Abstract
We first establish q-non uniform difference versions of the integral inequalities of Hölder, Cauchy–Schwarz, and Minkowski of classical mathematical analysis and then integral inequalities of Grönwall and Bernoulli based on the Lagrange method of linear q-non uniform difference equations of first order. Finally, we prove the Lyapunov inequality for the solutions of the q-non uniform Sturm–Liouville equation.
Similar content being viewed by others
1 Introduction
Considering the most general divided difference derivative [1, 2],
admitting the property that if \(f(t)=P_{n}(t(s))\) is a polynomial of degree n in \(t(s)\), then \(\mathcal{D} f (t(s))=\tilde{P}_{n-1}(t(s))\) is a polynomial in \(t(s)\) of degree \(n-1\), one is led to the following most important canonical forms for \(t(s)\) in order of increasing complexity:
When the function \(t(s)\) is given by (2)–(4), the divided difference derivative (1) leads to the ordinary differential derivative \(D f (t)=\frac{d}{dt}f(t)\), finite difference derivative
and q-difference derivative (or Jackson derivative [3])
respectively (see also [4,5,6]). When \(t(s)=x(q^{s})\) is given by (5), the corresponding derivative is usually referred to as the Askey–Wilson first order divided difference operator [7] that one can write as follows:
where \(x(z)=\frac{z+z^{-1}}{2}\) is the well-known Joukowski transformation and \(z=q^{s}\).
The calculus related to the differential derivative, the continuous or differential calculus, is clearly the classical one. The one related to derivatives (6),(7), (8) (difference, q-difference, and q-non uniform difference, respectively) is referred to as the discrete calculus. Its interest is twofold: On the one hand, it generalizes the continuous calculus; on the other hand, it uses a discrete variable.
This work is concerned with the q-non uniform difference calculus. We particularly aim to establish q-non uniform difference versions of the well-known in differential calculus integral inequalities of Hölder, Cauchy–Schwarz, Minkowski, Grönwall, Bernoulli, and Lyapunov. Another captivating work on the raised inequalities can be found in [8] for a calculus based on a derivative also generalizing (6) and (7), but one will remark that even if that work greatly inspired us, there is not any hierarchic relationship between the calculus considered there (see [9] for a general theory) and the one considered here (see also [10,11,12]) based on (8). We will note also that (8) is at our best knowledge the most general known divided difference derivative having the property of sending a polynomial of degree n in a polynomial of degree \(n-1\).
In the following lines, we first introduce basic concepts of q-non uniform difference calculus necessary for the sequel, and then study the mentioned integral inequalities. The functions that are considered in the q-non uniform difference calculus are clearly defined on the set
2 q-Non uniform difference calculus
In this section, we discus the essential elements of q-non uniform integral calculus and q-non uniform linear difference equations of first order.
2.1 q-Non uniform integral calculus
2.1.1 Integration
We consider the q-non uniform divided difference derivative defined by (from now on \(q\in \mathbb{R}^{+}\), \(s\in \mathbb{Z}^{+}\), as indicated below)
A function \(f(x(z))\) is said to be q-non uniform differentiable on \(x(z)\) iff the ratio on the r.h.s. (right-hand side) of (10) exists and is finite. Clearly, every continuous function on \(\mathbb{T}\) (in the topology of \(\mathbb{R}\)) is q-non uniform differentiable on that set.
Let us suppose that \(\mathcal{D}F(x(z))=f(x(z))\); \(z=q^{s}\) or equivalently
We have
then we have
By adding member by member, we get
Hence
This integral sends a polynomial (in \(x(z)\)) of degree n in a polynomial of degree \(n+1\) [12].
Replacing \(x(z)=\frac{z+z^{-1}}{2}\) in this last equation, we have
Let us stop for a moment on the appropriate writing of (14) to verify in particular whether in these equations the integral on the left-hand side has a lower bound, actually lower than the upper bound. Note first that the function \(x(z)=\frac{z+z^{-1}}{2}\), \(z \in \mathbb{R}^{+}\), is decreasing for \(0< z<1\) and increasing for \(1< z< \infty \).
Assume first that \(0< q<1\). In this case, according to the construction \(N\geqslant s \geqslant 0 \), one will have \(q^{N}\leqslant z=q^{s}\) and \(x(z) \leqslant x(q^{N}) \), and the convenient writing of (14) is
On the other hand, if \(1< q<\infty \), we will have \(z\leqslant q^{N}\) and \(x(z) \leqslant x(q^{N})\) and (14) becomes (15) again. On the other side, in (15), the factor
is always positive regardless of whether \(0< q<1\) or \(1< q<\infty \) (let us agree from now on that \(0< q<1\) and therefore \(0 \leqslant z \leqslant 1\)). This leads us to the following fundamental positivity property of the integral in (15).
Property 2.1
If \(f(x(z)) \geqslant 0\) and \(a=x(q^{\alpha }) \leqslant b=x(q^{ \beta })\), then
Corollary 2.1
If \(f(x(z))\geqslant g(x(z))\) and \(a=x(q^{\alpha }) \leqslant b=x(q ^{\beta })\), then
2.1.2 Connection between the q-integral and q-non uniform integral
We have
For \(N \rightarrow \infty \),
and the integral with lower and upper finite bound can be written as follows:
\(a=x(q^{\alpha }) \leqslant b=x(q^{\beta })\) (if the integrals on the r.h.s. of (18) exist).
A function \(f(x(z))\) defined on \(\mathbb{T}\) is said to be q-non uniform integrable on a finite interval \([x(z), x(q^{N})]\) iff the sum on the r.h.s. of (15) exists and is finite. If the upper bound is infinite, the q-non uniform integrability means the convergence of the infinite series on the r.h.s. of (17). Every continuous function on \(\mathbb{T}\) (in the topology of \(\mathbb{R}\)) is clearly q-non uniform integrable on any finite interval on that set.
Remark 2.1
It is not difficult to notice that to deal with the case where \(s\geqslant N \geqslant 0\) it would suffice to replace in the preceding formulas \(z=q^{s}\) by \(q^{N}\) and vice versa.
In this case, from (13), we obtain
or by replacing \(x(z)=\frac{z+z^{-1}}{2}\) in (19)
Passing to the q-integral, we will have
For \(N\rightarrow 0\),
Finally, if \(a=x(q^{\alpha })\leqslant b=x(q^{\beta })\),
2.1.3 Fundamental principles of analysis
-
(i)
We can formulate the statement of the fundamental principle of analysis as follows: “The q-non uniform derivative of the integral of a function is this function itself”. This corresponds to the formula
$$ \mathcal{D} \biggl[ \int ^{x(z)}_{x(q^{N})}f\bigl(x(z)\bigr)\,d_{q}x(z) \biggr] =f\bigl(x(z)\bigr). $$(24) -
(ii)
This is the q-non uniform version of the Newton–Leibnitz formula
$$ \int ^{x(z)}_{x(q^{N})} [\mathcal{D} f ]\bigl(x(z) \bigr)\,d_{q}x(z)=f\bigl(x(z)\bigr)-f\bigl(x\bigl(q ^{N}\bigr) \bigr). $$(25)
2.1.4 Integration by parts
By integrating the formula
and using the second fundamental principle of the analysis, one obtains
2.2 Linear q-non uniform difference equations of first order
The general linear q-non uniform difference equation of first order is given by
or
where \(a(x(z))\), \(b(x(z))\), \(\tilde{a} (x(z))\), and \(\tilde{b} (x(z))\) are known, while \(y(x(z))\) and \(\tilde{y} (x(z))\) are unknown functions to be determined.
Consider first the homogeneous equation corresponding to (28):
Detailing, we get
which gives
where
Using the recursion
we obtain
or
For \(N\longrightarrow \infty \) (or \(z_{0} \longrightarrow 0 \)), we obtain
Let us define the exponential function
It is clear that
and
Similarly, for the homogeneous equation corresponding to (29)
developing, we have
where
Using the recursion
we obtain
or
For \(N\longrightarrow \infty \) (or \(z_{0} \longrightarrow 0 \)), we have
Let us now define the second exponential function
We have
and
From (35) and (44) we obtain that:
If \(\tilde{a}(x(z))=-a(x(z))\), then \(\tilde{p}(x(z))=p(x(z))\) and \(y_{0}(x(z))\tilde{y}_{0}(x(z))=y_{0}(x(z_{0}))\tilde{y}_{0}(x(z _{0}))\). Hence we have the following.
Theorem 2.1
If \(y(x(z))\) and \(\tilde{y}(x(z))\) are solutions of
and
respectively, and satisfy \(y(x(z_{0}))\tilde{y}(x(z_{0}))=1\), then \(y(x(z))\tilde{y}(x(z))=1\).
Direct proof
As well \(y(x(z_{0}))\tilde{y}(x(z_{0}))=1\), then \(y(x(z))\tilde{y}(x(z))=1\). □
Corollary 2.2
Let us consider now the non-homogeneous equations (28) and (29). To find the solution of the non-homogeneous equation
we assume that \(y_{0}(x(z)) \) is the solution of the corresponding homogeneous equation
and use the Lagrange method (variation of constants method) by taking
as the solution of (47), where \(c(x(z)) \) is unknown. By placing (49) in (47), we have
or
Then, since
we get
or
then
This gives us the following relation:
Placing (52) in (49), we obtain
hence
where
We do the same with the equation associated to (28), equation (29), using the Lagrange method with
where \(\tilde{c}(x(z)) \) is an unknown function. Placing (57) in (29), we have
Placing (58) in (57), we obtain
hence
Now we can write the general solution of (29) like
where
3 q-Non uniform difference integral inequalities
In this section, we will first establish q-non uniform versions of some integral inequalities of classical mathematical analysis such as the integral inequalities of Hölder, Cauchy–Schwarz, and Minkowski. It can be seen that the techniques used in classical analysis remain valid here. Then we establish q-non uniform versions of some other integral inequalities based on the linear q-non uniform difference equations of first order and the corresponding Lagrange resolution method: the inequalities of Grönwall, Bernoulli; and finally we prove the Lyapunov inequality for the solutions of the q-non uniform Sturm–Liouville equation.
3.1 q-Non uniform Hölder and Cauchy–Schwarz inequalities
Theorem 3.1
(q-non uniform Hölder inequality)
Let \(a, b \in [1, \infty [\,\cap\, \mathbb{T}\). For all real-valued functions f, g, defined and q-non uniform integrable on \([a,b]\), we have
with \(\alpha >1 \) and \(\frac{1}{\alpha } +\frac{1}{\beta }=1\).
Proof
Let us first show that for \(A, B \in [1, \infty [\) we have
Indeed, since \(\frac{1}{\alpha } +\frac{1}{\beta }=1\), \(A,B \in [1, \infty [\) , \(\frac{A^{\alpha }}{\alpha }+\frac{B^{\beta }}{\beta } \) runs through the segment \([A^{\alpha },B^{\beta }]\), while \(\frac{\log A ^{\alpha }}{\alpha }+\frac{\log B^{\beta }}{\beta } \) runs through the segment linking the points \((A^{\alpha },\log A^{\alpha })\) and \((A^{\beta },\log A^{\beta })\). By concavity of the logarithm function, we conclude that \(\log (\frac{A^{\alpha }}{\alpha }+\frac{B^{ \beta }}{\beta } )\geqslant \frac{\log A^{\alpha }}{\alpha }+\frac{ \log B^{\beta }}{\beta }= \log (AB) \). By applying the exponential to the two members of this inequality, we obtain (64).
Let us now take
considering that
(otherwise, clearly \(f\equiv 0\) or \(g\equiv 0\) and (63) becomes equality). Due to Property 2.1 and its Corollary on the positivity of the integral, by substituting \(A(x(z))\) and \(B(x(z))\) in (64) and integrating on \([a,b] \), we have
which gives us directly the q-non uniform Hölder inequality
□
If we take \(\alpha =\beta =2 \) in the q-non uniform Hölder inequality (63), we have the q-non uniform Cauchy–Schwarz inequality.
Corollary 3.1
(q-non uniform Cauchy–Schwarz inequality)
Let \(a,b \in [1, \infty [\, \cap\, \mathbb{T}\). For all real-valued functions f, g, defined and q-non uniform integrable on \([a, b]\), we have
3.2 q-Non uniform Minkowski inequality
We can now use the q-non uniform Hölder inequality to deduce the q-non uniform Minkowski inequality.
Theorem 3.2
(q-non uniform Minkowski inequality)
Let \(a, b \in [1, \infty [\,\cap\, \mathbb{T}\). For all real-valued functions f, g, defined and q-non uniform integrable on \([a, b]\), we have
with \(\frac{1}{\alpha }+\frac{1}{\beta }=1\), where \(\alpha >1\) and \(\beta > 1\).
Proof
We have
Using the q-non uniform Hölder inequality with \(\beta =\frac{ \alpha }{\alpha -1}\), we will obtain
Dividing both sides of this inequality by \(\lbrace \int _{a}^{b} \vert (f+g)(x(z)) \vert ^{\alpha } \vert \,d_{q}x(z) \rbrace ^{\frac{1}{\beta }} \), we have
As \(1-\frac{1}{\beta }=\frac{1}{\alpha } \), this gives us the q-non uniform Minkowski inequality. □
3.3 q-Non uniform Grönwall inequality
Let us first introduce the following inequalities based on the Lagrange method for the linear q-non uniform difference non-homogeneous equations.
Lemma 3.1
Let y, f be real-valued functions defined and q-non uniform integrable on \([c, d]\), \(\forall c, d\in [1,\infty [\,\cap\, \mathbb{T}\). Let \(a(x(z))\) such that \(p(x(z))= 1+(q^{\frac{1}{2}}-q^{-\frac{1}{2}})(z-z ^{-1})a(x(z))>0\). That means \(a(x(z)) \geqslant 0\).
Suppose that \(y_{0}(x(z))\) is a solution of
such that \(y_{0}(x(z_{0}))=1\).
If
then
Proof
Let \(y_{0}(x(z)) \) be the solution of the homogeneous equation
such that \(y_{0}(x(z_{0}))=1\). By looking for the function \(y (x(z)) \) satisfying (73) by the Lagrange method
where \(c (x(z)) \) is indeterminate, we replace (76) in (73) and we have
or
Using (75), we have
Simplifying this gives
or
since \(a(x(z))\in \mathbb{R}^{+}\) implies \(y_{0}(x(z)) >0\). By integrating the two members of the equality from \(x(z_{0})\) to \(x(z) \), we will have
However, from (76) follows \(c(x(z_{0}))=y(x(z_{0}))\), given that \(y_{0}(x(z_{0}))=1\). Hence
which gives the desired result
□
Taking account of Corollary 2.2 and the fact that by the definition \(E_{a, q^{-\frac{1}{2}}}(x(z_{0});x(z_{0}))=E_{a, q^{ \frac{1}{2}}}(x(z_{0});x(z_{0}))=1\), we obtain the following.
Corollary 3.2
If y, f, and a are functions satisfying the conditions of Lemma 3.1, then
Lemma 3.2
Let y, f be real-valued functions defined and q-non uniform integrable on \([c, d]\), \(\forall c, d \in [1, \infty [\,\cap\, \mathbb{T}\). Let \(a(x(z))\) such that
This means that \(a(x(z)) \leqslant 0\).
Suppose that \(y_{0}(x(z))\) is a solution of
such that \(y_{0}(x(z_{0}))=1\).
In that case, if
then, for all \(x(z) \in [1,\infty [\), we have
Proof
Let \(y_{0}(x(z)) \) be a solution of the homogeneous equation
Looking for the function \(y (x(z)) \) satisfying (87) by the Lagrange method
where \(c(x(z)) \) is indeterminate, we replace the function \(y(x(z))\) in (87) and we have
or
Using (89), we have
Simplifying this gives
or
since \(a(x(z))\leqslant 0\) implies \(y_{0}(x(z)) >0\). By integrating the two members of the equality above from \(x(z_{0})\) to \(x(z) \), we will have
However, from (90) follows \(c(x(z_{0}))=y(x(z_{0}))\), given that \(y_{0}(x(z_{0}))=1\). This gives
Then we obtain
which gives the desired result
□
As well as for Corollary 3.2, we deduce the following.
Corollary 3.3
If y, f, and a satisfy the conditions of Lemma 3.2, then
Theorem 3.3
(q-non uniform Grönwall inequality)
Let y, f be real-valued and q-non uniform integrable functions on \([c, d]\), \(\forall c,d \in [1,\infty [\,\cap\, \mathbb{T}\), and \(a \geqslant 0 \). If
then
Proof
Let us define
Then \(\nu (x(z_{0}))=0\) and \(\mathcal{D} \nu = y(x(zq^{-\frac{1}{2}}))a(x(z))\).
Hence hypothesis (99) gives
and
or
From Lemma 3.1, inequality (104) gives
and inequality (102) implies that (with \(\nu (x(z_{0}))=0\))
which is the q-non uniform Grönwall inequality. □
As a direct consequence, we get the following results.
Corollary 3.4
Let y, f be real-valued functions defined and q-non uniform integrable on \([c, d]\), \(\forall c,d\in [1, \infty [\,\cap\, \mathbb{T}\) and \(a(x(z))\) such that \(1+(q^{\frac{1}{2}}-q^{-\frac{1}{2}})(z-z^{-1})a(x(z))>0\). That means \(a(x(z)) \geqslant 0\). Then
for all \(x(z)\) implies that \(y(x(z))\leqslant 0\).
Proof
This is due to Theorem 3.3 with \(f (x(z))\equiv 0 \) □
Corollary 3.5
Let \(a \geqslant 0\) and \(\alpha \in \mathbb{R} \). If
for all \(x_{0}=x(z_{0}) > 0\), then \(y(x(z))\leqslant \alpha E_{a,q ^{-\frac{1}{2}}}(x_{0},x(z)) \).
Proof
By the q-non uniform Grönwall integral inequality (100), if we take \(f(x(z))=\alpha \), then
□
3.4 q-Non uniform Bernoulli inequality
Theorem 3.4
(q-non uniform Bernoulli inequality)
Let \(\alpha \in \mathbb{R}\). Then \(\forall x(z)\), \(x_{0}=x(z_{0}) \in [1, \infty [\) with \(x(zq^{-\frac{1}{2}}) >x_{0} \), we have
Proof
Let us take \(y(x(z))= \alpha (x(z)-x_{0})\), \(x(zq^{-\frac{1}{2}})>x _{0} \), ∀z. Then \(\mathcal{D}y(x(z))=\alpha \) and we have
which implies that \(\mathcal{D}y(x(z)) \leqslant \alpha y(x(zq^{- \frac{1}{2}}))+\alpha \).
On the other hand, by Lemma 3.1, we get (with \(y(x_{0}) = 0 \))
That is why \(E_{a,q^{-\frac{1}{2}}}(x_{0},x(z))\geqslant 1+y(x(z))= 1+ \alpha (x(z)-x_{0} ) \). □
3.5 q-Non uniform Lyapunov inequality
Let us now turn to q-non uniform Lyapunov inequality regarding the derivative and integral introduced. For that, consider the following q-non uniform Sturm–Liouville equation:
where \(D^{+}\) is defined by \(\mathcal{D}^{+}f(x(z))= \frac{f(x(zq))-f(x(z))}{x(zq)-x(z)} \).
Let us define the function F by
Lemma 3.3
Let u be a non-trivial solution of the q-non uniform Sturm–Liouville equation (110). Then, for all y belonging to the domain of F, the following equality remains verified:
Proof
We have
(using \(f(x(z))u(x(zq^{\frac{1}{2}}))=-\mathcal{D}\mathcal{D}^{+} u(x(z))\)), or
□
Lemma 3.4
Let y be in the domain of F, then \(\forall a, b \in [1, \infty [ \,\cap\, \mathbb{T}\) and \(c, d \in [a,b]\cap \mathbb{T}\) such that \(a \leqslant c \leqslant d \leqslant b \), we have
Proof
Let us take
Then \(\mathcal{D}^{+} u(x(z))= \frac{y(d)-y(c)}{d-c}\) and \(\mathcal{D} \mathcal{D}^{+} u(x(z))=0\), which proves that \(u (x(z))\) is a solution of (110) with \(f (x(z)) = 0 \), \(\forall x(z) \in [1, \infty [\,\cap\, \mathbb{T}\) and
\(\forall y(x(z))\) from the domain of F. From Lemma 3.3, we get
Consequently,
This leads us to the following result:
□
Theorem 3.5
(q-non uniform Lyapunov inequality)
Let f be a real-valued function defined and q-non uniform differentiable on \([a, b]\), \(\forall a, b \in [1,\infty [\,\cap\, \mathbb{T}\), \(a< b\), and let \(u(x(z))\) be a non-trivial solution of equation (110) with \(u(a)=u(b)=0 \), then
Proof
From Lemma 3.3 with \(y=0\) and \(u(a)=u(b)=0\), we have
Let \(M=\max \lbrace u^{2}(x(zq^{\frac{1}{2}})); x(q^{\frac{1}{2}}z) \in [a,b]\cap \mathbb{T}\rbrace \) and \(c \in [a,b]\cap \mathbb{T}\) such that \(u^{2}(c)=M \). Then
and
which implies that \(\int _{a}^{b} f(x(z))\,d_{q}x(z) \geqslant M \frac{4}{b-a}\).
The q-non uniform Lyapunov inequality is thus proved. □
4 Conclusion
In this paper, q-non uniform difference versions of the integral inequalities of Hölder, Cauchy–Schwarz, and Minkowski, and also the integral inequalities of Grönwall and Bernoulli based on the Lagrange method of linear q-non uniform difference equations of first order were established. Finally, the Lyapunov inequality for the solutions of the q-non uniform Sturm–Liouville equation was proved.
References
Magnus, A.P.: Associated Askey–Wilson polynomials and Lageurre–Hahn orthogonal polynomials. In: Lectures Notes in Math., vol. 1329, pp. 261–278. Springer, Berlin (1988)
Magnus, A.P.: Special non uniform lattices (snul) orthogonal polynomials on discrete dense sets of points. J. Comput. Appl. Math. 65, 253–265 (1995)
Jackson, H.F.: q-Difference equations. Am. J. Math. 32, 305–314 (1910)
Mursaleen, M., Ansari, K.J., Khan, A.: Approximation by a Kantorovich type q-Bernstein-Stancu operators. Complex Anal. Oper. Theory 11, 85–107 (2017)
Mursaleen, M., Ansari, K.J., Nasiruzzaman, M.: Approximation by q-analogue of Jakimovski–Leviatan operators involving q-Appell polynomials. Iran. J. Sci. Technol., Trans. A, Sci. 41(4), 891–900 (2017)
Mursaleen, M., Ansari, K.J.: Approximation of q-Stancu-Beta operators which preserve \(x^{2}\). Bull. Malays. Math. Soc. 40(4), 1479–1491 (2017)
Askey, R., Wilson, J.: Some basic hypergeometric orthogonal polynomials that generalize the Jacobi polynomials. Mem. Am. Math. Soc. 54, 1–55 (1985)
Hamza, A.E., Sheheta, E.M.: Some inequalities based on a general quantum difference operator. J. Inequal. Appl. 2015, 38 (2015)
Hamza, A.E., Sarhan, A.S.M., Shehata, E.M., Aldwoah, K.A.: A General Quantum Difference Calculus. Adv. Differ. Equ. 2015, 182 (2015)
Bangerezako, G.: q-Non uniform difference linear control systems. J. Appl. Anal. 20, 133–143 (2014)
Bangerezako, G.: An introduction to q-non uniform difference equations (2011). Preprint, Triestre, Miramare
Bangerezako, G.: Variational calculus on q-non uniform lattices. J. Math. Anal. Appl. 306, 161–179 (2005)
Acknowledgements
The authors acknowledge the referees for noting some misprints and communicating some recent references on q-calculus. JPN acknowledges the Doctoral School of the University of Burundi for support.
Funding
The funding for this article comes from the Doctoral School of the University of Burundi in which the doctoral research of Jean Paul Nuwacu (one of the authors of this work) is conducted and supported.
Author information
Authors and Affiliations
Contributions
All authors contributed equally and significantly in writing this article. All authors read and approved the final manuscript.
Corresponding author
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 (http://creativecommons.org/licenses/by/4.0/), 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.
About this article
Cite this article
Bangerezako, G., Nuwacu, J.P. & Shehata, E.M. q-Non uniform difference calculus and classical integral inequalities. J Inequal Appl 2019, 28 (2019). https://doi.org/10.1186/s13660-019-1983-0
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s13660-019-1983-0