Some basic methods are:

1. maximum of 2nd derivative to identify systolic foot.

2. maximum of 1st derivative to identify systolic slope

3. intersecting tangent to identify systolic foot.

4. best fit with a tanh function to identify systolic slope.

In my limited experience, PAT from methods 1 and 3 correlate better with BP, however, they are more susceptible to noise in the pulse wave signal.  This can be overcome with signal averaging though. 4 Is more robust, but is patented, in case you are considering commercial applications later on. 

As for literature, this paper is somewhat relevant, although it goes a bit further by calculating pulse wave velocity.

1. Gaddum, N. R., et al. (2013). "A technical assessment of pulse wave velocity algorithms applied to non-invasive arterial waveforms." Ann Biomed Eng 41(12): 2617-2629.

Two more papers:

2. Mukkamala, R., et al. (2006). "Continuous cardiac output monitoring by peripheral blood pressure waveform analysis." IEEE Trans Biomed Eng 53(3): 459-467.

3. Buxi, D., et al. (2015). "A survey on signals and systems in ambulatory blood pressure monitoring using pulse transit time." Physiological Measurement 36(3): R1-R26.

Hope this helps.

To calculate the pulse arrival time (PAT) or pulse transit time (PTT) using ECG and PPG signal, measure the time delay between R peak of the ECG signal and any one critical point of the PPG signal. Critical points are few characteristic identification points of the blood pulse wave. You can choose anyone critical point among the listed three critical point definitions.

Critical points are few characteristic identification points of the blood pulse wave. You can choose anyone critical point among the listed three critical point definitions.

1. 2nd derivative maximum point  (Systolic foot) 

2. 1st derivative maximum (Systolic slope)

3. Intersecting tangent point (Systolic foot) 

I suggest you, test with all critical points individually to get the best correlation with BP for PAT from your measurement system. In my experiments, when I calculated time delay between two blood pulse signal ( without ECG signal ), the first derivative maximum point was more repeatable. Since you have ECG signal along with PPG, the intersecting tangent method will be the best in my opinion.

Before start the experiments and signal acquisition, make sure that the interchannel delay between your ECG channel and PPG channel is very small  compared to expected PTT or PAT.  Also, use digitizer with high sampling rate to get better correlation and to make it free from error due to signal acquisitions.