 1, A(2r) = t + 3n − r, A(2r + 1) = t + 3n − r + 1, . . . , A(3r − 3) = t + 3n − 3,
..
.
A((m − 1)(r − 1) + 1) = t + mn − r − m + 2, A((m − 1)(r − 1) + 2) = t + mn − r − m + 3, A((m −
1)(r − 1) + 3) = t + mn − r − m + 4, . . . , A(m(r − 1)) = t + mn − m, and
A(m(r − 1) + 1) = t + mn − m + 1, A(m(r − 1) + 2) = t + mn − m + 2, . . . , A( r+1


2 ) =

8
 r

t + mn + 2 − r(m − 1).

Let Mr×r be an r × r symmetric matrix that corresponds to the labeling of the Kr subgraph
such that the integers A(1) to A(r) are assigned to the main diagonal from bottom right to top
left, and the remaining integers from A(r + 1) onwards are assigned to Mj,j+1 to Mj,r for j = 1
to j = r consecutively. Thus, the row sums of Mr×r that corresponds to the total labeling of
the Kr -subgraph are:
S1 = A(r)+A(r+1)+A(r+2)+A(r+3)+A(r+4)+A(r+5)+A(r+6)+A(r+7)+. . . +A(2r−1);
S2 = A(r + 1) + A(r − 1) + A(2r) + A(2r + 1) + A(2r + 2) + A(2r + 3) + A(2r + 4) + A(2r +
5) + . . . + A(3r − 3);
S3 = A(r + 2) + A(2r) + A(r − 2) + A(3r − 2) + A(3r − 1) + A(3r) + A(3r + 1) + A(3r + 2) +
. . . + A(4r − 6);
S4 = A(r + 3) + A(2r + 1) + A(3r − 2) + A(r − 3) + A(4r − 5) + A(4r − 4) + A(4r − 3) + A(4r −
2) + . . . + A(5r − 10);
S5 = A(r + 4) + A(2r + 2) + A(3r − 1) + A(4r − 5) + A(r − 4) + A(5r − 9) + A(5r − 8) + A(5r −
7) + . . . + A(6r − 15);
S6 = A(r + 5) + A(2r + 3) + A(3r) + A(4t − 4) + A(5r − 9) + A(r − 5) + A(6r − 14) + A(6r −
13) + . . . + A(7r − 21);
..
. Pr−1
A((x + 1)r − x+1


Sr = A(1) + x=1 2 ).
In general, for 2 ≤ x ≤ r − 1, we have
Sx = A(r+x−1)+A(2r+x−3)+A(3r+x−5)+. . .+A((x−1)r− x−1


2 +1)+A(r − x + 1)+A(xr−
x
x
x+1

2 + 1) + A(xr − 2 + 2) + . . . + A((x + 1)r − 2 ). It is now obvious that S1 < S2 < · · · < Sr .
Moreover,
r−1  
X r+1
S1 = t + 3n − r − 1 + (t + 2n − r − 1 + y) = r(t + 2n − 1) − + n.
y=1
2

Note that for given m, n, r,

n
m X n
m X
!
X X hk − 1  
r i
f (vi,j vi,k ) = m − +j −1 +i
2 2
i=1 k=r+1 i=1 k=r+1
n r  
m+1
 h  i
r
2 2
= m − + (n − r) + m (n − r) j − 1 −
3 3 2 2

is an increasing function from j = 1 to j = r. Thus, we can now conclude that w(vi,1 ) <
w(vi,2 ) < · · · < w(vi,r ) and
n r  
m+1
  
r
2 2
w(vi,1 ) = S1 + m − + (n − r) − m (n − r)
3 3 2 2
         
r+1 2
 n r m+1 2 r
= r(t + 2n − 1) − +n+m − + (n − r) − m (n − r)
2 3 3 2 2

9
Also note that for r + 1 ≤ j ≤ n, we have
j−1
X n
X
w(vi,j ) = f (vi,j ) + f (vi,j vi,k ) + f (vi,j vi,k )
k=1 k=j+1
j−1
" #
X hj − 1   
r i
= t + (m + 1 − i)(n − 1) + 2 − j + m − +k−1 +i
2 2
k=1
n
" #
X hk − 1
 
r i
m − +j −1 +i
2 2
k=j+1
h j − 1   r  i  
j
= t + (m + 1)(n − 1) + 2 − j + m(j − 1) − −1 +m +
2 2 2

h n     
j i h r i
m − + m(n − j) j − 1 − .
3 3 2

It is routine to check that the terms involving the variable j form an increasing function so that
w(vi,r+1 ) < w(vi,r+2 ) < · · · < w(vi,n ). Since n ≥ r + 1, we get


 n r   
r+1 2 m+1
w(vi,1 ) − w(vi,n ) = r(t + 2n − 1) − +n+m − + (n − r) −
2 3 3 2
  hn − 1  r 
2 r i
m (n − r) − t + (m + 1)(n − 1) + 2 − n + m(n − 1) − −1 +
2 2 2
 !
n
m
2
1 h
= 2m2 r 3 − 3mr 3 − 3m2 nr 2 + 3mnr 2 + 3mr 2 − 3r 2 + 3mn2 r + 3m2 nr −
6
6mnr + 12nr − 5m2 r − 3mri− 9r + m2 n3 − 3mn3 − 3m2 n2 + 6mn2 +
5m2 n − 6mn + 6n + 6m − 6
> 0.

Thus, w(vi,r+1 ) < w(vi,r+2 ) < · · · < w(vi,n ) < w(vi,1 ) < · · · < w(vi,r ) so that f is a local
antimagic total labeling that induces n distinct vertex weights. Therefore, χlat (A(mKn , Kr )) ≤
n. Since χla (A(mKn , Kr )) ≥ χ(A(mKn , Kr )) = n, the theorem holds.

Remark 3.1: We note that when m = 2, we may have w(vi,1 )−w(vi,n ) < 0. We have checked by
computer computation that all the induced vertex weights are distinct for 3 ≤ r+1 ≤ n ≤ 15000.

Example 3.1: For Cases (1) and (2), we give the labeling matrix of A(4K7 , K3 ) denoted
A(i, 7, 3) for 1 ≤ i ≤ 4.

10
 A(1, 7, 3) v1,1 v1,2 v1,3 v1,4 v1,5 v1,6 v1,7 sum A(2, 7, 3) v2,1 v2,2 v2,3 v2,4 v2,5 v2,6 v2,7 sum
v1,1 84 89 90 1 13 29 49 355 v2,1 − − − 2 14 30 50 96
v1,2 89 78 83 5 17 33 53 358 v2,2 − − − 6 18 34 54 112
v1,3 90 83 77 9 21 37 57 374 v2,3 − − − 10 22 38 58 128
v1,4 1 5 9 94 25 41 61 236 v2,4 2 6 10 88 26 42 62 236
v1,5 13 17 21 25 93 45 65 279 v2,5 14 18 22 26 87 46 66 279
v1,6 29 33 37 41 45 92 69 346 v2,6 30 34 38 42 46 86 70 346
v1,7 49 53 57 61 65 69 91 445 v2,7 50 54 58 62 66 70 85 445

A(3, 7, 3) v3,1 v3,2 v3,3 v3,4 v3,5 v3,6 v3,7 sum A(4, 7, 3) v4,1 v4,2 v4,3 v4,4 v4,5 v4,6 v4,7 sum
v3,1 − − − 3 15 31 51 100 v4,1 − − − 4 16 32 52 104
v3,2 − − − 7 19 35 55 116 v4,2 − − − 8 20 36 56 120
v3,3 − − − 11 23 39 59 132 v4,3 − − − 12 24 40 60 136
v3,4 3 7 11 82 27 43 63 236 v4,4 4 8 12 76 28 44 64 236
v3,5 15 19 23 27 81 47 67 279 v4,5 16 20 24 28 75 48 68 279
v3,6 31 35 39 43 47 80 71 346 v4,6 32 36 40 44 48 74 72 346
v3,7 51 55 59 63 67 71 79 445 v4,7 52 56 60 64 68 72 73 445
The vertex weights are w(vi,1 ) = 355+96+100+104 = 685, w(vi,2 ) = 358+112+116+120 = 706,
w(vi,3 ) = 374 + 128 + 132 + 136 = 770, w(vi,4 ) = 236, w(vi,5 ) = 279, w(vi,6 ) = 346 and
w(vi,7 ) = 445 for 1 ≤ i ≤ 4.

Example 3.2: For (m − 1)(r − 1) < r+1

2 , we give the labeling matrix of A(2K6 , K3 ) denoted
A(i, 6, 3) for 1 ≤ i ≤ 2.
A(1, 6, 3) v1,1 v1,2 v1,3 v1,4 v1,5 v1,6 sum A(2, 6, 3) v2,1 v2,2 v2,3 v2,4 v2,5 v2,6 sum
v1,1 33 34 35 1 7 15 125 v2,1 − − − 2 8 16 26
v1,2 34 29 36 3 9 17 128 v2,2 − − − 4 10 18 32
v1,3 35 36 28 5 11 19 134 v2,3 − − − 6 12 20 38
v1,4 1 3 5 32 13 21 75 v2,4 2 4 6 27 14 22 75
v1,5 7 9 11 13 31 23 94 v2,5 8 10 12 14 26 24 94
v1,6 15 17 19 21 23 30 125 v2,6 16 18 20 22 24 25 125

The vertex weights are w(vi,1 ) = 125 + 26 = 151, w(vi,2 ) = 121 + 32 = 160, w(vi,3 ) = 134 + 38 =
172, w(vi,4 ) = 75, w(vi,5 ) = 94 and w(vi,6 ) = 125 for 1 ≤ i ≤ 2.

Theorem 3.2: For
m ≥ 2 and n ≥ r + 1 ≥ 2, if 3 ≤ m ≤ 4 and 2 ≤ r ≤ 3, or if

(m − 1)(r − 1) ≤ r+1
2 , then χla (A(mKn , Kr ) ∨ K1 ) = n + 1.

Proof. Suppose m = 4. Consider r = 2 and the local antimagic total labeling f in the proof of
Theorem 3.1. We have the sum of all vertex labels is
2
X 4 X
X n
f (v1,j ) + f (vi,j ) = (t + 2n − 2) + (t + n − 1) +
j=1 i=1 j=3
4 X n   
X n 
4 + (5 − i)(n − 1) − 2 − j
2
i=1 j=3
   
n n+1
= 2t + 16(n − 2) −4 + 10n2 − 27n + 21
2 2
> w(vi,2 ).

Since χlat (A(4Kn , K2 )) = n = χ(A(4Kn , K2 )∨K1 )−1, by Theorem 2.7, χla (A(4Kn , K2 )∨K1 ) =
n + 1.

11
For all other cases when m = 3, 4, one can check that f satisfies the conditions of Theorem 2.7.
The details are omitted.
1) < i r+1


Suppose (m − 1)(r h − 2 . It is routine to check that the sum of all vertex labels is
Sv = m2 (n − r) n2 − r2 + (n − r)(n − 1) m+1 + 2m(n − r) − 21 m(n − r)(n + r + 1) +




2
h
i h

mr n2 − 2r + n(r + 1) − r+1 n



2 while the smallest vertex labels is w(v i,r+1 ) = m 2 −
i h i h i
r
+ (m + 1)(n − 1) − r − mr + m r+1 + m n3 − r+1 + m(n − r − 1) r − 2r . Thus,






2 2 3
2
Sv − w(vi,r+1 ) = 21 m2 r 3 − 56 mr 3 − 12 m2 nr 2 + 21 mnr 2 − m2 r 2 + 2mr 2 − r2 − 21 m2 n2 r + 21 mn2 r +
m2 nr − 25 mnr + nr + 12 m2 r + 13 mr + r2 + 21 m2 n3 − 16 mn3 − 12 m2 n + mn 6 + m + 1 > 0. Thus, the
theorem holds.

Let e be an edge of A(mKn , Kr ) joining a vertex of the Kr subgraph and another vertex not in
Kr .

Theorem 3.3: For
m ≥ 2 and n ≥ r + 1 ≥ 2, if 3 ≤ m ≤ 4 and 2 ≤ r ≤ 3, or if

(m − 1)(r − 1) ≤ r+1
2 , then χlat (A(mKn , Kr ) − e) = n.

Proof. Without loss of generality, we may assume e = vi,1 vi,r+1 . Consider the local antimagic
total labeling f in the proof of Theorem 3.1. For Case (1) with m = 4 and r = 2, we have for
A(4Kn , K2 ), vertex vi,1 and vi,2 has degree 4n − 7, while vi,j has degree n − 1 for 3 ≤ j ≤ n. We
now have

(i) w(vi,1 ) − (4n − 7) 6= w(vi,2 ) − (4n − 7);

(ii) w(vi,j ) − (4n − 7) 6= w(vi,k ) − (n − 1) for j = 1, 2 and 3 ≤ k ≤ n;

(iii) w(vi,j ) − (n − 1) 6= w(vi,k ) − (n − 1) for 3 ≤ j < k ≤ n.

For all other cases, one can check that f satisfies the conditions of Lemma 2.13(ii) too. The
details are omitted. Thus, the theorem holds.

By an argument similar to that for Theorem 3.2, we also have the the following theorem. The
details are also omitted.

Theorem 3.4: For
m ≥ 2 and n ≥ r + 1 ≥ 2, if 3 ≤ m ≤ 4 and 2 ≤ r ≤ 3, or if

(m − 1)(r − 1) ≤ r+1
2 , then χla ((A(mKn , Kr ) − e) ∨ K1 ) = n + 1.

Theorem 3.5: For m, n ≥ 2, χlat (A(mKn , K1 )) = n.

Proof. Note that |V (A(mKn , K1 ))| + |E(A(mKn , K1 ))| = (mn − m + 1) + m n2 = m(n +

2)(n − 1)/2 + 1. Let u = v1,n = . . . = vm,n . Define f : V (A(mKn , K1 )) ∪ E(A(mKn , K1 )) →

[1, m(n + 2)(n − 1)/2 + 1] as follows.

(i) f (u) = m(n + 2)(n − 1)/2 + 1,

(ii) f (vi,j ) = m(n + 2)(n − 1)/2 − i(n − 1) + n − j for 1 ≤ i ≤ m, 1 ≤ j ≤ n − 1,

(iii) f (vi,j vi,k ) = m[n(j − 1) − 2j + k − j − 1] + i for 1 ≤ i ≤ m, 1 ≤ j < k ≤ n.

12
Observe that for 1 ≤ i ≤ m, 1 ≤ j ≤ n − 1,
j−1
X n
X
w(vi,j ) = f (vi,j ) + f (vi,j vi,k ) + f (vi,j vi,k )
k=1 k=j+1
j−1   !
X h k i
= m(n + 2)(n − 1)/2 − i(n − 1) + n − j + m n(k − 1) − +j −k−1 +i +
2
k=1
n   !
X h j i
m n(j − 1) − +k−j−1 +i
2
k=j+1
"       #
j−1 j j 2
= m(n + 2)(n − 1)/2 + n − j + m n − − + (j − 1) +
2 3 2
"   #
j
m(n − j) n(j − 1) − − j − 1 + m(n + j + 1)(n − j)/2. (3.1)
2

Moreover,
m X
X n−1
w(u) = m(n + 2)(n − 1)/2 + 1 + f (vi,j vi,n )
i=1 j=1
m n−1   !
X X j h i
= m(n+)(n − 1)/2 + 1 + m n(j − 1) − +n−j −1 +i
2
i=1 j=1
   !  
2 n 2 n−1 m
= m(n + 2)(n − 1)/2 + 1 + m n − (n − 1) − − (n − 1) .(3.2)
2 3 2

Therefore, w(v1,j ) = · · · = w(vm,j ) for 1 ≤ j ≤ n − 1. Suppose p = |V (A(mKn , K1 ))| and

q = |E(A(mKn , K1 ))|. Thus, the labeling matrix, denoted A(i, n, 1), is as follow where p + q is
assigned once in A(1, n, 1).

A(i, n, 1) vi,1 vi,2 vi,3 vi,4

vi,1 p + q − i(n − 1) + n − 2 i m+i 2m + i
vi,2 i p + q − i(n − 1) + n − 3 m(n − 1) + i mn + i
vi,3 m+i m(n − 1) + i p + q − i(n − 1) + n − 4 m(2n − 3) + i
vi,4 2m + i mn + i m(2n − 3) + i p + q − i(n − 1) + n − 5
.. .. .. .. ..
. . . .
.
vi,j m(j − 2) + i m(n + j − 4) + i m(n(j − 1) − 2c + 2) + i m(3n + j − 11) + i
.. .. .. .. ..
. . . . .
vi,n−1 m(n − 3) + i m(2n − 5) + i m(2n − 8) + i m(4n − 12) + i
vi,n m(n − 2) + i m(2n − 4) + i m(3n − 7) + i m(4n − 11) + i

13
 ... vi,j ... vi,n−1 vi,n
vi,1 ... m(j − 2) + i ... m(n − 3) + i m(n − 2) + i
vi,2 ... m(n + j − 4) + i ... m(2n − 5) + i m(2n − 4) + i
vi,3 ... m(2n + j − 7) + i ... m(2n − 8) + i m(3n − 7) + i
vi,4 ... m(3n + j − 11) + i ... m(4n − 12) + i m(4n − 11) + i
.. .. .. .. .. ..
. . . .
j

. j
.

vi,j ... p + q − i(n − 1) + n − j + 1 ... m(nj − 2 − j − 2) + i m(nj − 2 − j) + i
.. .. .. .. .. ..
. . . . . .
j


vi,n−1 ... m(nj − 2 −
j − 2) + i ... p + q − i(n − 1) m(n − 2)(n + 1)/2 + i
vi,n ... m(nj − 2j − j) + i ... m(n − 2)(n + 1) + i ⋆

Obviously, for 1 ≤ j ≤ n − 1, (i) there is exactly one identical entry in rows j and j + 1 given
by f (vi,j+1 vi,j ) = f (vi,j vi,j+1 ), (ii) f (vj+1,j+1) + 1 = f (vj,j ) for j + 1 < n, (iii) all other entries
in column k of row j + 1 are larger than the corresponding entries in row j for k 6= j, j + 1, and
(iv) f (u) in row n is the largest of all labels. Thus, w(vi,1 ) < w(vi,2 ) < · · · < w(vi,n ). Therefore,
f is a local antimagic total labeling that induces n distinct weights. Since χ(A(mKn , K1 )) = n,
the theorem holds.

Example 3.3: For m = 2, n = 5, the matrices A(i, 5, 1), i = 1, 2, are given below.
A(1, 5, 1) v1,1 v1,2 v1,3 v1,4 v1,5 w(vi,j ) A(2, 5, 1) v2,1 v2,2 v2,3 v2,4 v2,5 w(vi,j )
v1,1 28 1 3 5 7 44 v2,1 24 2 4 6 8 44
v1,2 1 27 9 11 13 61 v2,2 2 23 10 12 14 61
v1,3 3 9 26 15 17 70 v2,3 4 10 22 16 18 70
v1,4 5 11 15 25 19 75 v2,4 6 12 16 21 20 75
u = v1,5 7 13 17 19 29 ⋆ u = v1,5 8 14 18 20 ⋆ ⋆
f (u) + 4k=1 f (v1,5 v1,k ) = 85
P P4
k=1 f (v1,5 v1,k ) = 60
w(u) = 145.

Theorem 3.6: Let e be an edge of

(I) A(mK2 , K1 ), m ≥ 2, then χlat (A(mK2 , K1 ) − e) = 2;

(II) A(mKn , K1 ), m ≥ 2, n ≥ 3, that is not incident to the vertex u = v1,n = · · · = vm,n , then
χlat (A(mKn , K1 ) − e) = n,

(III) A(mKn , K1 ), m, n ≥ 3, that is incident to the vertex u = v1,1 = · · · = vm,1 , then

χlat (A(mKn , K1 ) − e) = n.

Proof. (I). Note that A(mK2 , K1 ) − e ∼

= K1,m−1 + K1 . Let V (K1,m−1 + K1 ) = {x, u, vi | 1 ≤
i ≤ m − 1} and E(K1,m−1 + K1 ) = {uvi | 1 ≤ i ≤ m − 1}. For 1 ≤ i ≤ m − 1, define
f : V (K1,m−1 +K1 )∪E(K1,m−1 +K1 ) → [1, 2m] such that f (x) = 2m, f (vi ) = i, f (uvi ) = 2m−i,
f (u) = m. Thus, w(x) = w(vi ) = 2m 6= w(u) = m(3m − 1)/2. Thus, χlat (A(mK2 , K1 ) − e) = 2.
(II). Suppose e = v1,1 v1,2 not incident to u = v1,n = · · · = vm,n . Let f be the local antimagic total
labeling of A(mKn , K1 ) in Theorem 3.5. Define h : V (A(mKn , K1 ) − e) ∪ E(A(mKn , K1 ) − e) →
[1, m(n + 2)(n − 1)/2] such that g = f − 1 for every vertex and edge of A(mKn , K1 ) − e.
Thus, wg (vi,j ) = wf (vi,j ) − n for 1 ≤ j ≤ n − 1, and wg (u) = wf (u) − m(n − 1) − 1. Note

14
that f (v1,1 v1,2 ) = 0 means edge e = v1,1 v1,2 is deleted. Clearly, wg (x) = wg (y) if and only if
wf (x) = wf (y) for x, y 6= u. Since wg (vi,n−1 ) > wg (vi,j ) for j < n − 1, by Lemma 2.13, suffice to
show that A = wf (u) − wf (v1,n−1 ) − (m − 1)(n − 1) > 0. From Equations 3.2 and 3.1, we have

(    !  )
n 2 2 n−1 m
A = m(n + 2)(n − 1)/2 + 1 + m n − (n − 1) − − (n − 1) −
2 3 2
( "       #
j−1 j j 2
m(n + 2)(n − 1)/2 + n − j + m n − − + (j − 1) +
2 3 2
"   # )
j
m(n − j) n(j − 1) − − j − 1 + m(n + j + 1)(n − j)/2 − (m − 1)(n − 1)
2
m2 n3 mn3 m2 n2 mn2 m2 n 4mn m2 5m
= − − + − + +n+ − − 1.
3 3 2 2 3 3 2 2

Clearly, A > 0 if m = 2 or n = 3. Suppose m ≥ 3 and n ≥ 4, then

1 4mn m2 5m
A = (2m2 n3 − 2mn3 − 3m2 n2 + 3mn2 − 2m2 n) + +n+ − −1
6 3 2 2
1 2 3
> (m n − 2mn3 + m2 n3 − 3m2 n2 + 3mn2 − 2m2 n)
6
1
≥ (4m2 n2 − 3m2 n2 + 3mn2 − 2m2 n) > 0.
6
Thus, g is a local antimagic total labeling that induces n distinct weights. Since χ(A(mKn , K1 )−
e) = n, we have χlat (A(mKn , K1 ) − e) = n.
(III). Suppose e = v1,1 v1,2 incident to u = v1,1 = · · · = vm,1 so that e is incident to u. Define
f : V (A(mKn , K1 )) ∪ E(A(mKn , K1 )) → [1, m(n + 2)(n − 1)/2] as follows.

(i) f (u) = m(n + 2)(n − 1)/2,

(ii) f (vi,j ) = m(n + 2)(n − 1)/2 − i(n − 1) + n − j for 1 ≤ i ≤ m, 2 ≤ j ≤ n,

(iii) f (vi,j vi,k ) = m[ k−1

2 + j − 1] + i − 1 for 1 ≤ i ≤ m, 1 ≤ j < k ≤ n.

Note that f (v1,1 v1,2 ) = 0 means edge e = v1,1 v1,2 is deleted. Observe that for 1 ≤ i ≤ m, 2 ≤
j ≤ n,

15
 j−1
X n
X
w(vi,j ) = f (vi,j ) + f (vi,j vi,k ) + f (vi,j vi,k )
k=1 k=j+1
j−1
"  # !
X j−1
= m(n + 2)(n − 1)/2 − i(n − 1) + n − j + m +k−1 +i−1 +
2
k=1
n
"  # !
X k−1
m +j −1 +i−1
2
k=j+1
h j − 1  i
= m(n + 2)(n − 1)/2 − j + 1 + m(j − 1) +n−j −1 +
2
     
j n j
m +m −m . (3.3)
2 3 3
Moreover,
m X
X n
w(u) = m(n + 2)(n − 1)/2 + f (vi,1 vi,k )
i=1 k=2
m X n
 ! 
k−1
X
= m(n+)(n − 1)/2 + m +i−1
2
i=1 k=2
   
2 n m
= m(n + 2)(n − 1)/2 + m + (n − 1) . (3.4)
3 2

Listing the labels in matrix form as in the proof of Theorem 3.5, we have w(vi,2 ) < · · · < w(vi,n ).
Suffice to show that w(u) > w(v1,n ). Since m, n ≥ 3, we have
    hn − 1  
2 n m n
i
w(u) − w(v1,n ) = m + (n − 1) + n − 1 − m(n − 1) −1 −m
3 2 2 2
m n2 3 mn 3 2
m n 2 3mn 2 2
5m n 3mn m 2 m
= − − + + − +n− + −1
6 2 2 2 6 2 2 2
> 0

Thus, f is a local antimagic total labeling that induces n+1 distinct weights. Since χ(A(mKn , K1 )−
e) = n, we have χlat (A(mKn , K1 ) − e) = n.

Example 3.4: For Theorem 3.6(III) with m = 3, n = 4 and e = v1,1 v1,2 incident to u = v1,1 =
· · · = v2,1 , the matrices A(i, 4, 1), i = 1, 2, 3, are given below. Note that the label of the common
vertex only appears in A(1, 4, 1).
A(1, 4, 1) v1,1 v1,2 v1,3 v1,4 w(v1,j ) A(2, 4, 1) v2,1 v2,2 v2,3 v2,4 w(v2,j )
u = v1,1 27 0 3 9 ⋆ u = v2,1 ⋆ 1 4 10 ⋆
v1,2 0 26 6 12 53 v2,2 1 23 7 13 53
v1,3 2 4 25 15 58 v2,3 3 5 22 16 58
v1,4 6 8 10 24 67 v2,4 7 9 11 21 67
f (u) + 4k=3 f (v1,1 v1,k ) = 39
P P4
k=2 f (v2,1 v2,k ) = 15
w(u) = 72.

16
 A(3, 4, 1) v2,1 v2,2 v2,3 v2,4 w(v3,j )
u = v3,1 ⋆ 2 5 11 ⋆ P4
v3,2 1 20 8 14 53 k=2 f (v3,1 v3,k ) = 18
v3,3 3 5 19 17 58
v3,4 7 9 11 18 67

Suppose m = 2. By computer search, we found that the labeling defined for Theorem 3.6(III)
gives w(u) 6= w(v1,j ) for 2 ≤ j ≤ n ≤ 10000.

Theorem 3.7: Let e be an edge of

(a) A(mKn , K1 ), m, n ≥ 2, then χla (A(mKn , K1 ) ∨ K1 ) = n + 1.

(b) A(mK2 , K1 ), m ≥ 2, then χla ((A(mK2 , K1 ) − e) ∨ K1 ) = 3;

(c) A(mKn , K1 ), m ≥ 2, n ≥ 3, that is not incident to the vertex u = v1,n = · · · = vm,n , then
χla ((A(mKn , K1 ) − e) ∨ K1 ) = n + 1,

(d) A(mKn , K1 ), m, n ≥ 3, that is incident to the vertex u = v1,1 = · · · = vm,1 , then χla ((A(mKn , K1 )−
e) ∨ K1 ) = n + 1.

Proof. (a) Let G = A(mKn , K1 ), m ≥ 2, n ≥ 3. Consider G and the local antimagic total
labeling f of G as in the proof of Theorem 3.5. We note that f (v1,n ) is the largest assigned
label. Moreover, in each Mi , 1 ≤ i ≤ m and 1 ≤ k ≤ n−1, f (ui,k ) > f (vi,n vi,k ). Thus, sum of all
the vertex labels under f is larger than all the induced vertex weights under f . Since χlat (G)) =
χ(G∨K1 )−1 = (n+1)−1, by Theorem 2.7(b), we conclude that χla (A(mKn , K1 )∨K1 ) = n+1.
(b) Let G = A(mK2 , K1 ) − e, m ≥ 2. Consider G and the local antimagic total labeling f
of G as in the proof of Theorem 3.6(I). It is easy to show that χla (A(mK2 , K1 ) − e) = 3 for
m = 2, 3. If m ≥ 4, we note that the sum of all vertex labels under f is m(m + 5)/2 which is
not equal to all the induced vertex weights under f . Since χlat (G) = χ(G ∨ K1 ) − 1 = 3 − 1, by
Theorem 2.7(b), we conclude that χla ((A(mKn , K1 ) − e) ∨ K1 ) = 3.
(c) Let G = A(mKn , K1 ) − e, m ≥ 2, n ≥ 3, where e = v1,1 v1,2 is not incident to the vertex
u = v1,n = · · · = vm,n . Consider G and the local antimagic total labeling f of G as in the proof
of Theorem 3.6(II). It is easy to show that the sum of all vertex labels under f is 12 m(n − 1)(n +
1)[m(n − 1) + 1] which is larger than all the induced vertex weights under f . Since χlat (G) =
χ(G∨K1 )−1 = n−1, by Theorem 2.7(b), we conclude that χla ((A(mKn , K1 )−e)∨K1 ) = n+1.
(d) Let G = A(mKn , K1 ) − e, m, n ≥ 3, where e = v1,1 v1,2 is incident to the vertex u =
v1,1 = · · · = vm,1 . Consider G and the local antimagic total labeling f of G as in the proof
of Theorem 3.6(III). It is easy to show that the sum of all vertex labels under f is 12 m(n −
1)(n + 1)[m(n − 1) + 1] which is larger than all the induced vertex weights under f . Since
χlat (G) = χ(G ∨ K1 ) − 1 = n − 1, by Theorem 2.7(b), we conclude that χla ((A(mKn , K1 ) − e) ∨
K1 ) = n + 1.

Theorem 3.8: For m ≥ 2 and even n ≥ 2, χlat (mKn ) = χlat (mKn − e) = n. Otherwise,
n ≤ χlat (mKn ) ≤ m + n − 1.

Proof. We first note that for G = mKn , χlat (G), χlat (G − e) ≥ n, and that χla (G ∨ K1 ), χla ((G ∨
K1 ) − e) ≥ n + 1. For 1 ≤ i ≤ m, let V (mK2 ) = {ui , vi } and E(mK2 ) = {ui vi }. Define

17
f : V (mK2 ) ∪ E(mK2 ) → [1, 3m] such that f (ui vi ) = i, f (ui ) = 2n + 1 − i and f (vi ) = 3n + 1 − i.
Thus, f is a local antimagic total labeling that induces 2 distinct vertex weights.
Consider even n ≥ 4. Clearly, |V (mKn )| = mn and |E(mKn )| = mn(n + 1)/2. Let [t] =
{3(t − 1) + 1, 3(t − 1) + 2, . . . , 3(t − 1) + m}. Let M be a symmetric n × n matrix. Let Mi
be the labeling matrix of the i-th Kn . We define a total labeling of the i-th Kn by assigning
appropriate integers in [1, mn(n + 1)/2] as follows.

(i) For matrix M, assign [1] to [n(n+1)/2] beginning with the main diagonal entry horizontally
to the right and consecutively from row 1 to row n such that column k is the transpose of
row k.

(ii) Beginning with the (1,1)-entry, bold the entries alternately for every row and column.

(iii) For the i-th labeling matrix, Mi , the entry that corresponds to a bold [t] is the i-th integer
in [t] whereas the entry that corresponds to a non-bold [t] is the (m + 1 − i)-th integer in
[t].

(iv) For each row, observe that the sum of the first, second, . . . , m/2-th pair of entries are equal
respectively. Thus, M1 to Mm always have equal row sum for row 1 to row n respectively.

(v) Since the numbers are assigned consecutively from row 1 to row n, the row sums form an
increasing sequence.

Consequently, the labeling matrices represent a local antimagic total labeling of the mKn that
induces n distinct vertex weights. Thus, χlat (mKn ) ≤ n.
Observe that the largest row sum is R = mn + nj=2 (m(jn − 2j ) − m + 1) = m[n n+1
P


2 −
n+1

Pn 2 j−1

m+1


3 ]−n(m−1)/2, and the sum of all vertex labels is S = j=1 (m [(j −1)n− 2 ]+ 2 ) =
n+1 m+1
2
2m 3 +n 2 . Thus, S−R = m2 n3 /3−mn3 /2−mn2 /2+m2 n/6+mn−n/2+m4 /6−m2 /6 >

0. By Theorem 2.7(b), we have χla (mKn ∨ K1 ) ≤ n + 1.

We now consider mKn − e. If n = 2, the theorem clearly holds by Lemma 2.11. We now modify
the labeling of mKn by swapping the vertex label i and the incident edge label 2m + i for each
1 ≤ i ≤ m. Thus, all vertex labels remained unchanged except that the third vertex label of
each Kn is now reduced by 2m. Since the original second row sum is less than the original third
row sum by more than 2m, we now have a local antimagic total labeling of mKn that induces
n distinct weight. Since the edge joining the first and third vertex of the first Kn is labeled by
1, by Lemmas 2.12 and 2.13, we conclude that χlat (mKn − e) ≤ n.
Consider odd n ≥ 3. By modifying the labeling function for A(mKn , K1 ) in the proof of
Theorem 3.5, it is easy to get a local antimagic total labeling for mKn , m ≥ 2, n ≥ 3, that
induces m + n − 1 distinct weights. The theorem holds.

We note that mKn ∨ K1 ∼

= A(mKn+1 , K1 ) as in Theorem3.2.

Example 3.5: We take m = 3 and n = 6. The labeling matrices of all the cases with exact
χlat and χla are given below.

18
 M v1,1 v1,2 v1,3 v1,4 v1,5 v1,6 M1 v2,1 v2,2 v2,3 v2,4 v2,5 v2,6 sum
v1,1 [1] [2] [3] [4] [5] [6] v2,1 1 6 7 12 13 18 57
v1,2 [2] [7] [8] [9] [10] [11] v2,2 6 19 24 25 30 31 135
v1,3 [3] [8] [12] [13] [14] [15] v2,3 7 24 36 37 42 43 189
v1,4 [4] [9] [13] [16] [17] [18] v2,4 12 25 37 48 49 54 225
v1,5 [5] [10] [14] [17] [19] [20] v2,5 13 30 42 49 55 60 249
v1,6 [6] [11] [15] [18] [20] [21] v2,6 18 31 43 54 60 61 267

M2 v3,1 v3,2 v3,3 v3,4 v3,5 v3,6 sum M3 v4,1 v4,2 v4,3 v4,4 v4,5 v4,6 sum
v3,1 2 5 8 11 14 17 57 v4,1 3 4 9 10 15 16 57
v3,2 5 20 23 26 29 32 135 v4,2 4 21 22 27 28 33 135
v3,3 8 23 35 38 41 44 189 v4,3 9 22 36 37 42 43 189
v3,4 11 26 38 47 50 53 225 v4,4 10 27 37 48 49 54 225
v3,5 14 29 41 50 56 59 249 v4,5 15 28 42 49 57 58 249
v3,6 17 32 44 53 59 62 267 v4,6 16 33 43 54 48 63 267

The above matrices give χlat (3K6 ) = 6. Let v be the vertex of K1 . If the main diagonal labels
are the edge labels of 3K6 ∨ K1 , then v has label 666. Thus, the matrices give χla (3K6 ∨ K1 ) = 7.
Deleting edge vv1,1 of label 1 of 3K3 ∨ K1 and reducing all edge labels by 1, we get a local an-
timagic labeling of (3K6 ∨ K1 ) − vv1,1 . Thus, by symmetry, χla ((3K6 ∨ K1 ) − e) = 7 for e not
belong to any K6 . If we swap the labels of 1 and 7, 2 and 8, and 3 and 9, delete the edge v1,1 v1,3
that has label 1 and reduce all other labels by 1, we get χlat (3K6 − e) = 6. Now, if the main
diagonal labels are the edge labels of 3K6 ∨ K1 , then we have χla (3K6 ∨ K1 ) − e) = 7 for e that
belongs to any K6 .

Let v be the central vertex of the wheel Wn = Cn ∨ K1 of order n + 1 ≥ 4. Denote by A(mWn , v)

the amalgamation of m ≥ 1 copies of Wn at vertex v where A(Wn , v) = Wn . In [10], the authors
claimed that χlsat (A(mWn , v)) = 3 (or 4) for even (or odd) n. However, the labeling function
defined in the proof contains errors that can be corrected as follow.

Theorem 3.9: For m ≥ 1, n ≥ 3,

(
3 if m is even,
χlat (A(mWn , v)) =
4 if m is odd.

Proof. Let G = A(mWn , v), V (G) = {v, vi,j | 1 ≤ i ≤ m, 1 ≤ j ≤ n} and E(G) = {vvi,j | 1 ≤ i ≤
m, 1 ≤ j ≤ n} ∪ {vi,1 vi,n , vi,j vi,j+1 | 1 ≤ i ≤ m, 1 ≤ j ≤ n − 1}. Suppose n is odd. Define a total
labeling f : V (G) ∪ E(G) → [1, 3mn + 1] as follows.

(i) f (v) = 1 and f (vi,1 ) = m(n − 1) + i + 1,

(ii) f (vi,j ) = m(j − 2) + i + 1 for 1 ≤ i ≤ m, 2 ≤ j ≤ n,

(iii) f (vvi,j ) = mn + 1 + m(j − 3) + i for 1 ≤ i ≤ m and odd j = 3, 5, . . . , n − 1,

(iv) f (vvi,j ) = m(n − 1) + 1 + mj + i for 1 ≤ i ≤ m and even j = 2, 4, . . . , n − 2,

(v) f (vvi,1 ) = m(2n − 1) + 1 + i for 1 ≤ i ≤ m,

(vi) f (vi,1 vi,n ) = 2mn + m + 2 − i for 1 ≤ i ≤ m,

(vii) f (vi,j vi,j+1 ) = 3mn + m(1 − j) − i + 2 for 1 ≤ i ≤ m, 1 ≤ j ≤ n − 1.

19
It is easy to check that w(v) = 1 + 3 mn+1


2 , w(vi,1 ) = 8mn − m + 6, w(vi,j ) = 7mn − 2m + 6 for
odd j > 1, and w(vi,j ) = 7mn + 6 for even j. Thus, f is a local (super) antimagic total labeling
of G. Since 4 = χ(G) ≤ χlat (G) ≤ 4, the theorem holds.
Suppose n is even. Define a total labeling f : V (G) ∪ E(G) → [1, 3mn + 1] such that the labels
are as for odd n for all the vertices and the edges vi,1 vi,n . Moreover,

(i) f (vvi,j ) = 2mn − mj + i + 1 for 1 ≤ i ≤ m, 1 ≤ j ≤ n,

(ii) f (vi,j vi,j+1 ) = 2mn + m(j + 1) + 2 − i for 1 ≤ i ≤ m, 1 ≤ j ≤ n − 1.

It is easy to check that w(v) = 1 + 3 mn+1

2 , w(vi,j ) = 7mn + m + 6 for odd j, and w(vi,j ) =
7mn − m + 6 for even j. Thus, f is a local (super) antimagic total labeling of G. Since
3 = χ(G) ≤ χlat (G) ≤ 3, the theorem holds.

Note that Wn − e is the fan graph Fn if e is an edge not incident to the vertex v.

Corollary 3.10: For n ≥ 3, χlat (Fn ) = 3 if n is even, and 3 ≤ χlat (Fn ) ≤ 4, otherwise.

Note that the above theorem also holds for χlsat (A(mKn , v)). By Theorem 2.7, Lemma 2.12 or
Lemma 2.13, we also have

Corollary 3.11: For n ≥ 3 and e an edge not incident to vertex v,

(
3 if n is even,
χlat (A(mWn , v) − e) =
4 if n is odd.

Corollary 3.12: For n ≥ 3 and e an edge of A(mWn , v),

(
4 if n is even,
χla (A(mWn , v) ∨ K1 ) = χla ((A(mWn , v) ∨ K1 ) − e) =
5 if n is odd.

We now determine χlat (Km ⊙ Kn ) for m ≥ 1, n ≥ 2.

Theorem 3.13: For n ≥ 2, χlat (K1 ⊙Kn ) = n+1 and max{m, n+1} ≤ χlat (Km ⊙Kn ) ≤ m+n
if m ≥ 2.

Proof. Let G = Km ⊙Kn , m, n ≥ 2. Note that we can view G as a graph obtained from m copies
of Kn+1 by taking a vertex of each Kn+1 , and join these vertices by edges pairwise to form a
Km . Thus, we may have V (G) = {vi,j | 1 ≤ i ≤ m, 1 ≤ j ≤ n + 1} and E(G) = {vi,j vi,k | 1 ≤ j <
k ≤ n + 1} ∪ {vi,n+1 vj,n+1 | 1 ≤ i < j ≤ m}. Note that vertices vi,n+1 , 1 ≤ i ≤ m, form the Km
subgraph with |V (G)| = m(n + 1) and |E(G)| = m n+1 m



2 + 2 .
n+1
m

Define f : V (G) ∪ E(G) → [1, m(n + 1) + m 2 + 2 ] as follows.

(i) f (vi,j vi,k ) = m[ k−1

2 + j − 1] + i for 1 ≤ i ≤ m, 1 ≤ j < k ≤ n + 1;

(ii) f (vi,j ) = m n+1

2 + (m − i + 1)n − j + 1 for 1 ≤ j ≤ n;

(iii) f (vi,n+1 ) = m n+1

2 + mn + i for 1 ≤ i ≤ m;

(iv) f (vi,n+1 vk,n+1 ) = m n+1 + m(n + 1) + k−1

2 2 + i for 1 ≤ i < k ≤ m.

20
Observe that for 1 ≤ i ≤ m,
i−1
X m
X n
X
w(vi,n+1 ) = f (vi,n+1 ) + f (vi,n+1 vk,n+1 ) + f (vi,n+1 vk,n+1 ) + f (vi,j vi,n+1 )
k=1 k=i+1 j=1
i−1
      !
n+1 X n+1 i−1
= m + mn + i + m + m(n + 1) + +k +
2 2 2
k=1
m n
    ! "   #
X n+1 k−1 X n
m + m(n + 1) + +i + m + m(j − 1) + i
2 2 2
k=i+1 j=1
     !  
n+1 n+1 i−1 i
= m + mn + i + (i − 1) m + m(n + 1) + + +
2 2 2 2
  !      
n+1 m n n
(m − i) m + m(n + 1) + i + + mn +m + ni
2 3 2 2
    !  
n+1 n+1 i−1
= m + n(m + i) + i + (m − 1) m + m(n + 1) + (i − 1) +
2 2 2
     
i m n
+ (m − i)i + + m(n + 1)
2 3 2
    !    
n+1 n+1 m n
= m + n(m + i) + (m − 1) m + m(n + 1) + mi + + m(n + 1) +
2 2 3 2
i3 /2 − 5i2 /2 + 3i − 1 (3.5)

which is an increasing function.

Moreover, for 1 ≤ i ≤ m, 1 ≤ j ≤ n,
j−1
X n+1
X
w(vi,j ) = f (vi,j ) + f (vi,j vi,k ) + f (vi,j vi,k )
k=1 k=j+1
j−1
  ( "  # )
n+1 X j−1
= m + (m − i + 1)n − j + 1 + m +k−1 +i +
2 2
k=1
n+1
( "  # )
X k−1
m +j−1 +i
2
k=j+1
       
n+1 j−1 n+1 j
= m + (m + 1)n − j + 1 + mj +m −m +
2 2 3 3
m(n − j + 1)(j − 1)
   
n+1 n+1
= m + (m + 1)n − j + 1 + m + mn(j − 1) +
2 3
" #
m j 3 /3 − 2j 2 + 8j/3 − 1 (3.6)

also an increasing function.

21
Thus, we have w(vi,1 ) < w(vi,2 ) < · · · < w(vi,n ) and w(v1,n+1 ) < w(v2,n+1 ) < · · · < w(vm,n+1 ).
It is routine to check that
mn m
w(v1,n+1 ) − w(vi,n ) = (mn + n + 3m − 7) + n + (m2 + 3m + 8) − 1 > 0.
2 6
Therefore, f is a local antimagic total labeling that induces m + n distinct weights so that
χlat (Km ⊙ Kn ) ≤ m + n. Since χ(Km ⊙ Kn ) = max{m, n + 1}, we have max{m, n + 1} ≤
χlat (Km ⊙ Kn ) ≤ m + n. When m = 1, we have max{m, n + 1} = n + 1. The theorem holds.

In what follows, we consider some sparse graphs. Let Pn = v1 v2 · · · vn be the path of order
n ≥ 2.

Theorem 3.14: For n ≥ 2, χlat (Pn ) = 2 except that χlat (P4 ) = 3.

Proof. Obviously χlat (P2 ) = 2. By [7, Theorem 3.5], χla (P4 ∨ K1 ) = 4 implies that χlat (P4 ) 6=
2. Moreover, every local antimagic labeling of P4 ∨ K1 induces 4 distinct vertex colors also
corresponds to a local antimagic total labeling of P4 that induces 3 distinct vertex weights. One
such labeling that labeled the vertices and edges of P4 alternately is given by sequence 7, 2, 6,
3, 5, 1, 4.
Assume n ≥ 6 is even. Observe that by Theorem 2.7 and the local antimagic labeling of Wn
obtained in [1,5], we can get a local antimagic total labeling of Cn with an edge labeled 1. Since
Cn is regular, we can delete this edge and reduce all other labels by 1. Consequently, we get
a path Pn that admits a local antimagic total labeling that induces exactly 2 distinct vertex
weights. Since χlat (Pn ) ≥ χ(Pn ) ≥ 2, we have χlat (Pn ) = 2. The details are omitted. We note
that similar conclusion can also be obtained from the labeling of Wn , n even in Theorem 3.9 for
m = 1 after relabeling the vertices and edges of Wn in reverse order.
Consider odd n ≥ 5. Suppose n = 4k + 1. For n = 5, a required labeling sequence that labeled
the vertices and edges of P5 alternately is 6, 4, 7, 3, 2, 5, 8, 1, 9 with distinct vertex weights
10 and 14. By computer search, we are able to obtain all the 12 different labelings. For n ≥ 9,
define f : V (Pn ) ∪ E(Pn ) → [1, 8k + 1] as follows.

(i) f (v1 ) = 8k; f (v4k−1 ) = 6k; f (v4k+1 ) = 8k + 1;

(ii) f (v4i+1 ) = 7k − 3 + i for i ∈ [1, k − 1];

(iii) f (v4i−1 ) = 3k + i for i ∈ [1, k − 1];

(iv) f (v4i+2 ) = 7k + i for i ∈ [0, k − 1];

(v) f (v4i+4 ) = 4k + 1 + i for i ∈ [0, k − 1];

(vi) f (v4k v4k+1 ) = 2k; f (v4k−1 v4k ) = 4k;

(vii) f (v2i v2i+1 ) = 2k − i for i ∈ [1, 2k − 1];

(viii) f (v4i+1 v4i+2 ) = 2k + 1 + i for i ∈ [0, k − 1];

(ix) f (v4i+3 v4i+4 ) = 5k + 1 + i for i ∈ [0, k − 2].

22
It is not difficult to check that
(
10k + 1 for odd i,
w(vi ) =
11k for even i.

Thus, χlat (P4k+1 ) = 2.

Suppose n = 4k + 3 ≥ 3. A required labeling sequence for n = 3, 7, 11 are 1, 5, 3, 4, 2; 5, 13,
1, 9, 7, 2, 10, 11, 4, 3, 8, 12, 6; and 8, 21, 1, 14, 11, 4, 15, 17, 10, 2, 16, 18, 6, 5, 12, 19, 7, 3,
12, 20, 9. The corresponding distinct vertex weights are 6 and 12; 18 and 23; and 29 and 36
respectively.
For k ≥ 3, we define f : V (Pn ) ∪ E(Pn ) → [1, 8k + 5] as follows.

(i) f (v1 ) = 3k + 2; f (v2 ) = 1; f (v3 ) = 4k + 3; f (v2k+3 ) = 2k + 2; f (v4k+3 ) = 3k + 3;

(ii) f (v2i+3 ) = 3k + 3 + i for i ∈ [1, k − 1];

(iii) f (v2k+2i+1 ) = 2k + 1 + i for i ∈ [1, k];

(iv) f (v2i+2 ) = 5k + 4 + i for i ∈ [1, k];

(v) f (v2k+2i+2 ) = 4k + 3 + i for i ∈ [1, k];

(vi) f (v1 v2 ) = 8k + 5; f (v2 v3 ) = 5k + 4;

(vii) f (v2i+1 v2i+2 ) = 2k + 2 − 2i for i ∈ [1, k];

(viii) f (v2k+2i+1 v2k+2i+2 ) = 2k + 3 − 2i for i ∈ [1, k];

(ix) f (v2i+2 v2i+3 ) = 6k + 4 + i for i ∈ [1, 2k].

It is not difficult to check that

(
11k + 7 for odd i,
w(vi ) =
13k + 10 for even i.

Thus, χlat (P4k+3 ) = 2.

Example 3.6: The labeling sequence for P12 is 16, 10, 18, 3, 12, 11, 19, 1, 20, 5, 17, 9, 15, 2,
22, 7, 13, 6, 21, 4, 14, 8, 23 with 2 distinct vertex weights 26 and 31. The labeling sequence for
P16 is 22, 13, 24, 5, 16, 14, 25, 3, 17, 15, 26, 1, 27, 7, 23, 12, 21, 2, 30, 10, 19, 6, 28, 8, 18, 9, 29,
4, 20, 11, 31 with 2 distinct vertex weights 35 and 42.
The labeling sequence for P14 is 24, 13, 16, 1, 27, 9, 19, 2, 23, 12, 15, 3, 26, 8, 18, 4, 22, 11, 14,
5, 25, 7, 17, 6, 21, 10, 20 with 2 distinct vertex weights 37 and 30.
The labeling sequence for P13 is 24, 7, 21, 5, 10, 16, 13, 4, 19, 8, 22, 3, 11, 17, 14, 2, 20, 9, 23,
1, 18, 12, 15, 6, 25 with 2 distinct vertex weights 31 and 33.
The labeling sequence for P15 is 11, 29, 1, 19, 15, 6, 20, 23, 13, 4, 21, 24, 14, 2, 22, 25, 8, 7, 16,
26, 9, 5, 17, 27, 10, 3, 18, 28, 12 with 2 distinct vertex weights 40 and 49.

23
In [7, Theorems 3.6 and 3.7], the authors showed that χla (Fn ) = 3 for even n ≥ 4, and 3 ≤
χ (Fn ) ≤ 4 for odd n ≥ 3. From the proof of Theorem 3.14, when n = 4kP+ 1, we have
Pla4k+1 2 4k+3
i=1 f (ui ) = 22k + 12k + 1 6= 11k 6= 10k + 1. For n = 4k + 3, we have i=1 f (ui ) =
16k2 + 19k + 6 6= 11k + 7 6= 13k + 10. By Theorems 3.14 and 2.7, we have the following.

Corollary 3.15: For n ≥ 2, χla (Fn ) = 3.

Theorem 3.16: For n ≥ 3,

(
2 if n is even,
χlat (Cn ) =
3 otherwise.
Proof. It is obvious that χlat (C3 ) = 3. Assume n ≥ 4. In [1, 5], the authors showed that
(
3 if n is even,
χla (Wn ) =
4 otherwise.
Since (
2 if n is even,
χ(Cn ) =
3 otherwise,
by Theorem 2.7, we conclude that the theorem holds.

Theorem 3.17: For odd n ≥ 3, 4 ≤ χlat (Cn ∨ 2K1 ) ≤ 5, and for even n ≥ 6, 3 ≤ χlat (Cn ∨
3K1 ) ≤ 5.

Proof. Suppose n ≥ 3 is odd. Clearly, χlat (Cn ∨ 2K1 ) ≥ χ(Cn ∨ 2K1 ) = 4. In [7, Theorem
3.1], the authors proved that for odd n ≥ 3, χla (Cn ∨ 3K1 ) = 4. Moreover, the corresponding
local antimagic labeling g induces g+ (u1 ) = g+ (u2 ) = g+ (u3 ) = 5n(n + 1)/2, g+ (v1 ) = 8n + 3,
g+ (vi ) = (17n + 7)/2 for odd i ≥ 3, and g + (vi ) = (17n + 5)/2 for even i ≥ 2.
Define f : V (Cn ∨ 2K1 ) ∪ E(Cn ∨ 2K1 ) → [1, 4n + 2] such that f (vi ) = g(vi u3 ), and f (e) = g(e)
for e ∈ E(Cn ) or e = vi uj , j = 1, 2. Moreover, f (uj ) = 4n + j for j = 1, 2. Now, w(vi ) = g+ (vi )
and w(uj ) = g+ (uj ) + 4n + i for i = 1, 2. Thus, f induces 5 distinct vertex weights and
χlat (Cn ∨ 2K1 ) ≤ 5.
Suppose n ≥ 6 is even. Clearly, χlat (Cn ∨ 3K1 ) ≥ 3. In [7, Theorem 3.3], the authors proved that
χla (Cn ∨ 4K1 ) = 3. Moreover, the corresponding local antimagic labeling g induces g+ (vi ) =
9n + 3 for odd i, g + (vi ) = 17n + 3 for even i, and g + (uj ) = n(6n + 1)/2 for 1 ≤ j ≤ 4.
Define f : V (Cn ∨3K1 )∪E(Cn ∨3K1 ) → [1, 5n+3] such that f (vi ) = g(vi u4 ), and f (e) = g(e) for
e ∈ E(Cn ) or e = vi uj , j = 1, 2, 3. Moreover, f (uj ) = 5n + j for j = 1, 2, 3. Now w(vi ) = g+ (vi )
and w(uj ) = g+ (uj ) + 5n + i for i = 1, 2, 3. Thus, f induces 5 distinct vertex weights and
χlat (Cn ∨ 3K1 ) ≤ 5.

Problem 3.1: Determine χlat (Cn ∨ 2K1 ) for odd n ≥ 3, and χlat (Cn ∨ 3K1 ) for even n ≥ 4.

In [7, Theorem 3.9], the authors proved that for n, m ≥ 3,

(
m + 2 if m, n are even;
χla (Km ∨ Cn ) =
m + 3 if m, n are odd.
By Theorem 2.7, the following theorem holds.

24
Theorem 3.18: For m, n ≥ 3,
(
m+1 if m, n are even;
χlat (Km−1 ∨ Cn ) =
m+2 if m, n are odd.

In [10, Theorem 4], the authors also proved that a family of cubic bipartite graph denoted CB2k
obtained from C2k , k ≥ 3 has χlsat (CB2k ) = 2. Thus, by Lemma 2.11, we have

Corollary 3.19: For k ≥ 3, χlat (CB2k ) = χlat (CB2k − e) = 2.

It is easy to verify that the given labeling has sum of all the vertex labels not equal the induced
vertex weights. By Theorem 2.7 (b), we have

Corollary 3.20: For k ≥ 3, χla (CB2k ∨ K1 ) = 3.

By Lemma 2.4 in [7], it is easy to check that the following corollary holds.

Corollary 3.21: For k ≥ 3, χla ((CB2k ∨ K1 ) − e) = 3 where e is not an edge that belongs to
the induced C2k subgraph.

Note that when k is odd, CB2k is commonly known as Möbius ladder M2k or the circulant graph
C2k (1, k).

4 Conclusion and Open Problems

In this paper, we first prove that every graph is local antimagic. The proof gives a good
bound for us to determine χlat (G) (or χla (G ∨ K1 )) using a local antimagic labeling of G ∨ K1
(or a local antimagic total labeling of G). The local antimagic (total) chromatic number of
many family of graphs are determined. Particularly, we showed that there are graphs G with
χ(G) = χ(G−e) = χlat (G) = χlat (G−e) = χ(G∨K1 )−1 = χla (G∨K1 )−1 = χla ((G−e)∨K1 )−1
or χ(G) = χlat (G) = χla (G) − 1. The following problems arise naturally.

Problem 4.1: Characterize G such that χ(G) = χ(G − e) = χlat (G) = χlat (G − e) = χ(G ∨
K1 ) − 1 = χla (G ∨ K1 ) − 1 = χla ((G − e) ∨ K1 ) − 1.

Problem 4.2: Characterize G such that χ(G) = χlat (G) = χla (G) − 1.

Problem 4.3: Determine the exact values of χlat (A(mKn , Kr ) and χlat (A(mKn , Kr ) − e) for
m, n, r not satisfying Theorems 3.1 and 3.3.

Problem 4.4: For n ≥ 3, prove that if e is an edge adjacent to the vertex u of A(2Kn , K1 ),
then χlat (A(2Kn , K1 ) − e) = n.

Problem 4.5: Determine the exact values of χlat (mKn ) and χlat (mKn − e) for m ≥ 2 and
odd n ≥ 3.

Problem 4.6: Determine the exact values of χlat (Km ⊙ Kn ) and χla ((Km ⊙ Kn ) ∨ K1 ) for
m, n ≥ 2.

Problem 4.7: Determine the exact values of χlat (Fn ) and χla (Fn ∨ K1 ) for odd n ≥ 3.

25
Problem 4.8: Determine the exact values of χlat (Cn ∨ 2K1 ) and χla ((Cn ∨ 2K1 ) ∨ K1 ) for odd
n ≥ 3, and χlat (Cn ∨ 3K1 ) and χla ((Cn ∨ 3K1 ) ∨ K1 ) for even n ≥ 6.

In [4, Theorem 3.4], the authors showed that there are infinitely many circulant graphs (with at
most an edge deleted) of χla = 3. Since cycles are the simplest circulant graphs with χlat = 2,
we have

Problem 4.9: Determine the exact values of χlat (C) and χlat (C − e) for each circulant graph
C∼6 Cn , C2n (1, n), n ≥ 3.
=

In [6], the authors proved that any graph G of k ≥ 1 pendant(s) has χla (G) ≥ k+1. In [1,4,5,7–9],
the authors obtained many families of graph G having k ≥ 1 pendants with χla (G) = k + i, i =
1, 2. It is also obvious that χla (C3 ⊙O2 ) = 9. Besides χla (Kn ) = n, n ≥ 3, we are not aware of any
graph G ∼ 6 Kn containing no pendant vertices but having arbitrarily large χla (G) = χ(G). Our
=
results above show the existence of infinitely many such graphs G. This gives partial solution
to [1, Problem 3.2]: Characterize the class of graphs G for which χla (G) = χ(G).
Since every known result has χlat (G) ≤ χla (G), we end this paper with the following.

Conjecture 4.1: For each graph G of order at least 3, χlat (G) ≤ χla (G).

References
[1] S. Arumugam, K. Premalatha, M. Bacǎ and A. Semaničová-Feňovčı́ková, Local antimagic
vertex coloring of a graph, Graphs and Combin., 33 (2017), 275-285.
[2] J. Bensmail, M. Senhaji and K. Szabo Lyngsie, On a combination of the 1-2-3 Conjec-
ture and the Antimagic Labelling Conjecture, Discrete Math. Theoret. Comput. Sc., 19(1)
(2017) #22.
[3] J. Haslegrave, Proof of a local antimagic conjecture, Discrete Math. Theor. Comp. Sc.,
20(1) (2018), #18.
[4] G.C. Lau, J. Li, H.K. Ng and W.C. Shiu, Approaches which output infinitely many
graphs with small local antimagic chromatic number, (2020), submitted to Discrete Math,
arXiv:2009.01996.
[5] G.C. Lau, H.K. Ng, and W.C. Shiu, Affirmative solutions on local antimagic chromatic
number, Graphs and Combin., 36 (2020), 1337-1354.
[6] G.C. Lau, W.C. Shiu and H.K. Ng, On local antimagic chromatic number of graphs with
cut-vertices, (2020) submitted to Iran. J. Math. Sci. Inform., arXiv:1805.04801.
[7] G.C. Lau, W.C. Shiu and H.K. Ng, On local antimagic chromatic number of cycle-related
join graphs, Discuss. Math. Graph Theory 41 (2021) 133152 DOI : 10.7151/dmgt.2177.
[8] G.C. Lau, W.C. Shiu and C.X. Soo, On Local Antimagic Chromatic Number of Spider
Graphs, (2020) submitted to J. Discret. Math. Sci. Cryptogr., arXiv:2008.09754.
[9] K. Premalatha, S. Arumugam, Yi-Chun Lee and Tao-Ming Wang, Local antimagic
chromatic number of trees - I, J. Discret. Math. Sci. Cryptogr., (2020) DOI :
10.1080/09720529.2020.1772985.

26
[10] Slamin, N. O. Adiwijaya, M. A. Hasan, Dafik, and K. Wijaya, Local Super Antimagic Total
Labeling for Vertex Coloring of Graphs, Symmetry, 12(11) (2020) 10.3390/sym12111843.

[11] D. Zuckerman, Linear degree extractors and the inapproximability of Max Clique and Chro-
matic Number, Theory of Computing, 3 (2007) 103-128, doi:10.4086/toc.2007.v003a006

27