TRIACs (and all thyristors) can have difficulty turning off when driving an inductive load including the leakage inductance of a transformer.
We don 't have an inductive load here. Our load is a diode in series with a resistor (DLoad, RLoad).
A snubber network is still advisable, because even with a pure non-inductive resistive load, the TRIAC 'sees' all the wiring inductance, to the element and also back to the wall and from there to nearby appliances with a class X filter cap between L and N. Also, most mains soldering irons use a wound element so, depending on the winding structure, may not be non-inductive.
-------------------------------------------
I've put together a hopefully improved BR100/03 DIAC model:
* Parameterised three layer DIAC model
* for LTspice to match cuve of BR100/03
* without dicontinuities.
* Hand-crufted by Ian.M 22/08/2018
*
.subckt BR100 T1 T2 params: Vbr=30.5V, Ibr=30uA, Vfr=10.2V, If=10mA, Rs=1, Cp=1nF
*Precalculate consts.
.param K1={1.0141*Vbr}, K2={2.1481/Ibr}, K3={Vfr/Vbr}, K4={-7.0/If}, K5=(Vbr-Vfr)/Vbr
Rdiac N001 N002 {Rs}
Vdiac T1 N001 0 ; ammeter
Cdiac N001 T2 {Cp}
B1 N002 T2 v={K1}*tanh({K2}*I(Vdiac))*({K3}*exp({K4}*abs(I(Vdiac)))+{K5})
.ends BR100
I've tested the model under LTspice in the classic series dimmer circuit.
I don't think it can be improved without a selection of BR100/03 DIACs and a curve tracer with a low duty cycle pulsed mode to avoid self-heating.
The subcircuit parameters are similar to those in the Philips datasheet, but actual values apart from Ibr matched to Hugo Coolens curve data.
The magic numbers in the precalculate constants .param line are:
K1, 1.0141, match peak voltage to Vbr
K2, 2.1481, match 98% Vbr @ Ibr - increase to increase %
K4, 7.0, Match exponential to curve data - increase to sharpen
Attachments are the dimmer sim and a test jig sim to trace (and compare) model DIAC curves that has this model, the LTspice DIAC model from examples\educational\dimmer.asc, and Hugo Coolens' table based model (the one LTspice barfs on in an actual dimmer circuit).