Consider the integral
I := \iint_{\mathbb{R}^+_0 \times \mathbb{R}^+_0} \big(\lambda - 2d_1 \cosh(2\pi b k) - 2d_2 \cosh(2\pi b y)\big)_+ \, dk \, dy,
where d_1, d_2, b > 0 are as above and \lambda is such that the integrand is not identically zero, and in particular such that \lambda>2d_1+2d_2. The upper bound (\pi b)^{-2}\lambda\log^2\lambda Observe that for k,y\geq 0, we have the following analysis. The integrand is positive in the region of the (k,y)-plane where
\lambda - 2d_1 \cosh(2\pi b k) - 2d_2 \cosh(2\pi b y) > 0.
Fix k \ge 0 and \lambda large enough. Then the condition in y becomes
\cosh(2\pi b y) < \frac{\lambda - 2d_1 \cosh(2\pi b k)}{2d_2}.
Since \cosh(2\pi b y) \geq 1, it is necessary that
\lambda - 2d_1 \cosh(2\pi b k) > 2d_2,
which defines the domain in k:
0 \leq k < k_0 := \frac{1}{2\pi b} \operatorname{arccosh}\left( \frac{\lambda - 2d_2}{2d_1} \right).
For fixed k, define
y_{\max}(k) := \frac{1}{2\pi b} \operatorname{arccosh}\left( \frac{\lambda - 2d_1 \cosh(2\pi b k)}{2d_2} \right).
Then,
\begin{aligned}
&\int_0^{y_{\max}(k)} \big(\lambda - 2d_1 \cosh(2\pi b k) - 2d_2 \cosh(2\pi b y)\big) \, dy\\
&= \left(\lambda - 2d_1 \cosh(2\pi b k)\right) y_{\max}(k) - \frac{2 d_2}{2\pi b} \sinh\big(2\pi b y_{\max}(k)\big).
\end{aligned}
Using the identity \sinh(\operatorname{arccosh}(u)) = \sqrt{u^2 - 1} and setting
u(k) := \frac{\lambda - 2d_1 \cosh(2\pi b k)}{2 d_2},
we have
y_{\max}(k) = \frac{1}{2\pi b} \operatorname{arccosh} u(k), \quad \sinh(2\pi b y_{\max}(k)) = \sqrt{u(k)^2 - 1}.
Hence,
I \leq \int_0^{k_0} \frac{1}{2\pi b} \left[ \big(\lambda - 2d_1 \cosh(2\pi b k)\big) \operatorname{arccosh} u(k) - 2 d_2 \sqrt{u(k)^2 - 1} \right] dk,
where
k_0 = \frac{1}{2\pi b} \operatorname{arccosh} \left( \frac{\lambda - 2 d_2}{2 d_1} \right).
Using the change of variables u = \cosh(2\pi b k), with
dk = \frac{1}{2\pi b} \frac{du}{\sqrt{u^2 - 1}},
and the integration limits
u \in \left[1, \frac{\lambda - 2 d_2}{2 d_1}\right],
we get
I \leq \frac{1}{(2\pi b)^2} \int_1^{\frac{\lambda - 2 d_2}{2 d_1}} \frac{
(\lambda - 2 d_1 u) \operatorname{arccosh}\left(\frac{\lambda - 2 d_1 u}{2 d_2}\right)
- 2 d_2 \sqrt{\left(\frac{\lambda - 2 d_1 u}{2 d_2}\right)^2 - 1}
}{\sqrt{u^2 - 1}} \, du
\leq \frac{1}{(2\pi b)^2} \int_1^{\frac{\lambda - 2 d_2}{2 d_1}} \frac{
1}{\sqrt{u^2 - 1}} \, du \times \lambda \operatorname{arccosh}\left(\frac{\lambda }{2 d_2}\right)
\leq \frac{1}{(2\pi b)^2} \operatorname{arccosh}\left(\frac{\lambda-2d_2 }{2 d_1}\right) \times \lambda \operatorname{arccosh}\left(\frac{\lambda }{2 d_2}\right)
\leq \frac{1}{(2\pi b)^2} \operatorname{log}\left(\frac{\lambda-2d_2 }{ d_1}\right) \times \lambda \operatorname{log}\left(\frac{\lambda }{ d_2}\right).
In consequence we have proved that
\begin{aligned}
\sum_{j\geq 1}(\lambda-\lambda_j)_+\leq I_1= &4{\iint}_{\mathbb{R}^+_0\times \mathbb{R}^+_0} \big(\lambda-2d_1\cosh(2\pi bk)- 2d_2\cosh(2\pi by)\big)_+d kd y\\
&\leq \frac{1}{(\pi b)^2} \operatorname{log}\left(\frac{\lambda-2d_2 }{ d_1}\right) \times \lambda \operatorname{log}\left(\frac{\lambda }{ d_2}\right).
\end{aligned}
In consequence, when \lambda is large enough, we get
\sum_{j\geq 1}(\lambda-\lambda_j)_+\leq \frac{1}{\pi^2b^2}\lambda\log^2\lambda.
The dyadic analysis of the phase space: an upper bound Now, we are going to prove that
\begin{aligned}
\varkappa(\lambda) &\asymp \frac{1}{\pi^2b^2}(\lambda \log^2(\lambda)-(\lambda- \log_2(\lambda))\log^2(\lambda-\log_2(\lambda))\\
&-2\lambda\log(\lambda)+2(\lambda-\log_2(\lambda))\log(\lambda-\log_2(\lambda))\\
&+\log^2(\lambda)+\log_2(\lambda)).
\end{aligned}
Let us analyse I_1. So, let us take d=\min\{d_1,d_2\}, let us assume that \lambda>4d-1, and let us use that
\begin{aligned}
I_1 &\leq 4 \iint_{\mathbb{R}^+_0\times \mathbb{R}^+_0} \big(\lambda-2d_1\cosh(2\pi bk)- 2d_2\cosh(2\pi by)\big)_+d kd y \\
&\leq 4 \iint_{\mathbb{R}^+_0\times \mathbb{R}^+_0} \big(\lambda-2d\cosh(2\pi bk)- 2d\cosh(2\pi by)\big)_+d kd y.
\end{aligned}
Let us adopt the perspective of the Littlewood-Paley theory for pseudo-differential operators, and let us decompose the set \mathbb{R}^+_0\times \mathbb{R}^+_0 in the family of dyadic regions
\Omega_{j}=\{(y,k):2^{j-1}\leq \lambda-2d\cosh(2\pi bk)- 2d\cosh(2\pi by)\leq 2^j \},\quad j\in \mathbb{Z}.
Then, we have that
\begin{aligned}
I_1 &\leq 4 \iint_{\mathbb{R}^+_0\times \mathbb{R}^+_0} \big(\lambda-2d\cosh(2\pi bk)- 2d\cosh(2\pi by)\big)_+d kd y\leq 4\sum_{j\in \mathbb{Z}}\iint_{ \Omega_{j}} 2^jdkdy.
\end{aligned}
Note that for any j\in \mathbb{Z}, we have that
\begin{aligned}
\Omega_{j}=\{(y,k):\frac{\lambda-2^j}{2d}\leq \cosh(2\pi bk)+ \cosh(2\pi by)\leq \frac{\lambda-2^{j-1}}{2d} \}.
\end{aligned}
Note that
\forall (y,k),\quad 2\leq \cosh(2\pi bk)+ \cosh(2\pi by).
In consequence,
\frac{\lambda-2^{j-1}}{2d} <2\Longleftrightarrow j>\frac{\log(\lambda-4d)}{\log 2}+1,\textnormal{ and for these values of } j,\,\quad \Omega_j=\{\emptyset\}.
In consequence, the decomposition of \mathbb{R}^+_0\times \mathbb{R}^+_0 in the sets \Omega_j runs over j\in \mathbb{Z} such that j\leq \frac{\log(\lambda-4d)}{\log 2}+1.
In consequence, the change of variables (u,v)=(\cosh(2\pi bk),\cosh(2\pi by)) leads to
\begin{aligned}
&4\sum_{j\in \mathbb{Z}}\iint_{ \Omega_{j}} 2^jdkdy \\
&= \frac{1}{\pi^2b^2} \sum_{j\leq \frac{\log(\lambda-4d)}{\log 2}+1} 2^j \iint_{ \{(u,v): \frac{\lambda-2^j}{2d}\leq u+v\leq \frac{\lambda-2^{j-1}}{2d},\,u,v\geq 1 \} }\frac{dudv}{\sqrt{u^2-1}\sqrt{v^2-1}}.
\end{aligned}
Let us analyse the integral
R_j:= \iint_{ \{(u,v): \frac{\lambda-2^j}{2d}\leq u+v\leq \frac{\lambda-2^{j-1}}{2d},\,u,v\geq 1 \} }\frac{dudv}{\sqrt{u^2-1}\sqrt{v^2-1}}.
Item. Case 1: \frac{\lambda-2^j}{2d}<1. In this case \frac{\log(\lambda-2d)}{\log 2}<j, and with B=\frac{\lambda-2^{j-1}}{2d}, we have that
\begin{aligned}
R_j &= \iint_{ \{(u,v): \frac{\lambda-2^j}{2d}\leq u+v\leq \frac{\lambda-2^{j-1}}{2d},\,u,v\geq 1 \} }\frac{dudv}{\sqrt{u^2-1}\sqrt{v^2-1}}\\
&\leq \int\limits_{1}^B\frac{1}{\sqrt{u^2-1}}\int\limits_{1}^{B-u}\frac{dvdu}{\sqrt{v^2-1}}\\
&\leq \int\limits_{1}^B\frac{du}{\sqrt{u^2-1}}\int\limits_{1}^{B-1}\frac{dv}{\sqrt{v^2-1}}\\
&= \int\limits_{1}^B\frac{\cosh^{-1}(B-1)du}{\sqrt{u^2-1}}\\
&= \int\limits_{1}^B\frac{\log(B-1+\sqrt{(B-1)^2-1})du}{\sqrt{u^2-1}}\\
&\leq \log(2B)\int\limits_{1}^B\frac{du}{\sqrt{u^2-1}}= \log(2B)\cosh^{-1}(B)\leq (\log(2B))^2.
\end{aligned}
Case 2: \frac{\lambda-2^j}{2d}\geq 1, and then \frac{\log(\lambda-2d)}{\log 2}\geq j. Then, with A= \frac{\lambda-2^j}{2d} and with B=\frac{\lambda-2^{j-1}}{2d}, we have that
\begin{aligned}
R_j &= \iint_{ \{(u,v): \frac{\lambda-2^j}{2d}\leq u+v\leq \frac{\lambda-2^{j-1}}{2d},\,u,v\geq 1 \} }\frac{dudv}{\sqrt{u^2-1}\sqrt{v^2-1}}\\
&\leq \int\limits_{1}^A\frac{1}{\sqrt{u^2-1}}\int\limits_{1}^{B-u}\frac{dv du}{\sqrt{v^2-1}}+ \int\limits_{A}^B\frac{1}{\sqrt{u^2-1}}\int\limits_{1}^{B-u}\frac{dv du}{\sqrt{v^2-1}}\\
&\leq \cosh^{-1}(A)\log(2B)+ (\cosh^{-1}(B)-\cosh^{-1}(A))\log(2B)\\
&\leq (\log(2B))^2.
\end{aligned}
The analysis above allows us to estimate
\begin{aligned}
& \sum_{j\leq \frac{\log(\lambda-4d)}{\log 2}+1} 2^j R_j\\
&= \sum_{j\leq \frac{\log(\lambda-4d)}{\log 2}+1 : -\infty<j\leq 0 } 2^j R_j+\sum_{0 <j\leq \frac{\log(\lambda-4d)}{\log 2}+1 } 2^j R_j\\
&\leq \sum_{j\leq \frac{\log(\lambda-4d)}{\log 2}+1 : -\infty<j\leq 0 } 2^j \log^2\left(\frac{\lambda-2^{j-1}}{d}\right)+\sum_{0 <j\leq \frac{\log(\lambda-4d)}{\log 2}+1 } 2^j R_j\\
&\leq 2 \log^2\left(\frac{\lambda}{d}\right) +\sum_{1\leq j\leq \frac{\log(\lambda-4d)}{\log 2}+1 } 2^j R_j
\\
&\leq 2\log^2\left(\frac{\lambda}{d}\right)+\sum_{1\leq j\leq \frac{\log(\lambda-4d)}{\log 2}+1 } 2^j \left(\log\left(\frac{\lambda-2^{j-1}}{d}\right)\right)^2.
\end{aligned}
Observe that in the interval
1\leq x\leq \frac{\log(\lambda-4d)}{\log 2}+1 ,
we have that
\frac{\lambda}{d}-\frac{\log(\lambda-4d)}{d\log 2}-\frac{1}{d}\leq \frac{\lambda-x}{d}\leq \frac{\lambda-1}{d},
and we can assure that the lower bound \frac{\lambda}{d}-\frac{\log(\lambda-4d)}{d\log 2}-\frac{1}{d}>1, if \lambda is large enough. In this intervale the function f(x)= \left(\log\left(\frac{\lambda-x}{d}\right)\right)^2 is decreasing. We can make the estimate of the last series as a Riemann summation on the interval [1, \frac{\log(\lambda-4d)}{\log 2}+1] with the interior points being x_j=2^{j-1} and with \Delta x_j\leq 2^{j-1}-2^{j-2}=2^{j-2}. Indeed,
\begin{aligned}
& \sum_{1\leq j\leq \frac{\log(\lambda-4d)}{\log 2}+1 } 2^j \left(\log\left(\frac{\lambda-2^{j-1}}{d}\right)\right)^2 \\&=4\sum_{1\leq j\leq \frac{\log(\lambda-4d)}{\log 2}+1 } 2^{j-2} \left(\log\left(\frac{\lambda-2^{j-1}}{d}\right)\right)^2\\
&\leq 4\int_1^{\frac{\log(\lambda-4d)}{\log 2}+1}\left(\log\left(\frac{\lambda-x}{d}\right)\right)^2dx.
\end{aligned}
A change of variables in the last integral allows us to compute
\begin{aligned}
&\int_1^{\frac{\log(\lambda-4d)}{\log 2}+1}\left(\log\left(\frac{\lambda-x}{d}\right)\right)^2dx= d\int_{\frac{\lambda}{d}-\frac{\log(\lambda-4d)}{d\log 2}-\frac{1}{d}}^{\frac{\lambda}{d}-\frac{1}{d}}\left(\log\left(t\right)\right)^2dt\\
&=d\left(\left(\frac{\lambda}{d}-\frac{1}{d}\right)\log^2\left(\frac{\lambda}{d}-\frac{1}{d}\right)-2\left(\frac{\lambda}{d}-\frac{1}{d}\right)\log\left(\frac{\lambda}{d}-\frac{1}{d}\right)+2\left(\frac{\lambda}{d}-\frac{1}{d}\right)\right)\\
&- d\left(\left(\frac{\lambda}{d}-\frac{\log(\lambda-4d)}{d\log 2}-\frac{1}{d}\right)\log^2\left(\frac{\lambda}{d}-\frac{\log(\lambda-4d)}{d\log 2}-\frac{1}{d}\right)\right)\\
&-d\left(-2\left(\frac{\lambda}{d}-\frac{\log(\lambda-4d)}{d\log 2}-\frac{1}{d}\right)\log\left(\frac{\lambda}{d}-\frac{\log(\lambda-4d)}{d\log 2}-\frac{1}{d}\right)+2\left(\frac{\lambda}{d}-\frac{\log(\lambda-4d)}{d\log 2}-\frac{1}{d}\right)\right)\\
&=\left(\lambda-1\right)\log^2\left(\frac{\lambda}{d}-\frac{1}{d}\right)-2\left(\lambda-1\right)\log\left(\frac{\lambda}{d}-\frac{1}{d}\right)+2\left(\lambda-1\right)\\
&- \left(\lambda-\frac{\log(\lambda-4d)}{\log 2}-1\right)\log^2\left(\frac{\lambda}{d}-\frac{\log(\lambda-4d)}{d\log 2}-\frac{1}{d}\right)\\
&+2\left(\lambda-\frac{\log(\lambda-4d)}{\log 2}-1\right)\log\left(\frac{\lambda}{d}-\frac{\log(\lambda-4d)}{d\log 2}-\frac{1}{d}\right)-2\left(\lambda-\frac{\log(\lambda-4d)}{\log 2}-1\right)\\
&=\left(\lambda-1\right)\log^2\left(\frac{\lambda}{d}-\frac{1}{d}\right)-2\left(\lambda-1\right)\log\left(\frac{\lambda}{d}-\frac{1}{d}\right)\\
&- \left(\lambda-\frac{\log(\lambda-4d)}{\log 2}-1\right)\log^2\left(\frac{\lambda}{d}-\frac{\log(\lambda-4d)}{d\log 2}-\frac{1}{d}\right)\\
&+2\left(\lambda-\frac{\log(\lambda-4d)}{\log 2}-1\right)\log\left(\frac{\lambda}{d}-\frac{\log(\lambda-4d)}{d\log 2}-\frac{1}{d}\right)+\frac{2\log(\lambda-4d)}{\log 2}.
\end{aligned}
In consequence,
\begin{aligned}
I_1 &\leq 4\sum_{j\in \mathbb{Z}}\iint_{ \Omega_{j}} 2^jdkdy\\
&\leq \frac{1}{\pi^2b^2}\sum_{j\leq \frac{\log(\lambda-4d)}{\log 2}+1} 2^j R_j\\
&\leq \frac{2}{\pi^2b^2} \log^2\left(\frac{\lambda}{d}\right) \\
&+ \frac{4}{\pi^2b^2}\left(\left(\lambda-1\right)\log^2\left(\frac{\lambda}{d}-\frac{1}{d}\right)-2\left(\lambda-1\right)\log\left(\frac{\lambda}{d}-\frac{1}{d}\right)\right)\\
&- \frac{4}{\pi^2b^2}\left(\lambda-\frac{\log(\lambda-4d)}{\log 2}-1\right)\log^2\left(\frac{\lambda}{d}-\frac{\log(\lambda-4d)}{d\log 2}-\frac{1}{d}\right)\\
&+\frac{8}{\pi^2b^2}\left(\left(\lambda-\frac{\log(\lambda-4d)}{\log 2}-1\right)\log\left(\frac{\lambda}{d}-\frac{\log(\lambda-4d)}{d\log 2}-\frac{1}{d}\right)+\frac{\log(\lambda-4d)}{\log 2}\right).
\end{aligned}
So, when \lambda is large enough, one can estimate
\begin{aligned}
I_1 &\lesssim \frac{1}{\pi^2b^2}(\lambda \log^2(\lambda)-2\lambda\log(\lambda)-(\lambda- \log_2(\lambda))\log^2(\lambda-\log_2(\lambda))\\
&+2(\lambda-\log_2(\lambda))\log(\lambda-\log_2(\lambda))+\log^2(\lambda)+\log_2(\lambda)).
\end{aligned}
The dyadic analysis of the phase space: a lower bound For the analysis of I_2, let us take 2d'=\max\{2/d_1,2/d_2\}, and then
\begin{aligned}
I_2 &\geq 4 \iint_{\mathbb{R}^+_0\times \mathbb{R}^+_0} \big(\lambda-\frac{2}{d_1}\cosh(2\pi bk)- \frac{2}{d_2}\cosh(2\pi by)\big)_+d kd y \\
&\geq 4 \iint_{\mathbb{R}^+_0\times \mathbb{R}^+_0} \big(\lambda-2d'\cosh(2\pi bk)-2 d'\cosh(2\pi by)\big)_+d kd y.
\end{aligned}
Let us decompose the set \mathbb{R}^+_0\times \mathbb{R}^+_0 in the family of dyadic zones
\widetilde\Omega_{j}=\{(k,y):2^{j-1}\leq \lambda-2d'\cosh(2\pi bk)- 2d'\cosh(2\pi by)\leq 2^j \},\quad j\in \mathbb{Z}.
We have the estimate
\begin{aligned}
I_2 &\geq 4 \iint_{\mathbb{R}^+_0\times \mathbb{R}^+_0} \big(\lambda-2d'\cosh(2\pi bk)- 2d'\cosh(2\pi by)\big)_+d kd y\asymp 4\sum_{j\in \mathbb{Z}}\iint_{ \widetilde\Omega_{j}} 2^jdkdy.
\end{aligned}
Note that for any j\in \mathbb{Z}, we have that
\begin{aligned}
\widetilde \Omega_{j}=\{(y,k):\frac{\lambda-2^j}{2d'}\leq \cosh(2\pi bk)+ \cosh(2\pi by)\leq \frac{\lambda-2^{j-1}}{2d'} \}.
\end{aligned}
So, in our further analysis we fix \lambda>4d'-1. Note that
\forall (y,k),\quad 2\leq \cosh(2\pi bk)+ \cosh(2\pi by).
In consequence,
\frac{\lambda-2^{j-1}}{2d'} <2\Longleftrightarrow j>\frac{\log(\lambda-4d')}{\log 2}+1,\textnormal{ and for these values of } j,\,\quad \widetilde\Omega_j=\{\emptyset\}.
Observe that the decomposition of \mathbb{R}^+_0\times \mathbb{R}^+_0 into the sets \widetilde\Omega_j runs over j\in \mathbb{Z} such that j\leq \frac{\log(\lambda-4d')}{\log 2}+1.
The change of variables (u,v)=(\cosh(2\pi bk),\cosh(2\pi by)) leads to
\begin{aligned}
&4\sum_{j\in \mathbb{Z}}\iint_{ \widetilde\Omega_{j}} 2^jdkdy \\
&\asymp \frac{1}{\pi^2b^2} \sum_{j\leq \frac{\log(\lambda-4d')}{\log 2}+1} 2^j \iint_{ \{(u,v): \frac{\lambda-2^j}{2d'}\leq u+v\leq \frac{\lambda-2^{j-1}}{2d'},\,u,v\geq 1 \} }\frac{dudv}{\sqrt{u^2-1}\sqrt{v^2-1}}.
\end{aligned}
Let us analyse the integral
R_j:= \iint_{ \{(u,v): \frac{\lambda-2^j}{2d'}\leq u+v\leq \frac{\lambda-2^{j-1}}{2d'},\,u,v\geq 1 \} }\frac{dudv}{\sqrt{u^2-1}\sqrt{v^2-1}}.
Item. Case 1: \frac{\lambda-2^j}{2d'}<1. In this case \frac{\log(\lambda-2d')}{\log 2}<j. As above let us define A= \frac{\lambda-2^j}{2d'} and B=\frac{\lambda-2^{j-1}}{2d'}. Since \lambda\rightarrow\infty, we can fix a constant M_0\gg 2d', such that \lambda\geq M_0+1. Then, for j\leq 0, we can chose \alpha<1, such that
B= \frac{2d'}{\lambda-2^{j-1}}\leq \frac{2d'}{\lambda-1}\leq \frac{2d'}{M_0}<\alpha<1.
Note that this choice of \alpha allows us to compare 1/B<\alpha<1, or equivalently, 1<B\alpha<B, with \alpha independent of \lambda and of j\neq 0. In general, observe that for any j with \widetilde\Omega_j\neq \{\emptyset\}, that B\geq 2, and then one can always choose 1/B\leq 1/2\leq \alpha<1, independent of \lambda and of j\leq \frac{\log(\lambda-4d)}{\log 2}+1, satisfying that 1\leq B\alpha<B.
Keeping in mind this choice of \alpha we have that
\begin{aligned}
R_j &= \iint_{ \{(u,v): \frac{\lambda-2^j}{2d'}\leq u+v\leq \frac{\lambda-2^{j-1}}{2d'},\,u,v\geq 1 \} }\frac{dudv}{\sqrt{u^2-1}\sqrt{v^2-1}}\\
&\asymp \int\limits_{1}^B\frac{1}{\sqrt{u^2-1}}\int\limits_{1}^{B-u}\frac{dvdu}{\sqrt{v^2-1}}\\
&\geq \int\limits_{1}^{\alpha B}\frac{\cosh^{-1}(B-u)du}{\sqrt{u^2-1}}\\
&\geq \int\limits_{1}^{\alpha B}\frac{\cosh^{-1}(B-B\alpha)du}{\sqrt{u^2-1}}\\
&= \cosh^{-1}(B(1-\alpha)) \cosh^{-1}(\alpha B)\geq \log^2(B\varkappa) ,
\end{aligned}
where 0<\varkappa=\min(\alpha,1-\alpha). Note that the choice \alpha=1/2, optimises the value of \varkappa=1/2. In this case we have estimated
R_j\gg \log^2(B/2).
Case 2: \frac{\lambda-2^j}{2d'}\geq 1, and then \frac{\log(\lambda-2d')}{\log 2}\geq j. With A= \frac{\lambda-2^j}{2d'} and with B=\frac{\lambda-2^{j-1}}{2d'}, we have that
\begin{aligned}
R_j &= \iint_{ \{(u,v): \frac{\lambda-2^j}{2d'}\leq u+v\leq \frac{\lambda-2^{j-1}}{2d'},\,u,v\geq 1 \} }\frac{dudv}{\sqrt{u^2-1}\sqrt{v^2-1}}\\
&\asymp \int\limits_{1}^A\frac{1}{\sqrt{u^2-1}}\int\limits_{1}^{B-u}\frac{dv du}{\sqrt{v^2-1}}+ \int\limits_{A}^B\frac{1}{\sqrt{u^2-1}}\int\limits_{1}^{B-u}\frac{dv du}{\sqrt{v^2-1}}\\
&= \int\limits_{1}^B\frac{1}{\sqrt{u^2-1}}\int\limits_{1}^{B-u}\frac{dv du}{\sqrt{v^2-1}}\\
& = \int\limits_{1}^B\frac{\cosh^{-1}(B-u)du}{\sqrt{u^2-1}}.
\end{aligned}
Observe that for any j with \widetilde\Omega_j\neq \{\emptyset\}, one has that B\geq 2, and then one can always choose 1/B\leq 1/2<\alpha<1 independent of \lambda and of j\leq \frac{\log(\lambda-4d')}{\log 2}+1, satisfying 1\leq B\alpha<B.The analysis above allows us to estimate
\int\limits_{1}^B\frac{\cosh^{-1}(B-u)du}{\sqrt{u^2-1}}\geq \int\limits_{1}^{\alpha B}\frac{\cosh^{-1}(B-u)du}{\sqrt{u^2-1}}\geq \cosh^{-1}(B(1-\alpha)) \cosh^{-1}(\alpha B)
\geq \log^2(B/2).
In consequence,
\begin{aligned}
& \sum_{j\leq \frac{\log(\lambda-4d')}{\log 2}+1} 2^j R_j\\
&= \sum_{j\leq \frac{\log(\lambda-4d')}{\log 2}+1 : -\infty<j\leq 0 } 2^j R_j+\sum_{0 <j\leq \frac{\log(\lambda-4d')}{\log 2}+1 } 2^j R_j\\
&= \sum_{j\leq \frac{\log(\lambda-4d')}{\log 2}+1 : -\infty<j\leq 0 } 2^j \log^2\left(\frac{\lambda-2^{j-1}}{4d'}\right)+\sum_{0 <j\leq \frac{\log(\lambda-4d')}{\log 2}+1 } 2^j R_j\\
&\geq \log^2\left(\frac{\lambda-1}{4d'}\right) +\sum_{1\leq j\leq \frac{\log(\lambda-4d')}{\log 2}+1 } 2^j R_j
\\
&\gg \log^2\left(\frac{\lambda-1}{4d'}\right) +\sum_{1\leq j\leq \frac{\log(\lambda-4d)}{\log 2}+1 } 2^j \left(\log\left(\frac{\lambda-2^{j-1}}{4d'}\right)\right)^2.
\end{aligned}
As above, we can make the estimate of the last series as a Riemann summation. Hence,
\begin{aligned}
\sum_{1\leq j\leq \frac{\log(\lambda-4d)}{\log 2}+1 } 2^j \left(\log\left(\frac{\lambda-2^{j-1}}{4d'}\right)\right)^2 &=4\sum_{1\leq j\leq \frac{\log(\lambda-4d')}{\log 2}+1 } 2^{j-2} \left(\log\left(\frac{\lambda-2^{j-1}}{4d'}\right)\right)^2\\
&\asymp \int_1^{\frac{\log(\lambda-4d')}{\log 2}+1}\left(\log\left(\frac{\lambda-x}{4d'}\right)\right)^2dx.
\end{aligned}
A change of variables in the last integral allows us to compute
\begin{aligned}
&= 4d'\int_{\frac{\lambda}{4d'}-\frac{\log(\lambda-4d')}{4d'\log 2}-\frac{1}{4d'}}^{\frac{4\lambda}{d'}-\frac{1}{4d'}}\left(\log\left(t\right)\right)^2d't\\
&=4\left(\lambda-1\right)\log^2\left(\frac{\lambda}{4d'}-\frac{1}{4d'}\right)-8\left(\lambda-1\right)\log\left(\frac{\lambda}{4d'}-\frac{1}{4d'}\right)\\
&- 4\left(\lambda-\frac{\log(\lambda-4d')}{\log 2}-1\right)\log^2\left(\frac{\lambda}{d'}-\frac{\log(\lambda-4d')}{4d'\log 2}-\frac{1}{d'}\right)\\
&+8\left(\lambda-\frac{\log(\lambda-4d')}{\log 2}-1\right)\log\left(\frac{\lambda}{d'}-\frac{\log(\lambda-4d')}{4d'\log 2}-\frac{1}{d'}\right)+\frac{8\log(\lambda-4d')}{\log 2}.
\end{aligned}
In consequence,
\begin{aligned}
I_2 &\geq 4\sum_{j\in \mathbb{Z}}\iint_{ \Omega_{j}} 2^jdkdy\\
&\asymp \frac{1}{\pi^2b^2}\sum_{j\leq \frac{\log(\lambda-4d')}{\log 2}+1} 2^j R_j\\
&\gg\frac{1}{\pi^2b^2} \log^2\left(\frac{\lambda-1}{4d'}\right) \\
&+ \frac{1}{\pi^2b^2}\left(\left(\lambda-1\right)\log^2\left(\frac{\lambda}{4d'}-\frac{1}{4d'}\right)-2\left(\lambda-1\right)\log\left(\frac{\lambda}{4d'}-\frac{1}{4d'}\right)\right)\\
&- \frac{1}{\pi^2b^2}\left(\lambda-\frac{\log(\lambda-4d')}{\log 2}-1\right)\log^2\left(\frac{\lambda}{4d'}-\frac{\log(\lambda-4d')}{4d'\log 2}-\frac{1}{4d'}\right)\\
&+\frac{2}{\pi^2b^2}\left(\left(\lambda-\frac{\log(\lambda-4d')}{\log 2}-1\right)\log\left(\frac{\lambda}{4d'}-\frac{\log(\lambda-4d')}{4d'\log 2}-\frac{1}{4d'}\right)+\frac{\log(\lambda-4d')}{\log 2}\right).
\end{aligned}
In consequence, when \lambda\rightarrow \infty, we acomplish the two conditions on \lambda imposed above, namely that, \lambda>4d-1,4d'-1, and we allow for the estimate
\begin{aligned}
I_2 &\gg \frac{1}{\pi^2b^2}(\lambda \log^2(\lambda)-2\lambda\log(\lambda)-(\lambda- \log_2(\lambda))\log^2(\lambda-\log_2(\lambda))\\
&+2(\lambda-\log_2(\lambda))\log(\lambda-\log_2(\lambda))+\log^2(\lambda)+\log_2(\lambda)).
\end{aligned}
The proof of Theorem 1.1 is complete.