It is straightforward to see thatWe can express the additive twistsform $g(\tau)$ as $$ f\left(\tau+\frac{M}{N}\right)=\sum_{n\in \mathbb{N}}a(n)\mathrm{e}^{2\pi \mathrm{i}\frac{M}{N}}\,q^n,\qquad M\in\mathbb{Z},$$$$ g(\tau)=\sum_{d\mid N}\mu(d)\sum_{\substack{n\in\mathbb{N}\\d\mid n}} a(n)\,q^n.$$ lie in $M(k,\Gamma_1(N^2))$, becauseIn the notation of Atkin-Lehner $\left(\begin{smallmatrix}1 &\frac{M}{N}\\ 0 & 1\end{smallmatrix}\right)$ normalizes(Hecke operators on $\Gamma_1(N^2)$$\Gamma_0(m)$, Math. Taking linear combinations of these additive twistsAnn. 185 (1970), all134-160), the forms $$ f_m(\tau)=\sum_{\substack{n\in \mathbb{N}\\n\equiv m\ (N)}}a(n)\,q^n,\qquad m\in\mathbb{Z}, $$ lie ininner sum is $M(k,\Gamma_1(N^2))$$(f\mid U_d)\mid B_d$, hence so does yourwhich lies in $g(\tau)$$M(k,\Gamma_0(dN))$ by Lemmata 2, 6, 14 in the paper. This implies that $$ g\in M(k,\Gamma_0(N^2)). $$