The etch rate of lpcvd deposited nitride depends on the silicon content in it.
If the nitride you get is silicon rich that means the gas flow rate of SiH2Cl2
is several times larger than that of NH3, its etch rate in HF in very slow.

>From: "Andrea Mazzolari" 
>Reply-To: mazzolari@fe.infn.it, General MEMS discussion 
>To: "General MEMS discussion" 
>Subject: Re:[mems-talk] silicon nitride HF resistant
>Date: Mon, 14 Dec 2009 17:23:24 +0100 (CET)
>
>Hi all,
>
>i have some silicon wafers coated with 100nm LPCVD silicon nitide.
>I need to patter the silicon nitride. I realized photolitograhy and
>immersed the wafer in BHF. Usually removal of silicon nitride takes about
>4 hours, but this time after 7 hours silicon nitride was still present.
>I removed photoresist and immersed the wafer in concentrated HF. I can see
>that the color of silicon nitride slowly changes with time, so probably
>there is some etching, but after 1 hour silicon nitride was still
>present!!
>
>After that i immersed in the same HF a second wafers coated with 100nm
>LPCVD silicon nitide (this second wafer is coming from another source).
>Silicon nitride was removed in some minutes.
>
>Where could be the problem with the silicon nitride on the first wafer ??
>
>Best regards,
>Andrea