Remark 2.1
KV23If (X, d) is an F-symmetric space, then any convergent sequence is a Cauchy sequence. For N-symmetric spaces and for quasimetric spaces this assertion is not more true.
We introduce and study a general concept of multiple fixed point for mappings defined on partially ordered distance spaces in the presence of a contraction type condition and appropriate monotonicity properties. This notion and the obtained results complement the corresponding ones from [Choban, M., Berinde, V., A general concept of multiple fixed point for mappings defined on spaces with a distance (submitted)] and also simplifies some concepts of multiple fixed point considered by various authors in the last decade or so.
If (X, d) is an F-symmetric space, then any convergent sequence is a Cauchy sequence. For N-symmetric spaces and for quasimetric spaces this assertion is not more true.
A distance space (X, d) is an H-distance space if and only if any convergent sequence in X has a unique limit point.
([ChoB]) Let (X, d) be a distance space. Then:
1. If d is a symmetric, then (X^m, d^m) and (X^m, \bar{d}^m) are symmetric spaces, too.
2. If d is a quasimetric, then (X^m, d^m) and (X^m, \bar{d^m}) are quasimetric spaces, too.
3. If d is a metric, then (X^m, d^m) and (X^m, \bar{d}^m) are metric spaces, too.
4. If d is an F-distance space, then (X^m, d^m) and (X^m, \bar{d}^m) are F-distance spaces, too.
5. If d is an N-distance space, then (X^m, d^m) and (X^m, \bar{d}^m) are N-distance spaces, too.
6. If d is an H-distance space, then (X^m, d^m) and (X^m, \bar{d}^m) are H-distance spaces, too.
7. If (X, d) is a C-distance space, then (X^m, d^m) and (X^m, \bar{d}^m) are C-distance spaces, too.
8. If (X, d) is a complete distance space, then (X^m, d^m) and (X^m, \bar{d}^m) are complete distance spaces, too.
9. If d is an s-distance space, then (X^m, d^m) and (X^m, \bar{d}^m) are s-distance spaces, too.
10. The spaces (X^m, d^m) and (X^m, \bar{d}^m) share the same convergent sequences and the same Cauchy sequences. Moreover, the distances d^m and \bar{d^m} are uniformly equivalent, i.e., for each \varepsilon > 0, there exists \delta = \delta (\varepsilon ) > 0 such that:
- from d^m(x,y) \leq \delta it follows \bar{d^m}(x,y) \leq \varepsilon;
- from \bar{d^m}(x,y) \leq \delta it follows d^m(x,y) \leq \varepsilon.
Let (X, d, \preceq ) be a monotonically complete distance space. Then (X^m, d^m, \preceq _L) and (X^m, \bar{d^m}, \preceq _L) are ordered monotonically complete distance spaces, too.
It is obvious.
Let a \in X^m be a multidimensional fixed point of the operator of F: X^m \longrightarrow X. If any of the Conditions \Omega _i, i \in\{1, 2, 3, 4\}, is satisfied, then the operator F has a unique multidimensional fixed point.
Obviously, (X^m, \preceq _L) is a lattice, too. Let \rho = d^m, for i \in \{1, 2\}, and \rho = \bar{d}^m, for i \in \{3, 4\}. Then for x, y, z\in X^m and x \preceq _L y \preceq _L z we have \rho (x,y)+ \rho (y,x)\leq \rho (x,z) + \rho (z,x). Assume that b\in X^m is a multidimensional fixed point of the operator F with b \not= a.
Case 1. The points a and b are comparable.
Assume that a \preceq _L b. Then \rho (a,b) + \rho (b,a) = \rho (\lambda F(a),\lambda F(b)) + \rho (\lambda F(b),\lambda F(a)) < \rho (a,b) + \rho (b,a), a contradiction.
Case 2. The points a and b are not comparable.
Fix c = \max \{a, b\} \in X_n. We put d = \lambda F(\lambda F(c)). By construction, a \preceq _L d and b \preceq _L d. Hence, c \preceq _L d and \rho (a,c) \leq \rho (a,d), \rho (b,c) \leq \rho (b,d). Therefore \rho (a,c) + \rho (c,a) \leq \rho (a,d) + \rho (d(a). By virtue of the conditions \Omega _i, we then have \rho (a,d) + \rho (d,a) < \rho (a,c) + \rho (c,a), a contradiction.
Let (X, d) be an H-distance space and let a \in X^m be a multidimensional fixed point of the operator of F: X^m \rightarrow X. Then in any of the Conditions MK_i, i\in \{1, 2\}, the operator F has a unique multidimensional fixed point.
Obviously, in (X^m, \preceq _L), for any two points x, y \in X^m, there exist an upper bound and a lower bound. Let \rho = d^m. In this case for any two points x, y \in X^m, from the condition \rho (x,y) < r + \delta (r) and x \preceq y, it follows that \rho (x,y) < r.
Assume that b\in X^m is a multidimensional fixed point of the operator of F and that b \not= a.
Case 1. The points a and b are comparable.
Assume that a \preceq L b and \rho (a,b) = r > 0. Since \rho (a, b) < r + \delta (r), we have r = \rho (a,b) = \rho (\lambda F(a),\lambda F(b)) < r, a contradiction.
Case 2. The points a and b are not comparable. We put r = \inf\{max\{\rho (a,c), \rho (b,c)\}: c \in X^m, a \preceq _L c, b \preceq _L c\}. We claim that r = 0. Assume that r > 0. Then \delta (r) > 0 and there exists c such that \max\{\rho (a,c), \rho (b,c)\} < r + \delta (r), a \preceq _L c, b \preceq _L c. We put e = \lambda F(\lambda F(c)). Then a \preceq _L e and b \preceq _L e. Since \lambda F(\lambda F(a)) = a and \lambda F(\lambda F(b)) = b, we have \max\{\rho (a,e), \rho (b,e)\} < r, a contradiction. Thus r = 0. For each n \in \mathbb N there exists a point c_n \in X^n such that a \preceq _L c_n, b \preceq _L c_n and max\{\rho (a,c_n), \rho (b,c_n)\} < 2^{-n}. We can construct a sequence \{c_n: n \in \mathbb N\} for which a = \lim_{n \rightarrow \infty } c_n and b = \lim_{n \rightarrow \infty } c_n, a contradiction.
Let (X, d, \preceq ) be an ordered metric space, m \in \mathbb N, F: X^m \rightarrow X be an operator. Suppose that:
a) there exists a function \delta : (0, +\infty ) \longrightarrow (0,+\infty ) such that from r > 0, x, y \in X^m, d^m(x,y) < r + \delta (r) and x \preceq y it follows that d^m(\lambda F(x),\lambda F(y)) < r;
b) for any two points x, y \in X there exists an upper bound and a lower bound.
Suppose also that one of the following sets of conditions is satisfied:
1. (X, d, \preceq ) is monotonically complete; from x, y \in X^m and x \preceq _L y it follows that \lambda F(x) \preceq _L \lambda F(y); there exists a \in X^m such that a \preceq _L \lambda F(a).
2. (X, d, \preceq ) is complete; from x, y \in X^m and x \preceq _L y it follows that \lambda F(y) \preceq L \lambda F(x); there exists a \in X^m such that a \preceq _L \lambda F(a) or F(a) \preceq _L a.
Then there exists a unique multidimensional fixed point of the operator of F.
Let (X, d, \preceq ) be an ordered metric space, m \in \mathbb N, F: X^m \rightarrow X be an operator. Suppose that:
a) there exists a function \delta : (0, +\infty ) \longrightarrow (0,+\infty ) such that from r > 0, x, y \in X^m, \bar{d}^m(x,y) < r + \delta (r) and x \preceq y it follows that \bar{d}^m(\lambda F(x),\lambda F(y)) < r;
b) for any two points x, y \in X, there exist an upper bound and a lower bound; for any i \in \{1,2,\dots,m\}, the mapping \lambda _i is a surjection or, more generally, |\cup \{\lambda _i^{-1}(j): 1\leq j \leq m\}| = m, for each i \in \{1,2,\dots,m\}.
Suppose also that one of the following sets of conditions is satisfied:
1. (X, d, \preceq ) is monotonically complete; from x, y \in X^m and x \preceq _L y it follows that \lambda F(x) \preceq _L \lambda F(y); there exists a \in X^m such that a \preceq _L \lambda F(a).
2. (X, d, \preceq ) is complete; from x, y \in X^m and x \preceq _L y it follows that \lambda F(y) \preceq L \lambda F(x); there exists a \in X^m such that a \preceq _L \lambda F(a) or F(a) \preceq _L a.
Then there exists a unique multidimensional fixed point of the operator of F.
The text on this page is from arXiv:1701.00518v1 by Mitrofan M. Choban and Vasile Berinde, under the licence CC0 1.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.
The record of each item is in analysis/objects/1701/1701.00518.jsonl.