Multiplicative decomposition and infinite divisibility of the mandel distribution

Infinite divisibility of the Mandel distribution that arises in quantum optics is proved. On the basis of this fact, the multiplicative decomposition of this distribution into the countable convolution of some Poisson distributions is constructed.

Gespeichert in:
Bibliographische Detailangaben
Datum:2005
Hauptverfasser: Virchenko, Yu.P., Vitokhina, N.N.
Format: Artikel
Sprache:Englisch
Veröffentlicht: Інститут математики НАН України 2005
Online Zugang:https://nasplib.isofts.kiev.ua/handle/123456789/4434
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Назва журналу:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Zitieren:Multiplicative decomposition and infinite divisibility of the mandel distribution / Yu.P. Virchenko, N.N. Vitokhina // Theory of Stochastic Processes. — 2005. — Т. 11 (27), № 3-4. — С. 131–139. — Бібліогр.: 9 назв.— англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1859837374333190144
author Virchenko, Yu.P.
Vitokhina, N.N.
author_facet Virchenko, Yu.P.
Vitokhina, N.N.
citation_txt Multiplicative decomposition and infinite divisibility of the mandel distribution / Yu.P. Virchenko, N.N. Vitokhina // Theory of Stochastic Processes. — 2005. — Т. 11 (27), № 3-4. — С. 131–139. — Бібліогр.: 9 назв.— англ.
collection DSpace DC
description Infinite divisibility of the Mandel distribution that arises in quantum optics is proved. On the basis of this fact, the multiplicative decomposition of this distribution into the countable convolution of some Poisson distributions is constructed.
first_indexed 2025-12-07T15:35:52Z
format Article
fulltext Theory of Stochastic Processes Vol. 11 (27), no. 3–4, 2005, pp. 131–139 UDC 519.21 YU. P. VIRCHENKO AND N. N. VITOKHINA MULTIPLICATIVE DECOMPOSITION AND INFINITE DIVISIBILITY OF THE MANDEL DISTRIBUTION Infinite divisibility of the Mandel distribution that arises in quantum optics is proved. On the basis of this fact, the multiplicative decomposition of this distribution into the countable convolution of some Poisson distributions is constructed. 1. Introduction The registration process of low-intensity electromagnetic radiation represents the ab- sorption of some energy portions named photons. The number ñ of registered photons during the time T is random according to its physical nature1). Therefore, the descrip- tion of the registration process is based on the probability distribution Pn ≡ Pr{ñ = n} of this random value. The determination of this distribution is the quantum optics prob- lem. In this section of theoretical physics, it is considered [1], [2] that, under some certain physical conditions, the distribution Pn is the composite Poissonian one, Pn = 1 n! E ( J̃n T exp[−J̃T ] ) , (1) where J̃T is the random variable representing the energy of the electromagnetic field absorbed by a photon counter during the registration time T . In quantum optics, it is referred to as the Mandel distribution [3]. In the work, we study the probability distribution (1) from the mathematical point of view in the physically most simple case where the electromagnetic radiation is the so-called one-mode and completely polarized. In addition, we consider that the electro- magnetic radiation is completely noisy. Mathematically, this case is described by the random variable J̃T which is defined by the formula [1] J̃T ≡ J [ζ̃] = T∫ 0 ∣∣∣ζ̃(s) ∣∣∣2 ds , where ζ̃(s) = ξ̃(s) + iη̃(s), s ∈ R, are trajectories of the complex process with real and imaginary parts ξ̃ = {ξ̃(t); t ∈ R}, η̃ = {η̃(t); t ∈ R} corresponding to Ornstein– Uhlenbeck processes which are stochastically equivalent and independent. The Ornstein–Uhlenbeck processes are Markovian and Gaussian ones, and they are completely characterized by these properties together with the stationarity condition. This class of processes is parametrized by two numbers ν > 0, σ > 0. In view of Markovian and Gaussian properties, each Ornstein–Uhlenbeck process is completely de- termined by the conditional probability density w(x0, t0|x, t) of the transition from the 2000 AMS Mathematics Subject Classification. Primary 60G35, 81K05. Key words and phrases. Infinite divisibility, Mandel distribution, Poisson distribution, Ornstein– Uhlenbeck process. This research has been partially supported by RFFI and Belgorod State University. 1The sign ”tilde” points out further that the corresponding mathematical object is random. 131 132 YU. P. VIRCHENKO AND N. N. VITOKHINA point x0 ∈ R at t0 ∈ R to the point x ∈ R at t ∈ R (see, for example, [4], III, §8, and also [5]). It depends on the parameters ν and σ and has the following form: w(x0, t0|x, t) = ( ν πσ ( 1 − e−2ν|t−t0| ) )1/2 exp ( −ν [ x − x0e −ν|t−t0|]2 σ ( 1 − e−2ν|t−t0| ) ) . (2) Then, fixing the parameters ν and σ completely determines the probability distributions of random variables JT [ξ̃], JT [η̃] and, thus, in view of independence and equivalence of the processes {ξ(t); t ∈ R}, {η(t); t ∈ R}, it defines the probability distribution density of the random variable JT [ζ̃]. The characteristic function Qξ(−iλ) = E exp(iλJT [ξ̃]), λ ∈ R, of the random variable JT [ξ̃] has been calculated in [6]: Qξ̃(λ) = E exp(−λJT [ξ̃]) = ( 4rν exp(νT ) (r + ν)2 exp(rT ) − (r − ν)2 exp(−rT ) )1/2 , (3) where r = √ ν2 + 2λσ. From this, the characteristic function is determined as a function of complex variable on a two-list Riemann surface. For the complex process {ζ(t); t ∈ R}, the similar function is determined by the formula Q(λ) = Qξ̃(λ)Qη̃(λ) = [ Qξ̃(λ) ]2 (4) in view of independence and equivalence of the processes ξ̃, η̃. Whence it follows that Q(λ) is meromorphic (see, Corollary 1 below). The availability of the explicit formula of the function Q(λ) allows us to use some methods of the theory of complex variable functions to study the properties of the rather complicated Mandel distribution Pn corresponding to this characteristic function. The purpose of the present work is to construct the multiplicative decomposition of the Man- del distribution on Poisson distributions and, in particular, to give the proof of its infinite divisibility. 2. Properties of the function Q(λ) According to (3) and (4), the function Q(λ) has the form Q(λ) = 4νr [F (r)]−1 eνT , (5) F (r) = (ν + r)2erT − (ν − r)2e−rT . (6) Lemma 1. The function [Q(λ)]−1 is an entire function of λ ∈ C. Proof. The function F (r) is obviously an entire function of the variable r. In addition, one can directly verify that it is odd. Hence, it can be represented in the form F (r) = rG(r2). Then G(x) is also an entire function of x ∈ C. Really, for the coefficients 〈ak; k ∈ Z+〉 of the decomposition F (r) = ∞∑ k=0 akr2k+1 determining the entire function F , we have lim sup k→∞ |ak|1/(2k+1) = 0 . Then the following decomposition of the function G(x) is valid: G(x) = ∞∑ k=0 akxk MULTIPLICATIVE DECOMPOSITION 133 . Here, the coefficients possess the property lim sup k→∞ |ak|1/k = ( lim sup k→∞ |ak|1/2k+1 )2 = 0 . Hence, the last power decomposition has the infinite radius of convergence. Let us notice that r2 = ν2 + 2νλ is the entire function of λ. Then, according to (5), [Q(λ)]−1 = 1 4ν G(r2)e−νT , (7) i.e. [Q(λ)]−1 is the superposition of entire functions and, hence, it is the entire function of λ. Lemma 2. The zeros {λn; n ∈ Z+} of the function [Q(λ)]−1 are real and negative. They are defined by the formula λn = −ν2 + x2 n 2σ , n ∈ Z+, (8) where xn are positive solutions of the equation tg xT = 2νx x2 − ν2 , x ∈ R . (9) They satisfy the inequalities π T ( n − 1 2 ) < xn ≤ π T ( n + 1 2 ) , n ∈ N, (10) and x0 satisfies the inequality ν ≤ x0 ≤ π/2T (x0 exists only if ν ≤ π/2T ). Proof. According to (7), the zeros of the function [Q(λ)]−1 are solutions of the equation G(r2) = 0. Then the set of zeros coincides with the set of solutions of the equation e2rT = ( ν − r ν + r )2 (11) except the point r = 0 since lim r→0 [Q(λ)]−1 = e−νT 4ν lim r→0 F (r) r = e−νT 4ν F ′(0) = νT 2 eνT �= 0 is valid at this point. Equating the modules of both parts of (11), we obtain ν2 + |r|2 − 2νRe r ν2 + |r|2 + 2νRe r = e2νT Re r . This equality is impossible at Re r �= 0 since, at Re r > 0, the left-hand side is less than 1, and the right-hand side is more than 1, ν �= 0. At Re r < 0, we have opposite inequalities. Thus, for the validity of (11), it is necessary that Re r = 0. Putting r = ix, x ∈ R, in (11) yields ±eixT = ν − ix ν + ix , whence ± cosxT = ν2 − x2 ν2 + x2 , ± sinxT = − 2νx ν2 + x2 , which is equivalent to (9). The solution x = 0 of this equation, i.e. r = 0, may not be taken into account. Further, if x is the solution of Eq. (9), then (−x) is also its solution. But they give the same value λ = −(x2 + ν2)/2σ. Then, to find zeros λn, n ∈ N, of the function [Q(λ)]−1, it is necessary to choose only one of them. Therefore, we consider further only positive solutions of Eq. (9). The function tg xT grows on x, and the function [−2νx/(ν2 − x2)] decreases since it has the derivative [−2ν(ν2 + x2)/(ν2 − x2)2]. Since the function tgxT is periodic with 134 YU. P. VIRCHENKO AND N. N. VITOKHINA period π/2T and varies from −∞ to ∞ on the period, there exists only one intersection of the graphs of these functions in each interval (π(2n − 1)/2T, π(2n + 1)/2T ]. Each intersection gives one solution of Eq. (9). It is a simple solution since the derivatives of both the above-mentioned functions are finite and not equal to zero. Hence, their graphs are intersected transversally. Hence, Eq. (9) has the infinite set of simple solutions xn, n ∈ N, satisfying inequalities (10). In addition, this equation has the solution x0 in the interval (0, π/2T ] satisfying the inequalities ν < x0 ≤ π/2T provided ν ≤ π/2T . From this analysis, it follows that Eq. (8) has the infinite set of simple zeros λn, n ∈ Z+, λn = −(2σ)−1(ν2 + x2 n), n ∈ Z+ and the zero λ0 is realized only when the condition ν ≤ π/2T is valid. Thus, |λ0| > ν2/σ. Corollary 1. The simple zeros of the function [Q(λ)]−1 are real and negative poles of the function Q(λ). Since they satisfy the condition |λn| → ∞ as n → ∞ in view of (8) and (10), the function Q(λ) is meromorphic. Since xn > a = min{ν, π/2T } , n ∈ Z+ , the function Q(λ) is analytic in a circle centered at λ = 0 with a radius not less than (ν2 + a2)/2σ. Lemma 3. The growth order of the entire function [Q(λ)]−1 does not exceed 1/2. Proof. The growth order θ of the entire function is defined by θ = lim x→∞ sup ln ln max{| [Q(λ)]−1 |; |λ| = x} ln x . (12) Due to (5) and (6), we get | [Q(λ)]−1 | ≤ e−νT 2ν|r| (ν + |r|)2e|r|T , whence lim |r|→∞ ln ln | [Q(λ)]−1 | ln |r| ≤ lim |r|→∞ ( 1 + ln T ln |r| ) = 1 . Since |r| ≤ (ν2 + 2σ|λ|)1/2 is valid, i.e. lim |λ|→∞ ln |r| ln |λ| ≤ 1 2 , relation (12) yields θ = [ lim |r|→∞ ln ln | [Q(λ)]−1 | ln |r| ] [ lim |λ|→∞ ln |r| ln |λ| ] ≤ 1 2 . Corollary 2. Since the growth order of the function [Q(λ)]−1 is less than 1, the Hada- mard decomposition [Q(λ)]−1 = [Q(0)]−1 ∏ n ( 1 − λ λn ) , (13) where Q(0) = 1, is valid for it (see, for example, [7]). 3. The multiplicative decomposition of the composite Poisson distribution In this section, we demonstrate that, under certain conditions, the composite Poisson distribution can be represented as the convolution of an infinite sequence of Poisson distributions p(l), l ∈ N with step l. MULTIPLICATIVE DECOMPOSITION 135 Definition. The lattice probability distribution p(l) = 〈p(l) n ; n ∈ Z+〉 with step l ∈ N is called the Poisson distribution with parameter αl ∈ R+ and step l ∈ N if p(l) n = {(αm l e−αl/m! ) , if n = ml, m ∈ Z+ ; 0; if n �= ml}. (14) The generator function of the probability distribution p(l) is presented in the following form: Hl(x) = e−αl ∞∑ m=0 αm l m! xlm = exp ( αl(xl − 1) ) . (15) Let ñ be the lattice random variable with values in Z+. It is distributed according to a composite Poisson distribution, i.e. the probabilities Pr{ñ = n}, n ∈ Z+, of its values are determined by the formula pn = 1 n! EJ̃n exp(−J̃) , (16) where J̃ is a random variable which takes its values on [0,∞). Let its characteristic function EeitJ̃ be analytic in a circle with the center at t = 0. Then the power series of this function converging in this circle is determined as EeitJ̃ = ∞∑ l=0 (it)l l! Ml , (17) where Ml = EJ̃ l, l ∈ Z+, M0 = 1 are the moments of the random variable J̃ . In addition, the power series of the logarithm of the characteristic function is defined, i.e. lnEeitJ̃ = ∞∑ l=0 (it)l l! Kl . (18) It converges in a circle that has, generally speaking, a smaller radius with the same center. The coefficients Kl, l ∈ Z+, K0 = 0 of this series are named cumulants of the random variable J̃ . All of them are finite. If there are no zeroes of the function EeitJ̃ , then the convergence radii of both decompositions (17) and (18) coincide. Thus, in the common convergence circle of both series, the following formula is valid: ∞∑ l=0 (it)l l! Ml = exp ( ∞∑ l=0 (it)l l! Kl ) . (19) Let H(z) = ∞∑ n=0 znpn (20) be the generator function of the random variable ñ. It is obviously analytic in the unit circle. According to the transformations H(z) = ∞∑ n=0 zn n! EJ̃ne−J̃ = E exp ( (z − 1)J̃ ) = = E ∞∑ n=0 (z − 1)n n! J̃n = ∞∑ n=0 (z − 1)n n! Mn and to the analyticity property of the function Q(λ), the generator function H(z) is analytic in a circle with the center at the point z = 1. Then there exists a negative real λ such that the series H(e−λ) = ∞∑ n=0 e−λnpn (21) 136 YU. P. VIRCHENKO AND N. N. VITOKHINA converges and, hence, the convergence radius of the power series (20) is greater than unity due to the fact that∣∣∣∣∣ ∞∑ n=0 pnzn ∣∣∣∣∣ ≤ ∞∑ n=0 pne−λn < ∞ , z ≤ e−λ. In particular, it follows from the above inequality that all the moments Eñl, l ∈ N, of the random variable ñ are finite, Eñl = ∞∑ n=0 pnnl < ∞ . Since H(0) = Ee−J̃ �= 0 and H ′(0) = EJ̃e−J̃ < M1, the analytic function ln H(z) is defined in the neighborhood of the point z = 0. Therefore, in this neighborhood, the decomposition ln H(z) = ∞∑ l=0 K̄l l! zl (22) is valid. We name the coefficients K̄l, l ∈ N as reduced cumulants. Further, let all the reduced cumulants K̄l with numbers l ∈ N be nonnegative. Then, from (22), we have the series ∞∑ l=1 K̄l l! zl−1 = 1 z (ln H(z) − K̄0) having nonnegative coefficients and converging in a circle with a nonzero radius. The modulus maximum of this analytic function is attained on the positive part of the real axis (see, for example, [9]). Then, if z∗ > 0 is the divergence point of this series, H(z∗) = ∞ is valid. Hence, z∗ > e−λ and, therefore, its convergence radius is not less than the convergence radius of series (20). From here, it follows that series (22) converges at the point z = 1, too. Since H(1) = 1, the identity ∞∑ l=0 1 l! K̄l = 0 (23) follows from (22) at z = 1. This means that − ∞∑ l=1 1 l! K̄l = K̄0 . Then formula (22) takes the form ln H(z) = ∞∑ l=1 zl − 1 l! K̄l . (24) Putting z = eit in (22), we obtain lnEeitñ = ∞∑ l=1 eitl − 1 l! K̄l . (25) This decomposition gives us the Kolmogorov representation lnEeitñ = itEñ + ∞∫ 0 ( eitx − 1 − itx ) dμ(x) x2 of the characteristic function of the infinitely divisible distribution corresponding to the nonnegative random variable with the finite second moment. Here, the measure μ(x) is a monotone nondecreasing function determining the finite measure on [0,∞) (see, for MULTIPLICATIVE DECOMPOSITION 137 example, [8]). In our case, this measure is determined by the formula μ(x) = ∞∑ l=1 lK̄l (l − 1)! χ(x − l) , where χ(x) is the indicator function of the set [0,∞). The finiteness of the measure μ, i.e. the convergence of the series μ(∞) = ∞∑ l=1 lK̄l (l − 1)! < ∞ , is guaranteed by the finiteness of the second moment Eñ2 of the random variable ñ. At last, let us consider the arbitrary pair of distributions (14) p(l) and p(m), l, m ∈ N, with the generator functions (15) which have αl = ( K̄l/l! ) , αm = ( K̄m/m! ) . The lattice distribution which is formed by the operation ◦ of convolution applying to this pair lattice distributions has components determined by the formula (p(l) ◦ p(m))n = n∑ k=0 p (l) k p (m) n−k , n ∈ Z+ . Therefore, its characteristic function is the product Hl(x)Hm(x). Then the distribution p corresponds to the generator function H(x) = exp ( ∞∑ l=1 xl − 1 l! K̄l ) = ∞∏ l=1 Hl(x) . Thus, this distribution is performed as the infinite convolution p = p(1) ◦ p(2) ◦ ... ◦ p(l) ◦ ... (26) of the Poisson distributions (14), where p(l) has step l and parameter αl = ( K̄l/l! ) . The infinite convolution (26) converges component-wise, since the Poisson distribu- tions p(l) with step l > n have the first nonzero component (p(l))k at such k > 0 that k = l. Then, for each fixed n ∈ Z+, the following formula is valid: pn = ( p(1) ◦ ... ◦ p(n) ) n ( p (n+1) 0 p (n+2) 0 ... ) . The expression in the first bracket is a finite sum, and the infinite product of null com- ponents in the second bracket converges since ∞∏ l=n+1 e−αl = exp ( − ∞∑ l=n+1 αl ) = exp ( − ∞∑ l=n+1 K̄l l! ) . We summarize the above analysis in the following statement. Theorem 1. Let the random variable J̃ be concentrated on R+, let its characteristic function be analytic in a circle with a nonzero radius, and let its reduced cumulants K̄l, l ∈ N be nonnegative. Then the composite Poisson distribution p defined by (16) and constructed on the basis of the random variable J̃ is infinitely divisible. It is represented as the component-wise converging infinite convolution (26) constructed on the basis of Poisson distributions p(l) which have steps l and parameters αl = ( K̄l/l! ) , l ∈ N, respec- tively. 4. The multiplicative decomposition of the Mandel distribution In this section, we demonstrate that Theorem 1 may be applied to the Mandel distri- bution that will lead us to the basic result of the work. 138 YU. P. VIRCHENKO AND N. N. VITOKHINA Theorem 2. If the parameters ν, σ of the random variable J̃T satisfy the condition ν2/2σ > 1, then its reduced cumulants are positive, and they are determined by the formula K̄l = ∞∑ n=1 (l − 1)! (1 + |λn|)l > 0 . Proof. From expression (13) for the function [Q(λ)]−1, taking into account the negativity of zeros λn, n ∈ Z+, we obtain lnEeλJ̃ = lnQ(−λ) = − ∞∑ n=1 ln ( 1 − λ |λn| ) = ∞∑ m=1 1 m ∞∑ n=1 ( λ |λn| )m . Hence, according to (18), the cumulants of the random variable J̃ are Km = ∞∑ n=1 (m − 1)! |λn|m . (27) Let the condition ν2/2σ > 1 be fulfilled. According to Corollary 1, all the moduli of the zeros λn, n ∈ Z+ exceed 1, and therefore the function Q(λ) is analytic in the circle centered at λ = 0 and with radius being more than 1. Then the decomposition ∞∑ n=0 (z − 1)n n! Mn = H(z) converges at |z − 1| ≤ 1 + ε at an ε > 0 being sufficiently small. Since there are no zeros of the function Q(λ), the power series of ln Q(λ) has the convergence radius greater than 1. Applying the logarithm operation to both sides of the last equality and using the definition of reduced cumulants, we obtain the equality of decompositions converging as |z − 1| ≤ 1 + ε: ∞∑ m=0 (z − 1)m m! Km = ∞∑ l=0 zl l! K̄l . The decomposition in the left-hand side of the equality converges in the neighborhood of the point z = 0. Therefore, by differentiating both sides of this equality with respect to z l times and by putting z = 0, we obtain K̄l = ( dl dzl ∞∑ m=0 (z − 1)m m! Km ) z=0 . Due to the fact that the power series converges uniformly in its convergence region, all differentiations commute with the summation. As a result, we obtain the following formula for reduced cumulants: K̄l = ∞∑ m=0 Km+l (−1)m m! . (28) The calculation of the reduced cumulants on the basis of formula (28) gives K̄l = ∞∑ m=0 (−1)m m! Km+l = ∞∑ m=0 (−1)m m! ∞∑ n=1 (m + l − 1)! |λn|m+l = = ∞∑ n=1 |λn|−l ∞∑ m=0 (−1)m m! (m + l − 1)! |λn|m = = ∞∑ n=1 |λn|−l [( d dy )l−1 ∞∑ m=0 (−1)mym+l−1 ] y=|λn|−1 = MULTIPLICATIVE DECOMPOSITION 139 = ∞∑ n=1 (−1)l−1 |λn|l [( d dy )l−1 ∞∑ m=0 (−1)mym ] y=|λn|−1 = = ∞∑ n=1 (−1)l−1 |λn|l [( d dy )l−1 (1 + y)−1 ] y=|λn|−1 = = (l − 1)! ∞∑ n=1 |λn|−l ( 1 + |λn|−1 )−l = (l − 1)! ∞∑ n=1 (1 + |λn|)−l > 0 . Then it follows that K̄l ≥ 0 and K̄l = ∞∑ n=1 (l − 1)! (1 + |λn|)l ≤ Kl , l ∈ N . From the proved statement, on the basis of Theorems 1 and 2, we come to the basic result of the work. Theorem 3. If the condition ν2/2σ > 1 for the parameters ν, σ determining the proba- bility distribution of the random variable J̃T is fulfilled, then the Mandel distribution is infinitely divisible. Thus, it decomposes into the infinite convolution P = p(1) ◦ p(2) ◦ ... ◦ p(n) ◦ ... of Poisson distributions p(l), l ∈ N, with steps l and parameters αl = 1 l ∞∑ n=1 (l − 1)! (1 + |λn|)l . Acknowledgement: We are grateful to RFFI and Belgorod State University for the financial support of this work Bibliography 1. M. Lax, Fluctuation and Coherence Phenomena in Classical and Quantum Physics, Gordon and Breach, New York, 1968. 2. R. Glauber, Optical Coherence and Photon Statistics, in Quantum Optics and Electronics. eds. C. de Witt, A. Blandin, C. Cohen-Tannoudji. Lectures at Les Houches during 1964 session of the Summer School of Theoretical Physics at University of Grenoble. Gordon and Breach, New York (1965), 91–279. 3. L. Mandel, Progress in Optics, Vol.2, North-Holland, Amsterdam, 1963. 4. W. Feller, An Introduction to Probability Theory and its Applications, vol.2, 2d ed., Wiley, New York, 1971. 5. C.W. Gardiner, Handbook of Stochastic Methods for Physics, Chemistry and the Natural Sci- ences. 2d ed, Spriger Series in Synergetics. vol. 13, Springer-Verlag, Berlin, 1983. 6. A.J.F. Ziegert, A systematic approach to a class problems in the theory of noise and other random phenomena. Part II, examples, Trans. IRE IT (1957), no. 3, 38–44. 7. A.I. Markushevich, Theory of Analytic Functions, vol.2, Nauka, Moscow, 1968. (in Russian) 8. E. Lukacs, Characteristic Functions, Griffin, London, 1970. 9. M.V. Fedoryuk, Saddle–Point Method, Nauka, Moscow, 1977. (in Russian) E-mail : virch@isc.kharkov.ua
id nasplib_isofts_kiev_ua-123456789-4434
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 0321-3900
language English
last_indexed 2025-12-07T15:35:52Z
publishDate 2005
publisher Інститут математики НАН України
record_format dspace
spelling Virchenko, Yu.P.
Vitokhina, N.N.
2009-11-09T15:40:26Z
2009-11-09T15:40:26Z
2005
Multiplicative decomposition and infinite divisibility of the mandel distribution / Yu.P. Virchenko, N.N. Vitokhina // Theory of Stochastic Processes. — 2005. — Т. 11 (27), № 3-4. — С. 131–139. — Бібліогр.: 9 назв.— англ.
0321-3900
https://nasplib.isofts.kiev.ua/handle/123456789/4434
519.21
Infinite divisibility of the Mandel distribution that arises in quantum optics is proved. On the basis of this fact, the multiplicative decomposition of this distribution into the countable convolution of some Poisson distributions is constructed.
This research has been partially supported by RFFI and Belgorod State University.
en
Інститут математики НАН України
Multiplicative decomposition and infinite divisibility of the mandel distribution
Article
published earlier
spellingShingle Multiplicative decomposition and infinite divisibility of the mandel distribution
Virchenko, Yu.P.
Vitokhina, N.N.
title Multiplicative decomposition and infinite divisibility of the mandel distribution
title_full Multiplicative decomposition and infinite divisibility of the mandel distribution
title_fullStr Multiplicative decomposition and infinite divisibility of the mandel distribution
title_full_unstemmed Multiplicative decomposition and infinite divisibility of the mandel distribution
title_short Multiplicative decomposition and infinite divisibility of the mandel distribution
title_sort multiplicative decomposition and infinite divisibility of the mandel distribution
url https://nasplib.isofts.kiev.ua/handle/123456789/4434
work_keys_str_mv AT virchenkoyup multiplicativedecompositionandinfinitedivisibilityofthemandeldistribution
AT vitokhinann multiplicativedecompositionandinfinitedivisibilityofthemandeldistribution