Exploration of Thermal Field-Flow Fractionation for the Separation of Polyamides

(1)

Minor Thesis

Exploration of Thermal Field-Flow Fractionation

for the Separation of Polyamides

By

Bram van de Put

February 2020

Student number Examiner

12258156 dr. A. Astefanei

Research Institute Daily Supervisor Van ’t Hoff Institute for Molecular Sciences I. K. Ventouri, MSc.

Research Group Second Reviewer

Analytical Chemistry dr. R. Haselberg

DSM Geleen Industry Supervisor

(2)

2

Abstract

The suitability of thermal field-flow fractionation for the analysis of polyamides using HFIP-based binary solvent systems has been explored in this feasibility study.

Cloud points for different polyamides in binary mixtures of HFIP and several non-solvents including acetonitrile, water and methanol were recorded to assess their predictive capability for ThFFF retention enhancements.

Pure HFIP, as well as a range of binary compositions of HFIP with acetonitrile, water and methanol were found to provide limited to no retention. The use of 40% acetonitrile, however, provided ample retention especially for the tested semi-aromatic polyamide.

The selective gain in retention for the semi-aromatic polyamide may be exploited for the compositional analysis of copolymers of semi-aromatic and aliphatic monomers. However, the reason as to why this gain in retention occurs must be investigated further as the used 40% acetonitrile was above measured cloud point.

Effects of mass loading on the peak shape and recovery have been assessed. The mass load at which overloading based recovery loss occurred was found to be low (10 μg) compared to those found in literature (≈ 200 μg).

Our findings indicate that thermal field-flow fractionation may be a valuable asset in the compositional characterization of co-polyamides. However, there are several technical issues to overcome, such as unknown channel contaminations and lacking knowledge of the thermophoretic behavior of binary solvents.

Experimentation was eventually hampered by an unknown contamination in the flow channel which drastically reduced the optical transmission of the detector flow cell to where no light was transmitted at the relevant wavelength (200 nm) whatsoever. An explanation for the contamination is under investigation.

(3)

3

Introduction

Polyamides (PA) are of industrial and economic interest due to their tunable physical properties. Understanding the relationship between polyamide properties of interest and its molecular structure is imperative in the development of new materials and effective optimization of production processes. Polyamides are often structurally complex due to the existence of multiple property distributions including molecular weight distributions (MWD), chemical composition distributions (CCD), functionality type distributions (FTD) and branching distributions (BD), among others. These distributions are often quantified separately from one another requiring powerful and versatile analytical techniques [1]. The poor solubility of polyamides hugely limits the selection of suitable solvents for analysis. Historically, dangerous solvents were required at elevated temperatures [2]. More recently, the use of 1,1,1,3,3,3-Hexafluoroisopropanol (HFIP) as a solvent for polyamides at room temperature has seen a steady rise, enabling the use of a larger variety of solution phase analysis techniques [3][4][5].

Thermal field flow fractionation (ThFFF) is a separation technique which separates polymers based on their molecular weight and composition. The FFF family of techniques rely on the parabolic flow profile in a narrow, ribbon-like channel for differential migration of the components to be separated. For each component, the average height in the flow profile and therefore its velocity is based on the counteracting forces of an external field to one wall of the flow channel and diffusion from that wall. In ThFFF the external field is a temperature gradient, where one channel wall is heated while the other is cooled, this creates thermal diffusion towards the cold wall. The thermal diffusion effect is not well understood, though it is known to be dependent on the chemical composition while being virtually independent of molar mass. The retention in ThFFF depends on the sample specific diffusion coefficient (D) and thermal diffusion coefficient (DT) as well as the instrumental temperature gradient (ΔT) [6]. This leaves ΔT as the only reasonable way to increase retention. The maximum value of ΔT is however highly limited by the boiling point of the used solvent (bp HFIP = 57 °C [7]), potentially preventing the polymer from being properly retained.

Another possibility for increasing the retention in ThFFF is the use of binary solvent mixtures: When the two solvents have sufficiently different thermal diffusion coefficients of their own, they form a gradient across the flow channel [8]. If the solvent thermally diffusing to the hot wall is a non-solvent for the polymer in question, the polymer will be more closely confined to the cold wall due to the decreasing solubility against the non-solvent gradient. This has been shown to effectively increase ThFFF retention, allowing for the separation of much smaller (less retained) polymers or the use of lower ΔT values [9].

Experimental

Materials

Different polyamide (PA) homopolymers including: PA46, PA410, PA6, PA66 with approximate number average molecular weights (Mn) ≈ 25 kDa, and a semi-aromatic polyamide (PA-SA) with Mn ≈ 6 kDa, were provided by DSM Materials Science Center (Geleen, the Netherlands). Data processing and visualization was performed on a windows 10 workstation using Microsoft Excel and Matlab R2019b.

Used solvents included freshly distilled HFIP (Kindly supplied by DSM), Methanol (ULC-MS grade, Biosolve), acetonitrile (LC-MS grade, Biosolve, Valkenswaard, the Netherlands), and deionized water with a resistivity of 18 MΩ*cm.

(5)

5

Thermal Field-Flow Fractionation (ThFFF)

The thermal FFF system used was a TF2000 equipped with a PN3211 UV detector (Postnova Analytics, Landsberg am Lech, Germany) and a Shimadzu RID-10a differential refractometer (Shimadzu Benelux, ‘s-Hertogenbosch, the Netherlands). The flow channel was 250 μm in thickness, 2 cm in breadth and 45.6 cm in length. The applied flow rate was 0.3 mL/min. The used spacer was purpose made based on a PTFE-enhanced carbon composite material, which limited the applicable temperature gradient to 80K due to its high thermal conductivity. Used carrier liquids included pure HFIP and binary mixtures of HFIP with acetonitrile (ACN), methanol (MeOH) and water. A Shimadzu FCV-10alVP low pressure gradient mixing valve controlled by an LC-10adVP was used for mixing binary solvents on-the-fly unless stated differently. a Shimadzu variable volume mixer in the 2.6 ml configuration was placed after the pump for increased mixing efficiency. UV detection was carried out in dual wavelength mode at 200nm (absorption maximum of amide carbonyl) and 235nm (absorption maximum of aromatic ring) unless stated differently.

Injections were performed using a PN5300 autosampler with a 5 ml loop operating in partial loop-fill mode. All sample solutions were 0.5 mg/mL prepared from a 20 mg/mL stock solution, to have the same solvent composition as the used eluent. Samples were injected using an injection volume of 10ul unless stated differently.

Estimation of D and D

T

The retention parameter in ThFFF, defined as the reduced mean layer thickness, is the effect of a balance between the mass dependent translational diffusion coefficient (D) and the composition dependent thermal diffusion coefficient (DT) following the equation:

𝜆 = 𝐷

𝐷𝑇Δ𝑇

where λ is the dimensionless retention parameter, and ΔT is the temperature difference between the hot and cold wall. λ can then be used to calculate retention times, assuming a parabolic flow profile, using the equation:

𝑡0

𝑡𝑟

= 𝑅 = 6𝜆 (coth (1

2𝜆) − 2𝜆)

In which t0 the elution time of a void compound and tr the retention time of a compound with retention parameter λ. Inversely, the relation between D and DT can be derived from a measured retention time. The factor D across the entire MWD can be estimated using empirical constants from the literature which relate D to the molecular weight in the form of the following equation:

𝐷 = 𝐴𝑀−𝐵

In which M is the molar mass and A and B empirical constants.

Since the molecular weight distributions of all samples have previously been characterized using SEC relative abundancies of each molar mass can be calculated by assuming the molecular weights to follow a normal or Flory type distribution.

A retention model can be established by considering the retention parameter for each molecular mass with a hypothetical DT. Taking non-equilibrium effects as the only contributor of band broadening, we obtain a reasonable approximation of a real life fractogram. A few simplifications and assumptions are

(6)

6 made such as: omission of relaxation and longitudinal band broadening, and assuming the MWD to represent a normal or Flory distribution.

By taking the retention time of peak maximum from a measurement, the measured DT can be estimated by iteratively changing DT in the model until the difference between the retention time at the apex of the model and the experimental retention time converges. It should be noted that this way of approaching DT would mostly be applicable for comparing the relative DT of samples with largely different MWDs and does not give absolute DT values unless further corrections to the model are made, such as including other forms of band broadening and more accurately describing the molecular weight distribution. This approach was tested with reasonable success for a model polystyrene-co-methyl methacrylate system but could not be applied to the polyamides since the relationship between M and D is unknown.

Results and discussion

Although the ThFFF based separation of polyamides in HFIP had not been reported before, it was noted from the literature that polar polymer-solvent pairs that can form hydrogen bonds exhibit very little thermal diffusion and thus show little to no retention [10]. The necessity for binary solvent systems was thus evident to either weaken the hydrogen bonding interaction or to increase retention despite the hydrogen bonding effect. Acetonitrile, methanol, and water were chosen as candidate non-solvents due to their varying properties and good miscibility with HFIP. For a preliminary testing of what non-solvent concentrations can be used, cloud points were determined. ThFFF fractograms were then recorded using pure HFIP and HFIP in combination with the three non-solvents.

Cloud point determination

Solvent cloud points were determined by titrating solutions of 10 mg/ml of each polyamide in HFIP with the three different non-solvents that would later be used in conjunction with HFIP as solvent in ThFFF. Along with giving an approximate idea of what non-solvent concentrations can be used in ThFFF, they may also indicate how strongly the non-solvents confine the different polymers to the cold wall, and may thus correlate with the relative increase in retention. The results, presented as the percentage of non-solvent at which cloudiness occurred, are shown in table 1.

PA 46 PA 410 PA 6 PA 66 PA semi-arom. CP (%) MeOH 59.2 % (-) 52.0 % (1.5) 68.6 % (12.5) 56.5 % (2.4) 24.0 % (3.7) CP (%) ACN 43.0 % (2.1) 34.4 % (2.7) 53.3 % (2.8) 38.1 % (2.4) 21.3 % (7.9) CP (%) H2O 42.4 % (6.3) 34.9 % (2.7) 39.8 % (-) 39.6 % (2.3) 29.6 % (6.3)

The cloud points were significantly different between the polymers, possibly increasing the differences in the apparent DT when binary solvent systems are used, and thus increasing the capability of ThFFF to separate based on compositional differences. A more comprehensive report on the cloud point determinations is provided in attachment 1. The preparation of stock solutions included herein was also used for the subsequent ThFFF measurements.

Table 1: Cloud points for different polyamide non-solvent systems in HFIP, expressed in % non-solvent. All values are means of two repeats. Values between brackets are the 95% confidence intervals, for missing values repeats were the same within measurement accuracy.

(7)

7

Thermal Field-Flow Fractionation

Solvent composition

Preliminary ThFFF measurements were conducted using solely HFIP as solvent. An overlay of the resulting fractograms is displayed in figure 1. As expected, there was hardly any retention in pure HFIP, with all eluting distributions grouping at t0 which was about 5.3 min, as derived from a blank. The baseline fluctuation observed at about 1.5 minutes is caused by switching of the focusing valve.

ThFFF experiments were then continued using 20% and 40% of each of the selected non-solvents; water, methanol (MeOH) and acetonitrile (ACN). For all solvent combinations, no significant increase in retention was observed at 20% non-solvent. However, changes in peak shape were observed with 20% acetonitrile. For the aliphatic polyamides, a noticeable shoulder appeared around t0 when using acetonitrile concentrations of 20% both as eluent and solvent.

At a water concentration of 40%, retention did not increase notably, though a change in peak shape and loss in recovery was observed for the semi-aromatic polyamide.

No elution was observed using 40% methanol as solvent, instead the baseline dropped during each run, periodically, but not consistent with the injection frequency. The pressure was found to increase during these baseline drops. Possibly, the solvent boiled during these runs, resulting in deposition of the injected

0 0.1 0.2 0.3 0.4 0.5 0 2 4 6 8 10 12 U V signa l, 200nm (V)

Retention time (min)

Fractograms: PA 46, 20% non-solvent

HFIP

20% Acetonitrile 20% Methanol 20% water

Figure 2: Fractograms of polyamide 46 using 20% of the different non-solvents in HFIP and pure HFIP.

0 0.2 0.4 0.6 0.8 1 0 2 4 6 8 10 12 UV sign al, 200n m (V)

Fractograms: all polyamides, 100% HFIP

PA 46 PA 410 PA 6 PA 66 PA SA

Figure 1: Preliminary ThFFF fractogram of all homopolymers, solvent composition: 100% HFIP. Right: blank, showing t0 at about 5.3 minutes.

45.2 45.4 45.6 45.8 46 46.2 0 5 10 U V signa l (m v)

(8)

8 polymer onto the channel walls. This may have been caused by the formation of a low boiling eutectic, since the back pressure was relatively high compared to that of the other solvent combinations.

At 40% acetonitrile however, an increase in retention was observed for all polyamides. Especially the semi-aromatic polyamide (PA SA) gained greatly in retention, the entire polymer distribution was separated from t0. Fractograms for the different polyamides at 40% acetonitrile are provided in figure 3.

Identification of the void peak

Since the semi-aromatic polymer distribution is detached from the peak at t0 the nature of this void peak could now be identified. Injections of the PA SA sample using 40% ACN were performed by injecting 20ul of a 0.5mg/ml solution as well as 10 ul of a 1mg/ml solution, ensuring an equal amount of injected polymer while the amount of injected solvent is varied. The area of the void peak was found to be doubled with the higher injection volume, whereas the area of the polymer distribution remained approximately equal, indicating the void peak to be caused primarily by the solvent.

0 0.05 0.1 0.15 0.2 0.25 0.3 0 5 10 15 20 25 30 35 40 45 U V signa l, 200nm (V)

Fractograms: 40% Acetonitrile

PA 46 PA 6 PA 66 PA SA

Figure 3: Fractograms of the different polyamides using 40% acetonitrile as solvent. A fractogram of PA410 was not obtained.

0 0.1 0.2 0.3 0.4 0.5 0 5 10 15 20 25 30 35 40 U V signa l, 200nm (V)

PA SA: Injection volume and void peak

0.5mg/ml 20ul 1mg/ml 10ul

Figure 4: Fractogram of the semi-aromatic polyamide using with changing injection volume but similar mass load, showing the injection volume dependence of the void peak. Solvent

(9)

9

Effect of mass loading

The effect of the introduced analyte mass was tested by injecting different volumes between 5ul and 50ul of the 0.5mg/ml PA SA solution, again at 40% ACN.

Approximate peak areas were calculated by summing the intensities across the span of the eluted peaks shown in figure 5. Due to partial overlap of the polymer distribution and the void peak. The peak areas were found to behave linearly up to an injected volume of 20ul, corresponding to a sample mass of 10μg. It was found that the apex of the polymer distribution shifted to lower retention times with increasing mass load, indicating overloading at these higher concentrations. Notably, these values are very low compared to the typical masses to cause overloading in ThFFF, which is often in the hundreds of micrograms [11]. It should be noted that the increased volume of the sample plug is a secondary effect. However, this could not explain the large observed shift in retention and recovery since the solvent and eluent compositions are the same.

0.04 0.09 0.14 0.19 0.24 0.29 3 8 13 18 23 28 33 U V sig n al, 20 0nm (V )

PA SA 40% ACN: effect of the mass load

5ul 10ul 20ul 50ul

Figure 6: fractograms of the semi-aromatic polyamide at 40% ACN, using different injection volumes. Increasing skew to lower retention times is observed for the higher injection volumes.

100 120 140 160 180 200 220 240 0 10 20 30 40 50 60 Are a (s u m in ten sity ) injected volume (μl)

PA SA 40% ACN: injected volume

versus peak area

Figure 5: Injection volume versus sum-intensity estimated peak area for the semi-aromatic polyamide at 40% acetonitrile. Showing approximately linear response until an injection volume of 20 μl.

(10)

10

Solvent peak correction

Comparing the ThFFF fractograms of PA46 at 200nm and 235nm, it is evident the absorption of the polymer at 235nm is negligible compared to that of the solvent (see figure 7).

This presents a convenient way of correcting for the coeluting solvent peak in the 200nm trace. For this, a blank of each non-solvent concentration was recorded and the ratio between the peak areas in the 200nm and 235nm trace was calculated, resulting in a correction factor. The solvent peak could then be removed from the fractograms of the aliphatic polyamides by multiplying the 235nm trace with the earlier obtained correction factor and subtracting the scaled intensity values at each time point from the 200nm trace. Figure 8 shows the reasonable efficacy of the correction procedure. The remaining void peak might be due to low MW oligomers which will elute collectively around t0.

0.03 0.08 0.13 0.18 0.23 3 8 13 U V signa l (V)

PA 46, 40% ACN

200nm 235nm 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 3 5 7 9 11 13 15 U V signa l (V)

PA46, 40% ACN Solvent correction

235nm*factor 200nm 200nm -235nm*factor

Figure 7: Comparison of the PA46 40% ACN fractogram at 200nm and 235nm.

Figure 8: Example of blank-based solvent peak correction. Top: Scaled solvent peak in 235nm trace. Middle: Raw fractogram at 200nm. Bottom: Solvent peak subtracted fractogram at 200nm.

(11)

11

Recovery

To assess sample losses during the analysis, recovery values were calculated. Start and end points of the polymer distributions, including void peak, were manually selected, after which a linear curve was fitted between them, as well as an 8th_{order gaussian curve on all the datapoints between the selected start and} end. An example fit is displayed in figure 9.

Both the gaussian and linear fit were then integrated over a million time points using the built-in MATLAB “integrate” function and the resulting values at each datapoint were summed to obtain the area under the curves. The obtained area of the baseline was then subtracted from that of the gaussian fit. Fits were performed for both the 200nm and 235nm traces.

Peak areas for aliphatic polyamides with acetonitrile were corrected for solvent absorption using blank-based correction factors as described in the previous section. Surprisingly, significantly smaller solvent peaks were observed when using methanol or water as non-solvent. Recoveries were calculated for the different polymer-solvent systems, assuming full recovery using 100% HFIP. Ideally, recovery factors would have been derived using 100% HFIP without thermal gradient as reference.

PA 46 PA 410 PA 6 PA 66 PA SA HFIP 100 100 100 100 100 20% ACN 75.4 110.1 103.1 84.4 80.5 40% ACN 43.0 - 83.9 43.8 56.0 20% MeOH 97.5 90.0 85.9 81.5 69.5 20% Water 105.0 102.1 110.1 88.3 87.6 40% Water 120.1 87.9 105.1 90.1 9.2

For some polymer-solvent systems, the recovery was higher than that for 100% HFIP, which can have a number of reasons. The polymer can show less wall adsorption than in 100% HFIP, an overlapping solvent peak can have strong UV absorbance while going unnoticed in the blanks, the polymer can behave solvatochromically due to the strongly changing hydrogen bonding environment [12], or it might be within the measurement error. A conclusive cause cannot be found at this point.

Table 2: Percent recovery values for the different polymer-solvent systems. A fractogram for PA 410 at 40% ACN was not available.

PA 46 PA 410 PA 6 PA 66 PA SA HFIP 100 100 100 100 100 20% ACN 75.4 110.1 103.1 84.4 100.8 40% ACN 43.0 - 83.9 43.8 18.0 20% MeOH 97.5 90.0 85.9 81.5 69.5 20% Water 105.0 102.1 110.1 88.3 87.6 40% Water 120.1 87.9 105.1 90.1 9.2

Table 3: Percent recovery values for the different polymer-solvent systems. A fractogram for PA 410 at 40% ACN was not available.

Figure 9: Example elution profile and baseline fits for the 200nm trace of PA46, as used for integration.

(12)

12

Conclusions

While pure HFIP does not result in any retention of polyamides in ThFFF, a binary solvent system consisting of 40% acetonitrile in HFIP was found to provide retention for the polyamides in question. The retention increase for the aliphatic polyamides was marginal, whereas the semi-aromatic polyamide gained significant retention. Retention was characteristic for each of the tested polyamides, forming a foundation for differentiation in polymer blends as well as composition-based separations of copolymers. The strong reaction of the semi-aromatic polyamide is notable, in that the observed retention is notably high for a polymer of such low molecular weight (6 kDa). A previous study on binary solvent systems for the retention enhancement of an 11.2 kDa polystyrene sample achieved similar retention factors at a temperature drop of 100 °C [13], which to our knowledge was the highest retention ever achieved for polymers with Mn < 15 kDa. Polystyrene is relatively well retained, even using single component solvent systems, possibly making the retention increase observed here the highest ever reported.

The retention increase should be regarded with scrutiny though, since the acetonitrile concentration of 40% is above the previously determined cloud point, and many of the samples were notably cloudy. In case a solvent gradient does establish as expected one might think the polymer will get redissolved and be retained as normal. However, if the polymer does not redissolve, retention may progress via a much different mechanism, though this would warrant further investigation in its own right. Implementation of Multi Angle Light Scattering detection (MALS) or Dynamic Light Scattering detection would quickly provide insight in the solvated state of the polymer.

Since the retention enhancement in binary solvent systems is related to the lower solubility in the non-solvent enriched zone, it would logically correlate with the cloud points of the different non-solvents. No such relation was directly found, admittedly an existing relation might be hugely complex, for which the current data would not be adequate.

Further experimentation should be commenced including: the further increase in acetonitrile concentration, optimizing the temperature gradient, as well as the cold wall temperature and flow rate and inquiring upon the dependence of retention on the composition of (block-)copolymers for which the technique may be uniquely well-suited. The determination of thermal diffusion coefficients should be considered by either employing DLS or preparative SEC.

Further experiments could not be conducted within the duration of this project due to a particularly stubborn contamination. A brief overview of the troubleshooting procedure for the arising problem is outlined in attachment 2.

(13)

13

Acknowledgements

I would like to thank Iro Ventouri for providing me with this project and for her great guidance throughout. Great gratitude goes to DSM and in particular to Dr. Harry Philipsen for developing the concept in the first place, providing feedback and insight in the progression and providing funding and materials for the project. I would also like to thank the service and support team at PostNova for their help in setting up the system and supporting us with technical issues. I would further like to thank Dr. Alina Astefanei and Dr. Rob Haselberg for reviewing this thesis and providing feedback where necessary. Also, I would like to thank the engineers of the UvA for helping me realize the project and giving me the freedom to use and tinker with their systems. Lastly, I would like to thank my colleagues at the UvA for making every day in the office informative and above all fun.

(14)

14

References

[1] H. J. A. Philipsen, “Determination of chemical composition distributions in synthetic polymers,” J. Chromatogr. A, vol. 1037, no. 1–2, pp. 329–350, 2004.

[2] Y. Mengerink, “Non-exclusion separation techniques for polyamides,” Eindhoven University of Technology, 2001.

[3] C. C. Terant, E. M. Macchi, and R. V Figini, “Solution Properties and Molecular-Weight Distribution of Nylon 11.”

[4] S. R. Samanta, “Intrinsic Viscosity and Molecular Weight Measurement of Nylon 66 Polymers.” [5] R. Mendichi, S. Russo, L. Ricco, and A. G. Schieroni, “Hexafluoroisopropanol as size exclusion

chromatography mobile phase for Polyamide 6,” J. Sep. Sci, vol. 27, pp. 637–644, 2004.

[6] M. E. Schimpf and J. C. Giddings, “Characterization of thermal diffusion in polymer solutions by thermal field‐flow fractionation: Dependence on polymer and solvent parameters,” J. Polym. Sci. Part B Polym. Phys., vol. 27, no. 6, pp. 1317–1332, 1989.

[7] J. P. Bégué, D. Bonnet-Delpon, and B. Crousse, “Fluorinated Alcohols: A New Medium for Selective and Clean Reaction,” Synlett, no. 1, pp. 18–29, 2004.

[8] C. A. Rue and M. E. Schimpf, “Thermal Diffusion in Liquid Mixtures and Its Effect on Polymer Retention in Thermal Field-Flow Fractionation,” Anal. Chem., vol. 66, no. 22, pp. 4054–4062, 1994.

[9] G. E. Kassalainen and S. K. R. Williams, “Lowering the molecular mass limit of thermal field-flow fractionation for polymer separations,” J. Chromatogr. A, vol. 988, no. 2, pp. 285–295, 2003. [10] D. Niether and S. Wiegand, “Thermophoresis of biological and biocompatible compounds in

aqueous solution,” J. Phys. Condens. Matter, vol. 31, 2019.

[11] G. Greyling and H. Pasch, Thermal Field-Flow Fractionation of Polymers. Cham: Springer Nature Switzerland, 2019.

[12] N. De Silva, S. Y. Willow, and M. S. Gordon, “Solvent Induced Shifts in the UV Spectrum of Amides,” J. Phys. Chem. A, vol. 117, pp. 11847–11855, 2013.

[13] G. E. Kassalainen and S. K. R. Williams, “Coupling thermal field-flow fractionation with matrix-assisted laser desorption/ionization time-of-flight mass spectrometry for the analysis of synthetic polymers,” Anal. Chem., vol. 75, no. 8, pp. 1887–1894, 2003.

(15)

15

Attachment 1: Cloud-point determination report

Outline:

Determination of binary solvent cloud-points for polyamides, to establish limit non-solvent concentrations for the use in Thermal Field-Flow Fractionation (ThFFF) experiments.

Few organic solvents are suitable to fully dissolve polyamides (PA), the selection is generally limited to: Formic acid or 1,1,1,3,3,3-Hexafluoroisopropanol (HFIP), of which only HFIP is suitable for most analyses. Most other solvents cannot dissolve PA on their own, however binary solvent mixtures of HFIP and other solvents can.

The limit of the non-solvent fraction that can still dissolve a polymer is called the Cloud-Point, owing to the dense dispersion that is observed once this threshold is overcome. The cloud point is however not only dependent on the relative solvent/non-solvent fraction but also on the temperature (T) and the fraction of polymer in the system. A ternary phase diagram of a typical solvent-nonsolvent-polymer system is given below.

(16)

16 At low polymer concentration, the non-solvent fraction at the cloud point (φCP) increases approximately linearly with the polymer fraction. At appreciable polymer fractions though, φCP decreases with increasing polymer fraction. Since we will be operating at low polymer concentrations we can assume to operate in the former regime.

The polymer molecular weight (MW) also drastically influences the solubility, where higher molecular weights correspond to lower solubility and subsequently lower φCP. For low MW compounds the non-solvent may be a (semi-)non-solvent at low concentration.

With increasing temperature, the polymer solubility tends to only increase, we therefore will perform our cloud-point determinations at the lowest temperature we assume our solutions to reach during subsequent ThFFF runs (room temperature ~22 °C).

Materials:

- Shaker

- Micropipette (5000 μl) - Stirplate and small stirbars - Vials (10, 20 ml)

- Burette (20 ml)

- Syringe pump (KDS 100) - Glass syringe (2,5 ml, Luer)

- Luer adapter with blue PEEK tubing

Chemicals:

- HFIP (Distilled, kindly provided by DSM) - ACN (LC-MS Grade, Biosolve)

- MeOH (ULC-MS Grade, Biosolve) - Deionized Water (milli-Q)

Samples:

- PA 46 - PA 6 - PA 66 - PA 410 - PA SA

(17)

17

Methods:

Of each polymer sample, a stock solution of 20 mg/ml was made by weighing in about 200 mg of polymer and adding up to 10 ml of HFIP. The final pipetting scheme, weigh-ins and concentrations are given below. Target volumes were halved for use with a 5 ml micropipette (1/2 V).

Table 4: Weigh-ins and final concentration of the different polyamide stock-solutions.

Sample PA 46 Mn: 30k PDI: 2 DSM-5 PA 410 Mn: 30k PDI: 2 DSM-6 PA 46 PA 410 PA 6 PA 66 PA SA PT After unit Target mass 200 200 200 200 200 200 200 200 mg Weigh-in 198 193.6 191.3 198.7 197.8 190.8 198.3 194.5 mg V HFIP (theor.) 9.9 9.68 9.565 9.935 9.89 9.54 9.915 9.725 ml 1/2 V (theor.) 4.95 4.84 4.7825 4.9675 4.945 4.77 4.9575 4.8625 ml 1/2 V used 4.95 4.84 4.78 4.97 4.945 4.77 4.96 4.86 ml C true 20 20 20.01 19.99 20 20 19.99 20.01 mg/ml

Six 1 ml solutions for cloud-point determination were made in 10 mg/ml concentrations for all samples, by pipetting 500 μl and 100 μl of the previously made stock solutions into 1.5 ml vials and adding 500 μl and 900 μl of HFIP respectively.

A basic titration set-up was built using a syringe pump, fitted with a 0,5 ml glass syringe to which a length of PEEK tubing was attached. Small stirbars were added to each polymer solution. During titration the solution in question was continuously stirred using a stirplate at 500 rpm.

Before titration, the PEEK tubing was rinsed with solvent multiple times, wiped on the outside with tissue and suspended just below the surface of the solution to be titrated.

For each sample, titration was performed in duplicate for the three non-solvents: water, methanol and acetonitrile.

The syringe pump was set to dispense at a rate of 0.5ml/min, during which the titrated solution was visually monitored for the formation of a white suspension, indicating the Cloud-Point was reached. When the cloud point was reached, the syringe pump was stopped and the dispensed volume of non-solvent (Vns), as indicated by the syringe pump, was reported.

From the obtained volumetric data, the non-solvent fraction at the cloud point (φCP) was calculated using the following equation:

Φ𝐶𝑃 =

V𝑛𝑠

(18)

18

Results

The addition of up to 0.8 ml water to a 2mg/ml PA46 solution did not result in any noticeable cloudiness. Several more syringefuls of water were added to force cloud formation though still nothing happened. Likely the concentration of 2mg/ml is either too low to form a noticeable suspension, or the formed solids collected on the glass surface or the stir-bar, or the concentration was so low that a supersaturated solution was formed.

Addition of up to 0.4 ml of methanol to a 20mg/ml of PA46 solution neither yielded a noticeable suspension formation. After manual addition of about one more ml of methanol was a suspension formed. With the cloud-points being at much higher non-solvent concentrations, the syringe was replaced by a larger one of 2.5 ml, delivering at a rate of 31 ml/h.

The semi-aromatic polyamide sample behaved differently to the other polymers, gradually forming a more and more cloudy solution upon addition of more non-solvent after the noticeable cloud point of 0.24 ml/ml (MeOH/ml). The aliphatic polymers suddenly turned cloudy and gradually flocculated with more stirring when the cloud point was reached (MeOH). It seemed the semi-aromatic polymer formed more akin to a stable colloidal suspension.

Table 5: Established cloudpoints for the different polyamides-nonsolvent combinations.

PA 46 PA 410 PA 6 PA 66 PA semi-arom. CP (%) MeOH 59.2 % 52.0 % 68.6 % 56.5 % 24.0 % CP (%) ACN 43.0 % 34.4 % 53.3 % 38.1 % 21.3 % CP (%) H2O 42.4 % 34.9 % 39.8 % 39.6 % 29.6 %

Conclusions

Cloud points in binary mixtures of HFIP with methanol, acetonitrile and water have been determined for five polyamides. The cloud points were distinctly different between the various polymers. The semi-aromatic polyamide behaved differently than the aliphatics in that it formed a stable, translucent suspension instead of a cloudy suspension.

(19)

19

Attachment 2: Troubleshooting

After acquisition of the methanol and water-based runs, the previously promising results as obtained for acetonitrile could not be reproduced anymore. An acquisition of polyamide 46 at 40% acetonitrile revealed loss in retention and a skewed elution profile, both in the 200nm and 235nm traces. Due to the low recoveries for the 40% methanol and water experiments, adsorption of polymer on the channel surface was suspected to be the cause.

The channel was disassembled presenting several abnormalities: Part of the surface was coated in a shiny white residue which did not dissolve in methanol and was expected to be the unrecovered polymer, a dark green tarnishing was present around the inlet of the channel, most notably on the hot wall, several yellow circular spots were present both inside and outside of the flow-channel area, and a grease-like substance covered the hot and cold-wall surfaces outside of the flow-channel area (around the spacer).

The white residue was removed using pure formic acid, as HFIP was too volatile to be practical. The circular yellow spots were removed using methanol. The green tarnish was thought to be a copper compound formed from exposed copper from inside the inlet drilling and could not be removed using a variety of solvents (methanol, acetone, formic acid, and concentrated ammonia). During cleaning, greasy smears were formed on the surface of the cold wall, which were hard to remove, applying several drops of pure formic acid on the afflicted spot, leaving it for a few seconds and then removing the formic acid in a single wipe with tissue was the only working approach.

Figure 10: Photographs of the hot and cold plate before cleaning. left: shiny residue on cold plate, suspected polymer. middle: yellow spots on cold plate, unidentified. right: dark green deposit on the inlet side of the hot plate, suspected to be a copper complex.

(20)

20 After cleaning the channel was reassembled and a test run of polystyrene using THF as solvent was conducted, which showed decent retention and peak shape. Though when returning to HFIP based solvents, the 200nm detector channel showed zero transmission. Flushing overnight on low flow with 100% HFIP would let some light transmit through the detector, though switching to any other solvent composition saturated the detector again.

Eluent consisting of 40% acetonitrile in HFIP was flushed through the system, sampled from different locations and analyzed with UV spectroscopy to localize the source of contamination. A distinct UV-spectrum was observed for the samples taken after the flow channel, while all samples taken in front of the flow channel did not exhibit such features.

Figure 11: UV spectra of 40% ACN in HFIP effluent from different locations in the ThFFF system

Following these findings, the circular yellow spots observed previously were suspected of being the cause. Since these spots were also observed outside of the flow channel, where the spacer blocks the solvent from passing, the possibility of the contamination arising from the spacer was considered. The spacer was replaced, though the problems persisted.

A piece was cut out of the spacer and submerged in a mixture of 40% acetonitrile in HFIP. The spacer was left to extract for several hours after which the extract was again analyzed with UV-spectroscopy. The resulting UV-spectrum strongly resembled that of the contaminated effluent samples. However, the spacer is Teflon based and should not contain organic leachables.

A final candidate to account for the contamination was the greasy substance that resided outside of the flow channel during the first cleaning procedure. This was thought to originate from the thermal grease, which prevents a thermal insulating block from irreversibly fusing to the hot block. Like the spacer, a sample of (new) thermal grease was extracted with 40% acetonitrile and analyzed with UV spectroscopy. The obtained spectrum was saturated, and again strongly resembled the spectrum of the contaminated effluents. 0 2 4 190 210 230 250 270 290 In ten sity (A.U.) Wavelength (nm)

UV-VIS spectra of samples from different points

_{From bottle} After auto sampler before channel After manual valve before channel After channel before PEEK tubing

(21)

21

Figure 12: UV spectra comparing the extracts of cut-outs from the used spacer and thermal grease with contaminated effluent.

Thermal grease being the cause adds up with the otherwise hard to explain fact that the contamination only happened after “cleaning” the flow channel. A thermally liquefiable compound in the grease may have seeped out from between the insulating plate and the hot block, making its way under the uncompressed part of the flow channel. During cleaning, the grease will have been wiped into the flow channel area without being effectively removed with the used polar solvents, instead being smeared out in a thin film. From this film, strongly absorbing components will have slowly leached, being concentrated enough to absorb all the light at 200nm once the detector is reached.

Although several of the recorded spectra are saturated and thus not fully comparable, the obtained evidence strongly suggests the thermal grease being the culprit in contaminating the system. Further steps should be taken in cleaning the flow channel using less polar solvents, as these will likely be more potent in removing the grease-born contaminant.

0 1 2 3 4 190 210 230 250 270 290 In ten sity (A.U.) Wavelength (nm)

UV-VIS spectra of contaminant and extracts

After channel before PEEK tubing

After detector Spacer extract

Exploration of Thermal Field-Flow Fractionation for the Separation of Polyamides

Minor Thesis