Continue from this paper
Indexing is not verification. Extracted statements should be checked against the source.
Results & Lemmas (7)
Each result is auto-extracted from the paper via OCR - the math is shown verbatim and may carry occasional transcription artifacts. Click to expand the full statement.
Theorem 1.2 · radius
Theorem 1.2. For, c = a + b and C = const. > 1, with, define the functions f and g on (0,1) by respectively. Then we have: (1) For any with…
Theorem 1.2. For $a, b \in (0,1)$ , c = a + b and C = const. > 1, with $c \le 1$ , define the functions f and g on (0,1) by
$$f(r) = (1 + \sqrt{r})F(a, b; c; r) - F(a, b; c; 4\sqrt{r}/(1 + \sqrt{r})^2),$$
$$g(r) = CF(a, b; c; r) - F(a, b; c; 4\sqrt{r}/(1 + \sqrt{r})^2),$$
respectively. Then we have:
(1) For any $a,b \in (0,1)$ with $c \leq 1$ (c < 1, respectively), f is (strictly, respectively) increasing from (0,1) onto $(0,(R-\log 16)/B)$ , where B=B(a,b) and R=R(a,b) are defined by (1.4) and (1.6), respectively. In particular, for $a,b \in (0,1)$ with $c=a+b \leq 1$ , and for all $r \in (0,1)$ ,
(1.8)
$$F(a,b;c;(2\sqrt{r}/(1+r))^2) \le (1+r)F(a,b;c;r^2)$$
$$\le F(a,b;c;(2\sqrt{r}/(1+r))^2) + [(R-\log 16)/B]$$
and
(1.9)
$$\frac{1+r}{2}F(a,b;c;1-r^2) \le F\left(a,b;c;\left(\frac{1-r}{1+r}\right)^2\right) \\ \le \frac{1+r}{2}\left[F(a,b;c;1-r^2) + \frac{1}{B}(R-\log 16)\right],$$
with equality in each instance if and only if a = b = 1/2.
(2) If $C \geq 2$ , then g(r) > 0 for all $r \in (0,1)$ . Moreover, if $1 < C \leq 2$ , then g is strictly decreasing from (0,1) onto $(C_1, C-1)$ , where
$$C_1 = \begin{cases} -\infty, & \text{if } 1 < C < 2, \\ (R - \log 16)/B, & \text{if } C = 2. \end{cases}$$
In particular, for all $r \in (0,1)$ ,
(1.10)
$$F\left(a, b; c; \left(\frac{2\sqrt{r}}{1+r}\right)^{2}\right) - 1 < 2F(a, b; c; r) - 1$$
$$< F\left(a, b; c; \left(\frac{2\sqrt{r}}{1+r}\right)^{2}\right) < 2F(a, b; c; r) - \frac{1}{B}(R - \log 16).$$
(3) If 1 < C < 2, then $F(a,b;c;(2\sqrt{r}/(1+r))^2)$ and $CF(a,b;c;r^2)$ are not directly comparable for all $r \in (0,1)$ , that is, neither the inequality
$$F(a, b; c; (2\sqrt{r}/(1+r))^2) \ge CF(a, b; c; r^2)$$
nor its reversed inequality holds for all $r \in (0,1)$ .
From Theorem 1.2, we see that the first and second identities in (1.7) are the special case of (1.8) and (1.9), respectively, when a = b = 1/2. Parts (2) and (3) of Theorem 1.2 imply that in the Open Problem 1.1 the best value of C, for which $h(2\sqrt{r}/(1+r)) \leq Ch(r)$ , is 2. It should be noted that in (1.8) and (1.9), the hypergeometric functions have the same sets of parameters, in analogue with (1.7) where $\mathcal{K}(r)$ is this function. The so-called quadratic transformation formulas (see e.g. [AS, 15.3.15–15.3.32]) also yield formulas which give (1.7) as a particular case, but those formulas involve hypergeometric functions with two different sets of parameters.
In [AVV1, p. 79], the following problem was put forward. (For the particular case a = b = 1/2 the answer is known to be affirmative.)
OPEN PROBLEM 1.3. Is it true that the function
$$W(r) \equiv F(a, b; a + b; r) + \frac{C}{r}\log(1 - r),$$
where C = 1/B(a, b), is monotone on (0, 1) for suitable a and b?
In [AVV1, Theorem 6], Gauss' asymptotic formula (1.3) was refined by finding the lower and upper bounds for W(r), while in [ABRVV, Theorem 1.4] the function W(r) was shown to be monotone when $a, b \in (0, 1)$ or $a, b \in (1, \infty)$ . Our second result gives a full solution to the Open Problem 1.3.
Theorem 1.4 · radius
Theorem 1.4. For, c = a + b, let, and define the function f on (0, 1) by where B = B(a, b). Then we have: - (1) If, then. - (2) If and,…
Theorem 1.4. For $a, b \in (0, \infty)$ , c = a + b, let $a_1 = 1 - ab, a_2 = 2ab - a - b, a_3 = |a_1| + |a_2|$ , and define the function f on (0, 1) by
$$f(r) = BF(a, b; c; r) + \frac{1}{r}\log(1 - r),$$
where B = B(a, b). Then we have:
- (1) If $a_3 = 0$ , then $f(r) \equiv 0$ .
- (2) If $a_3 \neq 0$ and $a_1 \geq \max\{0, a_2\}$ , then f is strictly increasing from (0, 1) onto (B 1, R), where R = R(a, b).
- (3) If $a_3 \neq 0$ and $a_1 \leq \min\{0, a_2\}$ , then f is strictly decreasing from (0, 1) onto (R, B 1).
- (4) In the other cases not stated in parts (1)–(3), that is, $a_2 < a_1 < 0$ , f is not always monotone on (0,1).
Remark 1.1. It seems that there would be two possible cases not stated in parts (1)–(3) of Theorem 1.4:
$$a_2 < a_1 < 0$$
if $a_2 < 0$ , and, $0 \le a_1 \le a_2$ if $a_2 \ge 0$ .
However, one can show that there are no values of a and b such that $a_2 \ge a_1 \ge 0$ with $a_3 \ne 0$ . In fact, if $a_2 > a_1 = 0$ , then ab = 1 and $a_2 = 2 - (a+b) > 0$ so that $2 > a + b = a + (1/a) \ge 2$ , a contradiction. If $a_2 \ge a_1 > 0$ , then
$$(1.11) a+b-3ab+1 \le 0 \text{ and } ab < 1.$$
Since if $a, b \le 1$ then $a_2 = a(b-1) + b(a-1) \le 0$ and since if $a, b \ge 1$ then $a_1 \le 0$ , we may assume that $0 < a < 1 \le b$ . Thus it follows from (1.11) that $a \ge (b+1)/(3b-1)$ and hence
$$a_1 = 1 - ab \le 1 - \frac{b(b+1)}{3b-1} = -\frac{(b-1)^2}{3b-1} \le 0,$$
a contradiction again.
It is not difficult for us to employ the Monotone l'Hôpital's rule [AVV2, Theorem 1.25] to prove that the function
$$f(r) \equiv [\mathcal{K}(r) - \log(4/r')][(r'/r)^2 \log(1/r')]^{-1}$$
is strictly decreasing from (0,1) onto $(1/4,\pi-\log 16)$ , (see Corollary 2.1). Here and in the sequel, we let $r'=\sqrt{1-r^2}$ for $r\in(0,1)$ . Consider the function
$$g(r) \equiv \frac{B(a,b)F(a,b;a+b;r) + \log(1-r) - R(a,b)}{[(1-r)/r]\log(1/(1-r))}.$$
Then for a = b = 1/2 we have $g(r^2) = f(r)$ . Therefore, it is natural to ask if the function g(r) is monotone on (0,1) for $a,b \in (0,\infty)$ . Our next result answers this question.
Theorem 1.5 · radius
Theorem 1.5. For, let,,, and define the function f on (0, 1) by where B = B(a,b) and R = R(a,b) are defined by (1.4) and (1.6),…
Theorem 1.5. For $a, b \in (0, \infty)$ , let $A_1 = A_1(a, b) = a + b + ab - 3$ , $A_2 = A_2(a, b) = a + b - 3ab + 1$ , $A = |A_1| + |A_2|$ , and define the function f on (0, 1) by
$$f(r) = \frac{BF(a,b;a+b;r) + \log(1-r) - R}{[(1-r)/r]\log[1/(1-r)]},$$
where B = B(a,b) and R = R(a,b) are defined by (1.4) and (1.6), respectively. Then we have the following conclusions:
- (1) If A = 0, then $f(r) \equiv 1$ .
- (2) If $A \neq 0$ and $A_1 \leq \min\{0, A_2\}$ , then f is strictly decreasing from (0, 1) onto (ab, B R). In particular, with this condition, for all $r \in (0, 1)$ ,
(1.12)
$$\begin{cases} ab\frac{1-r}{r}\log\frac{1}{1-r} < BF(a,b;a+b;r) + \log(1-r) - R \\ < (B-R)\frac{1-r}{r}\log\frac{1}{1-r}. \end{cases}$$
(3) If $A \neq 0$ and $A_1 \geq \max\{0, A_2\}$ , then f is strictly increasing from (0, 1) onto (B - R, ab). In particular, under this condition, for all $r \in (0, 1)$ ,
$$(1.13) \begin{cases} (B-R)\frac{1-r}{r}\log\frac{1}{1-r} < BF(a,b;a+b;r) + \log(1-r) - R \\ < ab\frac{1-r}{r}\log\frac{1}{1-r}. \end{cases}$$
(4) In the other cases not stated in parts (1)–(3), namely, $0 < A_1 < A_2$ , f is not always monotone on (0,1) for all a and b with $0 < A_1 < A_2$ .
Theorem 1.5 gives two sided estimates for the O-term in Ramanujan's asymptotic formula (1.4).
Corollary 2.1
Corollary 2.1. The function is strictly decreasing from (0,1) onto. We shall next compare Corollary 2.1 to some earlier bounds for the…
Corollary 2.1. The function
$$f(r) \equiv \frac{\mathcal{K}(r) - \log(4/r')}{(r'/r)^2 \log(1/r')}$$
is strictly decreasing from (0,1) onto $(1/4,\pi-\log 16)$ .
We shall next compare Corollary 2.1 to some earlier bounds for the function $\mathcal{K}(r)$ . For this purpose we need the following proposition.
Proposition 2.2
Proposition 2.2. The function is strictly increasing and convex from (0,1) onto. In particular, for all,
Proposition 2.2. The function $g(r) \equiv (\log(1/r'))/(r^2 \log(4/r'))$ is strictly increasing and convex from (0,1) onto $(1/\log 16,1)$ . In particular, for all $r \in (0,1)$ ,
$$1/\log 16 < g(r) < (1/\log 16) + [1 - (1/\log 16)]r.$$
Corollary 2.4 · radius
Corollary 2.4. Let and be as in Remark 2.1, and as in Remark 2.2, B = B(a, b) and R = R(a, b). Then for all, (2.37) if, and if.
Corollary 2.4. Let $D_1$ and $D_2$ be as in Remark 2.1, $D_4$ and $D_5$ as in Remark 2.2, B = B(a, b) and R = R(a, b). Then for all $r \in (0, 1)$ ,
(2.37)
$$\max \left\{ ab \frac{1-r}{r} \log \frac{1}{1-r}, B-R-1 + \frac{1-r}{r} \log \frac{1}{1-r} \right\}$$
$$< BF(a,b;a+b;r) + \log(1-r) - R$$
$$< (B-R) \frac{1-r}{r} \log \frac{1}{1-r}$$
if $(a,b) \in D_1$ , and
$$(2.38) \quad (B-R)\frac{1-r}{r}\log\frac{1}{1-r} < BF(a,b;a+b;r) + \log(1-r) - R$$
$$< \min\left\{ab\frac{1-r}{r}\log\frac{1}{1-r}, B-R-1 + \frac{1-r}{r}\log\frac{1}{1-r}\right\}$$
if $(a,b) \in D_5$ .
Theorem 2.5
Theorem 2.5. For with c < a + b, let d = a + b - c, and define the function f on (0,1) by Then (2.41) and we have: - (1) If c = a or c = b,…
Theorem 2.5. For $a, b, c \in (0, \infty)$ with c < a + b, let d = a + b - c, and define the function f on (0,1) by
$$f(r) = (1 - r)^d F(a, b; c; r).$$
Then
(2.41)
$$f'(r) = \frac{1}{c}(c-a)(c-b)(1-r)^{d-1}F(a,b;c+1;r),$$
and we have:
- (1) If c = a or c = b, then $f(r) \equiv 1$ .
- (2) If $c < \min\{a,b\}$ or $c > \max\{a,b\}$ , then f is strictly increasing from (0,1) onto (1,D), where D = D(a,b,c) is defined by (2.40). Moreover, if $ab \ge (d-1)(c+1)$ (resp. ab < (d-1)(c+1)), then f is convex (resp. concave) on (0,1).
- (3) If $\min\{a,b\} < c < \max\{a,b\}$ , then f is strictly decreasing from (0,1) onto (D,1). Moreover, if $ab \ge (d-1)(c+1)$ (resp. ab < (d-1)(c+1)), then f is concave (resp. convex) on (0,1).