The American Mathematical Monthly

ISSN: 0002-9890 (Print) 1930-0972 (Online) Journal homepage: http://www.tandfonline.com/loi/uamm20

The Singular Part as Fixed Point

Tamás Titkos

To cite this article: Tamás Titkos (2018) The Singular Part as Fixed Point, The American
Mathematical Monthly, 125:1, 77-80, DOI: 10.1080/00029890.2018.1393219

To link to this article: https://doi.org/10.1080/00029890.2018.1393219

Published online: 21 Dec 2017.

Submit your article to this journal

Article views: 26

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=uamm20

Download by: [Purdue University Libraries] Date: 06 January 2018, At: 08:36
 The Singular Part as Fixed Point
Tamás Titkos

Abstract. The aim of this note is to investigate the Lebesgue decomposition theorem of non-
negative finite measures from a new point of view. Using a standard iteration scheme, we
identify the singular part as a fixed point of a nonnegative finite measure-valued map.

1. INTRODUCTION. In his recent paper [4], Arlinskiı̆ presented a new approach

to establish the Lebesgue-type decomposition of bounded nonnegative operators (see
also [2]). The aim of this short note is to translate his approach for set functions. To
that end, we offer first some informal remarks about the connection between bounded
Downloaded by [Purdue University Libraries] at 08:36 06 January 2018

nonnegative operators and nonnegative finite measures that may be interesting for the
reader.
As is known, the collection of all signed measures on a fixed σ -algebra is a Banach
lattice with the total variation norm. We will restrict our attention to the cone of non-
negative finite measures, which is a lattice as well (for details, see [1, Section 36]). The
key observation is the following: Although the set of bounded nonnegative operators is
not a lattice (see [3]), there is a natural two variable operation (A, B) → A : B that be-
haves similarly to the operation (μ, ν ) → μ ∧ ν for nonnegative finite measures. The
operator A : B is called the parallel sum of A and B (see [8]), and μ ∧ ν denotes the
greatest lower bound of μ and ν. But what does “behaves similarly to” mean?
It is also known that vector lattice arguments can be used effectively to investigate
absolute continuity of finitely additive set functions (see for example [6, Section II]).
Despite the structural differences mentioned above, the following equivalences allow
us to characterize absolute continuity (and construct the absolutely continuous part in
the Lebesgue decomposition) of measures and operators in the same way:
μ  ν ⇐⇒ μ = sup μ ∧ nν A  B ⇐⇒ A = sup A : nB.
n∈N n∈N

For such a proof of the Lebesgue decomposition theorem in the measure case, see [1,
Theorem 37.7]. The elementary approach we want to offer in this note is based on
another interesting analogy. Namely, two nonnegative operators A and B are singular if
and only if A : B is the identically zero operator. And, as we shall see, two nonnegative
finite measures μ and ν are singular if and only if μ ∧ ν is the zero measure. Using
these facts, we can imitate the proof of Arlinskiı̆: we identify the singular part as a
fixed point of a nonnegative finite measure-valued map. For other significantly different
approaches, see [7, 9, 10, 12–16].

2. NOTIONS AND NOTATIONS. Let X be a nonempty set and let A be a σ -algebra

on it. A σ -additive set function μ : A → R is called a nonnegative finite measure if
0 ≤ μ(A) ≤ μ(X ) < +∞ holds for all A ∈ A. The identically zero measure will be
denoted by θ. We write μ ≤ ν if μ(A) ≤ ν(A) holds for all A ∈ A. Now, define absolute
continuity and singularity as follows: μ is said to be ν-absolutely continuous (μ  ν, in

doi.org/10.1080/00029890.2018.1393219
MSC: Primary 28A12

January 2018] NOTES 77

 symbols) if ν(A) = 0 implies μ(A) = 0 for all A ∈ A. Singularity of μ and ν (denoted
by μ ⊥ ν) means that there exists a set P ∈ A such that μ(P) = ν(X \ P) = 0. We are
going to prove that for every pair of nonnegative finite measures μ and ν there exist
two nonnegative finite measures μa and μs whose sum is μ. Moreover, μa  ν and
μs ⊥ ν. We remark that if such a decomposition exists, it must be unique. Indeed, if
μ = μa + μs and μ = μa + μs are two Lebesgue decompositions, then μa − μa is a
signed measure that is both absolutely continuous and singular with respect to ν, and
hence μa − μa = θ. For the corresponding notions and statements, see [1, Section 37]
or [9, Section 30].
Since singularity is the key notion of this paper, for the sake of completeness we
present here the sketch of a well-known characterization. Namely, the following lemma
states that singularity can be viewed as an order property. Recall first that the set of
nonnegative finite measures is a lattice (see [1, Theorem 36.1]), and the greatest lower
bound can be expressed as follows:
  

(μ ∧ ν )(A) = inf μ(A ∩ P) + ν(A \ P)  P ∈ A for all A ∈ A.
Downloaded by [Purdue University Libraries] at 08:36 06 January 2018

Lemma. Let μ and ν be nonnegative finite measures on A. Their greatest lower bound
is the zero measure if and only if there exists a P ∈ A such that μ(P) = ν(X \ P) = 0.

Proof. If such a P ∈ A exists, then ϑ ≤ μ and ϑ ≤ ν imply for all A ∈ A that

ϑ (A) = ϑ (A ∩ P) + ϑ (A \ P) ≤ μ(P) + ν(X \ P) = 0.
For the converse implication, assume that μ ∧ ν is the zero measure. Then,
    
μ ∧ ν (X ) = inf μ(X ∩ P) + ν(X \ P)  P ∈ A = 0
implies for every k ∈ N that there exists a measurable subset Sk ∈ A such that
1 1
μ(X ∩ Sk ) ≤ and ν(X \ Sk ) ≤ .
2k+1 2k+1
∞ ∞
Now if Pn := Sk , then P := Pn satisfies μ(P) = ν(X \ P) = 0, because
k=n n=1
∞
1 1
μ Sk ≤ , and k ≥ n implies ν(X \ Pn ) ≤ ν(X \ Sk ) ≤ . 
k=n
2n 2k+1

Now, we are going to prove the existence of the Lebesgue decomposition. For the
sake of simplicity, we recall two simple facts separately. These properties of decreasing
sequences of nonnegative finite measures play a crucial role in the proof. Let (μn )n∈N be
such a sequence (θ ≤ μ j ≤ μi if i ≤ j), and we denote by μ its setwise limit. First, μ is
a nonnegative finite measure that equals the greatest lower bound of the set {μn | n ∈ N}.
The second assertion is trivial if we know that μ is indeed a nonnegative finite measure.
To see this, observe that μ is finitely additive, and hence for every sequence (An )n∈N of
pairwise disjoint sets we have
∞ n−1 ∞ ∞
0≤μ Ak − μ Ak =μ Ak ≤ μ1 Ak → 0, as n → ∞.
k=1 k=1 k=n k=n

Second, if ν is an arbitrary nonnegative finite measure, then lim μn ∧ ν = μ ∧ ν. In-

n→∞
deed, denoting the limit by η, one can observe that η ≤ ν and η ≤ μn for all n ∈ N.
Consequently, η ≤ μ, which implies that η ≤ μ ∧ ν. The reverse inequality is trivial.

78 
C THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 125
 Theorem. Let μ and ν be nonnegative finite measures on A. Then, μ splits into the
sum of two nonnegative finite measures μ = μa + μs , where μa  ν and μs ⊥ ν.

Proof. According to the lemma, we can identify ν-singular nonnegative finite measures
as fixed points of the nonnegative finite measure-valued map s defined by
s(ϑ ) := ϑ − ϑ ∧ ν. (1)
First observe that 0 ≤ s(ϑ ) ≤ ϑ for all ϑ, and hence the following iteration
s[0] (μ) := μ, and s[n] (μ) := s(s[n−1] (μ)) for n ≥ 1 (2)
defines a decreasing sequence of nonnegative finite measures. Let us denote by μs its
setwise limit. Also, observe the following two consequences of (1) and (2):

n
s[n] (μ) ∧ ν = s[n] (μ) − s[n+1] (μ) and μ − s[n+1] (μ) = (s[k] (μ) ∧ ν ).
k=1
Downloaded by [Purdue University Libraries] at 08:36 06 January 2018

The first formula implies that μs and ν are singular, the second one guarantees the
ν-absolute continuity of μa := μ − μs . Indeed, μs ∧ ν = θ according to
   
μs ∧ ν = lim s[n] (μ) ∧ ν = lim (s[n] (μ) − μs ) − (s[n+1] (μ) − μs ) = θ,
n→∞ n→∞
 
and if A ∈ A such that ν(A) = 0, then s[k] (μ) ∧ ν (A) = 0 for all k ∈ N. Hence,
   
n
 [k] 
μa (A) = lim μ − s[n+1] (μ) (A) = lim s (μ) ∧ ν (A) = 0. 
n→∞ n→∞
k=0

3. REMARK. This approach also works for bounded charges (see [5, Theorem 2.2.1]
for the toolkit), but uniqueness fails in that setting. The reason is that this notion of ab-
solute continuity is too weak in the finitely additive case, and there are nonzero charges
that are both absolutely continuous and singular with respect to another charge. For an
example, see [11, Example 1.2, p. 780].

ACKNOWLEDGMENTS. The author was supported by the National Research, Development and Innova-
tion Office, NKFIH-104206.

REFERENCES

[1] Aliprantis, C. D., Burkinshaw, O. (1998). Principles of Real Analysis, 3rd ed. New York and London:
Academic Press.
[2] Ando, T. (1976). Lebesgue-type decomposition of positive operators. Acta Sci. Math. (Szeged). 38: 253–
260.
[3] Ando, T. (1999). Problem of infimum in the positive cone. In: Rassias, T. M., Srivastava, H. M., eds.
Analytic and Geometric Inequalities and Applications. Mathematics and Its Applications, Vol. 478. Dor-
drecht: Kluwer Academic Publishers, pp. 1–12.
[4] Arlinskiı̆, Yu. M. (2017). On the mappings connected with parallel addition of nonnegative operators.
Positivity. 21: 299–327.
[5] Bhaskara Rao, K. P. S., Bhaskara Rao, M. (1983). Theory of Charges. New York: Academic Press.
[6] Bochner, S., Phillips, R. S. (1941). Additive set functions and vector lattices. Ann. Math. 42(1): 316–324.
[7] Brooks, S. K. (1971). The Lebesgue decomposition theorem for measures. Amer. Math. Monthly. 78:
660–662.
[8] Eriksson, S.-L., Leutwiler, H. (1986). A potential-theoretic approach to parallel addition. Math. Ann.
274: 301–317.

January 2018] NOTES 79

 [9] Halmos, P. R. (1974). Measure Theory. Graduate Texts in Mathematics, Vol. 18. New York: Springer-
Verlag.
[10] Hewitt, E., Stromberg, K. (1965). Real and Abstract Analysis. Graduate Texts in Mathematics, Vol. 25,
New York: Springer-Verlag.
[11] König, H. (2012). Measure and Integration: Publications 1997–2011. Basel: Birkhäuser/Springer Basel
AG.
[12] Saks, S. (1964). Theory of the Integral, 2nd revised ed. New York: Dover. (English trans. by L. C. Young.
With two additional notes by Stefan Banach.)
[13] Schep, A. R. (2003). And still one more proof of the Radon–Nikodym theorem. Amer. Math. Monthly.
110: 536–538.
[14] Tarcsay, Zs. (2013). A functional analytic proof of the Lebesgue–Darst decomposition theorem. Real
Analy. Exchan. 39(1): 219–226.
[15] Titkos, T. (2015). A simple proof of the Lebesgue decomposition theorem. Amer. Math. Monthly. 122:
793–794.
[16] Yam Ting Woo, J. (1971). An elementary proof of the Lebesgue decomposition theorem. Amer. Math.
Monthly. 78: 783.

Alfréd Rényi Institute of Mathematics of the Hungarian Academy of Sciences, Budapest, Hungary
Downloaded by [Purdue University Libraries] at 08:36 06 January 2018

titkos.tamas@renyi.mta.hu

A Noninductive Proof of de Moivre’s Formula

The classical proof of de Moivre’s formula using mathematical induction is well
known. We have found another proof using the fact that if the derivative of a complex
function of a real variable is identically zero, then the function is constant.

Theorem. Let n be√an integer and t a real number. Then (cos t + i sin t)n = cos nt +
i sin nt, where i = −1.

Proof. The result clearly holds for n = 0. So, let n = 0. For convenience, we define
f (t) = cos t + i sin t. We need to show that f (t)n = f (nt) or, equivalently, f (nt) f (− t)n
= 1. Using the fact that f (t) = if (t) and the product rule for derivatives, we have, for
all real t,
d
( f (nt ) f (−t )n ) = n f (−t )n−1 { f (nt ) f (−t ) − f (nt ) f (−t )}
dt
= in f (−t )n−1 { f (nt ) f (−t ) − f (nt ) f (−t )}
= 0.
So, we have f (nt) f (− t)n = c for some fixed complex number c, which, by taking t
= 0, gives c = f (0) = 1. Therefore, f (nt) f (− t)n = 1 as desired. 

—Submitted by Jitender Singh, Guru Nanak Dev University Amritsar, India

doi.org/10.1080/00029890.2018.1392751
MSC: Primary 26A06, Secondary 26A24

80 
C THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 125