VOOZH about

URL: https://en.wikipedia.org/wiki/Prime_omega_function

⇱ Prime omega function - Wikipedia


Jump to content
From Wikipedia, the free encyclopedia
Number of prime factors of a natural number

In number theory, the prime omega functions πŸ‘ {\displaystyle \omega (n)}
and πŸ‘ {\displaystyle \Omega (n)}
count the number of prime factors of a natural number πŸ‘ {\displaystyle n}
. The number of distinct prime factors is assigned to πŸ‘ {\displaystyle \omega (n)}
(little omega), while πŸ‘ {\displaystyle \Omega (n)}
(big omega) counts the total number of prime factors with multiplicity (see arithmetic function). That is, if we have a prime factorization of πŸ‘ {\displaystyle n}
of the form πŸ‘ {\displaystyle n=p_{1}^{\alpha _{1}}p_{2}^{\alpha _{2}}\cdots p_{k}^{\alpha _{k}}}
for distinct primes πŸ‘ {\displaystyle p_{i}}
(πŸ‘ {\displaystyle 1\leq i\leq k}
), then the prime omega functions are given by πŸ‘ {\displaystyle \omega (n)=k}
and πŸ‘ {\displaystyle \Omega (n)=\alpha _{1}+\alpha _{2}+\cdots +\alpha _{k}}
. These prime-factor-counting functions have many important number theoretic relations.

Properties and relations

[edit]

The function πŸ‘ {\displaystyle \omega (n)}
is additive and πŸ‘ {\displaystyle \Omega (n)}
is completely additive. Little omega has the formula

πŸ‘ {\displaystyle \omega (n)=\sum _{p\mid n}1,}

where notation p|n indicates that the sum is taken over all primes p that divide n, without multiplicity. For example, πŸ‘ {\displaystyle \omega (12)=\omega (2^{2}3)=2}
.

Big omega has the formulas

πŸ‘ {\displaystyle \Omega (n)=\sum _{p^{\alpha }\mid n}1=\sum _{p^{\alpha }\parallel n}\alpha .}

The notation pΞ±|n indicates that the sum is taken over all prime powers pΞ± that divide n, while pΞ±||n indicates that the sum is taken over all prime powers pΞ± that divide n and such that n / pΞ± is coprime to pΞ±. For example, πŸ‘ {\displaystyle \Omega (12)=\Omega (2^{2}3^{1})=3}
.

The omegas are related by the inequalities Ο‰(n) ≀ Ξ©(n) and 2Ο‰(n) ≀ d(n) ≀ 2Ξ©(n), where d(n) is the divisor-counting function.[1] If Ξ©(n) = Ο‰(n), then n is squarefree and related to the MΓΆbius function by

πŸ‘ {\displaystyle \mu (n)=(-1)^{\omega (n)}=(-1)^{\Omega (n)}.}

If πŸ‘ {\displaystyle \omega (n)=1}
then πŸ‘ {\displaystyle n}
is a prime power, and if πŸ‘ {\displaystyle \Omega (n)=1}
then πŸ‘ {\displaystyle n}
is prime.

An asymptotic series for the average order of πŸ‘ {\displaystyle \omega (n)}
is [2]

πŸ‘ {\displaystyle {\frac {1}{n}}\sum \limits _{k=1}^{n}\omega (k)\sim \log \log n+B_{1}+\sum _{k\geq 1}\left(\sum _{j=0}^{k-1}{\frac {\gamma _{j}}{j!}}-1\right){\frac {(k-1)!}{(\log n)^{k}}},}

where πŸ‘ {\displaystyle B_{1}\approx 0.26149721}
is the Mertens constant and πŸ‘ {\displaystyle \gamma _{j}}
are the Stieltjes constants.

The function πŸ‘ {\displaystyle \omega (n)}
is related to divisor sums over the MΓΆbius function and the divisor function, including:[3]

πŸ‘ {\displaystyle \sum _{d\mid n}|\mu (d)|=2^{\omega (n)}}
is the number of unitary divisors. (sequence A034444 in the OEIS)
πŸ‘ {\displaystyle \sum _{d\mid n}|\mu (d)|k^{\omega (d)}=(k+1)^{\omega (n)}}
πŸ‘ {\displaystyle \sum _{r\mid n}2^{\omega (r)}=d(n^{2})}
πŸ‘ {\displaystyle \sum _{r\mid n}2^{\omega (r)}d\left({\frac {n}{r}}\right)=d^{2}(n)}
πŸ‘ {\displaystyle \sum _{d\mid n}(-1)^{\omega (d)}=\prod \limits _{p^{\alpha }||n}(1-\alpha )}
πŸ‘ {\displaystyle \sum _{\stackrel {1\leq k\leq m}{(k,m)=1}}\gcd(k^{2}-1,m_{1})\gcd(k^{2}-1,m_{2})=\varphi (n)\sum _{\stackrel {d_{1}\mid m_{1}}{d_{2}\mid m_{2}}}\varphi (\gcd(d_{1},d_{2}))2^{\omega (\operatorname {lcm} (d_{1},d_{2}))},\ m_{1},m_{2}{\text{ odd}},m=\operatorname {lcm} (m_{1},m_{2})}
πŸ‘ {\displaystyle \sum _{\stackrel {1\leq k\leq n}{\operatorname {gcd} (k,m)=1}}\!\!\!\!1=n{\frac {\varphi (m)}{m}}+O\left(2^{\omega (m)}\right)}

The characteristic function of the primes can be expressed by a convolution with the MΓΆbius function:[4]

πŸ‘ {\displaystyle \chi _{\mathbb {P} }(n)=(\mu \ast \omega )(n)=\sum _{d|n}\omega (d)\mu (n/d).}

A partition-related exact identity for πŸ‘ {\displaystyle \omega (n)}
is given by [5]

πŸ‘ {\displaystyle \omega (n)=\log _{2}\left[\sum _{k=1}^{n}\sum _{j=1}^{k}\left(\sum _{d\mid k}\sum _{i=1}^{d}p(d-ji)\right)s_{n,k}\cdot |\mu (j)|\right],}

where πŸ‘ {\displaystyle p(n)}
is the partition function, πŸ‘ {\displaystyle \mu (n)}
is the MΓΆbius function, and the triangular sequence πŸ‘ {\displaystyle s_{n,k}}
is expanded by

πŸ‘ {\displaystyle s_{n,k}=[q^{n}](q;q)_{\infty }{\frac {q^{k}}{1-q^{k}}}=s_{o}(n,k)-s_{e}(n,k),}

in terms of the infinite q-Pochhammer symbol and the restricted partition functions πŸ‘ {\displaystyle s_{o/e}(n,k)}
which respectively denote the number of πŸ‘ {\displaystyle k}
's in all partitions of πŸ‘ {\displaystyle n}
into an odd (even) number of distinct parts.[6]

Continuation to the complex plane

[edit]

A continuation of πŸ‘ {\displaystyle \omega (n)}
has been found, though it is not analytic everywhere.[7] Note that the normalized πŸ‘ {\displaystyle \operatorname {sinc} }
function πŸ‘ {\displaystyle \operatorname {sinc} (x)={\frac {\sin(\pi x)}{\pi x}}}
is used.

πŸ‘ {\displaystyle \omega (z)=\log _{2}\left(\sum _{n=1}^{\lceil Re(z)\rceil }\operatorname {sinc} \left(\prod _{m=1}^{\lceil Re(z)\rceil +1}\left(n^{2}+n-mz\right)\right)\right)}

This is closely related to the following partition identity. Consider partitions of the form

πŸ‘ {\displaystyle a={\frac {2}{c}}+{\frac {4}{c}}+\ldots +{\frac {2(b-1)}{c}}+{\frac {2b}{c}}}

where πŸ‘ {\displaystyle a}
, πŸ‘ {\displaystyle b}
, and πŸ‘ {\displaystyle c}
are positive integers, and πŸ‘ {\displaystyle a>b>c}
. The number of partitions is then given by πŸ‘ {\displaystyle 2^{\omega (a)}-2}
. [8]

Average order and summatory functions

[edit]

An average order of both πŸ‘ {\displaystyle \omega (n)}
and πŸ‘ {\displaystyle \Omega (n)}
is πŸ‘ {\displaystyle \log \log n}
. When πŸ‘ {\displaystyle n}
is prime a lower bound on the value of the function is πŸ‘ {\displaystyle \omega (n)=1}
. Similarly, if πŸ‘ {\displaystyle n}
is primorial then the function is as large as

πŸ‘ {\displaystyle \omega (n)\sim {\frac {\log n}{\log \log n}}}

on average order. When πŸ‘ {\displaystyle n}
is a power of 2, then πŸ‘ {\displaystyle \Omega (n)=\log _{2}(n).}
[9]

Asymptotics for the summatory functions over πŸ‘ {\displaystyle \omega (n)}
, πŸ‘ {\displaystyle \Omega (n)}
, and powers of πŸ‘ {\displaystyle \omega (n)}
are respectively[10][11]

πŸ‘ {\displaystyle {\begin{aligned}\sum _{n\leq x}\omega (n)&=x\log \log x+B_{1}x+o(x)\\\sum _{n\leq x}\Omega (n)&=x\log \log x+B_{2}x+o(x)\\\sum _{n\leq x}\omega (n)^{2}&=x(\log \log x)^{2}+O(x\log \log x)\\\sum _{n\leq x}\omega (n)^{k}&=x(\log \log x)^{k}+O(x(\log \log x)^{k-1}),k\in \mathbb {Z} ^{+},\end{aligned}}}

where πŸ‘ {\displaystyle B_{1}\approx 0.2614972128}
is the Mertens constant and the constant πŸ‘ {\displaystyle B_{2}}
is defined by

πŸ‘ {\displaystyle B_{2}=B_{1}+\sum _{p{\text{ prime}}}{\frac {1}{p(p-1)}}\approx 1.0345061758.}

The sum of number of unitary divisors is

πŸ‘ {\displaystyle \sum _{n\leq x}2^{\omega (n)}=(x\log x)/\zeta (2)+O(x)}
[12] (sequence A064608 in the OEIS)

Other sums relating the two variants of the prime omega functions include [13]

πŸ‘ {\displaystyle \sum _{n\leq x}\left\{\Omega (n)-\omega (n)\right\}=O(x),}

and

πŸ‘ {\displaystyle \#\left\{n\leq x:\Omega (n)-\omega (n)>{\sqrt {\log \log x}}\right\}=O\left({\frac {x}{(\log \log x)^{1/2}}}\right).}

Example I: A modified summatory function

[edit]

In this example we suggest a variant of the summatory functions πŸ‘ {\displaystyle S_{\omega }(x):=\sum _{n\leq x}\omega (n)}
estimated in the above results for sufficiently large πŸ‘ {\displaystyle x}
. We then prove an asymptotic formula for the growth of this modified summatory function derived from the asymptotic estimate of πŸ‘ {\displaystyle S_{\omega }(x)}
provided in the formulas in the main subsection of this article above.[14]

To be completely precise, let the odd-indexed summatory function be defined as

πŸ‘ {\displaystyle S_{\operatorname {odd} }(x):=\sum _{n\leq x}\omega (n)[n{\text{ odd}}],}

where πŸ‘ {\displaystyle [\cdot ]}
denotes Iverson bracket. Then we have that

πŸ‘ {\displaystyle S_{\operatorname {odd} }(x)={\frac {x}{2}}\log \log x+{\frac {(2B_{1}-1)x}{4}}+\left\{{\frac {x}{4}}\right\}-\left[x\equiv 2,3{\bmod {4}}\right]+O\left({\frac {x}{\log x}}\right).}

The proof of this result follows by first observing that

πŸ‘ {\displaystyle \omega (2n)={\begin{cases}\omega (n)+1,&{\text{if }}n{\text{ is odd; }}\\\omega (n),&{\text{if }}n{\text{ is even,}}\end{cases}}}

and then applying the asymptotic result from Hardy and Wright for the summatory function over πŸ‘ {\displaystyle \omega (n)}
, denoted by πŸ‘ {\displaystyle S_{\omega }(x):=\sum _{n\leq x}\omega (n)}
, in the following form:

πŸ‘ {\displaystyle {\begin{aligned}S_{\omega }(x)&=S_{\operatorname {odd} }(x)+\sum _{n\leq \left\lfloor {\frac {x}{2}}\right\rfloor }\omega (2n)\\&=S_{\operatorname {odd} }(x)+\sum _{n\leq \left\lfloor {\frac {x}{4}}\right\rfloor }\left(\omega (4n)+\omega (4n+2)\right)\\&=S_{\operatorname {odd} }(x)+\sum _{n\leq \left\lfloor {\frac {x}{4}}\right\rfloor }\left(\omega (2n)+\omega (2n+1)+1\right)\\&=S_{\operatorname {odd} }(x)+S_{\omega }\left(\left\lfloor {\frac {x}{2}}\right\rfloor \right)+\left\lfloor {\frac {x}{4}}\right\rfloor .\end{aligned}}}

Example II: Summatory functions for so-termed factorial moments of Ο‰(n)

[edit]

The computations expanded in Chapter 22.11 of Hardy and Wright provide asymptotic estimates for the summatory function

πŸ‘ {\displaystyle \omega (n)\left\{\omega (n)-1\right\},}

by estimating the product of these two component omega functions as

πŸ‘ {\displaystyle \omega (n)\left\{\omega (n)-1\right\}=\sum _{\stackrel {pq\mid n}{\stackrel {p\neq q}{p,q{\text{ prime}}}}}1=\sum _{\stackrel {pq\mid n}{p,q{\text{ prime}}}}1-\sum _{\stackrel {p^{2}\mid n}{p{\text{ prime}}}}1.}

We can similarly calculate asymptotic formulas more generally for the related summatory functions over so-termed factorial moments of the function πŸ‘ {\displaystyle \omega (n)}
.

Dirichlet series

[edit]

A known Dirichlet series involving πŸ‘ {\displaystyle \omega (n)}
and the Riemann zeta function is given by [15]

πŸ‘ {\displaystyle \sum _{n\geq 1}{\frac {2^{\omega (n)}}{n^{s}}}={\frac {\zeta ^{2}(s)}{\zeta (2s)}},\ \Re (s)>1.}

We can also see that

πŸ‘ {\displaystyle \sum _{n\geq 1}{\frac {z^{\omega (n)}}{n^{s}}}=\prod _{p}\left(1+{\frac {z}{p^{s}-1}}\right),|z|<2,\Re (s)>1,}
πŸ‘ {\displaystyle \sum _{n\geq 1}{\frac {z^{\Omega (n)}}{n^{s}}}=\prod _{p}\left(1-{\frac {z}{p^{s}}}\right)^{-1},|z|<2,\Re (s)>1,}

The function πŸ‘ {\displaystyle \Omega (n)}
is completely additive, where πŸ‘ {\displaystyle \omega (n)}
is strongly additive (additive). Now we can prove a short lemma in the following form which implies exact formulas for the expansions of the Dirichlet series over both πŸ‘ {\displaystyle \omega (n)}
and πŸ‘ {\displaystyle \Omega (n)}
:

Lemma. Suppose that πŸ‘ {\displaystyle f}
is a strongly additive arithmetic function defined such that its values at prime powers is given by πŸ‘ {\displaystyle f(p^{\alpha }):=f_{0}(p,\alpha )}
, i.e., πŸ‘ {\displaystyle f(p_{1}^{\alpha _{1}}\cdots p_{k}^{\alpha _{k}})=f_{0}(p_{1},\alpha _{1})+\cdots +f_{0}(p_{k},\alpha _{k})}
for distinct primes πŸ‘ {\displaystyle p_{i}}
and exponents πŸ‘ {\displaystyle \alpha _{i}\geq 1}
. The Dirichlet series of πŸ‘ {\displaystyle f}
is expanded by

πŸ‘ {\displaystyle \sum _{n\geq 1}{\frac {f(n)}{n^{s}}}=\zeta (s)\times \sum _{p\mathrm {\ prime} }(1-p^{-s})\cdot \sum _{n\geq 1}f_{0}(p,n)p^{-ns},\Re (s)>\min(1,\sigma _{f}).}

Proof. We can see that

πŸ‘ {\displaystyle \sum _{n\geq 1}{\frac {u^{f(n)}}{n^{s}}}=\prod _{p\mathrm {\ prime} }\left(1+\sum _{n\geq 1}u^{f_{0}(p,n)}p^{-ns}\right).}

This implies that

πŸ‘ {\displaystyle {\begin{aligned}\sum _{n\geq 1}{\frac {f(n)}{n^{s}}}&={\frac {d}{du}}\left[\prod _{p\mathrm {\ prime} }\left(1+\sum _{n\geq 1}u^{f_{0}(p,n)}p^{-ns}\right)\right]{\Biggr |}_{u=1}=\prod _{p}\left(1+\sum _{n\geq 1}p^{-ns}\right)\times \sum _{p}{\frac {\sum _{n\geq 1}f_{0}(p,n)p^{-ns}}{1+\sum _{n\geq 1}p^{-ns}}}\\&=\zeta (s)\times \sum _{p\mathrm {\ prime} }(1-p^{-s})\cdot \sum _{n\geq 1}f_{0}(p,n)p^{-ns},\end{aligned}}}

wherever the corresponding series and products are convergent. In the last equation, we have used the Euler product representation of the Riemann zeta function.

The lemma implies that for πŸ‘ {\displaystyle \Re (s)>1}
,

πŸ‘ {\displaystyle {\begin{aligned}D_{\omega }(s)&:=\sum _{n\geq 1}{\frac {\omega (n)}{n^{s}}}=\zeta (s)P(s)\\&\ =\zeta (s)\times \sum _{n\geq 1}{\frac {\mu (n)}{n}}\log \zeta (ns)\\D_{\Omega }(s)&:=\sum _{n\geq 1}{\frac {\Omega (n)}{n^{s}}}=\zeta (s)\times \sum _{n\geq 1}P(ns)\\&\ =\zeta (s)\times \sum _{n\geq 1}{\frac {\phi (n)}{n}}\log \zeta (ns)\\D_{h}(s)&:=\sum _{n\geq 1}{\frac {h(n)}{n^{s}}}=\zeta (s)\log \zeta (s)\\&\ =\zeta (s)\times \sum _{n\geq 1}{\frac {\varepsilon (n)}{n}}\log \zeta (ns),\end{aligned}}}

where πŸ‘ {\displaystyle P(s)}
is the prime zeta function, πŸ‘ {\displaystyle h(n)=\sum _{p^{k}|n}{\frac {1}{k}}=\sum _{p^{k}||n}{H_{k}}}
where πŸ‘ {\displaystyle H_{k}}
is the πŸ‘ {\displaystyle k}
-th harmonic number and πŸ‘ {\displaystyle \varepsilon }
is the identity for the Dirichlet convolution, πŸ‘ {\displaystyle \varepsilon (n)=\lfloor {\frac {1}{n}}\rfloor }
.

The distribution of the difference of prime omega functions

[edit]

The distribution of the distinct integer values of the differences πŸ‘ {\displaystyle \Omega (n)-\omega (n)}
is regular in comparison with the semi-random properties of the component functions. For πŸ‘ {\displaystyle k\geq 0}
, define

πŸ‘ {\displaystyle N_{k}(x):=\#(\{n\in \mathbb {Z} ^{+}:\Omega (n)-\omega (n)=k\}\cap [1,x]).}

These cardinalities have a corresponding sequence of limiting densities πŸ‘ {\displaystyle d_{k}}
such that for πŸ‘ {\displaystyle x\geq 2}

πŸ‘ {\displaystyle N_{k}(x)=d_{k}\cdot x+O\left(\left({\frac {3}{4}}\right)^{k}{\sqrt {x}}(\log x)^{\frac {4}{3}}\right).}

These densities are generated by the prime products

πŸ‘ {\displaystyle \sum _{k\geq 0}d_{k}\cdot z^{k}=\prod _{p}\left(1-{\frac {1}{p}}\right)\left(1+{\frac {1}{p-z}}\right).}

With the absolute constant πŸ‘ {\displaystyle {\hat {c}}:={\frac {1}{4}}\times \prod _{p>2}\left(1-{\frac {1}{(p-1)^{2}}}\right)^{-1}}
, the densities πŸ‘ {\displaystyle d_{k}}
satisfy

πŸ‘ {\displaystyle d_{k}={\hat {c}}\cdot 2^{-k}+O(5^{-k}).}

Compare to the definition of the prime products defined in the last section of [16] in relation to the ErdΕ‘s–Kac theorem.

See also

[edit]

Notes

[edit]
  1. ^ This inequality is given in Section 22.13 of Hardy and Wright.
  2. ^ S. R. Finch, Two asymptotic series, Mathematical Constants II, Cambridge Univ. Press, pp. 21-32, [1]
  3. ^ Each of these started from the second identity in the list are cited individually on the pages Dirichlet convolutions of arithmetic functions, Menon's identity, and other formulas for Euler's totient function. The first identity is a combination of two known divisor sums cited in Section 27.6 of the NIST Handbook of Mathematical Functions.
  4. ^ This is suggested as an exercise in Apostol's book. Namely, we write πŸ‘ {\displaystyle f=\mu \ast \omega }
    where πŸ‘ {\displaystyle f(n)=\sum _{d|n}\mu (n/d)\sum _{r|d}\left(\pi (r)-\pi (r-1)\right)}
    . We can form the Dirichlet series over πŸ‘ {\displaystyle f}
    as πŸ‘ {\displaystyle D_{f}(s):=\sum _{n\geq 1}{\frac {f(n)}{n^{s}}}=P(s),}
    where πŸ‘ {\displaystyle P(s)}
    is the prime zeta function. Then it becomes obvious to see that πŸ‘ {\displaystyle f(n)=\pi (n)-\pi (n-1)=\chi _{\mathbb {P} }(n)}
    is the indicator function of the primes.
  5. ^ This identity is proved in the article by Schmidt cited on this page below.
  6. ^ This triangular sequence also shows up prominently in the Lambert series factorization theorems proved by Merca and Schmidt (2017–2018)
  7. ^ Hoelscher, Zachary; Palsson, Eyvindur (2020-12-05). "Counting Restricted Partitions of Integers Into Fractions: Symmetry and Modes of the Generating Function and a Connection to Ο‰(t)". The PUMP Journal of Undergraduate Research. 3: 277–307. arXiv:2011.14502. doi:10.46787/pump.v3i0.2428. ISSN 2576-3725.
  8. ^ Hoelscher, Zachary; Palsson, Eyvindur (2020-12-05). "Counting Restricted Partitions of Integers Into Fractions: Symmetry and Modes of the Generating Function and a Connection to Ο‰(t)". The PUMP Journal of Undergraduate Research. 3: 277–307. arXiv:2011.14502. doi:10.46787/pump.v3i0.2428. ISSN 2576-3725.
  9. ^ For references to each of these average order estimates see equations (3) and (18) of the MathWorld reference and Section 22.10-22.11 of Hardy and Wright.
  10. ^ See Sections 22.10 and 22.11 for reference and explicit derivations of these asymptotic estimates.
  11. ^ Actually, the proof of the last result given in Hardy and Wright actually suggests a more general procedure for extracting asymptotic estimates of the moments πŸ‘ {\displaystyle \sum _{n\leq x}\omega (n)^{k}}
    for any πŸ‘ {\displaystyle k\geq 2}
    by considering the summatory functions of the factorial moments of the form πŸ‘ {\displaystyle \sum _{n\leq x}{\frac {\left[\omega (n)\right]!}{\left[\omega (n)-m\right]!}}}
    for more general cases of πŸ‘ {\displaystyle m\geq 2}
    .
  12. ^ Cohen, Eckford (1960). "The Number of Unitary Divisors of an Integer". The American Mathematical Monthly. 67 (9): 879–880. doi:10.2307/2309455. ISSN 0002-9890. JSTOR 2309455.
  13. ^ Hardy and Wright Chapter 22.11.
  14. ^ N.b., this sum is suggested by work contained in an unpublished manuscript by the contributor to this page related to the growth of the Mertens function. Hence it is not just a vacuous and/or trivial estimate obtained for the purpose of exposition here.
  15. ^ This identity is found in Section 27.4 of the NIST Handbook of Mathematical Functions.
  16. ^ RΓ©nyi, A.; TurΓ‘n, P. (1958). "On a theorem of ErdΓΆs-Kac" (PDF). Acta Arithmetica. 4 (1): 71–84. doi:10.4064/aa-4-1-71-84.

References

[edit]
  • G. H. Hardy and E. M. Wright (2006). An Introduction to the Theory of Numbers (6th ed.). Oxford University Press.
  • H. L. Montgomery and R. C. Vaughan (2007). Multiplicative number theory I. Classical theory (1st ed.). Cambridge University Press.
  • Schmidt, Maxie (2017). "Factorization Theorems for Hadamard Products and Higher-Order Derivatives of Lambert Series Generating Functions". arXiv:1712.00608 [math.NT].
  • Weisstein, Eric. "Distinct Prime Factors". MathWorld. Retrieved 22 April 2018.

External links

[edit]