Skip to main content

Some inequalities involving the polygamma functions


Let \(\psi _{n}(x)=(-1)^{n-1}\psi ^{(n)}(x)\), where \(\psi ^{(n)}(x)\) are the polygamma functions. We determine necessary and sufficient conditions for the monotonicity and convexity of the function

$$ F(x;\alpha ,\beta )=\ln \bigl(\exp \bigl(\alpha \psi (x+\beta )\bigr)\psi _{n}(x) \bigr)- \ln (n-1)!, \quad x>\max (0,-\beta ), $$

for α and \(\beta \in \mathbb{R}\), where \(\psi (x)\) is the psi function. Consequently, this yields not only some new inequalities for the polygamma functions, but also new star-shaped and superadditive functions involving them. In addition, we improve a well-known mean-value inequality for the polygamma functions.


We recall that the logarithmic derivative of \(\varGamma (x)\) is called the psi or digamma function denoted by

$$ \psi (x)=\frac{d}{dx}\ln \varGamma (x)=\frac{\varGamma ^{\prime }(x)}{\varGamma (x)} $$

for \(x>0 \), that the derivatives \(\psi '(x)\) and \(\psi ''(x)\) are, respectively, called the tri-gamma and tetra-gamma functions, and that the derivatives

$$ \psi ^{(n)}(x)= (-1 )^{n-1} \int _{0}^{\infty }e^{-xt}\frac{t ^{n}}{1-e^{-t}} \,dt= (-1 )^{n-1}n!\sum_{k=0}^{\infty } \frac{1}{(x+k)^{n+1}} $$

for \(n\in \mathbb{N} \) are called the polygamma functions.

A world of the most fundamental properties involving the gamma, digamma, and polygamma functions can be found in some books [1,2,3,4]. It is known that these functions are all key parts of special functions. Moreover, they also play a vital role in other areas like analysis, physics, inequality theory, and statistics. Due to their significance, they attract many scholars to explore some their useful properties. In particular, some properties such as monotonicity, convexity, and complete monotonicity yield numerous inequalities related to these functions; see, for example, [5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28].

The following recurrence and asymptotic formulas are often encountered in the literature:

$$\begin{aligned}& \psi ^{(n)}(x+1)=\psi ^{(n)}(x)+(-1)^{n} \frac{n!}{x^{n+1}}\quad (x>0, n=0,1, \ldots ), \end{aligned}$$
$$\begin{aligned}& \psi (x)\sim \ln x-\frac{1}{2x}-\frac{1}{12x^{2}}+ \frac{1}{120x^{4}}- \cdots\quad (x\rightarrow \infty ), \end{aligned}$$
$$\begin{aligned}& (-1)^{n-1}\psi ^{n}(x)\sim \biggl( \frac{(n-1)!}{x^{n}}+\frac{n!}{2x ^{n+1}}+\frac{(n+1)!}{12x^{n+2}}-\cdots \biggr)\quad (x \rightarrow \infty , n=1,2,\ldots ). \end{aligned}$$

For convenience, we define \(\psi _{n}(x)=(-1)^{n-1}\psi ^{(n)}(x)\) for \(n\in \mathbb{N}\). It follows from (1.2) that \(\psi _{n}(x)\) is strictly completely monotonic on \((0,\infty )\) (see [29, 30]). Moreover, from (1.3) and (1.5) it follows that

$$\begin{aligned}& \lim_{x \rightarrow 0^{+}}x^{n+1}\psi _{n}(x)=n!, \end{aligned}$$
$$\begin{aligned}& \lim_{x \rightarrow \infty }x^{n}\psi _{n}(x)=(n-1)!. \end{aligned}$$

which easily yields \(\lim_{x \rightarrow 0^{+}}\psi _{n}(x)=\infty \) and \(\lim_{x \rightarrow \infty }\psi _{n}(x)=0\).

It was shown in [31, Thm. 1.9] that \(\psi _{n}(x)\) is logarithmically convex on \((0,\infty )\). This and Jensen’s theorem yield the following functional inequality:

$$ 1\leq \frac{\prod_{m=1}^{n} \psi _{k}^{p_{m}}(x_{m})}{\psi _{k}(\sum_{m=1}^{n} x_{m} p_{m})} \quad (n, k\in \mathbb{N}) $$

for all \(x_{m}>0\) and weights \(p_{m}\) (\(m=1,2,\ldots ,n\)), that is, \(p_{m}>0\) with \(\sum_{m=1}^{n} p_{m}=1\). Further, Alzer [32, Thm. 1] provided a refinement of (1.8) with the inequality

$$ \biggl(\frac{\sum_{m=1}^{n} p_{m} x_{m}}{\prod_{m=1}^{n} x_{m}^{p_{m}}} \biggr) ^{\alpha }\leq \frac{\prod_{m=1}^{n} \psi _{k}^{p_{m}}(x_{m})}{\psi _{k}(\sum_{m=1}^{n} x_{m} p_{m})} \quad (\alpha \in \mathbb{R}) $$

for all \(x_{m}>0\) and weights \(p_{m}\) (\(m=1,2,\ldots ,n\)) if and only if \(\alpha \leq k\). Then, inspired by these results, we are devoted to look for \(\alpha _{1}\), \(\alpha _{2}\), \(\beta _{1}\), and \(\beta _{2} \in \mathbb{R}\) such that, for all \(n, k\in \mathbb{N}\), the double inequality

$$ \frac{\exp (\alpha _{1} A_{n}(x_{m},p_{m},\beta _{1}) )}{ \exp (\alpha _{1} G_{n}(x_{m},p_{m},\beta _{1}) )}\leq \frac{ \prod_{m=1}^{n} \psi _{k}^{p_{m}}(x_{m})}{\psi _{k}(\sum_{m=1}^{n} x _{m} p_{m})}\leq \frac{\exp (\alpha _{2} A_{n}(x_{m},p_{m},\beta _{2}) )}{\exp (\alpha _{2} G_{n}(x_{m},p_{m},\beta _{2}) )} $$

is valid for all \(x_{m}>0\) and weights \(p_{m}\) (\(m=1,2,\ldots ,n\)). Here and in what follows,

$$ A_{n}(x_{m},p_{m},\beta )=\psi \Biggl(\sum _{m=1}^{n} x_{m} p_{m}+ \beta \Biggr) \quad \text{and} \quad G_{n}(x_{m},p_{m}, \beta )=\sum_{m=1}^{n} p_{m}\psi (x_{m}+\beta ). $$

It was proven in [32, Thm. 2] that

$$ M_{n}^{[s]} \bigl(\psi _{k}(x_{m}),p_{m} \bigr)\geq \psi _{k} \bigl(M _{n}^{[r]}(x_{m},p_{m}) \bigr)\quad (n, k\in \mathbb{N} \text{ and } r, s\in \mathbb{R}) $$

for all \(x_{m}>0\) and weights \(p_{m}>0\) (\(m=1,2,\ldots ,n\)) if and only if \(s\geq 0\) and \(r\geq -sk\), or \(s<0\) and \(r\geq -s(k+1)\).

The power means \(M_{n}^{[t]}\) are defined as (see [33])

$$ M_{n}^{[t]}(x_{m},p_{m})=\Biggl(\sum _{m=1}^{n} p_{m}x_{m}^{t} \Biggr)^{1/t} \quad (t\neq 0), \qquad M_{n}^{[0]}(x_{m},p_{m})= \prod_{m=1}^{n} x_{m}^{p_{m}}, $$

for parameter \(t\in \mathbb{R}\) and all \(x_{m}>0\) and weights \(p_{m}\) (\(m=1,2,\ldots ,n\)).

In view of (1.10), it is natural to raise a new question: under which conditions on r and s, the reversed inequality of (1.10) is still valid for all \(x_{m}>0\) and weights \(p_{m}\) (\(m=1,2,\ldots ,n\))?

The main aim of this paper is to settle the two questions posed thereinbefore. Our main results are the following.

Theorem 1.1

Let α, \(\beta \in \mathbb{R}\) and \(n\in \mathbb{N}\). Define the function

$$ F(x;\alpha ,\beta )=\ln \bigl(\exp \bigl(\alpha \psi (x+\beta )\bigr)\psi _{n}(x) \bigr)- \ln (n-1)!,\quad x>\max (0,-\beta ). $$

Then \(F(x;\alpha ,\beta )\) is strictly increasing and concave for \(\beta \leq 0\) if and only if \(\alpha \geq n\) and strictly decreasing and convex for \(\beta \geq \frac{1}{2}\) if and only if \(\alpha \leq n\).

Applying the asymptotic formulas (1.4) and (1.5) to the function \(F(x;n,\beta )\), we get \(\lim_{x\rightarrow \infty }F(x;n, \beta )=0\). This in combination with Theorem 1.1 yields the following bounds for \(\psi _{n}(x)\).

Corollary 1

Let \(n\in \mathbb{N}\), \(\alpha \leq 0\), and \(\beta \geq \frac{1}{2}\). Then we have the double inequality

$$ (n-1)!\exp \bigl(-n\psi (x+\beta ) \bigr)< \psi _{n}(x)< (n-1)!\exp \bigl(-n\psi (x+\alpha ) \bigr) $$

for \(x>-\alpha \).

As a matter of fact, the double inequality (1.11) was first proved by Batir [34, Thm. 2.1]. In addition, using the concavity and convexity of \(F(x;n,\beta )\) and applying Jensen’s inequality give the following double inequality.

Theorem 1.2

Let \(k\geq 1\) and \(n\geq 2\) be integers, and let \(\alpha _{1}\), \(\alpha _{2}\in \mathbb{R}\), \(\beta _{1}\geq \frac{1}{2}\), and \(\beta _{2}\leq 0\). Then we have the double inequality

$$ \frac{\exp (\alpha _{1} A_{n}(x_{m},p_{m},\beta _{1}) )}{ \exp (\alpha _{1} G_{n}(x_{m},p_{m},\beta _{1}) )}\leq \frac{ \prod_{m=1}^{n} \psi _{k}^{p_{m}}(x_{m})}{\psi _{k}(\sum_{m=1}^{n} x _{m} p_{m})}\leq \frac{\exp (\alpha _{2} A_{n}(x_{m},p_{m},\beta _{2}) )}{\exp (\alpha _{2} G_{n}(x_{m},p_{m},\beta _{2}) )} $$

for all \(x_{m}>-\beta _{2}\) and \(p_{m}>0\) (\(m=1,2,\ldots ,n\)) with \(\sum_{m=1}^{n} p_{m}=1\) if and only if \(\alpha _{1}\leq k\) and \(\alpha _{2}\geq k\).

Theorem 1.3

Let \(k\geq 1\) and \(n\geq 2\) be integers, and let r, \(s\in \mathbb{R}\). Then

$$ M_{n}^{[s]} \bigl(\psi _{k}(x_{m}),p_{m} \bigr)\leq \psi _{k} \bigl(M _{n}^{[r]}(x_{m},p_{m}) \bigr) $$

for all \(x_{m}>0\) and \(p_{m}>0\) (\(m=1,2,\ldots ,n\)) with \(\sum_{m=1} ^{n} p_{m}=1\) if and only if \(s\geq 1\) and \(r\leq -s(k+1)\), or \(s<1\) and \(r\leq -sk-M(s)\), where

$$ M(s)=\sup_{x>0} \bigl(f_{k+1}(x)-f_{k}(x)+sf_{k}(x) \bigr)\quad \textit{and} \quad f_{k}(x)=x\frac{\psi _{k+1}(x)}{\psi _{k}(x)}-k. $$

Remark 1

In brief, Theorem 1.3 can be restated as follows: inequality (1.13) is valid for all \(x_{m}>0\) and weights \(p_{m}\) (\(m=1,2, \ldots , n\)) if and only if \(r\leq -sk-M(s)\).

In the last section, we give some sharp inequalities and star-shaped functions, which are consequences of Theorem 1.1.

Some lemmas

In this section, we establish monotonicity properties of some functions closely related to the function \(e^{x}\).

According to Theorem 2.2 in [35], we obtain the following:

Lemma 1

([35, Theorem 2.2])

Let \(z\in \mathbb{R}\). Then the function

$$ g(x)=\frac{e^{-zx}-e^{-(1-z)x}}{1-e^{-x}} $$

is strictly increasing on \((0,\infty )\) if and only if \(z\in (-\infty ,0)\cup (\frac{1}{2},1)\) and strictly decreasing on the same interval if and only if \(z\in (0,\frac{1}{2})\cup (1,\infty )\).

Lemma 2

Let a, \(x_{1}\), and \(x_{2}\in \mathbb{R}\) with \(x_{2}>x_{1}>0\).

  1. (1)


    $$ H_{a}(x)=\frac{e^{a x_{1} x}-e^{(a-1)x_{1} x}}{e^{a x_{2} x}-e^{(a-1)x _{2} x}}. $$

    Then \(H_{a}(x)\) is strictly increasing on \((0,\infty )\) if and only if \(a\leq 0\) and strictly decreasing on the same interval if and only if \(a\geq \frac{1}{2}\).

  2. (2)

    The function

    $$ h(x)=\frac{1-e^{-x_{1} (1-x)}}{1-e^{-x_{2} (1-x)}}\frac{1-e^{-x_{1} x}}{1-e ^{-x_{2} x}}\frac{e^{a x_{1} x}}{e^{ax_{2} x}} $$

    is strictly decreasing on \((0,\infty )\) for \(a\geq \frac{1}{2}\).


(1) Differentiating \(H_{a}(x)\) with respect to x leads to

$$ \begin{aligned} &H_{a}'(x)\\ &\quad =\frac{ [a(x_{1}-x_{2})e^{(x_{1}+x_{2})x}+(a x_{2}-a x _{1}-x_{2})e^{x_{1}x}+(x_{1}+a x_{2}-a x_{1})e^{x_{2}x}+(a-1)(x_{1}-x _{2}) ]}{e^{(x-a x)(x_{1}+x_{2})} (e^{a x_{2} x}-e^{(a-1)x _{2} x} )^{2}}, \end{aligned} $$

which implies

$$ H_{a}'(x)=\frac{e^{(a x-x)(x_{1}+x_{2})}}{ (e^{a x_{2} x}-e^{(a-1)x _{2} x} )^{2}}\sum _{n=2}^{\infty }P_{n}(a)\frac{x^{n}}{n!}, $$


$$ P_{n}(a)=(x_{2}-x_{1}) \Biggl[\sum _{k=1}^{n-1}x_{1}^{k} x_{2}^{n-k} \biggl(1-a \frac{n!}{k!(n-k)!} \biggr) \Biggr],\quad n=2,3,\ldots . $$

It is not difficult to obtain that

$$ P_{n}(a)>0\quad \text{for all } n\geq 2, a\leq 0, $$


$$ P_{n}(a)< 0\quad \text{for all } n\geq 2, a\geq \frac{1}{2}. $$

Hence the function \(H_{a}(x)\) is strictly increasing on \((0,\infty )\) for \(a\leq 0\) and strictly decreasing on the same interval for \(a\geq \frac{1}{2}\). Now, we assume that the function \(H_{a}(x)\) is strictly increasing on \((0,\infty )\). Firstly, we need to notice that

$$ \rho (x)=e^{x_{1}x}+e^{x_{2}x}-e^{(x_{1}+x_{2})x}-1< 0 \quad \text{for } x>0. $$

Then, for \(x>0\), we have:

$$ a< \frac{x_{2}e^{x_{1}x}-x_{1}e^{x_{2}x}+x_{1}-x_{2}}{(x_{2}-x_{1})(e ^{x_{1}x}+e^{x_{2}x}-e^{(x_{1}+x_{2})x}-1)}. $$

Letting x tend to ∞, we obtain \(a\leq 0\).

If \(H_{a}(x)\) is strictly decreasing on \((0,\infty )\), then we have \(a\geq \frac{1}{2}\) as x tends to 0 in the reversed inequality of (2.4).

(2) We write

$$ h(x)=H_{0}(1-x)H_{a}(x). $$

From part (1) of the lemma we conclude that, for \(a\geq \frac{1}{2}\), \(h(x)\) is the product of two positive decreasing functions on \((0,\infty )\). Therefore it is easy to get the desired result.

This completes the proof of Lemma 2. □

Lemma 3

Let α, \(\beta \in \mathbb{R}\) and m, \(n\in \mathbb{N}\). The function

$$ f(x)=\frac{\psi _{m+n}(x+\beta )}{\psi _{m}(x+\alpha )\psi _{n}(x+\beta )} $$

is strictly increasing on \((\max (-\alpha ,-\beta ),\infty )\) for \(\alpha -\beta \leq 0\) and strictly decreasing on \((\max (-\alpha ,- \beta ),\infty )\) for \(\alpha -\beta \geq \frac{1}{2}\).


Let \(k=m+n\), \(\theta =\alpha -\beta \), and \(y=x+\beta \). Then we have

$$ f(x)=L(y)=\frac{\psi _{k}(y)}{\psi _{m}(y+\theta )\psi _{n}(y)},\quad y> \max (-\theta ,0). $$

If we determine conditions for θ such that the function \(L(y)\) is strictly increasing and strictly decreasing on \((\max (- \theta ,0),\infty )\), then the desired result follows.

For convenience, we replace y by x in \(L(y)\). Using the well-known formula (1.2) and applying the convolution theorem for the Laplace transform, we get

$$ L(x)=\frac{\psi _{k}(x)}{\psi _{m}(x+\theta )\psi _{n}(x)}=\frac{\int _{0}^{\infty } q(t)e^{-xt} \,dt}{\int _{0}^{\infty } p(t)e^{-xt} \,dt}, $$


$$ q(t)=\frac{t^{k}}{1-e^{-t}} \quad \text{and} \quad p(t)= \int _{0}^{t} \frac{s^{m}e^{-\theta s}}{1-e^{-s}}\frac{(t-s)^{n}}{1-e ^{-(t-s)}} \,ds. $$

Differentiating \(L(x)\), we get

$$ L'(x)=\frac{\int _{0}^{\infty } tp(t)e^{-xt} \,dt \int _{0}^{\infty } q(t)e ^{-xt} \,dt-\int _{0}^{\infty } tq(t)e^{-xt} \,dt \int _{0}^{\infty } p(t)e ^{-xt} \,dt}{ (\int _{0}^{\infty } p(t)e^{-xt} \,dt )^{2}}, $$

so that using the convolution theorem for the Laplace transform again in (2.6) yields

$$ L'(x) \biggl( \int _{0}^{\infty } p(t)e^{-xt} \,dt \biggr)^{2}= \int _{0} ^{\infty } u(t)e^{-xt} \,dt, $$


$$ u(t)= \int _{0}^{t}p(s)q(t-s) (2s-t) \,ds. $$

We will prove that

$$ u(t)>0\quad \text{for } \theta \leq 0, t>0, $$


$$ u(t)< 0\quad \text{for } \theta \geq \frac{1}{2}, t>0. $$

Making the substitution \(s=\frac{t}{2}(1+y)\), we obtain

$$ u(t)=\frac{t^{2}}{2} \int _{-1}^{1} yp \biggl(\frac{t}{2}(1+y) \biggr)q \biggl(\frac{t}{2}(1-y) \biggr) \,dy. $$

For \(\delta >0\) and \(y\in (0,1)\), we define

$$ v(y)=yp\bigl(\delta (1+y)\bigr)q\bigl(\delta (1-y)\bigr) \quad \text{and} \quad w(y)=v(y)+v(-y). $$

Combining (2.10) with (2.11), we see that

$$ u(2\delta )=2\delta ^{2} \int _{-1}^{1}v(y) \,dy=2\delta ^{2} \int _{0}^{1}w(y) \,dy. $$

Hence, to prove (2.8) and (2.9), it suffices to show that

$$ w(y)>0\quad \text{for } \theta \leq 0, \delta >0, y\in (0,1), $$


$$ w(y)< 0\quad \text{for } \theta \geq \frac{1}{2}, \delta >0, y\in (0,1). $$

From (2.11) a direct computation immediately gives

$$ \begin{aligned}[b] w(y) &=y\delta ^{k} \frac{(1-y)^{k}}{1-e^{-\delta (1-y)}} \int _{0}^{ \delta (1+y)}\frac{s^{m} e^{-\theta s}}{1-e^{-s}}\frac{(\delta (1+y)-s)^{n}}{1-e ^{-(\delta (1+y)-s)}} \,ds \\ &\quad {}-y\delta ^{k}\frac{(1+y)^{k}}{1-e^{-\delta (1+y)}} \int _{0}^{\delta (1-y)}\frac{s ^{m} e^{-\theta s}}{1-e^{-s}}\frac{(\delta (1-y)-s)^{n}}{1-e^{-( \delta (1-y)-s)}} \,ds. \end{aligned} $$

Substituting \(s=\delta (1+y)x\) into the first integral and \(s=\delta (1-y)x\) into the second, we obtain

$$ \frac{w(y)}{y\delta ^{2k+1}(1-y^{2})^{k}}= \int _{0}^{1} x^{m}(1-x)^{n} \varphi (x;\theta ) \,dx, $$


$$ \begin{aligned}[b] \varphi (x;\theta ) &=\frac{1+y}{1-e^{-\delta (1-y)}} \frac{1}{1-e^{- \delta (1+y)x}} \frac{e^{-\theta \delta (1+y)x}}{1-e^{-\delta (1+y)(1-x)}} \\ &\quad{}-\frac{1-y}{1-e^{-\delta (1+y)}}\frac{1}{1-e^{-\delta (1-y)x}}\frac{e ^{-\theta \delta (1-y)x}}{1-e^{-\delta (1-y)(1-x)}}. \end{aligned} $$

It is sufficient to show that

$$ \varphi (x;\theta )>0 \quad \text{for } \theta \leq 0, \delta >0, x, y \in (0,1), $$


$$ \varphi (x;\theta )< 0\quad \text{for } \theta \geq \frac{1}{2}, \delta >0, x, y\in (0,1), $$

which imply that (2.13) and (2.14) are valid.

Firstly, we assume that \(\theta =0\). Clearly, (2.18) is equivalent to

$$ \phi (x)>0 \quad \text{for } \delta >0, x,y\in (0,1), $$


$$ \phi (x)=\frac{(1-e^{-\delta (1-y)(1-x)})(1-e^{-\delta (1-y)x})}{(1-e ^{-\delta (1-y)})(1-y)}-\frac{(1-e^{-\delta (1+y)(1-x)})(1-e^{-\delta (1+y)x})}{(1-e^{-\delta (1+y)})(1+y)}. $$

Differentiating \(\phi (x)\) with respect to x gives

$$ \phi '(x)=\delta \biggl(\frac{e^{-\delta (1-y)x}-e^{-\delta (1-y)(1-x)}}{1-e ^{-\delta (1-y)}}-\frac{e^{-\delta (1+y)x}-e^{-\delta (1+y)(1-x)}}{1-e ^{-\delta (1+y)}} \biggr). $$

Using Lemma 1, we get \(\phi '(x)>0\) for \(x\in (0,\frac{1}{2})\). Since \(\phi (0)=0\), we conclude that \(\phi (x)>0\) for \(x\in (0, \frac{1}{2})\). In combination with the fact that \(\phi (x)\) is symmetric about \(x=\frac{1}{2}\) on \((0,1)\), this establishes (2.20).

We suppose that \(\theta <0\). It is obvious that

$$ e^{-\theta \delta (1+y)x}>e^{-\theta \delta (1-y)x} \quad \text{for } \delta >0, x,y\in (0,1). $$

From (2.20) we see that

$$ \begin{aligned}[b] &\frac{1+y}{1-e^{-\delta (1-y)}}\frac{1}{1-e^{-\delta (1+y)x}}\frac{1}{1-e ^{-\delta (1+y)(1-x)}}\\ &\quad > \frac{1-y}{1-e^{-\delta (1+y)}}\frac{1}{1-e ^{-\delta (1-y)x}}\frac{1}{1-e^{-\delta (1-y)(1-x)}}. \end{aligned} $$

Hence, taking into consideration (2.21) and (2.22), we complete the proof of (2.18).

Next, we consider the case \(\theta \geq \frac{1}{2}\). Then (2.19) can be equivalently written as

$$ \omega (x)< \omega (0), $$


$$ \omega (x)= \frac{1-e^{-\delta (1-y) (1-x)}}{1-e^{-\delta (1+y) (1-x)}}\frac{1-e ^{-\delta (1-y) x}}{1-e^{-\delta (1+y) x}}\frac{e^{\theta \delta (1-y) x}}{e^{\theta \delta (1+y) x}}. $$

From part (2) of Lemma 2 it is easy to prove (2.23). Therefore the proof of (2.19) is complete.

The proof of Lemma 3 is complete. □

Using Lemma 3 and limit relations (1.6) and (1.7), we have the following:

Corollary 2

Let m, \(n\in \mathbb{N}\), \(\alpha \geq 0\) and \(\beta \geq \frac{1}{2}\). Then, for all \(x>0\), we have:

$$ 0< \frac{\psi _{m+n}(x+\alpha )}{\psi _{m}(x)\psi _{n}(x+\alpha )}< (m+n-1) (m+n-2)_{n-1}< \frac{ \psi _{m+n}(x)}{\psi _{m}(x+\beta )\psi _{n}(x)}< +\infty , $$

where \((m+n-2)_{n-1}=\frac{(m+n-2)!}{(n-1)! (m-1 )!}\). All bounds are optimal.

Remark 2

Setting \(m=n=1\) in Corollary 2, we get the well-known inequality

$$ \psi _{1}^{2}(x)-\psi _{2}(x)>0 $$

with the optimal lower bound. This inequality was proved in many papers by different methods (see [8, 12, 15, 16, 34]) and used to establish many significant results related to the gamma and polygamma functions in [12, 13].

Proofs of the main results

We give a proof of Theorem 1.1.


Differentiation yields

$$ F'(x;\alpha ,\beta )=\psi _{1}(x+\beta ) \biggl(\alpha -\frac{\psi _{n+1}(x)}{ \psi _{1}(x+\beta )\psi _{n}(x)} \biggr)=\psi _{1}(x+\beta )F_{\alpha , \beta }(x), $$

and therefore

$$ F''(x;\alpha ,\beta )=F_{\alpha ,\beta }'(x)\psi _{1}(x+\beta )-F_{ \alpha ,\beta }(x)\psi _{2}(x+\beta ). $$

From Lemma 3 we conclude that \(F_{\alpha ,\beta }'(x)>0\) for \(\beta \geq \frac{1}{2}\) and \(F_{\alpha ,\beta }'(x)<0\) for \(\beta \leq 0\). In view of Corollary 2, it follows that \(F_{\alpha ,\beta }(x)<0\) for \(\beta \geq \frac{1}{2}\) and \(\alpha \leq n\) and \(F_{\alpha ,\beta }(x)>0\) for \(\beta \leq 0\) and \(\alpha \geq n\). By relations (3.1) and (3.2) it is easy to see that \(F'(x;\alpha ,\beta )<0\) and \(F''(x;\alpha ,\beta )>0\) for \(\beta \geq \frac{1}{2}\) and \(\alpha \leq n\) and that \(F'(x;\alpha , \beta )>0\) and \(F''(x;\alpha ,\beta )<0\) for \(\beta \leq 0\) and \(\alpha \geq n\).

Next, we assume that \(F(x;\alpha ,\beta )\) is strictly increasing and concave for \(\beta \leq 0\). From (3.1) we obtain \(F_{\alpha , \beta }(x)>0\), so that Lemma 3 and \(\lim_{x\rightarrow \infty }F_{\alpha ,\beta }(x)=\alpha -n\) imply \(\alpha \geq n\).

Besides, from (3.2) it follows that

$$ \alpha -\frac{\psi _{n+2}(x)\psi _{n}(x)-\psi _{n+1}^{2}(x)}{\psi _{n} ^{2}(x)\psi _{2}(x+\beta )}>0. $$

Letting x tend to ∞ and utilizing (1.7), we get \(\alpha \geq n\).

By similar arguments the necessary condition of the case where \(F(x;\alpha ,\beta )\) is strictly decreasing and convex for \(\beta \geq \frac{1}{2}\) can be proved.

The proof of Theorem 1.1 is complete. □


of Theorem  1.2 Using the concavity and convexity in Theorem 1.1 and Jensen’s inequality, we can easily prove the sufficiency of the theorem. Assume that the left-hand side of (1.12) is valid and \(x_{m}\) are not all equal. Setting \(x_{2}=\cdots =x_{n}=y\), from (1.12) we get

$$ \alpha _{1}\leq \frac{p_{1}\ln \psi _{k}(x_{1})+(1-p_{1})\ln \psi _{k}(y)- \ln \psi _{k}(p_{1}x_{1}+(1-p_{1})y)}{\psi (p_{1}x_{1}+(1-p_{1})y+\beta _{1})-p_{1} \psi (x_{1}+\beta _{1})-(1-p_{1})\psi (y+\beta _{1})}. $$

Letting \(x_{1}\) tend to y and applying L’Hospital’s rule, we obtain

$$ \alpha _{1}\leq \frac{(p_{1}-1)\psi _{k+1}^{2}(y)-(p_{1}-1)\psi _{k}(y) \psi _{k+2}(y)}{(1-p_{1})\psi _{2}(y+\beta _{1})\psi _{k}^{2}(y)}, $$

so that (1.7) and (3.5) lead to \(\alpha _{1}\leq k\) as y tends to ∞. If the right-hand side of (1.12) is valid, then by letting y tend to ∞ in the reversed inequality of (3.5), we have \(\alpha _{2}\geq k\).

The proof of Theorem 1.2 is complete. □

Next we give a proof of Theorem 1.3.

Proof of Theorem 1.3

Firstly, we need to show that \(M(s)\) for \(s<1\) is well defined. It follows from (1.6) and (1.7) that \(\lim_{x\rightarrow \infty }f_{k+1}(x)-f_{k}(x)+sf _{k}(x)=0\) and \(\lim_{x\rightarrow 0}f_{k+1}(x)-f_{k}(x)+{sf_{k}(x)=s}\). According to [32, Lemma 2], it is obvious that \(0\leq M(s)<1\) for \(s<0\) and \(s\leq M(s)<1\) for \(0\leq s<1\).

Since the power mean is increasing with respect to its order (see [33, p. 159]) and \(\psi _{k}(x)\) is decreasing, it suffices to show that (1.13) holds for \(s\geq 1\) and \(r=-s(k+1)\) and for \(s<1\) and \(r=-sk-M(s)\).

In fact, it is still necessary to make use of the techniques of the proof of [32, Thm. 2], and we cannot present it in detail. In view of the proof of [32, Thm. 2], we will prove that \(\varPhi (x)\) is increasing on \((0,\infty )\) for \(s\geq 1\) and \(r=-s(k+1)\) and for \(0\neq s<1\) and \(r=-sk-M(s)\), where

$$ \varPhi (x)=x^{1-r}\psi _{k}^{s-1}(x)\psi _{k+1}(x). $$

Furthermore, we get

$$ \varPhi '(x)=\frac{\varPhi (x)}{x}\varPsi (x), $$


$$ \varPsi (x)=1-r-x\frac{\psi _{k+2}(x)}{\psi _{k+1}(x)}+(1-s)x\frac{\psi _{k+1}(x)}{\psi _{k}(x)}. $$

Case 1. \(s\geq 1\) and \(r=-s(k+1)\). From [32, Lemma 2] we conclude that \(\varPsi (x)\) is strictly increasing on \((0,\infty )\), so that \(\varPsi (x)>\lim_{x\rightarrow 0^{+}}\varPsi (x)=0\). Hence \(\varPhi (x)\) is increasing on \((0,\infty )\).

Case 2. \(0\neq s<1\) and \(r=-sk-M(s)\). Then \(\varPsi (x)\geq 0\) is equivalent to

$$ M(s)\geq f_{k+1}(x)+sf_{k}(x)-f_{k}(x), $$

and therefore \(\varPhi (x)\) is increasing on \((0,\infty )\).

We consider the case \(s=0\). From Theorem 2 of [32] it follows that \(r<0\). Let \(x_{1}\geq x_{2}\geq \cdots \geq x_{n}> 0\). Then we define

$$ F_{m}(x)=F(x,\ldots ,x,x_{m+1},\ldots ,x_{n}), \quad m=1,2\ldots ,n-1, $$


$$ F(x_{1},\ldots ,x_{n})=\sum_{m=1}^{n} p_{m} \ln \psi _{k}(x_{m})- \ln \psi _{k} \Biggl( \Biggl(\sum_{m=1}^{n} p_{m} x_{m}^{r} \Biggr)^{1/r} \Biggr). $$

Differentiating \(F_{m}(x)\) gives

$$ x^{1-r}F'_{m}(x)=\Biggl(\sum _{v=1}^{m} p_{v}\Biggr) \biggl((xz)^{1-r}\frac{\psi _{k+1}(xz)}{\psi _{k}(xz)}-x^{1-r}\frac{\psi _{k+1}(x)}{\psi _{k}(x)} \biggr), $$


$$ 0< z= \Biggl(\sum_{v=1}^{m} p_{v}+\sum_{v=m+1}^{n} p_{v}\biggl( \frac{x_{v}}{x}\biggr)^{r} \Biggr)^{1/r}\leq 1. $$

If \(r=-M(0)\), then we conclude that

$$ \biggl(x^{1-r}\frac{\psi _{k+1}(x)}{\psi _{k}(x)}\biggr)'\geq 0, $$

which, together with (3.8), implies \(F'_{m}(x)\leq 0\). By the techniques of the proof in [32, Thm. 2] we have \(F(x_{1}, \ldots ,x_{n})\leq 0\).

We assume that (1.13) is valid for all \(x_{m}>0\) and weights \(p_{m}\) (\(m=1,2,\ldots, n\)). We set \(x_{2}=\cdots =x_{n}=y\) and \(x_{1}=x\) in (1.13). Then, for \(s\neq 0\), we obtain

$$ f(x,y)=\psi _{k} \bigl(\bigl(p_{1}x^{r}+(1-p_{1})y^{r} \bigr)^{1/r} \bigr)- \bigl(p _{1}\psi _{k}^{s}(y)+(1-p_{1}) \psi _{k}^{s}(y) \bigr)^{1/s}\geq 0 $$

with \(f(y,y)=0\). In fact, a direct computation of the partial derivative of \(f(x,y)\) gives

$$ \frac{\partial f(x,y)}{\partial x}\bigg|_{x=y}=0. $$

In combination with \(f(y,y)=0\), this establishes

$$ \frac{\partial ^{2} f(x,y)}{\partial x^{2}}\bigg|_{x=y}\geq 0, $$

which implies that \(\varPsi (y)\geq 0\) is equivalent to

$$ -r\geq f_{k+1}(y)-f_{k}(y)+sy \frac{\psi _{k+1}(y)}{\psi _{k}(y)}. $$

If \(s\geq 1\), then letting y tend to 0 in (3.9) yields \(r\leq -s(k+1)\). For \(0\neq s<1\), from the definition of \(M(s)\) and (3.9) it follows that \(r\leq -sk-M(s)\). Finally, we prove the case \(s=0\). In the same way, from (1.13) we obtain

$$ f(x,y)=\psi _{k}^{p_{1}}(x)\psi _{k}^{1-p_{1}}(y)- \psi _{k} \bigl(\bigl(p_{1}x ^{r}+(1-p_{1})y^{r} \bigr)^{1/r} \bigr)\leq 0. $$

Since \(f(y,y)=0\) together with

$$ \frac{\partial f(x,y)}{\partial x}\bigg|_{x=y}=0, $$

we obtain

$$ \frac{\partial ^{2} f(x,y)}{\partial x^{2}}\bigg|_{x=y}=-r-f _{k+1}(y)+f_{k}(y)\geq 0. $$

Hence (3.10), together with the definition of \(M(s)\), leads to \(r\leq -M(0)\).

The proof of Theorem 1.3 is complete. □

Remark 3


$$ m(s)=\inf_{x>0} \bigl(f_{k+1}(x)-f_{k}(x)+sf_{k}(x) \bigr), $$

where \(f_{k}(x)\) is defined in Theorem 1.3. From [32, Lemma 2] we get

$$ \biggl(\ln \biggl(x\frac{\psi _{k+1}(x)}{\psi _{k}(x)} \biggr) \biggr)'< 0, $$

which implies \(f_{k+1}(x)-f_{k}(x)>0\). For \(s\geq 0\), it is easy to see that

$$ \bigl(f_{k+1}(x)-f_{k}(x)+sf_{k}(x) \bigr)>0. $$

Since \(\lim_{x\rightarrow \infty }f_{k+1}(x)-f_{k}(x)+sf_{k}(x)=0\), we conclude that \(m(s)=0\). For \(s<0\), from [32, Lemma 2] and (1.6) it follows that

$$ f_{k+1}(x)-f_{k}(x)+sf_{k}(x)>s, $$

which, together with \(\lim_{x\rightarrow 0}f_{k+1}(x)-f_{k}(x)+sf_{k}(x)=s\), implies \(m(s)=s\). As a consequence, the reversed inequality of (1.13) is valid for all \(x_{m}>0\) and weights \(p_{m}\) (\(m=1,2,\ldots, n\)) if and only if \(r\geq -sk-m(s)\).


In this section, we use Theorem 1.1 to yield some new inequalities for the polygamma functions.

We recall that a function f is said to be superadditive on an interval I if

$$ f(x)+f(y)\leq f(x+y)\quad \text{for all } x,y\in I \text{ with } x+y \in I. $$

If −f is superadditive, then f is called subadditive on I (see [36]). A function f is said to be star-shaped on \((0,\infty )\) if

$$ f(ax)\leq af(x) $$

is valid for all \(x>0\) and \(a\in (0,1)\). There is a close relation between the above two notions that star-shaped functions are superadditive (see [37,38,39]). Particularly, Trimble et al. [39] show that a function \(f(x)\) is star-shaped on \((0,\infty )\) if and only if \(f(x)/x\) is nondecreasing on \((0,\infty )\).

Theorem 4.1

For \(n\in \mathbb{N}\) and \(\beta \in \mathbb{R}\), let \(G(x)=e^{n \psi (x+\beta )}\psi _{n}(x)\).

  1. (1)

    For \(\beta \leq 0\), we have the inequality

    $$ \frac{1}{(n-1)!}< \frac{G(x+y)}{G(x)G(y)}< +\infty $$

    for \(x,y>-\beta \).

  2. (2)

    For \(\beta \geq \frac{1}{2}\), we have the inequality

    $$ 0< \frac{G(x+y)}{G(x)G(y)}< \frac{1}{(n-1)!} $$

    for \(x,y>0\). All bounds are optimal.


Since the proof of (4.1) is similar to that of (4.2), we provide the proof of (4.1). We define

$$ g(x,y)=\frac{G(x+y)}{G(x)}. $$

Partial differentiation yields

$$ \frac{\partial g(x,y)}{\partial x}=\frac{G(x+y)}{G(x)} \bigl(g_{1}(x+y)-g _{1}(x) \bigr), $$

where \(g_{1}(x)=G'(x)/G(x)\). For \(\beta \leq 0\), due to Theorem 1.1, we conclude that \(g_{1}(x)\) is decreasing on \((-\beta , \infty )\), so that \(\partial g(x,y)/\partial x<0\). By the double inequality

$$ \ln x-\frac{1}{x}< \psi (x)< \ln x-\frac{1}{2x}, $$

which can be derived from [40], we get \(\lim_{x\rightarrow \infty }G(x)=(n-1)!\) for all \(\beta \in \mathbb{R}\), \(\lim_{x\rightarrow 0}G(x)=\infty \) for \(\beta >0\), and \(\lim_{x\rightarrow -\beta }G(x)=0\) for \(\beta \leq 0\). Together with \(\partial g(x,y)/\partial x<0\) and Theorem 1.1, this establishes

$$ \frac{G(x)}{(n-1)!}< 1< g(x,y)< +\infty , $$

from which inequality (4.1) easily follows. Finally, we check that the bounds of (4.1) are optimal. From

$$ \lim_{x\rightarrow \infty }\lim_{y\rightarrow \infty } \frac{G(x+y)}{G(x)G(y)}= \frac{1}{(n-1)!} $$


$$ \lim_{x\rightarrow -\beta }\lim_{y\rightarrow \infty } \frac{G(x+y)}{G(x)G(y)}=+ \infty $$

we get the desired result. □

Theorem 4.1 immediately leads to the upper and lower bounds for the ratio \(\psi _{n}(x)\psi _{n}(y)/ \psi _{n}(x+y)\).

Corollary 3

Let \(n\in \mathbb{N}\). We have the double inequality

$$ \biggl(\frac{e^{\psi (x+y+1/2)}}{e^{\psi (x+1/2)+\psi (y+1/2)}} \biggr) ^{n}< \frac{\psi _{n}(x)\psi _{n}(y)}{(n-1)!\psi _{n}(x+y)}< \biggl( \frac{e ^{\psi (x+y)}}{e^{\psi (x)+\psi (y)}} \biggr)^{n} $$

for \(x, y>0\).

From Theorem 4.1 we observe that, for \(\beta \leq 0\), the function \(\xi _{1}(x)=\ln (e^{n\psi (x)}\psi _{n}(x-\beta ))-\ln (n-1)!\) is superadditive on \((0,\infty )\), and for \(\beta \geq \frac{1}{2}\), the function \(\xi _{2}(x)=\ln (e^{n\psi (x+\beta )}\psi _{n}(x))-\ln (n-1)!\) is subadditive on \((0,\infty )\).

In fact, \(\xi _{1}(x)\) and \(-\xi _{2}(x)\) are not only superadditive but also star-shaped on \((0,\infty )\).

Theorem 4.2

The functions \(\xi _{1}(x)\) and \(-\xi _{2}(x)\) are both star-shaped on \((0,\infty )\).


Since a function \(f(x)\) is star-shaped on \((0,\infty )\) if and only if \(f(x)/x\) is nondecreasing on \((0,\infty )\) (see [39]), it suffices to prove that \(\xi _{1}(x)/x\) and \(-\xi _{2}(x)/x\) is increasing on \((0,\infty )\). Since

$$ \biggl(\frac{\xi _{1}(x)}{x} \biggr)'=\frac{x\xi '_{1}(x)-\xi _{1}(x)}{x ^{2}}, \qquad \biggl(-\frac{\xi _{2}(x)}{x} \biggr)'=\frac{-x\xi '_{2}(x)+\xi _{2}(x)}{x ^{2}}, $$

by Theorem 1.1 we complete the proof of the theorem. □

Remark 4

Theorems 4.1 and 4.2 provide two distinct ways to prove that \(\xi _{1}(x)\) and \(-\xi _{2}(x)\) are superadditive.


In this paper, Theorem 1.1 gives necessary and sufficient conditions for \(\alpha \in \mathbb{R}\) such that the function \(F(x;\alpha ,\beta )\) is strictly increasing (decreasing) and concave (convex) on \((\max (0,-\beta ),\infty )\) for \(\beta \in (-\infty ,0]\) (\(\beta \in [1/2,\infty )\)). Consequently, Theorem 1.1 provides a alternative proof of [34, Thm. 2.1] and yields a double mean-value inequality (Theorem 1.2). In Theorem 1.3, it is stated that there exist necessary and sufficient conditions such that the reversed inequality of (1.10) is valid, which enhances Theorem 2 of [32].

Moreover, we find several new inequalities for the polygamma functions. In Theorem 4.1, we prove two sharp inequalities for the function \(G(x)\), which determine the upper and lower bounds for the ratio \(\psi _{n}(x)\psi _{n}(y)/\psi _{n}(x+y)\) (Corollary 3). Finally, two star-shaped functions are posed in Theorem 4.2.


  1. 1.

    Abramowitz, M., Stegun, I.A.: Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables. National Bureau of Standards Applied Mathematics Series. Dover, New York (1972)

    Google Scholar 

  2. 2.

    Fichtenholz, G.M.: Differential- und Integralrechnung, II. VEB, Berlin (1979)

    Google Scholar 

  3. 3.

    Magnus, W., Oberhettinger, F., Soni, R.P.: Formulas and Theorems for the Special Functions of Mathematical Physics. Springer, Berlin (1966)

    Google Scholar 

  4. 4.

    Luke, Y.L.: The Special Functions and Their Approximations. Vol. II. Mathematics in Science and Engineering. Academic Press, New York (1969)

    Google Scholar 

  5. 5.

    Alzer, H., Wells, J.: Inequalities for the polygamma functions. SIAM J. Math. Anal. 29(6), 1459–1466 (1998)

    MathSciNet  MATH  Google Scholar 

  6. 6.

    Alzer, H.: On some inequalities for the gamma and psi functions. Math. Comput. 66(217), 373–389 (1997)

    MathSciNet  MATH  Google Scholar 

  7. 7.

    Alzer, H.: A power mean inequality for the gamma function. Monatshefte Math. 131(3), 179–188 (2000)

    MathSciNet  MATH  Google Scholar 

  8. 8.

    Alzer, H.: Sharp inequalities for the digamma and polygamma functions. Forum Math. 16(2), 181–221 (2004)

    MathSciNet  MATH  Google Scholar 

  9. 9.

    Yang, Z.-H., Zheng, S.-Z.: Monotonicity of a mean related to polygamma functions with an application. J. Inequal. Appl. 2016(1), 216 (2016)

    MathSciNet  MATH  Google Scholar 

  10. 10.

    Qi, F., Li, X.-A., Chen, S.-X.: Refinements, extensions and generalizations of the second Kershaw’s double inequality. Math. Inequal. Appl. 11(3), 457–465 (2007)

    MathSciNet  MATH  Google Scholar 

  11. 11.

    Elezovic, N., Pecaric, J.: Differential and integral f-means and applications to digamma function. Math. Inequal. Appl. 3(2), 189–196 (2000)

    MathSciNet  MATH  Google Scholar 

  12. 12.

    Batir, N.: Some new inequalities for gamma and polygamma functions. J. Inequal. Pure Appl. Math. 6(4), 103 (2005)

    MathSciNet  MATH  Google Scholar 

  13. 13.

    Batir, N.: An interesting double inequality for Euler’s gamma function. J. Inequal. Pure Appl. Math. 5(4), 97 (2013)

    MathSciNet  MATH  Google Scholar 

  14. 14.

    Qi, F., Guo, S., Guo, B.-N.: Complete monotonicity of some functions involving polygamma functions. J. Comput. Appl. Math. 233(9), 2149–2160 (2010)

    MathSciNet  MATH  Google Scholar 

  15. 15.

    Qi, F., Guo, B.-N.: Necessary and sufficient conditions for functions involving the tri- and tetra-gamma functions to be completely monotonic. Adv. Appl. Math. 44(1), 71–83 (2010)

    MathSciNet  MATH  Google Scholar 

  16. 16.

    Guo, B.-N., Qi, F., Srivastava, H.M.: Some uniqueness results for the non-trivially complete monotonicity of a class of functions involving the polygamma and related functions. Integral Transforms Spec. Funct. 21(11), 849–858 (2010)

    MathSciNet  MATH  Google Scholar 

  17. 17.

    Koumandos, S.: Monotonicity of some functions involving the gamma and psi functions. Math. Comput. 77(264), 2261–2275 (2008)

    MathSciNet  MATH  Google Scholar 

  18. 18.

    Koumandos, S., Lamprecht, M.: Some completely monotonic functions of positive order. Math. Comput. 79(271), 1697–1707 (2010)

    MathSciNet  MATH  Google Scholar 

  19. 19.

    Yang, Z.-H.: Some properties of the divided difference of psi and polygamma functions. J. Math. Anal. Appl. 455(1), 761–777 (2017)

    MathSciNet  MATH  Google Scholar 

  20. 20.

    Li, A.-J., Chen, C.-P.: Some completely monotonic functions involving the gamma and polygamma functions. J. Korean Math. Soc. 45(1), 273–287 (2008)

    MathSciNet  MATH  Google Scholar 

  21. 21.

    Mortici, C.: Very accurate estimates of the polygamma functions. Asymptot. Anal. 68(3), 125–134 (2010)

    MathSciNet  MATH  Google Scholar 

  22. 22.

    Guo, B.-N., Qi, F.: Some properties of the psi and polygamma functions. Hacet. J. Math. Stat. 39(2), 219–231 (2010)

    MathSciNet  MATH  Google Scholar 

  23. 23.

    Guo, B.-N., Qi, F.: Refinements of lower bounds for polygamma functions. Proc. Am. Math. Soc. 141(3), 1007–1015 (2013)

    MathSciNet  MATH  Google Scholar 

  24. 24.

    Guo, B.-N., Qi, F., Zhao, J.-L., Luo, Q.-M.: Sharp inequalities for polygamma functions. Math. Slovaca 65(1), 103–120 (2015)

    MathSciNet  MATH  Google Scholar 

  25. 25.

    Yang, Z.-H., Zheng, S.-Z.: Monotonicity of a mean related to psi and polygamma functions with an application. J. Inequal. Appl. 2016(1), 216 (2016)

    MathSciNet  MATH  Google Scholar 

  26. 26.

    Yang, Z., Tian, J.-F.: A comparison theorem for two divided differences and applications to special functions. J. Math. Anal. Appl. 464(1), 580–595 (2018)

    MathSciNet  MATH  Google Scholar 

  27. 27.

    Yang, Z.-H., Tian, J.-F.: A class of completely mixed monotonic functions involving gamma function with applications. Proc. Am. Math. Soc. 146(11), 4707–4721 (2018)

    MathSciNet  MATH  Google Scholar 

  28. 28.

    Yang, Z.-H., Tian, J.-F.: Monotonicity rules for the ratio of two Laplace transforms with applications. J. Math. Anal. Appl. 470(2), 821–845 (2019)

    MathSciNet  MATH  Google Scholar 

  29. 29.

    Widder, D.V.: The Laplace Transform. Princeton University Press, Princeton (1946)

    Google Scholar 

  30. 30.

    Bernstein, S.: Sur les fonctions absolument monotones. Acta Math. 52(1), 1–66 (1929)

    MathSciNet  MATH  Google Scholar 

  31. 31.

    Artin, E.: The Gamma Function. Holt, Rinehart & Winston New York (1964)

    Google Scholar 

  32. 32.

    Alzer, H.: Mean-value inequalities for the polygamma functions. Aequ. Math. 61(1–2), 151–161 (2001)

    MathSciNet  MATH  Google Scholar 

  33. 33.

    Bullen, P.S.: Handbook of Means and Their Inequalities, Mathematics and Its Applications. Kluwer Academic, Dordrecht (2003)

    Google Scholar 

  34. 34.

    Batir, N.: On some properties of digamma and polygamma functions. J. Math. Anal. Appl. 328(1), 452–465 (2007)

    MathSciNet  MATH  Google Scholar 

  35. 35.

    Guo, B.-N., Qi, F.: Properties and applications of a function involving exponential functions. Commun. Pure Appl. Anal. 8(4), 1231–1249 (2009)

    MathSciNet  MATH  Google Scholar 

  36. 36.

    Rosenbaum, R.A.: Sub-additive functions. Duke Math. J. 17(3), 227–247 (1950)

    MathSciNet  MATH  Google Scholar 

  37. 37.

    Bruckner, A.M., Ostrow, E.: Some function classes related to the class of convex functions. Pac. J. Math. 12(4), 1203–1215 (1962)

    MathSciNet  MATH  Google Scholar 

  38. 38.

    Hardy, G.H., Littlewood, J.E., Polya, G.: Inequalities. Cambridge University Press, London (1952)

    Google Scholar 

  39. 39.

    Trimble, S.Y., Wells, J., Wright, F.T.: Superadditive functions and a statistical application. SIAM J. Math. Anal. 20(5), 1255–1259 (1989)

    MathSciNet  MATH  Google Scholar 

  40. 40.

    Anderson, G.D., Barnard, R.W., Richards, K.C., Vamanamurthy, M.K., Vuorinen, M.: Inequalities for zero-balanced hypergeometric functions. Transl. Am. Math. Soc. 347(5), 1713–1723 (1995)

    MathSciNet  MATH  Google Scholar 

Download references


The authors would like to express their sincere thanks to the editors and reviewers for their remarkable comments, suggestions, and ideas that helped to improve this paper. In particular, I sincerely appreciate that Lifei Zheng, as my supervisor, helped and encouraged me during the Master.


We declare that funding is not applicable for our paper.

Author information




All authors contributed to each part of this work equally, and they read and approved the final manuscript.

Corresponding author

Correspondence to Lichun Liang.

Ethics declarations

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (, which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Liang, L., Zhao, B. & Li, A. Some inequalities involving the polygamma functions. J Inequal Appl 2019, 54 (2019).

Download citation


  • Polygamma functions
  • Inequalities
  • Psi function
  • Star-shaped functions
  • Superadditive functions