Investigating nanoscale chemical heterogeneity in polyamide polymer membranes using STEM-EELS – a beam damage study
- Abstract number
- 183
- Presentation Form
- Submitted Talk
- Corresponding Email
- [email protected]
- Session
- Stream 1: EMAG - Soft and Hybrid Materials
- Authors
- Dr Catriona McGilvery (1), Dr Patricia Abellan (3), Dr Michal Klosowski (1), Prof Joao Cabral (1), Prof Andrew Livingston (1), Prof Quentin Ramasse (2, 4), Prof Alexandra Porter (1)
- Affiliations
-
1. Imperial College London
2. SuperSTEM Laboratory
3. University of Nantes
4. University of Leeds
- Keywords
beam damage
STEM-EELS
polymer
chemical mapping
- Abstract text
Polyamide reverse osmosis (RO) membranes are used on offshore platforms for seawater desalination. These membranes typically consist of a polyester backing layer, a 60-100 μm polysulfone (PSf) support film and a 100-700 nm thick polyamide (PA) film. This layered RO membrane has a complex structure with the tight PA separation layer containing porosity or density modulations on the nanoscale which control salt and ion selectivity. However, the structure of the PA layer and its contribution to ion selectivity, is poorly understood.
Whilst neutron reflectivity [1] and scanning transmission x-ray microscopy (STXM) [2] have been used to probe polymer membranes with high spatial resolution in the z direction and high energy resolution respectively, the only technique that is able to map functional chemistry with nanometre spatial resolution, is scanning transmission electron microscopy (STEM) combined with electron energy-loss spectroscopy (EELS). Recent advances in monochromators mean it is now possible to perform spatially resolved EELS with energy resolutions matching those of X-ray absorption spectroscopy.
Soft materials, such as polymers, are particularly challenging to study in the TEM due to the very low characteristic electron dose (characteristic electron fluence) required before damage occurs to the chemical bonds [3]. As a result, acquiring spatially resolved data in beam sensitive samples, is non-trivial. Here we present spatially resolved EELS maps with a resolution of <10 nm across PA membranes.
First, we carried out control beam damage experiments investigating the effect of parameters such as acquisition time, pixel size, sub-pixel scanning and probe size on the C K edge fine structure of plan view PA membranes. The membranes had a nominal thickness of 12 nm and data was acquired from an area of 800 x 800 nm2. Acquisition from such a large area allows lower doses to be used and the signal to be summed to improve the signal-to-noise ratio. This work was carried out at SuperSTEM on a Nion UltraSTEM 100MC “HERMES” monochromated electron microscope operated at 100 kV with a STEM probe size of 0.9 Å.
To investigate the effect of acquisition time/electron dose on the signal, a pixel spacing of 5.3nm was used and the acquisition time varied between 0.01-0.12 s, corresponding to an electron fluence of 7.03 – 42.2x 106 e/Å2. Electron fluences described here were calculated as a function of probe size rather than pixel size resulting in an increase of three orders of magnitude compared with those reported elsewhere. With a 0.02 s acquisition time three clear, sharp peaks were observed in the π* region of the C K edge at ~285.1, 286.4 and 288.0 eV. As the dwell time was increased peaks 1 and 3 reduced in intensity. Peak 1 moving to higher energy-loss and all three peaks merged forming one broad peak at around 286-288 eV. At this dose none of the features in the π* region can be distinguished. There was indication that a critical dose had been reached at an acquisition time of 0.02 s.
The influence of probe size was also investigated, comparing a probe of 0.09 nm with 1 nm whilst maintaining a pixel spacing of 5.3 nm. It was found that although the larger probe had a lower electron fluence, it actually caused more damage to the spectra. This indicates that the damage extends a significant distance from the probe position itself. To investigate this further the spacing between pixels was varied and data recorded for pixel sizes between 1 - 5.5 nm. For the same probe size and dose, the closer together the pixels the more damaged the spectra. This confirms that significant damage occurs outside of the electron probe such that when the electron beam probes the adjacent pixel its chemistry has already been altered.
Similar damage experiments were carried out for the resin and PSf. Whilst some changes in fine structure were observed, the large effects seen for PA were not there suggesting these polymers were already significantly chemically damaged at the electron fluences used for this study.
These studies are useful and essential. However, when probing chemical pathways across the depth the of the PA membrane i.e. in cross section, there is very little area from which to sum spectra to achieve high enough signal-to-noise ratio. Additionally, to observe changes in chemistry with sub-nm spatial resolution, as already discussed, the increased spectral damage can be significant. The dose required to achieve sufficient signal-to-noise ratio with this kind of spatial resolution, undoubtedly damages the chemical bonds. However, in this study we found that despite some damage and broadening observed in the C K edge, and in addition to mapping the elemental distribution, it is still possible to map spatial variation in functional chemistry across the membranes by integrating over windows corresponding to the main peaks found in the reference, minimally damaged spectra. Non-negative matrix factorization (NMF) machine learning algorithms were also used to gain further insights into the chemistry of the PA layer showing clear localisation of signal suggesting that the maps are representative of specific and real functional chemistry within the PA, PSf and resin layers.
In conclusion, it was found that by comparison with carefully acquired reference data it was possible to obtain useful information from ‘damaged’ data sets that could not be found by any other technique. It is noted that the PA membrane used had a much higher damage threshold than other polymer materials which aided in this study. The introduction of direct electron detectors and high frame rate cameras, as well as the ability to have fine control of the dose using electron monochromators, mean it should now be possible to further investigate these beam damage phenomena and acquire spatially resolved data at much lower electron doses.
- References
[1] Foglia, F. et al, Adv. Funct. Mater. 2017, 27 (37), 1701738
[2] Mitchell, G.E. et al. Polymer 2011, 52 (18), 3956-3962
[3] Egerton, R.F. Micron 2019, 119, 72-87