Strategies for the identification of curing and degradation products of oil binders in paintings using FIA H-ESI Orbitrap HRMS

(1)

Bachelor thesis Chemistry

Strategies for the identification of curing and

degradation products of oil binders in paintings using FIA

H-ESI Orbitrap HRMS

by

Liam Hilhorst

18/03/2020

Student number

11740906

Research institute

Cultural heritage agency of the Netherlands

Research group

Cultural heritage laboratory

Supervisor

Prof. dr. K.J. van den Berg

Daily supervisor

(2)

Keywords:

Degree of oxidation, Flow-Injection analysis, Heated-electrospray ionization, Lipid

analysis, Method optimization, Oil paintings, Orbitrap high resolution mass spectroscopy.

Abstract

This thesis presents the development of methods for a high throughput lipid screening technique using FIA-H-ESI Orbitrap HRMS. The thesis revolves around the analysis of multiple naturally aged paint samples which were used to gain MS spectra used to investigate whether the analytical methods are suitable for identifying the condition of the paint. The methods include alterations in the H-ESI tuning focusing on the effects of the spray voltage and the capillary temperature. It was found that a threshold located around 2.0 kV must be exceeded to allow for proper ionization of the sample. Alterations in capillary temperature did not produce a notable impact in the mass spectra. This was due to the absence of triacylglyceride signals which sensitivity reportedly are most affected by the capillary temperature. A characterization of mass spectra was also provided that exemplifies the data that can be generated using the methods. The spectra are also compared to earlier recorded mass spectra taken by F.G. Hoogland using qTOF HRMS acting as a reference. It was found that the spectra shared small resemblance indicating severe signal suppression and contamination. Carryover from previous experiments was also identified through the analysis of the blank runs. Reproducibility was also briefly investigated, but due to these flaws it yielded undesirable results making the technique unable to generate reliable and reproducible quantification in its current iteration. Qualification of lipids was successful and a list was formulated that contains both the encountered lipid species as the expected species which have not been detected due to signal suppression. An indication was also presented of improvements to the method.

(3)

Samenvatting VWO

Moderne olieverfschilderijen krijgen recent te maken met instabiele verflagen. Het opnieuw vloeibaar worden, afschilveren en het lekken van het draagmiddel wordt steeds vaker een probleem bij het behoud van moderne olieverf schilderijen. Om te begrijpen waarom dit gebeurt kan worden gekeken naar de chemische bestandsdelen van olie verf. Zoals de naam van het materiaal al doet vermoeden bevat olieverf olie dat wordt gebruikt om alle kleurstoffen en toevoegingen van de verf te dragen: De zogenoemde draagolie van de verf. Deze olie bestaat uit lipiden, dat zijn moleculen die dezelfde structuur hebben als vet moleculen. Men gebruikt olie wegens zijn unieke droogproces. Dit proces is niet alleen maar het opdrogen van de olie, maar er vind tegelijkertijd een complexe scheikundige reactie plaats. Door invloed van licht gaan de onverzadigde bestandsdelen, de moleculen met een dubbele koolstofbinding, van de olie met elkaar reageren en vormt het een complex netwerk van aan elkaar verbonden lipiden, dat ervoor zorgt dat de verf hard wordt. Waar het bij olieverf misgaat is dat moleculen uit de lucht, met name zuurstof en water, ook betrokken raken bij deze reactie en zorgen voor een ongewenste reactie. De lipiden maken daardoor zuurstofverbindingen en worden in kleinere stukjes opgebroken. Hierdoor wordt de verf weer zwak dat zorgt voor de problemen. In dit verslag wordt daarom geprobeerd een methode te ontwikkelen om te kunnen analyseren in hoeverre de verf ten prooi is gevallen aan deze nevenreacties met zuurstof en water. Op deze manier kan worden gezien of de verf stabiel is en kan op tijd worden ingegrepen voordat er schade optreedt. Dit wordt gedaan met behulp van massa spectrometrie. De lipiden worden met deze techniek geanalyseerd en er kan onderscheid worden gemaakt op basis van de molecuulmassa. Zo kun je dus precies zien welke verf is aangetast door zuurstof en water en welke minder. Voor het ontwikkelen van een analyse methode waar men schade wil voorkomen, is het belangrijk om snel metingen te kunnen doen. Daarom wordt specifiek flow injection analysis heated electrospray orbitrap high resolution mass spectrometry (FIA-H-ESI-orbitrap HRMS) gebruikt. Dit is een aparte techniek die het mogelijk maakt om via een loopvloeistof monsters in hun geheel te analyseren zonder deze eerst lang te hoeven voorbewerken. Dit zorgt er alleen wel voor dat je zeer zorgvuldig moet zijn met het verstellen van het apparaat om alle bestandsdelen te kunnen meten. In het verslag wordt daarom gekeken naar de effecten van het verstellen van de machine en in hoeverre de voorgestelde extractie methode in staat is om een volledig beeld te scheppen van de lipiden in de verf. Uit het onderzoek blijkt dat een minimaal voltage van 2.0 kV moet worden aangebracht op het sample om de signalen zichtbaar te maken. Ook bleek dat de temperatuur van weinig invloed was. Er waren daarnaast echter tekortkomingen geconstateerd. In de bestudeerde methodes werden de signalen die overeenkwamen met de zwaardere onderdelen van de verf niet zichtbaar en lekte vorige resultaten door naar nieuwere metingen waardoor sommige signalen sterk begonnen te schommelen in intensiteit. Daardoor kunnen twee identieke metingen andere resultaten geven. De techniek heeft daardoor dus nog enige aanpassingen nodig om bruikbaar te zijn voor het vaststellen van de stabiliteit van een verfmonster, maar het blijft potentieel geschikt als een analytische methode wanneer deze kunnen worden verholpen.

(4)

1 1. Introduction

Oil paint has taken a prominent role in the development in conservation techniques during the majority of its history. Even now, art conservation continues to be challenged by new problems concerning this popular art medium. In the past 180 years oil paint has gone through rapid transformations partially owing to the mass commercialization of the material through the first tin paint tubes in 1841.1,2_{As an}

effect, the modernization of oil paint underwent rapid innovations by the additions of various new additives and recipes.

A side effect of oil paint becoming a portable and preservable good is the rise of a chain of new art styles stemming from impressionism.3_{These industrial and artistic developments persisted well into}

recent times and has caused a myriad of new conservation problems, stemming from both new paint compositions as well as the innovative application techniques rooting from modern painters.4,5,6,7

Modern oil painting can also include additional materials and can deliberately be applied thicker than usual, causing the paint to partake in unique chemical environments. This in turn also calls for more unique solutions by art restorers and conservators. The obstacles that present-day conservators face are characterized by an unstable paint surface. An example of this can be found in the paintings from Frank van Hemert’s Seven series in which exudation of the paint medium caused liquid paint to drip from deeper layers and sparked an analytical investigation by J.J. Boon and F.G. Hoogland.6_{Due to these}

novel difficulties within art conservation, a thorough understanding of the chemical implications of paint curing has seen increased importance.

Instability and paint dripping may be attributed to oxidation of a poorly drying paint medium. Oil paint requires a binder in the form of an oil, usually poppy oil, linseed oil or safflower oil. These consist of lipids that can oxidize into a firm polymer structure. A drawback to this is that oxidation also occurs with atmospheric oxygen and water which can compromise the stability of the paint. The abundance of certain lipid species can give an indication of the nature of the sample source. Due to the information the lipid contents can give to the nature of an oil, it is also used in food lipid analysis and other applicable oil sources.8,9

Previous approaches to analyzing paint mediums include FTIR, SEM-EDX, liquid chromatography mass spectroscopy (LC-MS), gas chromatography mass spectroscopy (GC-MS) and direct temperature mass spectrometry (DTMS), among others.6,10,11,12,13_{However, due to the complexity of the lipid content}

within oil paint samples, interpretation of the lipid content often remains ambiguous, partially owing to the presence of isobaric compounds. In order to gain insights to quantifying the curing and oxidation of paint layers, high resolution analysis is required, while GC-MS and LC-MS can be applied for quantitative screening of lipids, these processes are often time consuming.

A technique for determining the degree of oxidation in paint mediums has been proposed by A. van den Doel (2015) and J.D.D. van den Berg, et al. (2019). 7,14_{However, these analytical techniques have}

(7)

2

The development of an analytical approach with high throughput and easy to understand information capable of identifying the condition of paint, analytical support can more easily be applied on unstable oil paintings and improve critical decision-making regarding conservation. Vichi et al. (2012) proposed a method for profiling food lipids which could be adapted towards the analysis of the chemical contents of paint binders.8_{High resolution orbitrap mass spectroscopy (Orbitrap-MS) serves as a promising}

analytical method towards investigating the lipid content of oil paint samples. This is because Orbitrap-MS offers high resolution and allows for easy in-source fragmentation for additional characterization.15

The high resolution is in most cases enough to characterize most lipids on exact mass alone. Paired with FIA, a fluid-based injection technique, high throughput and fast analysis time can be achieved without an impairing loss to sensitivity. Through these means, formulating a suitable protocol towards lipid analysis would open the door to crucial information regarding the degree of oxidation of a paint sample. Considering the need for an analytical methods that has fast analysis times, yet contains specific information to easily determine the condition of paint, this study investigates and optimize the capability of FIA H-ESI-Orbitrap-MS to determine the degree of oxidation of paint samples in a reliable and reproducible manner. With this the thesis investigates the question of what implementations will aid in the optimization of FIA H-ESI-Orbitrap-MS lipid analysis and can the method produce clear and reliable quantification of lipids and their oxidation products in paint samples.

This is achieved through running multiple tests runs on aged paint sample and comparing the MS data they generate. Usage of aged paint samples that resemble the material for which this method is developed for is essential to determine the effectiveness of the protocol. Reproducibility of the method is also briefly explored by repeat measurements and tested by comparing the relative intensity of key lipid components.

This report provides a theoretical background explaining the process behind the curing of paint according to literature. In the experimental section a thorough explanation of the explored analytical method is presented, accompanied by the resulting mass spectra.

Analytes were subjected to a variety of MS settings during analysis, in order to study aspects of reproducibility, carryover, mass discrimination, etc.

Paint manufacturers periodically produce small batches of paint and apply them to commercially prepared canvas to record the dry times of their paint, as a method of identifying faulty produces and ensure its quality. Winsor & Newton, a paint producer, donated a collection of these batches to the Tate gallery who made these available to the RCE.

Naturally aged ‘student grade’ paint samples manufactured by Winsor & Newton (1960-1965) and naturally aged low quality paint bought at a local general store (Action 2010) were extracted and analyzed to study the composition of these paints.

The interpretation of which, as well as a discussion of the described methods and suggestions towards further spectra optimization are formulated in the discussion section.

(8)

3 2. Theoretical background

2.1. The curing process

Oil paint requires the use of a paint binder usually linseed oil or safflower oil, that serve as a medium for pigments, paint stabilizers, drying agents and other additives. These drying oils are primarily composed of triacylglycerides of unsaturated fatty acids making them vulnerable to oxidation reactions. The curing and aging of paint layers can be summarized by three chemical processes: Oxidative polymerization, oxidative degeneration and hydrolysis illustrated by the degradation scheme proposed by Burnstock, van den Berg, et al. (2019) shown in Figure 1.7_{The curing of the paint can be attributed}

to the oxidation of the unsaturated triglycerides which, through the use of light induced free radicals, polymerizes into a complex matrix that provides a firm structure. However, deterioration can occur through oxidative degeneration and hydrolysis. Triacylglycerides (TAGs) are essentially esters of glycerol and fatty acid. When paint is exposed to atmospheric oxygen new radical compounds are formed that do not contribute to the formation of a matrix structure, but instead causes cleavage at the unsaturated sections. As a result, hydroxy fatty acids and diacids and other oxygenated compounds gradually start to develop over time, while short chain side product escape the medium as volatiles. Another means of deterioration is by hydrolysis. Triacylglycerides (TAGs) are essentially esters of glycerol and fatty acids making them vulnerable when exposed to water after hydrolysis. Water present in the oil medium will over time react with the ester bonds resulting in free fatty acids (FFAs), diacyl glycerides (DAGs) and monoacylglycerides (MAGs). Combined with oxidative degeneration, oxidized species of these compounds simultaneously, such as dicarboxylic acids, arise making the medium more water-soluble. Difficulty arrives when trying to analyze the structural integrity of the paint binder as the size, complexity and inability to be dissolved severely limits analysis methods. Consequently, the degradation products and materials that did not bind to the cross-linked matrix yet are more commonly investigated to understand the condition of a paint binder.

Figure 1. Oil paint model by Van den Berg et al. (2019) adapted to show the formation of various compounds

(9)

4 2.2. Instrumental background

2.2.1. Flow injection analysis

While chromatographic techniques like liquid chromatography (LC) and gas chromatography (GC) are typical techniques for lipid screening, due to its ability to separate compounds generating MS data in which analytes can be distinguished on their molecular characteristics.17,18_{However, two major}

drawbacks to this technique are its usually long analysis time and large amount of eluent or carrier gas. Flow injection analysis (FIA) is an analytical approach that injects a sample into a carrier solution. Sample is injected along the flow of the carrier solution to a detector.19_{The simplicity and ease of use}

allows for fast analysis times and high throughput. This paired with the characteristic to competently maintain accuracy and sensitivity makes it a popular injection method for purposes that require high sample throughput while maintaining satisfying analytical results.

It makes the technique an interesting approach towards constructing a method of analyzing paint samples. Afterall, when developing an analytical approach aimed at helping identify the condition of a paint medium, it is crucial to implement techniques that generate both useful, but also quickly accessible MS spectra giving it an edge over time consuming chromatographic techniques or less sensitive ambient mass spectrometric techniques. This could in turn give way for faster and more thorough decision making by art restorers and conservators to prevent or treat damage to oil paintings.

2.2.2. Heated Electrospray Ionization

Mass spectrometry is a useful tool for quantification and identification of an analyte’s chemical contents. The field knows a variety of approaches with a range of advantages, causing optimization a complex aim. As for all mass spectroscopic techniques an ionization source is necessary for the detection of any compound. Heated electrospray ionization (H-ESI) is known as a soft ionization technique, meaning fragmentation of the formed pseudo-molecular ions seldomly occurs.16_Similarly,

to its non-heated counterpart the technique induces ionization by applying a high voltage when entering the H-ESI needle. The ionized sample is then sprayed into miniscule droplets where the charge amasses at the surface. The solvent in the droplets then evaporate leading to an increase in ion density until a critical radius is reached where the repulsive forces exceed the surface tension. Consequently, the droplets rupture into smaller droplets repeating the cycle until small gaseous ions remain. Unlike ESI the heated variant aids the evaporation process by introducing heated gas for added control over the nebulization process. Finally, the ions are directed towards the ion transfer capillary that carries the ions to the orbitrap. This process is assisted by a ‘sweeping cone’ that, aided by sweeping gas towards the capillary, directs ions towards the capillary. A soft ionization technique is required in the first screening of the sample to achieve more clarity into the degree of oxidation. Fragmentation would obscure the characterization in a complex sample containing a multitude of different lipids. Since lipids have very similar structures finding the originating parent ion of fragmentation can be met with difficulty and is ultimately unnecessary when comparing oxidation products. Combined with the large selection of adjustable parameters for optimizing the ionization step, gave additional reason to employ this technique.

(10)

5

2.2.3. The workings of high-resolution Fourier transform Orbitrap mass spectroscopy

Orbitrap Fourier transform high resolution mass spectroscopy (Orbitrap-MS) is a useful spectroscopic technique characterized by its capability to quickly obtain high resolution mass spectra compared to other common mass analyzers.15_{It features the use of two cone-shaped electrodes facing each other}

split by a small gap supported by a third central spindle-shaped electrode resulting in a linear current when a current is applies. The resulting radial electric field forces ions towards the central electrode, while simultaneously the tangential component causes an opposing centrifugal force. A near perfect circular motion of the ions can then be achieved by properly adjusting these parameters. Due to the shape of the outer electrodes the resulting axial electric field will push ions towards the center of the ion trap and combined with the linearity of the current, causes purely harmonic oscillations to occur. A mass spectrum can then be produced using the Fourier transformed signals detected by the outer electrodes that receive a frequency signal acquired by the oscillating ions. The technique can achieve high resolution spectra due to the more easily distinct frequencies generated by varying mass-to-charge ratios, making it ideal for quickly differentiating wide arrays of lipids. It is therefore deemed fruitful to develop a protocol using this technique optimizing lipid analysis.

2.2.4 Mass spectrum data acquisition and pretreatment

Both the positive- and the negative ion mode mass spectra were used for they offer complimentary signals, detecting increased amount of lipid species. The negative mode can detect all acidic lipids such as free fatty acids and diacids with high clarity due to reduced chemical noise and lower signal density. The negative mode is unable to detect compounds that are unable to deprotonate, being most unoxidized acylglycerides. The positive mode can detect these range of lipids since ammonium and sodium adduct formation enable the more neutral lipid species such as the acylglycerides to ionize, but this mode is unable to detect free diacids.14

When using FIA, a typical broad ‘bump’ is observed in the injection curve. To acquire an average mass spectrum that best represents the contents of the sample, a time frame must be selected that contains a large portion mass spectra corresponding to the sample and mitigate the spectra that have a higher intensity of chemical noise or random contaminations. These kinds of signals are more likely to occur when little material passes the mass analyzer and spectra should theoretically improve when taking an average time spectra of the middle portion of the ‘bump’ seen shortly after injection. Firstly, a fixed timeframe was proposed which corresponded with the middle of most bumps, but it was later replaced using the more consistently applicable Full Width at Half Maximum (FWHM) technique. This implies using the area the enclosed by the two point that equal halve the maximum of the curve ensuring a good portion directly in the middle of the bump is used spectrum acquisition.

2.3. Recognition and use of key lipid signals in mass spectrometry

2.3.1. Nomenclature and abbreviations

When analyzing a sample containing lipids it is often difficult to recognize the nature of a specific lipid species when conforming to regular IUPAC notations. A system that is adapted for visual convenience is usually employed when describing these compounds. Lipids can essentially be divided in three groups:

Monoacylglycerides (MAG), Diacylglycerides (DAG) and Triacylglycerides (TAG), that are compounds comprised of one, two and three ester bonds with fatty acids respectively. While there is a multitude of possible fatty acids these compounds can be bound to, the most applicable when analyzing plant-based paint binders are palmitic acid (P), linolenic acid (Ln), linoleic (L), oleic and Stearic acid(S). When naming acylglycerides their esters are indicated in by noting the amounts of carbon

(11)

6

atoms, number of double bonds and oxygen atoms present in (e.g. DAG(C16, 2O+C18:1)).

Since this often leads to complicated formula, the ester bonds are abbreviated using the one letter abbreviations of the FA on which they are based, when applicable:

C16 = P C18 = S C18:1 = O C18:2 = L C18:3 = Ln

Dicarboxylic acids are abbreviated as:

DiCx, with x equals the amount of carbon atoms present. Suberic acid = DiC8 = Sub

Azaleic acid = DiC9 = A Sebacic acid = DiC10 = Seb

Oxidized acylglycerides are indicated by adding an ‘Ox-’ prefix when the lipid has been oxidized. The ‘Diacid’ prefix is added when the acylglyceride contains an ester bond with a dicarboxylic acid. 2.3.2. Anticipating lipid components

Due to previous attempts at lipid analysis in paint samples, lipid contents of paint binders are generally well understood, forming a recognizable cluster of compounds.20,21_{Popular plant-based oil binders such}

as poppyseed oil, safflower oil, linseed oil and walnut oil contain lipids comprised of a limited amount of fatty acids. Commonly only lipid species containing palmitic acid (P), linolenic acid (Ln), linoleic acid (L), Oleic acid(O) and stearic acid (S) are detected. Because the oxidation chemistry of these lipids is understood to such a degree that their major oxidation products can be easily deduced, a list of compounds can be constructed detailing the lipid species expected to be found in the paint samples. Regarding oxidized species of lipid compounds, it can be tough to predict every possible oxidized species as the radical chemistry can lead to unlikely compounds being formed. Therefore, only the more likely intermediate species of oxidative cleavage are considered. This includes the epoxide of oleic acid containing compounds (C18:1, 1O) and its dihydroxide product (C18, 2O). Multiple oxidations on linoleic acid, (C18:1, 3O) and (C18, 4O), are also anticipated in an aged paint sample.

A simplified list of all expected compounds can be consulted in Table 1, which focusses primarily on compounds with the highest expectancy of being encountered in the samples discussed in this report. Dimers, ion complexes and derivatives are not considered in the reference table, but naturally cause additional signals. In the positive mode, only [M+NH4]+ and [M+Na]+ are expected. All extracts have

ammonium acetate added to improve pseudo-molecular ion formation. Some compounds have better affinity with ammonium ions and other compounds are more prevalent as a sodium adduct. The addition of ammonium acetate is therefore beneficial for the signal expression of a broader range of lipids. It should also be noted that other adducts are being formed using this technique, such as lithium adducts and loss of water and ammonium adducts, which further complicates signal identification and are excluded in the characterization of the spectra.

(12)

7

Table 1a. List of expected positive lipid ions in aged paint samples.

Assigned structure Ion Chemical

formula Exact mass [g/mol] Positive mode Diacid MAG(Sub) [M+NH4]+ [M+Na]+ C11H24NO6+ C11H20NaO6+ 266.15981 271.11521 Diacid MAG(A) [M+NH4]+ [M+Na]+ C12H26NO6+ C12H22NaO6+ 280.17549 285.13086 Diacid MAG(Seb) [M+NH4]+ [M+Na]+ C13H28NO6+ C13H24NaO6+ 294.14651 299.14651 MAG(P) [M+NH4]+ [M+Na]+ C19H42NO4+ C19H38NaO4+ 348.31084 353.26623 MAG(Ln) [M+NH4]+ [M+Na]+ C21H42NO4+ C21H38NO4+ 372.31084 377.26623 MAG(O) [M+NH4]+ [M+Na]+ C21H44NO4+ C21H40NaO6+ 374.32649 379.28188 MAG(S) [M+NH4]+ [M+Na]+ C21H46NO4+ C21H42NaO4+ 376.34214 381.29753 Ox-MAG(C18:1, 1O) [M+NH4]+ [M+Na]+ C21H44NO5+ C21H40NaO5+ 390.32140 395.27680 Ox-MAG(C18, 2O) [M+NH4]+ [M+Na]+ C21H46NO6+ C21H44NaO6+ 408.33196 413.28736 Ox-MAG(C18:1, 3O) [M+NH4]+ [M+Na]+ C21H44NO7+ C21H40NaO7+ 422.31123 427.26662 Ox-MAG(C18, 4O) [M+NH4]+ [M+Na]+ C21H46NO8+ C21H44NaO8+ 440.32179 445.27719 Diacid DAG(Sub+A) [M+NH4]+ [M+Na]+ C20H38NO9+ C20H34NaO9+ 436.25411 441.2095 Diacid DAG(2A) [M+NH4]+ [M+Na]+ C21H40NO9+ C21H36NaO9+ 450.26976 455.22515 Diacid DAG(Sub+P) [M+NH4]+ [M+Na]+ C27H54NO7+ C27H50NaO7+ 501.38948 509.34488 Ox-Diacid DAG(Sub+C18:1, 1O) [M+NH4]+ [M+Na]+ C27H52NO8+ C27H48NaO8+ 518.36874 Diacid DAG(A+P) [M+NH4]+ [M+Na]+ C28H56NO7+ C28H52NaO7+ 518.40513

Diacid DAG(Sub+C18:1, 1O) [M+Na]+ _C

27H48NaO8+ 523.32414

Diacid DAG(A+P) [M+Na]+ _C

28H52NaO7+ 523.36051 Diacid DAG(Sub+Ln) [M+NH4]+ C29H54NO7+ 528.38948 Diacid DAG(Sub+O) [M+NH4]+ [M+Na]+ C29H54NO7+ C29H50NaO7+ 530.40513 535.36053 Diacid DAG(A+C18:1, 1O) [M+NH4]+

[M+Na]+ C28H56NO8+ C28H52NaO8+ 532.38439 Diacid DAG(Sub+S) [M+NH4]+ [M+Na]+ C29H58NO7+ C29H54NaO7+ 532.42078

Diacid DAG(Sub+Ln) [M+Na]+ _C

29H48NaO7+ 533.34488

Diacid DAG(A+C18:1, 1O) [M+Na]+ _C

28H52NaO8+ 537.33979

Diacid DAG(Sub+S) [M+Na]+ _C

29H54NaO7+ 537.37618

Diacid DAG(A+Ln) [M+NH4]+ C30H56NO7+ 542.40513

Diacid DAG(A+O) [M+NH4]+ C30H58NO7+ 544.42078

Diacid DAG(A+S) [M+NH4]+ C30H60NO7+ 546.43643

Diacid TAG(3Sub) [M+NH4]+ C27H44NO12+ 578.31710

Diacid TAG(3DiC8) [M+Na]+ _C

27H40NaO12+ 583.27250

DAG(2P) [M+NH4]+ C35H72NO5+ 586.54050

DAG(2P) [M+Na]+ _C

35H68NaO5+ 591.49590

Diacid TAG(2Sub+A) [M+Na]+ _C

28H46NaO12+ 597.28815 Ox-DAG(P+C18:1, 1O) [M+NH4]+ C35H70NO6+ 600.51977 DAG(P+O) [M+NH4]+ C37H74NO5+ 612.55615 Ox-DAG(2C18:1, 1O) [M+NH4]+ C35H68NO7+ 614.49903 DAG(P+S) [M+NH4]+ C37H76NO5+ 614.57180 Ox-DAG(P+C18:1, 1O) [M+NH4]+ [M+Na]+ C37H74NO6+ C37H70NaO6+ 628.55107 633.50646

(13)

8

DAG(2S) [M+NH4]+ [M+Na]+ C39H80NO5+ C39H76NaO5+ 642.60310 647.55850 Ox-DAG(P+C18, 2O) [M+NH4]+ [M+Na]+ C37H76NO7+ C37H72NaO7+ 646.56163 651.51703 Ox-DAG(S+C18:1, 1O) [M+NH4]+ [M+Na]+ C39H78NO6+ C39H74NaO6+ 656.58237 661.53776 Ox-DAG(P+C18:1, 3O) [M+NH4]+ [M+Na]+ C37H74NO8+ C37H70NaO8+ 660.54089 665.49628 Ox-DAG(S+C18, 2O) [M+NH4]+ [M+Na]+ C39H80NO7+ C39H76NaO7+ 674.59293 679.54833 Ox-DAG(P+C18, 4O) [M+NH4]+ [M+Na]+ C39H80NO9+ C39H76NaO9+ 678.55146 683.50686 Diacid TAG(S+2Sub) [M+NH4]+ [M+Na]+ C37H70NO10+ C37H66NaO10+ 688.49942 693.45482 Ox-DAG(S+C18:1, 3O) [M+NH4]+ [M+Na]+ C39H78NO8+ C39H74NaO8+ 688.57219 693.52759 Diacid TAG(S+Sub+A) [M+NH4]+ [M+Na]+ C38H72NO10+ C38H68NaO10+ 702.52759 707.47047 Ox-DAG(S+C18, 4O) [M+NH4]+ [M+Na]+ C39H80NO9+ C39H76NaO9+ 706.58276 711.53816 Diacid TAG(S+2A) [M+NH4]+ [M+Na]+ C39H74NO10+ C39H70NaO10+ 716.53072 721.48612 Diacid TAG(2P+Sub) [M+NH4]+ [M+Na]+ C43H84NO8+ C43H80NaO8+ 742.61915 747.57454 Diacid TAG(2P+A) [M+NH4]+ [M+Na]+ C44H86NO8+ C44H82NaO8+ 756.63480 761.59019 Diacid TAG(P+S+Sub) [M+NH4]+ [M+Na]+ C45H88NO8+ C45H84NaO8+ 770.65045 775.60584 Diacid TAG(P+S+A) [M+NH4]+ [M+Na]+ C46H90NO8+ C46H86NaO8+ 784.66610 789.62149 Diacid TAG(2S+Sub) [M+NH4]+ [M+Na]+ C47H92NO8+ C47H88NaO8+ 798.68175 803.63714 Diacid TAG(2S+A) [M+NH4]+ [M+Na]+ C48H94NO8+ C48H90NaO8+ 812.69740 817.65279 TAG(3P) [M+NH4]+ [M+Na]+ C51H102NO5+ C51H98NaO5+ 824.77017 829.72556 TAG(2P+O) [M+NH4]+ [M+Na]+ C53H102NO6+ C53H98NaO6+ 850.78582 855.74121 TAG(2P+S) [M+NH4]+ [M+Na]+ C53H106NO6+ C53H102NaO6+ 852.74121 857.75686 Ox-TAG(2P+C18:1, 1O) [M+NH4]+ [M+Na]+ C53H104NO7+ C53H100NaO7+ 866.78073 871.73613 TAG(P+O+S) [M+NH4]+ [M+Na]+ C55H108NO6+ C55H104NaO6+ 878.81712 883.77251 TAG(P+2S) [M+NH4]+ [M+Na]+ C55H110NO6+ C55H106NaO6+ 880.83277 885.78816 Ox-TAG(2P+C18, 2O) [M+NH4]+ [M+Na]+ C53H106NO8+ C53H102NaO8+ 884.79130 889.74669 Ox-TAG(2P+C18:1, 3O) [M+NH4]+ [M+Na]+ C53H104NO9+ C53H100NaO9+ 898.77056 903.72596 TAG(O+2S) [M+NH4]+ [M+Na]+ C57H112NO6+ C57H108NaO6+ 906.84842 911.80381 TAG(3S) [M+NH4]+ [M+Na]+ C57H104NO6+ C57H100NaO6+ 908.86407 913.81946 Ox-TAG(P+S+C18, 2O) [M+NH4]+ [M+Na]+ C55H110NO8+ C55H106NaO8+ 912.82260 917.77799 Ox-TAG(2P+C18, 4O) [M+NH4]+ [M+Na]+ C53H106NO10+ C53H102NaO10+ 916.78113 921.73652 Ox-TAG(2S+C18:1, 1O) [M+NH4]+ [M+Na]+ C57H112NO7+ C57H108NaO7+ 922.84333 927.79873 Ox-TAG(P+S+C18:1, 3O) [M+NH4]+ [M+Na]+ C55H108NO9+ C55H104NaO9+ 926.80186 831.75726 Ox-TAG(2S+C18, 2O) [M+NH4]+ [M+Na]+ C57H114NO8+ C57H110NaO8+ 940.85390 945.80929 Ox-TAG(P+S+C18, 4O) [M+NH4]+ [M+Na]+ C55H110NO10+ C55H106NaO10+ 944.81243 949.76782 Ox-TAG(2S+C18:1, 3O) [M+NH4]+ [M+Na]+ C57H112NO9+ C57H108NaO9+ 954.83316 959.78856 Ox-TAG(2S+C18, 4O) [M+NH4]+ [M+Na]+ C57H114NO10+ C57H110NaO10+ 972.84373 977.79912

Abbreviations are used as specified in Section 2.3.1.

(14)

9

Compound name Ion Chemical

formula Exact mass [g/mol] Negative mode Hexanoic acid (C6) [M-H]- _C 6H11O2- 115.07645 Heptanoic acid (C7) [M-H]- _C 7H13O2- 129.09210 Octanoic acid (C8) [M-H]- _C 8H15O2- 143.10775 Nonanoic acid (C9) [M-H]- _C 9H17O2- 157.12340

Pimelic acid (DiC7) [M-H]- _C

7H11O4- 159.06628

Decanoic acid (C10) [M-H]- _C

10H19O2- 171.13905

Suberic acid (DiC8), (Sub) [M-H]- _C

8H13O4- 173.08193

Azaleic acid (DiC9), (A) [M-H]- _C

9H15O4- 187.09758

Sebacic acid (DiC10), (Seb) [M-H]- _C

10H17O4- 201.11323 Dodec-3-enedioic acid [M-H]- _C 12H20O4- 228.13616 Diacid MAG(Sub) [M-H]- _C 11H19O6- 247.11871 Palmitic acid (P) [M-H]- _C 16H31O2- 255.23295 Diacid MAG(A) [M-H]- _C 12H21O6- 261.13436 Ox-FFA C16:1, 1O [M-H]- _C 16H29O3- 269.21222 Linolenic acid (Ln) [M-H]- _C 18H29O2- 277.21730 Linoleic acid (L) [M-H]- _C 18H31O2- 279.23295

Oleïc acid (O) [M-H]- _C

18H33O2- 281.24860 Stearic acid (S) [M-H]- _C 18H35O2- 283.26425 Ox-FFA C18:1, 1O [M-H]- _C 18H33O3- 297.24352 Ox-FFA C18, 2O [M-H]- _C 18H35O4- 315.25408 Ox-FFA C18:1, 3O [M-H]- _C 18H33O5- 329.23335 Ox-FFA C18, 3O [M-H]- _C 18H35O5- 331.24900 Diacid DAG(2Sub) [M-H]- _C 19H31O9- 403.19736 Diacid DAG(Sub+A) [M-H]- _C 20H33O9- 417.21301 Diacid DAG(2A) [M-H]- _C 21H35O9- 431.22866 Diacid DAG(P+Sub) [M-H]- _C 27H49O7- 485.34838 Diacid DAG(P+A) [M-H]- _C 28H51O7- 499.36403 Diacid DAG(O+Sub) [M-H]- _C 29H51O7- 511.36403 Diacid DAG(S+Sub) [M-H]- _C 29H53O7- 513.37968 Diacid DAG(O+A) [M-H]- _C 30H53O7- 525.37968

Ox-Diacid DAG(C18, 1O+Sub) [M-H]- _C

29H51O8- 527.35894

Diacid DAG(S+A) [M-H]- _C

30H55O7- 527.39533

Ox-Diacid DAG(C18, 1O+A) [M-H]- _C

30H53O8- 541.37459

Ox-Diacid DAG(C18, 2O+Sub) [M-H]- _C

29H53O9- 545.36951

Ox-Diacid DAG(C18, 2O+A) [M-H]- _C

30H55O9- 559.38516 Diacid TAG(3Sub) [M-H]- _C 27H43O12- 559.27600 Diacid TAG(2Sub+A) [M-H]- _C 28H45O12- 573.29165 Diacid TAG(Sub+2A) [M-H]- _C 29H47O12- 587.30730 Diacid TAG(3A) [M-H]- _C 30H49O12- 601.32295 Diacid TAG(P+2Sub) [M-H]- _C 35H61O10- 641.42702 Diacid TAG(P+Sub+A) [M-H]- _C 36H63O10- 655.44267 Diacid TAG(S+2Sub) [M-H]- _C 36H63O10- 669.45832 Diacid TAG(P+2A) [M-H]- _C 36H63O10- 669.45832 Diacid TAG(S+Sub+A) [M-H]- _C 37H65O10- 683.47397

(15)

10

Diacid TAG(S+2A) [M-H]- _C 38H67O10- 697.48962 Diacid TAG(2P+Sub) [M-H]- _C 43H79O8- 723.57804 Diacid TAG(2P+A) [M-H]- _C 44H81O8- 737.59804 Diacid TAG(P+S+Sub) [M-H]- _C 45H83O8- 751.60934 Diacid TAG(P+S+A) [M-H]- _C 46H85O8- 765.62499 Diacid TAG(2S+Sub) [M-H]- _C 47H87O8- 779.64064

Ox-Diacid TAG(S+Sub+C18, 1O) [M-H]- _C

47H85O9- 793.61991

Diacid TAG(2S+A) [M-H]- _C

48H89O8- 793.65629

Ox-Diacid TAG(S+A+C18, 1O) [M-H]- _C

48H87O9- 807.63556

Ox-Diacid TAG(S+Sub+C18, 2O) [M-H]- _C

47H87O10- 811.63047

Ox-Diacid TAG(S+A+C18, 2O) [M-H]- _C

48H89O10- 825.64612

2.3.3. Dividing mass spectra into convenient points of interest

Due to the sheer amount of distinguishable species in a lipid sample, the resulting mass spectra can be intimidating to inexperienced individuals. A standard spectrum is composed in Figures 2 and 3 in which the prominent lipid contents are characterized, allowing for an overview what can be expected when analyzing these spectra. Figures 2 and 3 additionally indicates mass ranges in which certain species of lipids are generally found. Section 4.1.1 further characterizes the lipids identified. The presence or absence of lipid species can be used to determine the viability of an analytical method.

Figure 2. Standard spectrum of Titanium white (W&N) using method A in the negative mode with labeling of

the most prominent lipid contents. Indicated areas express mass ranges where the [M-H]-_{ions of specific lipid} types are likely to be encountered.

(16)

11

Figure 3. Standard spectrum of titanium white (W&N) using method A in the positive mode with labeling of the

most prominent lipid contents. Indicated areas express mass ranges where the [M+NH4]+_{and [M+Na]}+ _{ions of} specific lipid types are likely to be encountered.

2.3.4. The use of key mass signal ratios to indicates the characteristics of the paint binder

In practice, it is often of secondary importance to determine every lipid compound when developing a method of determining the rate of oxidation and hydrolysis of a given paint sample. Fixating on the relation of a couple key compounds can provide more easily discernable characteristics of the paint sample. P/S and O/S ratios are generally used as a tool to determine the specific oil or oil mixture in mass spectrometry, but methods of determining the degree of hydrolysis and oxidation have also been proposed.7

3. Experimental section

3.1 Materials

3.1.1 Samples

A total of 6 paint samples were analyzed during measurements. All samples were collected using a surgical metal knife and to chip of a small amount of the sides and transporting them to a glass container using metal tweezers.

Four of the paint samples were ‘student grade’ paint swatches dating from 1964–1965 produced by Winsor & Newton which was used by the company to investigate drying times and overall quality of their respective paint batches.14

An additional two samples were taken from a modern low-quality artist’s paint at a general store (Action) and painted out in 2010. All samples were naturally aged.

(17)

12

Samples were taken from these paint swatches:

• Burnt Sienna – Winsor & Newton (BS W&N) • Cobalt blue – Winsor & Newton (CoB) • Raw Sienna – Winsor & Newton (RS)

• Titanium white – Winsor & Newton (TiWh W&N) • Burnt Sienna – general store (BS GS)

• Titanium white – general store (TiWh GS) Abbreviations given between (parentheses) and in bold.

3.1.2 Sample preparation

Sampling of the paint swatches was performed by removing approximately 0.5 mg of the sample using a surgical metal knife. The paint samples were then suspended in 1 mL EtOH (Sigma Aldrich, gradient grade for liquid chromatography LiChropur™) and pulverized using a clean metal needle. The samples were then left to extract for 60 minutes. The mixture would then be homogenized using a W. Krannich KG device. Afterwards, a portion of this mixture was centrifuged for 60 seconds (8000 rpm). A set volume of 20 mMol ammonium acetate (Sigma Aldrich, for LC-MS LiChropur™) was placed in a 4 mL vial with insert and spring fuse screw caps with PTFE septum and filled with an equal volume of the centrifuged extract. The mixture was then homogenized by flushing it 3–4 times and used for MS analysis. The paint extracts were subsequently stored at 10 ℃ covered in aluminum foil in between uses.

3.2. Methods

3.2.1 Instrumental details

The measurements were taken using the switching ionization mode, providing spectra both in the positive and in the negative mode. Using a ThermoFischer scientific Q exactive focus model orbitrap mass spectrometer, a (H-ESI II) probe and a Waters acquity H-Class HPLC to facilitate flow injection analysis.

Mass spectra were generated and interpreted using Chromeleon 7. 3.2.2 Tuning settings

to manipulate the conditions for proper ion formation during analysis, the following H-ESI parameters can be modified:

• Auxiliary gas flow/ temperature

The flow and temperature can be modified to influence the speed of evaporation concerning the formation of ions in the gas state

• Sheath gas

The pressure of the sheath gas used to facilitate the spaying the sample into aerosols can additionally be adjusted to manipulate the velocity and the size of the initial droplets

(18)

13

• Sweeping gas

Sweeping gas is an optional appliance that is directed along a sweeping cone to direct gas state ions towards the ion transfer capillary. The pressure of this gas can be increased to reduce contamination on the sweeping cone and reduce the need for maintenance to the system.

• Capillary temperature

The temperature of the ion transfer capillary can additionally be modified accommodate the flow rate of the analyte to optimize the number of ions that reach the mass analyzer.

• Spray voltage

The voltage applied to the sample when passing the capillary can be modified to increase the likely hood of ionizing compounds in the sample.

To investigate the influence of the tuning parameters on the ion signal of the MS system optimization can be done tuned. In total six methods were explored to give an estimation of the effects of altering the H-ESI parameters. The settings used are presented in Table 2. Method A functions as a default setting from which the other methods are alterations in one or more of the parameters to determine their effect on the mass signals.

Table 2. List of investigated methods and their corresponding H-ESI settings.

Auxiliary gas pressure [Z] Auxiliary gas temperature [℃] Sheathing gas pressure [Z] Capillary temperature [℃] Voltage applied [kV] Sweeping gas pressure [Z] description

Method A 10 200 30 250 3.5 1 Recommended settings

by the Chromeleon 7 software when using 10 µL/min injection speeds. It serves as the starting point for further alterations.

Method B 10 300 30 350 5.5 1 A higher voltage implies

easier formation of ions and the increased temperature speeds up nebulization. A risk with using these increasing these values is corrosion of the needle.

Method C 10 100 30 150 6.5 1 High voltage and low

temperature.

Method D 25 200 45 250 3.5 2 Increased gas settings

improving nebulization speed.

Method E 10 100 30 150 1.5 1 Low temperature and

low voltage

Method F 10 3 30 350 1.5 1 Low voltage and high

(19)

14

3.2.4 Injection method and system reequilibration

During all sample injections the total volume injected equals 10 µL over 5 min using 10 mM ammonium acetate in EtOH as eluent.

In between all measurements the system was reequilibrated by running a blank sample containing 10 mMol ammonium acetate in EtOH for 5 minutes. The blank samples were then investigated for containing any lipids to ensure reequilibration of the system. Some lipid signals, most notably diacids, maintained in the system as evidenced by Figure 7. After long repeated blank runs the amount of lipids visible in the blank samples remained mostly unaffected as shown in Figure 8. During measurements 5 minute blank runs were maintained in between injections. Total analysis time was therefore 10 minutes.

4. Results

4.1. Chemical contents of the samples

4.1.1. Standard results

Across all measurements it was not unusual to produce highly similar data. The standard spectra shown in Section 2.3.3 is an exemplary spectrum which is interpreted in detail. In both the positive and the negative mode the injection pattern shows a conventional curve indicating where most of the sample is detected as seen in Figure 4. Their respective spectra are typically characterized by high signals in the diacid and free fatty acid region indicated in Figure 2 & 3. What is also made visible are in these spectra are labeled peaks which are identified in Table 1 and 2. Not all peaks are identifiable using this method and most of them are difficult to correlate to lipid signals. For this reason many signals, in particular those in the positive mode, are excluded from interpretation.

Figure 4. Standard injection pattern of titanium white (W&N) using method A in the positive mode (top) and

(20)

15

Figure 5. Standard spectrum of Titanium white (W&N) using method A in the negative mode with labeling of

the most prominent lipid contents.

Figure 6. Standard spectrum of titanium white (W&N) using method A in the positive mode with labeling of the

most prominent lipid contents.

4.1.1. Compounds identified

A crucial criteria to determine whether a method is suitable for the determination of the degree of oxidation is its ability to detect most lipid species in the sample. A comparison can then be made between the fragmented compounds and their integral counterparts. Tables 1 & 2 present a brief characterization of the lipids in titanium white (W&N). It serves as a clear example of what compounds are easily detectable using method A, but also eludes many expected lipids which are shown in Appendix III. A list of the expected lipids is seen in Table 1. What is notable in the unidentified expected compound is that mostly triacylglycerides and compounds containing linoleic and linolenic acid. When using the negative mode, most prominent peaks were identified or can be attributed to common ion derivatization from parent ions. Unfortunately, some expected lipids of relatively high mass-to-charge values were unidentifiable. Spectra taken in the positive mode possess harder to identify peaks and while many expected ions are present in most spectra, they are often surrounded by unidentified compounds of similar intensity. Moreover, few compounds of an m/z value higher than 650 were able to be detected.

(21)

16

Table 2. Prominent lipid compounds identified in Titanium white (W&N) using Method A in the negative mode.

Ambiguous and low intensity compounds are excluded. Assigned structures are determined solely on exact masses.

Abbreviations are used as specified in Section 2.3.1

Mass signal [m/z] Relative intensity [%] Chemical formula

Assigned structure Ion Exact mass

[g/mol]

115.075 20.0 C6H11O2- Hexanoic acid, (C6) [M-H]- 115.07645

129.091 21.3 C7H13O2- Heptanoic acid, (C7) [M-H]- 129.09210

141.016 37.5 C4H6NaO4- Sodium diacetate

[2M+Na-2H]

-141.01693

143.107 40.4 C8H15O2- Octanoic acid (C8) [M-H]- 143.10775

159.065 1.6 C7H11O4- Pimelic acid (DiC7) [M-H]- 159.06628

171.139 4.7 C10H19O2- Decanoic acid (C10) [M-H]- 171.13905

173.081 5.7 C8H13O4- Suberic acid, (DiC8), (Sub) [M-H]- 173.08193

187.097 16.9 C9H15O4- Azaleic acid, (DiC9), (A) [M-H]- 187.09758

201.113 4.6 C10H17O4- Sebacic acid, (DiC10), (Seb) [M-H]- 201.11323

215.129 2.5 C11H19O4- Undodecanoic acid (DiC11) [M-H]- 215.12888

247.119 10.0 C11H19O6- Diacid MAG(Sub) [M-H]- 247.11871

255.232 47.7 C16H31O2- Palmitic acid, (P) [M-H]- 255.23295

261.135 57.6 C12H21O6- Diacid MAG(A) [M-H]- 261.13436

281.248 76.6 C16H29O3- Oleic acid, (O) [M-H]- 281.24860

283.265 19.5 C18H35O2- Stearic acid, (S) [M-H]- 283.26425 295.228 30.2 C18H31O3- FA (C18:2, 1O) [M-H]- 295.22787 297.245 87.6 C18H33O3- FA (C18:1, 1O) [M-H]- 297.24352 315.255 26.7 C18H35O4- FA (C18, 2O) [M-H]- 315.25408 329.235 30.8 C18H33O5- FA (C18:1, 2O) [M-H]- 329.23335 343.213 7.5 C18H31O6- FA (C18:2, O4) [M-H]- 343.21261 431.229 16.6 C21H35O9- Diacid DAG(2A) [M-H]- 431.22866 471.333 1.6 C26H47O7 Diacid DAG(P+DiC7) [M-H]- 471.33273 485.350 2.8 C27H49O7- Diacid DAG(P+Sub) [M-H]- 485.34838 499.366 6.7 C28H51O7- Diacid DAG(P+A) [M-H]- 499.36403 525.379 3.2 C29H53O7- Diacid DAG(O+Sub) [M-H]- 525.37968 527.395 2.8 C30H55O7- Diacid DAG(S+A) [M-H]- 527.39533 655.441 0.1 C36H63O10- Diacid TAG(Sub+A+P) [M-H]- 655.44267 669.459 0.1 C37H65O10- Diacid TAG(2Sub+S) [M-H]- 669.45832 697.489 0.5 C39H69O10- Diacid TAG(2A+P) [M-H]- 697.48962

(22)

17

Table 3. Prominent lipid compounds identified in Titanium white (W&N) using Method A in the positive mode.

Ambiguous intensity compounds are excluded. Assigned structures are determined solely on exact masses.

Mass- to-charge ratio [m/z] Relative intensity [%] Chemical formula

Assigned structure Ion Exact

mass [g/mol] 271.115 4.5 C11H20NaO6+ Diacid MAG(Sub) [M+Na]+ 271.11521

280.175 285.131 13.3 26.4 C12H26NO6+ C12H22NaO6+ Diacid MAG(A) [M+NH4]+ [M+Na]+ 280.17549 285.13086 348.308 353.268 2.1 27.5 C19H42NO4+ C19H38NaO4+ MAG(P) [M+NH4]+ [M+Na]+ 348.31084 353.26623 374.326 379.283 7.9 12.5 C21H44NO4+ C21H40NaO6+ MAG(O) [M+NH4]+ [M+Na]+ 374.32649 379.28188

381.298 6.3 C21H42NaO4+ MAG(S) [M+Na]+ 381.29753

395.277 25.4 C21H40NaO5+ Ox-MAG(C18:1, 1O) [M+Na]+ 395.27680

413.287 19.6 C21H44NaO6+ Ox-MAG(C18, 2O) [M+Na]+ 413.28736

427.268 25.4 C21H40NaO7+ Ox-MAG(C18:1, 3O) [M+Na]+ 427.26662

518.406 10.9 C28H56NO7+ Diacid DAG(A+P) [M+NH4]+ 518.40513 528.389 2.7 C29H54NO7+ Diacid DAG(Sub+Ln) [M+NH4]+ 528.38948 544.419 7.0 C30H58NO7+ Diacid DAG(A+O) [M+NH4]+ 544.42078 612.556 5.6 C37H74NO5+ DAG(P+O) [M+NH4]+ 612.55615

4.1.2. Limited detection of triacylglycerides

A comparison was made with the unpublished GC-qTOF MS spectra by F.G. Hoogland (2011) seen in Appendix II.24 _{Across all measurements, exemplified by method A, signal strength of triacylglycerides}

seem to be remarkably low compared to lipids with lower masses as indicated by Tables 1 and 2. This is to such a degree that many expected triacylglycerides are not detected at the TAG area seen in Figure 2 and 3. In the spectra by F.G. Hoogland (2011) signal strength in this area seem to show peak clusters of noticeable intensity contrary to the measurements taken in this investigation leading to the suspicion that these signals are suppressed in the currently proposed methods.

4.1.3. Contaminations

A cluster differing 74.0 m/z can be observed in Figure 9 at the combined sample marked as (C). These are characteristic of polysiloxane clusters and proceed to become progressively more prevalent when drawing injection from the same container. Multiple injections from the same vial were therefore reduced to a minimum. After 3–4 uses these patterns begin to prevalently occur in the TAG region from Figure 2.

Additionally, Carryover was observed between measurements. During experimentation a blank run of 5 minutes was maintained using ammonium acetate 10 mM in ethanol to rinse the system of any remaining lipids of the previous measurement. When analyzing these blank runs signals of lipid compounds can be detected. Figure 7 shows that blank runs before and after measurements differ in lipid content. Most notably, the dicarboxylic acid content and the signals between 400 and 600 m/z are more prominent after the first measurement with raw sienna using method A done after thorough flushing with water:methanol 1:1 for 20 min and a blank run. To resolve this issue three blank runs of 5 minutes were performed, followed by five blank runs of 2 minutes and their spectra were analyzed for lipid traces to see when the carryover was reduced to a minimum. After repeated blank runs using this method it becomes clear that some lipids, most noticeably palmitate, persist in the system indicated by Figure 8. Since carryover reduction by multiple blank runs showed poor improvements to the method, 5 min blank runs were maintained between all measurements.

(23)

18

Figure 7. Negative mode mass spectra of blank runs measured before and after the first measurement.

Figure 8. Negative mode mass spectra of blank runs measured after multiple blank measurement.

4.1.4. Use of analytical standards to investigate triglyceride detection

To investigate the detection of triacylglyceride compounds, chemical standards of tristearin and tripalmitin were investigated, as indicated by Figure 9. A 100-ppm solution of both compounds were prepared in EtOH containing 10 mM ammonium acetate. Additionally, 1:1 mixture of the tristearin and tripalmitin was made using the same concentrations. While detection is seamless at 50-100 ppm concentrations, detection of 1ppm concentrations was not possible.

(24)

19

Figure 9. Positive mode mass spectrum of tripalmitin, tristearin and a combined sample. Contaminations marked

as (C).

4.1.5. Bromine/chlorine isotope patterns

In the titanium white and burnt sienna spectra from paint bought at the general store (GS) shown in Figures 10 and 11. a high density of peaks is observed between 400 and 600 m/z, which consistently appear in the samples. Peak clusters typically differ 33.96 m/z and the shape of the peaks can be recognized as a combination of bromine and chlorine isotope patters. These compounds are therefore unlikely to be associated with lipids or their oxidation products. It is more likely that these are associated with fire retardants like chlorinated paraffins, but exact characterization was met with difficulty.

(25)

20

Figure 11. Repeated mass spectra of titanium white (GS) using method B in the negative mode.

4.2. Spectra comparisons and peak relationships

Due to the complexity of the MS spectra, comparing all signals will result in unclear and complicated data. Instead, key ratios can of the lipid compounds most associated with the stability should give a more clear indication regarding the stability of a paint sample. These ratios are discussed in Section 4.2.1.

To discern whether compounds are missing from the spectrum because it is simply absent from the sample or because the method is unable to detect it, reference data is used of comparable samples of earlier recorded MS spectra discussed in Section 4.2.2.

4.2.1. Important intensity ratios used for spectra comparison

By comparing the relative intensities of key lipid mass signal additional information regarding the effects of the instrumental settings as well as the differences between paint samples can be extracted. Firstly differences between paint samples allows for comparison between the paint sample. The effect of the method used can simultaneously be investigated as well. The key signals are inspired by A. van den Doel (2015) and Colombini (2002).14,10_{The reference spectra of F.G. Hoogland (2011) were also}

used as it measured the same sample and should give an indication of the true relative intensity values of these samples.24_{Unfortunately, only physical copies of these spectra could be used, resulting is some}

of their data being difficult to decipher. The P/S ratio of some samples was noted with certainty however. The ratios are noted in Table 4.

(26)

21

Table 4. Calculated signal ratio’s using the relative intensities of key lipid signals.

Sample Settings P/S MAG(A)/

MAG(P)

O/S A/P ∑Diacids DAG(2A)/

DAG(P+A) Titanium white (GS) Method A Method B Method C reference 2.5 2.8 3.5 n/a 25.9 18.1 16.7 ‘O’ not detected 7.1 8.6 8.9 175 172 177 6.6 6.5 7.1 Burnt Sienna (GS) Method A Method B Method C reference 2.7 2.4 2.1 n/a 6.6 4.2 5.3 ‘O’ not detected 3.8 1.5 3.7 41.1 39.0 41.9 2.5 2.0 - Titanium white (W&N) Method A Method B Method C 2.4 2.5 2.4 1.3 1.2 1.4 3.9 4.1 3.9 0.4 0.5 0.5 27.7 32.1 31.5 0.4 0.5 0.5 Cobalt blue (W&N) Method B 2.5 4.2 0.1 3.0 143 4.8 Raw Sienna (W&N) Method A Method B Method C reference 1.3 1.3 1.1 1.3 3.7 2.9 3.9 0.6 0.6 0.6 1.4 1.4 1.8 136 135 136 2.4 2.5 2.3 Burnt Sienna (W&N) Method A Method B reference 1.4 1.5 3.2 6.5 10.4 0.4 0.2 7.8 3.2 138 137 3.1 3.0

4.2.2. Use of reference spectra

GC-qTOF MS spectra recorded by F.G. Hoogland seen in Appendix II were used as reference to identify missing signals and as method comparison.

When the acquired spectra are compared to earlier spectra recorded with earlier GC-qTOF MS spectra of the same samples, differences in signal ratio can be observed. The degree of oxidation can be estimated by comparing characteristic peaks associated with oxidative degeneration to non-oxidized compounds. The ratios of these anchor points can be seen in Table 4.

4.3 Tuning H-ESI settings

This section investigates the effects of altering the parameters available by using a H-ESI ionization probe. Five methods are investigated to indicate their effect on signal intensities. It is important these settings produce reproducible spectra, but can also support high through-put experiments. To

summarize, these methods attempt to provide an indication in what matter ionization can be improved. The methods investigated can be seen in Table 4.

(27)

22

4.3.1 Spray voltage threshold

During the first few measurements sporadic injection patterns could be seen when a spray voltage lower than 2.0 kV was applied. This is exemplified by Figure 12. and as a result, Method E and method F, which both feature spray voltages lower than 2.0 kV, was regarded with increased scrutiny. Due to this, more attention was directed at spectra taken using the other methods. All other H-ESI settings showed an injection curve which is more typical when using FIA, as seen in Figure 13.

Figure 12. MS Injection curve of a titanium white (W&N) measurement using method D which is exemplary of

all measurement taken when the spray voltage is below 3.0 kV

Figure 13. MS Injection curve of a titanium white (W&N) measurement using method B which is exemplary of all

measurement taken when the spray voltage exceeds 3.0 kV.

4.3.2 Effects of voltage on signal intensity

Apart from the investigated methods spray voltages ranging from 1.0 to 8.0 kV have been examined separately having all other parameters corresponding to method A. As previously mentioned in Section 4.3.1. ionization becomes impaired when voltages under 2.0 kV are used. As for increased spray voltages the signal pattern of the samples remain mostly unchanged, exemplified in Figure 14. The figure shows that voltages just above the ionization threshold of 2.0 kV has very similar results to spectra taken at very high voltages. The intensity of sebacic acid seems to be most affected by the change in voltage.

(28)

23

Figure 14. Comparison spectrum of raw sienna recorded at a spray voltage of 3 kV and 8 kV. The remaining

settings follow those of method A. Abbreviations are used in accordance with Section 2.3.1.

4.3.3. Changes in relative intensities by altered capillary temperature

During experimentation methods which investigates the effects of capillary temperature is altered giving an indication of its effects on ionization of the samples. However, when comparing these differing values, little effect is observed. Figure 15 illustrates that higher temperature cause no visible differences in ionization. Lowering the capillary temperature results in the same to be true, as shown in Figure 16.

(29)

24

Figure 16. Comparison spectrum of titanium white (GS) taken using method A and method C.

4.3.4. Effects of the altered gas pressure settings

The pressure difference caused by raising the auxiliary gas pressure and the sheathing gas pressure by 15 arbitrary units (method D compared to method A) caused strong deviations in signal patterns as seen in Figure 17. These extreme results made it clear that optimizing these settings would require precision and many alterations must be made to find the ideal parameters for nebulization. In further experiments alterations were limited to the spray voltage and temperature to limit the amount of spectra, as the current state of these spectra is seen as severely limited. This is because the relative intensities are altered haphazardly and loss of signal is observed in mass ranges 100–150 m/z and 270–800 m/z, but this highly differing spectrum serves as an indication of what the efficacy of changing the gas pressure would be for method optimization.

Figure 17. MS spectra comparison of titanium white (W&N) measured with default settings (method A) and

(30)

25 4.4. Reproducibility

The titanium white (GS) and burnt sienna (W&N) samples were repeated with different extracts to determine reproducibility of the sample. Titanium white and raw sienna were tested for day reproducibility seen in Figure 19 and 20.

Initially, mass spectra were extracted from the injection curve using a fixed time interval. The negative ion spectra were then used to compare repeatability of the spectra. The differences in relative intensity between key components were then compared to identify the consistency of this method. Unfortunately, this approach yields undesirable reproducibility as seen in Table 5. A different approach was then used utilizing full width at half maximum (FWHM). The resulting consistency of the relative peak heights have not necessarily improved as also indicated by Table 5. However, the use of peak fractions seemed more applicable for differing peak shapes and FWHM intervals were continued to be used at following spectrum extractions.

Stability of the samples is illustrated in Figure 18. The negative mode was once again investigated to get a clear view of the signal ratios of the extracts and to limit the comparison to the majority of the more crucial intensity ratios of oxidized products. The spectra were taken from the same extract but performed with a two-week gap between measurements. During the period between measurements the titanium white sample was kept in the dark at 10 ℃. What immediately becomes apparent is that there is a noticeable difference in relative intensity. The ‘fresh’ samples possess a higher degree of free diacids compared to the ‘older’ sample. Few instances of new signals have been observed, providing a similar peak identity.

The reproducibility of the extraction was also investigated by taking a new extract from the same paint swatch and comparing its average relative peak intensities next to the previous repetition, additionally seen in Figure 18. On first sight the reproduced spectra provide very similar peak intensity, showing little deviating signals. Contrary to this, the numerical value of the relative intensities translate poorly. Tables 4 additionally reveals that the relation between free fatty acids (FFA) and free diacids can differ upwards to 50% upon repeated experiments.

Figure 18. Negative mode MS spectra of titanium white (GS) of multiple extracts to compare sample

(31)

26

Figure 19. Stacked spectra of repeated measurements of a duplicate extract of titanium white (GS) in the negative

mode. Abbreviations of the labels are in accordance with Section 2.3.1.

Raw Sienna has been analyzed over a broader time period providing a representation of the reproducibility of the method over a broad timeframe. The stacked representation of these measurements seen in Figure 20 immediately shows that the measurement possesses poor reproducibility, indicated by the most prominent key components. While all signals are uniformly present with no added unknown masses, the relative intensities differ to an extreme degree among these repetitions. In spite of the variation in relative intensities, the relation of palmitate (P) and stearate (S) remains relatively consistent in all measurements.

(32)

27

Figure 20. Stacked negative mode spectra of a single raw sienna extract measured at three different occasions. In

between measurements the sample was stored at 10 ℃ and was briefly stirred before preparing it for injection. Labels are abbreviated in accordance with Section 2.3.1.

Table 5. List of key ratios calculated from the titanium white (GS) extract using method B found in the negative

mode spectra extracted from the injection curve using a fixed time area near maximum and the FWHM method.

Titanium white Extract 1 Fixed timeframe Average deviation Titanium white Extract 1 Fixed timeframe [%] Average deviation Titanium white Extract 1 FWHM [%] Average deviation Titanium white Extract 2 FWHM [%] P/S 17.60 20.14 10.75

O/S No oleic acid detected No oleic acid detected No oleic acid detected

DiC8/9 4.72 6.45 4.95

∑diacids 2.31 1.52 1.67

DiC8/P 45.90 62.42 38.85

Strategies for the identification of curing and degradation products of oil binders in paintings using FIA H-ESI Orbitrap HRMS

Bachelor thesis Chemistry