Hello Arnon,

Please find below my answers/tips after our Monday Skype meeting :

1. Auto focusing

To identify the cause and to determine the most relevant processing, we have planned a Skype meeting Wednesday 16^th at 10 AM Paris time.

Therefore, I will guide you to perform several tests to have a better understanding of the instrument status.

2. Instrument sensitivity

The sensitivity of your spectrometers is limited by the gate. Due to the fact you cannot have a very fast gating spectrometer feature, your plasma is observed during 1 ms minimum. LIBS plasma lifetime is about of tens of µs, so there is a decrease of the signal. A plasma is an inhomogeneous phenomenon through its time life. The elements do not emit at the same intensity at the beginning and ending of the plasma. Some of them emit at the beginning like the state II emission and the most electronegative or stable elements. With these spectrometers, you have the full plasma lifetime. You have a decrease of intensity but a better representativity of the material.

However, you may increase the sensitivity by optimizing three parameters.

1. By increase the laser energy.

It will increase the ablated mass i.e. the number of excited elements inside the plasma. Be careful, when you go upper than a threshold (about 7 to 9 mJ), some plasma will formed on particles in the ambient air. A laser-plasma shielding effect can occur too. To avoid a decrease of reliability due to higher laser energy, decrease the number of laser shots (the best case is to use one shot per spectrum but it is a little time consuming).

Another good practice is to draw the intensity = f(laser energy) function for the elements of interest and make a compromise or duplicate the analyses at several laser energy.

2. By changing the delay (DG): time before the plasma observation.

A good practice is to draw the intensity = f(delay) function for the elements of interest and make a compromise or duplicate the analyses at several optimized delay.

This optimization must be performed when you change the laser energy and the material due to matrix effect.

3. By selecting the state II emission lines.

As you already know, some elements are more sensitive at state II of excitation like the alkaline earth metal (the column of Mg, Ca …).

3. Instrument repeatability

The repeatability of the signal (at consecutive laser shots) depends on two variables :

1. The plasma stability.

A plasma is not basically a stable phenomenon. To decrease the impact on the measure, it is advised to make several repetition and to consider a measure like an average (or sum) of spectra.

Good to know, the repeatability decreased following a square root function depending on the number of repetitions. We both know this rule.

2. The material homogeneity and spatial resolution of the analysis.

The plasma ablates a mass of about 50µm of diameter for few µm in depth. Depending on the material inhomogeneity, it is possible to measure a particular spot than contains an inclusion or a higher concentrations.

So, I have checked the data you send us (ARMI and NBS samples) and find below my tips.

All the spectra have the same shape. There is no obvious difference in signal shapes.

If I consider all the 25 spectra independently (without making a sum or an average), I obtain the following repeatability :

1. 2.7% and 5.2 % for the ARMI sample on Cu 327.4 nm, Fe 373.5 nm, Al 394.5 nm and Mn 403.1 nm
2. 3.0 and 5.4 % for the NBS sample on Mg+ 285.0 nm, Al 394.5 nm and Mn 403.1 nm
3. 1% for the ARMI and 3% for the NBS with the complete integration of each spectrum.

The only repeatability variation I observe clearly on your data is for pollutant elements like Na (588.8 nm) or K ( 766.473nm).

I think the shifted reliability you observed do not come from the instrument but from the material inhomogeneity.

By performing a sum of 3 spectra like you described to me, I obtained a better repeatability (this is logical due to the increase of repetitions) between 1 and 2%.

Maybe I did not observed the same emissions than you did ?

Good to know is LIBS has a heteroskedastic signal.

The variance increase when the concentration is very low or very high for one emission line. To obtain the wanted variance, the emission line used must be changed to a more sensitive (if concentration is low) or less sensitive (is concentration is high).

Another explanation could come from the difference of data processing.

I use the “Extract” function of Prisma to calculate for each spectrum the integrated signal on a wavelength range and I considered the intensity less background signal (I-B).

I used this formula to calculate the repeatability : standard deviation divided by average.

Let me know what data processing you used, or the emission lines for which you observed the variation and I will check again.

Hopping I answered properly to your points.

Regards,

Kévin.

OK Arnon. Let’s talk at 6pm with Skype. Best, Ronald

Does 6pm or later, works for you ?