While going through old notes, in relation to the Dewar-Chatt-Duncanson model, it states that the stretching frequency for N2 to the group 8 transition metals (vNN) proceeds as:

[(SiP3)Fe(N2)]- 1920 cm-1

[(SiP3)Ru(N2)]- 1960 cm-1

[(SiP3)Os(N2)]- 1931 cm-1

However, the notes fail to elaborate further. I would have perhaps assumed that the trend was due to increasing size of the 3d -> 4d -> 5d orbitals and thus increased pi-donation causing a weaker bond and thus lower frequency, however, this does not fit the trend observed. Thus leads me to my question: how does the Dewar-Chatt-Duncanson model relate to the observed trend?

While going through old notes, in relation to the Dewar-Chatt-Duncanson model, it states that the stretching frequency for $\ce{N2}$ to the group 8 transition metals ($\nu(\ce{NN})$) proceeds as:

\begin{array}{cc} \hline \text{Compound} & \nu,~\pu{cm-1} \\ \hline \ce{[(SiP3)Fe(N2)]-} & 1920 \\ \ce{[(SiP3)Ru(N2)]-} & 1960 \\ \ce{[(SiP3)Os(N2)]-} & 1931 \\ \hline \end{array}

However, the notes fail to elaborate further.

I would have perhaps assumed that the trend was due to increasing size of the $\ce{3d -> 4d -> 5d}$ orbitals and thus increased pi-donation causing a weaker bond and thus lower frequency, however, this does not fit the trend observed. Thus leads me to my question: how does the Dewar-Chatt-Duncanson model relate to the observed trend?