[step:Derive elementarity from existential preservation]We prove by induction on $L$-formulas $\psi(z_1,\dots,z_m)$ that for every tuple $a=(a_1,\dots,a_m) \in H^m$,
\begin{align*}
\mathcal{N} \models \psi(a_1,\dots,a_m)
\quad\iff\quad
\mathcal{M} \models \psi(a_1,\dots,a_m).
\end{align*}
For atomic formulas this follows from the definition of the induced substructure: function symbols are interpreted by restriction, constant symbols agree, equality is equality in the common universe, and relation symbols are restricted from $\mathcal{M}$ to $H$. The Boolean connectives are preserved by the induction hypothesis.
It remains to handle existential quantifiers. Let $\psi(z_1,\dots,z_m)$ be $\exists x\,\theta(x,z_1,\dots,z_m)$, and let $a \in H^m$. If
\begin{align*}
\mathcal{N} \models \exists x\,\theta(x,a),
\end{align*}
then some $h \in H$ satisfies $\mathcal{N}\models \theta(h,a)$, so the induction hypothesis gives $\mathcal{M}\models \theta(h,a)$ and hence $\mathcal{M}\models \exists x\,\theta(x,a)$. Conversely, if
\begin{align*}
\mathcal{M}\models \exists x\,\theta(x,a),
\end{align*}
then existential preservation gives $h \in H$ such that $\mathcal{M}\models \theta(h,a)$. The induction hypothesis gives $\mathcal{N}\models \theta(h,a)$, so $\mathcal{N}\models \exists x\,\theta(x,a)$.
Thus $\mathcal{N}\preccurlyeq \mathcal{M}$.[/step]